Analyzing and modeling land subsidence hazard in delta regions using time series InSAR and explainable artificial intelligence: A case study across the Nile Delta, Egypt

Land subsidence poses a significant risk to densely populated delta regions, including the Nile Delta in Egypt. We employ the Interferometric Synthetic Aperture Radar (InSAR) time series analysis to monitor and analyze subsidence trends across the entire Nile Delta from 2015 to 2023. We observe subsidence rates ranging from −3 mm/year to over −15 mm/year in major cities such as Alexandria, Damietta, and Rosetta, with higher subsidence observed along the coastline and reclaimed areas. We developed a deep learning model to quantitatively link subsidence rates to various environmental and anthropogenic factors. We relied on the available in situ datasets, complemented with satellite-derived information. This approach ensures that our model incorporates the best obtainable representation of subsidence drivers in the study area, despite limitations in direct field observations. The model demonstrates high prediction accuracy with a Peak Signal-to-Noise Ratio (PSNR) of 39.7 and a Structure SIMilarity (SSIM) index of 0.93, with the ability to make future predictions for up to 6 months. Moreover, it enables interpreting the key drivers behind subsidence using Shapley additive explanations (SHAP), a tool for explainable artificial intelligence. To our knowledge, this is the first implementation of applying the SHAP AI technique across both spatial and temporal domains. Therefore, this helps to understand the contribution of each potential causative factor to accelerated land subsidence trends in space and time. The results emphasize that land use, land cover, urban load, terrestrial water storage, and clay content in the high-productive aquifers are key driving mechanisms for the observed land subsidence. The study presents a novel approach to analyze and model land subsidence hazards in delta regions that can be utilized for any other susceptible regions in the world.