USF Libraries
USF Digital Collections

Twenty years later


Material Information

Twenty years later what we have learned since the Big Thompson flood : proceedings of a meeting held in Fort Collins, Colorado, July 13-15, 1996
Series Title:
Special publication / Natural Hazards Research and Applications Information Center ;
Portion of title:
What we have learned since the Big Thompson flood
Physical Description:
1 online resourcce (xxii, 205 p.) : ill. ;
University of Colorado, Boulder -- Natural Hazards Research and Applications Information Center
Natural Hazards Research and Applications Information Center, University of Colorado
Place of Publication:
Boulder, Colo
Publication Date:


Subjects / Keywords:
Flood damage prevention -- Congresses -- Colorado   ( lcsh )
Flood forecasting -- Congresses -- Colorado   ( lcsh )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
conference publication   ( marcgt )
non-fiction   ( marcgt )


Includes bibliographical references.
General Note:
Description based on print version record.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002025197
oclc - 432313804
usfldc doi - F57-00095
usfldc handle - f57.95
System ID:

This item is only available as the following downloads:

Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200325Ia 4500
controlfield tag 001 002025197
005 20090928143710.0
006 m d
007 cr bn|||||||||
008 090821s1997 coua sb s100 0 eng d
datafield ind1 8 ind2 024
subfield code a F57-00095
0 245
Twenty years later
h [electronic resource] :
b what we have learned since the Big Thompson flood : proceedings of a meeting held in Fort Collins, Colorado, July 13-15, 1996.
3 246
What we have learned since the Big Thompson flood
Boulder, Colo. :
Natural Hazards Research and Applications Information Center, University of Colorado,
1 online resourcce (xxii, 205 p.) :
1 490
Special publication / Natural Hazards Research and Applications Information Center ;
v no. 33
Description based on print version record.
Includes bibliographical references.
Flood damage prevention
z Colorado
Flood forecasting
2 710
University of Colorado, Boulder.
Natural Hazards Research and Applications Information Center.
i Print version:
t Twenty years later.
d Boulder, Colo. : Natural Hazards Research and Applications Information Center, University of Colorado, 1997
w (OCoLC)37902580
Special publication (University of Colorado, Boulder. Natural Hazards Research and Applications Information Center) ;
Natural Hazards Center Collection
4 856


.. r .. ...... 1 ,.' Twenty Years Later < . ," ':L ", Have ":"'learned' Since the ".,!.,.. 'f Big thompson flood .... '.. .. .. : .-, .. '. Proceedings of a Meeting Held in Fort Collins, Colorado July 13-15, 1996. > '.


TWENTY YEARS LA TER WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLOOD Eve Gruntfest Editor Proceedings of a Meeting Held in Fort Collins, Colorado July 13-15, 1996 Special Publication No. 33 Natural Hazards Research and Applications Information Center University of Colorado Boulder, Colorado


The opinions contained in this volume are those of the authors and do not necessarily represent the views of the funding or sponsoring organiza tions. The use of trademarks or brand these papers is not intended as an endorsement of any product. Published 1997. This volume is available from: The Natural Hazards Research and Applications Information Center Institute of Behavioral Science Campus Box 482 University of Colorado Boulder, CO 80309-0482 tel: (303) 492-6819 fax: (303) 492-2151 e-mail: WWW: http://www colorado. edu/hazards ii


TABLE OF CONTENTS Table of Contents . . . . . . . . . . . . . . .. iii Acknowledgments .............................. vii List of Abbreviations ............................ ix List of Participants . . . . . . . . . . . . . . . x INTRODUCTION AND OVERVIEW ................... 1 PART 1: FEDERAL PERSPECTIVE Barriers and Opportunities in Mitigation Richard W. Krimm ............................ 15 The Bureau of Reclamation and Dam Safety Howard Gunnarson ........................... 21 Flood Warning/Preparedness Programs of the Corps of Engineers Kenneth Zwickl .............................. 26 PART II: DAM SAFETY Olympus Dam Early Warning System David B. Fisher . . . . . . . . . . . . . . . 31 Dams, Defects, and Time Wayne J. Graham ............................ 40 1996 Willamette and Columbia River Flood Cynthia A. Henriksen .......................... 50 PART III: HUMAN DIMENSIONS OF DISASTER Emergency Communications: A Survey of the Century's Progress and Implications for Future Planning Bascombe J. Wilson ........................... 57 Coping Self-Efficacy Following Natural and Human-Caused Disasters Charles C. Benight . . . . . . . . . . . . . . 65 Church World Service and Lessons Learned for Mitigation Kristina J. Peterson and Richard L. Krajeski ........... 75


Table of Contents PART IV: METEOROLOGICAL CAPABILITIES AND CLIMATOLOGICAL ISSUES National Weather Service Advanced Capabilities in Flash Flood Forecasting Lee W. Larson 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 83 Comparison of Deficiencies Associated with the Big Thompson Flash Flood Event and Recent Flood Events in the Eastern United States Solomon Go Summer 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90 Climatology of Extreme Rain Events in the United States from Hourly Precipitation Observations Harold E. Brooks, David Jo Stensrud, and Daniel Vo Mitchell 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 103 The Flash Flood Forecaster Course at the National Weather Service Training Center: The Environmental Research Laboratories Component Harold Eo Brooks, Charles Ao Doswell III, Robert Ao Maddox, Dennis Ao Rodgers, and Barry Schwartz 0 0 0 0 0 0 0 0 0 0 0 0 0 111 PART V: WARNING SYSTEMS Caliente Creek ALERT Flood Warning System Audit Clark Farr and David C. Curtis 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 121 Evolution of Local Flood Warning Systems and Early Notification Procedures in Denver, Colorado Kevin Go Stewart 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 133 Putting Effective Flood Warning Systems in Place: The Process and Guidelines in Australia John Handmer and Chas Keys 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 141 PART VI: INTERNATIONAL EXPERIENCES Flash Floods in Mexico Mao Teresa Vazquez, Ramon Dominguez, Oscar Fuentes, and Jose Antonio Maza 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 153 Flooding and the Demise of the Moche Empire Kenneth Ro Wright and John Dracup 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 161 iv


Table of Contents PART VII: PALEOHYDROLOGICAL METHODS Problems with Use of Statistical Probability as a Tool for Prediction of Extreme Events Gregory G. Hammer . . . . . . . . . . . . .. 171 Bayesian Flood Frequency Analysis with Paleohydrologic Bounds for Late Holocene Paleofloods, Santa Ynez River, California Daniel R.H. O'Connell, Daniel R. Levish, and Dean A. Ostenaa ......................... 183 Paleohydrologic Bounds and the Frequency of Extreme Floods Daniel R. Levish, Dean A. Ostenaa, and Daniel R.H. O'Connell . . . . . . . . . . . .. 197 v




ACKNOWLEDGMENTS For 20 years I have been fortunate to have worked with an extraordi narily dedicated group of people devoted to lessening the impacts of flash floods. The 1996 Symposium re-emphasized that the Big Thompson flood had shaped the work of a generation of flood hazard mitigation experts. In this respect, the work of Bob Kistner, John Swanson, Larry Mooney, Robert Jarrett, Larry Stern, Wayne Graham, John Henz, Larry Larson, Kevin Stewart, Jack Truby, Len Boulas, and Patricia Hagan deserves special mention. Many people at the Federal Emergency Management Agency (FEMA) worked to develop the symposium. They particularly supportive on program ideas. John Swanson provided encouragement from the moment I mentioned the idea to him two years before the actual meeting. Mike Armstrong, Steve Olsen, Tony Mendes, Mary Ahlstrom, Karen Morman, Floyd Shoemaker, Norm Lizotte, and Jay Wilson worked tirelessly. Dick Krimrn from headquarters provided fine remarks at the opening session. Jim Knoy from EPA, and Erik Nilsson served on the steering committee and were very supportive. Fred Sibley and Bill Rakocy at the Colorado Office of Emergency Management helped throughout the planning process and particularly with the smooth operation of the audio visual equipment at the symposium. Tommy Greer, Ron Cattany, Jerry Smith, and Polly White also from OEM contributed significantly to the success of the Symposium. Marc Weber, Carol Foster, and Diana Buchanan from the University of Colo rado-Colorado Springs and Doug Leas of the University of Nebraska-Omaha were essential to the smooth mechanics of the meeting in Fort Collins. The 1986 Symposium had a wide variety of sponsors and significant funding. The 1996 Symposium, on the contrary, was funded by enthusiasm more than by dollars, meaning a great deal more work for volunteers. There was little money available for the brochure and for promoting the Sympo sium. The World Wide Web site for the Symposium was linked with the FEMA homesite, and many important participants learned of the meeting through the Web, bringing interest from all over the world in a way that traditional means might not. We even had two young newlyweds, the Raflo's, fit the Symposium into their planned Rocky Mountain honeymoon after they found our site on the Web. When some people asked why we needed another Symposium, Jerry Peterson of the U. S. Army Corps of Engineers encouraged planning for the


Acknowledgments meeting by pointing out that the Corps could have been intensively involved with warning efforts at the 1986 Symposium. His comment revealed that meetings can make a difference! He also arranged funding for the publication of these Proceedings. Bob Jarrett and Tom Yorke from USGS also gener ously provided some financial support. Lori Allen of Gallileo International is a consummate professional emergency manager. She may have sensed a looming disaster as the date of the Symposium approached. She leaped in with extraordinary organizational skills and managed the arrangements for the vendors (who were excellent additions to the Symposium) and worked closely with the hotel many crucial hours. Her dedication was essential to the success of the Symposium. The Saturday field trip through the Big Thompson Canyon added many dimensions to the Symposium experience; 30 people took part. Expert guidance from Bob Kistner, Bob Jarrett, Larry Stern, and particularly from Sharlynn Wamsley, a canyon resident, enabled us to recreate the flood stories and observe their lingering effects. Mary Fran Myers was a constant source of encouragement and assistance. Sylvia Dane did an excellent job assembling and editing the papers. Gilbert F. White unfortunately was unable to attend the Symposium, but his spirit and unwavering support for flood hazard mitigation efforts were present everywhere during the Symposium. Participants from five countries and 38 U. S. states joined to make this symposium a success. At the closing session on July 12th, when the notion of the 30th anniversary symposium was raised, I offered my support to the idea and turned the responsibility over to the next generation of flash flood mitigation specialists, including Dianne Brien, Pamela Pate, John England, and Patricia Gavelda. I look forward to working with them in 2006. Eve Gruntfest Symposium Organizer University of Colorado Colorado Springs, CO viii


ALERT AWHPS ASDSO ASOS AWIPS CFS CRP CSE CWA CWS EBS EMWIN ERL EWS E&SS FEMA FSL HPD ICOLD IFLOWS MAR NCDC NOAA NSSL NWR NWS QPF PAR PMF PMP LIST OF ABBREVIATIONS Automated Local Emergency in Real Time Area Wide Hydrologic Prediction System American Society for Dam Safety Officials Automated Surface Radar Observing System Advanced Weather Interactive Processing System Chronic Fatigue Syndrome Critical Rainfall Probability Coping Self-Efficacy County Warning Area Church World Service Emergency Broadcast System Emergency Managers Weather Information Network Environmental Research Laboratory Early Warning System Kern County Engineering and Survey Services Department Federal Emergency Management Agency Forecast Systems Laboratory Hourly Precipitation Data International Commission on Large Dams Integrated Flood Observing and Warning System Modernization and Associated Restructuring (National Weather Service) National Center for Climatic Data National Oceanic and Atmospheric Administration National Severe Storms Laboratory NOAA Weather Radio National Weather Service Quantitative Prediction Forecast Population at Risk Probable Maximum Flood Probable Maximum Prediction ix


LIST OF PARTICIPANTS Raul Acuna Arcus Data Security, Inc. P.O. Box 3785 Englewood, CO 80155 Christopher Adams CIRA Foothills Research Center Colorado State University Fort Collins, CO 80523 (970) 491-8448 e-mail: Mary Ahlstrom FEMA Region VIII P.O. Box 25267 Denver Federal Center Denver, CO 80225 (303) 235-4838 fax: (303) 235-4857 Larry Akers Div. of Disaster & Emergency Services 1100 N. Main Helena, MT 59604-4789 Lori Allen Galileo International 5350 S. Valentia Way Englewood, CO 80 III (303) 397-5731 fax: (303) 397-5199 Mike Armstrong FEMA Region VIII P.O. Box 25267 Denver Federal Center Denver, CO 80225 (303) 235-4812 fax: (303) 235-4849 x Rachel Badger URS Operating Services 1099 18th Street Denver, CO 80202 Nancy Barnett Western Insurance Information Services 6565 South Dayton Street Suite 2400 Englewood, CO 80111 Katie Baumann Assn. of Contingency Planners 100 Fillmore Street Denver, CO 80206 (303) 782-3082 fax: (303) 782-3544 Bob Baun Fort Collins Coloradoan Fort Collins, CO 80225 (970) 224-7742 Ken Beegles Colorado Division of Water Resources P.O. Box 1880 Durango, CO 81302-1880 Charles C. Benight Department of Psychology University of Colorado P.O. Box 7150 Colorado Springs, CO 80933-7150 (719) 262-3616 e-mail:


List of Panicipants Susan Bosworth Sociology Dept. College of William and Mary Williamsburg, VA 23187-8795 tel and fax: (804) 221-2390 e-mail: slbosw@malthus.morton.wm. edu Leonard A. Boulas Colorado Office of Emergency Management Camp George West 15075 South Golden Road Golden, CO 80401-3979 (303) 273-1622 Dianne Brien University of Wyoming Geology and Geophysics Laramie, WY 82071 e-mail: Harold Brooks NOAA/ERLINational Severe Storms Laboratory Norman, OK (405) 366-0499 fax: (405) 366-0472 e-mail: Tom Browning Colorado Water Conservation Board 1313 Sherman Denver, CO 80203 (303) 866-3441 Stan Brua U.S. Army Corps of Engineers Baltimore District Baltimore, MD (410) 962-4972 fax: (410) 962-4894 xi Diana Buchanan Geography Department University of Colorado P.O. Box 7150 Colorado Springs, CO 80933-7150 (719) 262-3513 fax: (719) 262-3019 Stan Bush Littleton Emergency Planning 2415 East Maplewood Avenue Littleton, CO 80121-2819 Antonio Cancelliere 1500 W. Plum Street, llB Fort Collins, CO 80521 e-mail: Dan Carlson FEMA Region VIII Mitigation Division Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267 Ron Cattany Culurado Department of Natural Resources Colorado Office of Emergency Management Camp George West 5075 South Golden Road Golden, CO 80401-3979 (303) 273-1775 fax: (303) 273-1795 Jan Christner URS Operating Services 1099 18th Street Denver, CO 80202 (303) 291-8284


Robert Clark Hydrology University of Arizona Tucson, AZ e-mail: Mark Crespin Rocky Mountain Catastrophe 5762 Lamar Street Arvada, CO 80002 David Crews P.O. Box 296 146 S. Grain Clearwater, KS 67026 Dave Curtis, Principal DC Consulting 9477 Greenback Lane, Suite 522A Folsom, CA 95630 (916) 988-2771 e-mail: Steve Denney Colorado Office of Emergency Management Camp George West 5075 South Golden Road Golden, CO 80401-3979 (303) 273-1622 Jim Disney Larimer County Commissioner Ft. Collins, CO (970) 498-7010 William Doal U.S. Army Corps of Engineers CEMRO-ED-HE 215 North 17th Street Omaha, Nebraska 68102-4978 (402) 221-4582 xii List of Participants Nolan Doesken Atmospheric Science Department Colorado State University Fort Collins, CO 80523 (970) 491-8545 e-mail: nolan@ulysses.atmos.colostate. edu Tom Donaldson Lower Colorado River Authority P.O. Box 220, S-501 Austin, TX 78767-0220 Charles Doswell III National Severe Storms Laboratory 1313 Halley Circle Norman, OK 73069 (405) 366-0439 fax: (405) 366-0492 Harry Dotson Hydrologic Engineering Center U.S. Army Corps of Engineers Davis, CA (916) 756-7718 e-mail: dotson@hec61.wrc-hec.usace. Rosalie Dukart Adams County Emergency Preparedness 450 South 4th A venue Brighton, CO 80601 (303) 289-5441 Donald Eddy 14185 West 21st Place Golden, CO 80401 Barbara Ellis FEMA Region VIII Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267


List of Participants John F. England Hydrologic Science and Engineering Colorado State University Ft. Collins, CO 80523 (970) 491-8395 e-mail: Ed Everaert U. S. Department of Interior Bureau of Reclamation 11056 West County Road #18E Loveland, CO 80537 David Fisher U. S. Bureau of Reclamation Denver Federal Center, Bldg. 67 P.O. Box 25007 (D-8470) Denver, CO 80225-0007 (303) 236-9000 fax: (303) 236-1070 Maureen Fordham Anglia Polytechnic University Geography Department East Road Cambridge CBl IPT, U.K. fax: 011 2236352973 Carol Foster 3829 Meadow Lane Colorado Springs, CO 80907 (719) 594-4183 1. M. Fritsch Department of Meteorology Pennsylvania State University University Park, PA 16802 fax: (814) 865-3663 e-mail: xiii Patricia Gavelda Colorado Office of Emergency Management 132 West B Street #260 Pueblo, CO 81003 (719) 544-6563 fax: (719) 545-1876 Michael T. Gelski Salvation Army 1370 Pennsylvania Street Denver, CO 80203 Robert T. Glancy National Weather Service 10230 Smith Road Denver, CO 80239 (303) 361-0661 fax: (303) 361-5508 Mark Gonzales Geography Department University of Denver Denver, CO 80208 Joyce Gordy Red Cross Representative to FEMA Region VIII P.O. Box 25267 Denver Federal Center Denver, CO 80225 (303) 235-4838 fax: (303) 235-4857 Wayne Graham U. S. Bureau of Reclamation Building 67, Denver Federal Center Denver, CO 80225-0007 (303) 236-0123 fax: (303) 371-5508


Tom Grier Colorado Office of Emergency Management Camp George West 15075 South Golden Road Golden, CO 80401-3979 (303) 273-1622 Mike Grimm Fort Collins Stormwater Utility 235 Mathews Street Fort Collins, CO 80522 (970) 224-6036 e-mail: Eve Gruntfest Department of Geography University of Colorado P.O. Box 7150 Colorado Springs, CO 80933-7150 (719) 262-3513 fax: (719) 262-3019 e-mail: Howard Gunnarson U.S. Bureau of Reclamation Building 67, Denver Federal Center Denver, CO 80225 (303) 236-9000 fax: (303) 236-1070 Patricia Hagan U.S. Bureau of Reclamation Building 67, Denver Federal Center Denver, CO 80255-0007 (303) 236-9000 ext. 265 fax: (303) 236-1070 xiv Gregory Hammer Dam Safety Branch List of Participants Division of Water Resources 800 8th Avenue Room 321 Greeley, CO 80631 (970) 659-0259 fax: (303) 659-0579 e-mail: John Handmer Flood Hazard Research Centre Middlesex University Queensway, Enfield EN3 4SF e-mail: Michael Hedberg 3289 Prospector Drive Casper, WY 82604 (307) 577-0196 Cynthia Henriksen North Pacific Division U.S. Army of Engineers P.O. Box 2870 Portland, OR 97208-0270 (503) 326-3745 e-mail: cynthia.a.henricksen@npdOl.usace. John Henz Henz Meteorological Services 2480 W. 26th Street, Suite 310B Denver, CO 80211 (303) 458-1464 fax: (303) 458-5309 Ed Herring Weld County Office of Emergency Management 910 Tenth Avenue Greeley, CO 80631-3873


List of Participants Mike Hittensburg FEMA Region VIII Denver Federal Center Denver, CO 80225 (303) 235-4812 Dave Holm Colorado Office of Emergency Management Camp George West, Building 120 15075 South Golden Road Golden, CO 80401-3979 (303) 273-1622 Joan Hopkins Centennial Chapter Red Cross 3105 Swallow Place Ft. Collins, CO 80525 (970) 282-0024 Brian Hyde Colorado Water Conservation Board 1313 Sherman Denver, CO 80203 (303) 866-3441 Paullwai U.S. Army Corps of Engineers Missouri River Division 12565 W. Center Road Omaha, NE 68144 Garrett Jackson Woodward-Clyde Consultants Stanford Place III, Suite 1000 4582 South Ulster Street Denver, CO 80237 (303) 740-2600 Steve Jamieson GEl Consultants 5660 Greenwood Plaza Boulevard Suite 202 Englewood, CO 80111 xv Robert Jarrett U.S. Geological Survey Box 25046, MS 412 Denver, CO 80225 (303 )236-6447 fax: (303) 236-5034 e-mail: Lyne Johnson 3289 Prospector Drive Casper WY 82604 (307) 577-0196 Jonathan M. Kelly Wright Water Engineers, Inc. 2490 West 26th Street Denver, Co 80211 (303) 480-1700 fax: (303) 480-1020 Matthew Kelsch NOAA/Forecast Systems Lab R/E/FSI 325 Broadway Boulder, CO 80303 (303) 497-6719 fax: (303) 497-7262 e-mail: Selma Kessler 3289 Prospector Drive Casper, WY 82604 (307) 577-0196 Robert Kistner Kirstner & Associates 750 South Kline Way, Suite 100 Lakewood, CO 80226-3923 (303) 985-1837 fax: (303) 985-1186 e-mail:


Robin Knappe American Red Cross 444 Sherman Street Denver, CO 80203 Michael Knipps U.S. Bureau of Reclamation Box 36900 316 North 26th Street Billings, MT 59107 (406) 247-7630 fax: (406) 247-7993 Jim Knoy Environmental Protection Agency 999 18th Street, Suite 500 Denver, CO 80202 (303) 312-6071 fax: (303) 312-6838 e-mail: knoy.j Gordon Knuckey Colorado Voluntary Organizations Active in Disaster 410 South High Street Denver, CO 80209 (303) 733-8742 tax: (303) 765-5971 Richard Krajeski Church World Service 114 High Street Mannington, WV 26582 (304) 986-1614 fax: (304) 986-3099 Darrel Kranse Bureau of Reclamation P.O. Box 250007, D-55oo Denver, CO 80226 xvi Dick Krimm FEMA List of Participants 500 C Street, S.W. Washington, DC 20472 (202) 646-3692 fax: (202) 646-4060 Howard Kutzer U S. Department of Housing and Urban Development 633 17th Street First Interstate Bank Denver, CO 80202 Larry Larson Assn. of State Floodplain Managers 4233 W. Beltline Highway Madison, WI 53701 (608) 266-1926 fax: (608) 264-9200 e-mail: Lee Larson 8879 Juniper Prairie Village, KS 66207 (301) 713-0619 fax: (301) 713-0963 e-mail: Douglas P. Leas University of Nebraska 1301 Avenue A Plattsmouth, NE 68048 (402) 296-6110 e-mail: Dan Levish U.S. Bureau of Reclamation P.O. Box 25007 D-8330 Denver, CO 80225 (303) 236-4195 x274 e-mail:


List of Participants John Liou FEMA Region VIII Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267 (303) 235-5995 fax: (303) 235-4857 Norman Lizotte FEMA Region VIII Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267 (303) 235-5995 fax: (303) 235-4857 Scott Logan FEMA Region VIII Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267 (303) 235-5995 fax: (303) 235-4857 Linda MacIntyre City of Boulder Public Works 1739 Broadway, Suite 415 P.O. Box 791 Boulder, CO 80306 Patricia Martin Adams County Emergency Preparedness 450 South 4th A venue Brighton, CO 80601 (303) 289-5441 Jon Mason U.S. Geological Survey 2617 E. Lincoln Way, Suite B Cheyenne, WY 82001-5662 (307) 778-2931 ext. 2717 fax: (307) 778-2764 e-mail: xvii David McComb Department of History Colorado State University Ft. Collins, CO 80523 John McKay FEMA 16825 South Seton A venue Emmitsburg, MD 21727 (301) 447-1286 fax: (301) 447-1589 Karen Medde City of Boulder 1739 Broadway, Suite 300 P.O. Box 791 Boulder, CO 80306 Tony Mendes FEMA Region VIII Denver Federal Center Denver, CO 80267 (303) 235-4790 fax: (303) 235-4857 Fred Metzler FEMA Region VIII Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267 Kirk Miller U.S. Geological Survey 2617 E. Lincoln Way, Suite B Cheyenne, WY 82001-5662 (307) 778-2931 ext. 2716 fax: (307) 778-2764 e-mail: Richard Minor Loveland Fire Department 410 East 5th Street Loveland, CO 80532 (970) 962-2497


Larry Mooney National Weather Service 10230 Smith Road Denver, CO 80239 (303) 361-0661 fax: (303) 361-5508 Richard Moore Parsons Brickerhoff 1660 Lincoln Street, Suite 2000 Denver, CO 80264 (303) 832-9091 fax: (303) 832-9096 Karen Morman FEMA Region VIII Denver Federal Center P.O. Box 25267 Denver, CO 80225-0267 (303) 235-4831 fax: (303) 235-4857 Burt Morrison U.S. Geological Survey 2617 E. Lincoln Way, Suite B Cheyenne, WY 82001-5662 Virginia Motoyama FEMA Region VIII Denver Federal Center, Bldg. 710 P.O. Box 25267 Denver, CO 80225-0267 Mary Fran Myers Natural Hazards Research and Applications Information Center University of Colorado Campus Box 482 Boulder, CO 80309 (303) 492-2150 fax: (303) 492-2151 e-mail: xviii List of Participants Erik Nilsson Larimer County Emergency Management P.O. Box 1190 Fort Collins, CO 80522 (970) 498-5310 fax: (970) 493-2795 e-mail: Dan 0' Connell U.S. Bureau of Reclamation Denver Federal Center P.O. Box 25007 D-8330 Denver, CO 80225 (303) 236-4195 ext. 275 e-mail: Steve Olsen FEMA Region VIII Denver Federal Center, Bldg. 710 P.O. Box 25267 Denver, CO 80225-0267 (303) 235-4814 fax: (303) 235-4849 Dean Ostenna U.S. Bureau of Reclamation Denver Federal Center P.O. Box 25007 Denver, CO 80225 Park Owens Pennington County Emergency Management 315 St. Joseph Street, B-31 Rapid City, SD 57701 e-mail: Pamela Pate Department of Geography University of Texas at Austin Austin, TX e-mail:


List of Participants Pete Peterka Community Alert Network 301 Nott Street Schenectedy, NY 12305-1039 Kristina Peterson Church World Service 114 High Street Mannington, WV 26582 (304) 986-1614 fax: (304) 986-3099 Thomas A. Pick 5047 South Yank Court Morrison, CO 80465 Alan Raflo Virginia Tech Museum of Natural History 428 North Main Street Blacksburg, VA 24061-0542 (540) 231-5307 e-mail: Bill Rakocy Colorado Office of Emergency Management Camp George West 15075 South Golden Road Golden, CO 80401-3979 (303) 273-1774 Brad Radstall Water Resources Engineer GEl Consultants 5660 Greenwood Plaza Boulevard Suite 202 Englewood, CO 80111 Kristy Ray American Red Cross 2100 Sandstone Drive Ft. Collins, CO 80524 xix Jim Redmond 9633 Silver Fox Cove Memphis, TN 38133 (901) 678-2852 Paul Ridlen Woodward-Clyde Consultants Stanford Place III, Suite 1000 4582 South Ulster Street Denver, CO 80237 (303) 740-2600 Bill Rogers Woodward Clyde Stanford Place III Suite 1000 4582 South Ulster Street Denver, CO 80237 (303) 740-2600 Beth Roman Denver Water 1600 West 12th Avenue Denver, CO 80254 (303) 628-6207 Jack Rose Omaha District Corps of Engineers 215 North 17th Street Omaha, NE 681022-4978 (402) 221-4148 Jose Salas Hydrologic Science and Engineering Colorado State University Ft. Collins, CO 80523 (970) 491-8395 fax: (970) 491-8671 Amy Schickentanz 3289 Prospector Drive Casper, WY 92604 (307) 577-0196


Glen D. Schlueter Ft. Collins Stormwater Utility 235 Mathews Ft. Collins, CO 80522 William Schneider Vestige Press 908 Lockview Court Fort Collins, CO 80524 Floyd Shoemaker FEMA Region VIII P.O. Box 25267 Denver, CO 80225-0267 (303) 235-4929 fax: (303) 235-4857 Mark Shotkaski Parsons Brinckerhoff 1660 Lincoln Street, Suite 2000 Denver, CO 80264 (303) 832-9091 fax: (303) 832-9096 Fred Sibley Colorado Office of Emergency Management Camp George West 15075 South Golden Road Golden, CO 80401 (303) 273-1622 Tracy Carter Sondeen FEMA Region VIII Denver Federal Center, Bldg. 710 P.O. Box 25267 Denver, CO 80225-0267 (303) 235-4994 fax: (303) 235-4857 e-mail: xx List of Participants Jim Soule Colorado Geological Survey 1313 Sherman, Room 715 Denver, CO 80203 (303) 866-2611 fax: (303) 866-2461 Mary Starr Colorado Office of Emergency Management 132 West B Street, #260 Pueblo, CO 81003 (719) 544-6563 fax: (719) 545-1876 Gary Stenson Nightengale Messaging 475 17th Street, Suite 920 Denver, CO 80202 Larry Stern Boulder County Sheriffs Department 1805 33rd Street Boulder, CO 80301 (303) 441-3637 Kevin Stewart Urban Drainage and Flood Control District 2480 West 26th Avenue, Suite 156-B Denver, CO 80211 (303) 455-6277 fax: (303) 455-7880 LaRue Stivers FEMA Region VIII Denver Federal Center, Bldg. 710 P.O. Box 25267 Denver, CO 80225-0267


List of Participants Chandran Subramaniam CIRA/ERL/FSL 325 Broadway RlE/FS5 Boulder, CO 80303 (303) 497-6015 fax: (303) 497-6301 David Sullivan Office of Emergency Services City of Denver Denver, CO 80202 Solomon Summer National Weather Service Eastern Regional Office 630 Johnson Avenue Bohemia, NY 11716 (516) 244-0111 fax: (516) 244-0167 e-mail: ssummer@smptgate.ssmc.noaa. gov John Swanson FEMA Region VIII P.O. Box 25267 Denver Federal Center Denver, CO 80225-0267 Alan Taylor City of Boulder 1739 Broadway, Suite 300 P.O. Box 791 Boulder, CO 80306 William Thomas Colorado State Patrol Fort Collins, CO Robert Tibi National Weather Service P.O. Box 1122 Salt Lake City, UT 84110 xxi Ed Tomlinson ENFO, Inc. P.O. Box 680 Monument, CO 80132-0680 (719) 488-9117 fax: (719) 477-9118 e-mail: Jack Truby 321 Cook Street Denver, CO 80206 (303) 377-8474 Donna Tucker FEMA Region VIII P.O. Box 25267 Denver Federal Center Denver, CO 80225 (303) 235-4838 fax: (303) 235-4857 Don Van Wie DIAD Inc. 780 Juniper Avenue Boulder, CO 80304 e-mail: Teresa Vasquez National Center for Disease Prevention Del. Coyoacan, Mexico D.F. 04360 (525) 606-9350 fax: (525) 606-1608 e-mail: Mark Wallace Golden Fire Department 911 10th Street Golden, CO 80401


Sharlynn Wamsley Big Thompson Canyon Association 1501 South County Road 23E Berthoud, CO 80513 (970) 532-3898 e-mail: Marc Weber Geography Department University of Colorado P.O. Box 7150 Colorado Springs, CO 80933-7150 (719) 262-3513 fax: (719) 262-3019 e-mail: Polly White Colorado Office of Emergency Management 15075 South Golden Road Camp George West, Bldg. 120 Golden, CO 8040 1 (303) 273-1622 John Wiener 875 33rd Street Boulder, CO 80303 e-mail: Chris Williams Woodward-Clyde Consultants Stanford Place 3, Suite 1000 4582 South Ulster Street Denver, CO 80122 (303) 740-2600 Don Wilmes Data Transmission Network 9110 West Dodge Road, Suite 200 Omaha, NE 68114 (402) 390-2328 fax: (402) 390-7188 xxii List of Participants Bascombe Wilson U.S. Air Force Denver Federal Center FEMA Region VIII Denver, CO 80225 (303) 752-4135 e-mail: Sandra Wilson Spenser Curtis Foundation 524 North Tejon Colorado Springs, CO 80903 (719) 630-8212 Bob Wold Colorado Office of Emergency Management 15075 South Golden Road Camp George West, Bldg. 120 Golden, CO 80401 Tom Yorke U.S. Geological Survey Reston, VA Sherryl Zahn FEMA Region VIII Denver Federal Center, Bldg. 710 P.O. Box 25267 Denver, CO 80225-0267 Kenneth Zwickl U.S. Army Corps of Engineers 20 Massachusetts Avenue, N.W. Washington, DC 20314 (202) 761-0169


INTRODUCTION AND OVERVIEW WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLOOD Eve Gruntfest University of Colorado-Colorado Springs Introduction The Big Thompson Canyon The Big Thompson Canyon is one of the most scenic in the Rocky Mountain region. U.S. Route 34 runs through the canyon, adjacent to the river in many spots, and it is the main link between the Colorado plains and Rocky Mountain National Park. In June 1976, just before the flood, the full time canyon population was estimated to be 600 and part-time residents numbered approximately twice that. Also, an undetermined number of tourists were present, attracted by trout fishing, Rocky Mountain National Park, stream-side motels, and campgrounds (Gruntfest, 1977). Three major communities reside in the 25-mile canyon that runs west from Loveland, Colorado: 1) Cedar Cove, which lies just above the Narrows; 2) Drake, the largest town, located at the confluence of the North Fork and the Main Fork of the Big Thompson; and 3) Glen Comfort, a smaller town on the North Fork of the Big Thomp son. The Flood On July 31, 1976, the Big Thompson Canyon was filled with residents and visitors. It was the Saturday of the weekend commemorating Colorado's Centennial and the last holiday weekend before the start of the school year. That night a flash flood ravaged the canyon, causing the worst natural disaster, in terms of documented lives lost, in Colorado state history. The death toll of a 1921 flood in Pueblo may have been larger; however, estimates of lives lost in that flood range from 100 to 350. Heavy rain fell over a 70-square-mile area in the central portion of the Big Thompson watershed between 6:30 and 11:00 p.m. The most intense rainfall,


2 WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLOOD between 12 and 14 inches, fell on slopes in the western end of the canyon. The flood washed out all stream and rain gages, so accurate measurements were not possible. Yet, the impact of the flood could have been worse. The North Fork peak streamflow occurred approximately 40 minutes later than the Main Fork peak. If the two peaks had coincided, the peak streamflow would have been even greater than the 31,200 cubic feet per second recorded at the mouth of the canyon. At least 139 people died in the flood, and eighty-eight people were injured. Seven people were listed as missing. The flood destroyed 316 homes, 45 mobile homes, and 52 businesses. Seventy-three mobile homes suffered major damage. The Symposium Idea After the Big Thompson flood, everyone concerned with natural hazards resolved that a disaster of this magnitude should never happen again. This resolve was particularly strong in Boulder, Colorado, where officials realized that they faced a worse catastrophe if a Big Thompson-like storm materialized over the Boulder Creek drainage area. Downtown Loveland is four miles from the mouth of the Big Thompson Canyon and was basically unaffected by the Big Thompson flood. Downtown Boulder, however, lies directly at the mouth of Boulder Canyon. Flash flood hazard awareness following the Big Thompson flood was high, especially since this flood occurred only four years after 237 lives were lost in Rapid City, South Dakota. These two events re-focused official attention on flash floods, particularly in the western United States. Ten Years After In the decade following the Big Thompson Canyon flood, many scientific, technological, and educational advances took place. The idea for the first symposium grew out of our interest in evaluating the notion of "disaster as opportunity." In what ways had we learned from the Big Thompson catastrophe? Were we more or less vulnerable and in what ways? These questions had many facets. After a major disaster, communities and governments pay scientific and policy attention to the seriousness of natural disasters. We had a catastrophic flash flood in Colorado in 1976-what difference has it made? The 1986 meeting, held in Boulder and sponsored by the National Science Foundation, the U.S. Army Corps of Engineers, and the National Oceanic and Atmospheric Administration, brought together 125 professionals with diverse perspectives to identify the areas of progress and lack of progress.


Gruntfest 3 The participants crossed many disciplinary boundaries and included forecasters, hydrologists, sociologists, geomorphologists, local civil defense officials, water engineers, members of the insurance industry, lawyers, and geographers. The symposium provided a rare opportunity to look back and assess the strengths and weaknesses of post -disaster research and policy actions. The post-audit provided a real event and time period focus for reflection on the commonly held premise that disasters are opportunities for change to reduce losses in the future. In 1986, after we spent a day-and-a-half discussing advancements and disappointments, we met to make research and policy suggestions. Findings from the Tenth Anniversary Symposium The discussion sessions following two days of the symposium led to the five specific recommendations. (More detail is available in the 1986 proceedings volume.) First, the need for transferring available flash flood hazard mitigation information was identified as greater than the need for the acquisition of new data. Second, the definition of "publics" for public awareness needed to be broadened. Third, better techniques for estimating costs, benefits, and losses were required so that accurate evaluations of mitigation strategies could be made. Fourth, the public and private sectors must work in a coordinated fashion to resolve important issues such as "how safe is safe enough" with regard to dam projects. And, fmally, the distinctions between flash floods and slow-rise floods must be recognized and clarified (Gruntfest, 1987). Twenty Years After In 1996, 10 years after the first symposium and 20 years after the flood, the initial question of vulnerability remained. Vulnerability to flash floods was increasing by virtue of the vast increases in population in the southwest ern U.S. Debris flows, mudslides, and alluvial fan flooding were all causing more damages as more people moved into hazardous areas. Flash flood deaths have not declined. A major policy dilemma continues to be how to get people to abandon their cars and climb to safety in flash floods. The public underestimation of the power of flowing water prevails. The 1996 symposium's mission followed our earlier meeting as a post audit to the flood and evaluation of lessons learned. As a group we again represented a wide range of disciplines, including meteorology, paleo hydrology, psychology, emergency management, geography, and floodplain


4 WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLOOD management. We also included a large representation from the press and residents from Big Thompson Canyon. As in 1986, professionals took the time to critically examine and learn from past events and to talk with people from different areas of expertise. Papers presented offered lessons learned from experiences in Larimer County, the Front Range, and other parts of the world, including West Virginia, California, England, Mexico, and Italy. Nationally, we have not had a major flash flood, in terms of loss of life, since the Shady Side, Ohio, disaster on June 14, 1990, when 26 people died. However, on a global scale, at the close of the 20th century the loss of lives from flooding continues. During the first seven months of 1996 alone, flash floods killed over 1,000 people around the world, many hundreds in flooding in southeast China. The State of the Flash Flood Hazard in 1996 Vulnerability is Increasing There is nothing unique about the Big Thompson Canyon in terms of its vulnerability to a severe flash flood. There are several other Front Range canyons just as vulnerable to similar or worse catastrophes. Vulnerability is increasing as population swells. The best detection system will not save lives unless the messages are delivered in a timely fashion and the people at risk know what to do-and do it promptly. We have been lucky that the Big Thompson flood was the last flood to kill more than 100 people in the U. S. but it is not due to our wise land-use decisions. There are increased possibilities for compounding the impacts of natural disasters by the co-location of hazardous materials in floodplains. For example 20,000-gallon propane tanks were found in the Missouri River floodplain during the 1993 floods in the Midwest (Gruntfest and Pollack, 1994). The hazardous materials question as well as the issue of dam safety must be addressed. Our catastrophe potential increases as infrastructure continues to age. People still die in their cars crossing flooded roads because they still don't want to get out of their vehicles. How can we convince people that they are better wet than dead? Unfortunately, the media and our own public awareness documents and videotapes reinforce the notion of relative safety through images of people being dramatically rescued from their cars while flood waters rush around them. Even when roads are closed, people drive around barriers. In the Susquehanna River floods in New York in January 1996, 30 people received


Gruntfest 5 tickets after the police posted signs indicating that the road was closed. The police actually had to stand in the water in waders and give people tickets for crossing the flooded, barricaded road! Importance of Low Tech Measures and Environmental Cues During the 1993 Mississippi/Missouri River floods, in spite of advanced hydrological and meteorological models, we were quite dependent on low tech adjustments such as sandbags and local knowledge. Data from the sophisticated hydrologic models from the National Weather Service, private meteorologists, and the Corps of Engineers were all constantly available and the local people kept tabs on them. But, there were still numerous difficulties with timeliness. People needed to know what the impacts of the five inches of rain currently falling in Kansas City would be at downstream Hermann, Missouri, later in the day. Or, what if an upstream levee did not hold? The most accurate information for the person who needed to know whether or not he or she would be able to cross the Missouri to get to work was provided by local knowledge and a very low tech measurement device: a measurement stick in the river that was checked by people sitting on a bench by the river or by word of mouth passed along by the road department employee who stopped at the cafe for a cup of coffee. While the Mississippi flooding experience was radically different from the Big Thompson flood in terms of lead time, the crucial roles of environmental cues, common sense, and local knowledge are as important in our high tech environment as they were 50 years ago. For the thousands of people who enjoy the beauties of flash flood-prone canyons, common sense remains essential. They must interpret the environ mental cues of a river getting louder, getting closer to the bank, and the rain falling harder than usual, and then abandon their cars and climb to safety. New Scientific Collaborations Flash flood information is available also from some unusual partnerships. Remote sensing efforts at the National Aeronautics and Space Administration (NASA), combined with the work of geographers at Dartmouth College, keep an up-to-the-minute archive of global flood events on the World Wide Web ( archive has unlimited potential for keeping up with events for educational purposes and providing lessons from flood experiences elsewhere. Before this website was


6 WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLOOD developed, students had much more difficulty keeping up with disasters around the world. Updated daily, the floods are graphically available through remote sensing and text is drawn from all the major news services around the world. The Dartmouth effort reduces the tendency toward nationalistic myopia that affects many people in the U.S. who rely mostly on newspapers that focus on events within our borders, rather than on those in the rest of the world. From the vantage point of paleohydrology, a discipline that was just beginning in 1986, it is suggested that the recurrence interval of the Big Thompson flood is 10,000 years. This means it was an event so rare that planning for the next one is impractical. Yet, there is still strong debate over the likelihood of another rainstorm of greater than 12 inches in the Colorado high country. How can we best communicate to populations at risk without confusing them? More Public Awareness One year after the Big Thompson flood, signs were placed at the entrances of Front Range canyons in Colorado with the purpose of giving people information on appropriate actions to take during the next major rainstorm that causes serious flooding. The signs are based on events in the Big Thompson Canyon on the night of July 31, where many people died trying to out-drive the flood. Other places have adopted the sign idea. The Arizona Flood Plain Managers and the National Weather Service have each developed videotapes aimed at reducing the number of people who drive through flooding roads, and consequently, the number of casualties. Findings from 1996 Symposium Twenty years ago the Corps of Engineers and the Bureau of Reclamation were leaders in flood control-mostly through dam construction and levee building. In 1996, they were lead actors in the flood detection realm, deeply involved in detection and warning systems for flash flood mitigation below dams and in floodplains. The National Weather Service was the only group making weather predictions in 1976. By 1986 private meteorologists were offering services to communities, corporations, and television stations. By 1996 the comple mentarity of public and private services seems nearly seamless, especially when users surf the Internet. Problems of flood forecasting are so complex that numerous actors can be involved and not duplicate their efforts. Also,


Gruntfest 7 vendors played an active role at the 1996 Symposium, which would have been nearly unimaginable in 1976 and 1986. There are thousands of Web sites instantly available. The National Weather Service reserves the legal obligation for forecasts, although oth professionals and nonprofessionals now have many more options for obtaining weather information. The Association of State Floodplain Managers, the Association of State Dam Safety officials, Automated Local Emergency in Real Time (ALERT) user groups, and numerous growing professional emergency management organizations are the essential creative groups committed to flash flood hazard mitigation. The partnerships between the ALERT users in many states and flood control districts and government agencies are flourishing, with excellent potential for improving the likelihood that timely warnings are received. ALERT systems are also being used for air pollution monitoring, fire weather forecasts, and water supply decision making. Since the Big Thompson flood, the need for detection and warning systems has been identified and acted upon. Twenty years ago, there were no automated stream and rain gage networks. Now, there are thousands, and they are not only accessible from central base stations at fire departments or emergency management offices, but those of us with modems and computers can keep abreast at home. Finally, in many cases, detection is being combined with the crucial elements of response. The reduced expense of personal computers and the increased speed of data transfer have radically altered the availability of data on real time river basin and rainfall. For example, in Maricopa County, Arizona, which includes Phoenix, residents and flood control engineers have access to real time stream and rain-gage data 24 hours a day (http://www.maricopa.govlflood/fcd.html). Individuals along many rivers in California can access Web sites that monitor stream gages to determine whether or not to evacuate (http://wwwdrw. Real-time radar access is also readily available. These sites were well visited during 1995, 1996, and 1997 floods and increased use can be expected as more people join the Web. Local emergency managers report that the Web data serves a vital public education purpose and significantly reduces phone traffic at emergency operations centers. In addition, NOAA's Forecast Systems Laboratory has developed a prototype where real time weather information is fully integrated in a geographical information system ( emwdp/emwdp.html). So far it is only available in Boulder, Colorado, but has great promise for applications elsewhere.


