Philadelphia Water Department’s Delaware SWAP: A Comprehensive Approach Mark Maimone, CDM Christopher S. Crockett, Philadelphia Water Department ABSTRACT The Philadelphia Water Department (PWD) is currently completing Source Water Protection Assessments (SWA) for eight drinking water intakes on the Delaware River. One of the intakes, PWD’s Baxter drinking water intake, is located on the Delaware River downstream of Trenton, in a portion of the river that is tidally influenced. Because of its location, the Baxter intake presents several unique challenges, and the resulting Source Water Assessment is one of the most complex of any assessment in the country. The contributing watershed covers over 3000 square miles, and includes numerous tributaries. Watershed land use ranges from heavily urban (Philadelphia, Trenton), to suburban areas north of Philadelphia, to rural and undeveloped areas near the New York/Pennsylvania border. The Source Water Assessment Program (SWAP) on the Delaware River has several unique aspects, including: The development of a comprehensive, point source database for the entire Delaware River watershed. The database is programmed to locate the thousands of potential point sources in relationship to the river and tributary, estimate potential contaminant loading for 10 contaminant categories from each source, estimate potential contaminant concentration at the intake from each of the sources, and estimate travel time to the intake under high water flow conditions from each of the sources. • • The development of one of the largest applications of a stormwater model using the USEPA SWMM code to estimate non-point source contaminant loading to the Delaware River for nine of the 10 contaminant categories. The SWMM model is a continuous simulation model. • The development and application of a tidal zone model to estimate time of travel and pollutant concentrations for releases to the river within the tidal zone. Because of the dynamics of tidally influenced flow in the lower Delaware River, contaminant releases downstream of the intake can be carried to the intake, and must be included in the analysis. • The use of sophisticated decision support software to screen the thousands of point sources, and to integrate the point sources and non-point sources into a single evaluation to identify the 100 highest priority sources for the Baxter intake. The tidal model, database calculations, and stormwater pollutant loading model all provide input into the decision support program to produce the priority lists. The complex approach needed to integrate all the diverse elements of the study into a coherent assessment was developed to meet the challenge of producing over 50 SWAP reports for intakes on both the Delaware River and Schuylkill River. It represents a template for developing a consistent, comprehensive assessment for large river systems, 1 and its use for the Baxter Intake SWA is an excellent case study to demonstrate a successful application. INTRODUCTION The Safe Drinking Water Act Reauthorization in 1996 included a specific component for source water protection called the Source Water Assessments (SWAs). The SWAs are a process involving water suppliers, watershed organizations and other stakeholders, which identify the protection priorities of the water supply. As part of its federal requirement to conduct the SWAs, the Pennsylvania Department of Environmental Protection (PADEP) sought to involve water suppliers and the community in the SWA process. Using a partnership approach that includes water suppliers working with the state, a Source Water Assessment Partnership was formed for the Delaware River Watershed. The Philadelphia Water Department (PWD) is leading the Source Water Assessment Partnership and is conducting source water assessments for 8 surface water supplies within the Delaware River Watershed. Originating in the Catskills (Schoharie County), New York, and flowing south to the mouth of the Delaware Bay in Philadelphia, Pennsylvania, the 330 mile-long Delaware River winds its way through four states on the eastern coast of the United States, encompassing 42 counties and 838 municipalities in the Mid-Atlantic Region of the country. The Delaware River flows southeast for 78 miles through rural regions along the New York-Pennsylvania border, heads southwest, along the border between Pennsylvania and New Jersey. It turns southeast again at Easton, PA, where the Delaware River is met by the Lehigh River (its second largest tributary). The Delaware then flows approximately 80 miles to the tidal waters of Trenton, New Jersey, thus completing about 200 miles of its 330-mile journey. About 30 miles downstream of Trenton, the river passes through the fifth largest metropolitan region in the nation—the heavily industrialized Philadelphia (PA)/Camden (NJ) area—and the mouth of the Schuylkill River, its largest tributary, which flows into the Delaware. From there, the river flows on past Wilmington, Delaware and through the more rural regions of Cape May, New Jersey on its eastern shore and Cape Henlopen, Delaware on the west, completing its course as it meets the Delaware Bay. Along its route from the headwaters to the mouth of the bay, the Delaware River drains a total of 13,539 square miles (0.4% of the land mass in the U.S.) in New York, Pennsylvania, New Jersey, and Delaware. The Delaware River, its bay, and 216 tributary streams play a significant role in sustaining life and the economy in these areas. Among other things, these bodies of water are used for fishing, transportation, power, cooling, recreation, and other industrial and residential purposes. Most importantly, though, they provide drinking water for about 17 million people, or almost 10% of the country's population. Figure 1 presents a map of the entire Delaware River Drainage Basin, its major subwatersheds, and its tributaries. 