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 Ka
controlfield tag 001 001910234
007 cr mnu|||uuuuu
008 070925s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001689
Clayback, Kim Beth.
Investigation of normalized streamflow in West Central Florida and extrapolation to ungaged coastal fringe tributaries
h [electronic resource] /
by Kim Beth Clayback.
[Tampa, Fla] :
b University of South Florida,
ABSTRACT: Deriving accurate streamflow estimates for ungaged watersheds provides a challenging task for water resource engineers. Traditional methods include correlation to the nearest USGS streamflow station or numeric simulation of watershed rainfall-runoff processes. Mean annual flow, ten percent exceedance and other streamflow indices can be normalized and non-dimensionalized by dividing by the watershed drainage area and the mean annual precipitation rate. Obtaining non-dimensional parameters can be especially useful for extrapolation of flows to downstream, ungaged, coastal fringe regions. Florida and other states along the Gulf Coast exhibit strong variability in the magnitude of streamflow fraction of precipitation. The irregular patterns created by the variance in magnitude do not correlate well with traditional statistical methods of parameter estimation. Using spatial and hydrologic factors, this study, through parameter sensitivity analysis, correlates land-use, slope, soil type, precipitation, and watershed area to a non-dimensional fraction that is to be applied to ungaged regions to determine the streamflow scaling. The study domain for the land-use correlation method is West-Central Florida. Strong trends in correlation to land-use were found but underlying geology must also be considered when defining the study domain. Urbanization, depth-to-water-table and grassland were the dominant parameters in the northern study domain yielding an 80 percent correlation to streamflow fraction for the combined factors. While in the southern domain, wetlands and depth-to-water-table combined to be an indicator with a 75 percent correlation.
Thesis (M.A.)--University of South Florida, 2006.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 59 pages.
Adviser: Mark A. Ross, Ph.D.
x Civil Engineering
t USF Electronic Theses and Dissertations.
Investigation of Normalized Streamflow in West Central Florida and Extrapolation to Ungaged Coastal Fringe Tributaries by Kim Beth Clayback A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Ci vil Engineering Department of Civil and Environmental Engineering College of Engineering University of South Florida Major Professor: Mark A. Ross, Ph.D. Mark Stewart, Ph.D. Ahmed Said, Ph.D. Kenneth Trout, Ph.D. Date of Approval: July 17, 2006 Ke ywords: watershed ratio, ungaged watersheds, streamflow fractionation, precipitation fraction, coastal sub basins Copyright 2006, Kim Beth Clayback
Acknowledgements In the spring of 1971, I decided I was going to get a Masters Degre e. Thirty five years later, its finally happening. To all of those who have given me the opportunity to realize my aspirations, I thank you. My continuing education would not have been possible without the love, confidence and support of my parents, Ro n and Carol. My two extraordinary children, Gabbie and Blake, have brought sunshine to my days. Rocco has provided unending patience. Many thanks to Dr. Ross for believing in me and sharing his enthusiasm and insight, and to Dr. Stewart, Dr. Trout and Dr Said for their suggestions and edits. Further thanks to all of my friends, Auristela, Ken, Nirjhar, Eduardo, Jing, Lisa, Makhan and Sandra, for their smiles, assistance and encouragement. Thank you to Rob for just being there. And thank you Michael for being my partner in this endeavor.
i Table of Contents List of Tables ii List of Figures iv Abstract v i Chapter One: Introduction 1 1.1 Introduction 1 1.2 Background Information 2 1.3 Objective of Study 7 Chapter Two: Data Collection and Me thodology 8 2.1 Study Area 8 2.2 Methods 16 2.2.1 Normalized Streamflow Fraction 16 2.2.2 Streamflow 17 2.2.3 Precipitation 18 2.2.4 Land Use 20 2.2.5 Soils 23 2.2.6 Slope 24 2.2.7 Evapotranspiration 27 2.2.8 Statistical Analysis 27 C hapter Three: Results and Findings 29 3.1 Depth to Water Table 29 3.2 Northern Domain 30 3.2.1 One to One Correlation with NSF and Land Use Classifications 30 3.2.2 Stepwise Regression 35 3.2.3 Multivariate Linear Regression 36 3.3 Southern D omain 39 3.3.1 One to One Correlation with NSF and Land Use Classifications 39 3.3.2 Stepwise Regression 39 3.3.3 Evapotranspiration 39 3.3.4 Charlotte Harbor Study of Ungaged Sub basins 42 3.4 Complete Study Domain 43 Chapter Four: Discussion and Conclusions 45 4.1 Northern Domain 45 4.2 Southern Domain 46 4.3 Complete Study Area 47
ii References 48 Appendices 51 Appendix A: Tables 52
iii List of Tables Table 1: Annual Precipitation Values 20 Table 2: Depth to Water Tabl e Statistics 24 Table 3: Estimated Annual Evapotranspiration Rates 27 Table 4: Results of Actual NSF Vs Predicted NSF for Northern Sub basins 37 Table 5: Results of Actual NSF vs Predicted NSF for Southern Coastal Basins 41 Table 6: Charlotte Harbor Study Using HSPF vs NSF Equation 42 Table 7: United States Geological Survey Gaging Station Locations 52 Table 8: Gaging Station Statistics 54 Table 9: Land Use Classification Percentage by Sub basin 56
iv List of Figures Figure 1: Study Do main: West Central Florida Counties 10 Figure 2: Karst Formations at Land Surface 11 Figure 3: USGS Gage Locations and Sub basin Areas 14 Figure 4: Recharge Discharge Zones 15 Figure 5: Precipitation Gages Locations 19 Figure 6: Land Use Class ification Map 22 Figure 7: Monitor Well Locations 25 Figure 8: Average Depth to Water Table Map and Wellfield Locations 26 Figure 9: One to One Correlation: Northern Domain: Urbanization with Wellfield 31 Figure 10: One to One Correlation: Northern Domain: Urbanization without Wellfield Pumping Sub Basins 32 Figure 11: One to One Correlation: Northern Domain: Grasslands without Wellfields 33 Figure 12: One to One Correlation: Northern Domain: Open Water and Wetlands without Wellfield Pumping Sub basins 34 Figure 13: One to One Correlation: Northern Domain: Open Water and Wetlands in Sub basins with Wellfields 34 Figure 14: One to One Correlation: Northern Domain: Shallow Depth to Water Table without Wellfields 35 Figure 15: Multivaria te Regression Results 36 Figure 16: Comparison of Regression Prediction vs Area Scaling Method 38 Figure 17: Test Gages and Sub Basins in the Northern Domain 38 Figure 18: Southern Domain: Regression Equation 40
v Figure 19: Southern Domain: Around the L ittle Manatee River: Predicted vs Area Scaling Method 41 Figure 20: Normalized Streamflow Fraction 44
vi Investigation of Normalized Streamflow in West Central Florida and Extrapolation to Ungaged Coastal Fringe Tributaries By Kim Beth Clayback ABS TRACT Deriving accurate streamflow estimates for ungaged watersheds provides a challenging task for water resource engineers. Traditional methods include correlation to the nearest USGS streamflow station or numeric simulation of watershed rainfall runo ff processes. Mean annual flow, ten percent exceedance and other streamflow indices can be normalized and non dimensionalized by dividing by the watershed drainage area and the mean annual precipitation rate. Obtaining non dimensional parameters can be e specially useful for extrapolation of flows to downstream, ungaged, coastal fringe regions. Florida and other states along the Gulf Coast exhibit strong variability in the magnitude of streamflow fraction of precipitation. The irregular patterns created b y the variance in magnitude do not correlate well with traditional statistical methods of parameter estimation. Using spatial and hydrologic factors, this study, through parameter sensitivity analysis, correlates land use, slope, soil type, precipitatio n, and watershed area to a non dimensional fraction that is to be applied to ungaged regions to determine the streamflow scaling.
vii The study domain for the land use correlation method is West Central Florida. Strong trends in correlation to land use were found but underlying geology must also be considered when defining the study domain. Urbanization, depth to water table and grassland were the dominant parameters in the northern study domain yielding an 80 percent correlation to streamflow fraction for the combined factors. While in the southern domain, wetlands and depth to water table combined to be an indicator with a 75 percent correlation.
1 Chapter One: Introduction 1.1 Introduction Quantifying streamflow from ungaged catchments is a cha llenge faced by the hydrological community. Due to the difficulties in gaging flows in tidal conditions, all observed streamflow gages are located well upstream of any tidal influences, leaving large portions of the river ungaged (Ross et al., 2005). Coa stal and estuary management relies heavily on extrapolation of fresh water flows from the nearest available gaging station to predict the fresh water flows in the ungaged catchments. While the area scaling method is commonly employed by the engineering co mmunity, a need exists for a more accurate representation of the rainfall runoff in addressing environmental health, community safety and industrial concerns. Streamflow impacts must be considered when planning for a new development, roadway, or drainage s ystem. One aspect of streamflow, direct runoff, is a primary component in predicting flood levels and calculating the storage requirements of a storm water pond. There is a strong need for a more accurate method for extrapolating streamflow to un gaged areas. This method would allow urban planners, engineering consultants, and other professionals to determine the estimated streamflow for an ungaged area in a straight forward analysis and without expensive mathematical modeling. Rainfall runoff mo deling consists of two different approaches: direct correlation of known physical parameters to extrapolate ungaged flows and
2 parameterization of general catchment characteristics inserted into a conceptual model to forecast ungaged flow. The most commonl y used method at present is to determine the streamflow for an ungaged area is to take the data from the nearest streamflow gaging station and estimate the streamflow for the desired area using an area scaling method (Hammett and DelCharco, 2005). The are a scaling method traditionally under predicts the streamflow values indicating that using area alone as a predictor is insufficient (Ross, 2006). Depending on the distance and commonalities between the station and the new area, this estimation can be accu rate within a few percent or a few hundred percent. The process is not rigorous and does not take into account any of the distinct land use differences between the gaged and ungaged basins. Using conceptual models requires more time and a priori catchmen t attributes but does not necessarily yield results with greater precision. This study seeks to directly relate the effects of land use, soil type, depth to water table, and slope on the fraction of precipitation runoff in ungaged catchments. 1.2 Backg round Information Hydrological literature shows several studies striving to extrapolate catchment characteristics to ungaged basins from hydrologically homogenous regions that are measured (Post and Jakeman, 1999; Nathan and McMahon, 1990). The region of influence approach and cluster analysis as well as regression analysis have all been attempted with limited success (Kokkonen et al., 2003). A 2003 study of the German Rhine River basin set out to find a relationship between catchment properties and weig hted model parameters using a Lippschitz continuous transfer function. The study suggested predictions of conceptual models are considerably uncertain, even for gaged catchments (Hundecha et al., 2003).
