Abstract
Gully erosion is an environmental problem that causes topsoil loss and ecological damage. Understanding its formative processes and predisposing factors is crucial for effective management. However, few studies have focused on the spatial modelling of these factors. In the current study, machine learning (ML)-based spatial statistics were used to identify the predisposing factors of gully formation processes in south-eastern Nigeria. A geospatial database was created using 262 gully locations, fragmented into four categories as dependent variables. A total of 20 predictor variables, including geomorphometry, terrain, geo-environmental factors, proximity, and climate, were fused into a multinomial logistic regression model. The model was then fitted using the nnet package in R after testing for multi-collinearity and spatial autocorrelation. The results revealed that the study area, which is a 7th-order drainage basin, is vulnerable to gullying, as expressed by Horton's laws of geomorphometry. The highest simulated probability range was observed for moderate gully (0.5512–0.7175), with terrain and climate as predictive factors, whereas the lowest simulated probability range was observed for full gully (0.1136–0.4107) as influenced by basin geomorphometry. Soil creep (0.3869–0.5317) and low gully (0.1025–0.5107), both driven by climate, proximity, and terrain factors, exhibited a moderate tendency for occurrence. The findings reported in this paper can aid in decision-making for sustainable environmental management of gully cycles and provide a novel approach for global gully-active drainage basins.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call