8 WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLOOD Introduction to the 1996 Proceedings Papers As with all proceedings, the papers in this volume represent a moment in time -20 years after a disastrous flood. It is interesting to compare the moments in time-to see how our professional perspectives have changed since 1976. The presentations from the 1996 symposium differ significantly from the 1986 contributions. The differences can be summarized in categories that are discussed briefly here and can be more clearly seen through the proceedings papers themselves. 1. The optimism that technological innovations would reduce or eliminate the need for traditional stream or rain gages has faded. The hope prevalent through the 1980s has been replaced by a more realistic recognition that technological innovations must be complemented by "old fashioned" rain and stream gages and that ground truthing is still essential 2. Paleohydrologic techniques are more frequently being applied to decisions about dam safety. Ten years ago conventional hydrology was beginning to see a challenge from paleohydrologists. However, there are improvements in methodology and an increased number of applied case studies including, the work by Robert Jarrett of the U.S. Geological Survey and Mike Grimm of the Fort Collins Stormwater Utility. Reduced federal funding for all types of projects has acceler ated the use of the new technique, which will ultimately reduce the safety requirements for large dams because paleohydrologists argue that traditional methods set unreasonably high expectations of flood flows. This debate continues with significance for flash flood forecast ing and dam safety requirements. 3. The 1996 symposium was unusual in several respects: First, it brought together a wide variety of people united by one disaster that occurred 20 years ago. It was striking how many careers and lives were deeply affected by that one event in Colorado. And, because the meeting is a special event-taking place every 10 years and providing a unique composition of participants-we experienced some very moving moments. In particular there was a fascinating interaction between the people of the Big Thompson Canyon and John Rold, who served with the Colorado Geological Survey as state geologist for many years before and after the


Gruntfest 9 flood. John tried very hard to explain why he made the recommendations for returning the road to its present site. Even 20 years after the event, the poignancy of the interaction indicated that the residents and the geologist remembered the moment as if was yesterday. Both spoke from their hearts regarding the disaster and the land use decisions that followed. Howard Gunnarson from the Bureau of Reclamation listened intensely as Theresa Vasquez reported on flash floods in Mexico, including the story of the flooding the same year as the Big Thompson flood (1976) that killed 2,000 people. He remarked that the Bureau must take into account how their dams and policies concerning their dams impact our neighbors to the South. Daunting Questions Tragically, on July l3, 1996, the day of the symposium field trip, three people were killed in a flash flood on Buffalo Creek in Colorado. The vegetation on the headwater area of Buffalo Creek about 140 miles southwest of the Big Thompson Canyon had been destroyed by a fire two weeks earlier. The lack of vegetation intensified the impacts of the heavy rain. Several homes were washed away in the small drainage and there were no official flash flood warnings in effect before the flood. It wasn't until dawn on August I, 1976, that the world realized that one thunderstorm killed 140 people and destroyed so much property. The next time more than 12 inches of rain falls at the top of a watershed in Colorado, New Mexico, Arizona, West Virginia, or elsewhere, will there be a timely official forecast? Will the campers, motel owners, homeowners, and motorists heed the warnings and do the right thing by climbing to higher ground? And, as we near the end of the 20th century, 20 years after the Big Thompson disaster, will the knowledge of the extraordinary amount of rainfall, the number of people killed, the extent of the destruction be known only the next day as it was on August 1, 1976?


10 WHAT WE HAVE LEARNED SINCE THE BIG THOMPSON FLoOD References Arizona Flood Plain Managers 1995 Flash Floods: A Warning to Beware. Video. Gruntfest, E. and D. Pollack 1994 "Warnings, Mitigation, and Litigation: Lessons for Research from the 1993 Floods." Update Water Resources 95 (Spring) :40-45. Gruntfest, E., ed. 1987 What We Have Learned Since the Big Thompson Flood Proceed ings of the Tenth Anniversary Conference, July 17-19, 1986. Special Publication No. 16. Boulder, Colorado: Natural Hazards Research and Applications Information Center, University of Colorado. Gruntfest, E. 1977 What People Did During the Big Thompson Flood. Natural Hazards Research and Applications Information Center, Working Paper No. 32. U.S. Department of Commerce, National Oceanic and Atmospheric Administration 1996 Hidden Danger: Low-Water Crossings. Video. National Weather Service Office of Hydrology. Note: A comprehensive bibliography on flash flooding has also been prepared and is available electronically through the Natural Hazards Research and Applications Information Center.


This page IS intentionally blank




This page IS intentionally blank


BARRIERS AND OPPORTUNITIES IN MITIGATION Richard W. Krimm Acting Associate Director for Mitigation Federal Emergency Management Agency (FEMA) One thing the Big Thompson flood demonstrated to our nation is that we as a society-at the local, state, and federal levels-need to work to reduce the risks we face from flooding and other types of disasters. Many of our communities face a risk similar to that experienced in the Big Thompson Canyon. And, unless we work together to implement mitigation measures to reduce or eliminate the risks we face, the Big Thompson flood will be repeated elsewhere. This argument seems to make perfect sense. Yet, while the concept of mitigation seems simple enough, in the real world it has always been a tough sell. The reasons for this can be grouped into four major categories: 1) a poor understanding of what mitigation is and how it can benefit people, 2) a lack of resources at the federal and state levels, 3) the need to quantify the savings associated with implementing cost-effective mitigation opportunities, and 4) the resistance by U.S. citizens to restrictions on the use of their land. Understanding Mitigation First of all, most people's knowledge and understanding of mitigation is limited at best. If we asked people what "mitigation" means, we would find that most Americans don't even understand the word, let alone how mitigation can benefit them and their communities. The problem is particularly vexing because mitigation is not a difficult concept to grasp. As Morrie Goodman, the Federal Emergency Management Agency's (FEMA) Director of Strategic Communications, said, "It's investing in a guardrail at the top of a mountain so you don't need to call an ambulance to the bottom." Mitigation involves protecting individuals, families, and communities from disaster. It's as simple as that, but still the problem persists.


16 BARRIERS AND OPPORTUNITIES IN MITIGATION Much of the difficulty, I believe, lies in the timing of when most mitigation activity commonly occurs. Mitigation requires planning, and as such, does not take place quickly after a disaster. Consequently, these activities are not as exciting or visible as disaster response and recovery. When the disaster is news, attention is focused on saving lives and showing images of damages. Little or no attention is given to why some homes and buildings survived the event due to strong building codes or the investment in other mitigation measures. Only weeks and months later, when the flood or earthquake has been overtaken by more current events, does mitigation get any attention. And by then, it's often too late to grab people's attention and communicate the mitigation message. It's also a problem related to the short-term focus that many Americans have, and the inherent difficulty in getting people to recognize the long-term risks they face. For example, I often have difficulty planning for next week, let alone the next 50 years. The same is true for people living in high-hazard areas. Many of those who live in the floodplain or along an earthquake fault tend to downplay the risks they face unless such an event is fresh in their minds. As a result, most people prefer to buy a new television or car rather than invest in mitigation measures that mayor may not be needed for years. To combat these problems and get people to start thinking more about protecting themselves, FEMA Director James L. Witt has moved aggres sively to increase people's knowledge about mitigation and provide incentives to help them protect themselves. He never gives a speech or testifies before Congress without discussing the advantages of mitigation. After seeking input from local, state, and federal officials nationwide, FEMA recently published the National Mitigation Strategy, which clarifies and outlines where we as a nation are headed in reducing risk. Weare invigorating our efforts to work with the private sector, private nonprofit groups, and the news media to spread the word about mitigation and seek their assistance in working with local communities. Mr. Witt even has spoken with the president and members of his Cabinet to identify ways federal agencies can more comprehensively promote hazard mitigation. But while this has produced real results, we as a community need to more clearly explain mitigation, or perhaps even develop a new term that is more easily understood by the average American. We also need to redouble our efforts to evaluate the costs and benefits of mitigation measures, and communicate their effectiveness in quantifiable terms. This last area is one that can be of great help to the emergency manage ment community. At this time, there continues to be little research targeted toward evaluating the true costs of disasters as well as the cost-effectiveness


Krimm 17 of mitigation measures. This information is critical to any effort to market mitigation to people and their elected officials. It would also provide the federal, state, and local governments with the ammunition they need to convince people that a little mitigation before a disaster is much more cost effective than cleaning up the mess after a disaster occurs. Limited Resources Now to the second constraint to implementing hazard mitigation: an extremely limited pool of resources. Ever-shrinking budgets and competing priorities describe the rules of the game we now play at FEMA, and the situation isn't much better at state and local levels. We have entered an era of diminishing government, and the cause of mitigation hasn't been spared the cut of the budget ax. The pie is getting smaller, and FEMA's ability to fund experimentation and basic research into mitigation is diminishing. As a result, it is more important than ever that we end up with tangible benefits after spending the dollars we still have. This type of budgetary environment requires us to rethink what we are doing with our limited resources to ensure we are getting our dollar's worth. As a result, we need to focus on achieving practical results with our research dollars. Instead of concentrating on basic research, the time has come to take all the technical data and research already out there and turn it into something useful for engineers, city planners, construction companies, and public officials. The research conununity needs to focus its efforts on how to reduce the costs of disasters and provide data that can be used to fight special interest groups who prefer to ignore mitigation. Although the scientific community must change its focus, it is the only way we can produce the results necessary to maintain mitigation funding in the future. Costs and Benefits Implementing mitigation is also difficult due to the lack of macro economic research quantifying the cost-effectiveness and benefits of mitigating natural and human-caused hazards. For a long time, hazard mitigation has been supported by its proponents in almost entirely abstract terms. By necessity, we've been using arguments like "It reduces human suffering when disasters occur" and "moving people out of harm's way reduces the costs of disasters." Yet, despite these claims, data supporting the costs and benefits of mitigation in high-risk areas has been lacking, especially on macro-economic and regional levels. As a result, it has been virtually


18 BARRIERS AND OPPORTUNITIES IN MITIGATION impossible for us to demonstrate the value of mitigation, and this, in tum, has hindered our ability to convince elected officials to use more of their scarce resources to mitigate the hazards they face. With all the support and resources at their disposal, the participants at this meeting are in one of the best positions to pursue this type of research. By identifying the overall costs of disasters and the comparative savings that can be realized through proven mitigation activities, information could be obtained that would provide us with a valuable tool to reduce disaster losses over the long term. Private Property and Land Use Finally, efforts to mitigate risks have been hampered by people's resistance to restrictions on the use of their land or property. Most people in the U.S. believe the Constitution grants the right to everyone to use their land as they deem proper. As a result, they often resent the government restricting the use of their property through land-use measures. Yet, after a disaster occurs, they expect the government-local, state, and federal-to take care of them and replace their damaged or destroyed possessions, as if it is their right by virtue of paying their taxes. Many government organizations adopt this same attitude. Local govern ment councils often fight sound floodplain management practices and mitigation measures, while land developers influence their decisions through patronage and favors. At the state level, officials often avoid making the tough decisions that would restrict their constituent's land-use decisions. Even at the national level of government, this attitude persists-many of those members of Congress who speak eloquently about reducing the costs of disasters through mitigation are the first to threaten amending the National Flood Insurance Program legislation to take care of a constituent that wants to build in the floodplain without meeting the requirements for federal flood insurance coverage. The impact of political pressure on land and property-use cannot be understated. In a recent report of the Association of State Floodplain Managers, I was discouraged to see how many states still have weak floodplain management programs and that funding for these programs has decreased over the last three years, in large part due to political pressures over land use. The Commonwealth of Pennsylvania has relegated their state floodplain manager to an obscure post in the Commerce Department, while complaining that they are not receiving enough disaster relief money from the federal government. And the attorney general in Texas has said that local


Krimm 19 coinmunities do not have to enforce floodplain ordinances, giving them freedom over land-use decisions. In 1973, when Gilbert White testified before Congress on the need for the Flood Disaster Protection Act, he summed up very well the need for wise use of the floodplain when he said: Continued land-use management that takes account of what so far seems to have been reasonable levels from experience on inland floods, is not to bring economic disaster to the communities affected. It is rather to avert disaster of a far greater sort to the nation as a whole. To the extent that communities have not engaged in land use, one must recognize that there has been a trade-off, and continues to be a trade-off, between the short-term benefits that are gained by a private developer and landowner and the long-term costs of the federal government in bailing out those people who subsequently occupy the property and then come to the federal government for relief, or for costly protection work. When I think that Gilbert White made that statement over 23 years ago, I find it discouraging that we are still fighting many of the same battles over land-use that we fought over two decades ago. It all may sound a bit depressing, but all hope is not lost, nor is mitigation destined to fade into the backdrop. Despite all the difficulties we now face in implementing mitigation, the fight is not over, and progress is still being made. States like Florida are making great strides in mitigation. Both the National Emergency Management Association and the Association of State Floodplain Managers are actively exploring how to further incorporate mitigation into their day-to-day activities and their professions. Mitigation success stories are increasing in number. And the actions we have been taking to protect critical facilities, buyout properties in the floodplain, elevate people's homes, and guide development decisions are beginning to payoff. But most importantly, I also see a political window of opportunity opening for mitigation. For the first time, we have a FEMA director with credibility in the emergency management community, who touts mitigation at every opportunity. We have a president who is fully supportive of our efforts to further the cause of mitigation and has said publicly, "the time has come to mount a nationwide effort focused on reducing the impact of disasters as well as reducing their economic consequences. As we continue to reach a balanced budget, reduce the deficit, and protect the vital interests of our citizens, the value of mitigation programs is clear." We also have the disasters of the last 10 or so years, which have raised many people's awareness of the hazards they face. This has led many elected officials to rethink their actions and look into better protecting their


20 BARRIERS AND OPPORTUNITIES IN MITIGATION communities and their constituents. These efforts have already started to yield tremendous gains-for example, after the Midwest Floods of 1993, over 8,000 homes in Missouri and Illinois were purchased and people were moved out of the floodplain, permanently removing them from that risk. This type of action would not have been possible without an acceptance of mitigation principles and the support and participation of officials at the federal, state, and local levels. Now, our job is to expand this political window of opportunity. And I believe you can playa critical role in making this Initiate research to quantify the costs and benefits of hazard mitigation, so that we can have the additional ammunition we need to convince hesitant officials of what they need to do to protect their communities. And by focusing on applied instead of basic research, you can also ensure that those officials who are more accepting of mitigation have the tools and supporting information they need to better explain the need to absorb the costs of mitigating the hazards they face. There are still a great number of obstacles in the way of creating a more disaster-resistant society. But despite all our problems, the time is right to move forward on mitigation. Through better research and the application of that research by various levels of government, we can accomplish a great deal. And if we can muster a willingness to capitalize on today's political environment to better educate elected officials and the public, we can make great progress in achieving our goals to reduce the costs of disasters. Big Thompson should serve as a lesson for us all. We must take responsibility for our actions. We need to plan in order to protect our communities from disasters. It's up to all of us. But most importantly, it's critical at the local level to get involved.


THE BUREAU OF RECLAMATION AND DAM SAFETY Howard Gunnarson Bureau of Reclamation Commissioner's Program Analysis Office Denver, Colorado Introduction This paper briefly touches on some of the changes that have taken place in the Bureau of Reclamation since the disasters of 1976 and addresses flood mitigation activities, or preparedness activities as we call them, that have been implemented throughout the Bureau of Reclamation. Recent History The Bureau of Reclamation has gone through major changes over the last 20 years. Our mission used to emphasize construction and maintenance of water resource projects for reclamation of the arid and semiarid lands of the west. Reclamation controls 472 dams and dikes throughout the 17 western states plus associated reservoirs, power plants, irrigation projects, and municipal and industrial (M&I) facilities. Our mission today centers around management of water and related resources in an environmentally and economically sound manner in the interest of the American public. We are no longer primarily a dam building agency. The Big Thompson One of the projects built by Reclamation between 1938 and 1959 was the Colorado-Big Thompson Project (CBT Project) for hydroelectric power production and supplemental irrigation. The project is operated by Reclama tion and the Northern Colorado Water Conservancy District. The Big Thompson flood of 1976 impacted several features of this project. Olympus Dam forms Estes Lake and is located at the head end of the Big Thompson River Canyon. The dam experienced some erosion damage during the flood from a tributary stream just downstream from the dam, but was in


22 THE BUREAU OF RECLAMATION AND DAM SAFETY no threat of failure. Flows through the town of Estes Park were stored in the reservoir during the flood and releases from the dam into the Big Thompson River were cut off. Another feature damaged was the Big Thompson siphon, a 9-foot diameter, 220-foot-Iong pipe located at the mouth of the canyon. It was destroyed during the flood, carried 600 yards downstream, and smashed into a house. Other project features such as the Big-Thompson Diversion Dam, located downstream in the canyon, sustained damage to lessor degrees. Total damage to Reclamation facilities in 1976 dollars was slightly more than $1 million. The Teton Dam There was another Reclamation project, under construction in 1976, that turned into a disaster just five weeks before the Big-Thompson flood. That event had major impacts on the Bureau of Reclamation; in fact, it turned our agency upside down, figuratively speaking. That disaster was the failure of Teton Dam in Southeast, Idaho, on Saturday morning, June 5, 1976. The resultant flood killed 11 people in the downstream valley. This was the first and only dam failure of a Reclamation designed and constructed dam. Construction was nearing completion on the 305-foot-high dam, and the reservoir was filling for the first time when the failure occurred. About 250,000 acre-feet, or 80 billion gallons of water, were rapidly released downstream. Peak flow was in excess of two million cubic feet per second in the downstream valley. The timing of the flood caused from Teton Dam failure was similar to most flash floods, as there was very little information available early in the event to indicate what was about to take place. The event time-line went from observing seepage downstream from the dam two nights before to seepage appearing on the lower face of the dam at 8:30 a.m. Saturday morning, loss of two dozers in the erosion hole higher on the dam's face at about 11 :20 a.m., and breach of the dam crest at 11 :55 a.m. Warning of the pending dam failure was given to the Fremont and Madison County Sheriffs' dispatchers at about 10:43 a.m., or a little over an hour before the breach of the dam crest. It was remarkable that only 11 lives were lost. Numerous internal and external reviews and investigations followed concerning how Reclamation designed, constructed, and operated dams. Hundreds of recommendations were made and adopted over the next couple of years that substantially changed the agency. The following are a few of the changes made in the late '70s and early '80s:


Gunnarson 23 Congressional authorization for Reclamation to repair and modify its deficient dams for dam safety purposes (1978 and revised in 1984), Reclamation Dam Safety Office established, Established a much stronger Emergency Management program (including Emergency Preparedness Plans, which are now called Emergency Action Plans, inundation maps, and Standing Operating Procedures for dams), (Note: the purposes of the two programs that Reclamation calls the Dam Safety and the Emergency Management Programs are very similar; to minimize downstream loss of life during a serious event. Safety of Dams is more geared to finding and correcting a deficiency at a dam while Emergency Management emphasizes preparedness actions to take during the operational life of the dam to ensure the safety of the public.) Requirements for instrumentation at dams were greatly changed, More elaborate controls were implemented in the design and construc tion of dams, Independent consultant reviews of dam designs were initiated, More public involvement was required during the design and construc tion of dams, Dam safety and dam operators training courses were established, Memos of Understanding were signed with each of 17 western States to share information related to Reclamation dams, and Research was initiated to evaluate the risk and probability of dam failure. Ongoing Program Changes Continuing changes took place in the Safety of Dams and Emergency Management Programs over the years. An example was that during the late 1980s, the emphasis on structural modification at dams for Safety of Dams deficiencies broadened to include nonstructural corrective actions such as early warning systems. With all five components of an early warning system in place, Reclamation's capability to provide timely warning to local authorities of life-threatening operational releases or dam failure is enhanced, which in turn facilitates a safe evacuation of the impacted population. The early warning system designed and installed at Olympus Dam and its


24 THE BUREAU OF RECLAMATION AND DAM SAFETY upstream drainage basin is discussed in a separate paper by Dave Fisher and Patricia Hagan. Last year Reclamation developed and is currently implementing, on an agency-wide basis, new policy and directives on emergency management. The new policy essentially says that Reclamation is to take reasonable and prudent actions to ensure the safety of the public and to protect environmental resources potentially affected by incidents at our facilities. Ensuring the safety of the public means that Reclamation will continue to take the following actions as required in the new Emergency Management Directives: Ensure that adequate Emergency Action Plans (EAP) are developed for Reclamation highand significant-hazard dams and that they are regularly reviewed and updated in a timely manner. New requirements say that the EAPs contain initiating conditions and emergency response levels. Develop and conduct an emergency exercise program to evaluate emergency response capabilities at each of Reclamation IS highand significant-hazard dams jointly with local jurisdictions. Ensure that all appropriate Reclamation and operating entity personnel acquire professional emergency management training. Ensure that appropriate technical information, such as inundation maps, are dcvclopcd and made available for use by the downstream local jurisdictions in their dam-specific warning and evacuation plans. Offer technical assistance to emergency management officials during their plan revisions or development. Coordinate annually with appropriate federal, state, and local emergency management officials to: 1) support their emergency management efforts, 2) ensure that the local dam-specific warning and evacuation plan response procedures are properly linked to the corresponding notification procedures in Reclamation IS EAPs, 3) encourage the scheduling of and the participation in joint exercises involving Reclamation dams, and 4) where no local plan exists, encourage the affected jurisdictions to develop their own "dam specific" warning and evacuation plan to properly respond to incidents at Reclamation dams. Maintain a redundant means for timely communication with emer gency management officials.


Gunnarson 25 Last year we also produced a double volume document to assist in the implementation of the emergency management directives called the Emergency Planning and Exercise Guidelines. These volumes contain suggested instructions on how to accomplish the required activities, prototypes of an Emergency Action Plan and a dam-specific warning and evacuation plan, and other planning tools. Another current activity given a high priority is the digitizing of all our inundation maps, including processes used to produce the maps. The processes currently being used or under development include: 1) scanning inundation areas from previously prepared maps and layering that information onto digitized raster graphic (DRG) quad sheets using a Geographic Information System (GIS), 2) using river modeling software to create the maps on a PC computer, and 3) using GIS-based software, called Dambreak Interface and developed by our Mid Pacific Regional Office in Sacramento, California, to produce the inundation maps. The software originally produced by the National Weather Service called DAMBRK is an integral part of each of these processes. All of our inundation maps will be produced in digital format on CD-ROM for use on GIS and emergency information systems (EIS).


FLOOD W ARNING/PREPAREDNESS PROGRAMS OF THE CORPS OF ENGINEERS Kenneth Zwickl u.s. Army Corps of Engineers Introduction In my position with the U.S. Army Corps of Engineers, I monitor and coordinate the Corps Civil Works activities in flood warning/preparedness (FW/P). This paper explains how the Corps gets involved in FW/P systems, the Corps flood damage reduction process, and other Corps programs. It also discusses the progress the Corps has made in recognizing and promoting FW/P as flood damage reduction measures. Flood Damage Reduction Planning Process The Corps of Engineers has a process that must be followed in order to plan, design, and construct a flood damage reduction project. The first step in this process is the reconnaissance study, which is 100 % federally funded. The purpose of the study is to identify the flooding problem and the possibility for a viable project. Normally, the Corps examines one or two alternative solutions to the flood problem-enough to show that there is federal interest in resolving the problem. Following the reconnaissance phase, a feasibility study is performed. The cost of this study is shared with a local sponsor, with 50 % federal funds and 50% nonfederal funds. The feasibility study examines a wide variety of alternative solutions and combinations of solutions and identifies the best plan for reducing future flood damages. A benefit/cost (B/C) analysis is per formed, and the recommended plan must have a B/C ratio of greater than 1.0. It is important to note that the feasibility study can be skipped if the reconnaissance study identified a FW /P system as the only feasible solution. When this occurs, the Corps moves directly into detailed design and construction. The design and construction phase of the Corps process is simply the development of detailed plans and specifications for and then construction of the project. Currently, this phase of the process is funded with 75 % federal


Zwickl 27 dollars and 25% nonfederal. The Water Resources Development Act of 1996 would change cost-sharing for structural projects to 50%-50%. However, nonstructural projects such as FW /P systems would not be affected by this change. The House of Representatives has suggested that the cost-sharing formula be 65%/35% for both structural and nonstructural projects, and the Senate has not made any suggestions on changing the cost-sharing percent ages. Flood Plain Management Services and Planning Assistance to States Programs Two other programs of the Corps of Engineers that involve FW /P systems are the Flood Plain Management Services Program (FPMS) and the Planning Assistance to States Program. The Flood Plain Management Services Program was authorized to allow the Corps to provide technical assistance and planning guidance to states and local governments for flood-related activities, such as flood warning, flood-proofing, flood hazard identification, and others. The FPMS program is 100% federally funded to conduct studies for state, local, and tribal governments. Many of the requests for assistance involve only information-type studies. However, if a study results in a plan for reduction of flood damages, the requesting agency must provide 100% of the funding for construction, operation, and maintenance of the plan. The Planning Assistance to States (PAS) Program was authorized to allow the Corps to provide technical assistance and planning guidance to states and local governments on water-related activities. This is much broader authority than that provided under FPMS in that the Corps can perform studies on water supply and distribution, water quality, and many environmental concerns not directly related to flooding. The program shares cost at a ratio of 50% federal, 50% nonfederal for studies, and like the FPMS Program, the cost for construction, operation, and maintenance is 100 % nonfederal funds. The Corps can develop a flood warning/preparedness system to include inundation mapping, number and location of gauges and other equipment, technical assistance in calibrating flood prediction software, and planning guidance for development of response plans. Under the Corps' flood damage reduction program, the Corps has completed 34 FW/P systems and has 35 systems under study or construction. Eighty-two studies have been com pleted, three studies are underway using the FPMS authority, and 18 studies have been completed and two are underway using the PAS Program.


28 FWOD W ARNING/PREPAREDNESS PROGRAMS OF THE CORPS Twenty Years of Progress A theme of this symposium is "20 years of progress," and the Corps has made good progress over the last 20 years in the FW IP arena. In the early 1970s, Congress mandated that we include a totally nonstructural plan in our feasibility studies for flood damage reduction. While this did not result in a flood of nonstructural construction projects, it did increase Corps awareness of these alternative solutions. Through the 1970s and early 1980s, the Corps learned more-expanding its expertise in nonstructural measures, including FW IP systems-and constructed a number of nonstructural projects. In the early 1980s, one brave Corps District and one very persistent local sponsor gave FW/P activities a big boost within the Corps. The Passaic River Basin had a very complex flood problem that was going to require many years of study and construction to solve. The sponsor insisted on interim help for the flooding. The New York District proposed an interim FW/P system that made it through the Corps review process and was implemented, opening the door for the Corps to get more involved in FW/P activities. In a few short years, the Corps changed the planning guidance to highlight FW IP systems as viable flood damage reduction measures, as stand-alone systems, components of a larger project, or as interim measures. We developed a comprehensive and very popular one-week training course which has been held at the Hydrologic Engineering Center for the last nine years. In the last few years, the Corps Engineering Division has insisted that a FW IP component be included in every project, whenever feasible. As a result, more and more FW IP projects were being proposed and implemented. The Corps needs to continue building on its past successes, and continue to stress Flood Warning/Preparedness in its project planning process to provide for future flood damage reduction and public safety.




This page IS intentionally blank


OLYMPUS DAM EARLY WARNING SYSTEM David B. Fisher Bureau of Reclamation Background Olympus Dam is located on the Big Thompson River two and one-half miles east of Estes Park, Colorado. The dam was constructed between 1947 and 1949 and is a composite structure consisting of a zoned earth embank ment and a concrete gravity spillway section. The dam has a total crest length of 1,951 feet, which includes a 1,631-foot-Iong embankment section and a 320-foot-Iong concrete section. The embankment has a structural height of 70 feet, with a crest elevation of 7,481 feet. A 3-foot-high parapet wall and curb were constructed on the upstream side of the 30-foot-wide crest. The concrete section of the dam, which is located at the right abutment, contains both a spillway and an outlet works. The spillway is an ogee crest with a crest elevation of 7,460 feet. Releases are controlled by five 20-foot by 17-foot radial gales with a discharge capacity of 21,200 ft3/S at a maximum water surface of 7,475. Lake Estes, which is formed by Olympus Dam, was designed as an afterbay for the Estes Powerplant. Lake Estes must maintain an operating level between elevations 7469.5 and 7475.0 for power production. At the maximum water surface elevation, 7475.0, the reservoir is approximately one mile long, has a surface area of 185 acres, and impounds 3,068 acre-feet of storage. There are no irrigation releases made directly into the Big Thompson River. The releases for power generation into Olympus Siphon are ultimately restored in Carter and Horsetooth Reservoirs, which provide releases directly to the irrigation districts. Deficiencies The hydrologic deficiency for Olympus Dam is related to the safe passage of the Probable Maximum Flood (PMF). The thunderstorm PMF is characterized by a peak discharge of 83,900 ft3/s and a two-day volume of


32 OLYMPUS DAM EARL Y WARNING SYSTEM 79,000 acre-feet. Floods exceeding approximately 12% of the PMF may create debris flows (logs, structural debris, etc.) of sufficient quantity to plug the spillway, causing overtopping and potential failure of Olympus Dam. Previous Dam Safety Activities During the dam safety study, several structural alternatives were studied as possible corrective actions to mitigate the consequences of dam failure. These alternatives included: 1) Construction of a side-channel spillway on the left abutment of the dam. 2) Placement of a roller compacted concrete (RCC) cap on the crest and upstream face of the embankment dam. 3) A combination of alternatives 1 and 2. A variety of studies were done between 1982 and 1985, including: 1) Downstream hazard assessments for loss of life and economic damages. 2) Studies on the potential for plugging the existing spillway with debris during large flood events (a concern after the Lawn Lake Dam failure). 3) Appraisal designs and cost estimates for various structural modifica tions to prevent failure of the dam caused by overtopping. Rationale For Selection of EWS There are several reasons for selecting the implementation of an Early Warning System (EWS) as the preferred corrective action at Olympus Dam. First, studies showed that by installing an adequate monitoring system in the basin above Olympus Dam, the Bureau of Reclamation can provide a warning to officials to evacuate the population that may be affected by flooding below Olympus Dam. The EWS would effectively reduce or eliminate the potential loss of life due to high releases from, or the failure of, Olympus Dam. An EWS is several orders of magnitude less expensive than most of the proposed structural alternatives. Further, the structural alternatives would not be effective in reducing the potential for loss of life due to flooding downstream. The structural alternatives would simply have passed the large flood flows on downstream. The only structural alternative that would be


Fisher 33 equivalent to the EWS in its effectiveness in reducing the potential for loss of life would be to raise the dam to store the entire flood. This alternative is not politically feasible, since it would cause the inundation of the town of Estes Park during the flood event, nor is it technically or economically feasible. Another reason for selecting a warning system as the preferred corrective action alternative is the need to warn downstream populations of high releases from the dam. The safe channel capacity below Olympus Dam is approxi mately 1,000 ft3/S. At this level, the amusement park immediately down stream from the dam begins to be inundated. Releases of approximately 1,500 ft3/S will begin to flood permanent residences in the Big Thompson Canyon. Most of the structures in Big Thompson Canyon are inundated at flows of 5,000 fe/so This is less than one-fourth of thepresent release capacity of Olympus Dam spillway. Therefore, it is imperative that the downstream population receive adequate warning of impending high releases from Olympus Dam. Early Warning System Description The Early Warning System (EWS) hardware and software were installed in the fall of 1993. The EWS hardware consists of eleven sites with multiple sensors as follows: 10 tipping bucket rain gauges, nine stream gauges, two weather stations, two reservoir level sensors, four temperature sensors, and a repeater site. These sites transmit data via line-of-site VHF radio to the repeater site that splits a microwave signal placed into the eastern Colorado area office microwave system. The data are received at three independent EWS base stations that are located at the eastern Colorado area office (microwave), the Estes Power Plant (VHF), and the Reclamation-Western Area Power Administration (W APA) Joint Operations Center (JOC) (microwave and VHF). Reservoir Elevation Information The Olympus Dam reservoir elevation gauge is an important source of information when making decisions concerning evacuation of the population at risk (PAR) located downstream from the dam. However, there will not be adequate time to carry out a successful evacuation during a severe thunder storm event based on the rate of rise in the reservoir alone. Streamflow data from the Big Thompson River gauge located just upstream from the dam will provide accurate inflow data for flows up to the top of the flume (approxi mately 1,700 ft3/s). This information is available both on the EWS and from


34 OLYMPUS DAM EARLY WARNING SYSTEM the Great Plains Region Hydromet System. Inflow data coupled with inflows from the Estes power plant, outflows through the spillway, and div'ersions through the Olympus siphon are required to calculate changes in storage at the reservoir. Streamflow Data There are nine stream gauges in the Olympus drainage basin, one on Black Canyon Creek, one on Fish Creek, four on the Big Thompson River or tributaries, and three on Fall River or tributaries. The term "upper gauges" refer to the stream gauges located at the Fall River at Endovalley picnic area, the Big Thompson River at Moraine Park, and Glacier Creek at Sprague Lake. The term "lower gauges" refers to the remaining sites. The travel time for high flows from the upper gauges to the dam range from 1.6 to 2.8 hours. The travel time for high flows from the lower gauges to the dam range from 0.3 to 1.1 hours. The travel times may vary depending on the magnitude of the flows. The slope, distance, and travel times from each stream gauge site to Olympus Dam are listed in Table 1. The lower stream gauges are valuable for confirming runoff from a rainfall or snowbelt event; however, in the case of a severe thunderstorm, they may not provide adequate warning time by themselves. Warning based on rainfall at the rain gauges may be required to obtain the necessary warning time for a rapidly progressing thunderstorm event. Streamflow monitoring will include summing flows on upstream stream gauges by groups. When it becomes evident, based on data from the stream gauges, that inflows to the reservoir are sufficient to force a spillway discharge of 1,500 ft3/s, specific actions must be undertaken. As spillway discharges increase, additional actions must be undertaken. The following groups are recommended: Group 1: Combined flow at Big Thompson River at the Power Plant, Black Canyon Creek, and Fish Creek. Group 2: Combined flow at Big Thompson River at East Portal Road, Fall River at Cascade Dam, Black Canyon Creek, and Fish Creek. Group 3: Combined flow at Big Thompson River at Moraine Park, Glacier Creek at Sprague Lake, Fall River at Endovalley, Chiquita Creek at Endovalley, Black Canyon Creek, and Fish Creek.


Fisher 35 Since there is very little storage available in Lake Estes for flood attenuation, the inflow shown on the stream gauges by groups (see above) will closely approximate the required outflow from the dam. Care should be taken not to add flows from stream gauges located on the same river (Le., Fall River at Cascade and Fall River at Endovalley). The travel times from the stream gauge sites to Lake Estes were calculated using information from two historic dam failures that occurred in the basin. A Bureau of Reclamation memorandum dated July 5, 1951, reports that Lily Lake Dam failed May 25, 1951, from overtopping of the embankment due to wave action. The flood wave traveled the 5.5 miles to Lake Estes in about one hour. The total elevation drop was 1,500 feet, which corresponds to an average slope of 273 feet per mile. Cascade Dam failed July 15, 1982, because of the piping failure of Lawn Lake Dam. The flood wave traveled the 6.0 miles to Lake Estes in 1.1 hours, as documented in the paper, The Lawn Lake Failure, dated December 1982. The total elevation drop was 1,020 feet, which corresponds to an average slope of 170 feet per mile. Elevation-discharge rating tables were developed for each of the previously nonrated stream gauge sites using surveyed cross-sections and streambed profiles, and estimates of channel and overbank roughness. This information was input into the PSEUDO computer model, a standard step backwater model with logic similar to HEC-2. The Big Thompson River at the power plant site has an existing U.S. Geological Survey rating through a stage of 7.0 feet. The rating was extended for use during extreme inflow events using results from the computer model. An important maintenance item for the EWS is for the area office staff to provide streamflow measure ments during a variety of flows to verify or adjust the synthesized ratings at the stream gauge sites. There are various outflows that affect the PAR that resides downstream from Olympus Dam. Current operations require contacting various entities at 750, 1,000, 1,100, and 1,500 ft3/s. Higher outflows would require continued coordination with downstream officials. Outflows of 2,000 ft3/s


Table 1. STREAM GAUGE SLOPE, DISTANCE, AND TRAVEL TIME INFORMATION OLYMPUS DAM DRAINAGE BASIN Sta Description or Full Station Name Slope Distance to ID To Dam Dam (Ft/mi) (Miles) Upper Gauges 3360 Big Thompson River @ Moraine Park 86 6.9 3380 Glacier Creek @ Sprague Lake 171 7.3 3240 Fall River @ Endovalley Picnic Site 128 9.5 Lower Gauges 3120 Black Canyon Creek @ Estes Park Water Plant 137 3.8 3320 Big Thompson River @ Power Plant 0 0.1 3340 Big Thompson River @East Portal Road 74 4.3 3220 Fall River @ Cascade Dam 170 6.0 3420 Fish Creek @ Golf Course 89 0.8 Historic Lily Lake Dam Failure (May, 1951) 273 5.5 Historic Lawn Lake Dam Failure (July, 1982) 170 6.0 (the failure of Lawn Lake Dam caused the failure of Cascade Dam) Estimated Travel Time (Hours) 2.1 1.6 2.8 1.0 0.0 1.2 1.1 0.3 1.0 1.1 W 0\ o til t; i c;') I'-l i


Fisher 37 would require evacuation of portions of the canyon, and outflows of 5,000 ft3/s are expected to overtop Highway 34, restricting evacuation routes. Because of this, the storm time at which outflows exceed these values were used to develop the decision criteria discussed later in this document. Rainfall Criteria A HEC-l rainfall-runoff model was developed to obtain an estimate of the runoff from rainfall depths using various basin infiltration scenarios. The model also computed the impacts on the reservoir level from those inflows. The model was developed using the subbasin delineation and runoff parameters from the PMF study dated May 7, 1981. After an initial calibration process, model runs were made for various rainfall events centered at different locations in the drainage basin for dry and wet soil conditions. Establishing the rainfall criteria required determining the one-hour and three-hour precipitation values, and comparing the time at which these values occurred with the time that the reservoir outflows exceeded 2,000 ft3/s and 5,000 ft3/s for each model run. A spillway release of 2,000 ft3/s impacts many homes and bridges across the river, and a spillway release of 5,000 ft3/s begins to overtop Highway 34 potentially blocking evacuation routes. An estimate of the criteria was determined by establishing a precipitation value that would keep the frequency of false alarms low, while still providing as much warning time as possible. The 25-year, one-hour precipitation value are ally reduced for 150 square miles is 1.2 inches, and the 25-year, three hour precipitation value is 1.7 inches. This rainfall depth has a 4% chance of occurring in any given year, which implies a frequency of false alarms of once every 25 years. The 100-year, one-hour precipitation value areally reduced for 150 square miles is 1.5 inches, and the 100-year, three-hour precipitation value is 2.2 inches. This rainfall depth has a 1 % chance of occurring in any given year, which implies a frequency of false alarms on the average of once every 100 years. The tradeoff in selection of rainfall decision criteria is warning time versus frequency of false alarms. A search of the historical precipitation data for Estes Park rain gauge shows the maximum rainfall amounts were 0.70 inches in one hour, and 0.80 inches in three hours from 1978 to 1993. A search of the historical precipita tion data for nearby gauges at similar elevations as the Estes Park rain gauge shows the maximum rainfall amounts were 1.60 inches in one hour, and 3.00 inches in three hours. This event occurred on July 22, 1991, at the Allens park Lodge rain gauge located at an elevation of 8,450 feet. This gauge has


38 OLYMPUS DAM EARLY WARNING SYSTEM a period of record of 45 years. Its second greatest rainfall amounts were 1.40 inches in one hour, and 1.80 inches in three hours. The next highest amount was at the Drake gauge, elevation 6,170 feet, which experienced 0.90 inches in one hour, and 1. 90 inches in three hours from 1975 to 1993. Because the 100-year one-hour and three-hour rainfall amounts were only exceeded at one gauge for the period of record, these criteria are assumed to have an acceptable frequency of false alarm. A series of rainfall events were modeled using the criteria of 1.5 inches in one hour, or 2.2 inches in three hours. For all the routings, the starting reservoir elevation was 7,474.0 feet (normal water surface elevation) with an outflow release condition limited to 2,000 ft3/s until the reservoir reaches elevation 7,475 feet, and releases staged up to 22,000 ft3/s at elevation 7,481.0 feet. The storm area was limited to the lower 66 mF near the dam. Saturated and dry soil moisture assumptions were modeled, with the saturated condition having no initial loss and a constant infiltration rate of 0.1 inches per hour, and the dry condition having a 1.0 inch initial loss and a 0.3 inches per hour constant infiltration rate. The warning time varies from 0.6 to 2.2 hours for all the storms modeled. It should be noted that the saturated basin condition is very conservative, and would reflect either frozen ground or heavy antecedent precipitation prior to the storm event. The more likely scenario is the "Dry" condition. It should also be noted that the storm events modeled are representative of the worst possible storm scenarios and are not events that are likely to occur. Observing the stream gauges associated with the rain gauges during an actual rainfall event will be important if the rainfall amounts are resulting in runoff amounts that could cause damaging spillway discharges to be required. Rainfall events from three to eight inches distributed uniformly over the entire 150 mF basin and for those depths occurring on each of the four individual stream subbasins for relatively dry conditions were also modeled. The storms were arranged with a depth-duration pattern as experienced at the Glen Haven rain gauge during the Big Thompson River flood of August 31, 1976. The infiltration assumptions for this series of model runs were 1.0 inch initial loss and a 0.3 inches per hour constant infiltration rate. The warning time varies from 0.8 to 2.8 hours for all the storms modeled. The rainfall decision criteria will require the rainfall threshold to be exceeded at half or more of the EWS rain gauge sites. This requires the storm area to be large enough to neglect highly localized storms that would not affect operations at the dam. Any high flows caused by this type of storm would be detected through the stream gauge network.


Fisher 39 Conclusions The streamflow criteria require that as it becomes evident based on data from the stream gauges that inflows to the reservoir are sufficient to force a spillway discharge of 1,500 ft3/S, specific actions must be undertaken. As spillway discharges increase, additional actions must be undertaken. All actions specified in the Standing Operating Procedures (SOP) and the Emergency Action Plan (EAP) will also be undertaken. The rainfall decision criteria will require the rainfall threshold to be exceeded at half or more of the EWS rain gauge sites. This requires the storm area to be large enough to neglect highly l?calized storms that would not affect operations at the dam. Any high flows caused by this type of storm would be detected through the stream gauge network.


DAMS, DEFECTS, AND TIME Wayne J. Graham U.S. Bureau of Reclamation Denver, Colorado Abstract Colorado has experienced approximately 100 dam failures since the 1800s. Information associated with dam failures throughout the United States is analyzed to show trends and patterns. Many of the dams that have failed have done so during their first few years of operation, indicating design or construction defects. Some dams that survive their early years have still failed. Have these failures been the result of original design defects or deterioration? During the eight years from 1970 to 1977, approximately 260 people, an average of 32 people per year, died as a result of U.S. dam failures. During the 18 years from 1978 to 1995 approximately 16 people, an average of one person per year, died as a result of U.S. dam failures. Possible reasons for this significant reduction in dam failure fatalities are presented. Introduction The 1993-1994 update of the National Inventory of Dams determined that there were 74,053 dams (generally more than 25 feet high or impounding more than 50 acre-feet of water) located in the United States. More than 23,700 of these dams are classified as having a "significant" or "high" hazard potential, indicating that failure of these dams could cause loss of life and/or much economic loss. These dams are maintained and operated by public agencies, utilities, private owners, and others. Colorado has 1,674 dams (2.3% of the U.S. total) according to the 19931994 update. Many of the dams built in the U. S., including Colorado, were built many years ago. Nationally, for dams where the age can be determined, 20% of the dams are more than 55 years old and 7 % are more than 85 years old. As this inventory of dams continues to age, should more dam failures be anticipated? Does aging contribute to increased likelihood of dam failure?