3 Boughton (1984) used a simple three parameter model to estimate water yield from ungaged catchments using precipitation and evaporation data. The parameters were surface storage capacity (land use and cover), infiltration capacity (soil type) and baseflow (streamflow characteristics). The results show an error of 10 percent in rainfall data yield and a 20 percent error in estimated runoff but the model had a very low sensitivity to the possible errors in choosing the parameter values. The parameter values were assigned based on catchment coverage, soil t ype and flow characteristics of the ungaged basin. To extrapolate runoff values of ungaged catchments, the SFB (Surface Infiltration Baseflow) model can be used for small ephemeral catchments and with catchments where baseflow forms a significant componen t of runoff. Servat and Dezetter (1993) modeled 20 catchments in the Ivory Coast using 2 conceptual models, GR3 (Edijatno and Michel, 1989) and CREC (Cormary and Guilbot, 1973), applied to ungaged catchments. Using area, rainfall and land use variables t o calibrate the model, the study found the correlation matrix showed almost no clear linear relationship between the model parameters and the selected variables. Woolridge et al. (2001) used a simple infiltration model combined with a landscape/climate r egionalization approach to provide meaningful physical parameters capable of simulating the effect of land use on hydrologic response at the regional scale. The study suggests that a greater reliance on spatial modeling combined with qualitative reasoning is needed to shift away from explicit modeling of small scale catchment processes. When using forested and non forested land use classifications in conjunction with the infiltration model, a robust calibration resulted, allowing for more confidence in ex trapolating model predictions to ungaged catchments.
4 Mwakalila (2003) provides a technique for relating catchment properties to hydrological responses in gaged basins which he used to predict hydrological responses on ungaged basins. The study used topo graphy, soil, climate and land use parameters. The strategy used to develop relationships between hydrological responses and physical characteristics was correlation analysis, principal components analysis, stepwise regression and multiple regression. Ni ne basin characteristics were used in 6 hydrological parameter equations. The equations correlate to measured basin characteristics with an R 2 range of 0.72 for effective storage capacity to 0.89 for discharge when recession commences. Pandey and Nyugen (1999) conducted a comparative study of regression based on methods in flood frequency analysis. Six methods were evaluated, one being ordinary least square (OLS) regression used in this investigation. Pandey and Nyugen conclude the OLS method gives unbi ased and minimum variance estimates of parameters provided the errors are independent and are normally distributed. Under these conditions the OLS estimate is also the maximum likelihood estimate of the parameters. Sefton and Howarth (1998), using a geog raphical information system (GIS), linked topography, soil type, climate and land cover to the hydrological model IHACRES (Jakeman et al., 1990; Littlewood and Jakeman, 1994). IHACRES calculates effective rainfall which contributes directly to streamflow and a second parameter that converts rainfall excess (runoff) into streamflow over 60 catchments in England and Wales. Dynamic response characteristics (DRC), describing the hydrologic response of the catchment, were correlated to the 30 unique physical c atchment descriptors. The four highest correlated parameters were standard annual average rainfall (R 2 = 0.83), groundwater or aquifer present (R 2 =0.77), percent of peaty soil (R 2 = 0.76), and percent of tilled soils (R 2 =0.76).
5 Hammett and DalCharco (2005) studied flood discharge in West Central Florida. Using a geographic information system to evaluate basin characteristics, the study found contributing drainage area, channel slope and the percent of total drainage area covered by lakes to be statisticall y significant in describing flood discharge at gaged stations. Regression equations were used to relate the three parameters at gaged stations. The equations were then used to estimate flood discharge at ungaged sites using drainage area and basin slope from the ungaged site as input values. The study looked at recurrence intervals of 2, 5, 10, 25, 50 100, 200 and 500 years and found none of the stations were significantly affected by regulation or urbanization. The limitation of the study in estimating flow in ungaged areas was that the ungaged coastal areas studied were heavily urbanized, more so than the sub basins that were gaged and used to create the regression equations. Regression analysis has been the tool of choice for attempting to extrapolat e the hydrologic response and physical parameters from a gaged catchment to that of an ungaged catchment (Kokkonen et al., 2003). The USGS has developed numerous regression equations to provide methods for estimating multiple streamflow statistics at unga ged sites (Ries and Gray, 2005). StreamStats (Ries and Gray, 2005) is a web based application that determines the watershed boundaries and measures physical and climatic characteristics of ungaged sites using a geographic information system (GIS). Stream flow estimates result when the basin characteristics are linked to the appropriate regression equation. The authors suggest StreamStats reduces the average time from hours to minutes to obtain streamflow statistics for ungaged regions. StreamStats assumes rural flow conditions only; the error for the ungaged sites is the same as known sites and a direct correlation with the upstream physical attributes, not the attributes of the ungaged sub basin. National implementation will take several years.
6 The surf ace water model HSPF (Bicknell et al., 2001) was used to estimate ungaged flows to upper Charlotte Harbor (Ross et al., 2005). The study used local rainfall, potential evapotranspiration (ET), estimated (ET), agricultural pumping and gaged streamflow to ex trapolate flow to ungaged sub basins. Groundwater was not considered in the study. Runoff to precipitation ratios were calculated after the model calibration reached acceptable levels. The study cautions high correlations between individual parameters m ay be a source of model uncertainty Using catchment attributes to parameterize conceptual models has also met with limited success (Carlile et al., 2002). Mathematical modeling has been popular to recreate physical conditions using any number of watersh ed parameters. Conceptual modeling of catchments based on an array of real parameters and theoretical mathematical attributes fill the literature. Two common messages are: the relationships between model parameters and catchment attributes are not alwa ys adequately investigated; and over parameterization of conceptual models is a problem (Kokkonen et al., 2003). Over parameterization can significantly reduce model reliability and predictive capability.
7 1.3 Objective of Study Understanding the influen ce of land use, depth to water table (DTWT) and other basin parameters on the streamflow regime of coastal areas will allow engineers and planners to better estimate streamflow in ungaged basins. This will allow evaluation of proposed changes and remediat ion in environmentally sensitive areas. The need for a more accurate representation of spatial variability is crucial to making educated and informed decisions. The objective of this study is to identify the factors that control streamflow and create a p rocedure to extrapolate the mean annual flow to ungaged, coastal regions. This thesis is designed to develop an easy, efficient and more accurate procedure to estimate streamflow in ungaged regions through the use of a dimensionless number called a norma lized streamflow fraction (NSF). Land use characteristics, soil type, depth to water table, slope and estimated evapotranspiration (ET) are the parameters to be analyzed to find a relationship between upstream, measured streamflow and downstream, ungaged streamflow. In creating a normalized streamflow fraction, the runoff contributing to streamflow is independent of precipitation and the area of the sub basin. A case study in West Central Florida is made to show that simple area extrapolation is unrelia ble and simple GIS based analyses can be made to better estimate streamflow in ungaged sub basins.
8 Chapter Two: Data Collection and Methodology 2.1 Study Area The study domain in West Central Florida is approximately 10,400 square miles encompassing 16 counties with a population of 3.1 million people (Figure 1). The average precipitation across the domain for the 11 year period used in the study is 52 inches per year, but precipitation shows substantial spatial and temporal variability (Scott, 2006) On average, the driest months of record are November and April, with the wettest being July and August (NOAA, 2006). The mean annual temperature is 73F. The mean annual open water evaporation rate for the region is 52 inches per year (Ruskauff et al. 2003). Significant geologic differences exist between the northern and southern parts of the study domain; the dividing line runs roughly along Interstate 4 (Figure 1). The northern part of the area has widespread karst features at land surface with an unconfined aquifer below (Figure 2). The southern area has confined aquifers with little surface expression of karst features. Three distinct aquifer systems, the surficial, the Intermediate and the Floridan, are found in the karst dominated framework. The spatial variation of the confinement suggests partitioning of the study area. The carbonate features at land surface affect the surface water runoff and recharge in the northern sub basin area. Carbonate rock formations constitute the major water bea ring unit with high permeability and extensive pore space. The high permeability found in these shallow carbonate formations allows for precipitation to immediately become part of the deeper groundwater system where it is subject to
9 potable supply, pumpin g and direct marine discharge and therefore, not part of streamflow in the presently highly developed aquifer. The Florida peninsula sits atop a mostly submerged carbonate platform that generally gets thicker to the south. Pleistocene sand and clay depos its of varying composition and thickness exist throughout the exposed peninsula. The occurrence of carbonate rocks range from land surface in the north to depths of greater than 500 feet below land surface to the south (Tihansky and Knochemas, 2001). The overburden deposits and the degree of confinement thickens to the south, in excess of 200 feet, with fewer karst features appearing at the surface. The surficial aquifer system is predominately sand, the intermediate aquifer system is interbedded silicic lastics and carbonates and the Floridan aquifer system is massive carbonates (Tihansky and Knochemas, 2001). The central part of the study domain shows carbonate units dipping and becoming overlain by the thickening Hawthorne Formation that forms the Inter mediate aquifer system south of Tampa Bay. Below the Intermediate Aquifer System is a confining unit for the Floridan Aquifer and the presence of the confining unit is the primary cause of the change in geologic environments between the north and south po rtions of West Central Florida.
