Groundwater is one of the most important freshwater reservoirs in the world that plays an important role in conserving ecosystem and ensuring human well – being. When groundwater systems are affected by drought, first the richarge, then the groundwater level and finally the possibility of discharge decreases, which can affect agriculture and water resources and lead to irreversible effects such as subsidence. Until the implementation of the Gravity Recovery and Climate Experiment (GRACE) project, there was no method for monitoring terrestrial water storage anomalies (TWSA), which includes both surface and subsurface water, and monitoring groundwater storage (GWSA) with traditional methods and piezometric wells was time-consuming, expensive, and unable to adequately cover large spatial and temporal ranges. In this study, the status of groundwater in 6 major river basins in Iran is eva‎luated using data from the GRACE project and its complementary project, the Global Land Data Assimilation System (GLDAS), during the period of 255 months from 2002 to 2023. However, GRACE data also has limitations, such as its large 0.5-degree scale that poses challenges for studies in extent of basins and auifers. In this study, the XGBoost machine learning algorithm is used for downscaling, which has shown excellent performance in modeling TWSA data. Calculated GWSA data is compared with observation data from the country's piezometric wells for the years 2003 to 2016, resulting in a correlation coefficient of 0.92 and an RMSE of 39 mm for the entire Iran. Another limitation of using GRACE data is the presence of data gaps for a period of 15 months due to the project's interruption and restart, which in this study, the SARIMA model is used to complete the GWSA time series, which has successfully modeled the time series and filled the existing gap with accurate predictions. From the completed time series, it can be found that on average, the wells in the country's aquifers have a level drop of about 29 cm per year and the highest rate of decrease is for to the Caspian Sea basin, and the lowest rate is related to the Persian Gulf basin and the Qaraqom basin. Furthermore, in this study, using the historical data and scenario projections of SSP2 of the canESM5 climate model, the GWSA is modeled and its changing trends are predicted until 2030 using machine learning methods. The created model has shown relatively good performance in all basins except for the eastern border basin and the Qaraqum basin. In addition, the status of groundwater drought is examined in the basins of Iran from 2002 to 2030 using the Groundwater Drought Severity Index (GGDI) and compared with the Standardized Precipitation Index (SPI). Overall, a correlation of 8 months is observed between meteorological drought and groundwater drought. Furthermore, the comparison of the GGDI with atmospheric-oceanic factors has shown that these factors have a significant impact on climate and drought variations in Iran, especially in basins adjacent to extensive surface water resources.

حميدرضا صفوي

محمدحسين گل محمدي , رامتين معيني