Graham 41 Data from the world and the U.S. are evaluated to help answer these questions. u.s. Dam Failure History History shows that loss of life due to dam failure has diminished with time. In the late 1800s and early 1900s, there were a large number of dam failures that caused considerable loss of life. The largest U.S. loss of life from a dam break occurred in May 1889, when the 72-foot-high South Fork Dam near Johnstown, Pennsylvania, failed. A flood caused the 36-year-old dam to overtop. Warnings were meager, and the widespread flooding that was already occurring in Johnstown before the dam failure prevented evacuation of the community. In some dam failures, buildings are damaged but not destroyed. In the case of this dam failure, however, a large number of buildings in Johnstown were washed away. More than 2,200 people lost their lives in this disaster (U.S. National Park Service, 1977). Another significant dam failure in the U. S. occurred north of Los Angeles, California, on March 12, 1928, a few minutes before midnight. The two-year-old St. Francis Dam, a 188-foot high concrete gravity structure, failed as a result of a foundation failure at one of the abutments. The death toll from this dam failure was about 420 people (Outland, 1977). Table 1 summarizes all dam failures in the United States that resulted in fatalities during the period 1960-1995. This table expands previously published information (U.S. Bureau of Reclamation, 1989; Brown and Graham, 1988). During this 36-year period, about 320 people died as a result of dam failure. There were about 260 fatalities during the 1970s. During the 1980s, when the number of dams in the U.S. was larger than ever before and the average dam was continuing to age, there were only seven fatalities. In the first half of the 1990s, there were 10 fatalities resulting from dam failure. During the 36-year period from 1960 to 1995, dam failures caused an average of nine fatalities per year in the United States. During the 16-year period from 1980 to 1995, dam failures caused less than one fatality per year in the United States. Some reasons for this reduction from losses in earlier times include: Improved design standards and construction techniques. Higher failure rates of dams during their first few years after comple tion; fewer dams were constructed during this period. Removal or modification of dams that were more prone to failure.


42 DAMS, DEFECTS, AND TIME Lack of conditions severe enough to cause failure. In Table 1, "Warning Time" is defined as the amount of time between initiation of the dam failure warnings and dam failure. Many of the entries in this column are zero, indicating dam failure warnings were not issued prior to failure. In many cases, some warning was begun after dani failure. The quality and quantity of these delayed warnings varied greatly. "People at Risk" is defined as the number of people in the dam failure floodplain at the time of failure. Economic losses also occur when dams fail. During the 36-year period from 1960 to 1995, economic losses caused by dam failure were 1.6% of the total economic losses caused by flooding in the United States. During the 16year period from 1980 to 1995, economic losses caused by dam failure were 0.2% (one-fifth of 1 %) of the total economic losses caused by flooding in the U.S. Stated in another way, during this period total flood losses were about 500 times as much as the flood losses resulting from dam failure. (Note that economic losses resulting from dam failure were only included for dam failures that caused at least one fatality. Therefore, the percentages do not include the impact of economic losses from dam failures that did not cause fatalities) During the 10-year period ending in 1993, average annual flood damages in the U.S. exceeded $3 billion (Interagency Floodplain Management Review Committee, 1994). Assuming that $5 billion represents current annual flood damages, dam failure economic losses would be $10 million per year based on the 16-year historic relationship between dam failure losses and total flood losses. This equals $133 per dam per year-strikingly less than common perception. Assuming that this cost of failure must be borne by the entire U. S. population of 250 million, this annual loss averages about four cents per person. Many other societal costs far exceed this amount. For example, compare the four cents per person cost with the amount people pay to insure residences and vehicles from crime, accidents, and environmental abnormali ties. The difference is substantial.


","", ., ->----Table 1. Dam Failures in the United States Resulting in Fatalities 1960 through May 1996 Dam Location Date of FailAge of Cause of Failure Dam Volume Re-Warning People at ure Dam Height leased Time Risk (Feet) (AF) (Hours) Electric Light Eagleville, 1960 n/a n/a 26 n/a n/a n/a Pond New York Park Norwich, 3/6/63 110 Piping during elevated 20 138 0 500 Connecticut 9:30 p,m, level caused by rain, Little Deer near Hanna, 6/16/63 1 Piping during normal 86 1150 0 50 Creek Utah 6:13 a,m, weather, Baldwin Hills Los Angeles, 12/14/63 12 Piping during normal 66 700 1 hour and 16,500 California 3:38 p,m, weather, 18 minutes Swift northwest 6/8/64 49 Overtopping during 157 34,300 unknown unknown Montana 10 a,m, major flood event. Lower Two northwest 6/8/64 51 Embankment washed 36 20,930 unknown unknown Medicine Montana 3:30 p,m, out next to concrete spillway walls. Lee Lake near East Lee, 3/24/68 3 Piping during normal 25 300 0 80 Massachusetts 1:25 p.m. weather. Loss of Life 1 heart failure 6 1 5 19 9 2 ,Ii> c;'l i:! I:l :::! .j:o.. w


Dam Location Date of FailAge of Cause of Failure ure Dam Buffalo Creek Logan County, 2/26/72 0 Slumping of dam face Coal Waste West Virginia 8:00 a.m. during 2-year rain event. Lake "0" Alaska April 1972 nla Unknown. Hills ::::anyon Lake Rapid City, 6/9/72 39 Overtopping during South Dakota 10:45 p.m. catastrophic flood; 245 total deaths from all flooding. Bear Wallow Buncombe 2/22/76 nla Rainfall; County, 2:30 a.m. probable overtopping. North Carolina Teton near Wilford, 6/5/76 0 Piping of dam core in Idaho 11:57 a.m. foundation key trench during initial filling. Laurel Run hear Johnstown, 7120/77 16 Overtopping. Pennsy I vania 2:35 a.m. Sandy Run hear Johnstown, 7120/77 63 Overtopping. Pennsy I vania Kelly Barnes near Toccoa 11/6/77 78 Slope failure. during Falls, Georgia 1:30 a.m. lO-year flood. ----_._.-Dam Volume Re-Warning Height leased Time (Feet) (AF) (Hours) 46 404 0 15 48 unknown 37 700 0 36 30 0 305 250,000 1 hour 15 minutes 42 450 0 28 46+ 0 40 630 0 People at Risk 4,000 unknown very large but unknown 8 25,000 150 unknown 250 Loss of Life 125 1 33 4 11 (6 from HzO) 40 5 39 t j


Location !iii Dam Date of Fail-I Age of ure Dam Cause of Failure Dam Height (Feet) 7/15/82 1791741Lawn Lake piping dur-I 26117 5:30 a.m.! ing normal weather/ 7:42 a.m. Cascade from overtopLawn Lake I near and then Estes Park, Cascade Lake Colorado ping. D.M.A.D. I near Delta, Utah 6/23/83 24.1 Backcutting caused by I 29 1 :00 p.m. i collapse of downI stream. diversion dam Nix Lake near Hender son, Texas 3/29/89 I 55 Overtopping 23 and then! Fayetteville, Lockwood I North Carolina 9115/89 9:30 p.m.! 10:00 p.m. Kendall Lake I Camden, I 10110/90 South Carolina 7:00 p.m. Georgia dams I Many dams failed 1994 Timber Lake I near LynchI 6/22/95 burg, Virginia 11 :00 p.m. Pondl Alton, iNew Hampshire 3/13/96 6:50 p.m. 23/30 I Each dam failed from overtopping. 90 69 2 Overtopping. Overtopping Failure occurred in the area of concrete spill way. Dam not overtopped. 18114 18 33 36 Volume Re-, Warning leased Time (AF) (Hours) 674/25 o 16,000 1+ 837 o 72/32 O? 690 o 1449 o 193 o People at, LOSS Risk of tt m 'rtf 5000 500 6 unknown but large unknown but large *4 lane highway 50 Life 3 2 4 3? 2 c;') i:l ;::,. ""'" VI


46 DAMS, DEFECTS, AND TIME At What Age Do Dams Fail? Information is available from numerous sources on dam failures. One of the more comprehensive data bases is that developed by the International Commission on Large Dams (ICOLD, 1995). ICOLD's statistical analysis provides information that can help answer the question: "At what age do dams fail?" The ICOLD data base is primarily for dams that are at least 15 meters (48 feet) high. ICOLD prepared statistical data and conclusions using worldwide data, excluding China. ICOLD's conclusions included the following: The percentage of failures of large dams has been falling over the last four decades; 2.2 % of dams built before 1950 failed, failures of dams built since 1951 are less than 0.5%. Of the failed dams built before 1950, 80% were no older than 36 years, and 50% were no older than 4 years. Most failures involved newly built dams, with 70% of failures occuring in the first 10 years, and more especially in the first year after commission ing. An earlier study, using worldwide data, developed the following conclusions: "The most frequent failures have occurred during the first complete filling of the reservoir, which usually takes place within five to seven years after construction. It seems that after that age very few dams have failed. It also seems that after the useful life of a dam (say 60 to 100 years), the percentage of accidents increases again. Thus, if the first fillings of new dams are carefully controlled, ... and if all old dams are the object of careful inspection and calculation, a great number of failures (possibly more than 70%) can be avoided" (Serafim, 1991). United States information from Table 1, used with the National Inventory of Dams data, can also be used to gain further knowledge in this area. Table 2 shows the relationship of the percentage of older dams that have failed in each decade since the 1960s to the percentage of older dams existing in the inventory. The dams that were about 50 years old or more included the following: Swift, Lower Two Medicine, Mohegan Park, Sandy Run, Kelly Barnes, Nix Lake, Lawn Lake, Cascade Lake, Kendall Lake, and Timberlake, a total of 10 dams. Only one of these, Lawn Lake, failed during normal weather


Graham 47 Table 2. Failure of Older u.s. Dams That Resulted in Loss of Life Decade during Percentage of dams in Percentage of failures which dam failU.S. inventory over during decade of dams ure took place: 50 years old: more than 50 years old: 1960s 14 50 1970s 15 33 1980s 20 50 1990s 27 33 Average 23 42 conditions, an indication that the dam had deteriorated to a point of failure. All of the other dams failed either as a result of rainfall that caused the reservoir to rise to higher than normal levels or overtopping. It can be concluded that older dams have a higher rate of failure than newer dams that have passed the initial test of being able to hold water. However, it can not be concluded with the data available that the aging of the U.S. dam inventory is adding to our risk. Most of the over-50-years-old dams that failed did so during a major flood event. These older dams, built using less demanding design floods, would be expected to have a higher failure rate than newer dams that generally have been designed using more demanding standards. Data is also available for the failure of dams in the United States (Hatem, 1985). This analysis indicated that during the period 1971 to 1980 there were 100 dam failures for which the age at the time of failure could be determined. The failure of dams more than 50 years old accounted for 32 % of all dam failures during this time. The number of dams in the U.S. during this time that were more than 50 years old was about 15 % of the total number of dams. Thus, dams that were more than 50 years old failed at a rate about double the expected rate if failures were evenly distributed based on age. Conclusion The evidence suggests that older dams are more prone to failure than newer dams that have passed the initial test of being able to hold water. The


48 DAMS, DEFECTS, AND TIME higher failure rate for older dams may be due to the less stringent design standards that went into their construction. The evidence is weak in supporting the belief that older dams are failing due to deterioration. One reason for this is that as deterioration is discovered, and it is felt that the dams integrity is being jeopardized, most prudent dam owners undertake action to correct the problem. References Ad Hoc Interagency Committee on Dam Safety of the Federal Coordinating Council for Science, Engineering and Technology 1979 Federal Guidelines for Dam Safety. ASDSO Newsletter, 1996 "Information on Bergeron Dam." May. Beard, Daniel P. 1994 Memorandum to James Lee Witt, Director, Federal Emergency Management Agency. January. Brown, Curtis A., and Wayne J. Graham 1988 "Assessing the Threat to Life from Dam Failure," Water Resources Bulletin 24 (6): 1303-1309. Federal Emergency Management Agency 1992 Floodplain Management. in the United States: An Assessment Report. Volume 2. 1994 National Dam Safety Program: 1992 & 1993, A Progress Report. Volume 1. Fread, Danny L. 1988 The NWS DAMBRK Model: Theoretical Background/User Docu mentation. Silver Spring, Maryland: National Weather Service.


Graham 49 Graham, Wayne J. and Chih Ted Yang, 1995 "Dam Safety and Nonstructural Damage Reduction Measures." Paper presented at the U.S. Korea Joint Seminar on Reduction of Natural Disaster in Water Environment. July. Hatem, Georges Antoine 1985 "Development of a Data Base on Dam Failures in the United States: Preliminary Results." Thesis submitted to the Department of Civil Engineering of Stanford University, December 1985. Interagency Floodplain Management Review Committee 1994 Sharing the Challenge: Floodplain Management into the 21st Century. Washington, D.C.: Federal Emergency Management Agency. International Commission on Large Dams 1995 Dam Failures-Statistical Analysis. Bulletin 99. Paris, France. Outland, Charles F. 1977 Man-Made Disaster: The Story of St. Francis Dam. Glendale, California: Arthur H. Clark Company. Serafim, J.L. 1981 "Safety of Dams Judged from Failures," Water Power and Dam Construction. U S. Army Corps of Engineers 1994 National Inventory of Dams, Updated Data, 1993-1994. u. S. Bureau of Reclamation 1989 Policy and Procedures for Dam Safety Modification Decision making. U.S. National Park Service 1977 Johnstown Flood Brochure. Johnstown, Pennsylvania: Johnstown National Memorial.


1996 WILLAMETTE AND COLUMBIA RIvER FLOOD Cynthia A. Henriksen North Pacific Division, Corps of Engineers The flash flooding that occurred in the Willamette River Basin in February 1996 was part of a larger regional flood that encompassed the entire Columbia River Basin in the Pacific Northwest. Controlling the event involved cooperation among many federal, state, and private agencies. Although flood frequencies varied across the Willamette Basin, some locations experienced flooding at a frequency as high as a 200-year event. The North Pacific Division of the U.S. Army Corps of Engineers, in coordination with the National Weather Service River Forecast Center, used SSARR (Streamflow Simulation and Reservoir Regulation) for inflow forecasting in the Columbia River Basin. The Corps of Engineers opened emergency operations centers in Portland, Oregon, and Seattle and Walla Walla, Washington. The primary storage reservoirs in the Pacific Northwest are both federally and privately owned. The Corps of Engineers operates John Day Dam on the lower Columbia River and the Dworshak storage project in Idaho. Flood control storage projects in the Willamette Basin contain about 1.6 million acre-feet of storage and include 13 dams. Eleven of the dams have storage and two are strictly for re-regulation of power peaking from upstream projects. The Bureau of Reclamation also operates a major storage project, the Grand Coulee Dam on the mains tern Columbia River. Other major water storage facilities that were used during the February 1996 flood include the Hugh Keenleyside Dam in Canada, owned by BC Hydro and containing 7.1 million acre-feet of storage capacity in Arrow Lakes. Idaho Power Company also owns a storage project, called Brownlee, on the Snake River. All of these storage projects were instrumental in the flood control operation during this event.


Henriksen 51 Before the Flood The flood of February 1996 in the Pacific Northwest was preceded by a winter of above-normal precipitation throughout the basin. From November 1995 through January 1996, precipitation was 102 % to 182 % above normal through the 262,000-square-mile basin above The Dalles, Oregon. In the 11 ,600-square-mile Willamette Basin, precipitation was 173 % above normal for the winter. During the last week of January 1996, the Willamette Basin experienced an unusual event-snow on the usually warm Willamette Valley floor. Finally, during the weekend of February 3 and 4, the Willamette Basin was bombarded with an ice storm that left ice on the saturated valley floor. On Monday, February 5, the temperature warmed to about 60 F in a few hours, causing a rapid snowmelt. Meanwhile, a weather pattern was setting up to bring a belt of warm, moist clouds from Hawaii into the Pacific Northwest. Getting Ready By Monday, February 5, the Corps of Engineers Reservoir Control Center (RCC) examined the streamflow forecasts and expected a rise in the stages at the Portland harbor. The ultimate goal of the Columbia and Willamette River reservoir system is to protect the Portland Harbor from flooding, where flood stage is 18 feet. By Monday, February 5, the harbor stages were expected tu rise from seven feet to as high as 14 feet by February 8. The appropriate reservoir operation in this event is to draft the John Day reservoir to capture future flood peaks. The RCC responded by drafting John Day nearly one foot to elevation 264.0 feet. This was an evacuation of nearly 50,000 acre-feet. By Tuesday, February 6, the weather pattern from Hawaii strengthened and the harbor stages were expected to rise even higher than 14 feet. The Corps requested that BC Hydro in Canada reduce outflow from the Treaty project Arrow in Canada. BC Hydro cooperated and started daily reductions in outflow. The outflow was immediately reduced from 92,000 to 70,000 cubic feet per second (cfs). The reductions were agreed to continue at 15,000 cfs per day until the outflow reached 15,000 cfs on Thursday, February 8. Also on February 6, the John Day reservoir elevation was drafted even further to elevation 262.5 feet. By late in the day of February 6, precipitation was expected to be even greater than originally forecast. At about midnight on February 6, the Corps began a heavy draft of the John Day Reservoir to evacuate as much water as possible without pushing the Portland Harbor above the 18-foot flood stage.


52 1996 WILLAMETIE AND COLUMBIA RIVER FLOOD The John Day Reservoir was drafted to elevation 260.7 feet by 6:00 a.m. on February 7, forcing the harbor to its flooding level of 18 feet. On Wednesday, February 7, the Corps of Engineers called upon 'many regional entities to participate in the stage reduction operation. The Corps of Engineers reduced outflow from the Dworshak project in Idaho from 10,000 cfs to 1,000 cfs. On Thursday, February 8, the Corps asked the Idaho Power Company to reduce the outflow from Brownlee from 25,000 cfs to 15,000 cfs. Although they had not been asked to participate in a flood reduction event in many years, they agreed to cooperate. The Corps also called upon the Bureau of Reclamation to reduce outflow from the Grand Coulee project in Washington on the mainstem of the Columbia River. The flow was reduced from 135,000 cfs on Monday, February 5, to as low as 50,000 cfs outflow on Thursday, February 8. This reduction caused difficulty for the Bonneville Power Administration, which markets power from the federal dams in the northwest. Because the reduction in flow from Grand Coulee caused difficulty in meeting system load requirements, the Corps quickly brought generation units back on line on the Lower Snake River to compensate for the loss of flow at Grand Coulee. Beginning Thursday, February 8, the Corps began to refill the John Day Reservoir to capture as much of the flood peak from the Columbia River as possible. The full elevation of John Day is 268.0 feet, and the water level was at 260.7 feet, thus, refilling the reservoir would take from Thursday, February 8, through midnight, February 9. The total storage equaled nearly 400,000 acre-feet and reduced flow by about 90,000 cfs into the Portland Harbor while the reservoir filled. The Willamette projects are primarily flood-control reservoirs that are annually drafted to their minimum conservation pool by December and remain at these minimum elevations through the end of January in order to be available to contain winter flood events like the one that was occurring. Although the projects had begun to fill slightly before the event, the total amount of storage in the 11 projects was only 8 % of the total composite capacity. Although there is much reservoir regulation on the Columbia River, Willamette Project operations are simple. Once the control points downstream of the Willamette Projects reach bankfull and are continuing up toward flood stage, all the project outflows are reduced to zero or minimum outflow. The Willamette projects remained at zero or minimum outflow until they filled and were forced to pass inflow. The only project in the Willamette that filled and was forced to pass inflow was the Foster project, a small re-regulation project on the south fork of the Santiam River that only stores 55,000


Henriksen 53 acre-feet of water. The flooding that occurred in the Willamette basin was all due to local runoff below the storage projects. Downstream of Foster on the mainstem of the Santiam River is the Jefferson gauge. At this site, the flood hydrograph was rising very fast and indicating a 200-year event. Flooding became heavier in the north. During a six-hour period on Thursday, February 7, the precipitation gauge at Foster Dam registered two inches of rain during a six-hour period. In the southern portion of the Willamette Basin, the flooding and rainfall was less than in the North. South at Hills Creek Dam the maximum rain period was only 0.4 inch on February 8. After the flood was over, the Corps of Engineers estimated that the stage and damage reductions due to Corps of Engineers reservoir operations were significant. At the gage at Eugene, the Corps estimated that stage reduction was nine feet, reducing potential damage by $195 million. Further down stream at the Oregon state capitol of Salem, the Corps estimated a stage reduction 71/2 feet and damage reduction of $280 million. Within the state of Oregon and at the highly developed Portland-Vancouver metropolitan area, total damages were estimated to have been reduced by $1.14 billion. Forecast Modeling The North Pacific Division Corps of Engineers, in coordination with the National Weather Service River Forecast Center, uses the Streamflow Synthesis and Reservoir Regulation (SSARR) model, which is a hydrologic routing model that is initiated at 4:00 a.m. every day. The first three days of the forecast are in six-hour time steps, and the next two days are daily time steps. The River Forecast Center (RFC) uses six hourly quantitative precipitation forecasts (QPF) developed by the National Weather Service as input to the inflow forecast portion of the model. The RFC also uses routing parameters to simulate soil saturation in particular basins. Once the individual basin inflows are developed, the Corps of Engineers then overlays reservoir regulations at all the projects in the region to develop the best operational scenario to achieve maximum flood reduction in the Portland Harbor. Overall, forecasting was quite good. As the QPF was updated, the basin inflow forecasts may have been resubmitted. There was a discontinuity when the QPF was updated and the regulation resubmitted during evening or nighttime hours, since the regulation could not be re-initialized to any time other than 4:00 a.m. This was particularly troublesome on the lower Columbia River, where many side streams flowing into the Bonneville and


54 1996 WILLAMETIE AND COLUMBIA RIVER FLOOD John Day reservoirs were experiencing record peak flows. The model could not keep up with the activity during the day. On the Willamette Basin, the forecasts were better. This is a smaller basin and fairly well calibrated from the upper reaches in southern Oregon to the Corps of Engineers last downstream control point in Salem, Oregon. Beginning as early as Monday February 5, the SSARR forecasts were predicting the peak on the Willamette Basin to arrive at the Portland Harbor late February 8 or early February 9. Although the magnitude of the peak kept growing as the QPF was updated, the timing did not vary. Ultimately the peak of 28.6 feet occurred in the Portland Harbor at 6:00 a.m. on February 9 The estimated flow from the Willamette River was about 450,000 cfs and the flow from Bonneville Dam was 400,000 cfs. The total flow in Portland Harbor was approximately 1,000,000 cfs. Emergency Response Emergency operations centers were opened by the Corps of Engineers in Seattle, Walla Walla, and Portland. From February 7 through 9, the emergency operations centers coordinated with the RCC twice each day to share reservoir operations and flood-fighting information. The National Weather Service participated in the briefings, updated the weather reports, and projected flood warnings for the many unregulated streams in the region. These briefings were made available to the public and the news media, but were particularly helpful to emergency operations personnel, who had to decide how and where to go for flood-fighting and evacuations. The emergency operation center in the Portland District was open through February 27. Although the actual rain and flood event itself lasted only about five days, there was continued activity following the flood. Emergency operations centers assisted in clean-up events, and the Reservoir Control Center had to safely evacuate all the flood waters that were stored throughout the region. Throughout this stressful and dramatic event-before, during, and after-a fine example of regional cooperation was displayed by all federal, state, and private agencies.




This page IS intentionally blank


EMERGENCY COMMUNICATIONS A SURVEY OF THE CENTURY'S PROGRESS AND IMPLICATIONS FOR FUTURE PLANNING Bascombe J. Wilson, MPA, CEM Director, D ERA Disaster Resource Center Background Today, new electronic systems and applications are being developed faster than we can readily assimilate them into our personal and professional lives. Increasingly, even experts in the fields of information management and communications technology find it difficult to keep pace with new develop ments because of extreme complexity and narrowing specialization. The nation and local communities are best served when emergency managers and the elected leadership have a working knowledge of the ways communications and information management systems inter-operate, and are able to make informed decisions regarding cost-effective technical solutions to local problems. This study is intended to help bridge that gap by outlining for emergency managers some of the history of emergency communications and emergency management (the road behind us), a candid assessment of current systems (where we are now), a projection of where technology is leading us (a tentative guess at the road ahead), and a review of the lessons we've learned along the way. Many of those lessons, as we all know, were learned through loss of life, terrible suffering by many people, and devastating property loss. We should not forget those lessons, even in the face of a bright and optimistic future filled with dazzling technological wonders. This paper began as a review of communications/information manage ment/incident command trends developing over the last 20 years following the tragic Big Thompson Canyon flood disaster. It soon became apparent, however, that a 20-year perspective was not sufficient to provide a clear understanding of the trends and developments shaping our current state of emergency communications and incident command systems. As I looked backwards a decade at a time, I began to develop a new appreciation for the difficulties our predecessors had in integrating new technologies into their organizational structures and in keeping up with rapid change.


58 EMERGENCY COMMUNICATIONS The past 100 years have been marked with technological developments that revolutionized the structure of societies the world over, the world's collective view of itself, and the relationships among peoples and nations everywhere. Developments in communications and information management have been at the core of this global metamorphosis, and appear to be one of the central guiding forces shaping our own future. Twenty years ago, Larimer County, Colorado, experienced a devastating flash flood in the Big Thompson River Canyon between the towns of Drake and Loveland. More than 140 people died, and 316 houses, 56 mobile homes, and 52 businesses were destroyed. Large numbers of people died attempting to warn and rescue others. Residents, tourists, campers, and migrant workers were washed away without a trace. It will never be known exactly how many people died that day. In one of the most beautiful areas of North America, disaster struck unexpectedly and scarred forever the landscape and the lives of its people. This work is dedicated with respect and reverence to the memory of those whose lives were lost, to the courageous rescue and recovery teams who worked diligently in the days following the tragedy, and to the survivors who are still recovering. This research project reviewed the process through which emergency communications and information systems evolved over the past 100 years, and attempted tentative conclusions about strategies that might be helpful for future development of systems in the public interest. The following report is a brief overview of the conclusions of the study. Baseline Considerations The Big Thompson tragedy shocked us, because most of us thought our system of weather radars, forecasts, and warning systems could give us adequate time to prepare and evacuate. We were wrong then, and today-20 years later-the weather and human-caused events can still surprise us with situations evolving faster than our warning systems and the public can respond. Since that time, many of the communications devices we use have either changed or recently come into use. For example, the National Weather Radio System expanded to cover most of the nation, but most people still did not buy receivers. The Federal Communications Commission considered mandatory installation of receiver modules in all new television sets, but manufacturers lobbied against this due to the cost-about $5.00 per unit. Mobile Data Terminals (MOTs) were successfully tested in police and fire vehicles to relieve the congestion on voice channels and improve speed and


Wilson 59 accuracy of information exchange. Personal Computers (PCs) have come into almost universal use, and a large percentage of radio broadcast stations are running on autopilot and are unattended. Also, the Incident Command System has become widely (although not universally) adopted. For decades, the Amateur Radio Emergency Service (ARES) has provided a well-organized communications backup resource in most communities, and is frequently exercised through Simulated Emergency Tests. At their own expense, volunteer amateur radio operators provide the equipment and expertise needed to rapidly restore or expand critical local and long-distance communications during emergencies. A national movement toward centralized Public Service Answering Points (PSAPs) and a National Emergency Number (9-1-1) changed the way the public called for help. Depending upon jurisdiction, the 9-1-1 center could patch calls through to police and fire dispatchers, or dispatch could come directly from the 9-1-1 center. Many states required telephone companies to impose surcharges on all customers to pay for 9-1-1 service, which ran to millions of dollars. For a long time, customers confused the number 9-1-1 with repair service at 6-1-1, or directory assistance at 4-1-1. One could, however, always dial zero for an operator and get help. Serious limitations exist with the PSAP/9-1-1 system, but to avoid eroding public confidence in this concept, these are often not widely discussed. Limitations include: The systems were designed for day-to-day not disaster response. Both the switching equipment and the human operators/ dispatchers become overloaded during major emergencies, and a form of gridlock occurs during disasters. The public, however, has grown to expect and demand a high level of service and responsiveness from PSAPs, even during major disasters. The systems depend on the normal phone system. If callers can't get dial tones, they can't reach 9-1-1. Emergency line load control protocols sometimes deny dial tones to residential customers during periods of high demand. Also, if phone lines or switching problems exist between the PSAP and the police/fire dispatcher, it is not possible to pass needed information to the dispatcher unless backup radio systems are available. Increasingly, they are not. Often, faster response could be obtained by calling the fire or police station directly, rather than going through the PSAP operator, who then must patch the call to the appropriate dispatcher.


60 EMERGENCY COMMUNICATIONS Without absolutely accurate, current maps, it is difficult for PSAP operators to determine the correct fire or police jurisdiction to handle the call. Additionally, adjacent communities have similar street and subdivision names, and telephone prefix numbers do not always align with political boundaries. Calls to 9-1-1 are often routed a thousand miles or more through central switching centers. A disruption in long distance service could interfere with or shut down local 9-1-1 service. Calls to 9-1-1 from a Private Branch Exchange (PBX) or an Offsite Branch Exchange (OBX) do not necessarily display the correct address on the PSAP console. Often, the dispatcher will see only the street address of the PBX switching center, not that of the caller. Home reception of satellite TV broadcasts on the C-Band became popular in the late 1980s when dishes became price-competitive with cable TV. Instead of two dozen channels, the satellite viewer could access 50 or more channels. Since many of these C-Band broadcasts were intended as propri etary feeds to cable operators, and not as signals for public consumption, encryption systems were developed to protect the signals. Again, hackers broke the codes and illegal satellite "black boxes" appeared that could descramble codes and allow the user to receive the signal for free. Also, at least one hacker broke the uplink codes for a major C-Band satellite network and superimposed his own programming over the satellite's normal channels for an evening. He was later apprehended and convicted of several crimes relating to the incident, but his success underscored the vulnerability of the relatively unsophisticated C-Band satellite network. A Quick Look at Today's Challenges Public service and emergency communications are scattered across the radio spectrum-there are so many specialized services that it is difficult to define any meaningful boundaries. All of these services provide good day-to day efficiency, but terrible operating conditions when mutual aid is needed or normal operations are disrupted. Today, urban areas are "trunking" their radio systems, usually on the 800 MHz band. That makes good economic sense, and allows more discreet functions and a larger number of users to coexist on a given number of radio channels. Despite this daily efficiency, such arrangements create an unworkable system during disaster. When things go wrong, as they have recently done during major disasters, police and fire units lose nearly all


Wilson 61 communications, and emergency operations centers are thrown into chaos. Several fall-back schemes are being explored, but this is a major challenge for the future. Global Positioning System (GPS) satellites allow very accurate location and tracking. GPS receivers can be incorporated within portable two-way radio equipment to send continuously updated location information to central dispatch. Global communications out of disaster areas can be obtained by using satellite receiver terminals in briefcase size units for voice, data, and fax message exchange at moderate cost. Adjacent jurisdictions do not always have compatible radio equipment. Mutual aid efforts are seriously hampered when different agencies bring radio equipment that cannot intercommunicate. For example, in the Denver metro area alone, different agencies use Low Band VHF, High Band VHF, UHF, and 800 MHz trunked systems. As long as the separate emergency operations centers or mobile command posts can relay messages through the phone system or over liaison radios, the system can work, but often without efficiency. The Emergency Broadcast System is being replaced with the Emergency Alert System to take into account non-standard broadcasting media, such as closed circuit music channels. The new system will attempt to deal with the problem of unattended broadcast stations, but the costs to smaller local stations of converting will be a major challenge to making the Emergency Alert System universal and dependable. Enhanced color Doppler radar, wind protiler systems, and expanded networks of remote rain gauges are improving the ability of the National Weather Service to track weather. Still, human observers, including trained volunteer observers-particularly amateur radio SKYW ARN teams-continue to be critical to local storm tracking and warning. This is particularly true regarding fast-developing phenomena such as tornadoes, severe hail, microbursts, and flash flooding. The National Oceanic and Atmospheric Administration is making dramatic improvements in the national weather radio alerting system through the use of digital technology to pinpoint weather warnings to specific city blocks or county sectors, which will likely coincide with the U.S. Postal Service's nine digit zip code map. Direct Broadcast Satellites on the Ku-Band provide direct signals for paid subscribers through the use of 18" to 24" dish antennas. Access to well over 100 channels is common. Although federal regulations require that state and local emergency operations centers have the ability to directly send public warning messages over cable TV systems and local broadcast stations, there


62 EMERGENCY COMMUNICATIONS is no similar capability to reach the increasing number of households served by direct satellite broadcast service. For catstrophic situations, the Federal Emergency Management Agency (FEMA) maintains five Mobile Emergency Response Systems (MERS) at key federal centers in the U. S. A MERS detachment can be moved by air or land to any location to provide high-bandwidth satellite links for video tele conferencing, high speed data, and a mid-sized telephone Private Branch Exchange (PBX), in addition to heat, water, air conditioning, and electrical power generation for a command center. The Amateur Radio Emergency Service (ARES) has been incorporated into the emergency response plans of many communities in the U. S. Unfortunately, over half of all communities do not have current, comprehen sive emergency response plans, and many do not have plans for backup communications. The Federal Communications Commission has reallocated some radio spectrum from amateur/emergency use to commercial interests because of a perceived lack of need. And, at least one Central American country is considering the reallocation of radio frequencies used by amateurs for emergency communcations through satellites currently in orbit to commercial use. This conversion is important because, even if one country takes such action, emergency communications throughout the entire hemisphere may be jeopardized because of interference on satellite up-link frequencies. The Near Future-A Tentative Projection Personal communications devices will continue to be developed, and many will be able to access the telephone network, but may not necessarily operate directly over it in their normal mode. For example, if the electric company can find a way to send communications signals over their power lines, they will be able to offer PCS service. Almost for certain, PCS will operate much like a cellular system, although the handheld units will be smaller, lower power, and cheaper to use. Low Earth Orbiting satellites (LEOs) will provide seamless global communications for portable and mobile terminals, without the need for pointing antennas. The LEOs will act like cellular phone stations in the sky. Serious concerns are that the LEO nets will usurp radio frequencies now allocated to critical emergency services as well as determining who will answer when someone dials zero for an operator (the operator could be in Tokyo or Frankfurt, rather than the U.S.). Also, barring technological


Wilson 63 breakthrough, a 9-1-1 call through a LEO system will not be properly connected without the delays of manual call routing. A proliferation of new X-l-l numbers to meet various public service and commercial demands will further weaken the identity of 9-1-1 with the public. Many vehicles will be equipped with automatic telemetry systems and distress alarms operating over satellite networks and tied into monitoring stations, which can alert local emergency services to disabled vehicles, stolen vehicles, and accidents. The public's expectation for excellence in emergency service and response is probably going to increase, while systems will become more complex and expensive. Further, government and community response agencies cannot anticipate every possible contingency and cannot afford the high cost of in place backup systems for every critical telecommunications network. Therefore, amateur radio will continue to be an essential resource during major disasters. As such, emergency managers need to continue integrating amateur capabilities into local plans, while training and exercising amateur radio operators in their mission. Civil government cannot ignore the threat of information warfare. Military units of many nations, some unfriendly to the U.S., are specializing in highly advanced methods not only for disruption of communications and computer systems, but more significantly, the undetectable intrusion into a targeted system for the purpose of manipulating data and planting false information. Implications for Strategic Planning 1. The public must be kept informed and given realistic information about system capabilities and limitations. Public awareness and individual preparedness must remain a top priority. 2. Our commitment to quality of service cannot be allowed to erode. The road ahead will be difficult and demanding, and we must give our emergency managers the executive-level training they need to succeed. Most of them will need a full-time technical expert on staff to assist them in making good telecommunications decisions. 3. We will probably ignore the threat of information warfare until a successful terrorist or enemy act does significant damage. We need to find ways to harden our local emergency communications and information management systems against information warfare attack. 4. System redundancy is expensive. It is more expensive not to have it.


64 EMERGENCY COMMUNICATIONS 5. System recoverability is expensive. The public service mission will fail if we do not have it. 6. Fall-back capabilities are hard to develop or retain. It is very hard to keep people trained on the use of old, low-tech systems, but profi ciency must be maintained until it is absolutely certain that new systems will function under sustained stress and unanticipated failure. 7. Increasingly, public safety and quality of life will depend on sustained telecommunications and electric power distribution. At the same time, the increased complexity of such systems and rapidly growing demand will make failures more likely to happen. Our only strategy as emergency managers is to perform comprehensive risk analyses and make plans for emergency restoration and recovery through reliable backup resources. 8. As a nation, we must give high priority to maintaining and encourag ing the volunteers who provide vital life-saving service in time of disaster. Just as we must provide support and flexibility in interstate licensing of emergency medical personnel and workers compensation and professional training for volunteer firefighters, we must do everything we can to strengthen the amateur radio community. That means working locally to incorporate their organizations into local response plans and making sure we adequately train our volunteer operators and exercise their capabilities. Nationally, we need to work with Congress and the Federal Communications Commission to abSOlutely guarantee that valuable radio spectrum needed for vital emergency communications is protected from reallocation to commer cial interests with less importance to the welfare of the American people. We need to protect this irreplaceable radio resource-both the frequencies that our police, fire, medical crews, and response teams use day to day, as well as the frequencies that amateur radio operators need to support us when we call on them for help. The views expressed in this paper represent the opinions of the author and do not necessarily represent those of anyone else, including the symposium sponsors, the Federal Emergency Management Agency, or the DERA Disaster Resource Center, or anyone else with whom the author is associated. This paper remains a work in progress, and the author would appreciate comments, critiques, additional information, or other suggestions. Please address them to Jay Wilson, DERA Disaster Resource Center, P.O. Box 280795, Denver, CO 80225-0795.


COPING SELF-EFFICACY FOLLOWING NATURAl .. AND HUMAN-CAUSED DISASTERS Charles C. Benight University of Colorado at Colorado Springs This paper discusses research on the role of coping self-efficacy in post-disaster psychological and physical responses in five separate samples, including natural and human-caused disasters. Research on the mental health impact of disasters has rapidly proliferated in the last several decades and has demonstrated that serious psychological ramifications occur following a disaster (Adams and Adams, 1984; Rubonis and Bickman, 1991). The primary aim of these research projects was to evaluate the predictive power of subjective appraisals of coping self-efficacy (CSE) for psychological distress in recovering from a major disaster. CSE is defined as a person's subjective appraisal of his/her ability to cope with the environmental demands of a stressful situation. For example, following a natural disaster such as a hurricane, victims are faced with significant emotional, financial, and often physical demands related to getting life "back to normal." A person's self-appraisal of how capable he or she feels to successfully manage these demands is an example of CSE for post-hur ricane recovery. Enhanced levels of CSE have been related to better functioning in a number of ways. For example, higher levels of CSE have been related to improved coping with physical assault (Ozer and Bandura, 1990), dealing with the psychological effects of abortion (Meuller and Major, 1989), better immune function (Wiedenfeld et aI., 1990), lower catecholamine reactivity during stress (Bandura, Taylor, Williams, Mefford, and Barchas, 1985), and reduced blood pressure response during a stressful task (Bandura, Reese, and Adams, 1982). A few studies have also shown that the better a person's appraisal of CSE, the better he or she recovers from extreme environmental experiences. Murphy (1987) found this when studying victims of the Mt. Saint Helens eruption, and it was also the case for Israeli soldiers who had faced military combat (Solomon et al., 1989). Thus, perceived coping


66 COPING SELF-EFFICACY FOLLOWING DISASTERS efficacy has been shown to directly effect levels of psychological distress following trauma, and may be critical for maintaining a sense of personal control during recovery from a major disaster (Freedy et al., 1992). The following discussion will describe five different disaster studies where CSE was investigated as an important predictor variable of psychologi cal distress. The disasters include Hurricane Andrew, Hurricane Opal, and the Oklahoma City Bombing. General healthy popUlations were studied in each of the disasters and two ill populations (HIV + and Chronic Fatigue Patients) were studied in Hurricane Andrew. Hurricane Andrew The first three studies were all completed following Hurricane Andrew, which blasted South Florida on August 24, 1992, leaving a devastating toll of 250,000 homeless, over $15 billion in damage, and unforgettable images of destruction. The recovery phase from Hurricane Andrew was, and probably still is, a challenge of often superhuman proportions. Removing debris, securing electricity and water, and soliciting contractors for rebuilding are just a few of the many problems victims of this disaster faced. The CSE measure utilized for all of the hurricane studies was designed to reflect these types of demands and asked participants to judge how capable they felt in managing the various challenges (see Table 1). The first study focused on the reactions of the general population of southern Dade county. The second study looked at the reactions with a group of HIV-infected gay men, and the third with a small sample of women diagnosed with Chronic Fatigue Syndrome. Study methods for the three investigations were virtually the same with minor differences related to the type of popUlation being studied. In general, participants responded to an in-depth interview, filled-out question naires, and provided a blood and urine sample at time 1 (approximately 1-3 months post-hurricane); and, for the neighborhood group, filled-out questionnaires at time 2 (approximately 7-9 months post-hurricane). Subjects from the first study (N = 180, male=62 and female = 118) were recruited from the general population in storm-ravaged neighborhoods. The mean age of those who were sampled was 39 years, and 43 % were Cauca sian, 34% were African American, 18% were Hispanic, and 4% were Asian American. Results from this first study indicated that CSE was an important predictor of psychological distress even after controlling for other factors such as income, education, gender, age, damage, threat of death, and lost resources. This study also showed that CSE beliefs two months after the hurricane were related to distress eight months later. This relationship was


Benight 67 strong enough that it still remained even after including early psychological distress levels. The second and third studies attempted to determine the importance of CSE in individuals who were facing the compounding stress of dealing with hurricane recovery on top of coping with a serious disease. We recruited 36 HIV -positive, mildly symptomatic (non-AIDS) gay men between the ages of 18 and 50 who underwent the stress of Hurricane Andrew and were assessed within six months of the storm. Mean age for these participants was 36.4 years (SD=9.6). Ninety-one percent had at least some college education. Results suggested again that greater levels of CSE were related to less emotional distress and post-traumatic stress disorder (PTSD) symptoms. In addition, greater CSE was associated Table 1. Coping Self-Efficacy Scale for Hurricane Recovery 1 I'm not at all capable Perceptions of Coping Self-efficacy 2 3 4 5 6 I'm moderately capable To What Extent Are You Capable of. ... 7 I'm totally capable 1. Maintaining personal security-protecting yourself and your property. 2. Maintaining financial security-Dbtaining financial resources either through employment or assistance. 3. Maintaining housing and food-negotiating insurance claims or FEMA claims, dealing with contractors or landlords, keeping food fresh, etc. 4. Maintaining intimacy and calm within the family-feeling close and avoiding conflict with loved ones. 5. Dealing with personal losses caused by the storm-loss of connec tions to loved ones, loss of treasured belongings, and so on. 6. Going back to normal routine-grocery shopping, banking, schools, gas stations, work, and so on? 7. Dealing with the emotions you've experienced since the stormsuch as anger, anxiety or depression?