10 Figure 1: Study Domain: West Central Florida Counties
11 Figure 2: Karst Formations at Land Surface
12 The coastal springs, found in the north, originate from elevated potentiometric heads from inland areas. The springs d ischarge millions of gallons per day of groundwater from the Upper Floridan Aquifer (Tihansky and Knochemus, 2001). The karst springs add freshwater to Tampa Bay and coastal embayments to the north. To the south, streamflow consists mostly of runoff with negligible direct significant spring flow. The three largest rivers in the study domain are the Withlacoochee, Hillsborough and Peace Rivers. The combined drainage area of the three watersheds is 5100 square miles, more than half of the study area. In the north, the Weeki Wachee, Chassahowitzka, Homosassa and Crystal Rivers all originate from coastal springs and are not considered as part of the runoff regime in the study, (Hammett and DelCharco, 2005). In the south, the Alafia, Little Manatee and Mana tee Rivers all terminate into Tampa Bay along with the Hillsborough River. The Myakka and Peace Rivers terminate into Charlotte Harbor. The land surface elevations for the study domain range from just over 200 feet above sea level to sea level. Ridge sy stems are found along the eastern boundary of the study domain, the Lake Wales Ridge, and in the northeastern corner of the domain, the Brooksville Ridge. The United States Geological Survey (USGS) maintained gaging stations used were at the outf low of the sub basins. The smallest drainage basin used in the study is 5.22 square miles and the largest drainage basin is 352.83 square miles (Figure 3). Pinellas County was excluded from the domain due to the tidal influence on fresh water streamflow t hroughout the entire county. Seventy two streamflow gages collected data across more than 6100 square miles in the domain, resulting in 60% of the area having recorded streamflow. Sixty six gages collected streamflow data for a period of record of seve n or more
13 years spanning 1993 2003. Six gages used in the study had a streamflow record of less the five years (Appendix, Table 8). Basins not adjacent to the coast and without a direct, active outflow gage were merged with gaged basins to produce the a rea contributing to the streamflow gage. Of the 72 gages, 30 were in the northern area, covering 1990 square miles and 42 gages in the southern area, covering 4150 square miles. In general, downstream basins had a shallower depth to water table. A GIS coverage map of the study area shows all of the coastal basins in the model domain are within groundwater discharge areas based on a comparison of potentiometric heads and topography (Geurink et al., 2000; SWFWMD, 2006) (Figure 4).
14 Figure 3: USGS Gag e Locations and Sub basin Areas
15 Figure 4: Recharge Discharge Zones (Source: SWFWMD, 2006)
16 2.2 Methods Data collection for the study consisted of two distinct elements: streamflow and spatial geographic data. The acquisition of streamflow data was str aightforward in that the information is available at http://www.water.usgs.gov The United States Geological Survey (USGS) web site provides information for daily, monthly and annual streamflow. This study used calendar year annual streamflow records from the individual gaging stations. A Geographic Information Systems (GIS) was used to evaluate spatial geographic data files (Ormsby et al., 2004). Shape files for land use classification, location of gaging stati ons, soil classification, elevation and precipitation are all public domain GIS files from federal and local government agencies. Basin delineations were used from a previous study (Geurink et al., 2000). A total of 145 basins resulted from the spatial de lineation of that study (Figure 3). 2.2.1 Normalized Streamflow Fraction Normalized streamflow fraction (NSF) is defined as the runoff fraction of precipitation; it is obtained by dividing the average annual flow (Q mean ) by the product of the contributi ng area and average annual precipitation. Non dimensionalization allows representation of the variables not in terms of specific units, but instead, represents the variables relative to the parameters of the problem. For the streamflow fraction, it allows analysis of the basin characteristics to be independent of the area of the basin and the mean annual rainfall. For mean flow, Q flow, has units of [L 3 T 1 ]; A area, has units of [L 2 ]; P precipitation has units of [LT 1 ]; the normalized streamflow fract ion is: NSF = Q A 1 P 1 (1)
17 An 11 year streamflow and precipitation record was used for the study. The land use information utilized was from 1999 (SWFWMD, 2006). Using an eleven year period provided six years of record for streamflow and precipitation data prior to and four years of record following the land use record. 2.2.2 Streamflow The first step toward the calculation of the normalized streamflow fraction is to locate g aging stations that are at the outlets of the sub basin areas. Starting with over 100 potential stations, the number was reduced to 70 for several reasons. Due to temporal variability, station records showing 11 years of record, from 1993 2003, were pr eferable. Sixty one of the stations used in the study had a complete eleven year record. Less than a five year period of record were considered to be poor indicators of spatial and temporal variations of flow trends. The time span for the flow record wa s established based on the availability of the most recent (1999) land use classification shapefile. The study method requires a gaged sub basin; therefore stations located within a basin were eliminated. Several stations, although at the outlet of the s ub basin, accounted for less than 10 percent of the basin area, and were eliminated from the study. Two gages were representative of spring flow only and were also removed from the study. Streamstage gages allow the total amount of upstream flow passing through a cross section to be measured as streamflow. The unique contributing flow for each gage was isolated by subtracting the flow at that gage from any upstream gages. The area of each sub basin was calculated in GIS. Spring flow represents a signif icant portion of streamflow in the Hillsborough, Alafia and Weeki Wachee Rivers. Major spring flow in the river is also measured by the USGS. To analyze only normal streamflow (runoff and baseflow) from the
18 contributing basin, spring flow was removed fro m the annual average flows listed in Appendix, Table 8 by subtracting the measured spring flow from the USGS reported streamflow. Some streamflow gages had several sub basins flowing to a single outlet. In this case, multiple sub basins were aggregat ed (merged) into a new sub basin. The new sub basin was assigned land use attributes based on the ratio of each individual sub basins area to the total area of the new sub basin (i.e. relative percent area). 2.2.3 Precipitation Precipitation data from available NOAA stations were analyzed for the same time period as the streamflow records (Scott, 2006). The average precipitation for the study domain was found to be 52 inches (132 cm) per year, but some variability between the north and south domain wa s evident. Therefore, precipitation averages for the eleven year period were calculated for the northern and southern domains, which are roughly divided by Interstate 4 or a line between Tampa and Orlando. In the north part of the domain the annual prec ipitation for the study period was 51 inches (130 cm), while in the south the average for the same period of record was 55 inches (139 cm) (Table 1). In addition to the hydrogeologic variations between the two regions (Tihansky and Knochemus, 2001), the r ainfall variability also led to a distinction in the analysis of the two areas (Figure 5).
19 Figure 5: Precipitation Gages Locations (Source: Scott, 2006)
20 Table 1: Annual Precipitation Values Year North Annual Precipitation South Annual Precipitat ion 1993 41.74 50.25 1994 56.01 59.19 1995 55.60 62.36 1996 50.55 46.88 1997 60.30 57.50 1998 50.00 58.29 1999 42.36 49.06 2000 33.20 36.80 2001 48.09 54.93 2002 64.64 63.67 2003 57.50 61.23 11 year average 50.91 54.56 2.2.4 L and Use The 1999 land use distribution was obtained from the Southwest Florida Water Management District on line GIS database ( http://www.swfwmd.state.fl.us ). The land use and land cover features are categori zed according the Florida Land Use and Cover Classification System (FLUCCS) ( FDOT, 2006). The 53 FLUCCS codes were reduced to seven hydrologically significant classifications: urban, water/wetland, agricultural, grass and pastureland, disturbed and mining and forested. The first evaluation of land use characteristics was based on the area of the individual polygons (features) calculated by GIS for each land use type and exported to a spreadsheet for analysis. The total areas of the land use polygons wer e divided by the total area of the sub basins to obtain the percent of the sub basin for each land use category. This method (polygon analysis) was extremely inefficient. The second method used a raster image of land use from which the total number of ras ter cells of a specific classification were calculated and then divided by the total number of raster cells in the sub basin. The raster cells are 30 x 30 meters
21 and provided enough discretization to be an accurate representation of the land use areas. Th is method provided a more efficient method of obtaining the actual percentages of each land use category (Figure 6). In calculating the amount of impervious surface of a given sub basin, the urban and disturbed/mining sub classifications of land use type a re assigned a percent impervious value from the FLUCCS. The percent impervious area values were weighted based on the contributing area affected within the sub basin. A total weighted value for the sub basin was then calculated.
22 Figure 6: Land Use Clas sification Map
23 2.2.5 Soils The National Resource Conservation Service (NRCS) conducts field surveying and testing to establish consistent descriptions of soil characteristics, i.e. type, grain size, ect. A seasonal high high water table depth and a seas onal high low water table depth are defined for each soil type by NRCS in their database. The average depth to water table is not measured directly and was derived. To establish an average water table depth, 263 surficial monitor wells in the study domai n with approximately 122,000 observations between 1989 and 2001 were evaluated for fluctuations in water levels (Figure 7). The wells are maintained by the local water management district (SWFWMD) and the USGS. After reviewing the record for anomalous da ta, the average depth to water table and the standard deviation for each well were calculated on an annual basis. By summing the standard deviations, dividing by the number of wells and then doubling the value, the result is the average water table fluctua tion range value for each year. Using half of the average of the annual standard deviations and adding the value to the seasonal high low water table depth yields an annual average water table depth value for each soil type, summarized in Table 2. By exec uting a table join in GIS, a spatial map representing average depth to water table can be completed (Figure 8). From this spatial view, representative values of regionally shallow, medium and deep depth to water table are established for each sub basin as well as the percent of each sub basin with shallow, medium and deep water table environments.
