Monitoring Groundwater Storage Based on Satellite Gravimetry and Deep Learning

Improper abstraction of groundwater in Iran has led to an average annual subsidence rate of 15 cm/yr. The management of Iran's water resources is essential due to its arid and semiarid climate and traditional agriculture. Monitoring groundwater storage (GWS) changes and their correct interpretation using deep learning (DL) methods can improve our understanding of groundwater systems. For this purpose, in this study, the GWS in Iran from 2003 to 2021 was downscaled using DL based on combining gravity recovery and climate experiment (GRACE) and GRACE-Follow on with a hydrological model. The GWS downscaling was performed from 1° to 0.25°. The GWS in the south of Tehran and northeast of Qazvin had the highest decrease of 15 mm/yr. A new GWS index was developed to correctly interpret the decline in GWS through the standardized precipitation index. The main reason for the decrease in GWS was the development of unsustainable agriculture from 2007 to 2012, which reached its lowest possible level after 2012–2018 with the intensification of climatic conditions. The calculated GWS index correlates more than 80% with 400 piezometric wells in Iran.