Machine Learning for Predicting Land Degradation in Indonesian Agricultural Regions

Abstract

Land degradation threatens agricultural productivity, food security, and environmental sustainability across Indonesia. Advances in machine learning provide new opportunities to improve the prediction of degradation hotspots and support proactive land management. This study develops predictive models using machine learning algorithms, including Random Forest, Support Vector Machine, and Extreme Gradient Boosting, to assess land degradation risk in major agricultural regions. Environmental variables such as rainfall, slope, soil characteristics, vegetation indices, and land-use patterns were integrated into the predictive framework using geographic information systems and remote sensing datasets. Model performance was evaluated using cross-validation and standard classification metrics. The results demonstrate that ensemble machine learning models achieved high predictive accuracy and successfully identified priority areas vulnerable to erosion and declining land productivity. Vegetation cover, rainfall intensity, and slope were consistently identified as the most influential predictors. The developed framework enables more efficient targeting of soil conservation measures and supports evidence-based agricultural planning. Integrating artificial intelligence with spatial environmental data can strengthen sustainable land management while improving resilience to climate variability. The proposed approach provides a scalable decision-support tool for policymakers responsible for conserving Indonesia's agricultural resources.