2 Figure 1 - Map of Delaware River Drainage Basin 3 SWAP METHODOLOGY Based on PADEP guidelines for the statewide SWAP, contaminant inventories of thousands of potential point sources and hundreds of non-point sources were developed. The inventory is an essential part of assessing the drinking water supply for the intake, because it compiles potential contaminant sources within the 5-hour, 25-hour, and beyond 25-hour time of travel delineation zones. The following federal databases were accessed for point sources in the study areas: Permit Compliance System (PCS); Resource Conservation and Recovery Act Information System (RCRIS); Comprehensive Environmental Response, Compensation, and Liability Act Information System (CERCLIS); and Toxic Release Inventory (TRI). Regulated aboveground storage tanks (ASTs) were also compiled from the PADEP Storage Tank Program. In addition to the point sources, non-point sources were included by dividing the watershed into subwatersheds and developing a stormwater runoff loading model to simulate pollutant loading from each subwatershed. A database management system was created to assist with storing parameter data, creating the model, and post-processing model outputs, as well as point source inventory data. Because of the large number of potential sources of contamination that have been identified, a process of successive screenings was required to narrow down the list of sources to the those that represent the greatest potential threat to the intake. These screenings helped focus the efforts of source water protection on those sources that have the greatest potential to affect the water quality of the source water at an intake. Figure 2 shows the process of successive screenings that helped narrow down the list of high priority sources from over 5000 to the 100 top priority sites. The process can be summarized as consisting of two types of screenings: geographic screening, and significance screening. 4 Figure 2 - Source Prioritization Flow Diagram Develop Initial Databases Landuse/Subwatershed Development Geographic Screening (Delineation Zones) SWMM Runoff Model Simulation Source Contaminant Data Runoff Contaminant Loads Tidal Zone Modeling Significance Screening Source Priorities Map Significant Sources Screening Processes. Potential source data comes from a number of data sources, and each database can contain hundreds of potential sources. Less significant point sources needed to be screened out, leaving only the most important sources for final ranking. Potential non-point sources were identified using the SWMM model and Event Mean Concentrations (EMCs) to calculate total annual pollutant loading for each subwatershed. A slightly different screening approach was needed for each type of source because of the data available and the structure of the databases. Figure 2 is a flow diagram of the screening and ranking process that was used to successively select the most important sites from each of the databases available, and combine them in an organized manner to produce a final list of high priority sites. The process can be compared to a playoff elimination process, with various divisions (data sources) providing a set number of teams (potential point and non-point sources) to the overall playoff. Like such playoff structures, it can occur that a site will not be included in the final list because it was eliminated in competition with other sites within its categories. (To follow the analogy, the 4th best team in a division is not invited to the playoffs, even if 5 it is better than the 3rd best team from another, weaker division, because only the top three teams are invited from each division.) Despite this fact, the process does provide the top sites from each database category, and provides valuable insight into the relative importance of each category of sites. Enough sites were included from each category to make sure that no highly ranked sites would be overlooked. The diagram shows that there were two general screening steps (or elimination rounds) leading to the final ranking. These are described briefly below. Geographic Screening. The PADEP zone concept was used to narrow the list of sources down to include only those with higher priority. The first screen applied to eliminate less important potential sources makes use of the zone concept recommended by PADEP for use in the SWAP: Zone A: the critical segment covering ¼ mile on either side of the stream upstream of the intake within a 5-hour travel time to the intake. All potential sources within this zone are included in the subsequent steps. Zone B: a second segment located within 2 miles of either side of the stream upstream of the intake, within a 25-hour travel time to the intake. All potential sources within this zone