Performance of Frequently Used Interpolation Methods to Predict Spatial Distribution of Selected Soil Properties in an Agricultural Watershed in Ethiopia

Abstract

A soil map of a watershed provides a wealth of knowledge and can be vital for implementing site-specific soil managements. Hence, watershed-based soil assessment was conducted to select an optimum spatial interpolation method, while aiming for sustainable soil managements. Intensive soil sampling was undertaken to investigate the performance of ordinary kriging (OK), inverse distance weighting (IDW), and radial basis functions (RBF) for predicting the spatial distribution of soil texture, pH, soil organic carbon (SOC), and available phosphorus (AP). The 72 ha study area was divided into a 100 x 100 m grid and approximately at the center of each grid, topsoil samples (roughly from 10-25cm depth) were collected from over 75 locations across the entire watershed. The exponential and Gaussian models were best fitted in the semivariogram of measured soil. The performance of each interpolation method was assessed quantitatively in terms of Nash-Sutcliffe efficiency (E), coefficient of determination (R2), and index of agreement (d). The interpolated maps generated based on the highest value of E displayed OK was best performed for SOC and sand. RBF was most suitable for mapping of AP and clay, while IDW gave better results when applied to pH. The highest value of R2, E, and d (0.51, 0.51, and 0.83, respectively) resulted from the spatial interpolation of AP. Overall, the cross-validation statistics for each interpolation method showed there was no single method that significantly outperformed the others. Therefore, one of the interpolation methods could be applied for surface map generation in future studies of similar regions.