|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
Towards the Development of a Coastal Pred iction System for the Tampa Bay Estuary by Heather Holm Havens A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Marine Science College of Marine Science University of South Florida Major Professor: Mark E. Luther, Ph.D. Paula G. Coble, Ph.D. Cynthia A. Heil, Ph.D. Steven D. Meyers, Ph.D. John H. Paul, Ph.D. Date of Approval: November 12, 2009 Keywords: Numerical modeling, Particle tracking, Estuaries, Algal blooms, Ammonium, Karenia brevis Copyright 2009, Heather Havens
Dedication I wish to thank everyone in the Ocean Monitoring and Pred iction Lab for their support. To my lab mates Monica Wilson and Sherryl Gilbert for always having their doors open. To Vembu Subramanian for his technical suppor t and comic relief. To Steve Meyers for his assistance with modeling issues and editing. To my committee members, Paula Coble, Cindy Heil and John Paul for their schol arly advice. A special thanks to my advisor, Mark Luther, for his compassion and guidance throughout my time at USF. For all of my friends and family for their suppor t to get me to this point, especially my parents. To my husband for his continuous en couragement and for being a partner in my journey.
Acknowledgments This work was supported in part by the Gr eater Tampa Bay Marine Advisory Council Physical Oceanographic Real-Time System, Inc. as well as the National Science Foundation biocomplexity program and th e National Oceanic and Atmospheric Administration (NOAA) Coastal Services Cent er. Thanks to Cindy Heil and Ryan Pigg at the Florida Fish and Wildlife Research In stitute and Chris Anastasiou at the Florida Department of Environmental Protection for pr oviding sampling data. Thanks as well to Brad Benggio at the NOAA Office of Respons e and Restoration (HAZMAT) for his input. Thanks to Paula Coble for her insi ght into colored diss olved organic matter photochemistry. Thanks also to Chuanmin Hu for his insight into satellite imagery.
i Table of Contents List of Tables ................................................................................................................ ..... iii List of Figures ............................................................................................................... ..... iv Abstract ...................................................................................................................... ....... vii Dissertation Introduction ..................................................................................................... 1 Chapter 1: Particle Tracking Simula tion of an Anhydrous Ammonia Spill ........................5 Introduction ..............................................................................................................5 Methods....................................................................................................................7 The coastal prediction system ......................................................................7 Anhydrous ammonia spill ............................................................................9 Results ....................................................................................................................12 Discussion ..............................................................................................................16 Chapter 2: Lagrangian Particle Track ing of a Toxic Dinoflagellate Bloom ......................26 Introduction ............................................................................................................26 Methods..................................................................................................................31 Numerical circulation model ......................................................................31 Particle tracking .........................................................................................32 Sampling ....................................................................................................34 Results ....................................................................................................................35 Discussion ..............................................................................................................40 Chapter 3: Dispersion of Colo red Dissolved Organic Matter ............................................51 Introduction ............................................................................................................51 Methods..................................................................................................................55 Results ....................................................................................................................59 Discussion ..............................................................................................................63 Conclusion .................................................................................................................... .....76 Future Work ...........................................................................................................77 Literature Cited .............................................................................................................. ....79 Appendices .................................................................................................................... .....88 Appendix A: Transport Quotient Calculation ........................................................89
ii Appendix B: Application of th e Coastal Prediction System ..................................90 Feather Sound Project ................................................................................90 Methods......................................................................................................91 Results ........................................................................................................94 Discussion ..................................................................................................95 Appendix C: Old Tampa Bay Karenia brevis samples ........................................102 About the Author ................................................................................................... End Page
iii List of Tables Table 1 Surface photobleaching rates (d-1) from the CDOM literature for wet and dry conditions at various estu arine and oligotrophic locations. ................75 Table 2 Water level harmonics used in the numerical circ ulation model. ..................100 Table 3 Average delta 15-N values from discharge points in Feather Sound. ............101
iv List of Figures Figure 1 Bathymetric map of the Tampa Bay estuary. ...................................................19 Figure 2 A National Oceanic and Atmos pheric Administrati on (NOAA) plot showing winds in Hillsborough Bay from 12-20 November 2007. .................20 Figure 3 Frames from a numerical model simulation initialized on 18 November 2007 following an anhydrous ammonia spill in the Alafia River. ................................................................................................................21 Figure 4 Water samples collected by the Florida Department of Environmental Protection (FDEP) immediately foll owing an anhydrous ammonia spill in the Alafia River. ...........................................................................................22 Figure 5 Water samples collected by the Florida Fish and Wildlife Research Institute (FWRI) after an anhydrous ammonia spill. .......................................23 Figure 6 Vertical profiles of model out put net velocities for three locations within Tampa Bay: across central Hillsborough Bay (aligned with the mouth of the Alafia River and bi secting two dredge islands in Hillsborough Bay), across southern Hillsborough Bay and central Middle Tampa Bay (aligned with the mouth of the Little Manatee River). ..............................................................................................................24 Figure 7 Vertical profile of model output net velocities for th e Alafia River. ...............25 Figure 8 Bathymetric map of the Tampa Bay estuary with the darkest cuts representing the dredged shipping channels. ...................................................45 Figure 9 Vertical profiles of the northsouth (v) component of the horizontal current flow averaged across four locations within Tampa Bay: across the mouth of Tampa Bay, across Middle Tampa Bay (aligned with the Little Manatee River), across th e mouth of Hillsborough Bay and across the mouth of Old Tampa Bay. ...............................................................46 Figure 10 Vertical profiles of the east-w est (u) horizontal curre nt flow averaged across two locations within Tamp a Bay: across Middle Tampa Bay (aligned with the Little Manatee River) and across the mouth of Old Tampa Bay. ......................................................................................................47
v Figure 11 Numerical model si mulations initialized at the beginning of (a) June, (b) July and (c) August 2005. ..........................................................................48 Figure 12 Transport quotients are a ra tio between the number of particles, representing Karenia brevis cells, in each individu al grid cell and the total number of particles in the model domain. ...............................................49 Figure 13 Figures showing concentrations of Karenia brevis collected from water samples at various locati ons throughout Tampa Bay for the months of (a) May (b) June (c) July (d) August 2005. ....................................50 Figure 14 Bathymetric map of the Tampa Bay estuary. ...................................................67 Figure 15 River di scharge rates (ft3 s-1) from the United States Geological Survey (USGS) for the Hillsborough River, Al afia River, Little Manatee River and Manatee River. ..........................................................................................68 Figure 16 Transport quotients, with a pplied post-processing photobleaching rate, from the Hillsborough River simulations. .......................................................69 Figure 17 Transport quotients, with a pplied post-processing photobleaching rate, from the Alafia River simulations....................................................................70 Figure 18 Transport quotients, with a pplied post-processing photobleaching rate, from the Little Manatee River simulations. .....................................................71 Figure 19 Transport quotients, with a pplied post-processing photobleaching rate, from the Manatee River simulations. ...............................................................72 Figure 20 Composite showing the tran sport quotients, w ith applied postprocessing photobleaching rate, from the four rivers (Hillsborough, Alafia, Little Manatee and Manatee) simultaneously. .....................................73 Figure 21 Bay-wide averaged monthly surface salinity. ..................................................74 Figure 22 Location of disc harge points in Feather Sound for nitrogen source tracking using stable isotopes. .........................................................................97 Figure 23 Snapshots of numerical model forecast simulation for nitrogen source tracking study (May 14-22, 2007). ..................................................................98 Figure 24 Probability di stribution for nitrogen s ource tracking study based on forecast simulation run from May 14-22, 2007. ..............................................99
vi Figure 25 Karenia brevis samples collected in Old Tampa Bay during a 2006 K. brevis bloom...................................................................................................103
vii Towards the Development of a Coastal Prediction System for the Tampa Bay Estuary Heather Havens ABSTRACT The objective of this research is to evaluate a coastal prediction system under various real world scenarios to test the efficacy of the sy stem as a management tool in Tampa Bay. The prediction system, comprised of a thre e-dimensional numerical circulation model and a Lagrangian based partic le tracking model, simulates oceanographic scenarios in the bay for past (hindcast), present (nowcast) and future (forecast) time frames. Instantaneous velocity output from the numerical circul ation model drives the movement of particles, each representing a fraction of the total material, within the model grid cells. This work introduces a probability calculation th at allows for rapid analysis of bay-wide particle transport. At every internal time step a ratio between the number of particles in each individual model grid cell to the tota l number of particles in the entire model domain is calculated. These ratios, herein cal led transport quotients, are used to construct probability maps showing locations in Tamp a Bay most likely to be impacted by the contaminant.
viii The coastal prediction system is first evalua ted using dimensionless particles during an anhydrous ammonia spill. In subsequent stud ies biological and chemical characteristics are incorporated into the transport quotie nt calculations when constructing probability maps. A salinity tolerance is placed on particles representing Karenia brevis during hindcast simulations of a harmful algal bl oom in the bay. Photobleaching rates are incorporated into probability maps constructed from hindcast simulations of seasonal colored dissolved organic matter (CDOM) transport. The coastal prediction system is made more robust with the inclusion of biological parameters overlaid on top of the circulat ion dynamics. The system successfully describes the basic physical mechanisms underl ying the transport of contaminants in the bay under various real world scenarios. The calculation of trans port quotients during the simulations in order to develop probability maps is a novel concept when simulating particle transport but one wh ich can be used in real-time to support the management decisions of environmental agencies in the bay area.
1 Dissertation Introduction Tampa Bay is the largest open water estuary in Florida (Hu et al., 2004) and is home to the 10th largest port system in the United States (Lewis et al., 1999). Many anthropogenic stresses to the water quality in Tampa Bay are the result of growing urban and agricultural watersheds (Bricker et al., 2007) surrounding the bay. Anthropogenic stresses include dredging to accommodate co mmercial shipping vesse ls (Bricker et al., 2007; Vincent, 2001), non-point source polluti on resulting in excess nitrogen loading (Cross, 2007; Greening and Janick i, 2006; Morrison et al., 20 06) and hazardous material spills (Lewis et al., 1999; Owens and Michel, 1995). As a result, Tampa Bay has been the focus of major water quality restoration e fforts in recent years (Greening and Janicki, 2006). Natural stresses to the water quality in the bay are present in the form of annual harmful algal blooms of Karenia brevis initiated offshore in th e Gulf of Mexico and brought into Tampa Bay by complex circulation features (Steidinger et al., 1998; Walsh et al., 2001). Several monitoring programs regularly sa mple in Tampa Bay for water quality parameters such as excess nutrients (Flori da Department of Environmental Protection (FDEP)) and harmful algae (Florida Fish and Wildlife Conservation Commissions Fish and Wildlife Research Institute (FWRI)). These sampling programs constitute separate
2 studies carried out by different agencies and are primarily event response in nature. A need exists to better understand the transport and fate of these water quality parameters in Tampa Bay through the development of a real -time water quality monitoring program. For a coastal monitoring program to exist in re al-time a prediction system that is capable of ingesting real-time meteor ological (wind speed and direc tion) and circulation (water level, salinity and fresh wate r flux) data needs to be deve loped. The data acquisition, processing and quality control should be au tomated. Numerical modeling of this data should be periodically ground-truthed with sa mpling data to determ ine the accuracy of the model at predicting the distribution of bay-wide water quality parameters. Results should be easily accessible to the water management community via the internet. This dissertation details research towards th e development of a coastal prediction system and its capability as a predictive management tool for Tampa Bay. Predictions made by the system encompass the simulation of oceanogr aphic parameters in the past (hindcast), present (nowcast) and future (forecast) time frames (Vincent, 2001). The prediction system is comprised of a nume rical circulation model coupled with a Lagrangian particle tracking model. A Lagrangian tracking scheme prevents artificial diffusion and allows for realistic particle movement versus an Eulerian approach which is overly diffusive (Burwell, 2001). The circulation model simulates the physical dynamics within the estuary using real-time oceanographic forcing conditions. The particle tracking model simulates the transport of
3 material within the bay using dimensionless particles or with the incorporation of nonconservative behaviors. Proba bility maps are generated from the particle tracking simulations to show probability distributions of material in Tampa Bay based on location. Four studies are carried out to examine the efficacy of the coastal prediction system as a management tool. First, the coastal prediction sy stem is utilized in real-t ime during a hazardous material spill with the purpose of alerting authorities to potential high impact areas in Tampa Bay. Forecast particle distributions are used by FWRI scientists to guide sampling for increased algal concentrations th at could result from the flux of nutrients into the bay. The effectiveness of the prediction system as an event response tool is examined and the data are ground-truthed with samples collected by the FDEP and the FWRI. Second, the spatial distribution of a harmful al gal bloom is simulated and compared with samples collected by the FWRI in 20 05 during the peak of a previous Karenia brevis bloom in Tampa Bay to determine the capacity of the model to capture general features of the observations. Particles are given a post-processing salinity tolerance based on the measured salinity range of K. brevis in the field. The capability of the model to reproduce the dispersion of an ev ent in hindcast mode is crit ical in the determining the accuracy of the prediction system. Third, the distribution of colored dissolved organic matter (CDOM) from the four largest freshwater riverine sources in Tampa Bay is examined during both wet and dry season
4 conditions. Daily decay rates are imposed on the particles to simulate seasonal CDOM photobleaching rates measured in the field. The ability to incor porate non-conservative behaviors into the prediction system distri bution maps is a powerful tool to more precisely describe freshwater content in Tampa Bay. Finally, the prediction system is evaluated du ring a FDEP study to forecast the advection of nitrogen from an urbanized region of Tampa Bay. This study examines the extent to which circulation in this region affects wate r quality and subsequen tly seagrass growth. The combination of a forecast simulation with field work is an example of adaptive sampling and is used as a method for ve rifying the prediction system results. Together these studies constitute initial para meterizations of a coastal prediction system developed for Tampa Bay. The prediction system assisted managers in real-time during a hazardous material spill and succe ssfully predicts the location of a resulting algal bloom. Hindcast simulations, generated from the pr ediction system, of a harmful algal bloom correlate well with samples collected at the time of the event. The distributions of freshwater and nutrient fluxes are mapped and ex amined in relation to seagrass coverage. Each of these studies are c onducted with the purpose of be tter understanding the transport and fate of water quality parameters in the Tampa Bay estuary while also supporting management decisions for environmental issues affecting the bay. To this end, an online component of the coastal predic tion system, that incorporates results from the studies that follow, is in development to better manage response and mitigation efforts in Tampa Bay.
5 Chapter 1: Particle Tracking Simula tion of an Anhydrous Ammonia Spill Introduction The hydrodynamics of the Tampa Bay estuary are influenced primarily by astronomical tides, winds and river runoff (Vincent, 2001). Classical estuarine ci rculation (Pritchard, 1967) for a partially to well-mixed estuary su ch as Tampa Bay, has a two-layered flow with fresh water flowing towards the mouth of the estuary at the surface and saline water flowing landward at depth. Bay-wide estuarine circulation varies depending on environmental conditions and is the driving fo rce behind the advection of material within the bay (Weisberg and Zheng, 2006). Reliable and accurate observati ons and predictions of bay-wide circulation assist with commercial shipping, hazardous material res ponse and environmental management. The potential for accidental or intentional contam ination within Tampa Bay is significant due to the types of hazardous materials (e.g. petr oleum products, sulfuric acid and anhydrous ammonia) regularly transported through the Port of Tampa by commercial vessels. In the event of a hazardous material spill within Tampa Bay operational prediction models can be used to accurately predict, or forecast, the transport of the pollutant (Vincent, 2001).