24 Table 2: Depth to Water Table Statistics Soil Category North: % total area North: Sub basin % Range of Values South: % total area South: Sub b asin % Range of Values Shallow DTWT (less than 3 ft) 30 3 60 30 5 93 Medium DTWT (between 3 & 6 ft) 45 15 74 57 7 88 Deep DTWT (greater than 6 ft) 25 0 92 13 0 92 2.2.6 Slope Through GIS analysis, the slope across the domain has a range of 0.03% to 2.54% with only six of the 145 sub basins with greater than 1% slope; three of these are in the Withlacoochee River Basin. The mean (average) slope of the entire study domain is 0.38% with a standard deviation of 0.32. This low value indicates slope is n ot a significant contributor to runoff, although at the local scale, the Withlacoochee Basin in particular, slope may influence localized runoff processes in the West Central Florida environment.
25 Figure 7: Monitor Well Locations
26 Figure 8: Average D epth to Water Table Map and Wellfield Locations
27 2.2.7 Evapotranspiration Evapotranspiration (ET) is the combined process of evaporation of water directly from a surface and transpiration of water by vegetation. Some of the factors influencing ET are rainf all, temperature, vegetation type and depth to water table. Ruskauff et al. (2003) developed estimated long term ET rates for West Central Florida (Table 3) for use in an integrated groundwater surface water model. The study uses the same depth to water t able ranges as shown above in Table 2. A detailed estimated annual evapotranspiration (ET) analysis was completed for the study domain. Each land use type is assigned an estimated ET value; based on the percent of land use in a given sub basin, a sp atially averaged estimated ET is calculated for each of the sub basins. Dividing the estimated ET by the average precipitation yields a ratio of ET to rainfall. This ratio represents the percentage of precipitation that can not become part of the normali zed streamflow. Table 3: Estimated Annual Evapotranspiration Rates Evapotranspiration Rates (in/yr) Land Use Classifications Depth to Water Table Urban Pervious Urban Impervious Agricultural Grassland Forest Wetland Mining Shallow <3 ft 38 15 38 45 3 4 45 52 42 Medium 3 6 ft 34 15 34 45 30 40 52 42 Deep > 6 ft. 30 15 30 45 26 35 52 42 2.2.8 Statistical Analysis Various statistical methods are utilized in working with the data sets to establish relationships between the NSF and the test parameters: a direct one to one correlation, a stepwise regression, and a multivariate regression. Additionally, to
28 relate the significance of the NSF to the test parameters, correlation coefficients are computed for the resulting NSF equations. Stepwise regression uses a sequence of t tests to evaluate the significance of a variable. It alternates between adding and removing variables, checking the significance of individual variables within and outside the model. Variables that are significant when entering the model may be eliminated if later they test to be insignificant. Stepwise regression does not test all possible regression models (Helsel and Hirsch, 2002). A parametric approach to statistics relies on the data being normally distributed. Nonparametric m ethods can be employed on any data set. Nonparametric methods should be used only when the underlying distribution is unknown or cannot be transformed to make it normal (Berthouex and Brown, 1994). The results obtained in this study are considered nonpara metric. Kendalls tau, T k is a measure of correlation between the strength of the relationship between two variables, regardless of whether the relationship is increasing or decreasing. Tau measures the strength of the monotonic relationship between an ordered paired observation, X and Y A monotonic relationship shows one variable increasing wile the other variable always increases or always decreases. Tau is a rank based procedure and is therefore resistant to the effect of a small number of unusual (nonparametric) values. Tau is dependent on the ranks of the data, not the values themselves and can be used where the data is limited (Helsel and Hirsch, 2002). The T k values will generally be lower than values of the traditional correlation coefficient r a strong value of r is 0.9 or higher, the tau value corresponding to the same data set is about 0.7 (Helsel and Hirsch,2002).
29 Chapter Three: Results and Findings Dividing the study domain into two regions was considered appropriate due to the h ydrogeologic and rainfall variability. A variety of correlation statistics were used to find parameter relationships. One to one correlation between the normalized streamflow fraction and individual land use parameters, evapotranspiration estimates, and depth to water table variability yield limited results; therefore, multiple parameters were considered together in a stepwise regression to find relative strengths of the individual parameters when grouped together. A stepwise regression of multiple param eters and a multivariate regression gave strong results in the northern part of the study domain. The southern domain did not yield meaningful results with any of the above methods. However, evapotranspiration rates based on average depth to water table and land use types did yield meaningful results in part of the southern coastal domain. 3.1 Depth to Water Table The 13 year history of the water table fluctuations shows an average fluctuation across the study domain of approximately four feet, 3.6 fee t precisely. Both the northern and southern parts of the study area show 30 percent of the entire land area has an average depth to water table from land surface to three feet deep. Therefore, the possibility exists for 30 percent (approximately 3000 squar e miles) of the study domain to have the water table at or above land surface during the wet season.
30 3.2 Northern Domain 3.2.1 One to One Correlation with NSF and Land Use Classifications Forested Areas: In the northern domain, forested land cover rang es from one percent to 65 percent of the sub basin area. Thirteen of the sixty sub basins have greater than 25 percent forest coverage. Of the 13, seven sub basins are in the Withlacoochee River Basin, an interior region of the domain with an urbanization of less than eight percent of the land area. Five of the 13 sub basins are found in the spring fed Weechi Wachee River Area. No direct correlation was found between the NSF and the percentage of forested area within a sub basin. Agricultural Areas: T he range of agricultural land use in the northern domain varies from less than one percent in 6 sub basins to 24 percent, with only two sub basins south of the Hillsborough River having agriculture greater than 20 percent of the total land use. The averag e was five percent agricultural land use. Agricultural land use did not show any correlation with the NSF in this domain. Mixed/Disturbed/Mining Area : The range of mixed/mining land use in the northern domain is from less than one percent in 30 of the 60 sub basins to 13 percent. Only one basin shows greater than six percent of the land use in mining. This land use category did not impact the study in this domain. Urbanized Area: The degree of urbanization in the study domain varies from less than one percent in a Withlacoochee River interior basin to 89 percent along the Hillsborough River. The ungaged coastal areas in the northern study area range from 24 to 82 percent urbanization. When correlating the NSF to urbanization greater than 25 percent a simple linear regression shows a 67 percent correlation between
31 the parameters (Figure 9) and a rank correlation, T k of 0.357, showing a less than 90% probability that the variables are correlated. T(k)=0.36 R 2 = 0.67 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.20 0.40 0.60 0.80 1.00 Fraction of Urban Area Normalized Streamflow Fraction Figure 9: One to One Correlation: Northern Domain: Urbanization with Wellfields Nine sub basin areas in the northern domain show significant physical responses to the wellfields and supplemental pumping that takes place within the basin boundaries. Wellfield pumping export s water from the sub basin and is not found along the coast in the ungaged sub basins. In the investigation of northern domain, removing the sub basins directly influenced by pumping from the analysis resulted in the relationship between normalized stream flow fraction and percent of urban area increasing to a simple linear correlation of 81 percent (Figure 10) and a rank correlation, T k of 0.515, showing 99% confidence the variables are correlated.
32 T(k)=.52 R 2 = 0.81 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.20 0.40 0.60 0.80 Fraction of Urban Area in the Sub-basin Normalized Streamflow Fraction Figure 10: One to One Corr elation: Northern Domain: Urbanization without Wellfield Pumping Sub basins Grassland and Pasture Area: Grass and pastureland varies from one percent to 55 percent of the land use in the northern study area. The coastal basins range from three percent to 21 percent of land use in this classification. The ungaged coastal basins are not pumped directly, so in assessing likely relationships, areas with pumping stresses were not considered. In evaluating the sub basins that do not have direct pumping stre sses due to wellfields, an inverse relationship was found between the percent of grassland and the normalized streamflow fraction of 65 percent using a simple linear correlation (Figure 11) and a T k rank correlation of ( 0.477), an inverse correlation sho wing a 90% probability the variables are correlated.
33 T(k)=-0.48 R 2 = 0.65 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 Fraction of Grassland or Pasture Normalized Streamflow Fraction Figure 11: One to One Correlation: Northern Domain: Grassland without Wellfields Water and Wetlands Area: Sub basins in the domain vary in open water/wetlands land use f rom 1 1/2 percent to 54 percent of the total area. The ungaged coastal basins vary from 17 to 45 percent. A one to one correlation between the NSF and water/wetlands in non pumping sub basins yields an inverse relationship linearly correlated to 65 per cent (Figure 12) and a rank correlation of ( 0.692), an inverse correlation showing greater than a 99% probability the variables are correlated. In evaluating sub basins containing, the opposite relationship is found; a greater water/wetland area indicate s an increase in the normalized streamflow fraction with a one to one linear correlation of 84 percent (Figure 13) and a T k rank correlation of 0.905, showing a 99% probability the variables are correlated.