68 COPING SELF-EFFICACY FOLLOWING DISASTERS with lower norepinephrine to cortisol ratios, suggesting that the better the appraisal of coping capability, the better the physiological function. The third study with Chronic Fatigue patients also confirmed this effect. Sixteen CFS patients (80% female, average age of 39 years) were recruited during the first four weeks following the storm. CSE added significantly to the prediction of generalized psychological distress. Interestingly, given the relationship of depression to CFS, CSE also added significantly to the prediction of depression. Finally, CSE added significantly to the prediction of symptom exacerbation and illness burden. Even after controlling for current levels of psychological distress, elevated levels of CSE were associated with reduced reports of physical impairment, psychosocial impairment, and total impairment scores from the Sickness Impact Profile. Collectively, these studies supported the idea that how one appraises his or her coping capabilities following a natural disaster is important in how well they are coping right after the storm, and in the case of the first study, how well they are managing over time. Based on these initial investigations, two additional studies have been completed in an attempt to replicate these findings under different conditions (i.e., Hurricane Opal and the Oklahoma City Bombing). Hurricane Opal Hurricane Opal hit the panhandle of Florida on October 4, 1995, hammering coastal communities with winds reaching 144 mph. Although not as strong as South Florida's Hurricane Andrew, Opal inflicted millions of dollars in losses to many communities along the coast. We entered the field on December 1, 1995. Sixty-six participants were recruited in and around the community of Niceville. Average age of the participants was 55 years old, 48 % were male and 52 % female. This sample was predominantly Caucasian and was well-educated, with approximately 86% reporting at least some college. This sample was much different than the Hurricane Andrew sample, where participants were much more diverse ethnically and socioeconomically. It was of interest to us whether the CSE construct would still provide important information on recovery in such a different group and following a storm that was not as devastating. Results demonstrated very clearly that CSE is related to psychological distress even after taking into account the important control factors of gender, amount of social support and loss of resources.


Benight 69 Oklahoma City The last study was an investigation of whether CSE appraisals would be valuable in determining psychological distress following a very different type of disaster-the bombing of the Alfred P. Murrah building in Oklahoma City. With this study, a new CSE measure was created to reflect the types of demands that victims of this tragedy would be facing (see Table 2). Twenty-seven victims were recruited two months after the bombing, found through local businesses within a five-mile radius around the bombing site. The mean age of this sample was 41 years, and, of these individuals, 48% were men and 52 % were women. The mean income range reported for sample one was between $40,000 and $45,000 per-year. Educationally, 7% reported a high school education, 26 % some college, 37% college graduate, and 30% graduate education. Ethnically, almost the entire sample was Caucasian, with only 3 % African American and 3 % Native American. As with Hurricane Andrew, participants responded to an interview, fIlled out a series of questionnaires, and provided urine specimens. As with the other four studies, perceptions of CSE were found to be highly predictive of reported psychological distress. Thus, in five different samples under a variety of traumatic situations, perceptions of CSE have been found to be important in understanding the psychological reactions of people as they attempt to "get things back to normal." What implications do these findings have for the disaster response community? And, more specifically, what do these results suggest for mental health response teams attempting to intervene following a major traumatic event? Table 2: Coping Self-Efficacy Scale for a Terrorist Bombing This assessment is designed to have you think about important issues related to dealing with the bombing. For each of the situations described below, you are asked to rate how confident you are that you can successfully deal with them. Because people differ from each other in the way that they are dealing with the crisis there is no single correct response. The following items refer to specific behaviors. Please think about yourself currently, not as it was the day of the bombing. Using the following scale, please rate how capable you think you are to:


70 COPING SELF-EFFICACY FOLLOWING DISASTERS 1 2 3 4 5 6 7 I'm not at all Capable I'm moderately Capable I'm totally Capable 1. Do my job skillfully. 2. Comfort children. 3. Maintain religious/spiritual beliefs. 4. Comfort friends. 5. Return to life as it was before. 6. Have conversations about the bombing. 7. Accept what happened. 8. Find some meaning in what happened. 9. Express my feelings about what happened. 10. Not bring my stress reactions home. 11. Be able to concentrate. 12. Not lose my temper. The following items refer to thoughts related to the bombing. Please think about yourself currently, not as it was the day of the bombing. Using the following scale, please rate how capable you think you are in managing .... 1 2 3 4 5 6 7 I'm not at all Capable I'm moderately Capable Thoughts of people dying. 1. 2. Thoughts of babies dying. I'm totally Capable 3. Controlling distressing thoughts about the bombing. 4. Painful memories of the event (e.g., people screaming, people in pain, etc.). 5. Thoughts about the pain of the families who lost loved ones. 6. Thoughts about my own vulnerability. 7. Thoughts about this happening again. 8. Thoughts about getting back at the individual(s) who did this. 9. Thoughts of personal injury. 10. Thinking optimistically. 11. Memories of the stench or odors emitting from the area. The following items refer to visual images related to the bombing. Please think about yourself currently, not as it was the day of the bombing. Using the following scale, please rate how capable you think you are imagining ....


Benight I 2 3 4 5 I'm not at all Capable I'm moderately Capable Painful images of the event. 1. 2. Distressing dreams. 6 71 7 I'm totally Capable 3. Visual reminders of the bombing (e. g., T. V., Pictures, seeing the building etc.). 4. Controlling images of the event that come into mind. 5. Images of the dead or injured people. 6. Images of distressed family members who lost loved ones. 7. Images of the children who died. The following questions refer to emotions related to the bombing. Please think about yourself currently, not as it was the day of the bombing. Using the following scale, please rate how capable you think you are in managing ... 1 2 3 4 5 6 7 I'm moderately Capable Feelings of inadequacy. I'm totally Capable I'm not at all Capable 1. 2. Feeling of being on the edge of losing emotional control. 3. Worries about personal vulnerability. 4. Feelings of helplessness. 5. Restlessness. 6. Feelings of rage. 7. To not "lose it" emotionally. 8. My anger toward the person(s) responsible for the bombing. 9. Being strong emotionally. 10. 11. Feelings of anxiety. Sad feelings. 12. Depressive feelings. 13. Feelings of grief. For the general disaster response community, these results suggest that a variety of agencies (e.g., FEMA, insurance agencies) can have a direct effect on emotional disaster recovery. The questions on the Hurricane Coping Self-Efficacy Measure addressed efficacy perceptions related to rebuilding one's home, maintaining financial security, maintaining personal security. These issues are heavily influenced by agencies such as FEMA, the Red Cross, the National Guard, and insurance companies. For example, insurance


72 COPING SELF-EFFICACVFOLLOWING DISASTERS companies varied widely in how quickly they provided initial funds to begin rebuilding following Hurricane Andrew. It was very clear in various neighborhoods which households had received an insurance check and which had not. Households with money were able to engage in active coping strategies such as buying a generator for electricity, renting a trailer for shelter, etc., and, once individuals are able to move forward in beginning the rebuilding process, coping self-efficacy perceptions will most likely improve. This type of situation, of course, refers to the middle and upper income population. Lower socio-economic groups, who often do not have insurance or are dependent on a landlord for housing needs, require other assistance that enables them to enact active coping behavior. For example, there were repeated stories following Hurricane Andrew of apartment complexes condemned by city and state inspectors. However, the only option for these individuals was to move all of their belongings into a tent city. It may make more sense to decentralize these types of housing alternatives into separate communities, thereby retaining some sense of the original living situation, again, realizing that the primary coping demand is to get things "back to normal" as quickly as possible. Obviously, it is logistically complex to accomplish; however, following Hurricane Andrew, the tent cities did not draw a huge population. Alternative strategies are necessary. In relation to mental health response, the current gold standard is Critical Incident Stress Debriefing following a major disaster. However, recent research is calling into question whether this is the most appropriate intervention strategy (see Kenardy, Webster, and Carter, 1996). It may be time to re-examine this intervention and see if other mental health techniques might be utilized to improve our strategies. For example, cognitive behav ioral therapy techniques such as goal-setting, cognitive restructuring, modeling, and reward systems might be creatively utilized to maximize individual appraisals of mastery. In the context of hurricane recovery, interventions might help individuals "re-set" their typical daily expectations, which are often unrealistic in the midst of the chaotic recovery demands, to more realistic goals that would enhance rather than decrease perceptions of coping self-efficacy. For example, daily goals may shift from getting to work, shopping, and picking up children to obtaining ice and spending one hour picking up debris.


Benight 73 Summary In five separate samples, our research has demonstrated a strong relationship between how capable individuals feel in dealing with disaster specific demands and how well they are able to adapt. The data suggest that the greater your perceived coping efficacy, the lower psychological distress levels and physical reactivity. Disaster response teams and mental health intervention strategists might be able to utilize these data to improve post-disaster response and lessen the psychological impact of these tragedies. References Adams, P.R. and G.R. Adams 1984 "Mount Saint Helen's Ashfall: Evidence for a Disaster Stress Reaction." American Psychologist 39: 252-260. Bandura, A., L. Reese, and N.B. Adams 1982 "Microanalysis of Action and Fear Arousal as a Function of Differential Levels of Perceived Self Efficacy." Journal of Personality and Social Psychology 43: 5-21. Bandura, A., C.B. Taylor, S.L. Williams, LN. Mefford, and J.D. Barchas 1985 "Catecholamine Secretion as a Function of Perceived Coping Self-Efficacy." Journal of Consulting and Clinical Psychology 53(3): 406-414. Freedy, J.R., D.L. Shaw, M.P. Jarrell, and c.R. Masters 1992 "Towards an Understanding of the Psychological Impact of Natural Disasters: An Application of the Conservation of Resources Stress Model." Journal of Traumatic Stress 5: 441-454 Kenardy, J.A., R.A. Webster, and G.L. Carter 1996 "Stress Debriefing and Patterns of Recovery Following a Natural Disaster." Journal of Traumatic Stress 9: 37-43. Meuller, P, and B. Major 1989 "Self-Blame, Self-Efficacy, and Adjustment to Abortion." Journal of Personality and Social Psychology 57: 1059-1068.


74 COPING SELF-EFFICACY FOLWWING DISASTERS Murphy, S. 1987 "Self-Efficacy and Social Support Mediators of Stress on Mental Health Following a Natural Disaster." Western Journal of Nursing Research 9: 58-86. Ozer, E.M. and A. Bandura 1990 "Mechanisms Governing Empowerment Effects: Self-Efficacy Analysis." Journal of Personality and Social Psychology 58: 472-486. Rubonis, A.V. and L. Bickman 1991 Psychological Impairment in the Wake of Disaster: The disas ter-Psychopathology Relationship." Psychological Bulletin 109: 384-399. Solomon, S. D. 1989 "Research Issues in Assessing Disaster's Effects." In R. Gist and B. Lubin (Eds), Psychosocial Aspects of Disaster. 308-340. New York: Wiley. Solomon, Z., M. Weisenberg, J. Schwarzwald, and M. Mikulincer 1988 "Combat Stress Reaction and Post-Traumatic Stress Disorder as Determinants of Perceived Self-Efficacy in Battle." Journal of Social and Clinical Psychology 6: 356-370. Wiedenfeld, S.A., A. O'Leary, A. Bandura, S. Brown, S. Levine,and K. Raska 1990 Impact of Perceived Self-Efficacy in Coping with Stressors on Components of the Immune System." Journal of Personality and Social Psychology 59: 1082-1094.


CmJRCH WORLD SERVICE AND LESSONS LEARNED FOR MITIGATION Kristina J. Peterson and Richard L. Krajeski Church World Service Introduction At the end of most TV editorials comes the phrase, "The views of the speaker are not necessarily the views of this station." Similarly, this paper contains our reflections on the activities of Church World Service Disaster Response (CWS) as we have experienced them over the past 15 years, not necessarily the views of CWS. We believe that CWS and its related groups have been most effective in their work in disaster recovery. Background Church World Service (CWS), which was formed after the Second World War as a cooperative effort to help restore and rebuild Europe, operates on behalf of 51 religious bodies and organizations in the u.S. CWS has both national and international programs as well as three program branches: 1) refugees, 2) hunger\development, and 3) disaster response, which we will address. Basic Philosophy The slogan for the Disaster Response office is "Prepared to Care," and its efforts are based on the understanding that disaster response takes place in a cycle that usually starts at the disaster impact and continues through the preparation for the next disaster. Activities encompass six phases: response, development, education, prevention, mitigation, and preparedness. The CWS response is driven by a set of value statements that focus on: seeking to identify and aid the most vulnerable people in disaster-the poor, elderly, children, physically challenged, ethnic minorities, and women. providing advocacy for the disenfranchised and the environment, and


76 CHURCH WORLD SERVICE AND LESSONS LEARNED FOR MITIGATION giving spiritual/psychological care and support for survivors, care givers, and communities. Other value assumptions of CWS and the groups with which it works include: empowering local peoples and communities (we believe that commu nity is defined from the "inside"), enabling long-term recovery and sustainable development, and seeking in partnership with local people the establishment of basic human rights. CWS is committed to operating on an open and nonexclusionary basis. Basic Operations CWS is active outside the U.S. and provides financial and technical assistance to local partner groups in the affected country. Within the U. S. support is given by providing funding through the CWS member groups, some material aid, and technical assistance and organizing support. CWS has paid staff and about 35 disaster response consultants who volunteer in the U. S., working for periods that range from three weeks to all the time they have in order to help communities recover from disasters. Consultants are sent into an area to: assess the situation and the needs of the affected areas; help identify those who are on the margin and often invisible; enable the organization of the affected areas, using the religious community as the base for building grassroots coalitions; and provide training and consultation for long-term recovery and develop ment. Working through state interfaith groups and local religious groups, the consultant helps build a formal and informal coalition that usually expands to include many community groups. A typical interfaith organization may include Christians, Jews, and Moslems (such as the group that came together following the Oklahoma City bombing), together with service and neighbor hood clubs, trade unions, and businesses. This organization will work to physically repair and rebuild the community, and more and more often also to address the pre-existing issues that turned an event into a disaster. Consultants in the field try to sensitive to both the local cultures (we never

PAGE 100

Peterson and Krajesld 77 assume that there is just one "culture") and the marginalized populations who may not be recognized and represented by local governments or other institutions. Marginalized groups are often reluctant to respond to formal government systems or other entities that may try to help during a disaster. People are often suspicious of individuals or agencies that hold over them the power of income or services. In our work, we have found that people are much more willing to trust a person from the religious community, even if that person is a stranger; a representative from the religious community can be an advocate without negatively impacting their day-to-day existence. The recovery and mitigation work of CWS and the local religious communities takes on a different form than that of a more formal group such as the Red Cross or the Federal Emergency Management Agency. Recovery and mitigation begin with the training of individuals in the vulnerable community in leadership skills, thus helping to open the doors of the "formal" structures to them. The local leadership can then work on the vulnerabilities of the local population. Although CWS and the religious communities are active in structural recovery and mitigation (rebuilding and elevation-for example), our concern also focuses on pre-existing nonstructural conditions. The barriers that produce vulnerability are correctable, some more easily than others. These barriers include, but are not limited to, housing, literacy, transportation, language, water, toxic dumps, jobs, and poverty. Following disasters, we hear the survivors, caregivers, and local government saying that they want to get the community back to normal. Normal seems to be a goal for most, but for vulnerable groups, normal may mean a return to oppressive and substandard conditions. Some recovery and mitigation policies actually worsen vulnerability. For example, a home that had a leaky roof before the disaster needs to be replaced in total, but some relief and recovery agencies will only grant repairs to restore the structure to "pre-existing conditions," that is, the state of the home prior to the disaster. In addition, mitigation for vulnerable populations must often include justice, jobs, and land-use reform. The religious community is generally more free to act than other recovery and mitigation agencies and can focus on doing what is necessary for the good of the survivors and their communities. The religious community only needs to think about what is right, just, and ethical; therefore, it has the freedom and flexibility to employ creative mitigation measures. It does not always live up to this call and sometimes is as short-sighted as others. But because it tends to be community based, representative, and inclusive, the

PAGE 101

78 CHURCH WORLD SERVICE AND LESSONS LEARNED FOR MITIGATION' religious community does have a freedom to act with and for the survivors and their communities. Some examples of creative mitigation on the part of local interfaith groups follow: Following an earthquake, many survivors were evicted from their rentals by a slum landlord. Rather than putting these folks back into inadequate housing, money was sought from the public sector and used by volunteer labor from the religious community to rehabilitate homes that were safer, healthier, and at less risk of damage or collapse due to earthquakes. After a major hurricane, the mayor of a southern town wanted to dispose of toxins collected in the clean up by dumping them near a very poor rural community, contaminating the community's water and more than likely inundating homes with toxic fumes over a long time period. Mitigation in this case involved stopping the illegal toxic dumping and monitoring its safe disposal. In a buy-out program in southern Louisiana to remove residences from floodplain areas, mitigation by the religious community meant putting homes and trailors on elevated bases and strapping down roofs to withstand strong winds. Following the Midwest floods and the frequent flooding in the Appalachian region, mitigation involved getting affordable housing out of the floodplain. After most recoveries that involve the interfaith community, new local grassroots organizations are created that work for the protection of women and children, support sustainable development, and empower people, These grow almost spontaneously out of grassroots recovery groups. Lessons We have learned the following from our work: People on the margin and the poor are experts at mitigation, although they may have never heard the word. We from the outside must remember that, for many people, "mitigation" is their daily bread for surviVal. We must learn that the survival skills of these popUlation are

PAGE 102

Peterson and Krajeski 79 very good, and we must find ways to enhance their skills and support them. They do not want to live that way. We still hear from "outsiders," that is, certain professionals from other helping organizations, that "they like to live that way," or that "they do not know any better or simply don't care." Views like these are untrue and more often than not based on intentional or unintentional racism, sexism, and classism. Those who try to be sensitive to and respect local cultures and local enpower ment and those who try to be academically objective must beware that their "hands-off-Iet-the-Iocals-do-it" attitude can be racist, sexist, and classist, particularly if their lack of involvement perpetuates the status quo. Mitigation is more than structural change-it is also about economic and social justice. Local communities will often function through informal networks and not through agencies and institutions. Religious beliefs are just as often powerful motivators for liberation and creating just communities as they are for oppression and maintain ing the status quo. Giving technical assistance (i.e., basic organizational skills, resource identification) and support to local people after a disaster enables them to recover and make nondisaster-related changes in their lives and communities. Local folks are going to be there after we leave, with whatever we leave behind. They need our help-not our garbage. Kristina J. Peterson holds degrees in Urban Planning, Theology, and Ethics. Her doctoral studies were in Peace Making. Richard L. Krajeski holds degrees in Philosophy, and Theology, and his doctoral studies were in Technological Transfer and Ethics. They are disaster specialists for CWS and the major authors of three disaster recovery manuals.

PAGE 103

This page IS intentionally blank

PAGE 104


PAGE 105

This page IS intentionally blank

PAGE 106

NATIONAL WEATHER SERVICE Anv ANCED CAPABILITIES IN FLASH FLOOD FORECASTING Lee W. Larson Hydrologic Research Laboratory National Weather Service, NOAA Silver Spring, Maryland Introduction The modernization of the National Weather Service (NWS) includes three major systems: the Doppler Weather Surveillance Radar (WSR-88D), the Advanced Weather Interactive Processing System (A WIPS), and the Automated Surface Observing System (ASOS). These advanced technologies are providing significant data and processing capabilities that are directly applicable to the flash flood problem. In particular, the Weather Forecast Office (WFO) Hydrologic Forecast System (WHFS) provides forecasters with unparalleled access to real-time data and the capability to process and identify potential flash flood situations. All of these technologies contribute to improved capabilities of the NWS to provide early and useful flash flood products to cooperators and the public. On July 31, 1976, more than 12 inches of rain fell in the Big Thompson Canyon in Colorado. The resulting flood left more than 140 people dead and destroyed homes and businesses. There has been a significant effort within the NWS over the last 20 years to improve our capabilities to respond to these types of events. Background Following the Big Thompson flash flood, in an effort to provide the WSFO with the tools necessary to effectively forecast these types of events, the Forecast Systems Laboratory (FSL), with support from the NWS, began the Program for Regional Observing and Forecast Services (PROFS), which

PAGE 107

84 NATIONAL WEATHER SERVICE CAPABILITIES was a proof of concept project for an NWS field office advanced work station environment. The PROFS work stations added considerable functionality in data handling, graphical display capability, and additional local model generation capability. The NWS and FSL, as a risk-reduction activity, then cooperated in the Denver A WIPS Risk Reduction and Requirements Evaluation (DAR3E) in the mid-1980s. The DAR3E project was designed to put a series of the PROF developed work stations in an operational Weather Service Forecast Office (WSFO). PROFS and the early DAR3E implementa tions had limited functionality to address the hydrologic operations at the WSFO. Over the next few years, through 1992, some limited success was achieved in adding hydrologic displays and applications to the DAR3E system and a later pre-A WIPS system. By mid-1993, the hydrologic application development on the pre-A WIPS system and the WSFO hydrologic develop ment effort was moved to the Office of Hydrology (OH). Figure 1 WFO Hydrologic Database Site Specific Pam Break db t l ______ l ____________ J ______ .J Station Reports Observer Hydrologic forecast Applications Precipitation River I "" ",i }. It :: f.:' ;1.,)(:;);; .. ,., .. :" Incorporated Pre AWIPS Patabase

PAGE 108

Larson Weather Forecast Office Hydrologic Forecast System 85 The WHFS features an integrated data management approach, employing a relational database management system (RDBMS) for storing the large volume of data necessary for hydrologic forecast operations. The WHFS database incorporates many data elements ranging from modernized data sets, such as NEXRAD precipitation estimates and GOES satellite imagery, to more traditional hydrologic data sets provided by automated reporting stations and cooperative observers. The supporting River Forecast Center (RFC) is the primary source of hydrologic guidance for the WFO, providing river stage forecasts on a daily and event-oriented basis. RFC guidance is also provided in the form of modernized flash flood guidance products that indicate current soil moisture conditions and associated rainfall thresholds necessary to induce flood activity. A collection of tools is provided within WHFS to manage the WFO hydrologic program through a series of graphical user interfaces. HydroBase, one of these tools, provides a method of managing station data, allowing for definition of various station attributes. Much of the data utilized by forecast applications are also defined through HydroBase. Program management tools, such as automated generation of monthly flood stage reports, are also part of HydroBase. WHFS Capabilities During a typical hydrologic situation, the forecaster may employ many aspects of the WHFS in combination to evaluate the current hydrologic conditions, evaluate data, and issue products notifying the public of flood situations. The Stage and Display (HydroView) application provides the forecaster with a method of monitoring and tracking the situation in real time. This application provides a geographic depiction of the WFO County Warning Area (CW A) with the ability to overlay an array of Hydrometeorological data. Station icons may be overlaid in combination with hydrologic or geopolitical boundaries such as rivers, river basins, county outlines, or major towns and highways. River station icons are color coded to indicate the proximity of the latest observation to action or flood stage. Precipitation stations are color coded to represent a precipitation accumulation for a selected time duration. This display is automatically refreshed at 15 minute

PAGE 109

86 NATIONAL WEATHER SERVICE CAPABILITIES Figure 2 Stage and Precipitation Display Data edit capability Geo-political overlays Data Overlays Observed river stages Observed precipitation Stage I precip Gridded FFG WFOQPF intervals, using the most recent observations and forecasts available. The forecaster may also view a time series display of river stage and precipitation observations for a period of up to 21 days. Forecast data is provided for a five-day time period. The Area Wide Hydrologic Prediction System (A WHPS) provides the forecaster at the WFO with an analysis of a flash flood threat in the WFO forecast area. A WHPS uses data from NEXRAD and gridded flash flood guidance from the servicing RFC to provide a graphical depiction of: 1) Critical Rainfall Probability (CRP), 2) one-hour rainfall projection, and 3) a difference display. The NEXRAD product that is used in the A WHPS system is the Hourly Digital Precipitation (HDP), which provides a gridded accumulation of precipitation for the previous hour each volume scan of the radar. The modernized flash flood guidance from the RFC indicates, for each HRAP grid, the amount of rainfall required in a particular duration to cause over-bank flood of small streams. The common durations for the rainfall in the flash flood guidance computations are 1, 3, and 6 hours. Two CRPs are computed for each duration: the first is the CRP based on the radar estimated rainfall, and the second is the CRP based on the radar estimated rainfall plus the one-hour projection. The CRP gives a statistical

PAGE 110

Larson 87 Figure 3 Area Wide Hydrologic Prediction System probability that the rainfall in a particular HRAP grid has exceeded the flash flood guidance for that grid square. The difference fields are a graphical depiction of the quantitative difference between the flash flood guidance and the radar estimated rainfall for each duration. A second difference graphic will depict the same information for the radar estimated plus one-hour projected rainfall totals. Utilizing the CRP products, the forecaster will be able to outline the potential flash flood area. At that point, the forecaster can issue the appropriate public product, either a flash flood watch or a flash flood warning. The Site-Specific Hydrologic Prediction System (SSHPS) is a local hydro logic model provided to allow the WFO forecaster to supplement RFC river forecast guidance by generating forecast river stages for fast response headwater and river basins. River stage observations and precipitation estimates are provided as input to a simplified rainfall runoff model, which produces an estimate of streamflow rise due to runoff reaching the river channel. Initial soil moisture conditions are accounted for through model state variables provided by the RFC. Dependent upon the model definition, other inputs such as snowmelt runoff and potential evapotranspiration may also be considered. Model definitions for individual basins are calibrated by the

PAGE 111

88 RFC Support Model Calibration Daily Model Carryover NATIONAL WEATHER SERVICE CAPABILITIES Figure 4 Site Specific Model River Level Past Future RFC, employing the NWS River Forecast System (NWSRFS) hydrologic models as a baseline. Gridded or point precipitation estimates may be used as model input, and may be selected by the forecaster prior to the execution of the model. Each of these forms of estimates' is ingested through a precipitation preprocessor that calculates basin average precipitation values for an amount of time specified by the model definition. Gridded estimates are utilized on a best-available basis employing Stage III, Stage II, and Stage I NEXRAD estimates. Future precipitation estimates may be incorporated through the assimilation of gridded Quantitative Precipitation Forecasts (QPF) products generated by the WFO, the RFC, or a national center. The forecaster interacts with the SSHPS through a graphical user interface that allows for interactive review and adjustment of model results and input. Conclusions WFO hydrologic forecast operations in the A WIPS era will differ dramatically from those in the pre-modernized NWS. The advent of more powerful computing technologies provides the opportunity to implement sophisticated hydrologic modeling, analysis, and forecast tools in a manner

PAGE 112

Larson 89 suitable for dealing with the wide range of possible hydrologic conditions and situations. Significant portions of the initial WHFS capability will be fielded to WFOs beginning in the fall of 1996, with full hydrologic forecast capability available shortly thereafter. This WHFS implementation will provide the WFO forecaster with the tools necessary to meet the goals of the NWS hydrologic services program, serve as the baseline for future enhance ments, and dramatically enhance the WFO's ability to identify and respond to short lived hydrologic events such as flash floods. Acknowledgments Significant contributions by Edwin L. May, NWS, Fort Worth, and Dale R. Shelton, formerly of the NWS, is gratefully acknowledged. References Shelton, Dale R. and Edwin L. May 1996 "Modernized Hydrologic Forecast Operations at National Weather Service Forecast Offices," 12th International Conference on Interactive Information and Processing System (IIPS) for Meteorology, Oceanography, and Hydrology. American Meteorological Society, Atlanta, Georgia, January 28-February 2, 1996.

PAGE 113

COMPARISON OF DEFICIENCIES ASSOCIATED WITH THE BIG THOMPSON FLASH FLOOD EVENT AND RECENT FLOOD EVENTS IN THE EASTERN UNITED STATES Solomon G. Summer NOAA/National Weather Service Eastern Region Bohemia, New York Introduction Since the 1976 Big Thompson flash flood, there has been significant progress toward improving the National Weather Service (NWS) flood and flash flood program. This is the result of new technology, a better under standing of the meso-scale hydrometeorological features that cause flash flooding, and the significant attention focused on flood forecasts and warnings as part of the NWS Modernization and Associated Restructuring (MAR). Yet, despite these efforts, there still remain stubborn areas where significant progress in overall warning capabilities has yet to be realized. This paper provides a service perspective, comparing the deficiencies associated with the Big Thompson Flash Flood with several flash floods that occurred in the Eastern region of the U. S. from 1982 to 1995. A comparative analysis of the deficiencies shows those areas where deficiencies have been addressed and identifies others that are still problem areas. Based on this analysis, general recommendations are offered for future improvements in support of the NWS Flash Flood program. Deficiencies In the Big Thompson Canyon Flash Flood of 1976, several deficiencies were identified in the NOAA Natural Disaster Survey Report for this event. They include: 1) Sparsity of real-time rainfall/river data in the river basin,

PAGE 114

Summer 91 2) Unfamiliarity among staff of hydrologic operations, 3) Inadequate radar coverage for precipitation estimates, 4) Lack of site-specific information and conveyance of the urgency of the event in warnings, 5) Insufficient warning dissemination-NOAA Weather Radio 6) Need for increased local awareness and preparedness activities, and 7) Difficulties in accurately forecasting excessive rainfall capable of producing flash flooding. Eastern Region Flood Events Six recent floods that occurred in the eastern United States were selected for analysis. The events all share the common features of being associated with significant flood-related deaths and/or damage and also provide a cross section of the types of flash flooding experienced in the East. They include: events with tropical origins; those resulting from slow-moving or training of thunderstorms; prolonged heavy rainfall events; terrain-enhanced rainfall; and wintertime events as a result of both rainfall and melting snows. The events chosen for analysis are: 1) June 1982 Southern New England floods; 2) November 1985 flood in central Appalachians; 3) May 1986 Little Pine Creek flash flood; 4) June 1990 Shadyside, Ohio, flash flood; 5) June 1995 Madison County, Virginia flash flood; and 6) January 1996 floods, Northeastern U.S. The June 1982 Southern New England Floods Synopsis: Prolonged and excessive rainfall that fell from June 4-6, 1982, caused severe, and in some cases, record, flooding in southern New England. The major damage occurred along small streams in coastal Connecticut and Rhode Island, where 48-hour rainfall amounts exceeded 12 inches, with one unofficial report of 17 inches. There were 12 deaths attributed to the flooding in Connecticut, and three in Rhode Island. Damage estimates from this event were $277 million in Connecticut and $3.3 million in Rhode Island. More

PAGE 115

92 DEFICIENCIES ASSOCIATED WITH FLASH FLOOD WARNINGS than 20 small dams were reported destroyed or partially breached. Five dams failed along the Falls River alone in and around Essex, Connecticut. Analysis. Flood potential statements preceding the event, based on general forecasts for heavy precipitation, heightened the awareness to flooding. However, the recognition of the magnitude of this event was not sufficiently conveyed to the general public and disaster officials, except in a few cases where direct telephone contact was made by the NWS. Lack of data from the most seriously affected areas was a major factor in assessing the flood threat. Adding to the problems were general inconsistencies in usage of hydrologic products between offices with responsibilities in the area. Widespread dissemination of warnings and forecasts were communicated via conventional mass-media, the NOAA Weather Wire, and NOAA Weather radio. No unusual problems were encountered. Aside from standard communications, the Rhode Island Civil Defense office was alerted by telephone, early in the event, as a result of the flood potential statement released by the NWS. This allowed the state to mobilize for weekend disaster operations. The Connecticut Office for Community Preparedness was not directly notified before or during the event. Flood damage was mitigated locally in the Connecticut town of Norwich as a result of a local self-help warning system. The November 1985 Appalachian Floods Synopsis. Starting in the headwaters on November 4 and 5, 1985, and continuing downstream for several days, record floods, some estimated to be 100-year to 500-year flood frequencies, occurred on several rivers in Virginia and West Virginia. Record floods occurred in portions of the James and Roanoke basins in Virginia, and in the upper Monongahela, upper Potomac, and Greenbrier rivers in West Virginia. Fifty-six people lost their lives, and total damages exceeded $1.3 billion. These floods were caused by the combination of two separate storms. The remnants of Tropical Storm Juan passed over the area during the first days of November 1985, causing moderate to heavy rainfall. However, on November 4, a strong low pressure system from the Gulf of Mexico deepened as it slowly moved northward into West Virginia and eastern West Virginia. Peak rainfall amounts of up to 18 inches and 14 inches were reported for Montebello, Virginia, (the James basin) and Milan, West Virginia (the Potomac basin), respectively. Analysis. In this event, due to lack of real-time automated rainfall and river data in the headwater areas and an underestimation of forecast rainfall, the

PAGE 116

Summer 93 urgency of the event was not initially portrayed. However, once the magnitude was realized, subsequent releases portrayed the severe nature of the flooding. Despite the use of NOAA Weather Wire, NOAA Weather Radio, and NA WAS, it was noted that "in this day of sophisticated telecommunication technology, it was shocking to see the dependency on the telephone to alert key officials and the news media of impending danger." The FEMA Interagency Hazard Mitigation report stressed that many communities do not have the tools to enforce basic floodplain management standards. They reported, "public awareness of flood hazards seems to be limited to recent memory and many were simply unprepared for the magnitude and dangers of the November flood." The May 1986 Little Pine Creek Flash Flood in Pennsylvania Synopsis. On May 30, 1986, a stationary thunderstorm dropped up to eight inches of rain in a little over two-and-a-half hours over a small portion of the North Hills area in metropolitan Pittsburgh, Pennsylvania. The resultant flash flood down the Little Pine Creek claimed nine lives. The basin area of Little Pine Creek is just 6.1 square miles. All the deaths were car-related; victims were traveling along a thoroughfare paralleling the creek. Analysis. Rainfall associated with this event was highly localized. Automated rainfall reports were available from the Integrated Flood Observing and Warning System (lFLOWS) to the local Weather Service Forecast Office (WSFO), but none of these were in the immediate area of the event. Only a limited amount of radar rainfall estimates were available to the forecasters in real time due to the proximity of the ground-clutter pattern. Vague reports of flooding were received in the office, but attempts to pin down the degree of flooding were unsuccessful. A flash flood warning was issued about one hour after the extreme flooding began on Little Pine Creek, based on scattered reports of flooding. The flash flood warning received excellent dissemination over the Emergency Broadcast System (EBS) however, no radio stations interviewed could recall broadcasting information contained in the special weather statements about the possibility of flooding. The local NOAA Weather Radio (NWR), which was not operational during much of the day, was restored to low power by late in the afternoon. Eye witness accounts of the flash flood indicated at least three of the nine-flood related deaths occurred when victims climbed on top of their vehicles to evade flood waters rather than leaving their vehicles for higher ground.

PAGE 117

94 DEFICIENCIES ASSOCIATED WITH FLASH FLOOD WARNINGS The June 1990 Shadyside Ohio Flash Flood Synopsis. On the evening of June 14, 1990, severe flash flooding occurred on Wegee and Pipe creeks (approximately 12-square-mile drainages) near Shadyside, Ohio. There had been no recent flash flooding on either of these two creeks. Recollections of past flooding were mainly of the backwater effects of the nearby Ohio River. Thus, public awareness of the possibility of such a devastating flash flood was nonexistent. Unofficial estimates of five inches of rain, up to four inches in one hour, fell on the headwaters of these two small creeks. The resultant flash flood occurred within an hour, cascading a 10-to 30-foot wall of water, according to eyewitness accounts. The flash flood resulted in 26 fatalities. Analysis. A flood watch was issued approximately two hours before the flash flood at Shadyside and was based on an analysis of synoptic and mesoscale conditions. The flood watch was given timely distribution by the local media to the residents in Belmont County, and many people in the flood area reported seeing the watch on television and hearing it on commercial radio. NOAA Weather Radio was ineffective due to poor reception in the area. A flash flood warning was not issued for this event. Neither radar data nor observed rainfall reports prior to or during the flood indicated the magnitude of the actual rainfall. Reports of flooding did not reach any NWS office until several hours after the event, following reports of bodies and debris floating on the Ohio River. The NWS survey team summarized the Shadyside flood by stating that, due to the small scale nature of the event and the rapidity at which it evolved, it was beyond the detection and warning capability of current NWS field technology and may even approach the limits of improved capabilities expected in the near future. The June 1995 Madison County, Virginia, Flash Flood Synopsis. Severe flash flooding, and in some cases record river flooding, occurred across portions of west-central Virginia during the week of June 25, 1995. The hardest hit area was centered over southern Madison County, where as much as 20 inches of rain may have fallen. One report indicated that as much as 10 inches of rain had fallen in two hours. The Rapidan River in this event approached the 500-year flood recurrence. Despite the magnitude of the event, only one death was reported in Madison County. Analysis. The NWS staff was very proactive during this event. Strong wording was used in their issuances. The June 27 midnight public forecaster,

PAGE 118

Summer 95 prior to the significant flash flooding that occurred in Madison County, conveyed the seriousness of the threat and virtually pinpointed the location of the most devastating flooding through sound meteorological analysis, backed up by the use of several data sources. The WSR-88D radar was key to the location of maximum rainfall as well as providing useful information from the V AD wind profile in showing the strengthening low-level flow. IFLOWS automated rainfall data provided ground-truth rainfall amounts prior to and during the flooding that helped calibrate the radar rainfall estimates. The IFLOWS system was also used by local county and state officials in their disaster assistance efforts. Skywarn reports from amateur radio operators provided valuable additional rainfall reports. Flash flood watches and warnings were issued for Madison County with extensive lead time. The watch preceded the flooding by 14 hours. The warning preceded the onset of flooding by two hours and the time of serious flooding by five hours. Emphasizing the flood threat further, a flood potential statement was issued that included a call-to-action statement for emergency managers because significant flooding was possible. As the flash flooding became more and more life-threatening, radio stations were called and asked to activate EBS, and continuous contact was made with emergency service personnel. River flood warnings and follow-up statements were routinely issued for the Rapidan and Rappahannock rivers. There was excellent coordination between the WSFO staff and the RFC staff in providing accurate river forecasts. The Hydrometeorological Analysis and Support (HAS) function at the RFC coordinated with the WSFO to input four inches of forecast rainfall (QPF) into the hydrological models. This was translated into timely warnings of major flooding on these river basins. January 1996 Northeastern U.S. Flood Synopsis. Although this flood can be more accurately described as a river flood event, the elements that contributed to the flooding resulted in rapid rises on major rivers more closely associated with flash flooding. The Susquehanna River Basin Commission referred to this event as a basin-wide flash flood over the entire Susquehanna watershed. The last time this occurred on such a large scale was Hurricane Agnes in 1972, and not surprisingly, the January 1996 floods in the Northeastern U.S. produced the most extensive river flooding since Agnes. One example of the devastation on the smaller streams was in Lycoming County in west-central Pennsylvania, where six deaths were reported.

PAGE 119

96 DEFICIENCIES ASSOCIATED WITH FLASH FLOOD WARNINGS Loyalsock Creek reached a record flood crest, exceeding Agnes by three feet, and the six deaths were reported on another hard-hit stream, Lycoming Creek. Particularly hard-hit was Old Lycoming Township at the lower part of the Lycoming Creek basin, where trailers were washed away and homes were heavily flooded. The conditions causing this event were a nearly unprecedented January snowmelt thaw (resulting from near-record to record snowpack), accompa nied by heavy intense rains (two to four inches), and adding to the recipe for disaster, significant ice jams on the rivers and streams. Total losses were 33 deaths and nearly $2 billion in damage. Many of the deaths occurred in stranded autos. Analysis. Most offices provided timely flood potential statements, watches, and short-fused warnings. Quantitative precipitation forecasts (QPF) were under-forecasted, and hydrologic models did not handle well the nearly unprecedented rate of snowmelt runoff. As a consequence, initial river flood warnings underestimated the magnitude of the event. However, when the effects of the rapid rate of runoff were noted in area streams, updated forecasts more accurately specified the magnitude as a major flood. As a result of timely and accurate forecasts at Wilkes-Barre, Pennsylva nia, on the Susquehanna River, more than 100,000 residents were safely evacuated. Emergency management response, particularly at the state and county level, was favorable. At the local level, this was not evident in some cases. Generally, the awareness level of the event, knowledge of NWS flood forecast and warnings, and correct response measures were higher at the county and state levels and deteriorated by the time information reached smaller communities. There was some misinformation passed on by the new media during this event that resulted in public confusion on what was actually occurring on the rivers and streams. The performance of the WSR-88D radars varied from office to office with respect to the estimates of precipitation. The meteorological conditions causing the heavy rain were primarily synoptic-scale, but exhibited some tropical characteristics. Automated satellite-telemetered river and rainfall gauges installed as part of the Susquehanna flood forecast initiative were also valuable in providing real-time data to the NWS offices, including the RFC. Comparative Analysis Table 1 shows a comparative summary of the six flood events relative to the deficiencies discussed in the Disaster Survey Report associated with the Big Thompson flash flood event. For each event, if a deficiency has been

PAGE 120

Summer 97 fully addressed, it is rated "Yes." If it has not been addressed and is still a problem area, it is rated "No." If a particular element has been partially addressed, it is noted as "partial." Table 1. Deficiencies Addressed Events 1 2 3 4 5 6 7 June, 1982 No No No No Partial No Partial Nov.,1985 No Partial No Partial No No Partial May, 1986 No Yes No No No No No June, 1990 No Partial No No No No Partial June, 1995 Partial Yes Yes Yes Partial Yes Yes Jan., 1996 Partial Yes Partial Partial Yes Partial Partial NWS modernized operations began in the eastern region in 1994-95. For the four events prior to modernization, only one element was fully addressed. For the two events following the modernized operations, all of the deficien cies were at least partially addressed. Among the individual deficiencies, the two clements that have shown the most progress are: Element 2-knowledge of hydrologic operations; and Element 7-use of QPF in hydrologic forecast operations. This is not surprising, since these areas have been emphasized in eastern region operations since 1980. The one deficiency that was somewhat surprising was Element I-the availability of real-time automated data. On further reflection, however, this can be explained by the fact that, as a direct result of the first four events, automated flood warning systems were either newly implemented or expanded for the areas of concern. Current and Future Operational Improvements The most significant improvements in the NWS Flash Flood Warning program are the result of NWS modernized operations and technologies. Despite these improvements, problems still exist where deficiencies have only been partially addressed. The use of WSR-88D radars has been a key component in pinpointing areas of excessive rainfall. Significant improvements have been made in radar/rainfall estimates and areal coverage of precipitation with the advent of

PAGE 121

98 DEFICIENCIES ASSOCIATED WITH FLASH FLOOD WARNINGS the WSR-88Ds in forecast operations. Nevertheless, considerable progress still needs to be made in improving precipitation estimates through calibration with gauge data and enhanced bias adjustment algorithms (i.e., tropical rainfall). There has been an increasing effort to familiarize NWS staff with hydrologic operations and make hydrology an integral part of overall station operations. This is mainly due to the emphasis in coupling hydrology with meteorology and an integrated "team" approach in implementing the NWS modernization and restructuring project. High-resolution mesonet data has been on the increase in flash flood areas due to the expansion of ALERT and IFLOWS networks. This is especially true of rain gauge data. There is, however, an increasing concern over the availability of river gauge data due to budget cuts imposed on the U.S. Geological Survey, which has as part of its mission a directive to gauge the nation's rivers, and for cooperative gauge data funded in part by state and local agencies. Decreasing budgets for river gauging have not only led to decreases in new gauge sites, but have in some instances led to gauge closures at existing forecast locations. It is ironic that, with the increased availability of mesonet data, the basic data used for providing information on river flows is threatened. Overall, there are still many flash flood-prone areas without mesonet data. Although the implementation of the WSR-88Ds may reduce the need for extensive rainfall data, calibration of radar data with gauge data will be essential to provide accurate point and areal estimates of rain. This is particularly true in areas of significant terrain, where radar coverage may be affected. The expansion of mesonets will require a concerted effort toward a public-private partnership. Improved rainfall forecasts can provide additional lead time for flash flood warnings. Annual median lead-time verification statistics were calculated for all eastern region flash floods from 1992 to 1995. In 1992 and 1993, median lead-time for each of these years was zero minutes. Described another way, the majority of events was preceded by zero lead-time. In 1994 and 1995, WSR-88D radars were implemented at most sites. Improvements in flash flood warning lead time resulted. The median lead time jumped to 6.25 minutes in 1994, and to 11.16 minutes in 1995. Although these results were positive, further improvements in warning lead time are necessary to afford greater protection of life and property. Incremental increases in flash flood warning lead time can be made through improvements in our skill to forecast excessive rainfall capable of producing flash floods. Methods of applying probabilities to define the uncertainty associated with QPF and its use in a flash flood decision system need to be further re-

PAGE 122

Summer 99 searched, leading to more objective techniques in determining the thresholds for issuing flash flood watches. Flash flood warning decision systems can take into account the individual characteristics of a river basin, the lead time for successful evacuation, and the risk factors to people in the floodplain. It could also factor in the time of year, which may influence the number of people vulnerable to dangerous flooding. One area that needs to be focused on is the delivery of products that convey in more site-specific terms the magnitude and urgency of the flash flood event. The NWS/Office of Hydrology (OH) is currently developing WFO Hydrometeorology applications that will provide forecasters with tools to run site-specific models for small streams and headwater areas. The Office of Hydrology is also developing an "area-wide" hydrologic software model to better define the flash flood potential and delineate areas where flash flood guidance has been exceeded and flooding is possible. This model depends on the use of the modernized flash flood guidance that is being implemented nationwide. The modernized guidance is being developed for small water sheds and will provide data on a gridded basis (4 x 4 km.). This will enable forecasters to compare high-resolution WSR-88D data with flash flood guidance on small scales, comparable to those that produce local flash floods. To maximize the effectiveness in using these models operationally will require RFCs to deliver accurate and timely state variables as guidance to the WFOs, and a well-trained WFO staff in effectively running the WFO Hydromet System. Product formats for flash flood warnings have not changed appreciably over the last 20 years. However, efforts are now underway to employ graphical representations, depicting flood inundation on small scales through the use of geographic information system (GIS) technology. Increased use of GIS display capabilities by local emergency management officials, and eventually the general public, will usher in a new way of pinpointing areas where flooding is expected. While improvements in warning dissemination continue with improved communications technology, its use in an overall flash flood program varies from location to location. In the events cited, there were times when direct telephone communication was vital, others where NOAA Weather Radio or amateur radio communications were important, and occasions where only directed local communications of a serious flood threat could have prevented loss of life. Redundancy of communications is an important consideration, as is a greater awareness and education of the affected user and communications media in dealing with life-threatening flash flood situations, down to the local level. One recent example of a new dissemination path that is being

PAGE 123

100 DEFICIENCIES ASSOCIATED WITH FLASH FLOOD WARNINGS implemented by the NWS is the Emergency Managers Weather Information Network (EMWIN). EMWIN makes core NWS forecasts, warnings, and other information available to emergency managers via a satellite uplink:. Greater attention has been given to warning preparedness through the implementation of dedicated warning coordination meteorologists (WCM) at all modernized weather offices. The WCMs can play a major role in improving preparedness and dissemination efforts as part of the total warning program. However, more coordinated efforts are needed among WCMs, intraand interagency efforts directed at mitigating flooding, local emergency officials, the news media, and the general pUblic. Summary Future improvements in the flash flood program will depend on an integrated approach in dealing with all the elements involved in the warning process. Efforts involved in addressing deficiencies must emphasize the total flash flood warning program and involve contributions from all levels of government, universities, and the private sector. Today's team approach lends itself to a horizontal integration of programs among federal, state, and local agencies, universities, the media, and the private sector. To fully realize maximum benefits in the flash flood warning program, the efforts of all these diverse entities in tackling the flash flood warning problem must be better integrated and focused on the total warning process. References National Weather Service, Eastern Region 1983 Southern New England Flash Floods. June 5-7. 1982. Natural Disaster Survey Report. Washington, D.C.: National Oceanic and Atmospheric Administration. 1987 Little Pine Creek Flash Flood of May 30. 1986. Natural Disaster Survey Report. Washington, D.C.: National Oceanic and Atmos pheric Administration.