6 Circulation models have been developed to simulate the dispersion of pollutants (Dimou and Adams, 1993; Gomez-Gesteira et al., 1999; Scott, 1997) wh ere transport and fate are best described by the movement of individual pa rticles rather than their concentration. Particle transport models generally c onsist of two components: a hydrodynamic component to define the kinetics of the flow and a particle tracking component to define the transport of the pollutant (Benkhaldoun et al., 2007). A coastal prediction system has been developed for Tampa Bay that interfaces a realistic numerical circulation model with meteorological forecasts to predict the hydrodynamics with in the bay. Overlaid on top of the circulation model is a Lagrangian particle tracking model which simulates the transport of contaminants by assigning a pa rticle to a water mass and following the particle as it is advected by the instanta neous model velocity field (Burwell, 2001; Meyers and Luther, 2008). Similar prediction systems have been developed to simulate the transport of contaminants in locations where two-dimensional barotropic circulation is sufficient (Gomez-Gesteira et al., 1999; Periez, 2004); fully three-dimensional simulations are necessary to resolve the grav itational circulation cont rolling the advection of materials in Tampa Bay (Weisberg and Zheng, 2006). The Tampa Bay coastal prediction system is evaluated during a hazardous material spill. The particle tracking model predicts the di spersion of material based on forecast circulation dynamics and records the distributi on of particles at each internal time step. From these particle distributions probability maps are constructed showing areas in the bay with the highest potential for contam ination. Periez (2004) simulated the dispersion of contaminants in the Strait of Gibraltar using a simila r model configuration
7 (hydrodynamics plus particle tracking). No evaluation of their model was performed during a spill event. The objective of this paper is to describe the Tampa Bay coastal predication system and to determine whether the forecasting compone nt of the system can assist responders during an anhydrous ammonia spill in Tampa Bay. The numerical model and observational parameters are discussed in th e following section. Comparisons between the particle transport simulation and in situ ammonium measurements are discussed. The utility of the particle transport simulation in alerting responders to areas of Tampa Bay with the highest probability of being affected by the spill is evaluated based on sampling conducted using model output. Conclusions are drawn as to the physical transport mechanisms affecting the spill and the effec tiveness of the model forecast at predicting this transport. Methods The coastal prediction system A numerical circulation model, based on the Princeton Ocean Model (Blumberg and Mellor, 1987), is forced with real-time oceanographic observations of the physical forcing functions for Tampa Bay to produce three-dimensional fields of circulation, temperature, salinity and water level in past (hindcast) and future (forecast) time frames. The hindcast model uses quality controlled data for boundary conditions and the forecast
8 model is initialized from the most recent hindcast output fields a nd holds constant the final model boundary conditions (except for wind speed and direction which are held at predicted values obtained from the National Weather Service). The circulation model used in this study is that developed by Meyers et al. (2007) which divides the bay into a 70 by 100 grid of cells in the horizontal and 11 sigma levels in the vertical. The model is forced at the bay m outh with water level obtained from the Tampa Bay Physical Oceanographic Real-Time System (TB-PORTS) and sali nity obtained from the Environmental Protection Commission (EPC ). Winds, evaporat ion and precipitation are forced uniformly over the entire model surface. The mode l simulation is also forced at discrete points using daily observations of fresh water flux (rivers and canals) obtained from the US Geological Survey (USGS) National Water Information System and monthly averaged wastewater discharge obtai ned from treatment plants. A detailed description of the model hydrodynamics and ev aluation can be found in Meyers et al. (2007). The instantaneous model velocity fiel ds generated by the circulation model are used to drive the particle tracking model. An algorithm developed by Burwell (2001) is used to advect dimensionless particles according to the simulated three-dimensional ci rculation model velocity field. Particles, each representing a fraction of the total haza rdous material, are ge nerated by the tracking model evenly throughout the water column within the grid cell closest to the site of the spill. A random walk technique (Dim ou and Adams, 1993; Korotenko et al., 2004; Periez, 2004; Proctor et al ., 1994) is used to simulate the dispersion of material by
9 linearly interpolating the pos ition of each particle betw een time steps based on the velocities of neighboring grid cells and summed with a rando m vector function computed independently at every time step for each particle (Meyers and Luther, 2008). Lagrangian particle tracking c ode records the time and loca tion of particles within the model grid cells at each time step; for this study the locations of pa rticles are written to files every 30 min durin g a 7 day-long simulation. At each time step the ratio between the number of particles in any i ndividual model grid cell to the total number of particles in the m odel domain is calculated. This value is called the transport quotient, and is a meas ure of the likely distribution of hazardous material within Tampa Bay (see Appendix A). Cells with the highe st transport quotients contain particles for the greatest amount of time during the simulation. Conversely, cells with low transport quotients during the simulatio n rarely contain par ticles. From these transport quotients a probability map can be constructed predicti ng the areas in Tampa Bay most likely to be affected by the transport of hazardous material. Anhydrous ammonia spill Bulk quantities of liquid anhydrous ammoni a, a compound used in fertilizer production, are transported under high pressure into th e Port of Tampa by ship and from there by pipeline to fertilizer produc tion facilities in surrounding c ounties. The transport of material in this manner poses a potential heal th and pollution hazard should there be an accidental or intentional release of amm onia into surface waters. Anhydrous ammonia
10 rapidly dissolves upon contact with wate r into ammonium h ydroxide; the amount dissolved into solution is dependent on pH temperature and salinity of the water. Ammonium hydroxide remains at the water surf ace much like an oil slick (Raj et al., 1974) The night of 12 November 2007 a rupture in an anhydrous ammonia pipeline occurred releasing an estimated 5-30 tons of ammoni um hydroxide into th e Alafia River over a period of two days. The Alafia River feeds directly into Hills borough Bay (HB) in the northeast portion of Tampa Bay and subs equently into Middle Tampa Bay (MTB) (Figure 1). Two days after the spill strong northwesterly winds (order 20 knots) were recorded over a 24 h period within HB (Fi gure 2) influencing the transport of the hazardous material within Tampa Bay. The model simulation was initialized on calendar day 18 November 2007 following a request for assistance by event re sponders. The modeling effort is split into hindcast and forecast phases. The Lagrangian tracking m odel first uses hindcast model output over model days 13-17 November 2007 to estimat e spill position on the day the of model initialization (Figure 3). The source function (i.e. particle re lease point) for the model is the easternmost grid cell in the Alafia Ri ver (Figure 4). The tracking model then continues in forecast mode beginning on 18 November and continues for a 48-h period (Figure 3.) to predict the tran sport of ammonia in Tampa Bay.
11 The concentration of ammonium particles in the model is cons erved throughout the hindcast/forecast simulation. The non-conserva tive properties of ammonium in seawater are ignored and ammonium is considered to be in solution with water however, it should be noted that ammonium concentrations are affected by chemical and biological activity and physical processes (Conomos et al., 1979) For example, ammonium hydroxide is readily consumed by phytoplankton and also in teracts with sediment through adsorption and desorption (Chao et al., 2007). Some toxigenic species of Pseudo-nitzschia spp. have been shown to increase in number to form a bloom (>100,000 cells L-1) in coastal waters rich in ammonium (Bates et al ., 1998). Diatoms of the genus Pseudo-nitzschia spp. have been observed in Tampa Bay since the 1960s an d are persistent in the bay at background concentrations (<1000 cells mL-1) (Badylak et al., 2007). Therefore the anthropogenic introduction of ammonium into the bay is expected to impact phytoplankton growth. In order to determine the impact the spill had on the ecosystem of the bay, water samples were collected on two occasions in order to determine (1) the st arting concentration levels of ammonium in the Alafia River a nd (2) if there was a resulting change to phytoplankton biomass in Tampa Bay. In the days immediately following the spill, wa ter samples were colle cted at the surface along the Alafia River by scientists at the Florida Department of Environmental Protection (FDEP) to measure ammonium conc entrations in the river. Samples were collected from stations starting the day af ter the anhydrous ammoni a spill and continuing for 2 days (Figure 4). Station 1 was located about 1 km west of the spill site. The
12 remaining 9 stations continued westward down the Alafia River toward Tampa Bay. Station 9 was located in HB at the mouth of the river and station 10 was between two dredge spoil islands in HB. Scientists at the Florida Fish and Wildlif e Research Institute (FWRI) investigated phytoplankton concentration and community co mposition as a result of the increased ammonium levels in the bay. One water samp le was collected from 12 different stations along the eastern coast of Tampa Bay (Figur e 5) on 19 November 2007 at a depth of 0.5 meters and each of the 12 samples were analyzed for algal composition. The FWRI defines a low algal count as >1x104 cells L-1, a medium count as >1x105 cells L-1 and a high count as >1x106 cells L-1 (for diatom species). The 12 sample sites were chosen by the FWRI due to the high probability of incr eased levels of ammonium in those areas based on the transport quotients genera ted by the coastal prediction system. Results Model results for 13-20 November 2007 show a plume of ammonium (represented by Lagrangian particles) moving down the Alafia River and into Tampa Bay (Figure 3). On the first day tidal currents carry the particle s westward to the mout h of the Alafia River and then back towards the east. This oscill ating tidal advection re peats for another day before the particles enter HB. The ammonium remains localized near the Alafia River for another day until it is carried southwar d. The ammonium is heavily concentrated offshore of Apollo Beach 2 days later. At the end of the simulation the particles are
13 distributed from Apollo Beach to the mouth of the Little Manatee River. The ammonium particles remained within 2 km of the eastern coast of Tampa Bay for the duration of the simulation. Three representative cross sections of the ba y are chosen to examine current flow: across central HB (aligned with the mouth of the Alafia River and bisecting the two dredge islands), across southern HB and across cen tral MTB (aligned with the mouth of the Little Manatee River) (Figure 6) At each of these locations the average vertical structure of the currents over the 7-day simulation is calculated following the methods of Meyers et al. (2007). An outward (n egative) flow is present at all depths along the eastern boundaries of each of the three cross sections with speed generally increasing toward the surface, ranging from about 2 cm s-1 near the bottom to 10 cm s-1 near the surface. In central HB the current is flowing inward (positive) within a nd above the shipping channel, except for a small area at the surface, with a maximum speed greater than 8 cm s-1. Flow is also positive to the west of the channel at all dept hs. In southern HB the flow is uniformly negative within the upper tw o meters of the water column. Maximum outflow is about 15 cm s-1 on the surface of the eastern co ast and decreases with depth. Two areas of subsurface inflow are present: one is slightly offset to the east of the channel the other is located along the wester n edge of the channel. Their maximum speeds are only about 2 cm s-1. In central MTB the vertical structure of the estuarine circulation breaks down and a horizontal gradie nt is found with outward flow within and to the east of the shipping channel. Again, the maximum speed is about 10 cm s-1. Weak inflow (<2 cm s-1) within the channel is confined to a few small areas along the eastern
14 slope and in the middle of the water column a bove the channel. A larger, but still weak, area of inflowing current is located to the west of the channel and extends from the bottom to about one meter below the surface. The mean horizontal velocity in the Alafia River during the simulati on is calculated as a function of depth and horizontal position (Figure 7). At the mouth of the river, flow is outward at all depths. The maximum outfl ow is at the surface and about 10 cm s-1. A layer of inward flowing water with a maximum speed around 4 cm s-1 is found to the east of the mouth and follows the bathymetry of the river upstream cr eating a two-layered circulation pattern within the river. The net transport is upstream along the river bottom and downstream in the surf ace layers of the river. A probability map is generated from m odel day 19 November 2007 of the forecast simulation (Figure 5). The model grid cells with the highest transport quotient values (concentration of particles) on that particular model day are the easternmost cells in the Alafia River and the cells offshore of Apo llo Beach. The transport quotients decrease significantly to the south of the Little Manatee River and within MTB. The FDEP water samples collected the day after the anhydrous ammonia spill, between station 2 and station 9 on the Alafia Rive r, contained elevated concentrations of ammonium (>2 mg L-1) (Figure 4). Samples were coll ected from station 2 on the first day only. On the second day samples from the upstream stations and the westernmost station (half of the samples collected) contained <1 mg L-1 of ammonium. Station 9,
15 located directly at the mouth of the ri ver, was the only station where ammonium concentrations were higher on the second da y of sampling than on the first day. The water samples collected 3 days after the spill contained concen trations of ammonium that averaged <0.5 mg L-1. Samples from station 10, located within HB between two dredge spoil islands, never contained greater th an trace amounts of ammonium (~0.02 mg L-1) during the 3-day period. Similar con centrations of ammonium (0.01 mg L-1) were recorded during another Pseudo-nitzschia spp. bloom (Bates et al., 1998). Concentrations of ammonium m easured at each of the other stations were at least an order of magnitude higher than those recorded from station 10. All algal samples were collect ed by the FWRI along the easte rn coast of the bay on 19 November 2007, based on the forecast model estimates of ammonium distribution for that day (Figure 5). The Pseudo-nitzschia spp. cell counts collected by the FWRI for the first 3 stations were in the medium range. Sa mples from station 4, located to the north of Apollo Beach, contained cell counts of Pseudo-nitzschia spp. that constituted a large bloom. Cell counts from stations 1-3 to the north of Apollo Beach, station 5 to the south of Apollo Beach and station 6 at the mouth of the Little Manat ee River were in the medium range for a Pseudo-nitzschia spp. bloom. Samples from the remaining stations constituted counts at low to background levels.
