GEE-Based Spatiotemporal Analysis of Rainfall Variability and Soil Erosion in Northern Bangladesh

Abstract

In this study, a new framework to evaluate the potential soil erosion spatiotemporal response to monsoon anomalies based on dynamic RUSLE modeling and analysis of interannual rainfall deviation is developed and applied over the Rangpur Division in northern Bangladesh for 38 years from 1987 to 2024 using Google Earth Engine (GEE). As opposed to earlier ones that use RUSLE under static climatic conditions, R-factors are calculated for every year using CHIRPS pentadal precipitation data; this allows a direct linkage between rainfall anomalies and erosion severity at the administrative level. Mann-Kendall tests showed no monotonic rainfall trend in any of the eight districts over time, suggesting that the hydroclimatic signal is one of interannual variability, and not necessarily one of a long-term directional change. This framework adequately covers the monsoon conditions both during extreme drought (2018) and extreme surplus (2020), and reveals that the variability in soil loss between years is driven by the variability in rainfall, with a strongly statistically significant correlation between the rainfall deviation and soil loss (R = 0.51, p = 0.00086). Kurigram turned out to be the most sensitive district with r = 0.8687. Such results serve as an empirical foundation for adaptive, climate responsive land management planning in data-poor monsoon areas.