34 T(k)=-0.69 R 2 = 0.65 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 0.60 Fraction of Wetlands Area Normalized Streamflow Fraction Figure 12: One to One Correlation: Northern Domain: Open Water and Wetlands without Wellfield Pumping Sub basins T(k)=0.91 R 2 = 0.84 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.10 0.20 0.30 0.40 0.50 Fraction of Wetlands Area Normalized Streamflow Fraction Figure 13: One to One Correlation: Northern Domain: Open Water and Wetlands in Sub basins with Wellfields Shallow Depth to W ater Table: A shallow depth to water table (three feet or less from land surface) exists across one third of the total area of the northern study domain. The sub basins range from three to 60 percent of the area with a shallow DTWT; non pumping sub basin s vary between four and 34 percent. A simple linear
35 correlation shows an insignificant increasing relationship between shallow DTWT and the NSF and a T k rank correlation of 0.538. T(k)=0.54 R 2 = 0.47 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Fraction of Shallow Depth-to-Water Table Normalized Streamflow Fraction Figure 14: One to One Correlation: Northern Domain: Shallow Depth to Water Table without Wellfields 3.2.2 Stepwise Regression The sum of the percentages of urban, grass and water/wetlands was greater than 58 percent of the land use in all but two of the 60 northern domain sub basins. The ungaged coastal basins range from 69 to 95 percent of the total land use in the three classifications. With this as a guideline, a stepwise regression using a commercial statistical software package was performed to assess the rank of the multiple parameters on the known normalized streamflow fractions. The stepwise regression ordered the most influential parameters to the normalized streamflow fraction as the percent of urbanization, the percent of shallow depth to water table and the percent of grasslands wit hin the sub basin. Surprisingly, open water and wetlands did not appear as an influential parameter in the NSF and when DTWT is in combination with urbanization and grasslands, the relative importance is much greater than the simple linear correlation ind icates.
36 3.2.3 Multivariate Linear Regression Using the stepwise regression as a guide for influential factors, further analysis was yielded a strong correlation with multivariate linear regression, producing the following equation: y = 1.0415 x (2) where y represents the normalized streamflow fraction and x represents the weighed factors of U: urban percent, DTWT: shallow depth to water tabl e percent and G: grassland percent: x = 3446 *[U] 19 *[DTWT] 16 *[G] 19 (3) The equation yields a linear correlation value of 0.80 and a T k rank correlation of 0.714, indicating a confidence level of greater than 99%, when used with sub basins that do not have municipal wellfield pumping stresses in the northern domain (Figure 15). T(k)=0.71 R 2 = 0.80 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 0.60 Weighted Urban, DTWT, Grassland Normalized Streamflow Fraction Figure 15: Multivariate Regression Results
37 In using the regression equation with the limitations of non pumping basins in the norther n domain, the mean absolute error was 5.5 percent when applied to actual data from the domain (Table 4). The predicted values using the regression equation as compared to the area scaling method resulted in a stream flow value greater than the estimation of the area scaling except in one case (Figure 16). This was an anomalous basin in the northern domain that contains the Tampa Bypass Canal, a large Army Corps of Engineers storm water mitigation project. Table 4: Results of Actual NSF Vs Predicted NSF f or Northern Sub basins Sub basin Actual NSF Predicted NSF Absolute Difference 73 0.338 0.277 0.061 74 0.256 0.235 0.021 79 0.389 0.235 0.154 88 0.364 0.251 0.113 94 0.519 0.551 0.033 111 0.180 0.203 0.023 117 0.427 0.390 0.037 133 0.266 0.324 0.059 134 0.357 0.240 0.117 135 0.111 0.196 0.084 138 0.192 0.188 0.003 139 0.196 0.209 0.013 140 0.153 0.109 0.044 153 0.118 0.165 0.047 154 0.162 0.216 0.054 155 0.241 0.263 0.022
38 0 20 40 60 80 100 0 20 40 60 80 100 Area Scaling Predicted NSF Regression Equation Predicted NSF Figure 16: Comparison of Regression Prediction vs Area Scaling Met hod To verify the regression equation, seven USGS gaging stations not used in the study were tested. The stations were across five sub basins in the northern domain and yielded a T k rank correlation of 0.81, indicating a probability of greater than 90% that the variables will correlate, with an 8.3 percent variance in the actual NSF to the predicted value using the regression equation (Figure 17). 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 Actual NSF Predicted NSF Figure 17: Test Gages and Sub basins in Northern Domain
39 3.3 Southern Domain The southern domain had 6 sub basins with greater than 30 percent of the land use being disturbed or mining. The sub basins were removed from the regression analysis because this land use class significantly alters streamflow runoff patterns and evapotranspiration. Therefore, i t is not relevant to the physical attributes of the targeted ungaged coastal basins to include those sub basins. Seven sub basin areas in the southern domain had large municipal pumping stresses attributed to the area. 3.3.1 One to One Correlation with NS F and Land Use Classifications The southern portion of the study domain did not show any significant correlations between the normalized streamflow fraction and a single land use type using one to one analysis. A depth to water table and normalized strea mflow fraction analysis did not show any noteworthy relationships. 3.3.2 Stepwise Regression Using a stepwise regression to find relationships between multiple land use types, depth to water table and the NSF did not yield any significant results for the southern domain stations. 3.3.3 Evapotranspiration With almost 90 percent of the area having the water table within six feet of land surface, evapotranspiration (ET) rates directly affect streamflow. With an extensive evapotranspiration analysis, the southern domain does start to show some correlation. The ET analysis incorporates all of the land use categories as well as shallow, medium and deep depth to water table spatial information.
40 When looking at the gaged basins adjacent to the ungaged basin s, an R 2 = .79 and a rank correlation Kendall Tau of ( .71) was attained (Figure 18) reporting with a probability of 95% for the variables to correlate. The equation used to extrapolate the normalized streamflow fraction to the ungaged region was: y = 1.4 x + 1.288 (4) where y in the NSF and x is the ET Precipitation Ratio. A mean absolute error of 7.3 percent occurs when measuring the predicted values against the actual values of normalized streamflow (Table 5). T(k)=-0.71 R 2 = 0.79 0.00 0.10 0.20 0.30 0.40 0.50 0.55 0.60 0.65 0.70 0.75 ET/Precipitation Ratio Normalized Streamflow Fraction Figure 18: Southern Domain: Regression Equation
41 Table 5: Results of Actual NSF vs Predicted NSF for Southern Coastal Basins Sub basin Actual NSF Predicted NSF Absolute Difference 40 0.443 0.454 0.011 41 0.246 0.296 0.049 49 0.350 0.430 0.080 52 0.298 0.329 0.031 54 0.378 0.377 0.000 55 0.394 0.434 0.039 65 0.434 0.382 0.052 165 0.399 0.391 0.008 166 0.175 0.565 0.391 When comparing the predicted results from the regression equation to the predicted results from the traditional area scaling method (Figure 19), the regression equation predicts an average of 3 percent higher for ungaged basins surrounding the Little Manatee River. 0.00 0.10 0.20 0.30 0.40 0.50 0 0.1 0.2 0.3 0.4 0.5 Area Scaling Prediction of NSF Regression Equation Prediction of NSF Figure 19: Southern Domain: Around the Little Manatee River: Predicted vs Area Scaling Method
42 3.3. 4 Charlotte Harbor Study of Ungaged Sub basins The upper Charlotte Harbor study (Ross et al., 2005) estimated runoff to precipitation ratios similar to the normalized streamflow fraction. Precipitation rates used were localized to each basin. Runoff was found by using the Hydrologic Simulation Program (HSPF) (Bicknell et al., 2001) to obtain runoff volume and then dividing by the localized precipitation resulting in the runoff ratio. The Charlotte Harbor study used average DTWT across the domain with an average estimated ET. A comparison of the results from the Charlotte Harbor study to the NSF calculated using estimated ET with a variable DTWT yielded similar results for 5 of the 8 sub basins (Table 6); the range of all 5 of the predictions are within 7 percent of the estimated runoff fraction. The 3 sub basins with varied results are all located on the Gulf Coast, west of the Myakka River (Figure 20). The sub basins showed higher runoff fractions using the ET regression equation than the HSPF model p redicted but overall the results fell within an 8.4 percent difference in the two predicted values. Table 4 shows the NSF method giving higher values than the basin ratio method. This may be attributed to the assumption of the basin ratio that ungaged bas in yield is the same discharge rate as the gaged part of the basin (Levesque and Hammett, 1997). Table 6: Charlotte Harbor Study Using HSPF vs NSF Equation S ub basin NSF Basin Runoff Ratio Absolute D ifference 24 0.244 0.232 0.012 26 0.292 0.298 0.0 06 29 0.273 0.204 0.069 30 0.295 0.294 0.001 36 0.309 0.259 0.050 44 0.475 0.289 0.186 46 0.398 0.240 0.158 48 0.457 0.270 0.187
43 3.4 Complete Study Domain Using the multivariate regression equation (Equation 2) for the northern domain ungaged basi ns and the linear relationship between ET and the normalized streamflow fraction in the southern domain, a map showing the normalized streamflow fraction (Figure 20) of the entire study area shows trends of increasing streamflow fraction toward the coastli ne. The two anomalies in the north (dark blue sub basins, NSF: 0.6 0.7) are representative of spring flow basins. The yellow in the northern domain is consistent with wellfield pumping basins.
