Modelling climate-driven agricultural land use change in a data-limited region: a big data analytics framework for the Kabul River Basin (KRB)

Abstract

In regions characterized by data scarcity, understanding the intricate coupling between hydrological variability and land-use transitions is paramount for regional stability. This study introduces an integrated big data analytics framework to assess the impact of discharge from the Kabul River Basin (KRB) on agricultural sustainability in Pakistan. Leveraging the Google Earth Engine (GEE) platform, we implemented a supervised random forest classification of Landsat multi-temporal imagery (1996–2021) to quantify land use and land cover (LULC) dynamics. The results revealed significant landscape instability: agricultural land contracted by 16.45%, while built-up areas expanded by 25.94%. Furthermore, a critical 37% collapse in surface water extent was observed following its 2006 peak, providing physical evidence of increasing hydrological stress. To address the ‘precipitation paradox’ identified in the basin, where agricultural persistence shows a near-zero correlation with local rainfall (r = 0.011), this study used a cellular automata (CA)-Markov model integrated with a hydrological sensitivity analysis. Projections indicate a 30% decline in arable land by 2030, driven by a synergistic combination of non-elastic urban encroachment and an anticipated 18% reduction in river discharge. These findings, validated with high accuracy (Kappa coefficients: 0.85–0.89), suggest that a region's food security is increasingly vulnerable to threshold-based land abandonment. The study concludes that the KRB's environmental resilience is at high risk without integrated transboundary water cooperation and evidence-based sustainable resource policies to manage the non-linear water–food nexus.