Zeyuan Qiu, Tony Prato and Chris Fulcher

The Impacts of Data Resolution on Simulation of Water Quality Using the Agricultural Nonpoint Source Pollution Model and a Geographic Information System

The Agricultural Nonpoint Source Pollution Model (AGNPS) is a most widely-used biophysical models used to simulate water quality impacts of agricultural practices in a watershed. The modeling efficiency can be significantly improved by incorporating a Geographic Information System (GIS), which makes it feasible to use high-resolution data sets. As the application objectives change, the data resolution for generating AGNPS input parameters may change as well. Selection of data resolution can be an important issue for AGNPS model users in light of objectives.

This application is conducted in Goodwater Creek watershed, Missouri. WAMADSS is used to automatically generate a number of input parameters from land use, soil, hydrology and hypsography layers, run the AGNPS model, and present water quality simulation results. Two data resolutions are used: 200 by 200, 100 by 100 square meters grid cells. The simulated water quality results with the different data resolutions are compared within the watershed and at the watershed outlet. The purpose of this paper is to show the differences in water quality simulation caused by data resolution and their implications for AGNPS users.

Introduction

The U.S. Environmental Protection Agency (EPA) has identified agriculture as the leading source of nonpoint source pollution in the U.S. (USEPA, 1994). Agricultural nonpoint source pollution is diffuse in nature, has complicated spatial and temporal dimensions, and is driven by the vagaries of weather. Its characteristics make it very difficult and expensive to monitor on a continuous and widespread basis utilizing field treatments and experiments. Biophysical simulation models have been developed as an alternative to estimate the impacts of agriculture on water quality.

The Agricultural nonpoint source pollution model (AGNPS) (Young et al., 1987) is a widely-used biophysical simulation model used to estimate water quality impacts of agricultural practices in a watershed. AGNPS is a spatially distributed, single-event watershed simulation model that uses landuse, soil, hydrology and hypsography data and subdivides complex watersheds into grid cells. It is intended to provide basic information on water quality to be used to classify nonpoint source pollution problems in agricultural watersheds. The model provides outputs on runoff, upland erosion, channel erosion, and sediment yield, and nitrogen, phosphorus, chemical oxygen demand (COD) in runoff and sediment (Young et al., 1987). The new version of AGNPS estimates pesticide movement in soil and runoff. Development of an annualized version of AGNPS is in progress.

A geographical information system (GIS) is a system of hardware, software and procedures designed to support the capture, management, manipulation, analysis, modeling and display of spatially-referenced data for solving complex planning and management problems (Burrough, 1986). Since AGNPS uses a distributed parameter approach to quantify a watershed by dividing the area into small grid units within the watershed, it is appropriate to use a GIS to storage and process those spatial characteristics (Mitchell et al., 1993). Several efforts have been made to integrate GIS with AGNPS to evaluate agricultural nonpoint source pollution. Tim and Jolly (1994) developed an integrated GIS and hydrologic/water quality model using ArcInfo GIS and AGNPS to evaluate agricultural nonpoint source pollution. He et al. (1993) used an integrated AGNPS and GRASS GIS package, GRASS WATERWORKS, to evaluate the impact of agricultural runoff on water quality in the Case River, a subwatershed of Saginaw Bay. Mitchell et al. (1993) developed an integrated AGNPS/GIS (GRASS) system to validate the AGNPS for predicting runoff and sediment delivery from small watersheds of mild topography. Srinivasan et al. (1995) developed a spatial decision support system using AGNPS and GRASS GIS to assess agricultural nonpoint source pollution.

