Gravity Recovery and Climate Experiment (GRACE) satellites can effectively monitor terrestrial water storage (TWS) changes in large-scale areas. However, due to the coarse resolution of GRACE products, there is still a large number of deficiencies that need to be considered when investigating TWS changes in small-scale areas. Hence, it is necessary to downscale the GRACE products with a coarse resolution. First, in order to solve this problem, the present study employs modeling windows of different sizes (Window Size, WS) combined with multiple machine learning algorithms to develop a new machine learning spatial downscaling method (MLSDM) in the spatial dimension. Second, The MLSDM is used to improve the spatial resolution of GRACE observations from 0.5° to 0.25°, which is applied to Guantao County. The present study has verified the downscaling accuracy of the model developed through the combination of WS3, WS5, WS7, and WS9 and jointed with Random Forest (RF), Extra Tree Regressor (ETR), Adaptive Boosting Regressor (ABR), and Gradient Boosting Regressor (GBR) algorithms. The analysis shows that the accuracy of each combined model is improved after adding the residuals to the high-resolution downscaled results. In each modeling window, the accuracy of RF is better than that of ETR, ABR, and GBR. Additionally, compared to the changes in the TWS time series that are derived by the model before and after downscaling, the results indicate that the downscaling accuracy of WS5 is slightly more superior compared to that of WS3, WS7, and WS9. Third, the spatial resolution of the GRACE data was increased from 0.5° to 0.05° by integrating the WS5 and RF algorithm. The results are as follows: (1) The TWS (GWS) changes before and after downscaling are consistent, decreasing at −20.86 mm/yr and −21.79 mm/yr (−14.53 mm/yr and −15.46 mm/yr), respectively, and the Nash–Sutcliffe efficiency coefficient (NSE) and correlation coefficient (CC) values of both are above 0.99 (0.98). (2) The CC between the 80% deep groundwater well data and the downscaled GWS changes are above 0.70. Overall, the MLSDM can not only effectively improve the spatial resolution of GRACE products but also can preserve the spatial distribution of the original signal, which can provide a reference scheme for research focusing on the downscaling of GRACE products.

Data from the Gravity Recovery and Climate Experiment (GRACE) satellite mission can be used to monitor changes in terrestrial water storage (TWS). In this study, we exploit the TWS observations from a new temporal gravity field model, Tongji-Grace2018, which was developed using an optimized short-arc approach at Tongji University. We analyzed the changes in the TWS and groundwater storage (GWS) in each of the nine major river basins of the Chinese mainland from April 2002 to August 2016, using Tongji-Grace2018, the Global Land Data Assimilation System (GLDAS) hydrological model, in situ observations, and additional auxiliary data (such as precipitation and temperature). Our results indicate that the TWS of the Songliao, Yangtze, Pearl, and Southeastern River Basins are all increasing, with the most drastic TWS growth occurring in the Southeastern River Basin. The TWS of the Yellow, Haihe, Huaihe, and Southwestern River Basins are all decreasing, with the most drastic TWS loss occurring in the Haihe River Basin. The Continental River Basin TWS has remained largely unchanged over time. With the exception of the Songliao and Pearl River Basins, the GWS results produced by the Tongji-Grace2018 model are consistent with the in situ observations of these basins. The correlation coefficients for the Tongji-Grace2018 model results and the in situ observations for the Yellow, Huaihe, Yangtze, Southwestern, and Continental River Basins are higher than 0.710. Overall, the GWS results for the Songliao, Yellow, Haihe, Huaihe, Southwestern, and Continental River Basins all exhibit a downward trend, with the most severe groundwater loss occurring in the Haihe and Huaihe River Basins. However, the Yangtze and Southeastern River Basins both have upward-trending modeled and measured GWS values. This study demonstrates the effectiveness of the Tongji-Grace2018 model for the reliable estimation of TWS and GWS changes on the Chinese mainland, and may contribute to the management of available water resources.

Thaw slumps are well-developed within a 10 km wide zone along the Qinghai-Tibet engineering corridor, especially along the Qinghai-Tibet highway and railway. Previous studies have focused on thaw slump instability such as its origin development, headwall retrogression rate, failure scale, and thermal regime, yet the intrinsic dynamic process of surface movement is relatively less known. In this study, we used InSAR based on the L-band ALOS PALSAR images acquired from January 2007 to October 2010 to investigate the distribution of thaw-induced slope failures containing retrogressive thaw slumps and active layer detachment failures along the Qinghai-Tibet highway (QTH). Our InSAR analysis reveals that the maximum annual average sedimentation rates are even up to -35 mm·yr⁻¹ in the slope direction to the K3035 thaw slump, and the K3035W active layer detachment failure developed on the west side of K3035. The distribution, failure extent, and stability of the slope failures obtained by our InSAR analysis all agree well with the field investigations. Our study illustrates that InSAR is an effective tool for studying the distribution and processes of the thaw slump-derived thermokarst and provides useful references for evaluating permafrost degradation in response to climate warming and external disturbance on the Qinghai-Tibet plateau.

