Aizemaitijiang Maimaitituersun

Accurate estimation of shallow groundwater depth (SGWD) is crucial for ecological security in arid oases. However, sparse monitoring wells and strong spatiotemporal heterogeneity hinder reliable monthly estimation. Traditional interpolation methods struggle to capture complex environmental drivers in anthropogenically modified regions. To address this issue, multi-source remote sensing data were integrated with machine learning models to estimate SGWD in the Weiku Oasis, Xinjiang, China. Field-measured SGWD data from March to November 2020 were collected. Based on the driving mechanisms of SGWD, six categories of variables were extracted, including optical, radar, topographic, meteorological, soil, and human activity factors. Four modeling strategies were designed by combining these variables with random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost) models. The optimal variable-model combination was selected for monthly SGWD mapping. SHAP was employed to interpret monthly feature importance and variable influence mechanisms. The results showed the following: (1) Among the four modeling strategies, Strategy II and Strategy IV achieved the best performance in monthly SGWD estimation, indicating that either single optical remote sensing data or the synergy of optical and radar remote sensing data can effectively characterize the spatiotemporal heterogeneity of SGWD in similar irrigated oasis regions. (2) Under different “variable-model” combinations, the RF model performed best in March, April, May, and October, whereas the SVM model performed best in June, July, August, September, and November, with XGBoost showing the weakest overall performance. Furthermore, the machine learning models outperformed Ordinary Kriging, yielding a mean R² of 0.790 and a 5.23% MAE reduction compared to −0.325. (3) Monthly SGWD maps revealed shallower depths in central and southeastern farmlands and piedmont alluvial fans, and greater depths in northern and northwestern desert margins due to insufficient groundwater recharge. (4) The dominant controls on SGWD exhibited clear seasonal variation. Salinity indicators, radar-derived wetness, and evapotranspiration prevailed in spring, followed by topographic and thermal–evaporative drivers in summer and by topography, precipitation, and soil properties in autumn. Despite these seasonal shifts, valley depth remained among the top five predictors throughout the nine months. This study provides a scientific method and technical support for dynamic monitoring and refined management of groundwater resources in similar irrigated oasis regions.

The Gravity Recovery and Climate Experiment (GRACE) introduces a new approach to accurately monitor, in real time, regional groundwater resources, which compensates for the limitations of traditional hydrological observations in terms of spatiotemporal resolution. Currently, observations of groundwater storage changes in Jiangsu Province face issues such as low spatial resolution, limited applicability of the downscaling models, and insufficient water resource observation data. This study based on GRACE employs Random Forest Regression (RFR) and Geographically Weighted Regression (GWR) methods in order to obtain high-resolution information on groundwater storage change. The results indicate that among the established 66 × 158 local GWR models, the coefficient of determination (R²) ranges from 0.39 to 0.88, with a root mean squared error (RMSE) of approximately 2.60 cm. The proportion of downscaling models with an R² below 0.5 was 18.52%. Similarly, the RFR models trained on the above time series grid data achieved an R² of 0.50, with the RMSE fluctuating around 1.59 cm. In the results validation, the monthly correlation coefficients between the GWR downscaling results and the data of measured stations ranged from 0.37 to 0.66, with 53.33% of the stations having a coefficient greater than 0.5. The seasonal correlation coefficients ranged from 0.41 to 0.62, with 60% of the stations exceeding 0.5. The correlation coefficients for the RFR downscaling results ranged from 0.44 to 0.88, with seasonal correlation coefficients ranging from 0.49 to 0.84. Only one station had a correlation coefficient below 0.5 for both monthly and seasonal results. In the validation of the correlation accuracy between the downscaling results and the measured groundwater levels, the Random Forest model demonstrated better predictive performance, which offers distinct advantages in improving the spatial resolution of groundwater storage changes in Jiangsu Province.