16 Discussion The particle trajectory hindcas t/forecast simulation demonstrates the physical dynamics in Tampa Bay following the anhydrous ammonia spill. Particles in the Alafia River are initially subject to tidal action and are retain ed within the river due to mixing within the upstream flowing bottom layers of the river (F igure 7). Particles that reach the river mouth, where flow is outward at all depths are rapidly transported southward along the eastern coast of Tampa Bay due in part to the presence of a strong current, seen in each of the 3 velocity profiles (Figure 6) as well as strong northwesterly winds that act to pile water along the eastern boundary of the bay (Figure 2). In order to determine what impact the 2-day wind peak on 15 November 2007 has on the transport of ammonium, a separate simulati on is run holding winds constant at 5 m s-1 from the northeast. These conditions are those roughly found in the bay on 13 November. Results from this simulation (not shown) suggest that the wind peak plays a significant role in the transport of material within Tampa Bay. The particles emerge from the Alafia River on the second day of th e simulation and disperse more widely throughout central HB and MTB. Few particle s are found adjacent to the eastern coast. This demonstrates the need to use a real istic wind field when conducting the model prediction, even for a time period as short as 24 h. Analyses of the collected water samples show that the simulation accurately models the transport of ammonium from the Alafia Ri ver. During the FDEP sampling period (13-15
17 November) model particles are not transported near the two spo il islands in HB. This is consistent with the trace levels of a mmonium found at station 10 (Figure 4). The model particle distribution was used by the FWRI to determine the algal sampling region. Scientists from FWRI m easured a medium to large sized Pseudo-nitzschia spp. bloom on 19 November 2007 in the area wher e the model simulation indicates a high concentration of ammonium. Th e highest concentrations of Pseudo-nitzschia spp. were collected at station 4 just north of Apollo Beach where the transpor t quotient values are also high (Figure 5). Particle concentration in th e model decreases to the south and west of Apollo Beach, consistent with decreasing Pseudo-nitzschia spp. concentrations collected from the surrounding stations to the south. No sampling was done away from the eastern coast of MTB, so the narrowness of the model particle distribution cannot be verified. The Pseudo-nitzschia spp. concentrations continue to decrease, as with the simulated ammonium concentrations, in sout hern MTB. The bloom is hypothesized to have formed due to increased ammonium that was transported out of the Alafia River and into the bay. The nature of an event respons e study is such that control measurements are rarely taken in advance. Consequently, th ere were no samples collected prior to the anhydrous ammonia spill to rule out the previous presence of a Pseudo-nitzschia spp. bloom. Together the circulation mode l and the Lagrangian particle tracking model form the coastal prediction system for Tampa Bay which is capable of simulating the physical transport of pollutants in the bay. The simulations do not take into account the effects of
18 weathering or biological processes on th e pollutants. As a result the particle concentration estimates are likely overesti mated by an unknown factor. Future versions of the prediction system shoul d incorporate realis tic biological and chemical processes. The coastal prediction system is used to support management decisions for several environmental issues affecting the bay (see Appendix B), specifically to simulate the trajectory of hazardous material spills for the FDEP and the FWRI. The models are capable of rapidly producing forecast simulations that, in the event of a spill, can alert authorities to areas in Tampa Bay with a high probability of being affected by the hazardous material. The prediction system at present is only accessible to scientists in the Ocean Monitoring and Predic tion Lab (OMPL) at the University of South Florida. The forecast simulations are compiled into an animation that is provi ded to end users at their request. In the future, decision ma kers will be allowed access to an online component of the coastal pred iction system. Event responde rs and other end users will be able to describe a spill scenario by en tering criteria into an online form. The prediction system will ingest the criteria then, using real-time data compiled from TBPORTS, display a 48 h simulation predicti ng how winds and currents will move the material around Tampa Bay. The ability to quickly set up custom scenarios will help manage response and mitigation efforts in real-time during an actual spill.
19 Figure 1 Bathymetric map of the Tampa Bay estu ary. Tampa Bay can be divided into four quadrants: Old Tampa Bay (OTB), Hillsborough Bay (HB), Middle Tampa Bay (MTB) and Lower Tampa Bay (LTB). A sh ipping channel runs from LTB through the central axis of the bay into MTB before splitting into two channels, one going to OTB the other to HB. The Alafia River drains the Hillsborough County watershed and empties into the eastern side of HB. An anhydrous ammonia spill occurred within the Alafia River and was transported into HB.
20 Figure 2 A National Oceanic and Atmospheric Administration (NOAA) plot showing winds in Hillsborough Bay from 12-20 Nove mber 2007. The plot shows wind speed (m/s) and direction (true) of winds duri ng the week-long simulation. Wind speed (red) and wind gusts (blue) are overlaid on top of each other. The direction of winds is represented by hatch marks.
21 Figure 3 Frames from a numerical model simu lation initialized on 18 November 2007 following an anhydrous ammonia spill in the Alafia River. The frames show the transport of model particles, representing am monium, from the Alafia River in eastern Hillsborough Bay into Middle Tampa Bay. The simulation is run in hindcast mode from 13-17 November and in forecast mode from 18-20 November. The time stamp is in UTC. Wind speed (m/s) and direction (indi cated by arrows) are shown for each frame. The scale represents the depth of the particles in the water column with values to the left of the scale being at the surface progressing to values to the ri ght of the scale being at the bottom.
22 Figure 4 Water samples collected by the Flor ida Department of Environmental Protection (FDEP) immediately following an anhydrous ammonia spill in the Alafia River. The samples were collected by th e FDEP from the 10 stations shown on the satellite image over a period of three days : 13-15 November 2007. The grid overlaying the satellite image shows the model grid cells that encompass the FDEP sample sites.
23 Figure 5 Water samples collected by the Florida Fish and Wildlife Research Institute (FWRI) after an anhydrous ammonia spill. The samples were collected from 12 stations, shown on the figure, along the eastern coas tline of Tampa Bay on 19 November 2007. The bar graph shows th e concentrations of Pseudo-nitzschia spp. counts (cells x 106 L-1) at each station. Labels on the graph indica te the FWRI classifications for medium (>2x105 cells L-1) and high (>1x106 cells L-1) cell counts (for diatom species). Transport quotients (Q), shown underlying the FWRI sa mple locations, are calculated for each model grid cell on 19 November 2007 and range on a scale from zero (low probability of finding a particle in a grid cell) to one (hi gh probability of finding a particle in a grid cell). The highest concentrations of Pseudo-nitzschia spp. were collected from station 4 where the Q values are also high. Particle concentrations decrease south of station 5, consistent with decreasing Pseudo-nitzschia spp. concentrations.
24 Figure 6 Vertical profiles of model output net velocities for th ree locations within Tampa Bay: across central Hillsbor ough Bay (aligned with the mouth of the Alafia River and bisecting two dredge islands in Hillsbor ough Bay), across southern Hillsborough Bay and central Middle Tampa Bay (aligned with the mouth of the Little Manatee River). The shaded region shows the bathymetry at the gi ven locations, with the deep incision in each being the dredged shipping channel. Curre nts flowing inward (into the bay) are represented by positive velocities while outward currents (out of the bay) are represented by negative velocities. Ve locities are in cm s-1.
25 Figure 7 Vertical profile of model out put net velocities for the Al afia River. The shaded region shows the bathymetry in the river. Currents flowing upstream are represented by positive velocities while outward downstream currents are represented by negative velocities. Velocities are in cm s-1.
26 Chapter 2: Lagrangian Particle Trac king of a Toxic Dinoflagellate Bloom Introduction Toxic blooms, resulting from large numbers of the unarmored dinoflagellate Karenia brevis are an almost annual occurrence along th e West Florida Shel f (WFS) (Steidinger et al., 1998; Walsh et al., 2001) with observations of colored water and fish kills dating back to 1844 (Tester and Steidinger, 1997). Blooms of brevetoxin producing K. brevis are responsible for neurotoxic shellfish poisoni ng, fish kills and re spiratory irritation (Magaa et al., 2003). The economy of Florida is impacted from these recurring blooms in the form of shellfish bed cl osures, medical costs, loss of tourism and disposal of dead fish (Kirkpatrick et al., 2004). A forecast syst em to predict the transport of these blooms could help alleviate some of these economic impacts. In order to understand and pr edict the transport of harmful algae in Tampa Bay, some knowledge of the dynamics of bloom initia tion and development is necessary. Background concentrations of Karenia populations are ubiquitous in Gulf of Mexico waters (Geesey and Tester, 1993) and certain phys ical factors are necessary to initiate the development, maintenance and transport of K. brevis blooms (Steidinger and Haddad, 1981) once concentrations exceed background le vels. Optimum growth conditions for K.
27 brevis in the field occur in water with temperat ures between 20-28 C and salinities of 3137 (Steidinger and Ingle, 1972); this species does not typically bloom in salinities <24 (Maier Brown et al., 2006). During daylight hours K. brevis cells concentrate in the upper water column, due to a positive phototac tic response (Heil, 1986), where transport and dispersion are subject to local winds a nd currents (Tester a nd Steidinger, 1997). K. brevis blooms originate 18-64 km offshore of the Florida coast (Steid inger, 1975) and are concentrated and transported inshore by comp lex interactions between coastal currents, wind-driven circulation and sh elf features (Steidinger et al., 1998). Sampling from a 1971 bloom in Tampa Bay indicated that K. brevis cells enter the bay from the Gulf of Mexico via a dredged shipping chan nel (Steidinger and Ingle, 1972). Tampa Bay is a drowned river bed estuary, abou t 50 km in length and covering more than 103 km2 (Zervas, 1993). A dredged shipping ch annel runs along the axis of the bay extending from Lower Tampa Bay (LTB) into Middle Tampa Bay (MTB) before splitting, one fork going west into Old Tampa Bay (OTB) and one fork going east into Hillsborough Bay (HB) (Figure 8). The buoyanc y driven circulation within Tampa Bay is typical of an estuarine syst em, with mean flow of saline Gu lf of Mexico water into the bay along the bottom and mean flow of fresh water out of the bay at the surface (Meyers et al., 2007; Weisberg and Zheng, 2006). Wate r exchange with the Gulf of Mexico occurs at the mouth of the bay, with mean inflow through Egmont Channel and mean outflow through both Egmont and Southwes t Channels (Meyers et al., 2007). K. brevis cells are thought to accumulate at fron ts along the WFS duri ng upwelling favorable conditions (Stumpf et al., 2008) and enter Ta mpa Bay on the inflowing current through
28 Egmont Channel (Steidinger and Ingle, 1972). Monitoring programs have recently become operational in the Gulf of Mexico and Tampa Bay to predict the onset and transport dynamics of K. brevis blooms. A Harmful Algal Bloom (HAB) monitoring program in the Gulf of Mexico and Tampa Bay is overseen by scientists at the Florida Fi sh and Wildlife Research Institute (FWRI). As part of the program, water samples are coll ected by volunteers and se nt to scientists at FWRI for microscopic analysis to determine K. brevis cell counts (Heil and Steidinger, 2009). Cell counts are grouped into categor ies based on potential ecological effect: counts <103 cells L-1 are considered background level, shellfish beds are closed when counts exceed 5 x 103 cells L-1, fish kills can occur at concentrations of 5 x 104 cells L-1 or greater and water discol oration becomes apparent with concentrations >106 cells L-1 (Walsh et al., 2002). Sampling becomes mo re vigorous (i.e. increased frequency and number of samples collected) following reports of a bloom or bloom impacts. The process of determining ce ll concentrations from samples is laborious and costly; a more interactive monitoring system, that combines field data with dynamic modeling, would provide a more useful management tool. One such system, the NOAA Gulf of Mexico HAB Operational Forecast System, repor ts, in the form of weekly bulletins, the predicted spatial extent, move ment and intensification of HABs in the Gulf of Mexico (Fisher et al., 2006). The heuristic forecas t model utilizes a comb ination of satellite imagery and wind predictions to simu late the extent and impact of K. brevis blooms in the Gulf of Mexico (Stumpf et al., 2009). Mode l forecasts from this system are verified
29 with FWRI cell counts, however the only boundary condition used in their model to determine the transport of these cells is predicted wind sp eed and direction. Prediction systems that are capable of fully three-dimensional numerical simulations (i.e. (Decker et al., 2007)) using observations of multiple boundary conditions (e.g. daily winds, freshwater inflow, water level and sa linity) are needed to guide HAB monitoring which, in Tampa Bay, at present relies predominantly on the collection of water samples in response K. brevis impacts (e.g. fish kills, discolored seawater or reports of respiratory irritation) with little advanced warning (Heil and Steidinger, 2009; Schofield et al., 1999). A prediction system capable of accurately forecasting bloom transport, based on underlying circulation dynamics, would assi st health officials and environmental managers with the mitigation of h ealth concerns and economic loses. Numerical circulation models simulate the physical dynamics that transport toxic HABs into and within coastal waters. Franks and Signell (1997) use a circulation model developed from Blumberg and Mellor (1987) co upled with a biological model to simulate the initiation of toxic Alexandrium tamarense blooms in the Gulf of Maine. Lagrangian particle tracking models are al so accurate tools for determini ng the initiation (Chen et al., 2007a) and transport (Cerejo and Dias, 2007; Ya nagi et al., 1995) of harmful algae. Lanerolle et al. (2006) use a two-dimensional numerical mode l and Lagrangian particle tracking to simulate the transport of K. brevis in response to along-sh ore wind stresses in the Gulf of Mexico. Three-dimensional simulations are necessary to resolve the
30 gravitational circulation cont rolling the advection of materials in Tampa Bay (Weisberg and Zheng, 2006). Here, a coastal prediction system, comprised of a three-dimensional numerical circulation model coupled to a Lagrangian particle track ing model, is evaluated to determine the capacity of the model to capture the general features of a 2005 K. brevis bloom in Tampa Bay. The circulation model is driven with hindcast forcing parameters (surface wind stress and freshwater flux) to reproduce th e underlying circulation dynamics present in the bay during the 2005 K. brevis bloom. Vertical profiles of the instantaneous velocity fields generated by the circulation model show transport mechanisms for various sections of the bay. These velocity fields drive a La grangian particle track ing model (Meyers and Luther, 2008) which simulates the transport of K. brevis cells within Tampa Bay by following particles, each representing a fracti on of the biological material, as they are advected throughout the model domain. Probabil ity maps, constructed from the particle transport simulations, show locations in Tamp a Bay that are most likely to be impacted by the bloom. The resulting probability maps are compared with cell concentrations in K. brevis samples collected during th e peak of the 2005 bloom. A salinity tolerance is placed on the particles when constructing the probability maps to simulate somewhat realistic biological behavi or as particles enc ounter different water masses. This study does not incorporate an all-inclusive biological model and, as such, several biological parameters, including vertical movement by K. brevis with light, are not considered.
