Spatiotemporal analysis of land degradation and soil erosion dynamics in the Chambal badlands using machine learning and geospatial techniques

Abstract The accelerating land degradation processes, including soil erosion and ravine formation, pose a critical challenge to global food security and agricultural sustainability. By integrating CA-Markov, geospatial modeling, SARIMA, and machine learning, this study investigates the spatiotemporal dynamics of land degradation and soil erosion in the Chambal Badlands across a 64-year horizon (1992–2056). The analysis reveals that ravine areas decreased from 14.59% in 1992 to 7.42% in 2024, with projections suggesting a further decline to 2.22% by 2056. Total soil erosion decreased from 27.56 million t yr − 1 in 1992 to 17.93 million t yr − 1 in 2024 and is projected to decrease further to 12.11 million t yr − 1 by 2056. These shifts are primarily driven by land reclamation initiatives implemented across the region. However, soil erosion intensity has increased locally within high-slope ravine belts. This is driven by climate change, as evidenced by the rise in rainfall erosivity (R-factor) from 407 MJ·mm·ha − 1 ·h − 1 ·yr − 1 in 1992 to 463 MJ·mm·ha − 1 ·h − 1 ·yr − 1 in 2024, with a projected increase to 482 MJ·mm·ha − 1 ·h − 1 ·yr − 1 by 2056. Among the soil erosion drivers (soil, topography, rainfall, and land management), machine-learning results identify land management practices (P-factor) as the dominant control on erosion, with the highest contribution (32.8% in 2024). The results suggest that targeted land management interventions, combined with climate-adaptive planning, can substantially mitigate future erosion risks. The study provides valuable insights for sustainable land-use planning and supports progress toward the Sustainable Development Goals, specifically SDG 2 (Zero Hunger), SDG 13 (Climate Action), and SDG 15 (Life on Land).