Evaluating Land Subsidence Triggered by the 7.8 Mw Turkey-Syria Earthquake Using an Advanced Machine Learning Model
Keywords: Land Subsidence, Surface Deformation, Susceptibility, Turkey-Syria Earthquake, Extreme Learning Machine
Abstract. The 7.8 Mw Turkey-Syria earthquake of 2023 caused massive destruction in several cities near the earthquake epicenter. However, there is a potential for significant land subsidence to occur across a broader region. Land subsidence, which can lead to significant infrastructure damage and ground deformation, necessitates detailed investigation. This research uses an advanced machine-learning technique to analyze the spatial distribution of earthquake-induced land subsidence and the extent of surface deformation. Sentinel-1 Synthetic Aperture Radar (SAR) data were processed to detect surface deformation near the epicenter and quantify the affected region's vertical displacement. An extreme learning machine model was developed using nine parameters, including slope, curvature, sediment thickness, soil thickness on slopes, peak ground acceleration, hydrologic soil, Vs30, land cover, and landslide probability. The model accurately predicted land subsidence susceptibility (accuracy of 85%) established correlations with ground deformation and observed vertical displacement. The results demonstrate that the deformation phase value ranges between 2.66 to −2.63, and the vertical displacement analysis suggests that a large portion of areas subsided downward up to 75 cm. Effectiveness of extreme learning machine in rapid land subsidence assessment, providing critical insights for disaster response and urban planning in seismically active areas. This study offers a useful solution for post-earthquake land subsidence analysis and lays the groundwork for integrating artificial intelligence with land subsidence research.