31 The objective of this paper is to evaluate the hindcasting capability of a coastal prediction system to simulate the basic spatial patterns of an observed K. brevis bloom in the Tampa Bay estuary. The parameterizations of the numerical circulation model and the particle tracking model are discussed in the following section. Comparisons between the particle transport simulations and observations of in situ K. brevis concentrations are discussed. An evaluation of bay-wide probability maps which illustrate the likelihood of K. brevis occurrence in Tampa Bay is performed to determ ine the utility of the simulations as event response tools. Finally, conclu sions are drawn as to the physical mechanisms involved in the transport of the 2005 K. brevis bloom within Tampa Bay. Methods Numerical circulation model The circulation model is a primitive equation numerical model adapted from the Princeton Ocean Model (Blumberg and Mellor, 1987) for Tampa Bay (Galperin et al., 1991; Vincent, 2001). The circulation model domain is divided into a grid of 70 by 100 cells in the horizontal and 11 sigma levels in the vertical. The model is initiated with hindcast boundary conditions that have been qua lity controlled. The model is forced at the bay mouth with water level obtained fr om the Tampa Bay Physical Oceanographic Real-Time System (TB-PORTS) and the University of South Florida Coastal Ocean Monitoring and Prediction System (COM PS) and salinity obtained from the
32 Environmental Protection Commi ssion (EPC). The model simulation is also forced at discrete points using daily obs ervations of fresh water flux (rivers and canals) obtained from the US Geological Survey (USGS) National Water Information System and monthly averaged wastewater discharge obt ained from treatment plants. The model simulates the hydrodynamics and velocity fields in Tampa Bay; a deta iled description of the model hydrodynamics and evaluation ca n be found in Meyers et al. (2007). The vertical structure of the instantaneous hor izontal velocity fields is averaged over a three month period (June-August 2005) to examin e mean current flow across sections of the bay. Four representative cross sections are chosen: ac ross the mouth of Tampa Bay, across Middle Tampa Bay (aligned with the Li ttle Manatee River), across the mouth of Hillsborough Bay and across the mouth of Old Tampa Bay. For the north-south (v) component of the horizontal velocity, positive values represent northward current flow (into the bay); negative values represent current flow directed southward (out of the bay). For the east-west (u) component of the horizon tal velocity, positive values represent an eastward component to the current flow; negative valu es represent a westward component to the current flow. The velocity fields generated by the circulation model are used to drive the particle tr acking model of Burwell (2001). Particle tracking A Lagrangian based particle tracking method, versus an Eulerian method, has the advantage of realistic sub-grid scale motion and best approxim ates movement of particles
33 in the bay. Dispersion in the particle trac king model is accomplished with a random walk technique (Burwell, 2001) using a 4th order Runge-Kutta scheme with model velocity linearly interpolated to position particles at each model time step. The particle tracking code records the time and location of partic les within the model gr id cells at each time step; for this study the locations of particles are written to files every 60 minutes during each month-long simulation. For details of th is scheme see Havens et al. (2009) and Meyers and Luther (2008). Hindcast circulation model output is used to simulate particle transport during the peak months of the 2005 K. brevis bloom in Tampa Bay: June-August. The tracking model is initialized at the beginning of each month during a flood tide. Particles are distributed evenly throughout the water co lumn into the model grid cells across Egmont Channel based on prior inference that K. brevis blooms enter Tampa Bay through this channel. Particles mix unless they are flushed out of th e mouth at which point they are not allowed to re-enter the model domain. As particles are transported be tween various model grid cells they are assigned a salinity based on the salinity of the grid cell they occupy at each time step. A post-processing salinity restriction is placed on the particles when they are found in grid cells with salinities that are below the K. brevis tolerance of 24 (Mai er Brown et al., 2006; Steidinger and Ingle, 1972).
34 At every internal time step in the simulations a ratio between the number of particles in each individual grid cell and the total numb er of particles in the model domain is calculated. This ratio, called the transpor t quotient, is calculated according to the methods of Havens et al. (2009) but with th e additional incorporation of the salinity restriction (see Appendix A). Grid cells with the highest transport quotients contain particles for the longest amount of tim e during that model simulation. Surface probability maps are constructed from the monthly averaged transport quotients. Particles with a salinity restri ction are not included in the tr ansport quotient calculations. Probability maps therefore show areas in Tampa Bay that fall within the salinity tolerances of K. brevis and have the greatest probability of being affected by the bloom based on circulation dynamics. Sampling Water samples were collected during the 2005 K. brevis bloom by FWRI scientists as part of a routine state HAB monitoring program and in response to reported bloom sightings and fish kills. Samples were coll ected at the surface, mid-depth and bottom of the water column from stati ons outside and within Tampa Bay from May to August of 2005. Concentrations of K. brevis cells from the collected water samples will be referred to as background (<103 cells L-1), low (103-104 cells L-1), medium (104-105 cells L-1) or high (>106 cells L-1). Low concentrations of K. brevis cells result in commercial shellfish bed closures, medium concentrations ar e responsible for fish kills and large concentrations cause visi ble water discoloration.
35 Results The three month averaged horizontal current fl ow at the mouth of Tampa Bay (Figure 9) through the Egmont Channel is negative (outward) at all depths with a maximum of 7075 cm s-1 near the surface and around 35 cm s-1 near the bottom; flow through the Southwest Channel is positive (inward) al ong the western boundary, increasing from 10 cm s-1 near the bottom to 40 cm s-1 near the surface, and nega tive throughout the rest of the channel with a maximum speed of 30-35 cm s-1. Across MTB the average current is northward within and above the shipping channel increasing from around 2 cm s-1 at the bottom of the channel to 14 cm s-1 at the surface and extends west of the channel (Figure 9). Fl ow east of the channel is to the south with a maximum speed of more than 14 cm s-1 at the surface. There is a westward component to the flow in MTB to the west of the channel increasing from 2 cm s-1 along the bottom to 8 cm s-1 at the surface (Figure 10 ). There is a weak eas tward component to flow within and above the channel; currents generally are flowi ng to the north within the channel. A strong eastward component to the flow is present to the east of the channel with speed generally increasing towa rd the surface, ranging from 6 cm s-1 near the bottom to >12 cm s-1 at the surface. Current flow across the mouth of HB on average is predominately southward and increases in speed towards the middle of the wa ter column. East of the channel flow is
36 strongly to the south along the easte rn shore. Peak inflow (10 cm s-1) is well to the west of the channel near the Interbay Peninsula. Two areas of weak inflow (<2 cm s-1) are present in the bottom and upper 2 m of the shi pping channel and to the east of the channel near the bottom. Particles ente r HB within these two areas of weak inflow and along the Interbay Peninsula. Averaged current flow diverges across th e mouth of OTB; currents flowing northward into OTB from MTB are confined to the western boundary of OTB while currents flowing out of OTB to the south are confined to the eastern boundary of OTB (Figure 9). There is a strong westward component to the northward flow (maximum 14 cm s-1) from MTB deflecting currents along the western boun dary of OTB and while currents flowing out of OTB are deflected eastward by so mewhat weaker flow (maximum 10 cm s-1 ) (Figure 10). An animation of the surface pa rticle positions for June 2005 shows particles contained in LTB during the first week of the simulation (r esults not shown). On model day 15 of the simulation the majority of the particles are concentrated within the shipping channel to northeast of the Sunshine Skyway Bridge. Some particles are transported into the middle of HB and just south of the Gandy Bridge in OTB by model day 20. The last model day of the June simulation (Figure 11a) shows: 1) the majority of particles located within MTB, 2) some particles being transported throughout HB and 3) a small number of particles traveling no rth of the Gandy Bridge, none going north of the Howard Franklin
37 Bridge. Baywide salinity at the particle locat ions is generally high (> 29) at the end of the month. During the first few days of the July 2005 simu lation particles at th e surface are rapidly transported along the shipping channel to the north of the Sunshine Skyway Bridge. Halfway through the simulation the particles are dispersed throughout MTB. By model day 20 a large number of particles are carried into OTB, most located south of the Gandy Bridge and some mixed between the Gandy Br idge and the Howard Franklin Bridge; very few particles are transported into HB. This pattern persists through the end of the simulation (Figure 11b), with no particles tr ansported north of the Courtney Campbell Causeway in OTB and very few particles ev er making it into HB. Baywide salinity encountered by the particles at the end of July is becoming fresher, as compared to June, with a large portion of MTB just above the K. brevis salinity tolerance of 24. Rapid transport of surface pa rticles along the shipping channel from Egmont Key also occurs during the initi al days of the August 2005 simulation. On model day 10 particles are concentrated in MTB along th e shipping channel have begun to disperse to either side of the Interbay Peninsula. The majority of particles remain tightly contained within the shipping channel in MTB on model day 15; a small number of particles are transported into the middle of OTB and into the middle of HB. The majority of particles remain concentrated near the Interbay Peninsula by model day 25; some particles are transported north of the Howard Franklin Bridge in OTB. By the end of the simulation (Figure 11c) few particles were transported into either upper OTB or upper HB. Those particles in
38 OTB were concentrated along the coasts between the Gandy Bridge and the Howard Franklin Bridge. Model baywide salinity at the end of August remains near tolerance levels in MTB and OTB with in creasing salinities toward LTB. Transport quotients, averaged over the month of June (Figure 12a), are highest along the southwestern coastline near Conception Ke y. A lower proportion of particles are found north of the Sunshine Skyway Bridge within the shipping channel. Transport quotient values decrease significantly in the rest of the bay for the month of June. Averaged transport quotients for July (Figure 12b) rema in high along the southw estern coastline. High proportions of particles can also be s een under the Sunshine Skyway Bridge and contained along the shipping channel in MTB. The highest August transport quotients (Figure 12c) are tightly contained within the shipping channe l. The particle concentrations decrease slight ly to the north near the Inte rbay Peninsula and near the entrances to OTB and HB. Model grid cells in OTB and HB contained particles for proportionately much less time than grid cells in the shipping channel during the three simulations. High concentrations (>106 cells L-1) of K. brevis cells were first detected in January 2005 during routine sampling approximately 15 mile s offshore of Tampa Bay in the Gulf of Mexico. High concentrations of cells were observed in early May at several locations around Palma Sola and low concentrations (103-104 cells L-1) were observed at Anna Maria (Figure 13a). High concentrations pers isted into the middle of May at Palma Sola while medium concentrations (104-105 cells L-1) were detected the third week in May at
39 Anna Maria. One measurement within Tamp a Bay, at Indian Key, contained background concentrations (<103 cells L-1) of K. brevis High cell concentrations persis ted offshore of Palma Sola through June (Figure 13b) as did medium concentrations at Anna Maria Island. At the en d of June high concentrations were detected at Conception Key and in th e MTB shipping channel; low concentrations were measured near St. Petersburg and bac kground concentrations were measured near the mouth of the Little Manatee River. Only background concentrations of K. brevis were detected offshore of Palma Sola throughout the month of July (Figure 13c); me dium concentrations at Anna Maria and high concentrations in the MTB shipping channel persisted th roughout the month. Samples collected at the beginning of Ju ly showed that while previously high concentrations at Conception Key decr eased to medium concentrations, high concentrations were measured just to the nor th at Indian Key. Lo w concentrations were detected near St. Petersburg. The high c oncentrations at Indian Key decreased to medium concentrations in mid-July and persisted through the e nd of the month. One sample collected from the Little Manatee River in the middle of the month showed background concentrations of K. brevis Background concentrations were also found in OTB near the Gandy Bridge at the end of the month; this was the only sample collected from OTB.
40 Background concentrations persisted at Palm a Sola throughout August (Figure 13d). The medium concentrations at Conception Key were also present throughout August, while only background levels were detected at Indian Key. High concentrations were measured in samples collected at the beginning of Augus t at Anna Maria, while concentrations in the shipping channel of MTB decreased to th e medium range. At the end of August the concentrations of K. brevis in the shipping channel of MT B and near St. Petersburg had decreased to background levels. It should be noted that only one sample was collected from OTB during the 2005 bloom since there were no reports of fish kills in the area during the bloom and additionally low salinities typically preclude the survival of K. brevis in this region for any substantial length of time. Sampling in OTB was c onducted the following year during another K. brevis bloom in Tampa Bay. Those data were collected during August and September 2006 (see Appendix C). Discussion An extensive K. brevis bloom was present in Tampa Ba y during the summer of 2005. The bloom entered through the mouth of Tampa Bay and, based on the model results above, was transported into the bay along the dredged shipping channel; these results are similar to those of Steidinger and Ingle (1972) from the 1971 bloom.
41 The short-term averaged flow (June-August 2005) is outward across both the Egmont Channel and the Southwest Channel with the exception of a narrow band of inward flow along the western boundary of the Southwest Cha nnel. This differs from the longer-term averaged flow which is uniformly inward across the Egmont Channel and uniformly outward across the Southwest Channel (Meyers et al., 2007). These results suggest, at least during this short simulation, that high K. brevis cell concentrations first observed in early May 2005 at Palma Sola traveled nor thward and entered Tampa Bay through the Southwest Channel sometime around early June. This cannot be determined conclusively at this time however, because with the ope n boundary at the mouth, the circulation model does not accurately address exchange between Tampa Bay and the Gulf of Mexico (Weisberg and Zheng, 2006). More evalua tions are needed to determine how K. brevis cells enter Tampa Bay. Strong axial surface currents transport simula ted particles from th e mouth of the bay along the shipping channel, out of LTB, under the Sunshine Skyway Bridge and into MTB. Broad flow to the northwest extends we st of the shipping channel at all depths in MTB resulting in the westward dispersion of particles shown in the probability maps. This is agreement with the findings of Meye rs et al. (2007), that surface flow converges (not shown) to the shipping channel both to the south and north of the Sunshine Skyway Bridge and that a westward component of the flow exists to the south of the Interbay Peninsula. The westward component of the flow in western MTB explains the distribution of particle concen trations in the probability ma ps. The probability maps for July (Figure 12b) and August (Figure 12c) show increased concentra tions of particles
42 west of the shipping channel in MTB. Partic les are carried northward into MTB within and along the shipping channel and deflected to the west where they join with southwestward flow (not shown) along th e western shore (Meyer s et al., 2007). An equally strong counter-flowing cu rrent is present along the ea stern coast of Tampa Bay; this current was also observed by Havens et al. (2009). The southeasterly current explains the low concentrations of K. brevis cells found to the east of the shipping channel in both the probability ma ps and in the observations of in situ K. brevis concentrations. Particles entrained in the ne gative current are rapidly transported towards the bay mouth and out of the model domain. The highest transport quotients, and therefor e the areas with the highest probability of being affected by the bloom, are found along the shipping channel in MTB. Similarly, the highest K. brevis concentrations sampled within the bay were along the shipping channel in MTB. These in situ results are in agreement with results from the simulations. Both show K. brevis cells being concentrated along the shipping channel, mostly within MTB. Anecdotal reports suggest th at there were no signs of K. brevis impacts (i.e. fish kills, reports of respiratory distre ss, etc.) in either OTB or HB during the 2005 bloom. Only one water sample was collected by FWRI scie ntists from either OTB or HB during the three month period when the K. brevis bloom was at its peak in Tampa Bay suggesting that fish kills were not reported from either of these areas during this period. OTB is not routinely monitored for K. brevis due to the low salinities typi cally found in that region
43 (see Appendix C). In fact, across the mouth of HB the area of inflowing currents where particles could enter is very small and restricted to flow along the Interbay Peninsula. Particles are restricted from en tering OTB by divergent currents at the mouth that act to deflect particles entering from MTB towa rds the western boundary where they are entrained in southwestward flow (Meyers et al., 2007). The small number of particles that were transported into OTB and HB were able to survive in the model and did not encounter salinity restrictions. These results suggest that circulation features and not a salinity barrier prevented a bloom from formi ng in the northern parts of the bay. Further investigation is needed to de termine why the bloom was not tr ansported into either area. It should be noted that a qua ntitative comparison between th e model simulations and the in situ observations could not be performed due to the scarcity of observations. Sampling is currently limited due to lack of fundi ng and adequately trained personnel (Heil and Steidinger, 2009). The products generated by the numerical ci rculation and particle tracking models accurately reproduce the spat ial distribution of the in situ samples collected during the 2005 K. brevis bloom. This study is the first of many data calibrations to the models with the goal of evaluating the coastal prediction system under real world scenarios. With more robust field evaluations and incorporati on of real-time oceanographic data from the Tampa Bay Physical Oceanographic Real-t ime System (TB-PORTS), the coastal prediction system can serve as a useful for ecasting tool to accurately and rapidly predict future bloom events. This interactive HAB forecasting system, comprised of monitoring
44 and modeling, will provide greater insight into the transport of recurring K. breivs blooms in Tampa Bay and can help mitigate the economic impacts resulting from these HABs.
