Demonstration of ArcGIS as a Tool to Predict Pollutant Loading in Stormwater Discharges Jesse S. Dickson CE 547: GIS in Water Resources Dr. J. Coonrod May 6, 2011 Introduction In order to address the more restrictive goals of the EPA’s NPDES Phase 2 discharge permits, municipalities need a cheap and effective means to characterize pollutant loads generated within their watersheds. The purpose of this paper is to demonstrate how ArcGIS can be used to predict pollutant concentrations as a potential planning tool for municipalities and flood control authorities. See Appendix C – Background for additional information pertaining to NPDES regulations. Methods The methods used for determination of an ultimate pollutant load originating from non-point sources in a watershed are fairly straight-forward. While the algorithm for modeling transport of pollutants can be somewhat involved, as demonstrated by the transport model described below, the algorithm itself is made up of a series of simple calculations comprised mostly of vector algebra, conversion of polygons to rasters, and joining tables to feature classes. Data Sources Data sources for the project, although compiled in a zip file, are all either available from via the internet from data clearinghouses such as RGIS, USGS, and PRISM, or capable of being created as a fairly simple .dbf file. For instance, the EMC field (described in Appendix A, III.B below) for pollutant concentrations based on land use, can be readily approximated and compiled in tabular (i.e., .dbf) form from information available in various locations on the internet, such as the Spreadsheet Tool for Estimated Pollutant Load (STEPL), located at: http://it.tetratech-ffx.com/steplweb/models$docs.htm. Additionally, a more accurate runoff model could be generated for areas with detailed topographic information (as opposed to the regression equations used in this project for generation of runoff data). Page 1 Projection The projection used for this project is the ___________. There is no overriding reason for using this projection, other than it accurately displays and represents the large land area (approximately _______ square miles) analyzed in this project. If the project had been a smaller land area, particularly one with a predominantly north-south oriented floodplain, a UTM projection with state plane coordinates would have been a more appropriate. Software Environment Most of the procedures performed in the course of this project were executed in the ArcView environment, and drawing heavily from the Spatial Analyst features found in the ArcToolbox sidebar. Although the compiled data provided both the .dbf file and model toolbar used in the pollutant raster creation and transport model, these features would need to be created using ArcCatalog and Model Builder, respectively. It is noted that scripting (presumably Python) was used to create the .dll files needed to link data from the routed runoff portion of the model to the routed (i.e., degradation of pollutant concentrations based on various decay parameters) pollutant concentration portion of the model. Although the Model Builder feature is not strictly required to perform the calculations necessary the transport of contaminants from each watershed to their ultimate point of discharge, by automating the algorithm in Model Builder, ArcGIS facilitates the ease of replicating calculations with a minimum of effort. This allows for quickly calculating results for multiple scenarios, including a changes in pollutant concentration or maintaining a database that tracks historical pollutant loading. Results This project resulted in a combination of . Although the process of creating the various rasters and raster manipulations necessary to create pollutant load concentrations Unfortunately, however, the failure of the model to run properly did not provide the Page 2 Conclusions ArcGIS provides a powerful tool in assessing pollutant loads in stormwater discharges a wide geographical area. By characterizing pollutant discharges based on land use, ArcGIS can be used to model pollutant concentrations to specific locations in a watershed. Municipalities and flood control authorities can use this information as a planning tool to prioritize potential treatment facilities by determining where such facilities should be located to achieve maximum benefit when treating stormwater for pollutants. Additionally, after construction of treatment facilities ArcGIS can be used to track pollutant removal and demonstrate that measureable goals in stormwater treatment have been met. Future Research Future research that would be helpful to this end would include locally applicable studies of pollutant loading based on land use. For smaller – but more complex – watersheds such as the Albuquerque area, a detailed stormwater model that includes local stormwater facilities (e.g., conveyance structures, detention facilities, etc.) would provide a much more accurate representation of discharges that carry pollutants to the Rio Grande. Such a water model would provide much more accurate data for pollutant concentrations. Finally, the failure of the .dll files to properly link the various inputs in the Transport Model illustrate the need to have a understanding of the scripting language used to create an intricate model. As with any modeling software, it is imperative for the user to have a firm grasp of the algorithm at the core of the model. This serves two functions: