Objectives Traditional hydrological models are difficult to accurately simulate hydrological processes, and the low spatial resolution of time-varying gravity field satellite data makes it difficult to accurately estimate changes in regional terrestrial water storage (TWS)，a continuously operating global navigation satellite system (GNSS) reference station network can monitor in real-time the vertical deformation of the elastic crust caused by changes in surface hydrological loads.

Methods This method combines the vertical displacement time series of GNSS reference stations with the sparrow search algorithm-variational mode decomposition (SSA-VMD) nonlinear signal extraction algorithm, which has high precision and spatiotemporal resolution, to estimate the spatiotemporal distribution of TWS in Yunnan region from 2011 to 2020. The estimated results are then compared and analyzed with time-varying gravity field data, global land data assimilation system (GLDAS) data, and precipitation data.

Results (1) Hydrological load is the primary factor causing seasonal variations in the GNSS vertical displacement time series, with a correlation coefficient above 0.5. The second factor is non-tidal atmospheric load, and the least influential is non-tidal oceanic load. In response to the complex noise present in the GNSS time series, the SSA-VMD method is employed to mitigate high-frequency noise, resulting in a cleaner processed series. (2) The TWS variation in Yunnan region, exhibits a gradual increase from northeast to southwest. In a very small number of areas, due to other factors (e.g., evapotranspiration and runoff), discrepancies with precipitation data may occur. From the inversion results, the annual amplitude of TWS obtained from GNSS is the largest, followed by gravity recovery and climate experiment (GRACE) / GRACE Follow-On (GRACE-FO), while that from GLDAS is the smallest. (3) Compared with precipitation data, the TWS variations retrieved from GNSS, GRACE/GRACE-FO, and GLDAS show a certain temporal lag. From 2011 to 2018, the TWS variations were relatively stable with an overall increasing trend. However, from 2019 to 2020, the equivalent water height (EWH) time series exhibited a significant downward trend, and the continuous decrease in water reserves led to a prolonged severe drought. By August 2020, water reserves had basically returned to normal.

Conclusions The spatial and temporal distribution of TWS obtained from GNSS inversion is generally consistent with GRACE/GRACE-FO and GLDAS data, showing a gradual decrease from southwest to northeast overall. However, the annual amplitude of TWS changes from GNSS is larger than that from the other two, mainly because GNSS is more sensitive to local TWS changes. The seasonal variations in GNSS-EWH, GRACE-EWH, and GLDAS-EWH are significant, and when analyzed with precipitation data, a certain lag is observed.