In this paper, data resolution refers the size of grid cell used to delineate a watershed. Data resolution is a key factor in the AGNPS modeling process since the grid cells serve as a "cookie-cutter" for other spatial input data including soil, landuse, hydrology and hypsography. The smaller the grid cell, the more number of grid cells for a given watershed. AGNPS only permitted a watershed to be divided up to 1,900 cells in version 3.65 (2,800 cells in version 5.0) due to the memory constraints on the personal computer. The cell size limitation is overcome by compiling AGNPS on a UNIX workstation where the GIS resides (Fulcher, 1996). Using digital input data and integrating AGNPS with GIS makes it possible to readily delineate a watershed into a grid cells size dictated by the user. The smaller the grid cells, the higher the data resolution and accuracy. However, there is a higher cost in term of processing speed and storage requirements. As the application objectives change, the data resolution for generating AGNPS input parameters may change as well. Selection of data resolution can be an important issue for AGNPS model users in light of objectives. The purpose of this paper is to show the differences in water quality simulation caused by data resolution and their implications for AGNPS users.

Methodology

The Center for Agricultural, Resource and Environmental Systems (CARES) at University of Missouri-Columbia developed a watershed management decision support system (WAMADSS) by integrating ArcInfo GIS with AGNPS, the Soil and Water Assessment Tool (SWAT) (Arnold et al., 1994) and the Cost and Return Estimator (CARE) (USDA SCS, 1988) model on the same computing platform (an IBM RISC 6000 UNIX workstation) to aid in development of water quality management plans. WAMADSS is an interactive decision support system. Graphic user interfaces or a series of menues are created using ArcInfo Arc Macro Language (AML), FormEdit and C language to link all its components and streamline processes of environmental modeling and economic analysis. In this application, WAMADSS is used to divide the watershed into small grid cells, generate an input file for AGNPS, execute the model and view results. Data process functions of WAMADSS involved in this application are summarized as follows. The details about the interactive menus, processes, parameters, files and their format can be found in Fulcher (1996).

Spatial Data Process

WAMADSS cut the watershed into small grid cells based on certain data resolution using the watershed boundary layer. The central points of all grid cells were extracted as a point coverage in ArcInfo to represent the watershed. The point coverage were linked to other spatial data through the spatial location of those points, i.e. the coordinates of those points and other relational INFO files through their identification items. The digital elevation model (DEM) was created from the digital contour coverage using the same resolution. The slope and aspect coverage were generated from DEM and were then linked to each central point. Other spatial layers including soils, landuse and hydrology were first rasterized using the same data resolution to identify the soil type, landuse categories and stream type in each cell. The rasterized grids were then vectorized so that soil, landuse and stream information can be linked to each central point. The resulting INFO file of the central point coverage contained all spatial information, such as the cell number, landuse category, soil type, stream type, aspect, land slope, and so on. The INFO file provided the basic information to process the attribute data outlined as follows.

Attribute Data Process

INFO stores the input parameters in a header file and nine relational files. AML and C programs are used to generate an ASCII AGNPS input file with a specific format by extracting the parameters from these INFO files. There are a total of 168 variables or items in these files: 149 variables are AGNPS input parameters (13 watershed-level parameters, 22 cell parameters and 114 optional parameters ) for each cell; 10 variables are reserved for model development testing; and the remaining nine variables represent a common key item (cell number) in each of the nine files. After executing AGNPS in WAMADSS, it generates an ASCII output file. In order to view the output in ARCPLOT, AML and C programs are used to parse the output file into six ASCII files and use these files to populate their respective INFO relational files. There are a total of 146 variables in these files: 144 AGNPS output parameters and 2 relational key item (cell number) for the pesticide and feedlot files.

Application and Results

This application is conducted in Goodwater Creek watershed, which is located in north-central Missouri near Centralia. This watershed was selected as a study site for Missouri Management Systems Evaluation Area (MMSEA) project in 1990. It typically represents the Central Claypan Soils Major Land Resource Area (MIRA 113), an area of about 10 million acres in the Midwest. The topography is characterized by broad nearly flat divides, gentle sideslopes, and broad alluvial valleys often dissected with small streams (MMSEA Team, 1995). The goal of MMSEA is to improve surface and ground water quality in Missouri and throughout the Midwest by improving the understanding of the complexities of claypan soil/glacial drift hydrology; evaluating the impact of six alternative farming systems on ground and surface water quality; increasing knowledge regarding development of improved farming systems; developing new farming technologies, assessing the socioeconomic impacts of farming systems; and implementing far-reaching educational programs (MMSEA Team, 1995).