PAGE 124

Summer 101 Federal Emergency Management Agency 1985 In Response to the November 7, 1985 Disaster Declaration. Interagency Flood Hazard Mitigation Report, FEMA-753-DR-WV. Region III Interagency Flood Hazard Mitigation Team. Goldsmith, B., B. Watson, and M. Hall 1995 Flooding of Late June 1995 Over the Shenandoah Valley, Potomac Highlands, and Virginia Piedmont. Operational Assessment. Sterling, Virginia: National Weather Service Forecast Office. Krzysztofowicz, R. 1993 Probabilistic Hydrometeorological Forecasting System: A Concep tual Design. Post-Print Volume, Third National Heavy Precipita tion Workshop. Pittsburgh, Pennsylvania: National Weather Service. National Oceanic and Atmospheric Administration 1976 Big Thompson Canyon Flash Flood of July 31-August 1, 1976. Natural Disaster Survey Report 76-1. Washington, D.C.: U.S. Department of Commerce. National Weather Service 1991 Shadyside, Ohio Flash Floods, JUlie 14, 1990. Natural Disaster Survey. Washington, D.C.: National Oceanic and Atmospheric Administration. Shedd, R.C. ,md R.A. Fulton 1993 "WSR-88D Precipitation Processing and its Use in National Weather Service Hydrologic Forecasting." Proceedings of the International Symposium 011 Engineering Hydrology. San Fran cisco, California: American Society of Civil Engineers. Shelton, D.R. and E.L. May 1996 "Modernized Hydrologic Forecast Operations at National Weather Service Forecast Offices." Twelfth International Conference on Interactive Information and Processing Systems for Meteorology, Oceanography, and Hydrology. Atlanta, Georgia, January 29 to February 2, 1996. Washington, D.C.: American Meteorological Society.

PAGE 125

102 DEFICIENCIES ASSOCIATED WITH FLASH FLOOD WARNINGS Susquehanna River Basin Commission 1996 "January 1996 Flash Flood." Susquehanna Guardian 5: 1. Zevin, S.F. 1993 "Steps Toward an Integrated Approach to Hydrometeorological Forecasting Services." In Third National Heavy Precipitation Workshop. 1-17. Pittsburgh, Pennsylvania: National Weather Service.

PAGE 126

CLIMATOLOGY OF EXTREME RAIN EVENTS IN THE UNITED STATES FROM HOURLY PRECIPITATION OBSERVATIONS Harold E. Brooks and David J. Stensrud NOAA/Environmental Research Laboratories National Severe Storms Laboratory Daniel V. Mitchell NOAA/Environmental Research Laboratories National Severe Storms Laboratory Cooperative Institute for Mesoscale Meteorological Studies Introduction Flash flooding is frcquently associated with heavy precipitation in a short period of time. Much work has been done in defining the climatology of precipitation on a time scale of 24 hours (e.g., Smith et al., 1994), but this is longer than the time scale associated with flash flood-producing rains. In individual flooding events, "bucket surveys" are often done in any container that holds water to estimate precipitation. Unfortunately, quality control of bucket surveys is problematic, since the question of whether a container was empty at the beginning of a heavy precipitation event can rarely be answered with confidence. We describe an effort to define the climatology of heavy rains on time scales of three hours or less, using the Hourly Precipitation Dataset (HPD), archived at the National Center for Climatic Data (NCDC). The HPD provides hourly observations of precipitation from around the United States for more than 40 years (1948-1993). Data for approximately 5,000 sites are found in the archive, although few of the site records cover the entire period. The number of reporting stations grew from nearly 300 in the late 1940s to approximately 2,800 in the early 1980s. The latter number represents a station density approximately equivalent to a uniform network with stations

PAGE 127

104 CLIMATOLOGY OF EXTREME RAIN EVENTS IN THE U.S. spaced 50 km apart. Some stations report in hundredths of inches, while others report in tenths of inches, and, although coverage is not uniform in space or time, these data are by far the most complete and accurate set of measurements of precipitation. From a meteorological perspective, flash floods may be the most difficult forecast hazard associated with thunderstorms. Identifying potential flash flood situations is frequently difficult (Doswell et aI., 1996). An understand ing of the climatology of heavy precipitation is essential for preparing flash flood forecasts, particularly if probabilistic estimates are made. Since heavy precipitation is a rare event at any single location, the experience level of weather forecasters dealing with the problem is generally quite limited. Yet, accurate forecasts of the threat are crucial for the protection of life and property. Nature of the HPD The HPD consists of a series of records, with one record per station on a given day. Each record contains the station identifying number, the date, and a series of six digit values that indicate the hour in local standard time and the precipitation in hundredths of an inch. Each record ends with the total daily precipitation. As such, the HPD is well-suited for computing time series of precipitation at individual stations and for measuring precipitation values, particularly for hourly and total daily observations. Time series data can reveal important information in certain cases. For example, part of the warning problem with the July 19-20, 1977, Johnstown, Pennsylvania, flash flood is indicated in the time series data. Most of the more than eight inches of rain fell after the late evening news, which was probably the last opportunity to warn the public. On the other hand, many events are not sampled by the HPD. With the flood that occurred on June 9, 1972, in Rapid City, South Dakota, two HPD sites were within 30 km of an estimated 12" 13-hour rainfall; yet, neither site recorded more than two inches of precipitation. In an even more extreme case, no HPD sites recorded significant precipitation associated with the Big Thompson River, Colorado, flood of July 31, 1976. Gauges at Drake and Estes Park were not put into place until after the flood. As with any large observational data set, quality control is a significant concern. Some errors are easily detected and removed from the dataset, while the accuracy of a number of other questionable observations is difficult to determine. "Simple" errors include records in which the value in the hundredths of an inch column is reproduced in the tens of inches column,

PAGE 128

Brooks, Stensrud, and Mitchell 105 reporting hourly accumulations of 10.01 inches and 20.02 inches. These values are clearly unreasonably large and, due to a repetitive pattern, can be identified and eliminated. Somewhat less obvious are extremely large reports (e.g., 15.55 inches in an hour), but the extreme value still makes it possible to eliminate them automatically. Another set of errors involves the recording of hundredths of an inch values in the tenths of an inch column (e.g., 7.80 inches in one hour, followed by 6.50 inches with no precipitation on either side of those two hours). These are problematic, as are other isolated instances of large precipitation inches/hour), because there is no objective way to determine whether they are the result of "bad" data or extremely large "good" data. Indeed, this is a fundamental problem in using any dataset about any rare, extreme events. We individually checked every hourly report of greater than 4.5 inches (approximately 360 reports) with reports from meteorological and climato logical journals, and found that only a few were likely to be good reports; the rest fell into the kinds of errors mentioned above. The real difficulty comes in attempting to hand check the much larger number of reports at smaller values. The distinction between obviously bad and good data becomes blurred, and the volume of work becomes prohibitive. Observed Frequency of Hourly Precipitation One of our purposes in investigating the frequency of heavy precipitation is to estimate the number of times operational weather forecasters will have to deal with this problem and to understand its implications for a national forecast center. Hence, we confine ourselves to looking at the average number of events in the contiguous United States and do not consider the lack of spatial uniformity. We touch only on the highlights of the precipitation record, focusing on hourly accumulations of an inch or more and, particu larly, on the average number of events in July. Longer accumulation times and larger amounts are referenced from the July hourly base values. We organized data in liz-inch aggregates (e.g., I inch to 11lz inches, 11lz inch to 2 inches, etc.) for ease of analysis and to increase the number of samples at higher values. We found that the annual cycle of heavy precipita tion peaks in July and is symmetric about that month (see Figure 1). In January, the month in which heavy precipitation is least frequent, there are about 7% of the number of events as in July. In total, on average, there are approximately 2,400 reports of I inch to 11/z inches hourly rainfalls in the HPD each year and 3,200 reports of 1 inch or greater. Twenty percent of all observations occur in July and more than 50% occur in June through August.

PAGE 129

106 CLIMATOLOGY OF EXTREME RAIN EVENTS IN THE U.S. 500 400 300 200 100 0 I I I I I I I I I Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Figure 1: Annual cycle of average number of one inch in one hour observations in the HPD for the entire United States. The number of events decreases logarithmically as the precipitation value increases (Figure 2). The fit to the curve for the July observations is extremely good for accumulations of 1 inch to 4 inches. Comparison of the number of reported events, given by the black squares in Figure 2, to the logarithmic line gives us some confidence regarding the number of extreme events in the HPD that are likely to be "bad." Later, this will provide a powerful tool for estimating the "true" number of events that occur in the United States in a year. While each month follows a similar logarithmic decrease, the rate of decrease shows hints of a seasonal cycle. In the summer, the number of events observed in a given liz-inch increment decreases to approximately 7.5 % of the value one inch lower. In the winter, it decreases to approximately 6.5%. While this is a relatively small change, it could be an indication of the greater frequency of strong convection, and hence, high rain rates, in the summer. Although not shown, results for accumulation times of two and three hours display few surprises when compared to the one-hour observations. Eight thousand (11 ,400) reports of 1 inch (greater than 1 inch) rains in two hours occur on average in the HPD and 17,400 (25,000) reports of 1 inch (greater than 1 inch) rains in three hours occur on average in the HPD. The logarithmic decay with increasing amount is slower, although the seasonality is more pronounced. In summer, for the two-hour reporting time, observa

PAGE 130

Brooks, Stensrud, and Mitchell 107 1000 Q) Cl m 100 ... Q) ... > <1: 10 ......... m ::J ........ C c <1: 0.1 ""-en 0.01 c ""-Q) > W 0.001 "' 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 Observed Hourly Rainfall (Inches) Figure 2: Average number of heavy rainfall events in July in United States reported in HPD. Events are aggregated in half-inch intervals, with all events greater than six inches in the last category. Black squares represent reports. Line is least squares fit to data from one to four inches. tions fall to 13 % for each one-inch increase, while the value is 10% in the winter. For three hours, the rate decreases to 16% in summer and 12% in winter. This is possibly due to the impact of larger-scale weather systems that produce more sustained periods of heavy precipitation, leading to observa tions of more than one inch in two and three hours. The extreme precipitation values (e.g., 3 inches or more) still result predominantly from convective environments, which are most pronounced in the summer. Estimating Frequency of Extreme Events Although the HPD provides the most complete set of high temporal and spatial resolution observations of precipitation, it is clearly inadequate for capturing extreme events. Since extreme events create the greatest risk of a major disaster, it is important to have some basis for climatological estimates of risk. The logarithmic decline of the number of events with increasing precipitation allows us to make estimates of the number of more extreme

PAGE 131

108 CLIMATOLOGY OF EXTREME RAIN EVENTS IN THE U.S. events. This is particularly important in understanding the likelihood of heavy precipitation events at even higher spatial resolution. We assume that the observations in the HPD are representative of the true climatological frequency of extreme precipitation over the contigous United States. As an example of the power of the dataset, let us make a "back-of-the envelope" estimate of the "true" annual frequency of 6 inches/hour events in the United States. From the decline to 8 % per inch, we estimate that there are approximately 0.0005 times as many 6 inches/hour events as 3 inches/ hour events. Since about 20 3" /hour events are observed each year, that implies that approximately 0.01 6 inches/hour events will be observed per year by the stations in the HPD. The next important question involves the representativeness of the observations. Precipitation, particularly that identified with convection, is associated with extremely large spatial gradients. Smith et al. (1994) provide examples of the poor spatial correlations between observational sites for convective precipitation. As a starting point, we assumed that each rain gauge represents an area of one square kilometer. Since the contiguous United States has an area of approximately 7.5 x 106 km2 this means that the 3,000 gauges in the HPD cover only 4 x 10-4 of the total area in the United States. This implies that, if we had an observational data set with 1 km horizontal spacing, we would observe approximately 2,500 times as many events of any kind as we currently observe with the HPD. Thus, there should be approxi mately 25 6 inches/hour events per year in the United States. While no such extremely high spatial resolution rain gauge network exists, the deployment of the WSR-88D radars in a national network provides an opportunity to develop a radar-estimated precipitation climatology. This promises estimates of the spatial and temporal correlations of precipitation on a national scale. Assuming that improved precipitation estimation techiques will be developed (e.g., Zrnic, 1996), it may also allow the best estimates to date of extreme precipitation. Implications As mentioned in the introduction, one of the goals of this study was to estimate the frequency at which operational forecasters have to deal with heavy precipitation during their forecast shifts. To do this, we considered two hypothetical forecasters. The first, Forecaster A, works his or her entire career at a local National Weather Service Forecast Office, while the second, Forecaster B, works one year at a national center, such as the Storm Prediction Center.

PAGE 132

Brooks, Stensrud, and Mitchell 109 There are approximately 100 forecast offices in the U.S., and an individual forecaster works 1/5 of the shifts in a typical office. Assuming Forecaster A has a 20-year career, he or she will work about 4 % of the annual national average of events during the course of his or her career. Thus, for the estimated annual average of 256 inches/hour events, Forecaster A will encounter one such event during his or her entire career. Forecaster B, on the other hand, at a national center, works 1/5 of the shifts and, thus, is on duty for 1/5 of the events every year. As a result, Forecaster B will be on duty for five 6 inches/hour events each year or 100 times as many events as Forecaster A. Thus, the relative levels of experience for both Forecasters A and B have critical implications for their roles in successfully anticipating, forecasting, and warning about flash floods. In the absence of quality guidance, a local forecaster must get the biggest precipitation event in his or her career "right" the first time, and many forecasters never get a second chance. However, even though extreme precipitation events are rare at any single location, they are relatively common from the national perspective. Thus, the national center forecasters develop significant experience in dealing with rare, extreme events in a relatively short time. As a result, the national centers have primary guidance responsibility, identifying general regions of greatest threat, while the local office has primary forecast and warning responsibility, narrowing the area of threat in short-term forecasts and identifying the exact location of extreme precipitation in warnings. Discussion The nature of flash flooding requires an understanding of the climatology of extreme precipitation. The HPD represents the most complete description of short time-scale precipitation measurements covering a significant time period over the entire United States. However, as we have seen, it still misses most of the truly large precipitation events that actually occur. The data allow us to make reasonable estimates of the real frequency of heavy precipitation. These estimates should be of value, both for emergency managers and for weather forecasting concerns to allocate resources and plan for the inevitable flash flood events. National network radar estimates of precipitation can be used to refine the climatology presented here, although it will be years before a significantly long period of radar observations exists to allow for reasonable estimates.

PAGE 133

110 CLIMATOLOGY OF EXTREME RAIN EVENTS IN THE U.S; Acknowledgments We would like to thank Chuck Doswell for his helpful comments during this work. The HPD is available from the National Climatic Data Center. References Doswell. C.A. III. H.E. Brooks, and R.A. Maddox 1996 "Flash Flood Forecasting: An Ingredients-Based Approach." Weather Forecasting 12. In press. Smith, I.A., and A.A. Bradley 1994 "The Space-Time Structure of Extreme Storm Rainfall in the Southern Plains. Journal of Applied Meteorology, 33: 1402-1417. Zrnic, D. S. 1996 "Weather Radar Polarimetry: Trends Towards Operational Applications." Bulletin of the American Meteorological Society 33: 1529-1534.

PAGE 134

THE FLASH FLOOD FORECASTERS COURSE AT THE NATIONAL WEATHER SERVICE TRAINING CENTER THE ENVIRONMENTAL RESEARCH LABORATORIES COMPONENT Harold E. Brooks Charles A. Doswell III and Robert A. Maddox NOAA/Environmental Research Laboratories/ National Severe Storms Laboratory Dennis A. Rodgers and Barry E. Schwartz NOAA/Environmental Research Laboratories/ Forecast Systems Laboratory Introduction The Big Thompson River flash flood disaster of July 31, 1976, was notable from a weather forecasting system perspective because the event occurred with no warning. Near the time of the Big Thompson flood, similar failures of the forecasting and warning system occurred in the Rapid City, South Dakota, flash flood of June 9, 1972, and the Johnstown, Pennsylvania, flash flood of July 20, 1977. Faced with more than 400 deaths in these three events, with no warnings issued, the National Weather Service (NWS) developed a program to improve forecasting and warnings for fla&h floods. A major component was the Flash Flood Forecasters' Course (FFFC) at the NWS Training Center (NWSTC) in Kansas City, Missouri. From the beginning, the FFFC was unique in that at least half of the eight-day course was taught by invited personnel from outside the NWSTC. Because of staff interests in the flash flood problem (e.g., Hoxit et al., 1978; Maddox et al., 1978, 1979), particularly in the mountainous terrain of the

PAGE 135

112 FLASH FLOOD FORECASTERS COURSE West, research meteorologists from the Environmental Research Laboratories (ERL) were invited to help teach the class. Since the first class in the fall of 1978, the course has met 78 times with approximately 24 students in each session. It was held for the last time in the fall of 1996. During the nearly 20 years of the FFFC, about 1,900 members of the NWS staff participated as students. In the fall of 1997, the FFFC will be replaced by a more general hydrometeorological forecasting course. At this time, it seems appropriate to look back at the ERL portion of the FFFC, which was unique in that research meteorologists concerned with forecasting hazardous weather were involved in training operational meteorologists. The Early Years of the FFFC ERL scientists from the Atmospheric Physics and Chemistry Laboratory (APCL), C. Chappell, R. Hoxit, and R. Maddox, were asked to prepare sessions on mesoscale analysis associated with heavy precipitation. The first class was taught by Chappell and focused on detailed case studies of the Big Thompson and Rapid City floods. Hoxit and Maddox taught the next classes. All three came to the same conclusion: NWS forecasters of the late 1970s knew too little about the basics of convection in the atmosphere to utilize the kind of detailed information that the instructors were trying to convey in the case studies. As a result, a major revision of the course materials was undertaken, with an emphasis on the basics of convection. Maddox spear headed this effort and produced a set of slides that proviueu the bulwark of the course materials for a decade. The primary foci of the revised course were on questions of how a forecast should be made, why heavy precipitation is so hard to forecast, and how convection "works." The need for detailed diagnosis of the four dimensional state of the atmosphere and the interaction of human forecasters with numerical weather prediction models were stressed. In particular, the course emphasized that forecasters needed to anticipate the possibility of rare, severe events occurring at any time if they were to respond with adequate forecasts and warnings. Brief case studies were used to reinforce the notions of the basics of convection and the forecast process. Student participation was crucial to the success of the course and, therefore, each class responded differently to the material that was presented. There were three primary avenues by which students interacted with the instructors and each other. The first was a pair of pretests, given before each major section (convection and large-scale analysis), which served three primary purposes:

PAGE 136

Brooks, Doswell, Maddox, Rodgers, and Schwartz 113 1) to give the student an idea of what he or she knew, 2) to indicate to the instructors areas that needed emphasis (in fact, it was initial pretests that convinced instructors to change the primary focus of the course), and 3) to measure growth in the expertise of the NWS staff. The results of pretest performance from many years of classes were summarized by Doswell and Maddox (1993). Regrettably, they found little change in the answers given through the years, indicating that the base know lege level of NWS forecasters had not improved during that time. It also raised questions about the concept of distance learning and the way that knowledgable forecasters transfer information to others at their office. The second avenue for interaction was discussions in class. These came about because of two primary kinds of activities. "Homework" analysis exercises were given out the day before each day of the ERL course. Second, many short exercises were included in the lesson plan to illustrate specific points of the lectures. Finally, at the end of each course, the classes were divided into teams of three to five forecasters. The teams were given historic data from specific cases and charged with forecasting where and how much heavy rain was expected, if any. An important component of the exercise was to demonstrate the importance of interacting with colleagues when making a forecast, particularly of a rare, severe event. During the 1980s, the material in the ERL portion of the FFFC remained static. A first plan to update the material in the mid-1980s was discouraged, because the FFFC was to be replaced by a hydrometeorological forecasting course "soon." (As mentioned in the introduction, "soon" turned out to be more than a decade!) Although several people taught a small number of classes, the majority were taught by Maddox and C. Doswell, first of the Weather Research Laboratory and later the National Severe Storms Laboratory (NSSL). Although the basic materials stayed the same, each instructor emphasized points of particular interest to them; thus, the nature of each meeting depended upon the individual instructor as well as the students. It was generally a positive aspect of the in-class learning process. The Second Great Redesign (1992-1994) By the early 1990s, some materials in the FFFC were becoming outdated and the "replacement" course was not imminent. As a result, ERL instructors C. Doswell and H. Brooks decided to undertake a second major revision of

PAGE 137

114 FLASH FLOOD FORECASTERS COURSE the course materials. The pretests, team exercises, and many of the examples were updated, with additional emphasis on an ingredientsand process-based approach to forecasting (Doswell et al., 1996). It was designed to help forecasters focus on important details as the available data increases, stressing the need for understanding the physical phenomenon in order to make a successful forecast. The number of case studies presented in the course increased dramatically, although with less detail in each case. This was done to illustrate the wide range of conditions under which heavy precipitation occurs. Because there are numerous misconceptions about convective systems that produce heavy precipitation, several of the case studies addressed those misunderstandings. Second, the case studies illustrate the common factors found in all the events, reinforcing the basic ingredients (moisture, instability, and lift) and processes approach to forecasting heavy precipitation, and the variety of cases demonstrated various ways that atmospheric processes bring the ingredients together. Finally, the geographic diversity of the cases illustrated that heavy precipitation can occur anywhere, with ingredients common to all regions, and that it is a threat all forecasters must consider. Between the pretests, case studies, and other exercises, cases from one U.S. territory, 11 U.S. states, and three Canadian provinces were developed, providing scenarios that ranged from Puerto Rico to Quebec, Alaska to Nevada. Cases were added as new information became available. The Third Great Redesign (1994-present) As technology advanced to the point that remote access to trammg materials became possible, we began putting the ERL FFFC on the World Wide Web (WWW), primarily because flash flood forecasting problems are not limited to the NWS-other agencies have substantial interest in the problem. Further, it is important that meteorology students recognize the difficulties associated with forecasting rare, severe events. Third, a larger audience can be reached via the World Wide Web than through occasional sessions in Kansas City. In fact, without advertising the still incomplete course (located at http://www nssl. uoknor. edu/projects/fffc/outline. html), approximately half the number of people who have ever attended the classes in Kansas City have at least looked at the outline of the course in the last year. Clearly, it is possible to reach vastly more people with the World Wide Web than through the classroom and to reach an audience that would not even be eligible to come to the NWSTC. And, notably, many of the "hits" on the outline have come from foreign countries.

PAGE 138

Brooks, Doswell, Maddox, Rodgers, and Schwartz 115 Second, the Internet version can be used as a refresher for students who attended the course or have worked through the outline. This is important for maintaining skills and, in some cases, reviewing technical details that might not have been grasped in the first encounter. We hope to someday develop a large library of cases that would be available on the World Wide Web. In the limited time at the NWSTC, it is impossible to work through many cases and, generally, some of the case studies are skipped or skimmed over in the interests of time. It is possible, with appropriate human and computer resources, to put a large number of cases on to the Web with great detail, enabling forecasters to work through the cases that will help them the most. In this medium, cases can be added quickly so that something new is available all the time, helping forecasters and students to find new challenges in working through the problems. Once a forecaster has worked through an example, he or she will know the answer, limiting the value of the individual resource. Certainly, there are drawbacks to the World Wide Web as a training tool. Some materials, such as homework analysis exercises, don't transfer well to the electronic setting. Interaction between students and instructors is not as rapid and extensive as it is in a classroom, and discussion among students is next to impossible. One cannot tailor the course to anyone student's specific needs, and considerable attention must be given to the maintenance and upgrades of both the course materials and the hardware on which it resides. Thus, the role of researchers as trainers becomes strained. In general, the people who have taught the course have been full-time research meteorolo gists and, as a result, development and maintenance of training course materials takes away from their research efforts. Since part of the intent was to have current researchers training forecasters, a paradox develops. If the Web instructors are to do an adequate job of training, they will need assistance in putting materials together and on line, or they will not have the time to do the research that is the basis for having them participate in training! Feedback to Research Researchers often have had their research agendas set or changed by issues that come up in the FFFC. As an example, precipitation efficiency (the amount of rain that falls out of a storm divided by the amount of water vapor entering the storm) is an important topic for estimating precipitation potential of an environment. Recently, the broader question of moisture budgets of thunderstorms was investigated, due to discussions in the FFFC, using a

PAGE 139

116 FLASH FLOOD FORECASTERS COURSE numerical model. The results of that investigation indicate that concentrating on the reported observations of precipitation is misleading if the intent is to forecast the amount of rain that will fall out of the storm. In fact, the sign of the dependence on vertical wind shear is reversed when the more complete budget is analyzed (Brooks and Stensrud, 1996). Also, a more comprehensive and ambitious project to study heavy precipitation and flash flooding has been initiated at NSSL recently. Looking Back at the FFFC Nearly 20 years have passed since the first meeting of the FFFC. Nearly 2,000 students taken the course. It is tempting to say that the instructors have made a difference in the way that heavy precipitation is forecast in the U. S. ; however, we have no information, other than rare anecdotes, that can confirm or deny that statement. Essentially no monitoring of activities before or after the class has ever been done. As a result, we have no way of knowing if any of the participants changed their behavior after coming to the FFFC. It is conceivable, although we hope not likely, that the participation of ERL staff in the FFFC actually had a detrimental effect on forecaster performance. However, it is heartening that no events with a death toll of the magnitude of the big three events from the 1970s (Rapid City, Big Thompson, Johnstown) have occurred since the FFFC began. Some events, such as the 1995 Rapidan River flood in Virginia (Fritsch et aI., 1996) have had forecasts and warnings with significant lead time. We would like to think that efforts such as the FFFC have made a difference in those cases. Yet, we recognize that we merely have been lucky and that another disaster could occur at any time without adequate forecasts and warnings. The ERL portion of the course will continue on the World Wide Web, even after the FFFC ends. We hope that more forecasters, particularly early in their careers, can be influenced through its existence and that somewhere down the road a disaster can be averted. Acknowledgments We would like to thank all of the scientists who taught in the course. Along with the authors, this includes C. F. Chappell, L. R. Hoxit, and D. W. Burgess. We also thank the students and staff at the NWSTC for their participation and help in the sessions.

PAGE 140

Brooks, Doswell, Maddox, Rodgers, and Schwartz 117 References Brooks, H.E., and D.J. Stensrud 1996 "The Moisture Budgets of Numerially Modelled Thunderstorms. In Proceedings of the Sixth Atmospheric Radiation Measurement (ARM) Science Team Meeting, San Antonio, Texas, 4-7 March 1996. U.S. Department of Energy. In press. Doswell, C.A. III, H.E. Brooks, and R.A. Maddox 1996 "Flash Flood Forecasting: An Ingredients-Based Approach." Weather Forecasting 12 .. Doswell, C.A. III and R. A. Maddox 1993 "A Review of Student Performance on Pretests Given at the Flash Flood Forecasting Course. Preprints, Symposium on Flash Floods, Vienna, Virginia, 2-6 August 1993. 401-410. American Meteor ological Society. Hoxit, L.R., R.A. Maddox, C.F. Chappell, F.L. Zuckerberg, H.M. Mogi!' I. Jones, D.R. Greene, R.E. Saffle, and R.A. Scofield 1978 Meteorological Analysis of the Johnstown, Pennsylvania, Flash Flood, 19-20 July 1977. NOAA Technical Report ERL 401-APCL 43 [NTIS Accession no. PB-297412]. Washington, D.C.: National Oceanic and Atmospheric Administration. Maddox, R.A., L.R. Hoxit, C.F. Chappell, and F. Caracena 1978 "Comparison of Meteorological Aspects of the Big Thompson and Rapid City Flash Floods." Monthly Weather Review 106: 375-389. Maddox, R.A., C.F. Chappell, and L.R. Hoxit 1979 "Synoptic and Meso-Alpha Aspects of Flash Flood Events. Bulletin of the American Meteorological Society 60: 115-123.

PAGE 141

This page IS intentionally blank

PAGE 142


PAGE 143

This page IS intentionally blank

PAGE 144

Flood History CALIENTE CREEK ALERT FLOOD WARNING SYSTEM AUDIT Clark Farr Kern County Engineering and Survey Services Bakersfield, California David C. Curtis DC Consulting Folsom, California Introduction The Caliente Creek watershed lies at the southern end of the Sierra Nevada Mountains in Central Kern County, California. The total area of the watershed is approximately 435 square miles, and mountains within the watershed range in elevation from about 4,500 feet to 8,000 feet above sea level. Six major floods were reported and documented during the past 45 years that had estimated peak flow rates ranging from 600 cubic feet per second (cfs) to about 23,000 cfs. Prior to the moderate flooding in January and March of 1995, the last major flooding occurred on March 1, 1983, when a peak flow of 12,800 cfs was estimated at the point where the creek exits the Tehachapi Mountains and spreads out onto the San Joaquin Valley floor near State Highway 58. ALERT System The current Caliente Creek ALERT flood warning system was constructed and went on-line in September 1984, following the 1983 flood. This system, designed by the California-Nevada River Forecast Center of the National Weather Service (NWS), includes six precipitation gauges and one stream flow gauge. Precipitation gauge density is approximately one gauge per 72.S square miles.

PAGE 145

122 CALIENTE CREEK ALERT FLOOD WARNING SYSTEM AUDIT Expected Service The ALERT facility construction was funded through a cooperative agreement between eight local agencies. Each of these entities had different expectations and uses in mind for the information supplied by the ALERT system. The Arvin-Edison Water Storage District maintains the primary irrigation canal servicing the southern end of the San Joaquin Valley. This canal traverses the contours of the Caliente Creek alluvial fan. The canal is at direct risk of severe damage and possible catastrophic breach due to large floods on Caliente Creek. Based upon the flood warning, the canal is drawn down to minimize the effects of severe flooding. The Kern Delta Water District operates and maintains a secondary irrigation canal system located on the Caliente Creek alluvial fan. Although this canal is smaller than the Arvin-Edison canal, its location immediately east of the town of Lamont creates a greater risk to life and property if a catastrophic breach occurs. As with Arvin-Edison, Kern Delta lowers canal levels based upon degree of expected flooding. The Tehachapi-Cummings Water District, located in the southern third of the watershed, uses the data collected by the ALERT system to help determine its water supply estimates. The Southern Pacific Railroad operates the only rail line connecting northern and southern California. This major rail line crosses the Tehachapi Mountains via a path cut by Caliente Creek and follows lhe <.:reek for several miles. Southern Pacific requires accurate flood alarm information to guide rail traffic through the mountains. Overestimations mean that millions of dollars of rail traffic are needlessly stalled. Underestimation results in millions of dollars of additional losses if trains are damaged during a large flood event. The Lamont Storm Water District was established by the state legislature to provide a flood control funding mechanism for the unincorporated town of Lamont. The district works with the county in preparing for a flood event. The City of Arvin, located on the southeasterly edge of the Caliente Creek alluvial fan, uses the information supplied by the flood alarm system to initiate their flood-fighting response. Kern County has primary flood-fighting responsibility within the unincorporated areas affected by Caliente Creek. The Emergency Services Office expects the ALERT system to provide three to five hours of sand bag distribution time. The Kern County Water Agency was created by the state legislature to act as a wholesaler of state water project water and to develop regional water

PAGE 146

Farr and Curtis 123 .supplies. The Caliente Creek ALERT system data assists them in determining water supply needs and reservoir feasibility. Reality Check When the ALERT system was first brought on-line in September 1984, it was anticipated that it would pay for itself the first year. The devastating floods of 1983 were fresh and clear in the minds of residents and politicians. Not only was the ALERT system not put to the test in 1984, but very little rain fell for the next decade. During that time, watersheds dried up along with the memories of Caliente Creek floods. Consequently, system priority dropped as flood memories faded. In addition, the deep economic depression that gripped the region fostered a succession of increasingly restricted county budgets. When Kern County first entered into the standard cooperative agreement with the National Weather Service (NWS) for maintenance of the ALERT system, the county Board of Supervisors directed that the maintenance cost of the facilities be shared by the eight participating agencies, with the county performing the actual maintenance. It is not clear why, but over time, for one reason or another, each participant was forgiven their maintenance cost share, leaving the full cost of maintenance up to the county. Unfortunately, during this same period, maintenance costs began to rise. Budget pressure, shrinking demand (no rain, no floods), and splitting system responsibilities all conspired to reduce the maintenance effort. Maintenance became strictly reactive. No calibration. No cleaning. If a unit failed, it was repaired when a technician was available and was not given the highest priority. 1995 Floods Finally, in January 1995, Caliente Creek produced its first significant flood since 1983. However, imminent flooding was not recognized until flood flows were sighted crossing the lower reaches of the floodplain, reducing the estimated 12 hours (minimum) response time for the town of Lamont to less than four hours. No warning time at all was provided to Southern Pacific Railroad or Arvin-Edison. A review of the ALERT system data disclosed that only one ALERT system precipitation gauge indicated significant rainfall, and even that rainfall was below the alarm threshold. Another gauge, which had transmitted only minimal rainfall data, was later found to have had a brim full catch funnel due to a plugged drain.

PAGE 147

124 CALIENTE CREEK ALERT FLOOD WARNING SYSTEM AUDIT The failure of the ALERT flood warning system in January 1995 to provide an adequate warning could have resulted in significant avoidable flood losses in the Lamont and Arvin areas if the flood event had been more severe. After the January problems, county staff were not willing to depend upon the ALERT system as a sole source for flood declaration and mobiliza tion. Instead, staff kept closer contact with the NWS, followed weather satellite imagery supplied by a private weather data provider, and continually monitored the two ALERT gauges that appeared to have functioned in the January event. The county's lack of confidence in the flood warning system resulted in a significant increase of staff time, both during regular working hours and in overtime, for the remainder of the 1995 flood season. Independent Audit There were lingering questions about the adequacy of the facilities in the system (number and location), the maintenance of those facilities, and the warning process that followed. These issues were tangible. It was the unknown, the intangible, that concerned Kern County Floodplain Manage ment. The question of whether or not Floodplain Management knew enough about the ALERT system concept to even know what it didn't know, was the issue. Therefore, the department's expectations of an independent audit were simple. Were the quantities and locations of the facilities adequate? What should a minimum maintenance program be? How should an ALERT system function? Did Kern County have the necessary skills to do the job? The expectations of the process were simply stated. Floodplain Manage ment needed a document that spelled out where the problems were and suggested solutions. This document would then be used as a tool to first revive the priorities of each department regarding the ALERT system and then to forge a new cooperative agreement with the original funding agencies. Caliente Creek ALERT System Audit The primary objective of the audit was to determine if the ALERT System was capable of reliably warning of floods. The project included: 1) a review of the ALERT system and its requirements, 2) a review of ALERT system operations, 3) a review of ALERT system maintenance procedures, and 4) preparation of a final report with recommendations. The existing ALERT system, installed in 1985, consisted of five automated rain gauges, an automated weather station, a stream gauge, and a computerized central station. Each automated station was equipped with a

PAGE 148

Farr and Curtis 125 battery-operated radio transmitter to send reports to the central station. A radio repeater relayed signals from the monitoring station to the central computer. The rain gauges were the standard tipping-bucket type used in ALERT systems; the buckets were sized to tip after 1 mm (0.04 inches) of rain had been collected through the 12-inch diameter rain collection orifice. As the bucket tipped and emptied its contents, a second bucket was positioned to receive the next 1 mm of rain. When the second bucket tipped, the first bucket returned to its original position to collect the next 1 mm increment. Each tip turned on a battery operated radio transmitter to send the data to the central computer. Rain gauges were located at Walker Basin, Back Canyon, Tollgate Lookout, Tehachapi, and Orejano Canyon. The weather station, located at Piute Peak, consisted of a tipping bucket rain gauge and meteorological sensor suite that reported wind speed, wind direction, temperature, relative humidity, and barometric pressure. A single stream gauge was located on Caliente Creek approximately one mile below the confluence of Tehachapi Creek and Caliente Creek. Stream level was measured by a pressure transducer located in the stream bed, which transmitted pressure readings to the central computer every 15 minutes. Radio signals from the monitoring stations in the Caliente Creek watershed were relayed to the central computer by a radio repeater located near Breakenridge Camp. The central computer was located at the offices of Kern County Communications Control 5, which maintained around-the-clock operations. Until 1995, the radio signals were also received at the local National Weather Service (NWS) office. However, the National Weather Service moved its local office to Hanford and no longer received these signals. Although there is no definite timetable, the NWS plans to install the necessary equipment to re-establish communications links with the Caliente Creek ALERT System, and the data will be used to support NWS forecast and warning operations. The central computer at Control 5 consisted of a personal computer using the National Weather Service's Hydromet software and the QNX Operating System. The central computer radio receiver/decoder package received signals relayed by the Breakenridge repeater, decoded the messages, checked the data quality, and stored the resulting rain, weather, and stream informa tion on the computer's hard disk. The central computer continuously monitored the data to detect alarm conditions such as high rainfall rates or high stream elevations. When excessive rainfall or high water was detected, the computer sounded an

PAGE 149

126 CALIENTE CREEK ALERT FLOOD WARNING SYSTEM AUDIT alarm, notifying personnel at Control 5. Once alarm conditions were verified, Control 5 personnel automatically initiated a Stage 1 Flood Response according to the Caliente Creek Flood Response Plan. Data from other ALERT gauges located in the area were also received by the ALERT computer at Control 5, allowing Kern County officials to track potential storms before they hit the Caliente Creek watershed. The ALERT computer also allowed outside users to dial-in by telephone to access the raw data and monitor developing conditions in the watershed. Engineering and Survey Services (E&SS) personnel, the chief hydrological technical support group for the Caliente Creek Flood Response Plan, was the principal outside user. E&SS personnel used personal computers at their own offices or homes to keep abreast of watershed conditions. Site Review Each field monitoring station, the central computer station at Control 5, and the remote access terminal at Engineering and Survey Services were visited and evaluated. The rain gauge at Tehachapi (ALERT Station 1807, elevation 4,620 ft.) is located in an open area protected on three sides by nearby low hills. The fourth side opens on a shallow slope to the valley below. This location should provide rainfall estimates that are fairly representative of the surrounding areas. Although there were considerable bird droppings along the side of the gauge, the gauge itself was in good condition. (Note: Accumulated bird droppings are a common cause of plugged gauges.) The Tehachapi gauge was battery operated and had no solar panel for recharging. The rain gauge at the Tollgate Lookout Station (ALERT Station 1805, elevation 5,460 ft.) is located at an abandoned mountaintop, forest fire lookout station. This site is exposed to wind on all sides and is also exposed to direct upslope winds on two sides. Because site elevation is 5,460 feet, the site is subject to freezing temperatures and snow during colder winter storms, and ALERT tipping bucket gauges are not intended to operate in snow and freezing conditions. However, the gauge appeared to be in excellent physical condition. The station operated on battery power only. The rain gauge at Back Canyon (ALERT Station 1811, elevation 4,400 ft.) is located at a saddle point between two mountain peaks. The site is exposed to wind as well as direct exposure to upslope winds from two sides. The elevation of this site also yields freezing temperatures and snow in the winter. The gauge appeared to be in excellent physical condition, and a solar panel was present for battery charging.

PAGE 150

Farr and Curtis 127 The rain gauge at Orejano Canyon (ALERT Station 1803, elevation 2,440 ft.) is protected by low hills on all sides. Scattered trees near the gauge site also help protect the gauge site. The physical condition of the gauging station was good, although bird droppings were evident. A solar panel was present for battery recharging. Brush and hills protect the gauge site at Walker Basin (ALERT Station 1809, elevation 4,240 ft.) from wind on one side only. The remaining three sides are highly exposed to wind from the west, south, and east. Snow and freezing conditions are also possible at this site. The physical condition of the gauging station was good. No evidence of bird droppings was present. The station did not have a solar panel for battery recharging. Brush and low hills relative to the gauge at Piute Peak (ALERT Station 1813, elevation 6,560 ft.) protect the gauge site well. In addition, the nearby brush is a sufficient distance from the gauge and does not interfere with gauge catch. The weather station included a tipping bucket rain gauge, a cup anemometer to measure wind speed, a vane to measure wind direction, a temperature sensor, a relative humidity sensor, and a barometric pressure sensor. At elevation 6,560 ft., the Piute Peak tipping bucket rain gauge experiences significant problems with snow and ice, although the physical condition of the gauge is good. A solar panel recharges the battery. Engineering and Survey Services staff reported that a new humidity sensor is needed. The Caliente Creek stream gauging station (ALERT Station 1800, elevation 1, 180 ft.) is located approximately one mile downstream from the confluence of Tehachapi Creek and Caliente Creek in a dense growth of riparian brush and small trees. The stream bed is relatively stable and the stream bank opposite the gauge is protected by gabions. Given streambed conditions in and along Caliente Creek, this site is a relatively good location for water measurements, although there is not an obvious hydraulic control section nearby that could be used for rating curve development. The pressure transducer is located in the streambed and subject to ongoing efforts to keep the transducer from being covered with sediments. which can delay sensor response times. This portion of the stream channel is in a deeply incised narrow valley. Radio communications to the Breakemidge repeater site are not reliable. The physical appearance of the gauge was good. A solar panel was used to recharge the battery. The central computer station is located at Kern County Communications Control 5. The central station includes a receiving antenna, a radio re ceiver/decoder, a PC-compatible 386-class computer, a telephone modem (1200 baud), a dot matrix printer, and a stand-by power supply. The central

PAGE 151

128 CALIENTE CREEK ALERT FLoOD WARNING SYSTEM AUDIT station computer equipment is located on a work desk in the main dispatch area and is easily accessible. The equipment was clean and in good working condition. The remote access station was located at the Kern County Engineering and Survey Services Department (E&SS). The station consists of a PC-compatible 486 class computer and a 14,400 baud modem. Dial-in access to the ALERT central station is limited to 1200 baud due to the slow speed modem at Control 5. Engineering and Survey Services personnel use the station to monitor data collected by the ALERT central station at Control 5. The E&SS staff made evaluations and decisions based on theoretical rating curves developed for the Caliente Creek gauging location using HEC-2 and a library of hydrologic modeling results from prior applications of HEC-l. Findings Rain Gauge Network Perhaps the most notable finding for the Caliente Creek ALERT system was that the rain gauge network was undersized. The Caliente Creek watershed, which covered nearly 500 square miles in mountainous terrain, was too large for six rain gauges to adequately represent rainfall for flood warning. Annual rainfall ranging from 6 to 25 inches further indicated the highly variable nature of rainfall in the basin. Localized cloudbursts have also caused severe flooding. The existing network of six gauges was even less capable of defining the volume of rainfall entering the watershed for cloudburst conditions. The network was just too sparse to identify local areas of heavy rainfall. In addition, the effective size of the network was reduced due to the location and type of gauges. Three of the sites were in locations that are highly exposed to wind, affecting the rain gauge catch. Generally, rain gauge catch is reduced by about 1 % per mile per hour of wind speed at the rain gauge orifice. Winds in the range of 50 to 75 miles per hour or more are not uncommon at unprotected locations and high elevations during severe winter storms. Under these conditions, rain gauge measurements at Tollgate Lookout, Back Canyon, and probably Walker Basin were not reliable. An economic analysis of the Caliente Creek rain gauge network indicated that a minimum network of 10 gauges would be needed to adequately measure general rainfall events. Further, a network of 19 rain gauges would be needed to cover cloudburst events. The analysis also indicated that even a network of up to 24 rain gauges would be economically justified.