45 Figure 8 Bathymetric map of the Tampa Bay estuary with the darkest cuts representing the dredged shipping channels. Tampa Bay can be divided into four quadrants: Old Tampa Bay, Hillsborough Bay, Middle Tampa Ba y and Lower Tampa Bay. Bridges and causeways are labeled.
46 Figure 9 Vertical profiles of the north-south (v) co mponent of the horizontal current flow averaged across four locati ons within Tampa Bay: acro ss the mouth of Tampa Bay, across Middle Tampa Bay (aligned with the Li ttle Manatee River), across the mouth of Hillsborough Bay and across the mouth of Old Tampa Bay. The shaded region shows the bathymetry at the given locations. Current s flowing northward (into the bay) are represented by positive velocities while sout hward currents (out of the bay) are represented by negative velocitie s. Velocities are in cm s-1.
47 Figure 10 Vertical profiles of the east-west (u) ho rizontal current flow averaged across two locations within Tampa Bay: across Mi ddle Tampa Bay (aligned with the Little Manatee River) and across the mouth of Ol d Tampa Bay. The shaded region shows the bathymetry at the given locations. Positive ve locities represent an eastward component to the current flow; negative velocities represen t a westward component to the current flow. Velocities are in cm s-1.
48 Figure 11 Numerical model simulations init ialized at the beginning of (a) J une, (b) July and (c) August 2005. The three frames show the transport of model particles, representing Karenia brevis cells, on the last model day of each simulation. The time stamp is in UTC. The scale represents the salinity of the water parcel that the particles encounter.
49 Figure 12 Transport quotients are a ratio between the number of partic les, representing Karenia brevis cells, in each individual grid cell and the total number of particles in the model domai n. Transport quotients range on a scale from zero (low probability of finding K. brevis in a model grid cell) to on e (high probability of finding K. brevis in a model grid cell). The three frames show average transport quotient values for (a) June, (b) July and (c) August 2005.
50 Figure 13 Figures showing concentrations of Karenia brevis collected from water samples at various locations throughout Tamp a Bay for the months of (a) May (b) June (c) July (d) August 2005. The samp les were collected by scientists at the Florida Fish and Wildlife Research Institute in response to re ported bloom sightings and fish kills. The size of the circles indicat e the concentration of K. brevis cells in that location and range from background (<103 cells L-1), low (103-104 cells L-1), medium (104-105 cells L-1) and high (>106 cells L-1) concentrations.
51 Chapter 3: Dispersion of Colored Dissolved Organic Matter Introduction Tampa Bay, the largest open water estuary in Florida, drains a mixed-use watershed about 4500 km2 in size (Zervas, 1993) which plays a la rge role in determining the type of organic material deposited into the bay (Che n et al., 2007b). This organic matter enters into different regions of Tampa Bay through local rivers and othe r freshwater inputs. Tampa Bay is comprised of four main re gions: Old Tampa Bay (OTB), Hillsborough Bay (HB), Middle Tampa Bay (MTB) and Lower Ta mpa Bay (LTB) (Figure 14). A dredged shipping channel, extending from LTB into MTB, bisects the bay before splitting into OTB and HB. Freshwater enters the bay primarily through HB and MTB setting up a buoyancy driven circulation system with mean flow of fresh water out of the bay at the surface and mean flow of sa line Gulf of Mexico water into the bay along the bottom (Meyers et al., 2007; Weisberg and Zheng, 2006). High amounts of organic matter in estuarie s impact the levels of solar radiation penetrating the water column (Blough and De l Vecchio, 2002; Corbett, 2007; Moran et al., 2000). Tampa Bay is characterized by high or ganic content, theref ore solar fluxes are
52 rapidly attenuated with depth (Hu et al ., 2004; Kouassi et al., 1990). Colored or chromophoric dissolved organic matter (CDOM) is the primary factor controlling light attenuation in Tampa Bay (Chen et al., 2007b). CDOM in surface waters diminishes the amount of solar radiation able to penetrate the water column resul ting in a reduction in the quantity and quality of the light that re aches the benthic habitat (Corbett, 2007) while also shielding organisms from UV exposure (Stabenau et al., 2004). Exposure of CDOM to sunlight results in the photochemical de gradation, or bleaching, of the absorption and fluorescence of CDOM (B lough and Del Vecchio, 2002; Kouassi et al., 1990; Morris and Hargreaves, 1997). Ra tes of photobleaching at the surface are controlled by the amount of light absorbed by the CDOM molecules. Several authors have described marine CDOM photobleaching rates under varying conditions and at different locations (Kieber et al., 1990; Kouassi and Zika 1992; Miller and Zepp, 1995; Nelson et al., 1998; Shank et al., 2009; Shank et al., 2005). Coastal areas exhibit varying levels of CD OM concentration depending on seasonal river flow (Blough and Del Vecchio, 2002). Rivers are the dominant source of CDOM in Tampa Bay (Stovall-Leonard, 2003), but other pathways include freshwater inputs such as runoff and groundwater (Coble, 2007; Corbe tt, 2007). Four major rivers discharge the bulk (about 85%) of the freshwater supply into Tampa Bay: the Hillsborough River, Alafia River, Little Manatee River and Ma natee River (Boehme and Coble, 2000; Chen et al., 2007b; Swarzenski et al., 2007). Di stribution of CDOM in Tampa Bay is
53 dominated by conservative mixing between inputs from the Hillsborough River and Alafia River (Chen et al., 2007b). Climatological freshwater discharge rates s how seasonally high flow from approximately June-October during which CDOM abundance is four times greater than during dry conditions (Chen et al., 2007b). Photobleach ing acts as a sink for CDOM, especially during periods of increased freshwater i nput (Del Vecchio and Blough, 2002). During the wet season in Tampa Bay, the buoyancy i nput from freshwater reduces mixing and causes the water column to become highly stratified (Burwell, 2001) Decreased mixing results in a shallow mixed depth, prolonged expos ure of CDOM to sunlight at the surface and greater photobleaching (B lough and Del Vecchio, 2002; Chen et al., 2007b; Nelson et al., 1998), provided that th e light penetration depth is greater than the pycnocline depth. However, shorter residence times duri ng the wet season act to flush out particles (Burwell, 2001) and result in greater amounts of fresh , or un-bleached, CDOM. During periods of low freshwater flow, incr eased vertical mixing limits exposure of CDOM to solar radiation at the surface (Blough and Del Vecchi o, 2002; Chen et al., 2007b). Turbulent mixing acts to transport CDOM below the solar irradation depth, which can be shallow in estuaries depending on the organic content of the water (Kouassi et al., 1990). However, mixing also increa ses the light penetrat ion depth potentially leading to photobleaching below the surface. Longer residence times during low flow conditions which would act to retain CDOM locally for long er periods than during the wet season.
54 An inverse relationship between CDOM and su rface salinity in estuaries has been shown to vary seasonally and between rivers (St ovall-Leonard, 2003) indicating conservative mixing (Chen et al., 2007b; Coble, 2007; Del Vecchio and Blough, 2004; Hu et al., 2004). CDOM distribution in Tampa Bay in particular is determined by the concentration of riverine inputs and the s ubsequent mixing between river and estuarine waters (Chen et al., 2007b). For these reasons CDOM can be used as a proxy for mixing or to trace the freshwater inputs from different riverine sources (Coble, 2007). The use of CDOM in conjunction with its bleaching rate (on the seasonal time scale) may be an effective tracer for evaluating the reside nce time for surface water masses (Nelson and Siegel, 2002). Numerical models offer powerfu l capabilities for prediction and simulation (Huthnance et al., 1993) and have been used to study and trac k organic matter in the Gulf of Mexico and Tampa Bay. The water quality model WASP (Water Analysis Simulation Program) is used to quantify nutrient loads and water qu ality in Tampa Bay (Wang et al., 1999). The Navy Coastal Ocean Model (NCOM) is used as a particle tracking model to follow river discharge and bio-optics in the Gulf of Mexico (Arnone et al., 2005). A similar particle tracking model, if applied to Tampa Bay a nd combined with photobleaching rates, would provide a better understanding of CD OM distribution and seasonality. CDOM distribution maps are a t ool to assist managers in de termining the extent to which UV penetration, water quality and freshwater fl ux will affect estuar ies and coastal areas
55 (Granskog et al., 2007). Hu et al (2004) examined the potential for using satellite technologies for assessing water quality para meters (including CDOM) and to create distribution maps in Tampa Bay for m onitoring purposes. Knowledge of CDOM distribution from its source(s) enhances th e monitoring of water quality (Chen et al., 2007b) and provides a better understanding of w hy some areas of Tampa Bay are more affected than other areas. This study is part of the ongoing development of the Tampa Bay coastal prediction system. The prediction system, comprised of a three-dimensiona l circulation model coupled to a Lagrangian part icle tracking model, is applied here to simulate CDOM dispersion in the Tampa Bay estuary during bot h dry and wet conditions in the bay. The parameterizations of the coastal prediction sy stem are discussed in the following section. The incorporation of a CDOM photobleaching rate is discussed. Evaluations of distribution maps showing the likelihood of ba y-wide CDOM dispersion are performed. Finally, conclusions are drawn as to the seas onal patterns of CDOM transport in Tampa Bay. Methods A numerical circulation model, based on the Princeton Ocean Model (Blumberg and Mellor, 1987), was developed for Tampa Bay (Galperin et al., 1991; Vincent, 2001) and produces three-dimensional fields of circulation, salinity and water level in the bay using
56 quality controlled boundary forcing conditions. Detailed model hydrodynamics and evaluation of the model are repor ted in Meyers et al. (2007). The circulation model includes a particle trac king algorithm (Burwell, 2001) that advects dimensionless particles with in the model domain, a 70 by 100 grid of cells in the horizontal and 11 sigma levels in the vert ical (Meyers et al., 2007), according to the simulated three-dimensional circulation field. The particle tracking model incorporates random-walk diffusion (Dimou and Adams, 1993; Meyers and Luther, 2008) and follows the dispersion of material in tidal and meteorologica lly induced flows using a 4th order Runge-Kutta scheme with model velocity linea rly interpolated to position particles at each model time step. The addition of random displacement terms to the Lagrangian particle tracking model and the parameterizat ion of eddy diffusivity allow for effective modeling of sub-grid scale pa rticle motion (Burwell, 2001). Boundary conditions from 2007 are chosen to perform a hindcast particle transport simulation, each particle repres enting a fraction of total CDOM discharge from four rivers in Tampa Bay, following the methods of Havens et al (2009). The 2007 model data set has undergone extensive quality contro l and 2007 is the last complete year that historical river flow data is available from the United States Geological Survey (USGS). Averaged daily river flow rates for 2007 are obtained from USGS for the Hillsborough River, Alafia River, Little Manatee Rive r and Manatee River and are plotted against USGS historical averaged flow (Figure 15) to determine whether river flow was typical
57 or anomalous during 2007 as compared to the historical trend. From these plots the period of average lowest flow in 2007 (her e called dry season) and the period of average highest flow (here called wet seas on) in 2007 are determined. For purposes of this study differences in flow rates betw een the four rivers are neglected. Each of the model grid cells across the mouths of each of the four rivers (Figure 14) are initialized with particles uniformly thr oughout sigma (throughout the water column) and centered on each cell center. The model outputs the total number of particles in every grid cell each day by summing the particles in a given grid cell at every internal time step (Burwell, 2001). The particle tracking model is initialized both during the dry season and during the wet season. During dry season simulations 1000 partic les are released at the mouths of each of the four rivers (for a total of 4000 particles). As reported by Chen et al. (2007b) approximately four times more CDOM is re leased during the wet season than during the dry season therefore 4000 partic les are released during the wet season simulations from the mouths of each of the rive rs (for a total of 16,000 partic les). During the wet and dry season simulations the particles are allowed to mix for 60 days within the model domain. Particles flushed out of the model domain near the bay mouth are removed from the simulation and not allowed to re-enter the model domain. The particle tracking model records the time and location of each pa rticle within the model grid cells at each time step during th e simulation. These spatial-temporal details
58 are written to files every 60 minutes during the 60-day-long si mulations. Post-processing analyses of the particle counts and locations are performed on the out put files after each simulation. For each model day (every 24 h) post-proces sing decay rates are applied to particle counts contained within the surface levels of the model water column (two uppermost grid cells) to simulate CDOM photobleaching th at occurs at the ocean surface (Clark and Zika, 2000; Vodacek et al., 1997; Whitehe ad and De Mora, 2000). Specific photobleaching rates are not available for Tampa Bay. However, a number of other studies have examined surface photobleachi ng rates of CDOM absorbance during dry (Miller and Zepp, 1995; Nelson et al., 1998) and wet (Kieber et al., 1990; Kouassi and Zika, 1992; Nelson et al., 1998; Shank et al., 2005) c onditions in estuarine and oligotrophic settings (Table 1). Photobleachi ng rates for this study are determined from studies in areas with organic fluxes co mparable to Tampa Bay. A dry season photobleaching rate was obtaine d from work by Miller and Zepp (1995) in a similarly organic-rich estuary along the Georgia coast during winter conditions. Their photobleaching rate of 0.0072 d-1 is applied to dry season si mulations in this study. For wet season simulations, a photobleaching rate of 0.12 d-1 from a study by Shank et al. (2005) in the Cape Fear estu ary in North Carolina, also a highly organic system, is applied. Following the methods of Havens et al. ( 2009), post-processing transport quotients are calculated from the output files (see Appendix A) after photobleaching rates have been
59 applied. Transport quotients are ratios between the number of particles in each individual model grid cell and the total number of partic les in the model domain at any given time of the simulations. Transport quotients are calculated in three-dimensions (x, y and z) for each model grid cell at each time step in the simulations. The transport quotients are then averaged over z and displayed in two-dimens ions on maps to show locations in Tampa Bay with the highest CDOM abundance (t hroughout the water column) during the simulations. The results of post-processing analyses (a pplied decay rate and calculated transport quotients) on the model simulation output f iles are maps of CDOM distributions with incorporated seasonal photobleaching decay rate s. These probability maps show areas in Tampa Bay that are most likely to be affected by CDOM during dry versus wet conditions based on surface bleaching rates and circulation dynamics. Results USGS historical measurements from the f our rivers examined in this study show, on average, 55 years of river flow rates (Figure 14). The lowest average flow rates for the Hillsborough River and Alafia River have historically been from April-June and November. In 2007 the lowest averaged flow rates for these two rivers were from AprilJune. For the Little Manatee River and Mana tee River, the lowest average flow rates have historically been from April-May a nd from November-December. The lowest average flow rates in 2007 for these two rive rs were from May-June and from November-
