Erosion Prediction with USLE and RUSLE2

• Activity Number: 191
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Yang Li*+
• Companies: Iowa State University
• Keywords: USLE ; RUSLE2 ; multivariate adaptive regression splines

Abstract

Universal Soil Loss Equation is a model that predicts the long term average annual rate of erosion on a field. It is replaced by the recently released RUSLE2 model. It is necessary to impute USLE estimations after 2006 and RUSLE2 estimation before 2002. We present models to predict soil loss in one model using the variables in the other one. With the aid of cross validation, we analyze an Iowa data set on field erosion with three fitting methods: simple linear regression, multiple linear regression, and multivariate adaptive regression splines. We found that MARS has the best prediction power of RUSLE2 soil loss based on USLE variables, while MLR is the best choice for predicting USLE soil loss based on RUSLE2 variables. The coefficient of variation computed for each observational point and a county-level CV map is produced to identity abnormal counties and spatial pattern.