PAGE 152

Far,. and Curtis 129 Several of the gauge sites were at elevations higher than 4,000 feet. Freezing conditions and snow create severe measurement problems for the standard ALERT tipping bucket gauges. Frozen buckets cannot tip and/or measure precipitation properly. Freezing rain can build up on the gauge walls and orifice screens, making it unable to measure precipitation until the gauge thaws or perhaps not at all if the precipitation sublimated before melting into the tipping bucket. Similarly, snow can build up on the orifice screen at the top of the gauge, freeze, and plug the gauge. The Tehachapi, Tollgate Lookout, and Walker Basin gauges do not have solar panels to recharge the data transmitter batteries, reducing the reliability of these gauges and increasing the likelihood that one or more of them will go out of service due to battery failure and further reduce the effective network size. All things considered, the Caliente Creek rain gauge network was probably closer to three gauges than the original six, making reliable assessments of developing watershed conditions even more difficult. It is unlikely that Kern County's flood warning expectations have been met with the existing rain gauge network. Furthermore, the mere existence of the flood warning system in this state has bred a certain level of confidence and sense of security in the community that is unwarranted given its inadequacies. Stream Gauging The single stream gauge on Caliente Creek is the bare minimum for a watershed of this size. The gauge's effectiveness as a warning tool is limited due to its location at the lower end of the watershed and the intermittent radio communications from that site. Maintenance and Gauge Calibration Site inspections conducted during this audit revealed that the physical elements of the gauge sites were well maintained. All of the gauges were in good physical condition. The only noticeable physical problems noted were nuisance bird droppings at two or three sites. However, the current maintenance program is very limited. Until recently, maintenance personnel responded only when a gauge went out of service; no routine or scheduled preventative maintenance was performed. Now routine light maintenance is carried out, including cleaning rain gauge screens, checking/replacing batteries, and leveling the tipping bucket. Critical high level maintenance functions such as gauge calibration or radio frequency alignment are not performed.

PAGE 153

130 CALIENTE CREEK ALERT FLOOD WARNING SYSTEM AUDIT Central Computer Station The flood warning system base station at Control 5 is a single computer configuration. Now that the local NWS office has moved to Hanford and no longer receives the radio signals from the flood warning system, a single failure at the Control 5 base station could cause the entire flood warning system to fail. The National Weather Service Hydromet software package used on the computer at Control 5 has limited remote access capabilities. This limits the kinds of analysis, data displays, and reports that key remote users can perform during flood emergencies. In addition, the 1200-baud modem at Control 5 limits transmission speeds to less than 5-10% of those possible with modems currently used by E&SS. Remote Access Station at E&SS The remote access station at E&SS consists of a PC, a 14.4k modem, with a standard communications software package. Communications and analysis capabilities are limited by software and hardware constraints at Control 5. Recommendations After a review of the audit, the following recommendations were offered: Expand to 10 gauges to meet minimum requirements for basin wide general storms. Expand to 19 gauges to meet minimum requirements for cloud burst events. Limit long-term expansion to 24 rain gauges. Move high exposure rain gauges to more representative sites. Add radio repeaters to improve radio communications. Add two or three addition stream gauges at upstream locations. Add another base station to increase system reliability and utility. Make E&SS responsible for system maintenance. Implement a program of routine preventative maintenance. Make flood warning system maintenance a budget line item to improve visibility for this mission-critical service. Increase spare parts inventory.

PAGE 154

Farr and Curtis 131 Summary An audit of Kern County's Caliente Creek ALERT flood warning system was performed. Each monitoring station in the system was visited and evaluated. The base station operations at Control 5 and the remote access station at Engineering & Survey Services were also reviewed. The audit determined that the Caliente Creek ALERT flood warning system was unlikely to meet the county's requirements for providing reliable flood warning for downstream communities, including Arvin and Lamont. The rain gauge network of six gauges was not sufficient to adequately estimate the volume of rain falling on the Caliente Creek watershed for reliable flood warning. Recommendations included: increasing the number of monitoring stations to meet minimum guidelines for flood warning systems, improving communi cations, and improving the system maintenance program. A detailed analysis of the appropriate rain gauge network required for the Caliente Creek watershed was performed. The analysis indicated that a minimum network of lOrain gauges was needed for proper estimation of rainfall from general basin-wide storms. A minimum network of 19 gauges was needed to properly identify rainfall in cloud-burst situations. In addition, an economic analysis of network size was also performed indicating that, based on potential damages mitigated in downstream communities, the county could economi cally justify a network as large as 24 gauges. Additional radio repeaters, spare parts, and a second computer base station were recommended to improve system communications and reliability. An improved maintenance program was recommended that included making Engineering & Survey Services responsible for system operation and maintenance with radio and electronics support from Kern County Communi cations. It was also proposed that the county make flood warning system maintenance a budget line item as a mission critical function. Capital cost estimates were provided for the minimum system upgrade to 10 rain gauges and the maximum system upgrade to 24 gauges. These recommendations can serve as a multi-year master plan to upgrade the current system. The process of correction has begun. The department expects to take several years before the system reliably meets the expectations of the cooperative group that first funded its construction.

PAGE 155

l32 CALIENTE CREEK ALERT FLOOD WARNING SYSTEM AUDIT References Curtis, David. C. 1993 An Economic Rationale for Rain Gauge Network Size for Flood Warning. Solvang, California: California Association of Flood Plain Managers Conference. 1995 "Wind Effects of Rain Gauge Catch", ALERT Transmission, Anaheim, California: ALERT Users Group. Kern County 1992 Caliente Creek Flood Response Plan. Office of Emergency Services. U.S. Army Corps of Engineers 1980 Caliente Creek Stream Group Hydrology. Sacramento, California: Department of the Army, Sacramento District, Corps of Engineers. 1995 Economic Analysis Caliente Creek, Kern County, California. Sacramento, California: Department of the Army, Sacramento District, Corps of Engineers.

PAGE 156

EVOLUTION OF LOCAL FLOOD WARNING SYSTEMS AND EARLY NOTIFICATION PROCEDURES IN DENVER, COLORADO Kevin G. Stewart Urban Drainage and Flood Control District Denver, Colorado Introduction An effective flood warning system design must include a flood threat recognition system to detect and evaluate the threat as it develops and alert local authorities concerning potential dangers. With the prediction/forecast technical support in place, internal communications should be considered the most important component of the flood warning system, but are often identified as the weakest link when systems fail. By developing plans that emphasize early notification procedures and communications, public safety officials can take appropriate preparedness actions and involve additional technical support personnel with warning decisions before flood damages or deaths occur. The total flood warning system must also disseminate flood warnings to the public. The success of the system will ultimately be judged by the public's response to the warning, a realm over which local officials have the least control. This paper describes the flood warning system for Denver, Colorado, and the surrounding region. The program's early notification procedures, custom products, and supporting technology have evolved from 17 years of providing flash flood predictions to emergency managers and public works agencies within the Urban Drainage and Flood Control District (UDFCD). The District (Figure 1) serves six county governments, including 30 cities and towns, and operates one of the largest automated flood detection networks in the United States. UDFCD also routinely conducts flash flood exercises and training programs for those involved with preparing for and responding to flood emergencies.

PAGE 157

134 EVOLUTION OF FLOOD WARNINGS SYSTEMS IN DENVER DOUGLAS \,jILES o 10 15 20 ----o 10 20 30 I(ILOU(T(RS ELBERT LEGEND Weather Station f Rain Gage I Stage Gage Rain & Stage Gage Figure 1. General location map and flood detection network, Denver, Colorado. Background During the late evening of July 31 and early morning hours of August 1, 1976, a disastrous flash flood cascaded out of the Big Thompson Canyon between Estes Park and Loveland, Colorado, claiming 145 lives and causing extensive property damage. Because this event occurred less than 50 km north of the Denver/Boulder metropolitan area, the flood increased public awareness of mountain flash flood dangers to an all-time high. For residents of Boulder, Colorado, this near-miss was especially frightening because Boulder Creek flows through the heart of the city and its canyon mouth is at the city's western edge. Boulder Creek has long been considered Colorado's most dangerous flash flood stream, with the highest potential for loss of life. Very soon after the Big Thompson flood, the hazards research community investigated the warning process used for Big Thompson, made projections concerning the impact of a similar flood on Boulder Creek, and recom mended the development of a local flood warning system with a meteorologi cal support component tailored for Boulder. The District's Flash Flood Prediction Program (F2P2) was created in 1979 in response to this research.

PAGE 158

Stewart 135 This program uses a private meteorological service to early notifications to the Boulder Regional Communications Center and other 24hour emergency contact points concerning flash flood potentials. The F2P2 focuses specifically on flood problems and functions in close cooperation with the National Weather Service (NWS) Weather Forecast Office in Denver. The success of F2P2 is attributed to the direct involvement and commit ment of many local government officials. Each year the UDFCD Board of Directors, comprising 15 locally elected officials and two appointed engineers, commits funds for continued meteorological support, acquisition of real-time weather data, and equipment maintenance. UDFCD staff coordinates with the various flood response agencies and volunteer organiza tions and routinely seeks input from those most closely involved with field operations in an effort to continually improve services. Flood Warning Plan Development A written plan is a critical component of a local flood warning program. UDFCD developed its first basin-specific flood warning plan in 1977. Since then, this drainage basin planning approach has been used to develop a total of seven plans, including the one for Boulder Creek. All UDFCD-supported plans are reviewed, updated, and exercised annually. Problem area identifica tion and decision aid development are important first steps in preparing an effective flood warning plan. Special efforts are made to insure that accurate, timely and understandable communications occur both before and during a flood emergency. With special emphasis on communications, the opportunity for a successful emergency response is greatly enhanced. Flood warning plans address the three basic elements of early detection and evaluation, warning dissemination, and response. Experience and research have shown that reliable heavy precipitation forecasts and predic tions of flood potential are necessary to prompt early preparedness actions by public safety officials. Additional technical support personnel (engineers and hydrologists) can also be called to an increased state-of-readiness based on the early meteorological information, well before a public flood warning is needed. The response component is viewed from two perspectives: 1) the pre emergency or proactive response of technical personnel and public safety officials, and 2) the public response to the warning. Although flood warning plans cannot completely address or control the public response, knowledge of human behavior is considered when planning how to issue flood warnings and to conduct emergency field operations.

PAGE 159

136 EVOLUTION OF FLOOD WARNINGS SYSTEMS IN DENVER Flood Warning Dissemination Procedures As stated previously, the F2P2 works in close cooperation with the NWS and many local government agencies. It is important to understand that the NWS and local governments are ultimately responsible for warning the public, and UDFCD's role is to provide technical support to aid in that decision. The electronic news media (local television and radio) are the primary vehicles for warning the public, but other methods such as NOAA Weather Radio, fixed-location public address/siren systems, cable television audio interrupts, emergency vehicle loudspeakers, and door-to-door notifications are also used. Written agreements and internal operating guidelines define individual responsibilities and prescribe how coordination should occur. For example, the NWS has written procedures describing how forecasters will coordinate with the F2P2 meteorologist. UDFCD input was obtained by the NWS in developing their procedures. Similarly, procedures used by UDFCD were developed from recommendations provided by the NWS and local govern ment partners. F2P2 message dissemination may be categorized in two ways; 1) early internal notification of flash flood potential and 2) internal notification of flash flood watches and warnings. The second category typically involves a NWS decision to issue a public watch or warning statement. The F2P2 meteorologist is only responsible for contacting affected local authorities and providing a more detailed interpretation. Therefore, F2P2 dissemination of watches and warnings are still considered internal to the program. With regard to flash flood watches and warnings (Le., Message 2 and Message 3), two possibilities exist. The first case involves a decision by the NWS to issue either a watch or warning. Coordination between F2P2 and NWS meteorologists normally occurs before the information is released for public broadcast. The F2P2 meteorologist responds by immediately notifying affected emergency communications centers and passing along any additional information or recommendations pertinent to local authorities (e.g., anticipated flood problems, specific geographic areas and streams affected, estimated severity and probability of occurrence, predicted precipitation amounts, available field observations, etc.). Once notified, each contact relays the message according to their internal procedures. This action initiates the mobilization of appropriate field resources and key personnel from various response agencies. Each agency has their own emergency plan to follow. The F2P2 meteorologist has the opportunity to concur or disagree with the NWS decision, but must issue the appropriate message.

PAGE 160

Stewart 137 The second case occurs when the F2P2 meteorologist decides to issue a Message 2 or 3 (flash flood watch or warning equivalent) without NWS concurrence. The same procedure described for the first case is followed and coordination with the NWS is required. Strict technical criteria have been established for when this type of action is warranted. In reality, it is very unlikely that a unilateral F2P2 flood warning (Message 3) would ever be issued; this has never happened. However, the flexibility to do so has been requested by local emergency managers since they are ultimately responsible for public warning decisions within their jurisdictions. The internal alert (Message I) is by far the most common F2P2 message disseminated and might best be classified as a flood potential advisory. Rainfall amounts meeting M-I criteria would likely cause only minor flood problems and represent a low to moderate threat to life and property. The F2P2 meteorologist is solely responsible for the decision to issue Message Is. The NWS is notified and may choose to issue their own public statement in response. It is important to note that M-Is are not intended for public dissemination, but are used to inform local authorities of the potential for flood problems later in the day and to keep them advised regarding the status of the threat. Between 20 and 30 M-Is are issued every year and special care is taken to identify priority messages so that unnecessary communications do not occur. This assures relatively frequent contact with local officials and helps maintain high degrees of confidence in the program. When a communi cation problem occurs, the opportunity exists to immediately resolve the problem in a manner that best serves the local government. Internal communications and message dissemination procedures for a single local or regional authority can be quite complex and inVOlve many contacts, as illustrated in Figure 2. Clear, concise communication must be used by technical personnel responsible for disseminating flood predictions and other related information to local decision-makers. Technical jargon should be avoided as much as possible. Key phrases like "red flag" or some other appropriate "wake-up" message can be very effective at prompting the desired emergency or pre-emergency response. Emergency management professionals in the Denver area have provided valuable assistance over the years by recommending changes addressing how and when F2P2 meteorolo gists communicate with emergency dispatchers. When the time comes, we believe that our continued emphasis on communications will improve our chances for a successful flood response.

PAGE 161

138 EVOLUTION OF FwOD WARNINGS SYSTEMS IN DENVER Figure 2. Urban Drainage and Flood Control District Flash Flood Prediction Program (F2p2) Weather Message Dissemination [Jefferson County, Colorado] r-Adams County I--Arapahoe County I--Boulder County I--Denver City and County Emergency Management Headquarters Supervisor All Sheriffs Patrol Units Arvada Police Broomfield Police Edgewater Police Golden Police and Fire Lakeside Police and Fire Lakewood Police Littleton Police and Fire Morrison Police Mountain View Police Westminster Police & Fire Wheat Ridge Police & Fire Arvada Fire Coal Creek Fire *PMS I--Douglas County Edgewater Fire Elk Creek Fire Evergreen Fire Fairmount Fire Genessee Fire I--Jefferson County -Golden Gate Fire Idledale Fire City of Arvada I--City of Lakewood City of Wheat Ridge '-City of Aurora *Private Meteorological Service Indian Hills Fire Inter-Canyon Fire West Metro Fire Lookout Mountain Fire Mount Vernon Fire North Fork Fire Pleasant View Fire Trumbull Fire Rocky Flats Consolidated Mutual Water Co. Colorado State Patrol CCIC Computer (Colorado DPS) NOTE: The above represents a typical F2P2 message fan-out for the Jefferson County Enhanced 911 Communications Center. Subsequent message relays within Jefferson County are unknown.

PAGE 162

Stewart 139 Conclusions For flood warning programs to be effective in protecting lives and property, a proper perspective must be maintained by those responsible for evaluating and detecting threatening conditions. Communication barriers must be eliminated by avoiding excessive use of technical terms and codes. Standardized messages are needed to insure consistent information. Judgments must be made by local officials on when and how to disseminate public warnings. Decision-makers must be willing to risk false alarms by recognizing that delayed warnings may cost lives. Specific responses can be targeted to known problem areas if reliable predictions andconfmning reports are conveyed to the appropriate officials in a timely manner. Those involved with local flood warning operations must anticipate public response and understand that people tend to seek confirmation before perceiving personal danger. Therefore, actions like barricading flooded road crossings must occur before a motorist or pedestrian makes the wrong choice. Police and other public safety officials need to know the locations of hazardous stream crossings and other problem areas. Floodplain residents must also be warned ani may need to be evacuated. While meteorologists and hydrologists strive for"accurate" flood predictions and increased lead times, equal or greater importance needs to be placed on recognizing when initial actions should be taken to prepare for a possible flood and where to target emergency field resources. References Downing, Thomas E. 1977 Warning for Flash Floods in Boulder, Colorado. Working Paper No. 31. Boulder, Colorado: Natural Hazards Research and A pplications Information Center, Institute of Behavioral Science, University of Colorado. 1977 Flash Flood Warning Recommendations for Front Range Com munities. Denver: Urban Drainage and Flood Control District. Gruntfest, Eve C. 1977 What People Did During the Big Thompson Flood. Working Paper No. 32.Boulder, Colorado: Natural Hazards Research and Applications Information Center, Institute of Behavioral Science, University of Colorado.

PAGE 163

140 EVOLUTION OF FLOOD WARNINGS SYSTEMS IN DENVER Stewart, K. 1987 "Planning for the Inevitable-Urban Flash Flood Warning Pro grams in the Denver Metropolitan Area." Paper presented at Annual Meeting of the Association of State Floodplain Manag ers, Seattle, Washington. Stewart K. and L.S. Tucker 1993 "Flash Flood Prediction and Early Warning." Paper presented at the Republic of China Workshop on Natural Disaster Reduction, Taipei, Taiwan, June. Urban Drainage and Flood Control District 1977 Early Flood Warning Planning for Boulder Creek. Prepared by Leonard Rice Consulting Water Engineers, Inc. Denver, Colo rado. 1991 "Flash Flood Prediction Program and Related Activities." Flood Hazard News 21 (1). 1992 "Flash Flood Prediction Program and Related Activities. Flood Hazard News 22 (1). 1996 Boulder Creek Flood Warning Plan. Denver, Colorado. 1996 Ralston Creek Flood Warning Plan. Denver, Colorado.

PAGE 164

PuTTING EFFECTIVE FLOOD WARNING SYSTEMS IN PLACE THE PROCESS AND GUIDELINES IN AUSTRALIA John Handmer Flood Hazard Research Centre Middlesex University London Chas Keys State Emergency Service Wollongong, New South Wales Australia Introduction Institutional change within emergency management over the last decade has been inexorable: moving at a rapid pace toward greater cooperation among the various emergency services and engagement with other stake holders and toward broadening mandates away from simply response. These changes have been driven by the needs for improved effectiveness of service delivery and improved efficiency by delivering these services at lower costs-or at least by demonstrating that they are cost-effective. To a significant extent, these changes reflect broader changes in the approach to government, greater public scrutiny and expectations, greater self-criticism, and the influence of trends overseas. The institutional evolution has also made the emergency management system more open to change and learning. Major emergencies and the political demands for action they stimulate also provide opportunities for achieving change. However, to be sustainable, changes must generally occur in the culture of the relevant organizations; that is, in the way they habitually do things. These changes have also affected the institutional arrangements governing flood warnings in Australia. After a decade of uncertainty over responsibility for warnings, during which there was no new investment in warning systems,

PAGE 165

142 EFFECTIVE FLOOD WARNING SYSTEMS IN AUSTRALIA a compromise was reached. The Australian government clarified and reasserted the Bureau of Meteorology's primary role in flood warnings, after which the Bureau committed itself to warning system upgrading. The states agreed to contribute funds to this effort. As part of the upgrading process, the Bureau sought increased involve ment of various state-based agencies. An important initiative here was the establishment of Flood Warning Consultative Committees (FWCCs) in each state and territory in 1989 and 1990. These were set up by the Bureau to guide the expansion of the Bureau's warning system with advice from agencies with potential data inputs and from agencies representing users of the service. In addition to the Bureau, members included representatives from water authorities, emergency service organizations, and local government who focussed on improving the accuracy and timeliness of forecasts by the flood warning centres operated by the Bureau in each state. Almost simultaneously, however, severe flooding in eastern Australia led to a demand for re-evaluation of how warnings were issued, resulting in the production of a "best-practice" warning guide (EMA, 1995) that also reflected the evolutionary changes occurring in Australia's emergency management institutions. This paper explores the process by which lasting changes to flood warning system design and operation have been sought through the creation of this guide. The authors were members of the team that developed the guide. Developing Effective Flood Warning Systems Modern technology has led to substantial increases in data and forecasting reliability and, in some cases, to warning lead time. However, although very important, such changes are relatively easy to implement and do not by themselves improve flood warning systems. Thus, the ultimate goals must be improving safety for those at risk and reducing flood losses. Achieving the necessary changes may pose both conceptual and practical challenges for the organizations involved. Conceptually, organizations need to broaden their perspective and see their contributions in terms of the overall objective rather than their component of it-to shift toward benchmarking by performance of the total system, rather than its component parts. This "total system" includes those at risk, and, in order to be effective, must incorporate the processes of review and learning. Practically, operational interaction between the various organizations needs to become "seamless." The Guide contains the following definition (p. 5):

PAGE 166

Handmer and Keys 143 A total flood warning system integrates flood prediction, the assessment of likely flood effects, the dissemination of warning information, the response of agencies and the public in the threatened community, and review and improvement. These components must operate together for sound flood warning performance to be achieved. Critical to the total system is how the various components function together to serve the system: "to empower individuals and communities to respond appropriately to a threat in order to reduce the risk of death, injury, property loss and damage" (Bureau of Meteorology and Australian Emergency Management Institute, 1993). Critical concepts include: integration, cooperation, shared responsibility, and thinking broadly about problems. In practical terms these translate into, for example, the inclusion of all relevant organizations, integration with floodplain management activities and emergency management activities, ownership by all organizations involved, and cooperative work with others to improve operation. Substance and Process Guidelines or manuals of practice for emergency management abound. They are easy enough to produce, but are of little value unless enthusiasti cally adopted and used by both planning and operational staff. Thus, success in substantially improving flood warning systems requires more than a sound document, although the document-the substance of the guidelines-is very important. Although it is also relatively easy to construct in a technical sense, it is much more difficult to ensure that the guidelines capture the detailed requirements of those expected to implement them, giving the document its essential credibility and legitimacy. The production of guidelines cannot be seen as the end of the process either; the institutions involved need to create a learning environment for continuing evolution and improvement. We suggest four components to lasting change in flood warnings: 1) the institutional context must recognize the need for continuous change and accommodate it; 2) the process used to develop the guidelines, to ensure ownership; 3) the document containing the guidelines; and 4) the process to ensure acceptance and continuing development.

PAGE 167

144 EFFECTIVE FWOD WARNING SYSTEMS IN AUSTRALIA The Process The formal process leading to the final document took over three years, but this is definitely not the end point. Embedding, that is, ensuring that change is permanent and improvement continues, may take another few years. The first steps below describe the development of a national consensus that a document was necessary to guide change in flood warning system design and operation. Then, once writing began, the challenge changed to keeping the key players involved. In April 1990, the three mainland states of eastern Australia experienced extensive, severe flooding. In particular, the unsuccessful struggle to prevent flooding in the remote town of Nyngan (population 2,500), and its wholesale evacuation by helicopter after its levees were breached, were major media and political events. Much of the blame was placed on inadequate warnings, and even though far from a flash flood, what happened in Nyngan helped put warnings on top of the post-flood action agenda. Emergency Management Australia (EMA), the Australian equivalent to the U.S. 's Federal Emergency Management Agency (FEMA), took the first formal step in the process by convening a national conference to consider what could be learned from the 1990 floods. The 50 participants at this meeting represented the state and territory FWCCs and members of the broader flood warning community, including media and research interests. Participants identified flood warnings as an area in need of detailed examina tion, and the 1991 workshop on flood warnings followed. This workshop called for the production of some "best practice" guidelines. Following the call, the workshop working group (or steering committee) was directed to produce a report of the meeting, along with some normative material. The results were circulated for comment among workshop participants, and it became clear that something different was required. This feedback led to the decision to produce a handbook or manual on the best flood warning practices. A subgroup of the original working party took on this task; they were given some resources for meetings by AEMI (Australian Emergency Management Institute, the training alm of EMA) and retained the backing of the participants at the 1991 workshop. Although the establishment of a stable team to guide the process, collate material, and draft the guidelines was informal and evolutionary, it enabled completion of the task and ensured broader ownership than would be possible with one individual. The "team" informally represented the Bureau of Meteorology, the state emergency services, Emergency Management Australia, and research interests. The practitioners were senior people with the unambiguous backing

PAGE 168

Handmer and Keys 145 of their organizations, and the group composition provided a good balance between experience and stakeholders as well as practitioner interests and critical outside evaluations. Document production was a long iterative process among members of the writing group and their personal networks; between them and (some of) the original workshop participants, and a loose reference group of people in the broader warning and emergency management profession. All this helped maintain wider group interest and commitment, kept the FWCCs involved, and helped direct the style toward the guidelines' audience-practitioners, including volunteer state emergency service (SES) staff, police, and local government officials. Once a draft of the document was produced, we also circulated it to a few nonspecialists for comment on writing style and level. They felt that sections were difficult for a nonspecialist to comprehend and included other comments on contents and structure. Consequently, we tried to make the text more user friendly and easier to read, and inserted summaries of the key points in each chapter. These are in question and answer format and enable the Guide to be examined in full at a summary level. Eventually, after further detailed review by our informal extended reference group, we felt the document was complete. It was sent formally to all FWCCs to seek their endorsement, and the draft document was endorsed by all FWCCs and by EMA. However, we did not stop there, but continued to seek high level endorsement from the relevant organizations. In addition, although we had worked with a reasonably wide range of people throughout the long process of document preparation, most "rank and file" members of state and territory emergency planning and response organizations were largely unaware of the process. The next stage, therefore, was to have the material accepted by operational staff, who are critical in delivering warnings at the community level. The Document Flood Warning: An Australian Guide The aim underlying the call of the 1991 flood warning workshop for national guidelines was to develop a clearer understanding of the flood warning task among practitioners by defining best practice in the field and indicating how it can be approached in different environments (see Table 1). A crucial part of this understanding is the concept of a "total flood warning system" into which the myriad tasks carried out (usually) by a number of organizations must be incorporated. Another important notion is that

PAGE 169

146 EFFECTIVE FLOOD WARNING SYSTEMS IN AUSTRALIA practitioners need to plan their approach beforehand (Chapters 7 and 8) in order to apply the most appropriate warning strategies for their own area given the nature of their particular flood hazard (e.g., speed of onset and area of potential inundation) and the nature of the community (demographics, in the broad sense). Planning and maintaining the system are time consuming tasks. That said, general principles must also be inculcated. For example, warnings should add value to (i.e., be more than merely) height predictions, give advice consonant with the nature of the threat and its likely impacts, communicate in ways that can be understood, and use appropriate "layers" of dissemination techniques. While many of these elements involve investments of time and money, a substantial number are nontechnical and inexpensive to implement; for example the identification of tasks and procedures. Many of these matters, if not all, involve ongoing activities characterized by liaison with communities at risk and other organizations in the flood warning system. Self criticism and reflection are also required to ensure that performance is evaluated honestly-a prerequisite to instituting effective change. Table 1. Guide Contents: Chapter Titles and Selected Subheadings 1) The place and purpose of effective flood warning 2) The total flood warning system (including public education and extreme events) 3) Flood prediction (including user requirements and "informal" prediction systems) 4) The interpretation of flood predictions (including adding meaning and information requirements) 5) The design of warning messages (including understanding the flood, the communities, and message design) 6) The communication of flood warnings (including types of warnings and dissemination, and the informal "system") 7) System review and improvement 8) Conclusions

PAGE 170

Handmer and Keys 147 Embedding the Ideas Production of the Guide and its widespread acceptance by senior management are important factors in full implementation. By themselves, though, they do not necessarily mean that the Guide will find its way into practice, or that the necessary changes will become permanent features of flood warning practice. It is just as important that operational staff support the Guide and implement it. EMA made resources available to hold workshops around Australia (except for the states of Tasmania and Victoria) to explain and discuss the Guide. These half-day workshops involved the Bureau, police, state emergency service staff, volunteers, and so on. Workshops walked partici pants through the Guide, and small group exercises helped participants examine the local context. Although the seed has been sown, more workshops are needed, and that will be the job of the FWCC/SES in each state. In New South Wales the Guide has been "workshopped" at SES conferences around the state. The state's 18 emergency management divisions are discussing flood exercises and routinely spend a day defining the warning task and reiterating the principles, goals, and methods. Post-flood debriefing should also occur in order to break the old warning mindsets: that warnings are only about "postboxing the Bureau's flood height predictions," and that all that is needed is general information provided over local radio stations. Debriefings should also be used to encourage SES staff-who are volunteers-to focus on using and building up "flood intelli gence cards" (which identify the local consequences of flooding at various heights), appropriate dissemination modes, layered warnings, and so on as set out in the Guide. Institutional Change: High Level Support For The Guide A recent evaluation showed that flood-related agencies welcomed the work of the FWCCs (Bureau of Meteorology, 1995) and generally agreed th3:t the committees have encouraged a greater level of cooperation between the various flood warning-related agencies. The FWCCs have provided a new forum and achieved more than simply advising the Bureau about its investments. In essence, the committees helped build an understanding of the multi-agency role of flood warning and educated personnel about their roles in the warning process. Nevertheless, the review suggested that the committees' terms of reference be broadened to "cover the performance of all aspects of flood

PAGE 171

148 EFFECTIVE FLOOD WARNING SYSTEMS IN AUSTRALIA warning procedures" (BoM, 1995: v). This change has been accepted by the Bureau of Meteorology and takes the FWCCs well beyond the role of merely buttressing the predictive process-which is what the Bureau originally set them up to do. Following this review, the director general of EMA has reached agreement with the director of the Bureau of Meteorology that the new role of the FWCCs should include encouraging the notion of "best practice" and monitoring progress toward higher quality flood warnings, using Flood Warnings: An Australian Guide as the basis. Conclusions Putting effective flood warning systems in place in Australia is a matter of evolutionary institutional change, creating the right context for rapid or substantial changes following severe flooding. The institutional context enables opportunities to be taken, and importantly allows for "embedding" and permanent change. The changes necessary for effective flood warning systems cannot be seen as single events-they require ongoing constant attention; hence, the need for the right institutional framework. Of course, this framework itself must be flexible and adaptive to changing circumstances and ever ready to take advantage of opportunities. The development of learning organizations and a supportive institutional environment is (or at least should be) provided by the FWCCs. The training arm of EMA, AEMI, should help to provide the learning environment and act as a key mediator between research/best practice and ordinary practice. The Australian guidelines had their impetus in serious flooding, but the necessary institutional environment was already in place. The key factor in developing the Guide with a high degree of ownership by those in the flood warning business was the use of a long iterative process that was as inclusive as possible. Once the Guide was developed and endorsed by the peak flood warning body in each jurisdiction, the FWCCs, attention turned to how to ensure the process continued. In this the FWCCs were the key. Support from the highest levelled to a change in the mandate of the committees (which was occurring anyway) and to them taking on the task of encouraging and monitoring the implementation of the Guide. The document containing the Guide is now universally and correctly seen as only part of the broader picture of achieving change. How would we characterise the factors underlying what, so far at least, appears to be a successful process? It is difficult to argue that anyone factor is more significant than others, but we can identify a number of important elements. The process was open and transparent; we tried to be inclusive and

PAGE 172

Handmer and Keys 149 proactive about encouraging input and comments from a wide diversity of perspectives. The writing "team" had credibility among the Australian flood warning industry and brought a mixture of writing skills, expertise, relevant networks (for example, one member of the team was a member of all state FWCCs and a senior member of the national flood prediction agency), critical perspectives, and independence. The fact that the team members were senior people in their fields and had the unambiguous backing of their agency heads meant that they could provide leadership and give the task priority. However, these factors are mostly under the control of those guiding document production. The organizations in the flood warning process must support the activity and need to be seeking, or at least open to, change and improvement. This is much more of a challenge than it sounds, as the sort of change required may involve fundamental changes to the organizational culture and priorities: in particular to move from technical factors to consider community expectations. Similarly, the organizations must be open to active participation and acceptance of ownership of the process and product-and interested in its continuation. In view of the three years it took to prepare the guidelines, an obvious question is whether they could be prepared in less time. There is no doubt that they could have been. But even if the writing group always gave the task top priority, in an interactive inclusive process the pace is inevitably dictated by others who will see it as a lower priority than dealing with immediate problems. Rushing may result in half-hearted input and therefore low commitment by these individuals and their organisations. It is likely that proper planning involving the relevant stakeholders is an inherently slow process. Thoroughness is vital, and planning must be slow and involve continuous reinforcement and repetition to effect change at the practitioner level. Another way of viewing planning time is to see it as a permanent unceasing commitment. It would be possible to draw up a set of guidelines by having an intensive meeting for several days, but this would not allow the necessary time for reflection, and to ensure resilience in the face of constantly changing institutional and political priorities. What does need to be said here is that instituting change at a level of the local practitioner is difficult given that flooding is relatively rare at the grass roots. The natural tendency is for people to do the basic minimum that they remember from long ago, and it takes a sensitive form of intervention for additional or innovative practices to be introduced. Note too that most of the operational flood-response work in Australia is done by volunteers or people who don't breathe the flood game every day of their working lives.

PAGE 173

150 EFFECTIVE FLOOD WARNING SYSTEMS IN AUSTRALIA Acknowledgements We express our thanks to our co-authors on the Guide: Jim Elliott (Australian Bureau of Meteorology) and John Salter (Emergency Manage ment Australia). Many people-far too numerous to mention by name-helped with the Guide. We are grateful to them all, and in particular to the other members of the original "working party" (Dingle Smith, Australian National University; Roger Jones, Emergency Management Australia; and Ken Mackey, Victorian Police), our informal reference group, to all those who provided written comments on the various drafts, to the flood warning consultative committees, and to Emergency Management Australia for sponsoring the project. We appreciate the efforts of Professor Eve Gruntfest in organizing this conference. References Bureau of Meteorology 1995 Flood Warning Service Upgrade. Canberra: Bureau of Meteorology, Department of the Environment, Sport and Territories. Bureau of Meteorology and Australian Emergency Management Institute 1993 Guidelines for Effective Warning. Mt. Macedon: AEMI. Emergency Management Australia 1995 Flood Warning: An Australian Guide. Canberra: Emergency Management Australia.

PAGE 174


PAGE 175

This page IS intentionally blank

PAGE 176

FLASH FLOODS IN MEXICO Ma. Teresa Vazquez, Ramon Dominguez, and Oscar Fuentes Centro Nacional de Prevenci6n de Desastres (CENAPRED) Jose Antonio Maza Comisi6n Federal de Electricidad (CFE) Abstract In this paper the main characteristics and damages and some rainfall and flow data are presented for the localities that have been most affected by flash floods in Mexico. Some conclusions are derived. Introduction Flash floods mostly occur in the northwestern part of Mexico. This region is characterized as semiarid. Streams flow intermittently, with long periods without any discharge; the area has scarce vegetation and its catchments are relatively small, with steep slopes, an erosion-susceptible surface, and high elevations. In this zone, intense floods of rather short duration have taken lives and caused economic losses. Flash floods occur due to cold fronts, cyclones, and sometimes to meteorological phenomena that are difficult to predict. In semiarid regions, towns and cities are usually founded upon the flood plain. If there is not enough water-carrying capacity in rivers, damage can increase. For this reason, it is understandable why disastrous events occur in these regions, as in Los Cabos, State of Baja California Sur in November 1993, where a flash flood caused $63 million in damage, or in 1976 when EI Cajoncito Stream flooded the city of La Paz, in the same state, causing the largest number of deaths by floods ever registered in the history of Mexico. This paper describes in some detail several of the most important flash flood events in Mexico that have caused considerable damage. Some conclusions are presented.

PAGE 177

154 FLASH FLOODS IN MEXICO Important Flash Flood Events Floods in La Paz, BCS, in September 1976 The city of La Paz, capital of the state of Baja California Sur, is periodically subjected to floods due to the overflowing of EI Cajoncito and EI Piojillo streams. In order to divert EI Cajoncito flow into EI Piojillo stream, some protection works were constructed, encouraging human settlements along a stretch of EI Cajoncito. On September 30, 1976, Hurricane Liza generated heavy rains that caused the collapse of a 6-meter high dike that acted as a small dam and was used to protect the urban area inside the dry riverbed. With its collapse, a big flood occurred downstream and caused the greatest number of deaths ever recorded due to floods in Mexico. The city of La Paz is surrounded by the La Laguna Sierra in the southern part, with elevations of 1,250 meters. Near the city, there is a group of seven low hills. When cyclones occur, the steep slopes of EI Cajoncito and EI Piojillo streams turn the streams into a wild torrent of water and, therefore, the high speed current causes an intense particle transport mostly originated by strong weathering processes of the rock (granite) of the La Laguna Sierra. Most of the material in suspension is deposited along the banks in the city because here the EI Cajoncito river bed slope diminishes. The rainfall recorded at EI Cajoncito climatological station on September 30 was 180 mm. The rain caused a flow rate of 950 m3/s with an approximate duration of 7.5 hours. These data were calculated by indirect methods because at that time there were no hydrometeorological stations near or in the basin. On September 30, 1976, Liza's destructive effects were felt in the city in the afternoon (3:30 p.m. approximately), with winds reaching 150 lcm/h, tearing trees, posts, and billboards and leaving the city with no electricity or drinking water (El Heraldo newspaper, October 4, 1976). At approximately 7:30 p.m., the strong current and the large volume of water carried by EI Cajoncito stream caused the collapse of a dike. The rupture released a water avalanche of almost 2 meters in height over four poor neighborhoods, where most of the houses were built out of wood and cardboard. The wave carried away people, vehicles, houses, and trees and buried them under large volumes of sand almost two meters deep. A stretch of 2.3 x 5 lcm, in which almost 2,000 persons lived, was converted again into the natural river channel (El Heraldo newspaper, October 6, 1976).

PAGE 178

Vazquez, Dominguez, Fuentes, and Maza 155 Even though the official death toll was about 600, it is possible that more than 1,000 people were killed. Material damages reached $3 million as of 1976, and between 10,000 and 12,000 people were left homeless (Avante newspaper, La Paz, BCS, October 7, 1976). Floods in Monterrey, Nuevo Leon, in September 1988 From September 14 to 17,1988, Mexico was battered by Gilbert-one of the strongest hurricanes in its history. It first struck the northern part of the Yucatan Peninsula, then moved through the Gulf of Mexico, and finally reached the coast of the state of Tamaulipas, where it began to weaken as it moved inland. The total duration of hurricane Gilbert was five days. Most of the accumulated rainfall caused by Hurricane Gilbert was concentrated in the northeastern region. Over the city of Monterrey and its surrounding area, there were heavy rains reaching over 200 mm during September 15 and 16. The maximum rainfall recorded was 370 mm. The major impact occurred early in the morning of September 17, when Gilbert struck the Sierra Madre Occidental with hurricane wind velocities over 150 km/h (Absalon, 1989). Heavy floods were generated in the main rivers that cross the states of Coahuila, Nuevo Leon, and Tamaulipas. The maximum recorded flows in some affected currents are indicated in Table 1 (Rosengaus, 1989). Table 1. Maximum Flows Recorded In Some Affected Currents RIVER STATION AREA GILBERT FLOW HISTORIC (km2 ) (m3/s ) FLOW (m3/s ) Santa Catarina Puente Zaragoza 1333 1900 178 San Juan Tepehuajes 3594 5880 2302 Pesqueria Los Herrera 20023 900 1317 Potosi Cabezones 1166 6900 1128 Heavy rainfall and floods generated by Gilbert in rivers and streams cost human losses; destruction of houses, roads, fords, dams, irrigating channels, and public utilities. Among the most affected zones, the metropolitan area of Monterrey, which is located in the surroundings of the channel that carries the Santa Catarina River to its confluence with the San Juan River. The day

PAGE 179

156 FLASH FWODS IN MEXICO of the event, this river-almost always dry and lifeless-recorded a flood of 2,000 m3/s Along the Santa Catarina River and in some other rivers mentioned in Table 1, there are frequent and extensive irregular human settlements. The phenomenon destroyed more than 10,000 houses of these makeshift structures. The street works that run along both banks of the Santa Catarina River were partly destroyed, and some vehicles and four passenger buses were swept away by the strong current and buried under several meters of mud. The roads that were near the overflowing rivers were covered by mud almost 1.5 m deep. In the state of Nuevo Leon, damage was assessed at $85.8 million as of 1988. It is hard to define the death toll, due to several factors: the explosive proliferation of human settlements of marginal classes (particularly families with many children); the lack of information about the number of people traveling in cars and buses along the Santa Catarina River; the number of people trapped by the current; and the number of people who were able to escape. It is estimated that about 200 people died. Most of the deaths were the result of imprudence, because victims tried to cross the Santa Catarina River in Monterrey using fords that cross the riverbed; approximately 80 % of the death toll is attributed to these attempts. This is a dramatic example of what the lack of a disaster culture represents, among the people and authorities of low level, which, in this instance, make decisions that can be a matter of life or death. Floods in Los Cabos, RCS, in November 1993 During November 3, 4, and 5, 1993, unusual storms of extraordinary intensity occurred in the highlands near San Jose del Cabo and Cabo San Lucas in the State of Baja California Sur, catching its 50,000 inhabitants unaware. The storms caused three deaths (Excelsior newspaper, November 5, 1993), affected 10,000 people (most of them in San Jose del Cabo), caused damage of more than $63.4 million as of 1993 and the partial or total interruption of daily activities. Table 2 shows a summary of the reported damage. It is estimated that the storm brought 632 mm of rain in 24 hours. This is three times higher than the mean annual rainfall and twice the highest rainfall recorded (317 mm in 1941). The meteorological origin of the heavy rains that affected Los Cabos region in November 1993 is uncertain because the rain was not caused by a hurricane nor a cold front, but rather by a disorganized

PAGE 180

Vazquez. Dominguez. Fuentes. and Maza 157 Table 2. Damage Caused in San Jose del Cabo and Cabo San Lucas (Los Cabos Region) AFFECTED STRUCTURES Communication Facilities Houses Businesses Public Services Infrastructure Agriculture DAMAGE CAUSED One third of the Airport-Cabo San Lucas road was destroyed About eight neighborhoods in San Jose del Cabo and Cabo San Lucas were buried in sand. Walls of several houses collapsed. Buried in sand. 50,000 inhabitants of Los Cabos were left with no drinking water, communication by land or air, electricity, telephone, and waste water treatment systems. Streets were washed away. Two bridges collapsed. 40 stretch es of water mains broke. Opera tions were interrupted at Cabo San Lucas International Airport. Col lapse of tourist activities. Ports were closed, several vessels sank. Economic losses were about $1.6 million as of 1993. disturbance of low pressure combined with a winter system. Also, the storm had a very local presence and was short lived. On the other hand, the water flow carried away large volumes of sand that formed between 1.5 and 2 meters of deposits. The flood buried cars and properties that had been constructed at the banks and on the bed of the rivers. As a result of seafront tourist development in Baja California Sur and the presence of considerable high mountain chains near the sea, most of the population centers are located on floodplains and, in many cases, on riverbeds, with no protection against floods. Considering the interference of road infrastructure with water drainage; the lack of a minimum sewage infrastructure for pluvial waters in the localities; poor urban growth planning,

PAGE 181

158 FLASH FLoODS IN MEXICO and the presence of irregular settlements and more permanent neighborhoods on the beds and banks of the rivers, events like the one of November 1993 are likely to happen. Floods in Tijuana, Be, in January 1993 Between January 6 and 7 of 1993, there was a 100 mm rainfall in the city of Tijuana, state of Baja California, in less than four hours, causing more than 20 deaths and substantial damage and leaving more than 10,000 people homeless. Due to this storm, the city, with a little more than a million inhabitants, was totally paralyzed on January 7 and part of the following day (Excelsior newspaper, January 11, 1993). The heaviest precipitation occurred on the 6th and 7th; however, there were uninterrupted rains during the next 14 days that surpassed the maximum levels of pluvial precipitation over the last 30 years. This situation caused chaos in Tijuana. Due to the major storm of January 6 and related floods, 32 neighborhoods were affected. Rains occurred due to a polar front that caused damage mainly in Tecate, Mexicali, and Ensenada, and particularly in Tijuana. The Tijuana and Tecate rivers, as well as the Alamar, El Carrizo, and Matanuco streams overflowed, covering everything in their path with mud. Damage in Tijuana were estimated to be near $31.7 million as of 1993. The flooding caused interruptions in drinking water and electricity supplies; subsidence of the ground as much as six meters in four locations in the city (due to saturated soils), as well as landslides on hillsides. Several houses and walls collapsed because of soil softening, and in several places water and mud accumulated inside houses, stores, and on streets. Roads and bridges also failed and collapsed. This natural disaster could have caused less damages and loss of life. The damage increased because Tijuana lacks proper urban infrastructure, as well as adequate land use planning for urban growth. Most of the people affected by the storm that hit Tijuana lived on the slopes of hills, canyons, and gulches, brooks; and even on riverbeds. About 10,000 families "live" in wood, cardboard, and sometimes in adobe houses on the soil floor. These zones were initially irregular settlements that belong to Tijuana's riverbed. Later on, state properties on hills and rivers were regularized without considering the possibility of landslides or floods. Here, inhabitants do not have access to regular community services (e.g., electricity, sewage, pavement, law enforcement, and security services).