60 December. Based on the USGS data from all four rivers, a 60 day period in 2007, from April-May, is chosen to represen t the dry season in this study. The highest average flow rates measured by USGS at each river occur historically from August-September. The highest averaged flow rates for 2007 occur in August for the Hillsborough River and October for the other th ree rivers. The second highest flow rates in 2007 occur in August for each of the rive rs except the Hillsborough River where the second highest rates are in September. Based on the overall historical and 2007 river flow data, the wet season for this study was determined to be a 60 day period from September-October. Probability maps showing depth averaged transport quotients for the 2007 dry season generally show the highest transport quotients, and therefore the ar eas with the highest probability of containing CDOM, closest to the rivers from which the particles are released (Figures 16-19). Transport quotient s averaged over the firs t 30 days of the dry season (Figure 16a) for particles (CDOM) entering Tampa Bay from the Hillsborough River are highest in HB at the mouth of th e river, at a few loca tions along the western coast and above the northern dredge spoil is land. Some particles are transported into MTB. After 60 days (Figure 16b) the areas with the highest CDOM probabilities are western HB and northern MTB. CDOM enteri ng the bay from the Alafia River during the dry season is highest in southern HB and along northern MTB for the first 30 days of the simulation (Figure 17a). After 60 days CD OM is more heavily concentrated south of the river in HB (Figure 17b) and more wide ly distributed along northern MTB. CDOM
61 released from the Little Manatee River in th e dry season is concentr ated in northern MTB and extends southward into cen tral MTB along the shipping ch annel after 30 days (Figure 18a). The CDOM distribution is more widesp read in northern MTB after 60 days (Figure 18b) and has higher probabilities. Particles re leased from the Mana tee River in the dry season remain tightly contained within the ri ver and just north the mouth in LTB during the first 30 days of the simulation (Figure 19a) with a few particles being transported northward under the Sunshine Skyway Br idge. After 60 days (Figure 19b) the distribution of CDOM remains the same with slightly more particles entering MTB along the shipping channel. Probability maps averaged over the 2007 wet season generally show more widespread CDOM dispersion throughout the study area than the simulations from the 2007 dry season. Transport quotients 30 days after pa rticles are released from the Hillsborough River (Figure 16c) are highest along th e western coast of HB. High CDOM concentrations extend across MTB and southwes tward along the shipping channel. After 60 days (Figure 16d) the probability distribution is much more widespread with most of HB, MTB and portions of LTB containing very high to high concentrations of particles during the simulation. The probability of finding CDOM 30 days after it has been released from the Alafia River into Tampa Bay during the wet season is highest along the shipping channel and across northern MTB (Fi gure 17c). The proportion of particles in southeastern HB and throughout MTB increases after 60 days (Figure 17d). After 30 days particles released from the Little Manatee River (Figure 18c) are found in the highest proportion to the sout h of the river along the eastern coastline of MTB. High
62 concentrations are also found in LTB. The sa me is true after 60 days of the simulation (Figure 18d); particles are mo st likely to be found along the shipping channel to south of the river. The distribution of particles re leased from the Manatee River is tightly contained along the river and along the southern coast of LTB after both 30 days (Figure 19c) and 60 days (Figure 19d). Some part icles are transported into MTB along the shipping channel. Composite probability maps are constructed to show the total contributions from the four rivers during the dry (Figures 20a-b) and we t seasons (Figures 20 c-d) and the overall bay-wide distribution of CDOM averaged over 30 days and 60 days of model simulations. The total riverine CDOM contribution during the dry season shows the highest transport quotients near the mouths of the four rivers after 30 days (Figure 20a). Moderate concentrations of CDOM are found throughout HB and extending into northern MTB. Low concentrations can be seen throughout most of the study area excluding northern OTB. After 60 days (Figur e 20b) the CDOM distribution remains unchanged throughout the study area but with more transport into OTB. HB and northern MTB contain higher transport quotients. The 30 day wet season composite (Figure 20c) shows the highest pr obabilities near the river mouths and moderate to high probabil ities in western HB and in MTB along the shipping channel. CDOM di stribution after 60 days (F igure 20d) does not change
63 drastically although portions of the shipping channel in MTB show slightly increased concentrations of CDOM as does northern MTB. Surface salinity maps show average bay-wide sa linity during the simulations (Figure 21). During the dry season, beginning in April (Figure 21a) and con tinuing through May 2007 (Figure 21b), most of the study area is in the high salinity range ( 35) with the exception of northern OTB and portions of LTB. CD OM released from the rivers during these months encounter high salinity water due to low freshwater input during this period in 2007. Salinities drop to the 30-35 range thr oughout LTB, MTB and HB for first 30 days of the wet season simulations (Figure 21c). Salinities in OTB range from 25-30. At the end of the wet season (Figure 21d) the proportion of the study area wi th salinities in the 25-30 range has increased. Northern OTB and HB become much more fresh with some areas having salinities <25. A portion of western MTB also becomes significantly fresher in October 2007. Discussion CDOM distribution in Tampa Bay is primarily controlled by mixing (Chen et al., 2007b) and becomes diluted the further it is transp orted away from its ri verine source (Coble, 2007). The ability to discern CDOM away fr om the coastline is related to seasonal freshwater input (buoyancy) and the rate at which the CDOM photodegrades over time with exposure to radiation.
64 The surface probability maps, w ith applied photobleaching rates, show the effect mixing has on CDOM distribution during both low flow and high flow conditions. In the dry season salinities reflect the low amounts of freshwater entering the bay. Velocity profiles (not shown) show that partic les are well-mixed and therefor e have less exposure to the applied surface photobleaching rate. In the we t season stratification occurs due to an influx of freshwater (resulting in lower bay-wi de salinities) and part icles are exposed to surface radiation (simulated by the photobleachi ng rate) for longer periods at the surface. Higher CDOM concentrations are observed in the wet season than in the dry season and the distribution is more wide spread throughout the study area. This is follows what has been reported in Tampa Bay (Chen et al., 2007b; Conmy et al., 2004), that CDOM concentration is inversely proportional to salinity and thus river input. HB is an area known for its poor water clar ity (Hu et al., 2004) and reduced rate of seagrass expansion (Johansson and Greening, 2000). The highest transport quotients are found in HB specifically near the Hillsborough River, an ar ea with historical seagrass loss (Greening and Janicki, 2006) and wher e little mixing occurs. High transport quotients are also found in eastern HB near the Alafia River. This region encompasses an area known as the Kitchen where seagrass m eadows have not expanded in recent years possibly due to poor water quality (Johanss on, 2000). This area of eastern HB was shown by Burwell (2001) to have a significan tly longer residence time, in both high and low streamflow conditions, than along the sh ipping channel to the west, implying that CDOM accumulates in this region.
65 The transport of particles southward out of HB along the eastern coast of MTB follows that observed by Havens et al. (2009) in a study which documented the presence of a strong southwestward flowing current to the eas t of the shipping cha nnel. Water quality and color are optimal for seagrass growth in this region of MTB (Johansson, 2000) supporting the theory that strong currents prev ent CDOM concentration in that portion of the bay, especially dur ing the wet season. This study is a first step in identifying C DOM seasonal distributi on patterns from major freshwater sources, with the inclusion of a photobleaching rate, in order to determine which areas will most likely be affected by CDOM in Tampa Bay. The four largest freshwater sources in Tampa Bay (Hillsbo rough, Alafia, Little Manatee and Manatee Rivers) are chosen to represent the sources of CDOM although a significant amount of freshwater, in the form of runoff and groundw ater (Tomasko et al., 2005), enters OTB as well. Debate exists over the amount of fr eshwater that enters Tampa Bay through OTB therefore that area was not considered in this study as a CDOM sour ce. The scarcity of particles observed in OTB is likely a result of having no freshwater input into OTB; a separate study of CDOM transp ort in OTB is warranted (see Appendix B) given the lack of seagrass recovery in that portion of the bay (Greening and Janicki, 2006). Photobleaching (or decay) is a function of the amount of radiation absorbed by CDOM molecules, primarily at the surface, and gene rally decreases with depth (Nelson et al., 1998). Given the highly organic nature of Tampa Bay and the degree to which high CDOM absorption in estuarie s restricts photobleaching to a thin surface layer (Blough
66 and Del Vecchio, 2002), photoblea ching was not considered below the uppermost grid cells in the model. The biologi cal production (or growth) of CDOM in situ is not considered in this study. No comprehensive CDOM datasets were avai lable to ground truth the distribution maps. Sampling programs in Tampa Bay would need to begin to include CDOM measurements in their routine sampling in order to have in situ comparisons with the simulations (Chuanmin Hu, personal communication). Among other things, knowledge of CDOM sour ces and distribution enhances our ability to monitor water quality (Chen et al., 2007b) and continue seagrass restoration efforts (Tomasko et al., 2005). Composite CDOM probability distribu tion maps are useful tools to guide sampling and ground truth satellite imagery. Tools, such as the prediction system used in this study, can track the effec tiveness of efforts to restore water quality in Tampa Bay and move toward adaptive monito ring and ecosystem-based management in the bay.
67 Figure 14 Bathymetric map of the Tampa Bay estuary. Tampa Bay can be divided into four quadrants: Old Tampa Bay (OTB), Hillsborough Bay (HB), Middle Tampa Bay (MTB) and Lower Tampa Bay (LTB). A dre dged shipping channel, running from LTB to MTB, bisects the bay before splitting into OTB and HB. The Hillsborough River and Alafia River empty into HB, the Little Manatee River empties into MTB and the Manatee River empties into LTB.
68 Figure 15 River discharge rates (ft3 s-1) from the United States Geological Survey (USGS) for the Hillsborough River, Alafia River, Little Manatee River and Manatee River. The blue line represents monthly averaged daily discharge rates in 200 7. The red line represents historical monthly averaged daily river discharge rates over va rious date ranges. The length of the USGS climatological da taset varies between rivers.
69 Figure 16 Transport quotients, with applied po st-processing photoblea ching rate, from the Hillsborough River simulations. The top two panels (a and b) show averaged transport quotients from the dry season simula tions and the bottom two panels (c and d) from the wet season simulations. Transport quotients are calculated for each model grid cell and range in scale from zero (low probabili ty of finding a particle in a grid cell) to one (high probability of finding a particle in a grid cell).
70 Figure 17 Transport quotients, with applied post-processing phot obleaching rate, from the Alafia River simulations. The top two pa nels (a and b) show averaged transport quotients from the dry season simulations and the bottom two panels (c and d) from the wet season simulations.
71 Figure 18 Transport quotients, with applied post-processing phot obleaching rate, from the Little Manatee River simulations. The top two panels (a and b) show averaged transport quotients from the dry season simula tions and the bottom two panels (c and d) from the wet season simulations.
72 Figure 19 Transport quotients, with applied post-processing phot obleaching rate, from the Manatee River simulations. The top two pa nels (a and b) show averaged transport quotients from the dry season simulations and the bottom two panels (c and d) from the wet season simulations.
73 Figure 20 Composite showing the transport quoti ents, with applied post-processing photobleaching rate, from the four rivers (Hillsbo rough, Alafia, Little Manatee and Manatee) simultaneously. The top two pane ls (a and b) show averaged transport quotients from the dry season simulations and the bottom two panels (c and d) from the wet season simulations.
74 Figure 21 Bay-wide averaged monthly surface salinity. The top two panels show surface salinity in the dry season av eraged for the months of (a) April and (b) May 2007; the bottom two panels show surface sa linity in the wet season aver aged for the months of (c) September and (d) October 2007.
75 Table 1 Surface photobleaching rates (d-1) from the CDOM literature for wet and dry conditions at various estuarin e and oligotrophic locations. Location Bleaching rate (wet season) Bleaching rate (dry season) Authors Sargasso Sea 0.011 0.0023 Nelson et al. 1998 Everglades, FL 0.6 Kieber et al. 1990 Biscayne Bay, FL 0.15 Kouassi & Zika 1992 Cape Fear, NC 0.12 Shank et al. 2005 Sapelo Island, GA 0.0072 Miller & Zepp 1995
76 Conclusion This dissertation expands on the work of Vincent (2001) to fu rther develop and evaluate a coastal prediction system for Tampa Bay. The 3-D numerical circulation model underlying the prediction system has been ex tensively validated by Burwell (Burwell, 2001) and Meyers (2007). A pa rticle tracking subroutine within the numerical model was developed and validated by Burwell (2001). His work determined that a Lagrangian based particle tracking method best approximate s the true movement of particles in the bay. This study takes the theore tical application of the coas tal prediction system a step further by incorporating basic biological characteristics on to particle simulations and evaluating the efficacy of those simu lations in real world scenarios. The prediction system is shown to accura tely forecast the physical transport of a contaminant in Tampa Bay. Biological paramete rs are then incorporated one at a time in order to separately evaluate their accuracy for simulating the transport of biological material within the bay. The incorporation of both a tolerance parameter and a decay rate are successful in describing the basic tr ansport mechanisms for algae and CDOM respectively from their sources into the bay. This work introduces a probability calculation th at allows for rapid analysis of bay-wide particle transport. Transport quotients are calculated at each time step of the simulation
77 and are used to compile probability maps of bay-wide particle transport. Providing environmental managers with these maps enable s them to quickly assess areas of highest impact in the bay without re quiring the detailed program ming skills and oceanographic knowledge necessary to build the images. The probability maps can also be tailored to assist scientists in focusing thei r sampling efforts in the field. The previous studies are nume rical approximations to realit y and should be treated as alternatives to costly in situ sampling. The forcing scenarios presented represent real world conditions and provide realistic interp retations of biological transport in Tampa Bay. The end result is a coasta l prediction tool that can be used in real-time to support the management decisions of several e nvironmental agencies in the bay area. Future Work 1) Updated versions of the numerical m odel should include a larger model domain so as to incorporate some particle transpor t outside of the open boundary at the mouth of the bay. At present the open boundary condition is set at zero. This constraint traps particles at the open boundary and does not a llow particles to exit the bay and then reenter. 2) Further evaluations of the coastal prediction system, either by hindcasting previous events or a performing nowcast/forecast in real-time, will improve the accuracy
78 of model predictions. These evaluations s hould incorporate other biological parameters such as sedimentation and growth. 4) Further examination of the strong eastern current al ong the coastline of Tampa Bay is warranted given the pres ence of this current in the velocity profiles of all three studies. 3) An online component of the predicti on system should be developed to allow environmental managers the ability to describe a specific spill or bloom scenario by entering a few initialization parameters. Th is would provide immediate access to model simulations to assist managers with clean-up or mitigation.
