Simulating soil erodibility in southeastern China using a sequential Gaussian algorithm

Abstract

Soil erodibility (K) is a key factor of soil erosion, and its appropriate quantification and interpolation are vitally important for soil and water conservation. The traditional point-represent-polygon approaches and common kriging method for the estimation of K, however, do not sufficiently represent the original data. The objectives of this study were to simulate the spatial distribution of K using a sequential Gaussian algorithm and analyze the uncertainty in evaluating the risk of soil erodibility in southeastern China. We determined 101 sampling points in the area and collected disturbed soil samples from the 0–20 cm layer at each point. Soil properties were determined, and K was calculated using five common models: the EPIC (Erosion/Productivity Impact Calculator), approximate nomograph, Torri, Shirazi, and Wang models. Among the chosen models, the EPIC model performed the best at estimating K(KEPIC), which ranged from 0.019 to 0.060 t ha h (ha MJ mm)−1, with a mean of 0.043 t ha h (ha MJ mm)−1. The KEPIC was moderately spatially variable and had a limited spatial structure, increasing from south to north in our study area, and all spatial simulations using the cooperative kriging (CK) interpolation and the sequential Gaussian simulation (SGS) with 10, 25, 50, 100, 200, and 500 realizations had acceptable accuracies. The CK interpolation narrowed the range, and the SGS maintained the original characteristics of the calculated data. The proportions of the risk area were 38.0% and 10.1%, when the risk probability for K was 60% and 80%, respectively, and high risk areas were mostly located in the north. The results provide scientific guidance for managing the risk of soil erodibility in southeastern China.