LINKING REMOTE SENSING AND GIS TO SOIL CONSERVATION: GENERATING AGNPS INPUT USING REMOTE SENSING AND GIS

Remote sensing techniques and geographic information systems offer a good means of collecting and manipulation of the data required to assess conservation practices. We have developed the methodology to discriminate land cover including the amount of residue cover from Landsat TM image classification. We have developed the methodology for the automatic generation of most of the data required by the AGNPS erosion model from three files: 1) digital file with contour lines from the USGS Topographic Maps, 2) digital soil mapping units from the Soil Surveys, and 3) land cover from Landsat TM image classification.

Table of Contents

• Introduction
• Objective:
• Methodology:
  • Erosion model:
  • Generation of the data:
  • Landsat image classification:
• Summary:
• References:

INTRODUCTION:

According to the Food Security Act of 1985, farmers have to fulfill certain conservation requirements between 1990 and 1995 to qualify for subsidies and federal loans. Remote sensing techniques offer a good means of monitoring the adoption of these conservation practices. These techniques can also be a cost efficient way to estimate singleevent watershed erosion rates. Remote sensing has been used for many years to evaluate different agricultural features including amount of crop residue and conservation practices (DeGloria et al., 1986; Motsch, 1987; Schaal, 1986; Whiting et al., 1987), cropping systems (Morgan and Nalepa, 1982; Stephens et al., 1985), soil variation (AgRISTARS, 1984; Walch, 1985), soil erosion (Logan et al., 1982; Trolier and Philipson, 1986; Welch et al., 1984), and hydrological systems (Mintzer and Askari, 1980).

Development of digital image processing and geographic information systems (GIS) has increased the potential applications of remote sensing. These two technologies can enable us to analyze large or small areas, integrate numerous variables into the evaluation process, and easily update database information (Walch, 1985). These improvements result in a less timeconsuming and less expensive methodology to monitor soil conservation practices and predict erosion rates.

An advantage of remote sensing techniques such as Landsat image classification is that large areas can be evaluated to determine type of crop, crop canopy density, and amount of residue cover. One of the limitations of any erosion model is that the collection and input data is traditionally done manually. This manual acquisition and input of data it is time-consuming, and costly.

OBJECTIVE:

- to develop the technology to generate automatically most of the data required by an erosion model using readily available data or easy to collect data and the ERDAS image processing system or aerial photography.

METHODOLOGY:

Erosion model:

The Agricultural NonPoint-Source pollution model (AGNPS, version 3.65) was developed by the USDA Agricultural Research Service. The model was developed to analyze and provide estimates of runoff water quality from agricultural watersheds using an IBM compatible personal computer. It operates on a cell basis, where cells are uniformly square areas subdividing the watershed, allowing analyses of any point within the watershed. For big watersheds a large cell size will have to be used resulting in a decrease in the accuracy of the model.

Because of the memory limitation on an IBM personal computer, the maximum number of cells that AGNPS can handle is 1900. If a cell size of 10 acres is used the largest watershed that AGNPS can evaluate is 19000 acres. If 2.5 acres cell size is used the largest watershed will be 475 acres.

The problem that we have encountered now, is that even when the data can be generated easier and faster, there still is the limitation of maximum number of cells that AGNPS can process. The AGNPS model was ported into the Ohio SuperComputer to increase the maximum number of cell. Using the SuperComputer, also avoids the long run time and storage area limitations encountered when using an IBM personal computer. This computer will give us the capacity to evaluate large areas in a more efficient way.

Generation of the data:

We have developed the algorithms to automatically estimate most of the data required by the AGNPS erosion model (Table 1). These data were generated from three main sources: 1) digital file with contour lines from the USGS Topographic Maps, 2) digital soil mapping units from Soil Surveys, and 3) land cover from Landsat TM image classification. Figure 1 shows the flow of data from these three sources and from which of the three the GIS files are generated. In addition to the ground cover estimates obtained from the ERDAS classification of the Landsat image or photo interpretation, ERDAS routines were used to build GIS files for slope and aspect from USGS topographic maps, and for soil mapping units from soil surveys. The Ohio Capability Analysis Program (OCAP) of the Ohio Department of Natural Resources has more than 50% of Ohio's soils digitized. Additional advantages of using GIS is that the data can be easily modified and also can be obtained and converted from other GIS format such as GRASS (US. Department of Defense and currently used by the USDA SCS) and ARC/INFO (ESRI).

Table 1. Input required by the AGNPS erosion Model

 * cell number                     * conservation factor           
 * receiving cell                  * surface condition             
 * SCS curve number                * aspect                        
 * slope                           * soil texture                  
 * slope shape                     - fertilization level           
 * slope length                    - fertilizer availability       
 * channel slope                   - point source                  
 * channel side slope              - gully source indicator        
 * Manning's Roughness Coeff       * COD                           
 * erodibility factor              * impoundment factor            
 * cropping factor                 * channel indicator             

* data that we can automatically generate
- data that we can estimate with the help of the field personnel

Figure 2 shows the data flow of how the data are manipulated from GIS format to the final GIS files with the outputs from AGNPS. Once the required GIS files are generated, a computer program converts from the GIS files into the tabular file format required by AGNPS. Using AGNPS, the data is checked for errors and the amount and energy intensity for the storm event is entered. AGNPS is run producing a tabular output with soil loss, hydrology, and nutrient pollution (table 2).

