Soil-Landscape Estimation and Evaluation Program (SLEEP) to predict spatial distribution of soil attributes for environmental modeling

Feras M. Ziadat, Yeganantham Dhanesh, David Shoemate, Raghavan Srinivasan, Balaji Narasimhan, Jaclyn Tech. (17/3/2015). Soil-Landscape Estimation and Evaluation Program (SLEEP) to predict spatial distribution of soil attributes for environmental modeling. International Journal of Agricultural and Biological Engineering, 8 (3), pp. 158-172.

Abstract

The spatial distribution of surface and subsurface soil attributes is an important input to environmental modeling. Soil attributes represent an important input to the Soil and Water Assessment Tool (SWAT), which influence the accuracy of the modeling outputs. An ArcGIS-based tool was developed to predict soil attributes and provide inputs to SWAT. The essential inputs are digital elevation model and field observations. Legacy soil data/maps can be used to derive observations when recent field surveys are not available. Additional layers, such as satellite images and auxiliary data, improve the prediction accuracy. The model contains a series of steps (menus) to facilitate iterative analysis. The steps are summarized in deriving many terrain attributes to characterize each pixel based on local attributes as well as the characteristics of the contributing area. The model then subdivides the entire watershed into smaller facets (subdivisions of subwatersheds) and classifies these into groups. A linear regression model to predict soil attributes from terrain attributes and auxiliary data are established for each class and implemented to predict soil attributes for each pixel within the class and then merged for the entire watershed or study area. SLEEP (Soil–Landscape Estimation and Evaluation Program) utilizes Pedo-transfer functions to provide the spatial distribution of the necessary unmapped soil data needed for SWAT prediction. An application of the tool demonstrated acceptable accuracy and better spatial distribution of soil attributes compared with two spatial interpolation techniques. The analysis indicated low sensitivity of SWAT prediction to the number of field observations when SLEEP is used to provide the soil layer. This demonstrates the potential of SLEEP to support SWAT modeling where soil data is scarce.