Modeling Post-Disaster Urban Sprawl Trajectories Through ANN-Based Land Use/Land Cover Change Analysis and Scenario Projection

Urban expansion in the aftermath of natural disasters presents critical challenges to sustainable land use planning and environmental resilience. This study models post-disaster urban growth trajectories by analysing land use and land cover (LULC) changes from 2017 to 2025 and simulating future development scenarios up to 2040 using an Artificial Neural Network (ANN) integrated within the QGIS MOLUSCE module. Sentinel-2 satellite imagery and digital elevation models were employed to classify six LULC categories: built-up areas, croplands, rangelands, bare land, forested areas, and water bodies. The methodology was applied to the Central District of Elazığ, Türkiye a region significantly affected by the 2020 earthquake. Results reveal a substantial 60.89% increase in built-up areas, primarily driven by rapid post-disaster reconstruction. This expansion has coincided with notable reductions in cropland and rangeland coverage, while forest and water body areas have increased due to afforestation projects and water-related infrastructure investments. Scenario-based projections indicate that, if current urbanisation trends persist, pressure on ecologically sensitive and agriculturally valuable lands will likely intensify by 2040. The ANN model demonstrated high predictive accuracy, with an overall correctness of 97.52% and a Kappa coefficient of 0.96. Based on these findings, the study recommends integrated planning strategies that: (i) prevent development on fertile agricultural plains, (ii) incorporate ecological thresholds in urban site selection, (iii) avoid construction near seismic fault lines and water bodies, and (iv) promote green infrastructure solutions including ecological corridors, urban forests, and sustainable stormwater systems within post-disaster urban development frameworks.