are also included in the subsequent steps. Zone C: the rest of the upstream watershed. These sources remain listed in the database, but are eliminated from further analysis because they are deemed less significant than sources in zones A and B. Potential sources within Zone C sources are dropped from further analysis within this preliminary assessment, leaving those sources within zone A or B for the intake. Significance Screening Significance screening was used to identify those sources that could deliver enough contaminant to the river to have a measurable impact on water quality at the intake. This was done either by simple threshold screening, based on the amount of contaminants stored or used, or by a more complex evaluation using several criteria and a decision support software tool. The percent change in the concentration of a chemical at the intake due to releases from each site was roughly estimated and was used to screen PCS (NPDES) sites. This threshold screening was performed to select the largest dischargers. A cutoff of a 1 percent change in concentration at the intake was established, based on the percent increase by the discharged mass loading. For some of the point source types, as well as for the many sub-watersheds contributing stormwater pollutants to the stream, a more sophisticated screening was performed using EVAMIX. EVAMIX is a matrix-based, multi-criteria evaluation program that makes use of both quantitative and qualitative criteria within the same evaluation, regardless of the units of measure. The algorithm behind EVAMIX is unique in that it maintains the essential characteristics of quantitative and qualitative criteria, yet is designed to eventually combine the results into a single appraisal score. The overall appraisal score is 6 used to determine the final ranking of alternatives from best to worst, or most important to least important. Criteria for the ranking varied. Criteria examples included the expected concentration of contaminant at the intake, the age of the storage tank, storage volume of the tank, chemical ranking based on the mix of chemicals onsite, whether there had been leaks in the past, the location relative to the river, and the travel time to the intake. Finally, all the significant (those that passed the screening) point sources and runoff loads (entered as pseudo point sources) were prioritized, accomplishing the main goal of the assessment. There were two types of final rankings. The first ranking was a combined ranking of sites from all categories, compared against each other. The second ranking was by contaminant type, with all significant sources contributing to a particular contaminant category included. Runoff Loading Model One of the key features of this SWAP was the fact that stormwater impacts were evaluated against point sources. In order to develop estimates of stormwater pollutant loads on an annual basis, a stormwater model for a large portion of the Delaware River watershed was developed. The RUNOFF module of the U.S. EPA’s Storm Water Management Model (SWMM) was used to simulate rainfall-runoff quantities and quality at specified inlet locations. Figure 3 displays the structure of the SWMM RUNOFF model. The model inputs sub-watershed parameters, rainfall time-series, climate data, and event mean concentrations (EMCs) for the land use categories, and outputs annual and monthly pollutant loads for the length of the simulation period. The model incorporates infiltration, depression storage, and roughness to estimate runoff flow and ultimately, runoff pollutant quantities. Figure 3 - Watershed Loading Model Schematic Diagram PRECIPITATION (long-term gauge record) accounts for: • antecedent moisture • wet, dry and normal years SURFACE RUNOFF as a function of: • imperviousness • slope • depression storage • evaporation • soil infiltration • snowmelt Land use 1 EMC 1 Land use 2 EMC 2 Land use 3 EMC 3 etc. Stream Network Event Mean Concentration function of: • pollutant • land use 7 The model helped to identify subwatersheds contributing more or less pollutants to the stream. Pollutant loading varied depending on the landuse. The subwatersheds within the southern portion of the Delaware SWAP study area are much more developed than the subwatersheds in the northern portion up through New York. The land surrounding the Delaware River within the Middle Delaware Subwatershed is part of the National Scenic River Corridor and contains the Delaware Water Gap National Recreation Area, which remains protected from development under Federal regulation. The Lehigh River subwatershed and the tidal portions of the Lower Delaware contain the largest amount of developed land within the study area, as illustrated in Figure 4. The tidal area of the Delaware River near Philadelphia remains the most populated and densely developed subwatershed of the entire study area, despite the rapid development activities occurring in the other subwatersheds. The Delaware Watershed Study model is composed of 391 subwatersheds. The subwatersheds were further divided into land use categories to track the contributing pollutant loads from each land use category. The land use categories were based on the USGS’s NLCD dataset updated with 2000 Census data for residential and commercial areas. The amount of impervious surface in the watershed