44 Figure 20: Normalized Streamflow Fraction
45 Chapter Four: Discussion and Conclusions 4.1 Northern Domain Three analysis techniques were used to obtain estimated flow in ungaged coastal catchments for the northern study domain: 1) a one to one correlation between the normalized streamflow fraction and indi vidual parameters, 2) a stepwise regression to find the most significant parameters affecting runoff and 3) a multivariate regression matrix to obtain exponents for the most significant parameters. Geologic variations between the northern and southern d omains have given rise to different runoff patterns. In the north, areas with the limestone aquifer at or near land surface experiences increased recharge to the deep groundwater system, thereby reducing the normalized streamflow fraction. Increased isol ated surface depressions, by dissolution of subsidence, is the primary mechanism for removing surface water from the surface water runoff system, taking water directly into the groundwater system. Karst areas are problematic but were not removed from the analysis and the prediction method for the northern (karst) region seemed to work as well for karst areas as well as non karst areas in the north. However, the NSF analysis is for streamflow minus spring flow and is only applicable when the contributing s pring flow is not included in the streamflow quantity. Wellfield pumping creates lower than expected streamflow and skews the NSF significantly. Increased streamflow fractions are found in areas with greater than 25 percent urbanization. The innate i mperviousness associated with urbanization
46 decreases the available surface area for infiltration leaving precipitation with nowhere to go but to the surface water system. The northern domain shows very strong correlations to urban land use, a seasonal hig h shallow depth to water table environment and the percent of grass or pasture land. Surprisingly, wetland percentage does not play a significant role in the controlling factors of the normalized streamflow fraction. Extrapolation of the normalized strea mflow fraction to the ungaged regions using multivariate regression provided a more practical and accurate value to use versus the area scaling method. High variability in the NSF indicates that area scaling methods are less reliable closer to the coast a s the region transitions to a groundwater discharge. Area scaling methods under predict NSF. 4.2 Southern Domain The southern domain presented an inconsistent pattern of parameters to be identified as an indicator of the normalized streamflow fractio n. The methods employed in the northern domain, one to one correlation and stepwise regression, did not show any meaningful relationships with the NSF in the southern domain. Using sub basins that were geographically close and an ET to precipitation ratio yielded the most accurate results for this domain. Evapotranspiration, depth to water table and land use are all closely related to determining the NSF in the coastal sub basins in the southern domain. A multivariate regression matrix was used to obtain t he NSF equation for this area. The southern domain has a higher degree of aquifer confinement with less recharge, therefore evapotranspiration influences more directly the surface water system producing a better relationship with streamflow. Agricultural pumping could account for some of the variability, but was not considered in this study. This type of
47 pumping varies from municipal groundwater pumping in that it is not exported and stays as part of the water budget directly influencing runoff and ET in this relatively shallow water table environment. 4.3 Complete Study Area The normalized streamflow fraction values are similar across the entire study domain but the relationship to land use and depth to water table varies significantly. The slope var iability and percent of wetlands did not seem to play a significant role in runoff across the basin domain. Soil type is important from the standpoint of determining the depth to water table. All of the land use classifications play a role in determining the ET values in the south. In the north, only urban and grasslands are significant. The study was limited in scope to West Central Florida. Groundwater recharge and discharge areas were not directly correlated to the sub basins. Agricultural pumping im pacts were not addressed.
48 References Bicknell, B.R., Imhoff, J.C., Kittle, J.L., Jobes, T.H., Donigan, A.S., 2001. HSPF Users Manual, Aqua Terra Consultants, Mountain View, California. Berthouex, P.M., Brown,L.C., 1994. Statistics for Environment al Engineers, CRC Press, LLC. Boughton, W.C., 1984. A Simple Model for Estimating the Water Yield of Ungauged Catchments, Civil Engineering Transactions, p. 83 88. Carlile, P.W., Jakeman, A.J., Croake, B.F.W., Lees, B.G., 2002. Use of Catchment Attribu tes to Identify the Scale and Values of Distributed Parameters in Surface and Sub Surface Conceptual Hydrology Models, International Environmental Modelling and Software Society Conference. Cormary, Y., Guilbot, A. 1973. Etude des relations pluie dbit su r trois basins versants dinvestigation. Proceedings of the IAHS Madrid Symposium, IAHS Publ., 108 265 279. Edijanto, Michel, C., 1989. A Model Rainflow with Three Parameters, Hydro Electric Power, no. 2, p. 113 121. Florida Department of Transportatio n, 2006. http://www.dot.state.fl.us. Girgin, S., Akyurek, Z., Usul, N., XXXX. Development of GIS Based Streamflow Analysis System for Turkey, Middle East Technical University, Ankara, Turkey, Department of Civil Engineering. Grayson, R.B., Moore, I.D., M cMahon, T.A., 1992. Physically Based Hydrologic Modeling: Is the Concept Realistic?, Water Resources Research, vol. 26, no. 10, p. 2659 2666. Geurink, J.S., Nachabe, M., Ross, M.A., Tara, P., 2000. Development of Interfacial Boundary Conditions for the S outhern District Ground Water Model of the Southwest Florida Water Management District, Center for Modeling Hydrologic and Aquatic Systems, Report # CMHAS.SWFWMD.00.03. Hammett, K.M., DelCharco, M.J., 2005. Estimating the Magnitude and Frequency of Flood s for Streams in West Central Florida, 2001, U.S. Geological Survey Report 2005 5080. Helsel, D.R., Hirsch, R.M., 2002. Statistical Methods in Water Resources, Techniques of Water Resource Investigations of the U.S.G.S. http://water.usgs.gov/pubs/twri/twr i4a3.
49 Hundecha, Y., Zehe, E., Bardossy, A., 2003. Regional Parameter Estimation from Catchment Properties for the Prediction of Ungauged Basins, International Association of Hydrological Sciences Conference Workshop. Jakeman, A.J., Littlewood, I.G., Whit ehead, P.G., 1990. Computation of the Instantaneous Unit Hydrograph and Identifiable Component Flows with Application to Two Small Upland Catchments. Journal of Hydrology vol. 117, p. 275 300. Kokkonen, T.S., Jakeman, A.J., Young, P.C., Koivusalo, H.J., 2003. Predicting Daily Flows in Ungauged Catchments: Model Regionalization from Catchment Descriptors at the Coweeta Hydrologic Laboratory, North Carolina, Hydrological Processes vol. 17, no. 11, p. 2219 2238. Levesque, V., Hammett, K.M., 1997. Compariso n of Two Methods for Estimating Discharge and Nutrient Loads from Tidally Affected Reaches of the Myakka and Peace Rivers, West Central Florida, USGS,97 118. Littlewood, I.G., Jakeman, A.J., 1994. A New Method of Rainfall runoff Modelling and its Applicati ons in Catchment Hydrology. In: Environmental Modelling Computational Mechanics Publications Southampton vol. 2, p. 143 171. Liu, J., Ding, Y., Zhou, X., Wang J., 2002. Land Surface Hydrology Parameterization over Heterogeneous Surfaces for the study o f Regional Mean Runoff Ratio with its Simulations, Advances in Atmospheric Sciences, vol. 19, no. 1, p. 89 102. McNamara, J.P., Kane, D.L., Hinzman, L.D., 1998. An Analysis of Streamflow Hydrology in the Kuparuk River Basin, Arctic Alaska: A Nested Waters hed Approach, Journal of Hydrology, vol. 206, p. 39 57. Miller, S.N., Guertin, D.P., Goodrich, D.C., 2005. Investigating Stream Channel Morphology Using GIS, http:// uwadmnweb.uwyo.edu/ renewableResources/soil/Miller_Scott.htm. Mwakalila, S., 2003. Est imation of Streamflows of Ungauged Catchments for River Basin Management, Physics and Chemistry of the Earth, vol. 28, p. 935 942. National Oceanic and Atmospheric Administration. National Resources Conservation Service. Nathan, R.J., McMahon, T.A., 19 90. Identification of Homogeneous Regions for the Purposes of Regionalisation, Journal of Hydrology, vol. 121, p. 217 238. Ormsby, T., Napoleon, E., Burke, R., Groessl, C., Feaster, L., 2004. Getting to Know ArcGIS Desktop. ESRI Press, Redlands, Claiforn ia.
50 Pandey, G.R., Nguyen, V.T.V., 1999. A Comparative Study of Regression Based Methods in Regional Flood Frequency Analysis, Journal of Hydrology, vol. 225, p. 92 101. Post, D.A., Jakeman, A.J., 1999. Predicting the Daily Streamflow of Ungauged Catchemt ns in S.E. Australia by Regionalising the Parameters of a Lumped Conceptual Rainfall Runoff Model, Ecological Modelling, vol. 123, p. 91 104. Phipps, S.P., Robbins, C., XXXX. Integrating GIS for Better Hydrologic and Hydraulic Modeling, Woolpert, Inc., Da yton, Ohio. Ries, K.G., Gray, J.R., 2005. StreamStats: A U.S. Geological Survey Web Site for Stream Information, http://water.usgs.gov/osw/programs/streamstats.html. Ross, M.A., Said, A., Trout, K., Zhang, J., 2005. Hydrologic Modeling of Streamflow from Ungaged Areas in the Upper Charlotte Harbor Basin, University of South Florida, Tampa, Fl. Ruskauff, G., Aly, A., Ewing, J., Jobes, T., Donigan, A., Tara, P., Trout, K., Ross, M., 2003. The Integrated Northern Tampa Bay Hydrologic Model (INTB), Volume 3, Tampa Bay Water, Tampa, Fl. Scott. M.H., 2006. Precipitation Variability of Streamflow Fraction in West Central Florida, University of South Florida, Tampa, Fl. Sefton, C.E.M., Howarth, S.M., 1998. Relationships between Dynamic Response Characteristics and Physical Descriptors of Catchments in England and Wales, Journal of Hydrology, vol. 211, p. 1 16. Servat, E., Dezetter, A., 1993. Rainfall Runoff Modelling and Water Resources Assessment in Northwestern Ivory Coast. Tentative Extension to Ungauged Cat chments, Journal of Hydrology, vol. 148, p. 231 248. Southwest Florida Water Management District, http://www.swfwmd.state.fl.us Tihansky, A.B., Knochemus, L.A., 2001. Karst Features and Hydrogeology in West Central Florida A Field Perspective, U.S. Geological Survey Report, Tampa, Fl. United State Geological Survey, http://www.usgs.gov or http:// nwis.waterdata.usgs.gov/usa/nwis/annual/calendar_year Whelan, F., 1999, Assessment of Hydrologic Properties and Land Use/ Land Cover in Pacific Northwest Watersheds through Hydrologic Modeling Utilizing GIS, Oregon State University, Department of Geosci ences. Wooldridge, S., Kalma, J., Kuczera, G., 2001. Parameterisation of a Simple Semi distributed Model for Assessing the Impact of Land use on Hydrologic Response, Journal of Hydrology, vol. 254, p. 16 32. Woolheiser, D.A., 1996. Search for a Physical ly Based Runoff Model A Hydrologic El Dorado?, Journal of Hydraulic Engineering, ASCE, vol. 122, p. 122 129.
