Landscape Context Influences Accuracy of Predicted Georeferenced Soil Data: Implications for Research

Lazaro-Lobo, A., & Ervin, G. N. (2024). Landscape Context Influences Accuracy of Predicted Georeferenced Soil Data: Implications for Research. GRI Report. Mississippi State University: Geosystems Research Institute.

Background and Objective: Large soil databases built with prediction methods can impose multiple limitations associated with prediction uncertainty. The objective of this study was to evaluate the accuracy of predicting soil properties with a polygon-based prediction approach and to assess the influence of the edaphic and topographic landscape context on that prediction accuracy. Materials and Methods: Ground-verified soil data (sand, silt, and clay, organic matter, and pH) were collected from 443 sample sites throughout Mississippi, and GIS predicted soil data were downloaded from the SSURGO database. The influence of the surrounding landscape at different spatial scales (0-300 m and 0-3000 m) on the absolute differences between ground-verified and GIS predicted values was evaluated using generalized linear models (GLMs). Results: Landscapes with high variability in the evaluated edaphic attributes showed higher differences between ground-verified and GIS predicted data, which suggested that the prediction accuracy of soil properties with GIS techniques decreases in landscapes with more variable edaphic attributes However, differences between ground-verified and GIS predicted data were generally lower in landscapes where edaphic and topographic data were spatially more heterogeneous. This could be the result of there being greater samples taken to develop the SSURGO database from areas with more heterogeneous soils or topography. Furthermore, differences between ground-verified and GIS predicted soil data were higher when the sample sites were nearer transportation routes and/or utility ROWs. Conclusion: Results showed that the surrounding land use and edaphic and topographic landscape highly influence the prediction accuracy of soil attributes with GIS techniques.