significantly affects the quantity of surface runoff. The adjusted land use designations were used to assign calculated impervious cover values for each sub-watershed-land use planning unit. The percentage of impervious area for all land use categories, excluding residential, were estimated using literature values and adjusted during the calibration. The percentage of impervious area for residential areas were calculated using Stankowski’s methodology, which calculates the percentage of total impervious area as a function of the population density. The rainfall-runoff from impervious surfaces is also affected by slope, depression storage, and evaporation, and these factors were included in the model. The amount of surface runoff is primarily driven by the precipitation. Long-term climate and precipitation records were used to drive the hydrology of the system. Using a longterm record represents a wide range of hydrologic conditions that occur in a given climate. The hourly rainfall data were obtained from the National Weather Service (NWS) at stations in and surrounding the Delaware Watershed. The hourly data was further discretized into 15-minute increments. To account for snowmelt, the daily minimum and maximum temperatures and average monthly wind speeds were obtained for the period of simulation. 8 Figure 4 – Update Landuse in the Delaware River Watershed 9 Hydrograph Separation for Baseflow and Runoff Calibration. In order to assess the reliability of the pollutant loads from SWMM, a hydrograph separation analysis of stream gage data was performed to provide model calibration targets. A hydrograph separation program was created in SAS® to divide the total flow into baseflow and surface runoff. This program was modeled after the USGS’s HYSEP computer program, but assumes only one of its three hydrograph separation methods, the sliding-interval method. The hydrograph separation yields total flow, baseflow, and runoff values in daily, monthly, seasonal, and annual averages. The daily average flows were obtained from the USGS for gages located in the Delaware River Basin. An example of baseflow separation is shown in Figure 5, with the darker shaded areas showing runoff, and the lighter areas the baseflow. Figure 5: Hydrograph Separation Analysis for the Perkiomen Creek at Graterford for May 1990. USGS Gauge 01446500 (Delaware River at Belvidere) August 1992 7000 A = 4,535 sq. miles N = A 0.2 = 5.4 days 2N = 10.8 days 2N* = 11 days 0.5(2N*-1) = 5 days Total Flow Baseflow 6000 5000 Flow (CFS) Surface Runoff 4000 3000 2000 1000 8/31/92 8/30/92 8/29/92 8/28/92 8/27/92 8/26/92 8/25/92 8/24/92 8/23/92 8/22/92 8/21/92 8/20/92 8/19/92 8/18/92 8/17/92 8/16/92 8/15/92 8/14/92 8/13/92 8/12/92 8/11/92 8/9/92 8/10/92 8/8/92 8/7/92 8/6/92 8/5/92 8/4/92 8/3/92 8/2/92 8/1/92 0 Date The USGS streamflow hydrograph separation results were used to calibrate the results from the models. Comparing the simulated values with the hydrograph separation results, parameters in the runoff loading models were further refined. 10 Tidal Zone Hydrodynamic Modeling An important part of the process was the estimate of potential concentration at the intake of contaminants released from any of the thousands of potential sources. It was also important to estimate the time of travel along the river that a released contaminant would take before reaching the intake. On most rivers, this calculation only has to account for the velocity of the stream, and the expected dilution produced by the flow rate at the intake. Below Trenton, however, the tidal influences make the estimate of time of travel to an intake much more complex. Time of travel could only be estimated by using a three dimensional, time variable hydrodynamic and water quality models developed for the Delaware River Basin Commission (HydroQual, 1998). The hydrodynamic model is a version of the Estuarine, Coast and Ocean Model (ECOM) developed by Blumberg and Mellor (1980, 1987). This model shows that transport of contaminants within the tidal zone is a combination of advective transport (flow with the river water), and diffusion of the contaminant. The advective transport actually reverses direction with each tidal cycle, and net movement downstream is very slow. Diffusion is the primary mechanism that moves the contaminants from the point of entry to the intake. Based on a series of model simulations where contaminants were inserted at different sections of the river, the simulated time of arrival of the contaminants were plotted. In general, the net “velocity” of the contaminant based on the initial time of arrival at the intake divided by the distance from the insertion point and the intake was usually between 7 and 10 feet per second. This occurred whether the contaminant was moving upstream from below the intake, or downstream from above the intake. Most of this movement in the tidal zone was the result of diffusion. Based on these results, the velocity of contaminants moving along the river in the tidal zone was assumed to be an average of 8 feet per second, and all time of travel calculations used this average velocity for extreme or worst case conditions. Another of the criteria used in the screening and final evaluation required an estimate of the concentration of the contaminant once it reached the intake. Because the intake is located in the tidal zone, two distinct reductions in concentration will occur. The first is due to the mixing of the contaminant with river water. The second occurs in the tidal zone as the contaminant diffuses due to the back and forth tidal motion. To estimate this, the following approach was used. The contaminant mass was assumed to enter the river at the nearest point to the potential point source, or at the downstream end of a subwatershed for non-point sources. River dilution was calculated by dividing the mass of contaminant assumed to have entered the river by the median flow at the intake. In this case, this is based on the median flow at the Trenton gage. This provides the first dilution. 