Tree cover increases on agricultural land amid continued landscape-level deforestation: a case study from eastern Afghanistan

Understanding changes in tree cover within agricultural landscapes is critical for assessing the resilience of rural communities and consequent land use dynamics in conflict-affected regions. This study examines spatiotemporal land use and tree cover changes across four districts of Nangarhar Province in eastern Afghanistan using high-resolution (3 m/pixel) satellite imagery. The temporal window examined for this study (2016-2024) spans a period of significant geopolitical upheaval in Afghanistan, including the collapse of the republican government in 2021 and the resulting shift in political power and governance dynamics. An object-based image analysis framework combined with a support vector machine classifier was applied to map five land use classes within settlement areas: tree cover, cropland, water, built-up areas, and barren land. Classification accuracy was assessed using stratified random sampling and Cohen's kappa coefficient, with all maps achieving κ values above 0.80. Post-classification change detection and transition matrices were used to quantify class-to-class conversions in hectares, while Sankey diagrams were employed to visualise the magnitude and direction of land use transitions. The results reveal varying patterns of tree cover change across the study districts, with evidence of both tree cover gains linked to expansion of agroforestry and on-farm tree retention, and losses driven by agricultural expansion and development. These changes may influence local livelihoods, as trees on farms provide important resources such as fruit, fuelwood, and shade. The findings demonstrate the effectiveness of integrating high-resolution remote sensing, object-based classification, and Sankey-based visualisation for analysing complex land use dynamics.