PAGE 182

Vazquez, Dominguez, Fuentes, and Maza 159 As of January 17 (after more than one week of rains), official reports indicated that there were more than 25 deaths, 10,000 homeless, dozens of injured people, collapsed houses and walls, neighborhoods and towns with no communication, floods, and chaos. This case is a clear example of the lack of experience regarding disaster management and disaster culture among the population and authorities, which is reflected on the urban infrastructure and mainly occurs in provincial cities where the federal budget is small and, therefore, limits the planning of urban development and the construction of an adequate infrastructure to withstand these phenomena. Conclusions Flash floods occur in some regions of the Mexican countryside. When they happen in semiarid zones, they can cause severe damage as well as human and material losses. Damage can increase due to the irregular growth of cities and localities that leads poor people to settle in high-risk zones, with no control by authorities. When authorities finally become aware of such settlements, these communities have already been converted into real neighborhoods that have to be regularized, with no preventive measures. Another important aspect is the ignorance of people and authorities about the hazards to which they are exposed. Similarly, sometimes poverty is so pervasive that people do not leave their homes, even though they are aware of the danger-they have no other place to go. On the other hand, hydraulic works built to protect against floods sometimes are not adequately designed, because the regulations for construc tion of bridges, fords, levees, etc., do not consider the required features for this type of extraordinary event. For these reasons, it is necessary: To regulate community development, with expert identification of hazard zones and creation of land use maps associated with different return periods. To compel people not to settle in floodplains. To build adequate infrastructure, for example, sewage and water. To construct dams in order to retain loose soil, washed by the flood waters, to prevent it from reaching downstream where the

PAGE 183

160 FLASH FLOODS IN MEXICO majority of settlements are located. In La Paz, the Buena Mujer Dam (upstream from the city) was built for this purpose. To implement continuous campaigns to improve hazard awareness and develop a disaster culture among citizens and authorities. It is important that people are aware of potential hazards. To develop a building code in semiarid zones, in basins with steep slopes, small areas, and heavy rainfall in order to design better hydraulic works and other structures that, if poorly designed, can interfere with water flows (bridges, embankments, culverts, fords, drains, etc.). Maintenance is basic for an adequate performance of the hydraulic structures, especially when an extraordinary meteoro logical event occurs. To collect the largest number of studies in Mexico related to this topic. Given the nature of the phenomenon, it is difficult to forecast it well in advance, so it is convenient to have good warning and instrumentation systems, and an adequate design of the hydraulic works. For this purpose, Mexico started in 1993 the modernization of the rainfall measuring equipment and in general, of the National Meteorological System (SMN), in order improve forecasting and prediction of hydrometcorological phenomena. References Amaro, T. Absalon 1989 Cronicas de desastres en el estado de Nuevo Leon. Cd. Guadalupe Nuevo Leon, 1989, Mexico. Rosengaus, M. Michel, S{mchez-Sesma, Jorge 1990 "GilbertEjemplo de Huracanes de gran intensidad," Revista de Ingenieria Hidraulica en Mexico, vol. V, num. 1, II Epoca, IMTA, CNA Publication, Enero-Abril 1990, Mexico.

PAGE 184

FLOODING AND THE DEMISE OF THE MOCHE EMPIRE Kenneth R. Wright Wright Water Engineers, Inc. Denver, Colorado John Dracup Civil and Environmental Engineering Department University of California, Los Angeles Abstract The EI Nino flooding of the Moche River Valley in Northern Peru between 500 and 700 A.D. contributed to the collapse of the great Moche empire. Intense flooding devastated the capitol of Moche and stripped away the urban landscape and agricultural fields. When the capital city was moved from Cerro Blanco to Pampa Grande, the new capital also experienced flood problems. Later, when the area was occupied by the Sican people, flooding continued to destroy urban and agricultural development, with their capital at Batfm Grande. Cooperation between archeologists and paleohydrologists for the study of ancient civilizations and their demise is useful in better understanding the long-term impacts of climate and flooding on modern civilizations. Introduction In the northern coastal region of Peru, the Moche empire rose to its height of state-craft and power by about 500 A.D. Political consolidation had been completed over a coastal area extending 600 Ian in length, and the Moche empire grew to prominence and power. The Moche are most well known to modern people by their exquisite ceramic pottery, which is today sought by the world's great museums. The basis for the Moche culture developed over a long period of time. Civilization, as measured by the use of pottery, commenced in northern Peru about 1800 B. C., which was also the time of the spread of intensive farming. Then, by 1200 B.C. the largest architectural

PAGE 185

162 FwoDING AND THE DEMISE OF THE MOCHE EMPIRE monument in the western hemisphere was built at Sechin; it was the U -shaped ceremonial center of Sechin Alto. Irrigation of land to grow food gave rise to centralization of power through corporate agriculture, which meant food surplus could be created to free portions of the population for building monuments and staffing armies. At the same time, control and distribution of food production and its surpluses provided power to those who held the reins. The Moche capital was at Cerro Blanco, with two enormous temples of adobe called Huaca de la Luna and Huaca del Sol, around which were magnificent buildings with grand courts, low platforms, many residences, and workshops. The temples, the city, and the production of fine ceramics and artwork were the manifestation of one of the New World's finest early civilizations with a well organized social system. The Achilles tendon of the Moche capital was its reliance upon and proximity to the Moche River. The capital and the irrigation canals and fields were in the floodplain and susceptible to the ravages of floods. Significant EI Nino Southern Oscillation (ENSO) flooding occurred in the 6th century. The flooding devastated the Moche empire because the all important irrigation works were destroyed in addition to the damage to the capital city. The great civilization had intensively developed in the floodplain of the River Moche without adequate consideration of the ENSO phenomenon and its ability to cause periods of extreme regional precipitation. EI Nino Southern Oscillation Changes in ocean currents and sea surface temperatures (SST) in the eastern equatorial Pacific were first noted and named by Peruvian fishermen in the 1890s. They called this phenomenon "EI Nino" because it usually occurred during the Christmas season. The EI Nino event is linked to a large-scale atmospheric pressure fluctuation that is measured at Darwin, Australia, in the western Pacific and at Tahiti in the central Pacific. This atmospheric phenomenon was named the Southern Oscillation by Walker (1923) and is now termed the Southern Oscillation Index (SOl). A positive value (phase) of SOl (standard index pressure anomalies at Tahiti minus Darwin) causes cooler SSTs to occur in the eastern equatorial Pacific and is termed La Nina. Conversely, the negative phase of SOl results in warm SSTs occurring in the eastern equatorial Pacific and is termed El Nino. The EI Nino and SOl together have been termed ENSO events (Diaz et al., 1992).

PAGE 186

Wright and Dracup 163 It has been documented that the two phases of SOl cause climatic anomalies throughout the world. Climatologists have stated that the ENSO phenomena are the second most important influence on climate in the world after the four seasons. The EI Nino phase of ENSO causes flooding in the Peruvian Andes, while the La Nina phase of ENSO causes drought. The excessive precipitation in the Peruvian Andes associated with the EI Nino phase of ENSO results in greater than normal floods pouring out of the coastal canyons into the alluvial valleys downstream from the canyons. The floods would carry with them great quantities of silt, sand, gravel, and rocks. The climatological record of great droughts and periods of floods has been inscribed into the great glaciers and ice caps of the Andes. One such ice cap is called Quellccaya. Paleoclimatic Evidence of Climatic Variability The Quelccaya ice cap is located in the Cordillera Oriental mountain range of southern Peru approximately 200 Ian northwest of Lake Titicaca. Scientists from the Byrd Polar Research Center at Ohio State University drilled and analyzed ice cores from the Quelccaya ice cap during the 1980s. Their work resulted in a climatological record dating from 470 A.D. to 1984 A.D. (Thompson and MosleyThompson, 1989). The annual accumulation of ice recorded in the ice core can be used to determine the relative annual precipitation for the region, and consequently both drought and flood patterns. Ice core data indicate time periods of deviations from the mean. A specific review was made of one major deviation, the post-WOO A.D. period, when a definite climatic shift to a lower precipitation persisted for several centuries. A statistical analysis of the ice core data was performed to determine if the post-WOO A.D. climatic shift was statistically significant. Using a t-test, the annual ice accumulation was found to have a statistically significant change at around 1000 A.D. (Ortloff and Kolata, 1993). The severity of a drought is determined by the cross product of its duration and its magnitude. The average value of the precipitation anomaly determines the magnitude of the drought. The duration of the drought is the time length of the precipitation anomaly (Dracup et aI., 1980a and b). Using these definitions, an extremely dry year is not over-emphasized if the following years return to normal. The data obtained from the ice cores included isotopic 180 data as well as the annual ice accumulation data. The isotopic 180 measurements are valuable in that they enable researchers to infer prevailing temperature data.

PAGE 187

164 FLOODING AND THE DEMISE OF THE MOCHE EMPIRE Isotopic 180 measurements from the Quelccaya ice cap indicate a rise in mean annual temperature of between 0.50 and lC beginning around 1000 A.D. and persisting until at least 1400 A.D. (Ortloff and Kolata, 1993). The rise in temperature, corresponding with the decrease in precipitation, indicates that there was indeed a climatic shift during this period. A somewhat similar evaluation technique can be used to estimate periods of excessive precipitation. For example, EI Nino episodes occurred in A.D. 511-12, 546, 576, 600, 610, 612, 650, 681, and with the same type of frequency in later centuries (Mosley, 1991). These floods were bad enough, but a three decade long drought happened between 562-594 A.D., with precipitation 30 % below normal. A marked increase in atmospheric dust in the ice core samples was noted by Thompson, indicating blowing soil from parched lands due to lack of vegetation (Mosley, 1991). The long-term climatological record from the Quellcaya ice cap provides evidence of precipitation events in the 6th and 7th centuries that were unkind to the Moche empire because of their intensive use of alluvial floodplains for their city as well as their corporate agricultural development. The Moche Demise High precIpItation events are etched in the ice caps and glaciers, however, did flooding actually occur in the Moche River Valley as a result of these events? The answer lies in the archaeological ruins and the floodplain deposits of the northern Peru coastal areas. These areas have been studied by leading scientists looking for the archaeological and anthropological records. The alluvial deposits have been analyzed by geologists. By bringing together the work and efforts of numerous disciplines, the circle from the physical scientist for the ice cap data and the meteorology to the cultural scientists and geologists for the evidence on and in the ground has been closed; the cause and effect has been established. The capital of the Moche empire at Huaca del Sol and Huaca de la Luna represented Moche Phase IV. It was heavily damaged by one of the early Quelccaya recorded EI Nino events, when flood water stripped away urban landscape and irrigation systems and then deposited several meters of sediment (Mosley, 1991). Archaeologists have identified the stripping, damage, and sedimentation deposits independent of the ice core data; it was the ice core information that closed the loop with regional precipitation data and estimated dates with a plus or minus accuracy of 20 years. The Moche capital was repaired with heightened platforms. Then, excavations show sand dunes encroached on the capital either before or after

PAGE 188

Wright and Dracup 165 the 31-year drought. However, the Moche Phase IV Period ended with the abandonment of the capital city. Phase V of the Moche empire started with the construction of a new capital at a place called Pampa Grande in the Lambayeque River valley, about 50 kIn from the sea. However, EI Nino flooding occurred at the new capital. Finally, the Moche empire came to an end, and Pampa Grande was abandoned and its inhabitants moved away. Archaeological excavation in Section D of Pampa Grande show over one meter of silty sand-flood deposits that completely buried structures. The flood that caused the deposit occurred close to the time of abandonment between 650 and 700 AD (Craig and Shimada, 1986). A new Sican culture emerged in the Pampa Grande Lambayeque region, along with a new capital and political center known as Batan Grande in the Rio La Leche valley. Thus, the Moche empire met its demise in the 7ili century with the best evidence pointing towards climatic induced extreme flooding that weakened the ability to govern through environmental stress (Mosley, 1991). Afterwards, the people of the area tended to adopt the styles of the Huari empire. The new Sican capital at Batan Grande was precariously close to the mouth of the Rio La Leche, probably so as to exercise control over the irrigation diversion head works. This development declined with much of the area being abandoned by about 1050 A.D. Field studies by Craig (1986) showed a massive 2.5-3.0 m of Quaternary alluvium in the stratagraphic columns at Batan Grande that he concluded had been caused by a single slack-water sedimentary event that overwhelmed the inhabitants of the area; the sedimentary event occurring during a pre 1000 A.D. flood. Summary The demise of the Moche empire at about 700 A.D. was likely caused by environmental stress in the form of great floods aggravated by periods of drought. The work of archaeologists and geologists have identified flooding sediment deposits and water damage to structures at the Moche capital cities at Cerro Bianco and Pampa Grande. On the other hand, glaciologists have cored ice caps to identify precipitation events that provide a degree of correlation to extreme precipitation years that would have caused floods. Research by meteorologists into the ENSO phenomena help provide the scientific basis for evaluating the cause and effect between the ice cap data, field evidence of great floods, and abandonment of the cities.

PAGE 189

166 FLOODING AND THE DEMISE OF THE MOCHE EMPIRE Much can be learned about ancient empires by interdisciplinary efforts and pooling of scientific knowledge. Knowing more about ancient civiliza tions and their problems can help modern societies to anticipate and cope with environmental extremes and resultant stresses. References Craig, A.K., and I. Shimada 1986 "EI Nino Flood Deposits at Batan Grande, Northern Peru." Geo archaeology: An International Journal 1(1): 29-38. Diaz, H.F., and G.N. Kiladis 1992 "Atmospheric Teleconnections Associated with the Extreme Phases of the Southern Oscillation." El Nino: Historical and Paleoclimatic Aspects of the Southern Oscillation. H.F. Diaz and V. Markgraf, eds. 2-28. Cambridge, England: Cambridge University Press. Dracup, J.A., K.S. Lee, and E.G. Paulson, Jr. 1980a "On the Statistical Characteristics of Drought Events." Water Resources Research 16 (2): 289-296. 1980b "On the Definition of Droughts." Water Resources Research 16 (2): 297-302. Mosley, M.E. 1992 The Incas and Their Ancestors: The Archaeology of Peru. London, England: Thames and Hudson. National Oceanic and Atmospheric Administration, National Geophysical Data Center 1986 "Tropical Ice Core Paleoclimatic Records: Quelccaya Ice Cap, Peru: A.D. 470 to 1984, Ohio State University Byrd Polar Research Center." Boulder, Colorado.

PAGE 190

Wright and Dracup Ortloff, C.R. and A.L. Kolata 1993 "Climate and Collapse: Agro-Ecological Perspectives on the Decline of the Tiwanaku State." Journal 0/ Archaeological Science 29: 195-221. 167 Shimada, I., C.B. Schaaf, L.G. Thompson, and E. Mosley-Thompson 1991 "Cultural Impacts of Severe Droughts in the Prehistoric Andes: Application of a 1,500-year Ice Core Precipitation Record." World Archaeology 22(3): 247-270. Thompson, L.G. and E. Mosley-Thompson, 1989 "One-Half Millennia of Tropical Climate Variability as Recorded in the Stratigraphy of the Quelccaya Ice Cap, Peru." Climatic Change in the Eastern Pacific and Western Americas. D. Peter son, ed. 15-31. Washington, D.C.: American Geophysical Union. University Corporation for Atmospheric Research 1994 "EI Nino and Climate Prediction." Reports to the Nation (Spring) 1994: 21-23. Walker, G.T. 1923 "Correlation in Seasonal Variations of Weather VIII: A Prelimi nary Study of World Weather." MEN. Indian Meteorology Department 24: 75-131. Wells, L.E. 1987 "An Alluvial Record of EI Nino Events from Northern Coastal Peru." Journal o/Geophysical Research 92 (13): 14,463-14,470. Wright, K.R., 1.A. Dracup, and 1.M. Kelly 1996 "Climate Variability Impact on the Water Resources of Ancient Andean Civilizations." Paper presented at the North American Water and Environment Congress, American Society of Civil Engineers, Anaheim, California.

PAGE 191

This page IS intentionally blank

PAGE 192


PAGE 193

This page IS intentionally blank

PAGE 194

PALEOHYDROLOGIC BOUNDS AND THE FREQUENCY OF EXTREME FLOODS Daniel R. Levish, Dean A. Ostenaa, and Daniel R.H. O'Connell Bureau of Reclamation, Denver Abstract An expeditious way to evaluate the probability of large floods for the safety of critical structures is to identify and assign ages to geomorphic surfaces adjacent to a stream that serve as limits for the paleostage of large floods over thousands of years. These paleo stage limits can then be put into a step-backwater model to calculate the maximum discharge that would not significantly inundate, and therefore significantly modify, a particular geomorphic surface. This maximum discharge, together with the age of the surface, forms a conservative limiting bound on flood discharge through time that is input for flood-frequency analysis. These bounds are not actual floods, but instead are limits on flood magnitude over a measured time interval. In this way, these bounds represent stages and discharges that have not been exceeded since the geomorphic surface stabilized. For dam safety, the critical issue is not the accurate estimation of a complete record of floods well within the operating capacity of the structure, but rather the frequency of floods that could challenge the operating capacity. The key issues are the accuracy of the frequency estimate of such large floods and the probability that the operational capacity of the dam will not be exceeded. Floods near the magnitude of the paleohydrologic bounds are direct indicators of the frequency of large floods that might compromise dam safety. Examples from paleoflood studies conducted for Causey Dam on the South Fork Ogden River, Utah, and Bradbury Dam on the Santa Ynez River, California, illustrate the utility of paleohydrologic bounds in flood frequency analysis. In both cases, flood frequency analysis based on the record of annual peak discharge estimates and extrapolation to the Probable Maximum Flood (PM F) leads to substantial overestimates of the frequency of large floods and to the conclusion that dam overtopping may be likely. In fact, in

PAGE 195

172 PALEOHYDROLOGIC BOUNDS AND FLOOD FREQUENCY both cases with the inclusion of paleohydrologic bounds in the flood frequency analysis, floods with a magnitude equivalent to spillway capacity are extremely low probability events. For the past several decades the estimated Probable Maximum Precipita tion (PMP) and the calculated Probable Maximum Flood (PMF) have been used as measures of hydrologic dam safety. However, both the PMP and PMF are hypothetical maximums and by definition have no associated probability. This absence of probability limits the utility of these hypothetical indices for risk -based dam safety decisions. In the western U. S., short historical and gauge records afford little support for the hypothetical PMF, and hydrologic indices based on the PMP are thwarted by the numerous complex and poorly understood assumptions required to turn rainfall into runoff. Paleohydrologic techniques are a means to directly assess the probability of extreme floods and test the validity of the PMF and PMP-based models. The results of paleoflood studies in California, Oregon, Utah, and Wyoming demonstrate that discharges with calculated annual probabilities of one in 10, 000 are in the range of 5 % to 25 % of the hypothetical PMF. Paleohydrologic Bounds A paleohydrologic bound is a time interval during which a given discharge has not been exceeded. In stable reaches, stage can be converted to discharge through hydraulic modeling. Both properties of the paleohydrologic bound, time and discharge, are independently determined from objective criteria in the field. This approach is appropriate for hazard assessment because it improves estimates of the frequency of large floods by using data that directly describe the largest floods. Although it is not necessary to develop evidence of specific paleofloods to define paleohydrologic bounds, it can be convement for illustration. The objective of studies of flood risk based on paleohydrologic bounds is to identify and assign ages to geomorphic surfaces adjacent to a stream that serve as limits for the paleo stage of large floods. These paleostage limits can then be entered into a step-backwater model to calculate the maximum discharge that would not significantly inundate, and therefore significantly modify, a particular geomorphic surface. The depth of significant inundation is calibrated based on the properties of the particular reach and comparison with the geomorphic impact of extreme floods (e.g., Baker and Costa, 1987). This maximum discharge, together with the age of the surface, forms a conservative bound on peak discharge through time for use in flood frequency analysis. These bounds are not actual floods, but instead are limits

PAGE 196

Levish, Ostenaa, and O'Connell 173 on flood magnitude over a measured time interval. In this way, these bounds represent stages and discharges that have not been exceeded since the geomorphic surface stabilized. Paleoflood hydrology includes the study of the geomorphic and strati graphic record of past floods (e.g., Baker, 1989; Jarrett, 1991). This record is a direct, long-term measure of the ability of a stream to produce large floods and may often be at least 10 to 100 times longer than the conventional record of annual peak discharge estimates. Paleohydrologic techniques offer a way to lengthen a short-term data record and, therefore, to reduce the uncertainty in hydrologic analysis (Jarrett, 1991). Obviously, this allows for a higher degree of assurance when making decisions regarding floods with long return periods. Paleoflood studies allow a long-term perspective that can put exceptional annual peak discharge estimates in context and assist in reconciliation of conflicting historical records. Most conventional estimates of the frequency of large floods are based on extrapolation from short records of annual peak discharge estimates, sometimes with the addition of historic information. Most magnitude estimates for extreme floods are made by extrapolation of the statistical model selected for flood-frequency to a given return period or annual probability, or by hypothetically maximizing rainfall-runoff models. Frequency estimates for maximized rainfall-runoff models are either arbitrarily assigned or are based on extrapolating the flood-frequency curve to the calculated discharge. No matter how many of these short-term records are statistically combined, they can never accurately characterize the probability of very infrequent floods because estimates of statistical confidence are directly related to the length of record. Further, due to short record length, many statistical distributions may fit the data, but the extrapolation to low probability floods is highly dependent on the choice of the distribution. Because each basin is unique, regionalizing or substituting space for time to compensate for short record length cannot completely substitute for the accurate characterization of the properties of a specific site or region, and may result in unwarranted confidence. If the record of annual peak discharge estimates contains an exceptionally large flood(s), this event(s) is usually assigned an unrealistically short return period, omitted from the frequency analysis, or "weighted" in some arbitrary fashion. Thus, any estimate of a flood with a recurrence greater than several hundred years that is based only on short-term record of annual peak discharge estimates or even long historic records of a few hundred years will have a large inherent uncertainty. Paleoflood hydrology offers a means of verifying return periods that are

PAGE 197

174 PALEOHYDROLOGIC BOUNDS AND FLoOD FREQUENCY many times longer than the length of the gauge or historic records (Costa, 1978). There is a long history of paleoflood hydrology, in a wide variety of settings throughout the world (Costa, 1986; Patton, 1987; Baker et aI., 1988). One widely used technique, slackwater studies, uses fine-grained sediment that accumulates in backwater areas to construct a detailed history of past floods (e.g., Patton et aI., 1979; Kochel and Baker, 1988). Early studies by Mansfield (1938) on the Ohio River and Jahns (1947) on the Connecticut River demonstrate another approach. They recognized that historic floods had overtopped sites not previously inundated in thousands of years. Lacking evidence of recent inundation, the age of a geomorphic surface is an estimator of the minimum return period of a flood that could inundate that surface (Costa, 1978; O'Connell et aI., 1996). Incorporation oflong-term paleohydrologic information in flood-frequency studies does not depend on being able to reconstruct the complete record of all past floods. Statistical techniques that can incorporate paleohydrologic bounds are a useful way to take advantage of paleohydrologic information (Stedinger and Cohn, 1986). In this way, it is not important if floods of a specified recurrence are not recorded or included in the frequency analysis. What is important is that limits on flood magnitudes over time intervals can be identified. Sensitivity analyses show that the addition of only one or two paleohydrologic bounds that span a range of hundreds to thousands of years have a significant impact on the shape of the flood-frequency curve (Ostenaa et aI., 1996a). The field expression of paleohydrologic bounds, stable geomorphic surfaces, are floodplains that have been abandoned due to stream incision. Once abandoned, their surface characteristics change with time. Two of the most easily recognized changes involve the modification of surface morphol ogy and the development of soil. Through time, slope processes and weathering mute the expression of surface irregularities related to fluvial erosion and deposition. Once a surface has stabilized, that is, it is no longer episodically overtopped, soils form in a predictable sequence (Birkeland, 1984). Disruptions in soil profiles and geomorphic features, such as eroded channels, that result from significant inundation by large floods are generally easily recognized. This is why these former floodplain surfaces are reliable indicators of flood stage through time. The limits of the surface define a maximum channel width through which a maximum discharge can be modeled. The ages associated with the geomorphic surfaces that form bounds for flood magnitude are almost always minimum ages because of the

PAGE 198

Levish, Ostenaa, and O'Connell 175 problems related to dating the precise time when a particular surface was abandoned. The result is an estimate of the maximum discharge during the minimum time interval since stabilization. These estimates are made even more conservative because through time, channels may downcut and erode laterally, resulting in apparently larger cross-sections and discharges. Therefore, a study goal is to locate stable reaches with the minimum channel capacity adjacent to geomorphic surfaces that place limits on paleostage through the reach. Paleohydrologic Bounds and Flood Frequency: Two Examples Examples from paleoflood studies conducted for Causey Dam on the South Fork Ogden River, Weber County, Utah (Ostenaa et aI., 1996b) and Bradbury Dam on the Santa Ynez River, Santa Barbara County, California (Ostenaa et aI., 1996a) illustrate the utility of paleohydrologic bounds for estimating the frequency of extreme floods. In both cases, conventional flood frequency analysis based only on the short record of annual peak discharge estimates with extrapolation to the PMF leads to substantial overestimates of the frequency of large floods. With the inclusion of paleohydrologic bounds in the flood frequency analysis, floods with a magnitude equivalent to spillway capacity are extremely low probability events at both dams. Causey Dam impounds water in the mountainous upper 210 km2 of the South Fork Ogden River basin. The da.m is a 258-meter-Iong, 66-meter-high earthfill embankment with an ungated spillway capacity of about 214 m3/s. The calculated thunderstorm PMF for Causey Dam has a peak discharge of more than 3,000 m3/s, with the calculated threshold for dam overtopping of 677 m3/s. Based on standard engineering flood frequency analysis (e.g., NRC, 1985), a discharge equivalent to the estimated threshold overtopping discharge has a return period on the order of 2,000 to 30,000 years. Significant overtopping of Causey Dam could result in dam failure, and dam failure would result in substantial and unacceptable consequences down stream. Downstream from Causey Dam, the South Fork Ogden River is characterized by two groups of Holocene geomorphic surfaces that form paleohydrologic bounds. Ages for the bounds are based on the geomorphol ogy and stratigraphy of these surfaces and 19 radiocarbon ages. The discharges associated with these bounds are calculated from step-backwater modeling in a stable reach 6 kilometers downstream from the dam. The "Holocene 2" paleohydrologic bound is formed by a group of surfaces that

PAGE 199

176 PALEOHYDROLOGIC BOUNDS AND FLoOD FREQUENCY have not been significantly overtopped in 2500 years. Based on step backwater modeling, a discharge of 115 m3/s would significantly modify this surface and therefore is the discharge value for the paleohydrologic bound. The "Holocene 1" paleohydrologic bound is formed by a group of surfaces that have not been significantly overtopped in the last 400 years. A discharge of 70 m3/s would modify these surfaces based on step-backwater modeling. Bradbury Dam is a 1,020-meter-Iong, 85-meter-high embankment that impounds water in the mountainous upper 670 km2 of the Santa Ynez River basin. Bradbury Dam has a gated spillway with a capacity of 4,533 m3/s. The calculated PMF for Bradbury Dam is 13,060 m3/s. Based on standard engineering flood-frequency techniques (e.g., NRC, 1985), the PMF has an extrapolated return period as frequent as less than 100 years. If the PMF were possible, this flow would overtop the dam and could result in dam failure, causing substantial and unacceptable consequences downstream. Between Bradbury Dam and the town of Lompoc 65 km downstream, there are two Holocene surfaces that form useful paleohydrologic bounds. Ages for these bounds are based on the geomorphology and stratigraphy of these surfaces and 17 radiocarbon ages. The discharges associated with these bounds result from step-backwater modeling in stable reaches 2 km and 55 km downstream of the dam. The "tl" paleohydrologic bound is formed by a group of surfaces that have not been significantly overtopped in 2900 years. Based on step-backwater modeling a discharge of 2,550 m3/s would significantly modify this surface and therefore is the discharge value for the paleohydrologic bound. The "fp2" paleohydrologic bound is formed by a group of surfaces that have been significantly overtopped once in the last 700 years. A discharge of 1980 m3/s would modify these surfaces based on step backwater modeling. Based on historical information a third bound can be constructed that spans from 1862 to 1907 at a discharge of 1275 m3/s. For flood frequency calculations, the information from the paleohy drologic bounds is combined with the record of annual peak discharge estimates. The relative amounts of time spanned by the paleohydrologic bounds and the record of annual peak discharge estimates for both the South Fork Ogden River and the Santa Ynez River is shown on Figure 1. To calculate flood-frequency statistics, the maximum likelihood (MLH) method of Stedinger and Cohn (1986) and Stedinger et al. (1988) is modified and incorporated into a Bayesian approach (Tarantola, 1987) (O'Connell et al., 1996). For both the South Fork Ogden River and the Santa Ynez River, the impact of the paleohydrologic bounds is dramatic. For the South Fork Ogden River, the analysis indicates that a flow of spillway capacity has a probability of much less than one in 10,000, with an extrapolated probability of less than

PAGE 200

Levish, Ostenaa, and O'Connell 120 100 (i) M 80 E Q) Ol IT; 60 .c u U) 0 1-40 ro Q) (L 20 0 0 3000 2500 ME 2000 Q) Ol w -5 1500 U) o 1-g; 1000 (L 500 1000 2000 Time Before 1995 (years) ,-11 Paleohydrologic Bound 1000 2000 Time Before 1994 (years) Figure 1 177 Ai 1 I \ 1 j 1 ] J 3000 B 3000 Comparative time spanned by paleohydrologic bounds and the record of annual peak discharge estimates for South Fork Ogden River at Causey Dam (A) and Santa Ynez River at Bradbury Dam (B). For the South Fork Ogden River there are no historic floods near the paleohydrologic bound. For the Santa Ynez River the 1969 flood exceeds the fP2 paleohydrologic bound. These plots illustrate that paleohydrologic bounds provide significantly more information about low probability floods than the record of annual peak discharge estimates.

PAGE 201

178 PALEOHYDROLOGIC BOUNDS AND FLOOD FREQUENCY Table 1. Mean Annual Probability and Return Period of Discharges Exceeding the Capacity of the Bradbury Dam Spillway (4533 m3/s) Data Set Mean Annual Prob-Return Period ability (years) Gauge 5.74 x 10-4 1,740 Historic + Gauge 1.43 x 10-4 7,000 Paleoflood + Historic 1.57 x 10-7 >6,000,000 + Gauge one in 500,000. This indicates that overtopping of the dam is an extraordi narily unlikely event. For the Santa Ynez River the incremental advantage of adding more information can be demonstrated by calculating flood-frequency statistics from various data sets. The addition of paleohydrologic bounds has the greatest impact on the flood frequency calculations (Table 1). Table 1 illustrates the difference in the conclusion drawn from different sets of data. Once again, for the Santa Y nez River it shows that a discharge equivalent to spillway capacity of Bradbury Dam is a very remote event with an extrapo lated probability of less than one in 6,000,000. Paleohydrologic bounds influence flood frequency calculations by extending the length of record. Inferences about low probability floods based only on short gauge and historic records (less than 150 years in the western U.S.) depend on assumptions of the statistical distribution chosen to portray flood frequency (e. g., 0' Connell et al., 1996). However, these short -term records contain no information about the long term behavior of floods. In fact, many gauge and historic records are hampered by trapping a low probability or long return period event in a short record. This adds significant bias to flood frequency estimates and in many instances in the western U. S. leads to an over estimate of flood magnitude for a particular probability or return period. Paleohydrologic Bounds and the Calculated Probable Maximum Flood The Probable Maximum Flood (PMF) has been used as a standard for hydrologic analyses in dam safety for several decades (NRC, 1985). As originally defined, the PMF has no return period. However, this definition

PAGE 202

Levish, Ostenaa, and O'Connell 179 is not practical for dam safety decisions based on risk. As a practical matter, the PMF has often been arbitrarily assigned a return period of 10,000 to 1,000,000 years at the upper and lower confidence limits for flood frequency analysis (e.g., NRC, 1985). Paleoflood studies are a basis for testing whether the calculated PMF and the associated extrapolated return period are realistic. Because the fluvial geomorphology and stratigraphy of floodplains adjacent to streams are recorders of the most extreme floods, paleoflood records should contain extreme floods that are a large percentage of the PMF, if such floods are physically possible. The shorter the estimated return period assigned to the PMF, the more likely it becomes that such large floods should be included in paleoflood records that are thousands of years in length. Considering the number of drainage basins present in an area the size of the western U. S., if there actually have been floods comparable to the hypothetical PMF, the numerous multi-thousand-year paleoflood records present along western streams are likely to record mUltiple PMF-scale floods. The paleohydrologic bounds from the South Fork Ogden River and the Santa Ynez River are only a small percentage of the calculated PMF for Causey and Bradbury Dams. Data from other Reclamation paleoflood studies in the western u.S. show a similar relationship to calculated PMF estimates (Table 2). It is clear that in a variety of hydro meteorological settings, the paleoflood record does not validate floods as extreme as the PMF, nor do the paleoflood data validate estimates of PMF return period in the range of 10,000 to 1,000,000 years. Rather, the paleoflood data imply a potential upper limit for flood magnitude that is substantially smaller than implied by PMF calculations. The data in Table 2 indicate that in the western United States, peak discharges with an extrapolated return period of 10,000 years may be as little as 5 % to 25 % of the calculated PMF. These results have substantial impact when incorporated into dam safety decisions or criteria based on risk. Conclusion The most reliable way to obtain probability estimates of extreme floods, floods with return periods of thousands of years, is to study the geomorpho logic and stratigraphic record of extreme floods. Paleoflood hydrology is an event-based method for extending the length of the flood record in order to make realistic estimates of the probability of extreme floods. An expeditious way to gain paleoflood information is through the use of paleohydrologic bounds. For guiding hydrologic dam safety decisions, comparable levels of

PAGE 203

180 PALEOHYDROLOGIC BOUNDS AND FLOOD FREQUENCY Table 2. Calculated Probable Maximum Flood versus 10,OOO-Year Flood Estimated From Paleoflood Studies Location Drainage Estimated Probable 10,000 year Basin Area 10,000 year Maximum paleoflood (km2 ) paleoflood, Flood Peak as percent97.5 percenDischarge age ofPMF tile, dis-(m 3/s) charge (-m3/s) South Fork 210 150 3075 5 Ogden River, Utah Santa Ynez 1080 2550 l3,060 26 River, CA Ochoco 764 285 4785 6 Creek, OR Crooked 6825 1100 7225 15 River, OR confidence cannot be obtained from analysis of short-term records of annual peak discharge estimates and historic information alone. Compared to conventional frequency analyses, incorporation of paleoflood data provides high assurance that the spillway capacity of Causey and Bradbury Dams will not be exceeded even at long return periods. For many streams in the western U. S., paleoflood information does not validate PMF discharges derived from rainfall-runoff modeling, nor does it validate the range of PMF frequency commonly used in hydrologic risk decisions. Baker, V.R. 1989 References "Magnitude and Frequency of Palaeofloods." In Floods: Hydrological. Sedimentological. and Geomorphological

PAGE 204

Levish, Ostenaa, and O'Connell 181 Implications, edited by K. Beven and P. Carling, 171-183, New York: John Wiley and Sons. Baker, V.R. and Costa, J.E. 1987 "Flood Power." In Catastrophic Flooding, edited by L. Mayer am D. Nash, 1-21, Boston: Allen and Unwin. Baker, V.R., Kochel, R.C., and Patton, P.C., editors 1988 Flood Geomorphology. New York: John Wiley and Sons. Birkeland, P.W. 1984 Soils and Geomorphology. New York: Oxford University Press. Costa, J.E. 1978 1986 Jahns, R.E. 1947 Jarrett, R.D. 1991 "Holocene Stratigraphy in Flood Frequency Analysis." Water Resources Research 14: 626-632. "A History of Paleoflood Hydrology in the United States." Eos (Transactions of the American Geophysical Union) 67: no. 17, 425, 428-430. Geologic Features of the Connecticut Valley, Massachusetts, as Related to Recent Floods. U.S. Geological Survey Water-Supply Paper 996. Paleohydrology and its Value in Analyzing Floods and Droughts. U.S. Geological Survey Water-Supply Paper 2375: 105-116. Kochel, R.C., and Baker, V.R. 1988 "Paleoflood Analysis Using Slackwater Deposits", In Flood Geomorphology. Edited by V.R. Baker, R.C. Kochel, P.C. Patton, 357-376, New York: John Wiley and Sons. Mansfield, G.R. 1938 FloodDeposits of the Ohio River, January-February, 1937, a Stu4' of Sedimentation. In U.S. Geological Survey Water Supply Paper 838: 693-736.

PAGE 205

182 PALEOHYDROLOGIC BOUNDS AND FLoOD FREQUENCY NRC (National Research Council) 1985 Safety of Dams, Flood and Earthquake Criteria. Washington, D.C.: National Academy Press. Ostenaa, D.A., Levish, D.R., and O'Connell, D.R.H. 1996a Paleoflood Study for Bradbury Dam, Cachuma Project, Califomil. Seismotectonic Report 96-3. Denver, Colorado: U.S. Bureau of Reclamation, Seismotectonic and Geophysics Group. Ostenaa, D.A., Levish, D.R., O'Connell, D.R.H., and Cohen, E.A. 1996b Draft Paleoflood Study for Causey and Pineview Dams, Weber Basin and Ogden River Projects, Utah. Seismotectonic Report 96-6. Denver, Colorado: U.S. Bureau of Reclamation, Seismo tectonic and Geophysics Group. Patton, P. C. 1987 "Measuring Rivers of the Past: A History of Fluvial Paleohydrol ogy." In The History of Hydrology, edited by E.R. Landa, E.R., and S. Ince, History of Geophysics 3,55-67, Washington, D.C.: American Geophysical Union. Patton, P.C., Baker, V.R., and Kochel, R.C. 1979 "Slack Water Deposits: A Geomorphic Technique for the Irterpre tation of Fluvial Paleohydrology." In Adjustments of the Fluvial System, edited by D.D. Rhodes, and G.P. Williams, 225-253, Dubuque, Iowa: Kendal-Hunt. Stedinger, J.R., and Cohn, T.A. 1986 "Flood Frequency Analysis with Historical and Paleoflood Information." Water Resources Research 22: 785-793. Stedinger, LR., Surani, R., and Therival, R. 1988 The MAX Users Guide. Ithaca, New York: Department of Environmental Engineering, Cornell University. Tarantola, A. 1987 Inverse Problems Theory: Methods for Data Fitting and Model Parameter Estimation, New York: Elsevier.

PAGE 206

BAYESIAN FLOOD FREQUENCY ANALYSIS WITH PALEOHYDROLOGIC BOUNDS FOR LATE HOLOCENE PALEOFLOODS, SANTA YNEZ RIVER, CALIFORNIA Daniel R. H. O'Connell, Daniel R. Levish, and Dean A. Ostenaa U. S. Bureau of Reclamation Denver, Colorado Abstract Paleohydrologic bounds demonstrate that peak discharges of about 70,000 ft3/s and 90,000 ft3/s had not been exceeded on the Santa Ynez River in 700 years and 2,900 years, respectively. These paleohydrologic bounds are combined with gage and historic data in a Bayesian approach to estimate flood-frequency probabilities. The Bayesian analysis uses likelihood functions that incorporate both parameter and data (discharge and geologic age) uncertainties. High-speed workstations make it possible to calculate parameter and peak discharge frequency probabilities using systematic parameter-space searches and direct numerical integration. Bayesian peak discharge frequency calculations demonstrate the value of the paleohydrologic bounds for a hydrologic risk analysis of Bradbury Dam. If only gage and/or historic data are available, an inescapable conclusion is that a flow exceeding spillway capacity (160,000 ft3/S ) is likely to occur within a lO,OOO-year period. Adding the paleohydrologic bounds to the Bayesian analysis shows that the probability of a flow exceeding spillway capacity within a lO,OOO-year period is less than 1 in 50,000. The Bayesian analysis demonstrates the substantial statistical gain the paleohydrologic bounds provide, because they put large historic discharges in their proper, long-term contexts and substantially reduce the range of possible discharges associated with long return periods.