11 Figure 6: Tidal Zone Model Results Maximum Relative Concentration at Baxter Intake Intake 10.00 1.00 0.10 140.0 130.0 120.0 110.0 100.0 Concentration at Intake 100.00 0.01 90.0 River Mile of Spill Release (100% Conc.) Because the intake is in the tidal zone, a second reduction in concentration occurs, this time based on the diffusion occurring in the tidal zone, as well as additional dilution as tidal water moves in and out of the tidal zone. To calculate this, the hydrodynamic model results were analyzed. Figure 6 shows the concentration of a 100 ppm point discharge at various points along the river at the Baxter Intake. Based on the results of model simulations shown in the figure, a reasonably conservative estimate of the tidal dilution effect is that the concentration at the intake will be only 10% of the concentration in the river for all spills or discharges occurring upstream of the intake. For spills or discharges occurring downstream of the river, the concentration of the contaminant at the intake will be only about 1% of the concentration in the river. ASSESSMENT RESULTS EVAMIX output was also used to complete the assessment by ranking all the sites that passed the screening evaluation in descending order of importance and then selecting the top sites based on the results of the ranking. EVAMIX was used to rank all sources over the entire range of contaminant categories. 12 A number of evaluation criteria were used in the evaluation, including: Relative Impact (concentration at the intake compared to drinking water standard) Travel Time from source to intake Removal Capacity of Treatment System Impact of contaminant on treatment process at water treatment plant Health Impacts of contaminant Potential For Release/Control of contaminant at the site Potential Release/Frequency of releases in the past Violation Type/Frequency of violations in the past Location relative to the river Full ranking allowed us to compile a final list of sources, independent of contaminant class. Ranking by contaminant category was also completed using the multi-criteria evaluation program EVAMIX. Results from the evaluation of each of the ten contaminant categories resulted in a listing of high, medium, and low priority sites for that contaminant category. The final results of the rankings were broken down into six major categories according to PADEP’s SWA Plan. These were represented by designations A through F, with A representing sources of highest protection priority and gradually decreasing to F for sources of lowest protection priority. This designation process was initially designed for intakes with a limited number of sources where the whole inventory could be ranked. However, given the large number of sources and the ranking process, sources that were represented by designations D through F were screened out in the significance screening process. Therefore, the remaining sources are all considered potentially significant sources of contamination and fell into categories A through C. The designations are described in Table 1. As shown, the sources in categories A through C may require additional ground-truthing in order to provide a more accurate designation of their significance. Although not considered to be potentially significant, sources in category D may need to be evaluated as more information becomes available. 13 Table 1 - Contaminant Source Ranking Designations Designation Description Potentially Significant Sources of Contamination to Water Supply A Potentially Significant Source of Highest Protection Priority B Potentially Significant Source of Moderately High Protection Priority C Potentially Significant Source of Moderate Protection Priority Remaining Sources From Inventory Screened Out By Significance Screening Criteria D Potential Source of Moderately Low Protection Priority E Potential Source of Low Protection Priority F Potential Source of Lowest Protection Priority Figure 7 shows the final ranking of sites for the Baxter intake. These “combined” results are based on all sites and all potential contaminants. The results provide significant insight into the relative threat that various types of sources might have on the water quality at the intake. The key results are: All of the highest ranked sites are either NPDES sites from the PCS database or stormwater pollutant loading represented by various subwatersheds. The top 22 ranked sources are NPDES, as well as 28 of the top 30 sites included as priorities in this combined rankings. Stormwater or NPS loading appears to also represent a high priority. There are 2 subwatersheds with stormwater related loading in the top