Groundwater resource sustainability faces significant challenges due to groundwater overdraft and waterlogging. Here we propose a novel framework for evaluating the sustainability of groundwater resources. The framework incorporates a dynamic calculation of the ecological groundwater depth (EGWD) at the grid scale, considering multiple protective targets. To quantitatively evaluate the groundwater sustainability, we utilize reliability, resilience, and vulnerability, to measure the frequency, duration, and extent of unsatisfactory conditions. We apply this framework to the lower part of Tao’er River Basin in China. During the non-growth period and growth period, the upper thresholds of the EGWD range from 1.16 to 2.05 meters and 1.16 to 4.05 meters, respectively. The lower thresholds range from 6.28 to 33.54 meters and 4.87 to 30.72 meters, respectively. Future climate change improves reliability performances in regions with deep groundwater depths. Although the precipitation infiltration increases in future scenarios, prolonged duration and enhanced intensity of extreme climate events lead to decreased resilience and vulnerability performances under climate change. The proportion of areas with resilience values less than 1/12 expands to 2~3 times that of the historical scenario. Furthermore, we observe that more areas face the dual challenges of groundwater depletion and waterlogging under future climate change, particularly in high-emission scenarios. This study enhances understanding of groundwater resource sustainability by considering the spatial-temporal distribution of the EGWD, climate change impacts, and the identification of key regions for management. The insights can inform the development of effective strategies for sustainable groundwater resource management.

Despite advances in hydrological Deep Learning (DL) models using Single Task Learning (STL), the intricate relationships among multiple hydrological components and model inputs might not be comprehensively encapsulated. This study employed a Long Short-Term Memory (LSTM) neural network and the CAMELS dataset to develop a Multi-Task Learning (MTL) model, predicting streamflow and evapotranspiration across multiple basins. An optimal multi-task loss weight ratio was determined manually during the validation phase for all 591 selected basins with streamflow data-gaps under 5%. During test period, MTL showed median Nash-Sutcliffe Efficiency predictions for streamflow and evapotranspiration at 0.69 and 0.92, consistent with two STL models. The MTL’s strength appeared when predicting the non-target variable, surface soil moisture, using probes derived from LSTM cell states—representative of the internal DL model workings. This prediction showed a median correlation coefficient of 0.90, surpassing the 0.88 and 0.89 achieved by the streamflow and evapotranspiration STL models, respectively. This outcome suggests that MTL models could reveal additional rules aligned with hydrological processes through the inherent correlations among multiple hydrological variables, thereby enhancing their reliability. We termed this as “variable synergy,” where MTL can simultaneously predict varied targets with comparable STL performance, augmented by its robust internal representation. Harnessing this, MTL promises enhanced predictions for high-cost observational variables and a comprehensive hydrological model.