PAGE 207

184 BAYESIAN FLooD FREQUENCY ANALYSIS Introduction This paper describes a Bayesian flood frequency analysis that incorporates paleohydrologic bounds information to assess the frequency and magnitude of large floods for Bradbury Dam on the Santa Ynez River, Santa Barbara County, California. This study was conducted as part of a paleoflood study (Ostenaa et aI., 1996) intended to reduce the range of uncertainty in extrapolated frequency estimates of extreme floods on the Santa Ynez River and to provide additional limits, beyond the historical record, for deci sion-making on hydrologic dam safety issues affecting Bradbury Dam. The previously calculated Probable Maximum Flood (PMF) with a peak discharge of 414,000 ft3/S overtops Bradbury Dam (U.S. Bureau of Reclamation USBR) 1993). A new PMF has a peak discharge of 460,900 ft3/S (USBR, 1995). The estimated recurrence interval for the PMF, between the upper and lower 90 % confidence limits, based on extrapolation from 41 years of annual peak discharge records, extends from less than 100 years up to 1 million years (USBR, 1993). The high uncertainty in the extrapolated return period of the PMF results in a very large range of justifiable expenditures for hydrologic dam safety modifications. Abandoned floodplains or stream terraces that flank the Santa Ynez River range in age from several hundred to more than tens of thousands of years. Terrace surfaces are underlain by stream-transported floodplain sediment, and therefore are quite easily modified by shallow flood inundation. Because these surfaces are easily modified, they are reliable recorders of maximum flood stage through time. If ages can be derived for preserved and/or flood modified surfaces, the surfaces become conservative datums for the magnitude of large floods. Likewise, the absence of features indicative of significant inundation is positive evidence of nonexceedence of a specific, limiting flood stage over the time spanned by the surface. A paleohydrologic bound is a time interval during which a given discharge has not been exceeded (Ostenaa et aI., 1996; Levish et aI., 1996). Our approach focuses on defining nonexceedence bounds of time and discharge rather than on construction of a detailed record of past floods. This is accomplished by identifying and assigning ages to geomorphic surfaces that serve as limits for the paleostage of large floods. These bounds are not actual floods, but instead are limits on flood magnitude over a measured time interval and represent stages and discharges that have not been exceeded since the geomorphic surface stabilized. Through step-backwater modeling, stage can be converted to discharge, so that in the flood frequency analysis,

PAGE 208

O'Connell, Levish, and Ostenaa 185 a bound is set for a time interval during which a specific discharge has not been exceeded. This approach is appropriate for hazard assessment because long-term paleohydrologic information provides powerful bounds for estimates of flood frequency (Stedinger and Cohn, 1986) and because the data are a direct description of the likelihood of the largest floods along a stream. Including long-duration paleohydrologic bounds in flood frequency calculations significantly reduces predicted peak discharge uncertainties at long return periods. It is not necessary to develop evidence of specific paleofloods to define limits for paleostage, although it is often convenient for illustration. For dam safety, the critical issue is not the accurate estimation of a complete record of floods well within the operating range of the structure, but rather the frequency of floods that could challenge the operational capacity of the structure. The key issues are the accuracy of the frequency estimates of such large floods and the probability that the operational capacity of the dam will not be exceeded. Floods near the magnitude of the paleohydrologic bounds are direct indicators of the likelihood of large floods that might compromise dam safety. Method The primary goals of this statistical analysis are to determine the annual risk (probability) of a peak discharge exceeding the maximum spillway capacity of Bradbury Dam and to derive flood frequency statistics. Another goal is to quantify the value of incorporating historical and paleohydrologic data into the statistical analysis of flood frequency. A Bayesian methodology (Tarantola, 1987) and likelihood functions modified from Stedinger and Cohn (1986) are used to incorporate data and parameter uncertainties. Parameter and flood frequency likelihoods and probability intervals are calculated directly by numerical integration. Systematic parameter-space searches provide the most powerful method to determine flood frequency probabilities, provided by high-speed workstations. A systematic search of a parameter space of four or less can be completed without resorting to Monte Carlo methods of statistical sampling and integration. This approach is used with annual peak discharge, historical, and paleohydrologic data to develop flood frequency probabilities, to estimate annual probabilities of a peak discharge exceeding spillway capacity, and to quantify the statistical value of incorpo rating historic and paleohydrologic data into the analysis. Several other methods are traditionally used to develop flood frequency estimates from discharge records. The weighted-moments technique presented

PAGE 209

186 BAYESIAN FLOOD FREQUENCY ANALYSIS in Bulletin 17B (U.S. Water Resources Council, 1981) is often used with annual peak discharge records, but incorporation of historical and paleohydrologic data is awkward. Bulletin 17B provides procedures to calculate flood frequency confidence intervals. However, while it accounts for uncertainties in mean and variance due to sampling errors, it ignores errors in skew that can be important (Chowdhury and Stedinger, 1991). Regional index flood methods pool annual peak discharge estimates from a region to reduce peak discharge quantile standard errors (Stedinger et al., 1993; Pitlick, 1994). However, errors are only reduced if the physiographic features that influence flood characteristics are homogenous over a region and the hydrometeorologic events are independent. The maximum likelihood (MLH) method of Stedinger and Cohn (1986) and Stedinger et al. (1988) provides a formalism to calculate flood frequency probabilities by combining information from annual peak discharge records, and historical and paleohydrologic sources that can encompass long time periods. This method can yield robust "global like" MLH parametric distribution estimates, but the "local" error analysis assumes the linearized covariance calculated at the MLH model position accurately describes the "global" model parameter distribution. The MLH method's "local" error analysis may not produce realistic uncertainty estimates in some situations and only methods that evaluate probabilities over a "global" model distribution space can assure robust uncertainty estimates. The Bayesian approach used here explicitly acknowledges that the parameters and data are never perfectly known. Both parameter and data uncertainties are incorporated into risk and probability interval estimates of flood frequency. It directly measures how well the data constrain model parameters. The Bayesian paradigm is a special case of the more general information theory of Shannon (1948), Tarantola and Valette (1982), and Tarantola (1987). These approaches quantitatively rank how well particular models fit data sets. For example, the value of each data point is somewhat uncertain and the ranking or goodness of fit of each possible frequency function is proportional to how often the frequency function predicts data values close to the observed data (high likelihood or "good fit") or predicts values far from the observed values (low likelihood or "poor fit"). The Bayesian approach uses a "global" parameter integration grid as outlined below in a systematic quantitative framework to identify what ranges of frequency functions are consistent with the data at various probabilities. By selecting broad probability intervals, conservative evaluations of risk are obtained.

PAGE 210

O'Connell, Levish, and Ostenaa 187 Stedinger and Cohn (1986) and Stedinger et al. (1988) developed likelihood expressions for combining several data types that are modified here to incorporate data measurement uncertainties. Integrals are added to their log-likelihood expressions to include discrete probability density functions (pdf's) representative of peak discharge and geologic age uncertainties. Details of the modified likelihood expressions are provided in Ostenaa et al. (1996). The approach of Stedinger (1983) is used to estimate peak discharge quantiles and the annual probability of a peak discharge exceeding spillway capacity. Developing Bayesian flood frequency statistics is a two-part process. In the first part, a parameter grid is constructed that spans the nonzero likelihood portion of parameter space for a particular choice of the frequency function. The Log Pearson III (LP3) frequency function is used to allow direct comparison with other methods often used to develop flood frequency statistics (e.g., Bulletin 17B, U.S. Water Resources Council, 1981). Larger grids with coarse parameter spacing are used to determine the nonzero likelihood regions of the parameter space, and then a smaller, finely sampled parameter grid is used for the final calculations of the conditional posterior probability density function. Parameter statistics are calculated from this density function to provide MLH values of the parameters, and numerical integration is used to provide marginal density functions for individual parameters or combinations of parameters. In the second part of the analysis, the posterior distributions of peak discharges for various quantiles and annual probabilities of a peak discharge exceeding the spillway capacity are calculated. This approach is ideal for risk-based hypothesis testing because it provides probability estimates of peak discharges of interest for dam safety decision making. Data Bayesian frequency analyses are performed using three data sets (Ostenaa et al., 1996). The first data set, paleohydrologic, includes all annual peak discharge, historic, and paleohydrologic data representing about 2,920 years of observation. The paleohydrologic data include two bounds, t1 and fp2, that provide discharge limits of 90,000 and 70,000 ft3/s, that have not been exceeded in 2,920 and 700 years, respectively (Table 1). Historic data from the Lompoc area provide a limit on peak discharge since 1862 relative to the size of the large flood in 1907. The 57 year peak discharge record includes an exceptional flood in 1969 (peak of record). A second data set, historic, is derived by deleting the paleohydrologic data from the complete data set to

PAGE 211

188 BAYESIAN FLOOD FREQUENCY ANALYSIS provide 132 years of annual peak discharge and historic data. The third data set consists of the most recent 57 years of annual peak discharge estimates from the gauge just downstream from the Bradbury Damsite (1935-1952), and the adjusted sum of the gauges upstream of Lake Cachuma (1953-1993). Complete Bayesian frequency analyses are performed for each data set to evaluate the incremental value of acquiring historic and paleohydrologic data. The model space for LP3 is discretized on 150x150x150 grids represent ing 3.375x106 LP3 models for each of the three data sets. The zeta range is limited to values of exp(zeta) in the range defined by an envelope curve, 114,000 fi3/s (Meyer, 1994), and the PMF, 414,000 fi3/s (USBR, 1993). Wider ranges of alpha and beta parameters are used to refine the grid to encompass all relative likelihoods of 10-7 or larger. These parameter ranges define a uniform boxcar for the parameter prior probabilities. An integration interval of 20 fi3/s is used to discretize the frequency functions. The probability distribution parameters for paleohydrologic bounds, time before 1994 observations (time intervals of paleohydrologic bound non exceedence), and peak discharges within a range are shown in Tables 1 and 2, respectively. The "End Prob" columns in Tables 1 and 2 are the likelihoods at the minimum and maximum values relative to the MLH values. The times in Table 1 are total years prior to 1994. The 1862 historic bound and 1907 peak discharge estimates are assigned weak central tendencies (large "End Prob"s) because they are poorly constrained_ Conversely, the tl, fp2, and 1969 peak discharge estimates have strong central tendencies (small "End Prob"s), but the triangular distribution still implies a weaker central tendency (more probabilities at the extremes) than a Gaussian or exponential distribution. Results Starting with the paleohydrologic data set results, the MLH LP3 parameters are identified by searching the parameter grid for the MLH value. The MLH frequency function (alpha=3.95, beta=-1.0775, zeta = 11.644) is evaluated for goodness of fit to the observed peak discharge data. The method of Hirsch and Stedinger (1987) is used to derive the Weibull plotting positions of the observed data. The Filliben probability-plot correlation test yields a correlation value of 0.988 indicating that the observed data are consistent with the MLH LP3 distribution (Vogel and McMartin, 1991). A total of 107,518 LP3 models have relative likelihoods of 10-7 or greater. The paleohydrologic MLH flood frequency estimates and 0.95 probability region limits (lower 2.5 percentile to upper 97.5 percentile) are plotted in

PAGE 212

Bound t1 fp2 18621906 Table 1. Paleohydrologic Bounds, Durations, and Relative Probabilities Min End ( fe/s) Prob 70,000 0.05 60,000 0.05 37,000 0.80 -Year 1907 1969 MLH Max End Min MLH (ft3/s) ( fe/s) Prob (yr) (yr) 90,000 104,000 0.05 2800 2920 70,000 81,000 0.05 570 700 45,000 50,000 0.80 45 45 Table 2. Peak Discharges Within a Range and Relative Probabilities Min End MLH Max End (ft3/s) Prob (fels) (ft3/s) Prob 48,000 0.80 55,000 63,000 0.80 70,000 0.05 88,000 110,000 0.05 Max (yr) 3190 820 45 I End Prob I 0.8 0.8 I 1.0 I o g <1:> ,::::: -.:: -: I:l.. -00 \0

PAGE 213

190 BAYESIAN FLoOD FREQUENCY ANALYSIS Figure 1 along with the upper 97.5 percentile limit from the historic and annual peak discharge data set calculations. The upper 97.5 percentile limit for the historic and annual peak discharge data demonstrate the weak constraints on long return period discharges afforded by such short time samples of peak discharge. In contrast, the paleohydrologic bounds constraints produce a much more contracted upper 97.5 percentile. Note, that neither the weighted-moment method of Bulletin 17B nor the MLH error estimation method of Stedinger et al. 's (1988) MAX program would provide these asymmetric (but more realistic) probability limits. These probability limits are more realistic because they incorporate the global information about the model's nonlinear probability distribution instead of local linearized estimates of model covariance at the MLH model position. 300,000 ,...-................. ........,r--......... "T""I"' .......... -,........,.. .................... --. ........ "rr"I"TI"l'"O Annual Peak Discharge Historic 97.5 Percentile :0" 200,000 SpillwaY' Capacity (160,000 cfs) Q) O"l '-m .!:: u CIl .::t:. m 8: 100,000 t1 Bound l...,.-:;L..,..o<..-:::lr.::::----" fp2 Bound Bound MLH LP3 Model Paleohydrologic 2.5 Percentile 1 0.01 0.001 0.0001 (1) (100) (1000) (10,000) Annual Probability (Return Period (years)) Figure 1. Flood frequency for the Santa Ynez River at Bradbury Dam. Annual peak discharges are shown with the horizontal lines for plotting position ranges and vertical lines for 20 measurement uncertainties. Paleohydrologic bounds are boxes denoting geologic age and discharge uncertainties. The 0.95 flood-frequency probability region using all data is shown in grey.

PAGE 214

O'Connell, Levish, and Ostenaa 191 The substantial difference in the predictions of flood frequency by the three data sets can be understood in tenns of each data set's plotting position for the peak discharge of record that occurred in 1969. With paleohydrologic data, the plotting position of the 1969 discharge is 1,390 years (Table 1); 191 years for the historic data, and 118 years for the annual peak discharge data. Moving the plotting position of the 1969 discharge to the range of 100-200 years in Figure 1 explains the much wider upper probability limits of the annual peak discharge and historic data sets relative to the paleohydrologic data. A much steeper LP3 slope is required to fit the discharges if the 1969 discharge plotting position is less than 200 years. The difference between the paleohydrologic data predictions of flood frequency and the predictions of the historic and annual peak discharge data can explained by the inability of annual peak discharge and historic data to place the largest 1969 discharge in its proper context (plotting position). The paleohydrologic bounds revealed that the 1969 flood was a rare event, larger than any flood in the past 700 years. The mean annual probability (and return period) estimates of a discharge exceeding spillway capacity from each of the three data sets (paleohydrologic: 1. 57xlO-7/yr, (6.35xI06 yr), historic: 1.43xlO-4/yr, (7000 yr), and annual peak discharge: 5.76xl0-4/yr, (1740 yr)) reveal the consequences of not placing a rare flood (1969) in its proper context. It is clear that given only a annual peak discharge or historic record, it would be necessary to conclude that a peak discharge exceeding spillway capacity probably has a return period of less than 10,000 years and might be as frequent as a 1,000 year event. However, the paleohydrologic data clearly demonstrate that a peak discharge exceeding spillway capacity is extremely unlikely (0.004 probabil ity for return periods to be less than 100,000 years). The paleoflood data clearly provide a sound basis to interpolate flood frequencies to the range of 1 in 10,000 years. However, extrapolation beyond the range of the paleoflood data is imprudent because there is no physical basis to support assuming the LP3 distribution for more rare events and because data uncertainties and biases make extrapolations highly uncertain (Kuczera, 1996). Therefore, it is statistically more meaningful to evaluate the residual risk associated with a peak discharge exceeding spillway capacity with a 1 in 10,000 year probability or greater (e.g. within the range of actual observed data) than to focus on extrapolations to mean annual probabilities. The residual risks is easily calculated by totaling the probabilities associated with predictions of annual probabilities of discharges exceeding spillway capacity of 1 in 10,000 or larger. The total probability of a peak discharge exceeding spillway capacity with an annual probability of 1 in 10,000 or

PAGE 215

192 BAYESIAN FLoOD FREQUENCY ANALYSIS larger is less than 1 in 50,000. This represents the residual risk of such an event in the context of interpolating the available data. Conclusions The Bayesian flood-frequency analysis demonstrates that the single most important factor affecting the statistical conclusions is length of record, consistent with the results of Frances et al. (1994). Paleohydrologic bounds can provide direct limits on peak discharge spanning several thousands of years. Sensitivity analyses of the MLH estimates of peak discharge demon strate that these estimates are not sensitive to radiocarbon age or discharge uncertainties, or to potentially "missed" peak discharges (Ostenaa et al., 1996). The paleohydrologic bounds afford the capability to estimate peak-discharge frequency in the 1,000to 10,000-year return period range (Figures 1 and 2), with substantially smaller uncertainties than extrapolated estimates from much shorter records of annual peak discharge estimates. This analysis demonstrates that the annual probability of a flow exceeding the spillway capacity of Bradbury Dam, 160,000 ft3/s, is extremely low. Traditional flood-frequency procedures have focused on "best fitting" distributions that have infinitesimal total probability as illustrated by the thin line of the MLH flood-frequency estimate in Figure 1. For risk analyses, we need to consider the flood-frequency estimates with various total probabilities that are strongly dependent on data uncertainties and biases. Thus, it is irrelevant what method is used to construct "best fitting" distributions (MLH, various moment methods, etc.) because they all produce a deterministic result that is completely inadequate for probabilistic risk analyses. A complete probabilistic integration approach (grid or Monte Carlo), like the Bayesian analysis presented here, is necessary to produce sufficient flood-frequency information (probability intervals) for probabilistic risk analyses. The numerical analyses described here can be calculated in about a day using inexpensive PC computers and readily-available software. Computational demands are not an impediment. Typical products needed for risk analyses are discharges associated with very rare 1 in 10,000 year) floods. Traditional approaches have extrapo lated beyond the available data, often to the Probable Maximum Flood (PMF) using an arbitrarily assigned range of annual probabilities. These flood frequency extrapolations using arbitrary frequency functions are a physically baseless statistical subterfuge that provides no reliable information. Figure 2 shows that constructions like the Probable Maximum Flood (PMF) have no place on flood-frequency plots. PMF estimates may be useful for design

PAGE 216

O'Connell, Levish, and Ostenaa (j) o 500,000 400,000 -; 300,000 2> ro ..c u (f) 76 200,000 Q) n.. 100,000 Spillway Capacity Physical Understanding Gage Record I Large Bias I Reduced Bias Large -.l Uncertainies I aT ,II' f t t: I + Additional Assurance 10 0 10 -2 10 -3 10 -4 10 -5 10 -6 Annual Probability 193 Figure 2. Observed flood-frequency data for the Santa Ynez River at Bradbury Dame as in Figure 1. considerations, but it is inappropriate to assign annual probabilities to the PMF, particularly by extrapolation beyond the available data. Instead, it is necessary to develop a physical understanding of flood processes to quantify discharge-frequency behavior for extremely rare events. Statistics only have significance if there is physical understanding and no amount of mathe matistry can substitute for physical understanding (Klemes, 1987, Bardsley, 1994). Bayesian flood-frequency analyses incorporating paleohydrologic bound information provide information necessary to establish appropriate dam safety priorities when evaluating an inventory of structures. In contrast, arbitrary flood-frequency extrapolations can actually compromise risk analyses. For instance, PMF analyses often suggest that hydrologic deficiencies exist at a

PAGE 217

194 BAYESIAN FLOOD FREQUENCY ANALYSIS structure. Using PMF flood-frequency extrapolations in this manner obscures the actual, shorter term risks and could in fact increase public risk by biasing the scheduling of modifications for an inventory of structures. Bayesian flood-frequency analyses with paleohydrologic data spanning thousands of years provides a consistent basis for risk based decision making. This information should be coupled with realistic rainfall-runoff modeling using three-dimensional digital elevations models and modern finite-element computational techniques for fluid flow that include nonlinear behavior (Zienkiewicz and Taylor, 1991) to further quantify the hazards associated with very rare precipitation events. References Bardsley, W. E. 1994 "Against Objective Statistical Analysis of Hydrological Extremes." Journal of Hydrology 162:429-431. Chowdhury, J. U. and J. R Stedinger 1991 "Confidence Intervals for Design Flood with Estimated Skew Coefficient. Journal of Hydraulic Engineering 117: 811-831. Frances, F., J.D. Salas, and D.C. Boes 1994 "Flood Frequency Analysis with Systematic and Historical or Paleoflood Data Based on the Two-parameter General Extreme Values Models," Water Resources Research 30: 1653-1664. Hirsch, R.M., and J .R. Stedinger 1987 "Plotting Positions for Historical Floods and Their Precision." Water Resources Research 23:715-727. Klemes, V. 1987 "Hydrological and Engineering Relevance of Flood Frequency Analysis" in Hydrologic Frequency Modeling, edited by V.P. Singh. Dordrecht, The Netherlands: D. Reidel. Kuczera, G. 1996 "Correlated Rating Curve Error in Flood Frequency Inference." Water Resources Research 32:2119-2127.

PAGE 218

O'Connell, Levish, and Ostenaa 195 Meyer, RW. 1994 Potential Hazards from Floodflows Within the John Muir House National Historic Site, Franklin Creek Drainage Basin, California. U.S. Geological Survey WRI Report 93-4009. Ostenaa, D.A., D.R. Levish, and D.RH. O'ConneII 1996 Paleoflood Study for Bradbury Dam, Cachuma Project, California. U. S. Bureau of Reclamation Seismotectonic Report 96-3. Pitlick, John 1996 "Regional flood frequency analysis for the Santa Ynez River basin and adjacent regions." In Paleoflood Study for Bradbury Dam, Cachuma Project, California. U. S. Bureau of Reclamation Seismotectonic Report 96-3. Shannon, C.E. 1948 "A Mathematical Theory of Communication." Bell Systems Tech. Journal 27:379-423. Stedinger, J.R. 1983 "Design Events with a Specified Flood Risk." Water Resources Research 19:511-522. Stedinger, J.R., and T.A. Cohn 1986 "Flood Frequency Analysis with Historical and Paleoflood Information." Water Resources Research 22:785-793. Stedinger, J.R, R Surani, R, and R Therival 1988 The MAX Users Guide. Ithaca, New York: Department of Environ mental Engineering, CorneII University. Stedinger, J.R, RM. Vogel, E. Foufoula-Georgiou 1993 "Frequency Analysis of Extreme Events." In Handbook of Hydrol ogy, edited by D.R Maidment, New York: McGraw-HilI Inc. Tarantola, A. 1987 Inverse Problems Theory: Methods for Data Fitting and Model Parameter Estimation. New York: Elsevier.

PAGE 219

196 BAYESIAN FWOD FREQUENCY ANALYSIS Tarantola, A., and B. Valette 1982 "Inverse Problems = Quest for Information." Journal o/Geophysi cal Research 50: 159-170. U.S. Bureau of Reclamation 1993 "Probable Maximum Floods, General Storm and Thunderstorm, Bradbury Dam." Memorandum from Chief, Surface Water Branch to Chief, Dam Safety Office, Bureau of Reclamation, Denver Office, May 27, 1993. U.S. Bureau of Reclamation 1995 Probable Maximum Flood Study Using Hmr 58 for Bradbury Dam, California." Memorandum from Ken Bullard, Flood Hydrology Group Technical Service Center to Chief, Dam Safety Office, Bureau of Reclamation, Reclamation Service Center, July 13, 1995. United States Water Resources Council 1981 Guidelines for Determining Flood Flow Frequency. Bulletin # 17B of the Hydrology Committee, Revised September, 1981, Washing ton, D.C. Vogel, R.W., and D.E. McMartin 1991 "Probability Plot Goodness-of-fit and Skewness Estimation Procedures for the Pearson Type 3 Distribution." Water Resources Research 27:3149-3158. Zienkiewicz, O. C., and R.L. Taylor 1991 The Finite Element Method. Volume 2. New York: McGraw-Hill.

PAGE 220

PROBLEMS WITH THE USE OF STATISTICAL PROBABILITY AS A TOOL FOR PREDICTION OF EXTREME EVENT FLOODS Gregory G. Hammer Colorado Division of Water Resources Dam Safety Branch The Colorado Rules and Regulations for Dam Construction and Dam Safety allow for the determination of the 100-year flood as the peak flow recorded over a 100-year period on a stream having a gauged record of at least 100 years (Colorado Office of the State Engineer, 1988). The statistical definition is that a 100-year flood has a 1 % probability of occurrence in any given year. In the book of Genesis, the Bible suggests that the 100-year flood occurs after raining for 40 days and nights. (The rain began 100 years after Noah was told to build an Ark.) None of these criteria is absolutely definitive, and, consequently, great latitude is allowed in the development of the predicted magnitude of such an event. To the public, this event is useful to describe the rarity or frequency of occurrence of an extreme event. To the technical community, however, its use is to identify a level of risk that is designed into projects subject to the effects of floods. The occurrence of the Big Thompson Canyon flood in 1976 served to dramatically underscore the concept of a 100-year flood. As Colorado approached its centennial anniversary of statehood, this event presented real and tangible evidence of severe floods. Hardly anyone had imagined that a flood of 30,000 cubic feet per second (cfs) would sweep out of the canyon, yet it did, and with little warning. Nor can anyone imagine today that it could have been worse. Since its occurrence, this storm has been labeled as anything from a 100-year up to a 1O,000-year flood. The probability of frequency, however, is not what really matters. The real issue is that the storm did occur, and, having occurred once, it will occur again. The important question, then, is: How much worse could it have been? As engineers, we seek to design our structures to be safe enough that no loss of life or property damage results from conditions that deviate from those

PAGE 221

198 PROBLEMS WITH THE USE OF STATISTICAL PROBABILITY that occur routinely. We create adequate safety factors for such a purpose. Where conditions can be widely variable or design data is limited, we tend to apply statistical methods to allow us some understanding of the risk of our design choices or decisions, such as in the science of hydrology. Using limited data, we project the results to an extreme that seems to be suitably risk free. In the past, the accepted event was commonly referred to as a 10,000-year flood, a concept that seemed so remote that its occurrence was not conceiv able. This has been replaced by the present use of the Probable Maximum Flood (PMF). While somewhat dependent still upon statistical probabilities, the PMF has its roots in predictable reality, and is thus better suited to limiting risks through the design of dam spillways and other flood control measures. Reducing risks increases costs, however, and these costs have themselves become as much a dilemma as the predicted magnitude of extreme floods. Unfortunately, significant cost reductions are not always achievable in the construction process, but rather have been accomplished by allowing conservative designs that accept higher degrees of risk. While this seems conceptually worthwhile, who should be responsible for assuming such risks, and who would suffer if such a design is exceeded? Seldom do such designs consider the feelings or perspective of those who are placed at risk, but rather weigh only the financial impacts to the developer or owner. Shortly after being asked to participate in this symposium, a discussion followed about my perceptions of recent studies in paleohydrology. This discipline gained wider exposure after the Big Thompson flood and was further reinforced by the Lawn Lake Dam failure in 1982. Here were two examples of dramatically high flows, on the order of 30,000 cfs and 20,000 cfs respectively, with clearly observed channel effects. Similar channel evidence was lacking in other areas, particularly in the higher mountain regions. In dam design, the paleohydrologic concept of the 10,OOO-year flood is associated with the elapsed time since the last ice age. In light of the previous concept of the 10,OOO-year flood, some questioned how meteorolo gists and engineers could predict floods with high flow rates despite no apparent physical evidence to support such occurrences. I share those concerns, but am not yet ready to accept the simple premise that, because a flood has not yet occurred, it will never occur. Recent studies have examined this condition at higher elevations, and have presented evidence that floods as large as the Big Thompson event have no physical basis. Under the auspices of the dam safety program, the Colorado

PAGE 222

Hammer 199 state engineer has commissioned a study that should address some of these concerns. I once read that the only thing humans learn from history is that humans learn nothing from history. Having grown up in Williamsburg, Virginia, history is a major interest of mine. When I first learned about the recent discussion of paleohydrology, I recognized a process that was practiced in the past. Prior to 1900, engineers designed spillways using the identified high water mark for the stream in question (National Research Council, 1985). Often, engineers would make the spillway somewhat larger than the previous "flood of record" to improve safety. Over time, however, many of these dams washed out as floods exceeded the capacity of these spillways. This was sufficiently common that overtopping is the single greatest cause of failure of embankment dams. In the publication Dam Incidents, USA (ASCE/USCOLD, 1976) it is noted that no dam constructed after 1925 has failed due to overtopping.] The subsequent version, Lesson from Dam Incidents, USA-II (ASCE/USCOLD, 1988), supports this finding. During this period a transition was occurring in the design of spillways: designs began to be based upon predicted precipita tion rather that probability of streamflow. By 1940, meteorological effects were the primary consideration in the development of design storms. As better observational and computation techniques developed, an improved understanding of weather patterns resulted in the Probable Maximum Precipitation (PMP) being more widely accepted. The Hydrometeorological Reports (HMR) (U.S. Department of Commerce) published by the National Weather Service provide this data. A consequent effect has been less emphasis upon the probabilistic return period for floods, except to inform the public of the relative magnitude of an event. I see this trend of moving away from probability as a good direction. Although the concept of the 100-year flood seems a good standard of magnitude for small or low hazard dams, it can create a false sense of security in the general public. It is easily misinterpreted, causing the public to believe that after the occurrence of a l00-year flood, it will be another 100 years before its recurrence. Thus, what should be stressed-and emphatically 1. Since USCOLD does not inventory dams less than 50 feet in height, this statement does not therefore apply to smaller dams that have been constructed with less engineering analysis. Many small dams are designed to route a 1oo-year return flood, and have been overtopped even when designed after 1925.

PAGE 223

200 PROBLEMS WITII THE USE OF STATISTICAL PROBABILITY so-is that if the flood happens once, it most certainly can happen again at any time, and possibly even tomorrow. In light of a historical perspective of this problem, a couple of stories seem in order. I once worked on a dam in Oklahoma, where it seems that the major storms occur over Memorial Day weekend. I was assigned to design several rim embankments to handle a new PMF, developed in the early 1980s. Over a several year period in the 1970s, the project had experienced three "floods of record." The first was identified as a 100-year event. When the second occurred a couple of years later, it became a 20-year event. When a third storm occurred, the statistical return period was further reduced. The project is now 50 years old, and I am not aware of any such events having occurred again. In this case, the assignment of a statistical return period had no legitimate application other than to give the public and the news media a measure of the severity of the storm and its probability of returning in the near future. In another case, a story was related to me about a flood-control project in the Phoenix area. As construction began, the local river master ridiculed the designers for building such a large structure. During his 30 years, there had never been a flood of the magnitude predicted by the engineers; yet, while still under construction, a flood larger than the design flow washed out the project. As the subsequent redesign was nearing completion, another event occurred, again larger than the designers had predicted. Assuming the third time is a charm, the design was redone and I presume has been built. While I cannot vouch for any accuracy, the purpose of relating this story to me initially, and here as well, is that when we think we know where the edge of the "envelope" of extreme events is, Mother Nature has a habit of pushing it a bit farther. Lastly, the house I grew up in is located some 40 feet above a small stream that runs through the woods behind the house. In the 35 years that my family has lived there, on only a few occasions had the stream gone out of its banks. But one evening a large storm that produced 13 inches of rain overnight raised the stream level to the back door steps! It has been common in recent years to hear stories of floods that have dramatically exceeded the previously recognized record floods. Within the perspective of paleohydrology, should it be presumed that if an event hasn't occurred in 10,000 years, then it is not likely to occur? Further, should we preclude the occurrence of a greater magnitude event than may be supported by the geologic record? The basic concept of such a long period is in itself curious. Some 10,000 years ago the earth was wrapped within the ice age. The hydrologic cycle as

PAGE 224

Hammer 201 we understand it was distorted to allow for the creation of ice. As the predictions of global warming hang over our heads, and ice caps recede, clearly the conditions of the past no longer prevail. Assuming that the cycle is balanced and nothing is lost to outside influences, it follows then that the former volumes of ice are now available to the atmosphere as moisture. It should not seem surprising then that stream gages continue to record floods of increasing magnitude. It seems to follow also that the premise of basing designs on past history is not appropriate when natural and physical processes suggest greater effects. Consider the Cherry Creek, Colorado, flood of May 30 and 31,1935. At the location of the storm center, where 24 inches of rain was recorded over a 24-hour period, the 100-year event based on NOAA Atlas (Miller, 1973) data would be approximately 3.5 inches. Projecting the statistical storm data (Figure 1) for more frequent return storms toward the observed 24 inches suggests that the 1935 event has a return period of 1018 years. This would well exceed any paleontological record of 10,000 years. Comparing this event with data from the Plum Creek event of June 1965 shows that the town of Agate received 8 inches of rain during both events. Again, the 100-year return storm precipitation index shows a value of 3.5 inches. Clearly an 8inch rain is well in excess of the "lOO-year storm," but two such events occurred over a 30-year period. Similar data were generated for the Big Thompson storm (Figure 2), where the center produced 12 inches of rain over four hours. In this example, the event would have a return period of almost 20,000 years. Further analysis shows that this event closely approximates the predicted probable maximum precipitation (PMP) at the storm center. In neither of these examples has it been proven that a larger event is not possible, and evidence suggests in fact that a larger storm may be probable. Without ignoring the geologic record then, how can we resolve the apparent difference between the predicted results of large precipitation events against the lack of physical evidence for flood flows of that magnitude? Two possible explanations exist, and both are conceivably simple. The first may be that no storms of such great magnitude as a PMF have yet to occur over most basins, The second may be due to the perspective of the data available. The meteorologist is predicting the volume of rain that is possible given conceivable extreme conditions. The paleohydrology study on the other hand is useful for determining the magnitude of historic extreme flows recorded in the geologic record. The former is input to a computer model, the later is the output. What occurs in between the two seems then to be our source of concern. If the former explanation is the cause, then our only course of action

PAGE 225

202 PROBLEMS WITII THE USE OF STATISTICAL PROBABILITY -" -" .. .. a:. [] @Ia '" .. III c:: b "'" .2 ." :::J .2 Ol u::: (; :. Co :; ... ... e III c. .s:: (J 8 iii f-+-t-i-j--.. -1---1-----1----+-1-+--1---1------1--1--:.-+----1-----1 UleJ jO

PAGE 226

Hammer I: o III >-., I: U :' i'i c o 0 15 .c ., IU ., iii c. ! 8 : I, I' ., .. 203

PAGE 227

204 PROBLEMS WITH THE USE OF STATISTICAL PROBABILITY is to watch and be ready when the event does occur. If the latter is true, then resolving the parameters that generate the observed runoff with the predicted precipitation will give better insight into predicting flood events. As the Colorado State Engineer's Office has embarked upon a program of reviewing the spillways in the Colorado, our developed procedures for hydrologic modeling occasionally suggest basin parameters that do not seem correct. The common tool for hydrologic analysis is the HEC-1 program developed by the U. S. Army Corps of Engineers. We have data for moderate storms that have occurred over well-instrumented basins. In these cases good rainfall depth-duration data and good stream flow data have been available for the affected stream. Using the optimization procedures, the program can develop the infiltration parameters for the basin for various methodologies. In some cases the calculated parameters are significantly different than what might be routinely selected. It is also not uncommon to encounter dams with spillways so small that the structure might be expected to overtop in a thick fog; yet, the dams are over 50 years old and this has not happened. Certainly over the last 50 years significant precipitation events have occurred, and they have occurred without generating a flood that affected these dams. Conse quently, further research into infiltration and runoff parameters is needed. No clear answer is available when predicting the magnitude and severity of floods. Large floods can and do occur, and despite predictions that may seem inordinately excessive, we continue to be awed by the magnitude of storms that have occurred. As long as human lives remain at risk from our choices, however, our predictions must be sufficiently conservative to lessen the risk. And, as history suggests, we should learn to expect large floods and prepare for them. References American Society of Civil Engineers, U.S. Committee on Large Dams (ASCE, USCOLD) 1976 Lessons from Dam Incidents, U. S.A .. 1988 Lessons from Dam Incidents, U. S.A. II. Colorado Office of the State Engineer 1988 Rules and Regulations for Dam Safety and Dam Construction, Denver: Colorado Division of Water Resources.

PAGE 228

Hammer 205 Miller, J.F., R.H. Frederick, and R.J. Tracey 1973 Precipitation-Frequency Atlas of the Western United States, Volume Ill-Colorado. Washington, D.C.: U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Weather Service National Research Council 1985 Safety of Dams: Flood and Earthquake Criteria. U. S. Department of Commerce 1988 Probable Maximum Precipitation EstimatesUnited States: Between the Continental Divide and the l03rd Meridian. Hydro meteorological Report No. 55A.

PAGE 229

This page IS intentionally blank

PAGE 230

..."I ,.,. Related Publications of Interest from the Natural Hazards Center All items can be ordered from the Natural Hazards Research and Applications Information Center Campus Box 482 University of Colorado Boulder, CO 80309-0482 (303) 492-6819 e-mail: WWW: Monograph Series MG53 Coastal Erosion: Has Retreat Sounded? Rutherford H. Platt et al. 1992. 210 pp. $20.00. MG54 Partnerships for Community Preparedness. David F. Gillespie. 1993. 150 pp. $20.00. MG57 Disaster Evacuation and the Tourist Industry. Thomas E. Drabek. 1994. 282 pp. $20.00. MG58 Disaster Evacuation Behavior: Tourists and Other Transients. Thomas E. Drabek. 1996. 370 pp. $20.00. Special Publications SP25 Action Agenda for Managing the Nation's Floodplains. A Review of Floodplain Management in the United States: An Assessment Report. 1992.22 pp. $20.00. SP27 When the River Rises: Flood Control on the Boise River 1943-1985. Susan M. Stacy. 1993. 209 pp. $20.00

PAGE 231

SP28 Guidelines for the Uniform Definition, Identification, and Measurement of Economic Damages from Natural Hazard Events. Charles W. Howe and Harold C. Cochrane. 1993.28 pp. $20.00. SP30 NANIA: "All Together"-Comprehensive Watershed Management. Proceedings of the Eighteenth Annual Conference of the Association of State Floodplain Managers. May 8-13, 1994. Tulsa, Oklahoma. 472 pp. $20.00. SP31 From the Mountains to the Sea-Developing Local Capabilities. Proceedings of the Nineteenth Annual Conference of the Association of State Floodplain Managers. May 22-26, 1995, Portland, Maine. 1996. 490 pp. $20.00. SP32 Coast to Coast: 20 Years of Progress. Proceedings of the Twentieth Annual Conference of the Association of State Floodplain Managers. June 10-14, 1996. San Diego, California. 412 pp. $20.00. Working Papers in Print WP81 Sullivan's Island, South Carolina-The Hurricane Hugo Experience: The First Nine Months. Jamie W. Moore and Dorothy P. Moore. 1993.64 pp. $9.00. WP82 Biological Hazards and Emergency Management. Janet K. Bradford et ai. 1992. 27 pp. $9.00. WP83 Natural Hazard Trends in the United States: A Preliminary Review for the 1990s. Pamela Sands Showalter, William E. Riebsame, and Mary Fran Myers. 1993. 58 pp. $9.00. WP84 The Public Policy Response to Hurricane Hugo in South Carolina. Elliott Mittler. 1993. 72 pp. $9.00. WP85 The Evolution of Flood Hazards Programs in Asia: The Current Situation. James L. Wescoat, Jr. and Jeffrey W. Jacobs. 1993. $9.00. WP87 Insurance and Natural Disasters: An Examination of the New Zealand Earthquake and War Damage Commission. Arnold R. Parr. 1994. 27 pp. $9.00. WP88 Natural Disaster Management in Korea: An Analytic Study with Policy Implications. Wook-Joong Kim. 1994. 94 pp. $9.00. WP90 Dreading the Next Wave: Nontraditional Settlement Patterns and Typhoon Threats on Contemporary Majuro Atoll. Dirk H.R. Spennemann. 1995.42 pp. $9.00. WP91 The Hyatt Skywalk Disaster and Other Lessons in the Regulation of Building. William L. Waugh, Jr. and Ronald John Hy. 1995. 15 pp. $9.00. WP93 Renewing FEMA: Remaking Emergency Management. Richard Sylves. 1995. 36 pp. $9.00.

PAGE 232

Working Papers on the World Wide Web WP94 Hurricane Damage to Residential Structures: Risk and Mitigation. Jon K. Ayscue. 1996. 11/13/96. Free. (Not available in printed form.) Quick Response Reports in Print QR68 Risk Factors for Death in the 27 March 1994 Georgia and Alabama Tornadoes. Thomas W. Schmidlin and Paul S. King. 1994. 15 pp. $5.00. QR70 Children of 1niki: Effects of Evacuation and Intervention. Roger S. Hamada. 1994. 23 pp. $5.00. QR72 Immediate Emotional Response to the Southern California Firestorms. E. Allison Holman and Roxane Cohen Silver. 1994.23 pp. $5.00. QR73 Residential Loss and Displacement Among Survivors of the 1993 Altadena Fire. Norma S. Gordon et al. 1994. 15 pp. $5.00. Quick Response Reports on the World Wide Web QR76 Farmers' and Public Responses to the 1994-1995 Drought in Bangladesh: A Case Study. Bimal Kanti Paul. 1995. $5.00 for printed copy. QR77 Psychophysiological Indicators of PTSD Following Hurricane Iniki: The Multi-Sensory Interview. Kent D. Drescher and Francis R. Abeug. 1995. $5.00 for printed copy. QR78 Self Organization in Disaster Response: The Great Hanshin, Japan Earthquake of January 17, 1995. Louise K. Comfort. 1996. $5.00 for printed copy. QR79 Transition from Response to Recovery: A Look at the Lancaster, Texas Tornado. David M. Neal. 1996. $5.00 for printed copy. QR81 Newspaper Reporting in Wake of the 1995 Spring Floods in Northern California. Ute 1. Dymon and Francis P. Boscoe. 1996. $5.00 for printed copy. QR82 Early Response to Hurricane Marilyn in the U.S. Virgin Islands. Betty Hearn Morrow and A. Kathleen Ragsdale. 1996. $5.00 for printed copy. QR84 Impact of Hurricane Opal on the Florida/Alabama Coast. David M. Bush et al. 1996. $5.00 for printed copy. QR85 The Potential Impact of Information Technology on the Struciure of Inter organizational Relationships during Crisis Response: The Pennsylvania Floods of 1996. Diana Burley Gant. 1996. $5.00 for printed copy.

PAGE 233

QR86 The Politics and Administration of Presidential Disaster Declarations: The California Floods of Winter 1995, Richard Sylves. 1996. $5.00 for printed copy. QR87 Coping Self-Efficacy and Psychological Distress Following the Oklahoma City Bombing, Charles C. Benight. 1996. $5.00 for printed copy. QR88 Response to Severe Winter and Blizzard Conditions in Grundy and Buchanan County, Virginia in 1996: A Focus Group Analysis. Joseph B. Perry, Duane Dukes, and Randall Norris. 1996. $5.00 for printed copy. QR90 Tornadoes in the Districts of Jamalpur and Tangail in Bangladesh. Thomas Schmidlin and Yuichi Ono, 1996. $5.00 for printed copy. Topical Bibliographies TBI6 A Bibliography of Weather and Climate Hazards. Jolana Machalek. 1992. 335 pp. $30.00. TBI8 Epidemiology of Disasters: A Topical Bibliography. Eric K. Noji. 1994. 69 pp. $20.00. TBI9 The Socioeconomic Aspects of Flooding in the U.S.: A Topical Bibliography. John Wiener. 1996. 49 pp. $20.00. Also available via the Internet for free at The Natural Hazards Observer The Natural Hazards Center publishes a bimonthly newsletter, the Natural Hazards Observer, which covers disaster management, mitigation, and education programs; information sources; research and findings from completed projects, recent legislation; applications of research at federal, state, and local levels and by private organizations; recent publications; and future conferences. Subscriptions to the printed version of the Observer are free within the U.S., and foreign subscriptions cost $15.00 per year. Beginning with Volume XX, No.4 (March 1996), back issues of the Observer are also available at the Hazards Center's Web site:

PAGE 234

How to Obtain Our Publications Shipping and handling charges must be added to all orders. Based on the total number of pages in an order, they may be calculated from the chart below. For larger orders, contact the Publications Clerk, Natural Hazards Research and Applications Informntion Center, Campus Box 482, University of Colorado, Boulder, CO 80309-0482; (303) 492-6819;fax: (303) 492-2151; e-mail: All orders must be prepaid and checks should be payable to the University of Colorado. American Express, Visa, Mastercard, and Diners Club cards are accepted. To obtain a printed copy of the full 8-page list of Natural Hazards Center publications, send $3.00 to the Publications Clerk at the above address. To receive a free copy via the Internet, send an e-mail message to or access the Natural Hazards Center's Home Page at Shipping Charges DOMESTIC # of Pages Printed Matter First Class 0-35 $3.00 $3.00 36 80 $3.50 $4.00 81 450 $4.00 $5.00 CANADA AND MEXICO # of Pages Surface Printed Air Printed 0-35 $3.00 $3.00 36 80 $3.50 $4.50 81 450 $5.00 $6.00 INTERNATIONAL # of Pages Surface Printed Air Printed 0-35 $4.00 $5.00 36 80 $5.00 $6.00 81 450 .'\;6.00 CaB for mice