priority sites. TRI sites are generally ranked lower. There are no TRI sites as high priority sites, and all TRI sites are found in the “moderate” priority category. RCRA sites, with or without ASTs, are generally ranked the lowest of all of the types of sites. Only 9 of these sites made it into the top priority sites. Results indicate that with a balanced assessment, those contaminant sources that are actually discharging to the river (NPDES permitted point sources or stormwater runoff) represent the greatest concern. Those with only the potential to release contaminants through spills or leaks (TRI, RCRA, AST) are generally given a lower priority. Despite the low overall rankings, the highest potential relative impacts appear to occur with the TRI and AST sites. The relative impact numbers show that, were a catastrophic spill or leak to occur at these highly ranked sites, concentrations at the intake could potentially be very high. 14 Health Impacts, as scored in the assessment, had a large influence on the resulting rankings, with those sites ranked high on potential health impacts ranking as important sites. Treatment Impacts were also important in the final rankings, with those sites scoring high on potential impact to the treatment process also ending up highly ranked in the overall assessment. The geographic distribution of significant sources showed that most of the category A sources were from point sources in drainage areas of the Delaware River below Trenton and stormwater runoff sources in the upper Delaware River Watershed. A comparison of the types of sources indicated by the ranking process with the sources indicated by water quality analysis and impaired stream information corroborates that NPDES discharges and polluted runoff (non-point sources) from developed areas are the most important influences on water quality at the PWD Baxter Intake. In addition to assessing priorities of all the sites based on combined results (all contaminant categories considered), a separate assessment was made for each of 10 contaminant categories. These categories were developed to represent groups of contaminants of concern for an intake. The 10 categories included . Cryptosporidium Petroleum Hydrocarbons Disinfection by-Products Salts Metals/Heavy Metals Total Suspended Solids Conservative Nutrients (Nitrate) Total/Fecal Coliform Non-conservative Nutrients (Phosphorus) Volatile Organic Compounds A table of the highest priority sites, and a map was produced for each of the 10 contaminant categories. Figure 8 illustrates PWD’s Baxter Intake’s priority point sources and subwatersheds for fecal coliform in the lower Delaware River Watershed. 15 Figure 7 Priority Point Sources and Subwatersheds for PWD’s Baxter Intake in the Lower Delaware Watershed 16 Figure 8 - Priority Point Sources and Subwatersheds for PWD’s Baxter Intake for Fecal Coliform in the Lower Delaware River Watershed 17 The following are overall observations of the contaminant category results. Salts: The highest priority sources of chlorides were either stormwater runoff from urbanized watersheds, or potential releases of industrial salts from industrial sites as represented by sites listed in the TRI database. It should be noted that neither type of source individually appeared to provide sufficient loading to cause water quality impairments at the intake, but combined, especially during winter periods, the runoff may result in some impacts. Cryptosporidium: Sources of pathogens were either stormwater runoff from agricultural or urbanized watersheds, and permitted discharges from wastewater treatment plants. NPDES sources were represented in the high priority category, while NPS sites were in the lower category. Most sources appeared to be relatively minor contributors. However, there were some sources that could provide sufficient loads to have a cumulative impact on the water quality. The overflows of raw sewage during wet weather events were roughly estimated and compared to the other potentially significant sources. Fecal Coliform: Sources were either stormwater runoff from agricultural or urbanized watersheds, and permitted discharges from wastewater treatment plants. Although both sources were represented in the high priority category, the results suggest that periodic loading from stormwater was orders of magnitude higher than the loading from wastewater treatment plants. During dry weather flows, wastewater loading is insignificant at the intake, but during storm events, fecal coliform would be expected to increase by orders of magnitude. The overflows of raw sewage during wet weather events were roughly estimated and compared to the other potentially significant sources. There was a broad geographic distribution of potentially significant sources of fecal coliforms in the watershed. This may be due to the fact that die-off was not factored into the analysis. Metals: Results generally show that NPDES permitted discharges were the primary sources. Some TRI sites with significant storage or use of metals were also rated as high priority sources, primarily because a catastrophic leak or spill would result in extremely high concentrations. Non-point sources from urbanized watersheds were generally a medium priority. Most of the TRI and AST sites fall into the moderate protection priority category (category C). Nitrates: The high priority category was dominated by NPDES dischargers, primarily wastewater treatment plants. Most of the loading from these sites appeared to be relatively low, and was not likely to cause a cumulative impact that would cause an exceedance of the nitrate standard at the intake. Moderate priority sites were a mixture of NPDES sites, TRI sites, and non-point runoff from storm water. The potentially significant sources were located within both the 5-hour time of travel (zone A) and the 25-hour travel time (zone B). Therefore, efforts to reduce nitrate impacts will be necessary watershed wide. Petroleum Hydrocarbons: There were a limited number of significant sources of petroleum hydrocarbons. Only above ground storage tanks containing fuel, or stormwater runoff were identified as significant potential sources of petroleum hydrocarbon loading. Most of the high priority sites were either fuel storage facilities (with a low probability of release 18 but potentially very high concentrations), or stormwater runoff with lower concentrations but frequent occurrence. Phosphorus: Similar to nitrates, the high protection priority category was dominated by NPDES dischargers, primarily wastewater treatment plants. Most of the loading from these sites appeared to be relatively low, and was not likely to cause a cumulative impact that would cause significant water quality impairment at the intake. There are a few very large industrial sites that were also included in the high category, primarily due to the high potential concentrations should a spill occur. Moderate priority sites were mainly a mixture of TRI sites and non-point runoff from stormwater. Disinfection By-Product (Total Organic Carbon): Overall the high protection priority sites were NPDES discharges from wastewater treatment plants. In general, NPS sites appeared to have a lower total load and impact on water quality than do the NPDES sites. TRI and AST sites were all found in the low priority category. Turbidity (Total Suspended Solids): Turbidity was analyzed using total suspended solids (TSS) as a surrogate. Only stormwater runoff and NPDES discharges were identified as potentially significant sources of TSS. The stormwater runoff (NPS sites) tended to show much higher loading with less frequency. The NPDES sites had lower rates of TSS loading, however, they were more constant discharges. Loading rates from non-point sources appeared high enough to cause concern for cumulative impacts at the intake during storm events. VOCs: In this case, the only significant potential sources of VOCs were storage tanks (ASTs), industrial sites from the TRI database, or wastewater treatment plants. The high protection priority category was a mixture of AST, TRI, and NPDES sites. The moderately high and moderate protection priority categories were primarily AST and RCRA sites. The NPDES sites appeared to load VOCs at a low rate, and are not likely to cause water quality impairment at the intake. The AST, and TRI sites would require a spill to cause water quality impairment, but resulting concentrations would be very high. CONCLUSIONS Performing a SWA on a large river system can be a daunting undertaking. This case study shows, however, that it is not only possible to do a thorough and effective SWA, but that the size of the watershed does not necessarily mean that the assessment needs to oversimplify. In fact, the Delaware River SWA would not have provided any useful results without addressing the inherent complexities. The primary technical difficulties that were overcome included: Developing a database and collecting enough data throughout the watershed to identify potential sites Programming vital links to a GIS system to calculate the time of travel from each potential source to each intake 19 Assessing pollutants carried by stormwater by developing a stormwater model using SWMM Effectively dealing with the issue of intakes influenced by tidal oscillations within the estuary using a tidal zone water quality model Using a decision software package EVAMIX to successive screen point sources, and to simultaneously evaluate point and non-point sources in a single evaluation. Including stakeholder participation in the process of evaluation to solidify cooperation of upstream communities Identification and prioritization of significant sources of pollutants is valuable, not only to the SWA process and the water suppliers in the basins, but also for use in the wide-scale, generalized watershed management planning efforts of the Philadelphia Water Department’s Office of Watersheds. Experience to-date reinforces that the use of a comprehensive, complex watershed modeling and decision support systems lend credibility to the evaluation process and is considered a key success factor for the SWA program. REFERENCES Blumberg, A.F., 1977. "Numerical Model of Estuarine Circulation," ASCE J. Hydr. Div., 103:295310. Blumberg, A.F. and G.L. Mellor, "A Coastal Ocean Numerical Model," In: Mathematical Modelling of Estuarine Physics, Proceedings of an International Symposium, Hamburg, August 2426, 1978. J. Sundermann and K.P. Holz, Eds., Springer-Verlag, Berlin, 1980. HydroQual, 1998. Development of a Hydrodynamic and Water Quality Model for the Delaware River. Prepared for Delaware River Basin Commission, May 29, 1998. 20