AGNPS was used to estimate the water quality impacts in Goodwater Creek watershed during the corn season of farming system 1 defined in MMSEA project. Farming system 1 represented the prevailing farming system in the study area, and was a corn-soybean rotation with conventional tillage and high pesticide and fertilizer applications. The storm event used for running AGNPS was 2.4 inch rainfall in 24 hours. This application also assumed there were no point pollution sources (feedlot and non-feedlot), no gully erosion and no impoundments in the watershed. WAMADSS was used to automatically generate a number of input parameters from land use, soil, hydrology and hypsography layers, run the AGNPS model, and present water quality simulation results. Two data resolutions were used: 200 by 200 and 100 by 100 square meter grid cells. The simulated water quality results with the different data resolutions were compared within the watershed and at the watershed outlet.

Table 1 compared the water quality impacts at the watershed outlet simulated by AGNPS with different data resolutions. It seemed that data resolution did not cause significant differences in runoff, soil loss and pesticide pollution. As shown in Table 1, runoff with 200 meter resolution is 1.16 inches, which is slightly higher than 1.15 inches with 100 meter resolution, while total sediment yield with 200 meter resolution is 2,951 tons which is slightly lower than 2,982 tons with 100 meter resolution. However, there are significant differences in the estimation of nutrient pollution. The nutrient pollution indicators refer to total soluble nitrogen, phosphorus, and chemical oxygen demand (COD), and their concentration levels in runoff. The nutrient pollution estimated with 200 meter resolution data is generally 20 percent higher than with 100 meter resolution.

Also, the sediment movement is more clearly identified with more accurate input data. As shown in Table 1, with the two different resolution input data, even though the sediment yields at the watershed outlet are almost same, the sediment generated within cell and the sediment deposition rate within the cell are different. The sediment generated within cell and the sediment deposition rate within the cell are 2.35 tons per acre and 71.01 percent with 100 meter resolution, which are higher than 1.60 tons per acre and 62.63 percent with 200 meter resolution, respectively. With high resolution input data, AGNPS identified more sediment generation and deposition activities in the watershed.

Table 1. Comparison of AGNPS Outputs between the Two Data Resolution

Water Quality Indicators        Units      100 Meter     200 Meter    Percent
at Watershed Outlet                        Resolution    Resolution   Change

Runoff                          inches           1.15        1.16       +0.87
Total Sediment Yield            tons          2981.97     2951.13       -1.03
Average Sediment Yield          tons/acre        0.16        0.15       -6.25
Sediment within Cell            tons/acre        2.35        1.60      -31.91
Sediment Deposition in Cell     percent         71.01       62.63      -11.80
Sediment Attached Nitrogen      lbs/acre         0.58        0.54       -6.90
Sediment attached Phosphorus    lbs/acre         0.29        0.27       -6.90
Total Soluble Nitrogen          lbs/acre         1.27        1.51      +18.90
Soluble N. Concentration        ppm              4.60        5.48      +19.13
Total Soluble Phosphorus        lbs/acre         0.54        0.66      +22.22
Soluble P. Concentration        ppm              1.98        2.41      +21.72
Total Soluble COD               lbs/acre         7.18        8.70      +21.17
Soluble COD Concentration       ppm             26.56       31.97      +20.37
Average Soluble Atrazine        lbs/acre         0.05        0.05        0.00
Soluble Atrazine Concentration  ppm              0.18        0.19       +5.56
Average Atrazine Leached        lbs/acre         0.08        0.08        0.00

Summary and Conclusions

WAMADSS is an interactive watershed management decision support system. It integrates ArcInfo GIS with AGNPS, SWAT and CARE model on an IBM RISC 6000 UNIX workstation to conduct water quality modeling and economic analysis at the watershed scale, and to aid in development of water quality management plans. In this application, the water quality modeling component of WAMADSS was used to delineate the watershed, generate an AGNPS input file, execute the simulation model, and display the simulation results. It is shown that AGNPS modeling efficiency can be significantly improved by using the digital spatial data and integrating with a GIS. Two aspects were emphasized by Fulcher (1996): (i) WAMADSS significantly reduces the time and labor needed to process and manipulate the required input parameters; (ii) human error is avoided in interpreting landscape characteristics such as land slope, slope shape factor, field slope length and channel slope, and inconsistency in discerning these characteristics across the entire watershed by developing a DEM through GIS.

