Multi-level pedotransfer modification functions of the USLE-K factor for annual soil erodibility estimation of mixed landscapes

Abstract

This study presents new pedotransfer functions for realistic and dynamic predictions of the USLE soil erodibility factor (K). Multiple linear regression (MLR) models capturing the factors of the USLE of Fish river watershed in Alabama are developed using a dynamic remotely sensed land cover data from MODIS integrated into ArcSWAT model. The correlation matrix results of a forward stepwise MLR analysis for the period 2001–2011 show that the K-factor is profoundly affected by five variables namely, slope length and steepness, crop management factor, soil permeability, soil moisture content, and soil bulk density. The study demonstrates that the soil moisture content and soil bulk density factors that are commonly unaccounted for in the traditional USLE have a significant impact on K-factor. The predictions of the MLR models developed in the study are verified for the years 2006 and 2011. The results show that the MLR model containing all five variables is a good fit for K-factor (R2: 0.89 for 2006; 0.94 for 2011). The study shows that the mean values of K estimated for 2001–2011 using the USLE and MLR are respectively, 1.7 and 1.2 times the mean value of the measured K. The strong correlation between the estimated K from the MLR model and the measured K indicates that the use of soil moisture content (R2 = 0.84, p < 0.05) and soil bulk density (R2 = 0.77, p < 0.05) as predictor variables gives a better all-inclusive estimation of the soil erodibility factor in regions with noticeable temporal land cover transformations.