52 Appendix A: Tables Table 7: United States Geological Survey Gag ing Station Locations Station Station Name 2236350 GREEN SWAMP RUN NEAR EVA 2256500 FISHEATING CREEK AT PALMDALE 2267000 CATFISH CREEK NEAR LAKE WALES 2268390 TIGER CREEK NEAR BABSON PARK, FL 2269520 LIVINGSTON CREEK NEAR FROSTPROOF, FL 2270000 CAR TER CREEK NEAR SEBRING 2270500 ARBUCKLE CRREK NEAR DESOTO CITY 2271500 JOSEPHINE CREEK NEAR DE SOTO CITY, FL 2293987 PEACE CREEK DRAINAGE CANAL NEAR WAHNETA 2294217 SADDLE CREEK AT STATE HIGHWAY 542 NEAR LAKELAND, FL 2294491 SADDLE CREEK AT STRUCTURE P11 NEAR BARTOW 2294650 PEACE RIVER AT BARTOW 2294898 PEACE RIVER AT FT MEADE 2295013 BOWLEGS CREEK NEAR FORT MEADE, FL 2295420 PAYNE CREEK NEAR BOWLING GREEN, FL 2295637 PEACE RIVER AT ZOLFO SPRINGS 2296500 CHARLIE CREEK NEAR GARDNER 2296750 PEACE RIVER AT ARCADIA 2297100 JOSHUA CREEK AT NOCATEE, FL 2297155 HORSE CREEK NEAR MYAKKA HEAD, FL 2297310 HORSE CREEK NEAR ARCADIA, FL 2298123 PRAIRIE CREEK NEAR FORT OGDEN, FL 2298202 SHELL CREEK NEAR PUNTA GORDA 2298608 MYAKKA RIVER AT MYAKKA CITY 229 8830 MYAKKA RIVER NEAR SARASOTA 2298928 TRIBUTARY TO MYAKKA RIVER NEAR VENICE 2299120 DEER PRAIRIE SLOUGH AT POWER LINE NEAR NORTH PORT 2299410 BIG SLOUGH CANAL NEAR MYAKKA CITY 2299450 BIG SLOUGH AT TROPICARE BLVD 2299737 SOUTH CREEK NEAR VAMO 2299 780 PHILLIPPI CREEK NEAR BEE RIDGE 2299861 WALKER CREEK AT SARASOTA, FL 2299950 MANATEE RIVER NEAR MYAKKA HEAD 2300018 GAMBLE CREEK NEAR PARRISH, FL 2300032 BRADEN RIVER NEAR LORRAINE, FL 2300042 WARD LAKE OUTFALL NEAR BRANDON 2300100 LITTLE MANATEE RIVER NEAR FORT LONESOME, FL 2300500 LITTLE MANATEE RIVER NEAR WIMAUMA 2300700 BULLFROG CREEK NEAR WIMAUMA, FL 2301000 NORTH PRONG ALAFIA RIVER AT KEYSVILLE 2301300 SOUTH PRONG ALAFIA RIVER NEAR LITHIA
53 Appendix A: (Continued) Table 7: (Continued) 2301500 ALAFIA RIVER AT LITHIA 2301750 DELANEY CREEK NEAR TAMPA, FL 2301900 FOX BRANCH NEAR SOCRUM 2301990 HILLSBOROUGH RIVER AT CRYSTAL SPRINGS 2302500 BLACKWATER CREEK NEAR KNIGHTS, FL 2303000 HILLSBOROUGH RIVER NEAR ZEPHYRHILLS 2303205 BAKER CREEK AT MCINTOSH ROAD NEAR ANTIOCH, FL 2303330 HILLSBOROUGH RIVER AT MORRIS BRIDGE NEAR THONTASASSSA 2303350 TROUT CREEK NEAR SULPHUR SPRINGS 2303420 CYPRESS CREEK AT WORTHINGTON GARDENS, FL 2303800 CYPRESS CREEK NEAR SULPHUR SPRINGS 2304500 HILLSBOROUGH RIVER NEAR TAMPA 2306000 SWEETWATER CREEK NEAR SULPHUR SPRINGS, FL 2306647 SWEETWATER CREEK NEAR TAMPA, FL 2307000 ROCKY CREEK NEAR SULPHUR SPRINGS, FL 2307200 BROOKER CREEEK AT VAN DYKE ROAD NEAR CITRUS PARK 2307359 BROOKER CREEK NEAR TARPON SPRINGS 2309848 SOUTH BRANCH ANCLOTE RIVER NEAR ODESSA 2310000 ANCLOTE RIVER NEAR ELFERS 2310147 HOLLIN CREEK NEAR TARPON SPRINGS, FL 2310280 PITHLACHASCOTEE RIVER NEAR FIVAY JUNCTION 2310300 PITHLACHASCOTEE RIVER NEAR NEW PORT RICHEY 2310525 WEEKI WACHEE R IVER NEAR BROOKSVILLE, FL 2310545 WEEKI WACHEE RIVER NEAR WEEKI WACHEE, FL 2310678 HOMOSASSA SPRINGS AT HOMOSASSA SPRINGS 2310947 WITHLACOOCHEE RIVER NEAR CUMPRESSCO 2311500 WITHLACOOCHEE RIVER NEAR DADE CITY 2312000 WITHLACOOCHEE RIVER AT TRILBY 231 2180 L WITHLACOOCHEE RIVER NEAR TARRYTOWN 2312200 L WITHLACOOCHEE RIVER AT RERDELL 2312500 WITHLACOOCHEE RIVER AT CROOM
54 Appendix A: (Continued) Table 8: Gaging Station Statistics Station Basin Number Mean Annual Flow (cfs) Area (sq mile ) Region 2236350 135 17.4 41.830 North 2256500 110 285.5 308.59 South 2267000 100 36.8 46.05 South 2268390 101 38.3 53.06 South 2269520 103 61.4 118.32 South 2270000 104 20.1 38.96 South 2270500 105 306.7 231.25 South 2271500 161 65.3 113.22 South 2293987 3 98.4 170.74 South 2294217 1 40.4 59.54 South 2294491 2 41.4 85.82 South 2294650 4 133.4 88.57 South 2294898 5 46.9 74.91 South 2295013 6 29.8 46.29 South 2295420 9 138.8 125.2 South 2295637 167 180.4 188.02 South 2296500 162 297.3 326.47 South 2296750 163 189.1 207.39 South 2297100 25 155.8 120.94 South 2297155 21 32.3 40.93 South 2297310 164 200.5 176.4 South 2298123 27 251.8 223.02 South 2298202 28 167.7 145.83 South 2298608 31 197.4 124.06 South 2298830 32 118.5 101.47 South 2 298928 37 12.2 49.98 South 2299120 33 37.8 26.14 South 2299410 34 47.2 35.83 South 2299450 35 110.2 50 South 2299737 41 16.1 16.25 South 2299780 40 55.3 31.07 South 2299861 39 6.7 6 South 2299950 49 93.6 66.54 South 2300018 52 72.3 60.36 South 230 0032 54 38.3 25.21 South 2300042 55 52.8 33.29 South 2300100 59 37.5 30.9 South 2300500 165 193.9 120.84 South
55 Appendix A: (Continued) Table 8: (Continued) 2300700 65 49.8 28.55 South 2301000 67 148.5 136.03 South 2301300 68 106.5 112.25 South 2301500 166 63.9 91.13 South 2301750 98 10.1 14.21 South 2301900 73 11.7 9.280 North 2301990 74 73.4 76.470 North 2302500 155 89.2 98.610 North 2303000 79 62.8 42.980 North 2303205 81 19.7 21.630 North 2303330 157 177.6 54.620 North 2303350 88 23. 5 17.250 North 2303420 158 46.9 128.670 North 2303800 91 72.4 39.120 North 2304500 156 48.9 139.880 North 2306000 94 31.6 16.240 North 2306647 117 23.0 14.380 North 2307000 152 39.7 46.640 North 2307200 111 3.5 5.220 North 2307359 151 11.7 27.830 N orth 2309848 121 5.3 13.180 North 2310000 122 49.3 56.420 North 2310147 123 5.6 21.800 North 2310280 125 5.2 148.840 North 2310300 126 16.2 32.560 North 2310525 132 153.8 10.440 North 2310545 133 89.5 27.600 North 2310678 134 94.5 70.570 North 231 0947 153 156.1 352.830 North 2311500 138 48.1 66.830 North 2312000 139 112.5 152.960 North 2312180 140 50.6 87.970 North 2312200 141 83.3 50.610 North 2312500 154 70.9 116.920 North