WAMADSS was used to simulate the water quality impacts of agriculture in Goodwater Creek watershed, Missouri. The AGNPS outputs with two different data resolution (100 by 100 meter and 200 by 200 meter) were reported. The results show that data resolution did not cause significant difference in estimating the soil loss and pesticide pollution. However, The nutrient pollution estimated with 200 meter resolution data is generally 20 percent higher than with 100 meter resolution. Also, the sediment movement is more clearly identified with more accurate input data. These results are important for AGNPS users to select appropriate data resolution in light of their objectives.

References

Arnold, J.G., J.R. Williams, R. Srinivasan, K.W. King and R.H. Griggs. 1994. SWAT: Soil and Water Assessment Tool, USDA ARS, Temple, Texas.

Burrough, P.A. 1986. Principles of Geographical Information Systems for Land Resources Assessment. Oxford University Press Inc., New York. Fulcher, C.L. 1996. A Watershed Management Decision Support System (WAMADSS) - Economic and Environmental Impacts of Land Use Activities for Reducing Nonpoint Source Pollution. Unpublished Ph.D. Dissertation. University of Missouri, Columbia, Missouri.

He, C., J.E. Riggs and Y.T. Kang. 1993. Integration of Geographic Information Systems and A Computer Model to Evaluate Impacts of Agricultural Runoff on Water Quality. Water Resources Bulletin 29(6): 891-900.

Missouri MSEA Team. 1995. The Missouri Management Systems Evaluation Area Research and Education Report: 1990-1995.

Mitchell, J.K., B.A. Engel, R. Srinvasan and S.S.Y. Wang. 1993. Validation of AGNPS for Small Watersheds Using an Integrated AGNPS/GIS Systems. Water Resources Bulletin 29(5): 833-842.

Srinivasan, R., B.A. Engel. 1995. A Spatial Decision Support System for Assessing Agricultural Nonpoint Source pollution. Water Resources Bulletin 30 (3): 441-452.

Tim, U.S. and R. Jolly. 1994. Evaluating Agricultural Nonpoint Source Pollution Using Integrated Geographic Information Systems and Hydrologic/Water Quality Model. Journal of Environmental Quality 23: 25-35.

U.S. Department of Agriculture, Soil Conservation Service. 1994. Cost and Returns Estimator: Users Guide.

U.S. Environmental Protection Agency, Office of Water. 1994. National Water Quality Inventory: 1992 Report to Congress, Washington DC, EPA 841-R-94-001.

Young, R.A., C.A. Onstad, D.D. Bosch, and W.P. Anderson. 1987. AGNPS (Agricultural Non-point source Pollution Model): A Watershed Analysis Tool, Conservation Research Report, USDA ARS, Washington, DC.

Zeyuan Qiu
Post-Doctoral Fellow, CARES
Center for Agricultural, Resource and Environmental Systems
University of Missouri-Columbia
200 Mumford Hall
Columbia, MO 65211
Telephone: (573)884-8936
Fax: (573)882-3958

Tony Prato
Professor of Resource Economics and Management, and Director, CARES
Center for Agricultural, Resource and Environmental Systems
University of Missouri-Columbia
200 Mumford Hall
Columbia, MO 65211
Telephone: (573)882-0147
Fax: (573)882-3958

Chris Fulcher
Research Associate and Associate Director, CARES
Center for Agricultural, Resource and Environmental Systems
University of Missouri-Columbia
200 Mumford Hall
Columbia, MO 65211
Telephone: (573)882-6534
Fax: (573)882-3958