79 Literature Cited Arnone, R. et al., 2005. Physical and bio-optic al processes in the Gulf of Mexico: Linking real-time circulation models and sa tellite bio-optical and SST properties, 8th Int. Conference on Remote Sensing for Marine and Coastal Environments, Halifax, Canada. Badylak, S., Phlips, E.J., Baker, P., Fajans J. and Boler, R., 2007. Distributions of phytoplankton in Tampa Bay estuary, U. S.A. 2002-2003. Bull. Mar. Sci., 80(2): 295-317. Bates, S.S., Garrison, D.L. and Horner, R.A., 1998. Bloom dynamics and physiology of domoic-acid-producing Pseudo-nitzschia species. In: D.M. Anderson, A.D. Cembella and G.M. Hallegraeff (Editors), Physiological ecolo gy of harmful algal blooms. NATO ASI Series G 41. Sp ringer-Verlag, Berlin, pp. 266-292. Benkhaldoun, F., Elmahi, I. and Seaid, M., 2007. Well-balanced finite volume schemes for pollutant transport by shallow wate r equations on unstructured meshes. J. Comput. Phys., 226(1): 180-203. Blough, N. and Del Vecchio, R., 2002. Chromophor ic DOM in the coastal environment. In: D. Hansell and C. Carlson (Editors), Biogeochemistry of marine dissolved organic matter. Elsevier Science, London, pp. 509-546. Blumberg, A.F. and Mellor, G.L., 1987. A de scription of a three-dimensional coastal ocean circulation model. In: N.S. Heap s (Editor), Three-dimensional Coastal Ocean Models. American Geophysical Union, Washington, D.C., pp. 1-16. Boehme, J.R. and Coble, P.G., 2000. Charact erization of colored dissolved organic matter using high-energy laser fragmen tation. Environ. Sci. Technol., 34(15): 3283-3290. Bricker, S. et al., 2007. Effects of nutrient enri chment in the nation's estuaries: A decade of change, Silver Spring, MD.
80 Burwell, D.C., 2001. Modeling the spatial struct ure of estuarine residence time: Eulerian and Lagrangian approaches. Dissertation Thesis, University of South Florida, Tampa, FL, 251 pp. Cabana, G. and Rasmussen, J., 1996. Comparis on of aquatic food chains using nitrogen isotopes. Proc. Natl. Acad. Sci., 93: 10844-10847. Cerejo, M. and Dias, J.M., 2007. Tidal transport and dispersal of mari ne toxic microalgae in a shallow, temperate coastal lagoo n. Mar. Environ. Res., 63(4): 313-340. Chao, X., Jia, Y., Shields, J.F.D., Wa ng, S.S.Y. and Cooper, C.M., 2007. Numerical modeling of water quality and sediment re lated processes. Ecol. Model., 201(3-4): 385-397. Chen, X.-h., Zhu, L.-s. and Zhang, H.-s ., 2007a. Numerical simulation of summer circulation in the East China Sea and its application in estima ting the sources of red tides in the Yangtze River estuary a nd adjacent sea areas. J. Hydrodyn. Ser. B, 19(3): 272-281. Chen, Z., Hu, C., Conmy, R.N., Muller-Karge r, F. and Swarzenski, P., 2007b. Colored dissolved organic matter in Tampa Bay, Florida. Mar. Chem., 104(1-2): 98-109. Clark, C. and Zika, R., 2000. Marine organic photochemistry: From the sea surface to marine aerosols. In: P. Wangersky (Editor), Handbook of Environmental Chemistry. Springer, Berlin. Coble, P.G., 2007. Marine optical biogeochemist ry: The chemistry of ocean color. Chem. Rev., 107(2): 402-418. Conmy, R.N., Coble, P.G., Chen, R.F. and Gardner, G.B., 2004. Optical properties of colored dissolved organic matter in the northern Gulf of Mexico. Mar. Chem., 89(1-4): 127-144. Conomos, T., Smith, R., Peterson, D., Hage r, S. and Schemel, L., 1979. Processes affecting seasonal distributions of wate r properties in the San Francisco Bay estuarine system. In: T. Conomos (Edito r), San Francisco Bay: The urbanized estuary. Pacific Division AAA S, San Francisco, CA, pp. 493.
81 Corbett, C., 2007. Colored Dissolved Or ganic Matter (CDOM), Workshop Summary. Charlotte Harbor National Estuary Program, Punta Gorda, FL. Cross, L., 2007. Feather Sound seagrass recovery project: Final repo rt and management recommendations for Feather Sound, Old Ta mpa Bay, Florida, St. Petersburg, FL. Decker, M. et al., 2007. Predicting th e distribution of the scyphomedusa Chrysaora quinquecirrha in Chesapeake Bay. Mar. Ecol. Prog. Ser., 329: 99-113. Del Vecchio, R. and Blough, N.V., 2002. Photobleaching of chromophoric dissolved organic matter in natural waters: Ki netics and modeling. Mar. Chem., 78(4): 231253. Del Vecchio, R. and Blough, N.V., 2004. Sp atial and seasonal distribution of chromophoric dissolved organic matter and dissolved organic carbon in the Middle Atlantic Bight. Ma r. Chem., 89(1-4): 169-187. Dimou, K.N. and Adams, E.E., 1993. A random-walk, particle tracking model for wellmixed estuaries and coastal waters. Es tuar. Coast. Shelf Sci., 37(1): 99-110. Fisher, K. et al., 2006. Annual report of th e Gulf of Mexico Harmful Algal Bloom Operational Forecast System (GOM HAB-OFS). Franks, P. and Signell, R., 1997. Coupled phys ical-biological mode ls for the study of harmful algal blooms. Galperin, B., Blumberg, A.F. and Weis berg, R.H., 1991. A time-dependent threedimensional model of circulation in Ta mpa Bay. In: S.F. Treat and P.A. Clark (Editors), Tampa Bay Area Scientific Information Symposium. Tampa Bay Regional Planning Council, Tampa, FL, pp. 77-97. Geesey, M. and Tester, P.A., 1993. Gymnodinium breve : Ubiquitous in Gulf of Mexico waters? In: T.J. Smayda and Y. Shimiz u (Editors), Toxic phytoplankton blooms in the sea. Elsevier, New York, pp. 251-255. Gomez-Gesteira, M. et al., 1999. A two-dimensional particle tracking model for pollution dispersion in A Corua and Vigo Rias (NW Spain). Oceanologica Acta, 22(2): 167-177.
82 Granskog, M.A., Macdonald, R.W., Mundy, C.J. and Barber, D.G., 2007. Distribution, characteristics and potential impacts of chromophoric dissolved organic matter (CDOM) in Hudson Strait and Hudson Bay, Canada. Cont. Shelf Res., 27(15): 2032-2050. Greening, H., 2004. Factors influencing seagrass recovery in Feather Sound, Tampa Bay, Florida, Pinellas County Environmental Foundation. Greening, H. and Janicki, A., 2006. Toward reversal of eutrophic conditions in a subtropical estuary: Water quality and seagrass response to nitrogen loading reductions in Tampa Bay, Florida, US A. Environ. Manage., 38(2): 163-178. Hansson, S. et al., 1997. The stable nitrogen isotope ratio as a marker of food-web interactions and fish mi gration. Ecology, 78(7): 2249-2257. Havens, H., Luther, M.E. and Meyers, S.D., 2009. A coastal prediction system as an event response tool: Particle tracking si mulation of an anhydrous ammonia spill in Tampa Bay. Mar. Pollut. Bull., 58(8): 1202-1209. Heil, C.A., 1986. Vertical migration of the Florida red tide dinoflagellate Ptychodiscus brevis Thesis Thesis, University of South Florida, Tampa, 118 pp. Heil, C.A. and Steidinger, K.A., 2009. Mon itoring, management, and mitigation of Karenia blooms in the eastern Gulf of Mexico. Harmful Algae, 8(4): 611-617. Herbert, R., 1999. Nitrogen cycling in coastal marine ecosystems. Microbiology Reviews, 23: 563-590. Hu, C. et al., 2004. Assessment of estuar ine water-quality indicators using MODIS medium-resolution bands: Initial results from Tampa Bay, FL. Remote Sens. Environ., 93(3): 423-441. Huthnance, J.M. et al., 1993. Towards Water Quality Models [and Discussion]. Philosophical Transactions: Physical Sciences and Engineering, 343(1669): 569584.
83 Johansson, J., 2000. Water depth (MTL) at the d eep edge of seagrass meadows in Tampa Bay measured by differential GP phase processing: Evaluation and technique, Tampa Bay Estuary Program Technical Publications. Johansson, J. and Greening, H., 2000. Seagrass restoration in Tampa Bay: A resourcebased approach to estuarine management In: S. Bortone (Editor), Seagrasses: Monitoring, ecology, physiology and ma nagement. CRC Press, Boca Raton, pp. 279-293. Kieber, R.J., Zhou, X. and Mopper, K., 1990. Formation of carbonyl compounds from UV-induced photodegradation of humic subs tances in natural waters: Fate of riverine carbon in the sea. Limnol. Oceanogr., 35(7): 1503-1515. Kirkpatrick, B. et al., 2004. Lite rature review of Florida red tide: Implications for human health effects. Harmful Algae, 3(2): 99-115. Korotenko, K.A., Mamedov, R.M., Kontar, A. E. and Korotenko, L.A., 2004. Particle tracking method in the approach for predic tion of oil slick transport in the sea: Modeling oil pollution resulting from rive r input. J. Mar. Syst., 48(1-4): 159-170. Kouassi, A. and Zika, R., 1992. Light-induced de struction of the ab sorbance property of dissolved organic matter in seawater Toxicol. Environ. Chem., 35: 195-211. Kouassi, A., Zika, R. and Plane, J., 1990. Light-induced alteration of the photophysical properties of dissolved organic matter in seawater. Neth. J. Sea Res., 27(1): 33-41. Lanerolle, L.W.J. et al., 2006. Numerical i nvestigation of the effects of upwelling on harmful algal blooms off the west Florida coast. Estuar. Coast. Shelf Sci., 70(4): 599-612. Lewis, R. et al., 1999. The Rehabilitation of th e Tampa Bay Estuary, Florida, USA, as an Example of Successful Integrated Coasta l Management. Mar. Pollut. Bull., 37(812): 468-473. Magaa, H.A., Contreras, C. and Villareal T.A., 2003. A historical assessment of Karenia brevis in the western Gulf of Mexico. Harmful Algae, 2(3): 163-171.
84 Maier Brown, A.F. et al., 2006. Effect of salin ity on the distribution, growth, and toxicity of Karenia spp. Harmful Algae, 5(2): 199-212. Meyers, S.D. and Luther, M.E., 2008. A numeri cal simulation of residual circulation in Tampa Bay. Part II: Lagrangian reside nce time. Estuar. Coast., 31: 815-827. Meyers, S.D. et al., 2007. A numerical simula tion of residual circulation in Tampa Bay. Part I: Low-frequency temporal vari ations. Estuar. Coast., 30(4): 679-697. Miller, W. and Zepp, R.G., 1995. Photochemi cal production of dissolved inorganic carbon from terrestrial organic matter: Si gnificance to the oceanic organic carbon cycle. Geophys. Res. Lett., 22(4): 417-420. Moran, M.A., Sheldon, W.M., Jr. and Zepp, R. G., 2000. Carbon loss and optical property changes during long-term photochemical a nd biological degradation of estuarine dissolved organic matter. Limnol. Oceanogr., 45(6): 1254-1264. Morris, D.P. and Hargreaves, B.R., 1997. The Role of Photochemical Degradation of Dissolved Organic Carbon in Regulating the UV Transparency of Three Lakes on the Pocono Plateau. Limnology and Oceanography, 42(2): 239-249. Morrison, G., Sherwood, E., Boler, R. and Ba rron, J., 2006. Variations in water clarity and chlorophyll a in Tampa Bay, Florida, in response to annual rainfall, 1985 2004. Estuar. Coast., 29(6): 926-931. Nelson, N.B. and Siegel, D.A., 2002. Chro mophoric DOM in the open ocean. In: D. Hansell and C. Carlson (Editors), Biogeoc hemistry of marine dissolved organic matter. Elsevier Science, London, pp. 547-573. Nelson, N.B., Siegel, D.A. and Michaels, A.F., 1998. Seasonal dynamics of colored dissolved material in the Sargasso Sea. Deep-Sea Res. Pt. I, 45(6): 931-957. Owens, E. and Michel, J., 1995. Beach cleaning and the role of technical support in the 1993 Tampa Bay spill, International Oil Sp ill Conference. American Petroleum Institute, Washington, D.C. Periez, R., 2004. A particle-t racking model for simulating pollutant dispersion in the Strait of Gibraltar. Mar. Pollut. Bull., 49(7-8): 613-623.
85 Pritchard, D.W., 1967. Observations of circulation in coastal plain estuaries. Estuaries, 83: 37-44. Proctor, R., Flather, R.A. and Elliott, A. J., 1994. Modeling tides and surface drift in the Arabian Gulf-application to the Gulf o il spill. Cont. Shelf Res., 14(5): 531-545. Raj, P.K., Hagopian, J. and Kalelkar, A.S ., 1974. Prediction of hazards of spills of anhydrous ammonia on water, U.S. Coast Guard, Cambridge, MA. Robinson, A. and Glenn, S., 1999. Adaptive sa mpling for ocean forecasting. Nav. Res. Rev., 51: 26-38. Schofield, O. et al., 1999. Optical monitori ng and forecasting systems for harmful algal blooms: Possibility or pipe dream? J. Phycol., 35(6): 1477-1496. Scott, C.F., 1997. Particle tracking simulation of pollutant discharges. J. Environ. Eng., 123(9): 919-927. Shank, G., Nelson, K. and Montagna, P., 2009. Importance of CDOM distribution and photoreactivity in a shallow Texas es tuary. Estuar. Coast., 32(4): 661-677. Shank, G.C., Zepp, R.G., Whitehead, R.F. and Moran, M.A., 2005. Variations in the spectral properties of freshwater and estuarine CDOM caused by partitioning onto river and estuarine sediments. Estuar Coast. Shelf Sci., 65(1-2): 289-301. Stabenau, E., Zepp, R.G., Bartels, E. and Zika, R., 2004. Role of the seagrass Thalassia testudinum as a source of chromophoric dissolved organic matter in coastal south Florida. Mar. Ecol. Prog. Ser., 282: 59-72. Steidinger, K.A., 1975. Implica tions of dinoflagellate life cycles on initiation of Gymnodinium breve red tides. Environ. Lett., 9: 129-139. Steidinger, K.A. and Haddad, K.D., 1981. Biologic and hydrogra phic aspects of red tides. Bioscience, 31: 814-819. Steidinger, K.A. and Ingle, R.M., 1972. Obse rvations on the 1971 red tide in Tampa Bay, Florida. Environ. Lett., 9: 129-139.
86 Steidinger, K.A., Vargo, G.A., Tester, P. A. and Tomas, C.R., 1998. Bloom dynamics and physiology of Gymnodinium breve with emphasis on the Gulf of Mexico. In: D.M. Anderson, A.D. Cembella and G.M. Hallegraeff (Editors), Physiological ecology of harmful algal blooms. NAT O ASI Series G 41. Springer-Verlag, Berlin, Heidelberg, pp. 133-153. Stovall-Leonard, A., 2003. Characterization of CDOM for the study of carbon cycling in aquatic systems. M.S. Thesis, University of South Florida, St. Petersburg, FL, 216 pp. Stumpf, R.P., Litaker, R.W., Lanerolle, L. and Tester, P.A., 2008. Hydrodynamic accumulation of Karenia off the west coast of Florida. Cont. Shelf Res., 28(1): 189-213. Stumpf, R.P. et al., 2009. Skill assessment fo r an operational algal bloom forecast system. J. Mar. Syst., 76(1-2): 151-161. Swarzenski, P.W., Baskaran, M., Henderson, C.S. and Yates, K., 2007. Tampa Bay as a model estuary for examining the impact of human activities on biogeochemical processes: An introduction. Mar. Chem., 104(1-2): 1-3. Tester, P.A. and St eidinger, K.A., 1997. Gymnodinium breve red tide blooms: Initiation, transport, and consequences of surf ace circulation Limnol. Oceanogr., 42: 10391051. Tomasko, D.A., Corbett, C.A., Greening, H. S. and Raulerson, G.E., 2005. Spatial and temporal variation in seagrass coverage in southwest Florida: Assessing the relative effects of anthropoge nic nutrient load reductions and rainfall in four contiguous estuaries. Mar. Pollut. Bull., 50(8): 797-805. Vincent, M.S., 2001. Development, implementation and analysis of the Tampa Bay coastal prediction system. Dissertation Th esis, University of South Florida, Tampa, FL, 252 pp. Vodacek, A., Blough, N.V., DeGrandpre, M.D ., Peltzer, E.T. and Nelson, R.K., 1997. Seasonal variation of CDOM and DOC in the Middle Atlantic Bight: Terrestrial inputs and photooxidation. Limnol. Oceanogr., 42(4): 674-686.
