Forecasting urban shifts post-earthquake: LULC change analysis in Elazığ, Turkey using ANN and Markov models

Abstract

Understanding land use and land cover (LULC) dynamics in seismically active regions is crucial for risk-informed urban planning and sustainable post-disaster recovery. This study investigates the impact of the Mw 6.8 Elaz & imath;& gbreve; earthquake (24 January 2020) on LULC patterns in eastern Turkey by integrating high-resolution Sentinel-2 satellite imagery with geographic information systems (GIS), remote sensing (RS), artificial neural networks (ANNs), and Markov chain modelling. The methodology comprises four phases: establishing a pre-earthquake baseline (2015-2019), assessing post-earthquake changes (2015-2023), analysing transition probabilities to identify key LULC drivers, and forecasting land-use scenarios for 2030 and 2050 under seismic and non-seismic conditions. Results reveal that seismic activity significantly accelerates urban expansion, shifting development towards geologically stable zones. By 2050, artificial surfaces are projected to occupy 54.70% of the region under seismic influence, compared to 48.87% without it. Agricultural land is more preserved in the seismic scenario (26.54%) than in the non-seismic case (22.68%), while pasture and meadow areas decline sharply to 6.18%, raising concerns for biodiversity and ecosystem services. These findings emphasise the importance of integrating ecological considerations and seismic risk into land-use planning frameworks. By combining multicriteria decision-making with machine learning-based forecasting, the study offers a replicable and scalable model for balancing urban growth, environmental conservation, and resilience. Framed within interdisciplinary insights from disaster resilience theory, urban governance, and spatial risk modelling, this research contributes to the global discourse on sustainable urban transformation in the face of increasing natural hazards.