Prediction of groundwater level variations based on gravity recovery and climate experiment (GRACE) satellite data and a time-series analysis: a case study in the Lake Urmia basin, Iran

The proper management and planning of groundwater resources depends on the awareness of the process of spatial and temporal changes of groundwater levels in aquifers. For the Lake Urmia basin in northwestern Iran, average groundwater levels have been obtained from existing GRACE satellite data, including CSR, GFZ and JPL, in conjunction with the Google Earth Engine for the period 2002–2016. The Mann–Kendall test shows a significant decreasing trend in groundwater level for all three GRACE datasets during the study period. Parametric time-series models, including ARIMA, MARS, PROPHET and ETS, were fitted to GRACE data. For all three datasets, the PROPHET model had the highest coefficients of R² (0.81–0.85) and the lowest MAE model error coefficients and RMASE The residuals of the PROPHET model are normally distributed, which indicates this model could be reliable for predictive purposes. For all three GRACE datasets, the PROPHET model shows a declining trend of groundwater levels in the Uremia Lake basin during the period 2017–2024.