Obtaining high-resolution products that can accurately estimate the spatiotemporal changes in regional water resources is essential for the rational utilization of water resources. The temporal gravity field model derived from the Gravity Recovery and Climate Experiment (GRACE) satellite can effectively monitor the regional terrestrial water storage (TWS) changes. Nevertheless, the main shortcoming of GRACE products is the coarse spatial resolution which limits its applicability to small-scale regions. Herein, we propose a geographically weighted regression downscaled model (GWRDM) to improve the spatial resolution of TWS from 0.5° to 0.1° and validate it against the downscaled results provided by the spatial global regression downscaled model (SGRDM) and temporal grid regression downscaled model (TGRDM). The results show that the downscaled products based on GWRDM outperform SGRDM and TGRDM. The GWRDM not only effectively improves the spatial resolution of GRACE products but also maintains the intensity of the original signal. Similarly, the GWRDM is further used to downscale the groundwater storage (GWS) and validated against the in-situ observations. The downscaled GWS based on GWRDM agrees well with in-situ observations. On the monthly scale, the average Correlation Coefficient ( CC ) for both is 0.53 and above 0.70 for some in-situ observations. On the annual scale, 81.25% and 37.50% of the in-situ observations have CC values larger than 0.60 and 0.90, respectively, which further verifies the reliability of GWRDM. Consequently, the GWRDM provides a practical algorithm for obtaining high-resolution water storage estimates, which is expected to provide a reference for small-scale regional water resources management.

Densely distributed Global Navigation Satellite System (GNSS) stations can invert the terrestrial water storage anomaly (TWSA) with high precision. However, the uneven distribution of GNSS stations greatly limits the application of TWSA inversion. The purpose of this study was to compensate for the spatial coverage of GNSS stations by simulating the vertical deformation in unobserved grids. First, a new deep learning weight loading inversion model (DWLIM) was constructed by combining the long short-term memory (LSTM) algorithm, inverse distance weight, and the crustal load model. DWLIM is beneficial for improving the inversion accuracy of TWSA based on the GNSS vertical displacement. Second, the DWLIM-based and traditional GNSS-derived TWSA methods were utilized to derive TWSA over mainland China. Furthermore, the TWSA results were compared with the TWSA solutions of the Gravity Recovery and Climate Experiment (GRACE) and Global Land Data Assimilation System (GLDAS) model. The results indicate that the maximum Pearson’s correlation coefficient (PCC), Nash–Sutcliffe efficiency (NSE) coefficient, and root mean square error (RMSE) equal 0.81, 0.61, and 2.18 cm, respectively. The accuracy of DWLIM was higher than that of the traditional GNSS inversion method according to PCC, NSE, and RMSE, which were increased by 67.11, 128.15, and 22.75%. The inversion strategy of DWLIM can effectively improve the accuracy of TWSA inversion in regions with unevenly distributed GNSS stations. Third, this study investigated the variation characteristics of TWSA based on DWLIM in 10 river basins over mainland China. The analysis shows that the TWSA amplitudes of Songhua and Liaohe River basins are significantly higher than those of the other basins. Moreover, TWSA sequences in each river basin contain annual seasonal signals, and the wave peaks of TWSA estimates emerge between June and July. Overall, DWLIM provides a useful measure to derive TWSA in regions where GNSS stations are uneven or sparse.

Temporal gravity field models derived from the Gravity Recovery and Climate Experiment and its Follow-On mission [GRACE(-FO)] observations effectively detect global surface mass changes. However, the 11-month gap between the GRACE(-FO) missions introduces biases into the research analysis and limits further application. Therefore, ensuring the continuity of the GRACE(-FO) products is crucial for accurately analyzing regional terrestrial water storage anomalies (TWSA). Here, we predict the water storage changes in the Yangtze River Basin during the GRACE(-FO) missions gap by using a geographically weighted-based reconstruction method (GW-based) and compare its accuracy with the results obtained from Region-based and Grid-based methods. The results reveal that the GW-based method achieves an average root mean square error (RMSE) of 12.24 mm and a correlation coefficient (CC) of 0.94 across the 13 test periods. These results outperform those from the Region-based and Grid-based methods, underscoring the superior accuracy of the GW-based method. Specifically, the spatial distribution of the GW-based results is more reasonable than that of the Grid-based results, and the CC and RMSE of the GW-based results are improved by 25.15% and 33.53%, respectively. The comparison with soil water storage changes derived from the GLDAS exhibited that the accuracy of the GW-based results is consistent across both the gap and test periods. Furthermore, the GW-based results strongly agree with the three published reconstruction results. Except for the lower reaches of the Yangtze River, where the CC is comparatively lower, those for the remaining regions exceed 0.938. Hence, the preceding analyses indicate the exceptional reliability of the GW-based method for predicting the water storage change within the Yangtze River Basin during the gap period.