Damage assessment following M 7.8 and 7.5 earthquakes in Turkey in 2023 using space-borne synthetic aperture radar (SAR) imagery and CNN

Rapid and accurate damage assessment plays a pivotal role in mitigating the dire consequences of natural or man-made disasters. Foremost, one of the main reasons for swift damage assessment and disaster response is the potential to save lives. In this study, we exploit the advantage of synthetic aperture radar (SAR) imagery to achieve a quick response over southern Turkey and northern Syria, which experienced two large earthquakes (M 7.8 and M 7.5) on 6th February 2023. We used SAR images supplied from the Sentinel-1 mission to create initial damage maps using interferometric (e.g., coherence) and intensity (e.g., backscattering coefficient) information. The created initial maps together with other influential factors, such as Vs30 (speed of shear wave velocity from the surface to a depth of 30 m), geological and PGA (peak ground acceleration) maps are used to build a new CNN (convolutional neural network) model. The proposed CNN model achieves an overall accuracy of 86% and a kappa coefficient of 0.70 following iterative training and validation. The produced results cover approximately 90,000 km² and can be used for enhanced search-and-rescue operations.