56 Appendix A: (Continued) Table 9: Land Use Classificatio n Percentages by Sub basin Land Use Classification Sub basin Wetland/Water Urban Mining/Disturbed Grassland Forested Agricultural 1 0.243 0.274 0.220 0.096 0.032 0.136 2 0.260 0.313 0.147 0.143 0.056 0.082 3 0.313 0.219 0.003 0.246 0.046 0.173 4 0. 177 0.145 0.174 0.255 0.037 0.212 5 0.125 0.076 0.540 0.100 0.011 0.148 6 0.252 0.044 0.001 0.437 0.088 0.178 7 0.138 0.005 0.394 0.306 0.027 0.130 8 0.033 0.008 0.894 0.027 0.002 0.036 9 0.111 0.011 0.627 0.106 0.057 0.087 10 0.168 0.009 0.035 0.575 0.041 0.172 11 0.169 0.122 0.076 0.346 0.078 0.209 12 0.217 0.039 0.010 0.418 0.118 0.198 13 0.232 0.054 0.006 0.420 0.166 0.122 14 0.117 0.012 0.003 0.544 0.050 0.274 15 0.166 0.000 0.000 0.607 0.065 0.162 16 0.157 0.000 0.000 0.647 0.084 0.113 17 0.201 0.001 0.000 0.555 0.115 0.127 18 0.225 0.014 0.005 0.367 0.090 0.298 19 0.156 0.008 0.000 0.515 0.157 0.163 20 0.241 0.038 0.002 0.423 0.144 0.152 21 0.132 0.001 0.295 0.433 0.091 0.048 22 0.195 0.003 0.035 0.487 0.235 0.044 23 0.187 0.016 0.0 00 0.525 0.126 0.146 24 0.250 0.036 0.019 0.407 0.201 0.087 25 0.109 0.040 0.005 0.492 0.050 0.304 26 0.241 0.284 0.002 0.259 0.098 0.116 27 0.178 0.002 0.001 0.547 0.034 0.238 28 0.119 0.055 0.005 0.408 0.204 0.210 29 0.169 0.061 0.003 0.434 0.303 0 .030 30 0.370 0.146 0.011 0.255 0.200 0.017 31 0.167 0.025 0.018 0.478 0.140 0.172 32 0.291 0.062 0.001 0.531 0.063 0.052 33 0.368 0.005 0.000 0.363 0.265 0.000 34 0.218 0.005 0.002 0.570 0.125 0.080 35 0.194 0.122 0.000 0.459 0.185 0.040 36 0.189 0 .073 0.002 0.507 0.206 0.022 37 0.257 0.114 0.011 0.309 0.296 0.013 38 0.115 0.676 0.001 0.063 0.030 0.115 39 0.074 0.784 0.000 0.036 0.051 0.056 40 0.088 0.620 0.022 0.125 0.059 0.086 41 0.229 0.147 0.001 0.361 0.210 0.051
57 Appendix A: (Continued) T able 9: (Continued) 42 0.229 0.134 0.007 0.359 0.196 0.075 43 0.128 0.613 0.001 0.088 0.069 0.101 44 0.130 0.605 0.006 0.118 0.088 0.053 45 0.195 0.286 0.002 0.312 0.188 0.016 46 0.174 0.258 0.001 0.384 0.158 0.024 47 0.200 0.235 0.000 0.192 0.371 0. 002 48 0.154 0.300 0.015 0.324 0.161 0.045 49 0.113 0.004 0.059 0.527 0.177 0.120 50 0.096 0.002 0.000 0.596 0.050 0.256 51 0.173 0.076 0.000 0.286 0.153 0.312 52 0.209 0.034 0.002 0.370 0.062 0.323 53 0.160 0.145 0.002 0.350 0.050 0.293 54 0.132 0. 084 0.005 0.467 0.142 0.170 55 0.212 0.260 0.085 0.205 0.159 0.080 56 0.159 0.411 0.014 0.218 0.107 0.090 57 0.225 0.422 0.004 0.214 0.059 0.076 58 0.260 0.204 0.039 0.234 0.046 0.217 59 0.128 0.004 0.389 0.366 0.032 0.081 60 0.129 0.007 0.004 0.436 0.150 0.275 61 0.117 0.063 0.021 0.330 0.086 0.381 62 0.233 0.016 0.012 0.466 0.080 0.194 63 0.217 0.373 0.001 0.161 0.077 0.170 64 0.208 0.119 0.019 0.297 0.118 0.239 65 0.119 0.089 0.028 0.289 0.121 0.354 66 0.142 0.198 0.041 0.337 0.125 0.157 67 0.129 0.210 0.368 0.156 0.071 0.067 68 0.114 0.007 0.727 0.079 0.031 0.042 69 0.140 0.095 0.186 0.395 0.101 0.082 70 0.148 0.170 0.205 0.220 0.074 0.184 71 0.137 0.129 0.084 0.353 0.233 0.065 72 0.161 0.526 0.008 0.139 0.119 0.047 73 0.124 0.352 0.02 3 0.293 0.072 0.136 74 0.229 0.188 0.028 0.371 0.112 0.073 75 0.136 0.344 0.018 0.316 0.083 0.103 76 0.122 0.363 0.029 0.354 0.066 0.067 77 0.194 0.274 0.003 0.419 0.068 0.042 78 0.309 0.027 0.004 0.497 0.156 0.007 79 0.223 0.149 0.053 0.399 0.137 0. 039 80 0.103 0.404 0.022 0.199 0.069 0.203 81 0.179 0.450 0.015 0.152 0.059 0.146 82 0.303 0.149 0.005 0.240 0.055 0.248 83 0.138 0.263 0.026 0.340 0.073 0.160 84 0.226 0.109 0.001 0.468 0.159 0.036
58 Appendix A: (Continued) Table 9: (Continued) 85 0.204 0.115 0.010 0.552 0.092 0.028 86 0.493 0.105 0.014 0.246 0.118 0.024 87 0.345 0.150 0.006 0.281 0.197 0.021 88 0.274 0.165 0.004 0.405 0.105 0.049 89 0.222 0.047 0.012 0.500 0.123 0.096 90 0.375 0.176 0.014 0.269 0.116 0.051 91 0.395 0.244 0.00 4 0.234 0.091 0.031 92 0.340 0.480 0.001 0.039 0.099 0.041 93 0.109 0.709 0.014 0.061 0.040 0.066 94 0.140 0.762 0.002 0.022 0.028 0.045 95 0.038 0.895 0.001 0.014 0.015 0.038 96 0.152 0.559 0.012 0.168 0.062 0.046 97 0.070 0.774 0.002 0.108 0.032 0. 014 98 0.081 0.642 0.022 0.171 0.058 0.027 99 0.137 0.387 0.081 0.275 0.100 0.019 100 0.135 0.101 0.010 0.199 0.125 0.430 101 0.187 0.113 0.021 0.170 0.143 0.365 102 0.066 0.121 0.000 0.266 0.189 0.359 103 0.322 0.100 0.003 0.212 0.084 0.279 104 0.1 86 0.204 0.000 0.257 0.084 0.268 105 0.152 0.067 0.013 0.442 0.128 0.198 106 0.338 0.130 0.000 0.258 0.079 0.195 107 0.248 0.216 0.002 0.213 0.149 0.172 108 0.242 0.212 0.001 0.156 0.072 0.316 109 0.202 0.158 0.002 0.268 0.170 0.199 110 0.130 0.007 0 .000 0.668 0.182 0.012 111 0.336 0.155 0.029 0.246 0.197 0.036 112 0.330 0.221 0.008 0.220 0.063 0.158 113 0.398 0.254 0.007 0.151 0.159 0.032 114 0.372 0.292 0.000 0.026 0.237 0.073 115 0.346 0.335 0.014 0.195 0.042 0.068 116 0.222 0.587 0.012 0.113 0.022 0.045 117 0.170 0.718 0.010 0.069 0.010 0.024 118 0.085 0.817 0.003 0.029 0.039 0.027 119 0.312 0.287 0.027 0.207 0.120 0.047 120 0.457 0.439 0.001 0.032 0.019 0.053 121 0.316 0.151 0.032 0.373 0.083 0.045 122 0.291 0.087 0.021 0.367 0.223 0.0 12 123 0.293 0.206 0.009 0.369 0.084 0.040 124 0.220 0.547 0.008 0.121 0.069 0.035 125 0.133 0.135 0.021 0.332 0.349 0.031 126 0.369 0.084 0.014 0.222 0.303 0.009 127 0.171 0.623 0.007 0.056 0.107 0.036
59 Appendix A: (Continued) Table 9: (Continued) 128 0.306 0.376 0.018 0.083 0.202 0.015 129 0.243 0.590 0.002 0.042 0.109 0.014 130 0.192 0.495 0.021 0.087 0.185 0.019 131 0.015 0.407 0.136 0.164 0.256 0.022 132 0.055 0.589 0.002 0.034 0.259 0.062 133 0.353 0.245 0.003 0.100 0.288 0.011 134 0.492 0.061 0.058 0.136 0.253 0.000 135 0.492 0.062 0.005 0.268 0.062 0.112 136 0.543 0.020 0.010 0.284 0.078 0.065 137 0.373 0.085 0.004 0.391 0.119 0.027 138 0.246 0.123 0.009 0.383 0.161 0.077 139 0.389 0.052 0.001 0.212 0.328 0.019 140 0.468 0.003 0.0 00 0.209 0.306 0.014 141 0.322 0.020 0.001 0.388 0.266 0.003 142 0.146 0.175 0.050 0.349 0.271 0.010 143 0.168 0.047 0.010 0.282 0.362 0.131 144 0.039 0.068 0.011 0.219 0.653 0.009 145 0.058 0.183 0.011 0.377 0.340 0.030