first, it allows the user to adjust the framework of the model when a processing error occurs. Second, and perhaps most importantly, it allows the user to ascertain—particularly in the absence of a processing error—when the algorithm has run correctly. In the case of this project, if the transport model had run to completion without Page 3 generating an error, but used a .dll file incorrectly (e.g., linked pollutant decay coefficients incorrectly), my unfamiliarity with the script used to generate the .dll file would likely cause the error to go unnoticed. While this is a concern that is important to any software that utilizes programming scripts, it is particularly noteworthy in ArcGIS since so many open source Python scripts are available (like the one used in the project) via the internet. . Page 4 Appendix A – Procedure The following procedures were applied to obtain the reported results: I. Acquire Data A. Data used in the presented results were compiled and stored in a zip file at the primary source for the research of the project. This file also contains the transport algorithm used to model transport of pollutants and is located at: http://www.ce.utexas.edu/prof/maidment/giswr2005/ex6/BacterialModel.zip B. Watershed Data: USGS Hydrologic Unit Code mapping: http://water.usgs.gov/GIS/huc.html C. Rainfall Data: PRISM Climate Group http://www.prism.oregonstate.edu/ D. Land Use Data: NRCS Data http://www.nrcs.usda.gov/technical/NRI/maps/aboutmaps/coverages.html E. Pollutant Data Based on LULC – “Estimation of Fecal Coliform Loadings to Galveston Bay” (Zoun, 2003) F. After extracting the zip file, install the DLL files required for the transport model using “Install.BAT” II. Calculate Runoff A. Runoff is calculated using derived empirical equations particular to the area. Other regression equations for all states can be found at: http://water.usgs.gov/software/NFF/ B. Convert PRISM rainfall data to raster using the FEATURE TO RASTER command in ArcToolbox C. Use RASTER CALCULATOR to create precipitation raster for each LULC raster. D. Use RASTER CALCULATOR and regression equations to create runoff raster from each precipitation raster (generated in step C) E. Use MOSAIC TO NEW RASTER to compile runoff rasters into one raster. F. Use ZONAL STATISTICS and watershed feature as an input and “Junction ID” as zone field to create runoff per basin. III. Create Pollutant Concentration Raster A. Using the previously created .dbf file, join the Concentration table to LULC polygon B. Create pollutant raster based on EMC field. IV. Create Non-Point Pollutant Load Raster A. Use RASTER CALCULATOR to multiply the runoff raster by the pollutant concentration raster to get a raster for the mass loading of pollutant. B. Note that results of the initially created raster contained negative numbers for concentration when applied to step C. It was determined that the values were larger than allowed, even using “float” domain. Therefore, the values were divided by 106 using RASTER CALCULATOR to get smaller concentration values. C. Use ZONAL STATISTICS AS TABLE with watershed feature to generate pollutant load per year for each basin. V. Model Transport of Pollutant Load A. Populate the “IncVal” field of the previously created chemaNode feature by JOIN with table created in IV.C. and watershed. B. Add the previously created transport model to the ArcToolbox. C. Run the Model. Page A D. Note that the model runs and appears to populate the SchemaLink feature class, but fails to carry generated data to the SchemaNode feature class. E. Appendix B – Datasets (attached digitally) Appendix C – Background The EPA’s Clean Water Act (CWA) applies not only to wastewater, but as of amendments enacted in 1987, also to stormwater runoff. The CWA created the National Pollutant Discharge Elimination System (NPDES), which grants permits to discharge into waters of the US. Phase 1 of the 1987 amendments to the CWA regulated stormwater discharged from industrial sites and from “Municipal Separate Storm Sewer Systems” (MS4) for cities with populations greater than 100,000 people. The first NPDES stormwater permits were issued in the early 1990s and for the most part, the only requirements of the MS4 permits were to conduct studies and monitor stormwater quality. In Phase 2 of the NPDES stormwater regulations, permits were to become much more restrictive, identifying specific and quantifiable treatment goals. Discharge permits in Phase 2 of the NPDES are specific to the MS4 requesting permission to discharge, and can apply to several MS4 whose stormwater originates in a single watershed. For instance, the permit to which Albuquerque must adhere will likely be a regional permit, which will apply to several MS4 in the Middle Rio Grande Valley, including Corrales, Bernalillo, and Sandia Pueblo. Since Phase 2 permits are watershed-specific, the treatment goals of each permit will depend on the contaminants of concern to a particular to a given watershed. Although many Phase 2 permits have not yet been issued, municipalities with MS4 have an idea of the treatment goals that will be required, based on analysis of stormwater that was completed in Phase 1. As stated above, the character of stormwater is unique to the watershed that generated it, but can be generalized based on land use. Table 1 below shows average contaminant concentrations for stormwaters generated by several MS4 across the US. Page B Table 1: Average Contaminant Concentrations Page C