87 Walsh, J.J. et al., 2002. A numerical analysis of landfall of the 1979 red tide of Karenia brevis along the west coast of Flor ida. Cont. Shelf Res., 22(1): 15-38. Walsh, J.J., Penta, B., Dieterle, D.A. a nd Bissett, W.P., 2001. Predictive ecological modeling of harmful algal blooms. Hum. Ecol. Risk Assess., 7(5): 1369-1383. Wang, P.F., Martin, J. and Morrison, G., 1999. Water quality and eutrophication in Tampa Bay, Florida. Estuar. Coas t. Shelf Sci., 49(1): 1-20. Weisberg, R.H. and Zheng, L., 2006. Circula tion of Tampa Bay driven buoyancy, tides, and winds, as simulated using a finite volume coastal ocean model. J. Geophys. Res., 111(C01005): C01005. Whitehead, R. and De Mora, S., 2000. Ma rine photochemistry and UV radiation. Environ. Sci. Technol., 14: 37-60. Yanagi, T. et al., 1995. A numeri cal simulation of red tide formation. J. Mar. Syst., 6(3): 269-285. Zervas, C., 1993. Tampa Bay oceanography project: Physical oceanographic synthesis, NOAA Technical Report, Silver Springs, MD.
89 Appendix A: Transport Quotient Calculation For any model grid cell within Tampa Bay, th e transport quotient (Q) is defined as TZt z y x p y x Q ) , ( ) ( (1) N t z y x n t z y x p ) , ( ) , ( (2) where, x,y,z are spatial indices within the model grid; t is the time; n is the number of particles in a given cell at any time and N is the total number of particles released.
90Appendix B: Application of th e Coastal Prediction System Feather Sound Project Nutrient loading is a significant problem fo r coastal regions throughout Tampa Bay. The water quality in one region in particular has been determined to be worse than in the rest of the bay (Greening an d Janicki, 2006). Feather Sound is a body of water located in the northwestern portion of Old Tampa Bay (OTB), between the Courtney Campbell Causeway and the Howard Franklin Bridge (Figure 22). The watershed surrounding Feather Sound is prim arily urban and residential with large areas of impervious cover. Hi gh levels of nutrients enter western Feather Sound via discharge from rivers runoff from fertilizers appl ied to lawns and golf courses and sewage treatment plant outfall (Cro ss, 2007). Nutrients entering the region, especially nitrogen which is a limiting nutrien t in coastal marine sy stems (Herbert, 1999), contribute to increased algal growth. Decomposition of the algal blooms causes eutrophication, potentially l eading to hypoxic or anoxic regi ons. The result is that seagrass beds have not recovered as well in Feather Sound as in the rest of Tampa Bay despite efforts over the last several decades to reduce point and non-point sources of nitrogen into the bay (Bricker et al., 2007). The acquisition of real-time data and the abil ity to perform nowcast/forecast simulations are becoming the standard for coastal observations and predictions. Referred to as
91 Appendix B (Continued) adaptive sampling by Robinson and Glenn ( 1999), this predictive method combines observations with model forecasts and has many applications in the marine user community. This study details an applicati on of adaptive sampling and demonstrates the utility of a numerical model as part of a coastal observing network in Tampa Bay, Florida. The purpose of this study is to de termine the extent of nitrogen transport in Feather Sound. From primary discharge locat ions the transport and fate of nitrogen is modeled to examine which regions of Feathe r Sound are most impacted by increases in nutrient input. A discussion of the results follows. This information will be beneficial to understanding the cause of seagrass frag mentation and loss in that region. Methods The Florida Department of Environmental Protection (FDEP) has been monitoring the water quality in Feather Sound for several years. They recently conducted field work using algal sentinels to de tect nitrogen from different sources within Feather Sound (Cross, 2007). Five fresh water discharge locat ions in Feather Sound we re selected in the FDEP study as the sites to anchor the algal sentinels (Figure 22). One site was selected to measure the outfall near the Clearwater wastew ater treatment plant. Two sites were selected to measure the discharge from separa te rivers, Alens Cr eek and Cross Bayou. The final two sites measured runoff from two golf courses located along the Feather Sound coast. The algal sentinels were deploye d at each of the five sites on May 14, 2007 and were recovered on May 22, 2007.
92 Appendix B (Continued) The goal of the FDEP algal sentinel study was to determine in what proportion nutrient sources (i.e. organic versus inorganic) were contributing to nitrogen enrichment in Feather Sound. By performing an isotopic anal ysis of the algal tissue a ratio between two isotopic forms of nitrogen, called the Delta 15-N ( 15N) ratio, was determined. Delta 15N ratios have been shown to increase with population density in a watershed (Cabana and Rasmussen, 1996) and are related to sewage inputs (Hansson et al., 1997). Delta 15-N ratios were calculated by biol ogists at the FDEP based on isot opic analysis of the algal sentinels. A 15N ratio close to zero indicates an inorganic nitrogen source (e.g., fertilizer); a 15N ratio greater than 10 indicates an organic nitrogen source (e.g., animal or human waste) (Cross, 2007). A coastal prediction system, comprised of a numerical circulation model and a particle tracking model, is used to simulate the transport of nitrogen from the five fresh water discharge points used in the FDEP study. Skill assessment of the circulation model as a forecasting tool was performed by Vincent (2001). A forecast simulation is run approximately one week prior to the deployment of algal sentinels at the five study sites. The results from the forecast are to be used to assist biologists from the FDEP with their nitrogen sampling project in Feather Sound and as an evalua tion of the coastal prediction system as a forecasting tool. An input file is constructed containing all of the boundary conditions necessary to run the model in a forecasting mode based on the method of Meyers et al. (2007). The predicted
93 Appendix B (Continued) open boundary water level is computed from local tidal harmonics (Table 1). The calculated amplitudes and phases are verified to be accurate with the water level from the Physical Oceanographic Real-Time System (P ORTS) Egmont Key station. The forecast simulations are initially forced with long term averages fo r temperature ( 25C), salinity (surface: 33.8, middle: 34.1, botto m: 34.3), precipitation (0.42 cmd-1), and winds (1.4 ms1 from 67 true N). Fresh water fluxes are initialized at the beginning of the simulation with the most recent nowcast data transmission. Eight hundred particles, each representing a fraction of the tota l nitrogen input, are simultaneously released from each of the five discharge locations at the beginning of the forecast. The particles are used to simulate nitrogen transport from those locations. The particles are released throughout the water column a nd are advected throughout the model domain by instantaneous model velocity fi elds. The spatio-temporal stamps of the particles are written to a data file every hour during the eight day simulation. It should be noted that the particles did not have a sediment ation rate associated with their transport. Transport quotients are calculated for the model simulation to determine a probability distribution of nitrogen in Feather Sound (Appe ndix A). Cells with the highest transport quotient contain particles for the greatest amount of time during the simulation.
94 Appendix B (Continued) Results Snapshots taken during the forecast simulation show that nitrogen released from the treatment plant and Alens Creek quickly mix together throughout the water column (Figure 23). Similarly nitrogen from the two golf course runoff sites mixes together within a day after entering Feather Sound. Over time nitrogen is transported north and then west from all of the sites in Feather Sound. Some nitrogen is transported into northern OTB through a small gap under the western portion of Courtney Campbell Causeway. The transport quotients from both simulations show that eight days after simulated nitrogen particles are released into Feather Sound the pa rcels of water containing nitrogen are most likely to remain close to the shore and are confined to we stern OTB (Figure 24). The areas to the north of the treatment plan t outfall and Alens Creek are forecasted to have the highest probability of containing nitroge n. A high probability also exists that nitrogen would be found in the area just nor th of the two golf c ourses during the study period. The 15N average ratio near the Clearwater treatmen t plant outfall site is greater than ten and elevated values were found at the mouth of the two creeks (Table 2). The lowest average 15N ratios are found near the golf course sites.
95 Appendix B (Continued) Discussion The model forecast simulation of nitrogen tran sport supports the circ ulation patterns in OTB during conditions that were dryer than no rmal for that month historically (National Weather Service). The overall surface circ ulation in Feather Sound is reduced during periods of low rainfall and the weakened transport is westward into the coast (Meyers et al., 2007). Longer residence times (Burwell, 2001) are a possible explanation for the degraded water quality in Feather Sound as compared with the rest of OTB. The results from the 15N ratios in northern OTB near the wastewater treatment facility suggest organic nitrogen sources, perhaps from animal or human waste found in runoff or wastewater systems. The ratios near the golf courses indicate that nitrogen in this area originated from inorganic nitrogen sources, such as from fertilizers used on the golf courses, in addition to organic sources. This joint study with the FDEP confirms resu lts from previous studi es (Bricker et al., 2007; Burwell, 2001; Cross, 2007; Greening, 200 4) suggesting that local retention of nutrient inputs from various sources (rivers land runoff, and wastewater) could be impeding the successful resurgence of seagrass in Feather Sound. The nitrogen sources ranged from predominately organic to a mixture of organic and inorgani c. This is to be expected in a mixed use watershed like the one surrounding Feather Sound. Further
96 Appendix B (Continued) circulation and flushing scenarios in Feathe r Sound are needed al ong with more rigorous sampling to fully understand the relationship be tween the transport of nutrients and the slow recovery of seagrass in western OTB. Further evaluations of the coastal prediction system as a forecasting are needed for the purpose of establishing sampling strategies prior to work in the field.
97 Appendix B (Continued) Figure 22 Location of discharge points in Feat her Sound for nitrogen source tracking using stable isotopes.
98 Appendix B (Continued) Figure 23 Snapshots of numerical model forecas t simulation for nitrogen source tracking study (May 14-22, 2007). The scale represents the depth of the part icles in the water column with red being at the su rface and blue being at depth.
99 Appendix B (Continued) Figure 24 Probability distribution for nitrogen source tracking study based on forecast simulation run from May 14-22, 2007.
100 Appendix B (Continued) Table 2 Water level harmonics used in th e numerical circulation model. NOS Tidal Harmonics for St. Petersburg, FL Tidal Constituent Amplitude (m) Phase (deg) Speed (deg/hr) M2 0.175 197.0 28.9841042 S2 0.057 211.7 30.0000000 N2 0.030 191.3 28.4397295 K1 0.167 49.9 15.0410686 O1 0.155 37.7 13.9430356 P1 0.049 57.6 14.9589314
101 Appendix B (Continued) Table 3 Average delta 15-N values from di scharge points in Feather Sound. Sample Location Avg d15N Clearwater treatment plant 11.61 Alen's Creek 8.18 Cross Bayou 8.05 Golf Course #1 7.50 Golf Course #2 7.43
102Appendix C: Old Tampa Bay Karenia brevis samples During the 2005 Karenia brevis bloom no samples were coll ected within Old Tampa Bay (OTB) since no reported fish kills occurred in that area during the bloom. Samples were collected in OTB during a 2006 K. brevis bloom for comparison purposes with the 2005 K. brevis data. These samples, collected in August and September of 2006 from the mouth of OTB to the Howard Franklin Br idge, showed background concentrations of K. brevis during the 2006 bloom.
103 Appendix C (Continued) Figure 25 Karenia brevis samples collected in Old Tampa Bay during a 2006 K. brevis bloom. The samples are classifi ed as background concentrations.
About the Author Heather Havens received a Bachelor of Arts degree in biology from Agnes Scott College in 2001. While attending Agnes Scott, she pa rticipated in a marine science summer study program offered through Boston Universitys School for Field Studies. Through this program Heather spent a summer in the Turks and Caicos studying management of marine protected areas. She went on to pur sue an advanced degree in marine science from the University of South Carolina wher e she earned a Master of Science degree in 2004. Under the advisement of Dr. Bjrn Kj erfve her masters research focused on the dispersion of snapper eggs from a spawning si te within the Belize Barrier Reef. Heather entered the Ph.D. program at the Univers ity of South Florida in 2004. Under the advisement of Dr. Mark Luther her doctor al research focused on the evaluation of a coastal prediction system for guiding water qua lity monitoring in the Tampa Bay estuary.
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 2200409Ka 4500
controlfield tag 001 002069312
007 cr mnu|||uuuuu
008 100421s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003193
Havens, Heather Holm.
Towards the development of a coastal prediction system for the Tampa Bay estuary
h [electronic resource] /
by Heather Holm Havens.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 103 pages.
Dissertation (Ph.D.)--University of South Florida, 2009.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
ABSTRACT: The objective of this research is to evaluate a coastal prediction system under various real world scenarios to test the efficacy of the system as a management tool in Tampa Bay. The prediction system, comprised of a three-dimensional numerical circulation model and a Lagrangian based particle tracking model, simulates oceanographic scenarios in the bay for past (hindcast), present (nowcast) and future (forecast) time frames. Instantaneous velocity output from the numerical circulation model drives the movement of particles, each representing a fraction of the total material, within the model grid cells. This work introduces a probability calculation that allows for rapid analysis of bay-wide particle transport. At every internal time step a ratio between the number of particles in each individual model grid cell to the total number of particles in the entire model domain is calculated.These ratios, herein called transport quotients, are used to construct probability maps showing locations in Tampa Bay most likely to be impacted by the contaminant. The coastal prediction system is first evaluated using dimensionless particles during an anhydrous ammonia spill. In subsequent studies biological and chemical characteristics are incorporated into the transport quotient calculations when constructing probability maps. A salinity tolerance is placed on particles representing Karenia brevis during hindcast simulations of a harmful algal bloom in the bay. Photobleaching rates are incorporated into probability maps constructed from hindcast simulations of seasonal colored dissolved organic matter (CDOM) transport. The coastal prediction system is made more robust with the inclusion of biological parameters overlaid on top of the circulation dynamics.The system successfully describes the basic physical mechanisms underlying the transport of contaminants in the bay under various real world scenarios. The calculation of transport quotients during the simulations in order to develop probability maps is a novel concept when simulating particle transport but one which can be used in real-time to support the management decisions of environmental agencies in the bay area.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Advisor: Mark E. Luther, Ph.D.
x Marine Science
t USF Electronic Theses and Dissertations.