Table 2. GIS Files that can be Generated from AGNPS

Hydrology/Soil Loss               Nutrient Pollution 
 - drainage area                    - For N and P
 - runoff area                          . in sediment within cell
 - peak flow                            . in sediment at cell outlet
 - cell erosion                         . water soluble within cell
 - sediment generated*                  . water soluble at cell outlet
     . cell erosion                 - COD
     . above cell                       . water soluble within cell
     . within cell                      . at cell outlet
     . yield                            . concentration
 
*sediment analysis available for:
     . clay      . small aggregates
     . silt      . large aggregates
     . sand      . total

The AGNPS erosion model has graphic capabilities that allows the graphical display of the watershed. This display has some limitations. The output generated by the model can be printed on a plotter but this printout is not to scale, which makes it difficult to find areas or fields in the watershed. Each cell can be selected individually to view the input data and also some of the output data.

We developed a computer program that converts either all or selected AGNPS output data into GIS files. These GIS files can be used for further analyses, converted into other GIS formats. They can be also printed to scale to fit maps such as the USGS Topographic Maps. With the ERDAS system we can overlay the output GIS files over a Landsat image of the area and on top of this the digital line graphs with roads, railroads, and streams. This type of overlay, facilitate the location in the field of problematic areas.

Landsat image classification:

We have developed the methodology to discriminate surface cover classes of 0-15, 15-45, 45-75 and > 75% using Landsat TM imagery with accuracy approaching 60% if we restrict ourselves to these specific ranges (Motsch, 1987; Olivieri et al. 1991). If we allow some overlap we can obtain higher accuracy.

SUMMARY:

We have developed the methodology to automatically generate most of the data required by the AGNPS erosion model from three sources of data using the ERDAS image processing system and some programs that we have developed:

1. file with contour lines from the USGS Topographic Maps,
2. soil mapping units from Soil Surveys, and
3. land cover from Landsat TM image classification or aerial photo interpretation

This technology provides a faster and more cost effective way to generate the data than traditional methods. It also avoid the human errors associated with manual methods.

REFERENCES:

• AgRISTARS Program Management Group. 1984. Agricultural and Resources Inventory Surveys Through Aerospace Remote Sensing., Research Report Fiscal Year 1983. Washington, D.C.
• DeGloria, S.D., S.L. Wall, A.S. Benson and M.L. Whiting. 1986. Monitoring conservation tillage practices using Landsat multispectral data., J. Soil & Water Conserv. 41(3):187 190.
• Lillesand, T.M. and R.W. Kiefer. 1987. Remote Sensing and image interpretation. John Wiley & Sons, New York.
• Logan, T.J., D.R. Urban, J.R. Adams and S.M. Yaksich. 1982. Erosion control potential with conservation tillage in the Lake Erie Basin: Estimates using the universal soil loss equation and the Land Resource Information System (LRIS). J. Soil & Water Conserv. 37(1):5055.
• Mintzer, O. and F. Askari. 1980. A remote sensing technique for estimating watershed runoff., Water Resources Center., Eng. Exp. Station, The Ohio State University, Columbus, OH.
• Morgan, K.M. and R. Nalepa. 1982. Application of aerial photographic and computer analysis to the USLE for areawide erosion studies. J. Soil & Water Conserv. 37(6):347350.
• Motch, B. 1987. Comparison of thematic mapper for crop residue maps and field information in Seneca County, Ohio. Ohio Department of Natural Resources, Columbus, OH.
• Olivieri L. J., G. M. Schaal, T. J. Logan, W. J. Elliot, and B. Motsch. 1991. Generating AGNPS input using remote sensing and GIS. Paper No. 91-2622, Am. Soc. Agr. Eng., St. Joseph, MI.
• Schaal, G.M. 1986. Analysis of Thematic Mapper classification of tillage practices in Seneca County, Ohio. Ohio Department of Natural Resources, Columbus, OH.
• Stephens, P.R., J.K. MacMillan, J.L. Daigle and J. Cihlar. 1985. Estimating universal soil loss equation factor values with aerial photography. J. Soil & Water Cons. 40(3):293296.
• Trolier, L.J. and W.R. Philipson. 1986. Visual analysis of Landsat Thematic Mapper images for hydrologic land use cover. Phot. Eng. and Remote Sensing. 52(9):15311538).
• Walch, S.J. 1985. Geographic information system for natural resources management. J. of Soil and Water Conserv. 42(2):202205.
• Welch, R., T.R. Jordan and A.W. Thomas. 1984. A photogrammetric technique for measuring soil erosion. J. of Soil and Water Conserv. 39(3):191194.
• Whiting, M.L., S.D. DeGloria, A.S. Benson and L. Wall. 1987. Estimating conservation tillage residue using aerial photography. J. Soil & Water Conserv. 42(2):130132.
• Young R.A. and C.A. Onstad. 1987. AGNPS, Agricultural Non-Point-Source Pollution Model; A watershed Analysis Tool. USDA. Report 35. USDA, Washington, DC.