Exploring driving factors of soil erosion using a Multiscale GWR model: a case study at Central Citarum Watershed, West Java, Indonesia

Abstract

To address challenges in soil erosion management and ecological rehabilitation, understanding the determinants of soil erosion is crucial. This research aims to achieve two primary objectives: (1) delineating the spatial patterns of soil erosion within the designated region and (2) identifying the influential factors using the Multiscale Geographical Weighted Regression (MGWR) methodology. The methodological framework involved the creation of grid datasets, with soil erosion as the response variable and a combination of physical and socioeconomic attributes as predictors. We extracted 550 data points from raster datasets, specifically centered on village locations, using the ‘extract multi-value to point’ tool in ArcGIS. The R Studio environment was utilized to select the relevant factors influencing soil erosion. The geographical detector technique was applied to determine the explanatory power of each determinant concerning the spatial patterns of soil erosion. Subsequently, data from the Ordinary Least Squares (OLS) model underwent MGWR analysis. The findings reveal that the Central Citarum Watershed experiences an estimated annual soil erosion of 23.16 million tons, averaging 102.01 tons per hectare. The analysis identified LS (slope length and gradient) and CP (vegetative cover and supportive practices) as the primary variables influencing the spatial variability of soil erosion. Notably, the MGWR model demonstrated enhanced explanatory capacity and effectiveness compared to both the OLS and Geographically Weighted Regression (GWR) methodologies.