Abstract Climate change is profoundly altering hydrological processes in alpine regions, which are acutely sensitive to warming; however, accurate projections of hydrological fluxes, notably runoff, remain challenging. In this study, six hydrological conceptual models and their combinations were first evaluated through comparative experiments for daily runoff simulation. Downscaled CMIP6 climatic variables were then used to drive the selected model/combination for future runoff variations using the Datong River Basin on the northeast Tibetan Plateau, China as a case study. The results indicate that: (1) among the single and combined models, the IHACRES-CMD and GR4JSG combination has shown satisfactory performance for runoff simulation in alpine basins; (2) future projections of climate and hydrological variables based on the selected model driven by CMIP6 outputs reveal a consistent warming-wetting tendency, with mean annual temperature, precipitation and runoff increasing by 1.6°C–3.8°C, 25.5–55.8 mm, and 0.8 – 1.4 billion m 3 , respectively, under various Shared Socioeconomic Pathways (SSP) scenarios between 2015 and 2100 relative to the historical period; and (3) runoff changes also demonstrate significant future hydrological shifts, showing increased interannual runoff but a more uniform intra-annual distribution, manifested through reduced runoff in July but increase in June. In addition, results also show larger runoff variability with increased occurrence of above-normal runoff in the future. These changes imply an elevated flood-risk potential during early summer, which poses a significant challenge to existing reservoir operation rules, necessitating adaptive strategies and climate-resilient allocation frameworks for water resource systems in alpine basins.
Abstract Groundwater seepage is influenced by various anthropogenic factors during urban construction, among which subsurface structures and surface hardening are two key factors. This study employed numerical simulations to investigate the impacts of rainfall infiltration on groundwater seepage under the effect of urban construction. A numerical model was established to predict groundwater level (GWL) in the phreatic aquifer (Aq0), incorporating the actual distribution of impervious surfaces and subsurface structures for a representative urban area in Shanghai. The model accounted for land surface conditions through rainfall infiltration coefficients, while subsurface structures were parameterized using equivalent hydraulic conductivity ( K eq ) derived from effective medium theory. Validation results show that the model achieved a maximum error of only 1.7%, demonstrating reliable performance in predicting groundwater seepage. There is 23.23% area with GWL declining more than 0.4 m in the actual scenario with 93% impervious surface hardening comparing without surface hardening. Prediction scenarios with varying reduction rates of impervious surface compared with the actual scenario ( R a ) were conducted to investigate the impact of surface hardening on GWL. The results indicate average GWL of the study area presents exponential decay with the impervious surface ratio. Further, the variations in GWL of Aq0 were investigated by reducing hardened surfaces in different locations with the same R a . As the proportion of underground structures in the reduced impervious surface increases, the area with the same increment of GWL decreases. Prioritizing low impact development in areas with lower subsurface structure density can significantly enhance the efficiency of groundwater recharge by rainfall infiltration.
Abstract Monthly groundwater level records in monitoring networks frequently include gaps that complicate drought assessment, pumping management, and groundwater-model calibration. This paper presents GWMonthlyKNN, an open-source workflow that imputes missing monthly water-surface-elevation (WSE) records using k -nearest neighbors (KNN) in a feature space encoding seasonality and short-term persistence, while limiting candidate donor wells to hydrologically plausible groups. The workflow was evaluated on a California Department of Water Resources dataset of 515 wells using blocked 12- and 24-month masked-gap experiments that emulate monitoring outages. Full-network tests identified eligible 12- and 24-month outage blocks for 461 and 410 wells, respectively. Under the default configuration, median coefficients of determination were 0.83 and 0.77, median RMSE values were 0.97 and 0.85 m, and median signed biases were + 0.20 and + 0.15 m . Sensitivity runs across neighborhood sizes showed similar median performance. GWMonthlyKNN outputs completed station series, diagnostics, and a provenance map for auditable groundwater data completion.
Abstract Accurate groundwater level (GWL) prediction is crucial for water resource management. The nonlinear response of GWL to climate change and human activities makes daily scale prediction challenging. This study proposed three strategies to enhance daily scale GWL prediction accuracy by integrating physical knowledge into a long short-term memory (LSTM) model: incorporating hydrogeological parameters (LSTM-HP), considering precipitation recharge delay (LSTM-LAG), and embedding the relationship of GWL and precipitation (LSTM-MR-D). Results showed that LSTM-LAG achieved the most improvement, increasing the Nash–Sutcliffe efficiency coefficient (NSE) value by 0.01–0.42 during testing, followed by LSTM-MR-D and then LSTM-HP. However, combining strategies did not always enhance prediction accuracy. Approximately 50% of the observation wells that could have achieved accuracy improvements experienced a decrease instead. Additionally, hydrogeological parameters identified by LSTM can help calibrate a physics-based model, achieving satisfactory GWL prediction results with an NSE of 0.95. These findings offer practical solutions for improving GWL predictions and aiding groundwater management decision-making.
Abstract Non-Gaussian characteristics are common in geological formations formed in complex depositional environments with distinct facies boundaries and highly contrasting hydrological properties. These non-Gaussian features pose significant challenges for characterizing CO 2 behaviors during geological carbon sequestration (GCS) projects. Because traditional physics-based numerical simulators often demand extensive computational resources for uncertainty quantification, this study develops a spatiotemporal surrogate model to expedite the process. The surrogate model is constructed upon a spatiotemporal convolutional-recurrent architecture, serving as a mapping operator that relates static non-Gaussian geological characteristics (permeability and porosity), and variable injection well configurations to dynamic evolution of CO 2 pressure and saturation. Subsequently, the proposed surrogate model substitutes the numerical simulator to facilitate uncertainty quantification through the iterative ensemble smoother (IES) algorithm. Results indicated that the proposed surrogate model effectively handles non-Gaussian geological characteristics and variable injection well configurations. It accurately replicated the performance of the MATLAB Reservoir Simulation Toolbox (MRST) simulator, achieving a coefficient of determination ( R 2 ) above 0.96 and a Pearson correlation coefficient (CC) above 0.98, and improving computational efficiency by a factor of 3,145 in forward modeling. Furthermore, integrating the surrogate model with the IES approach significantly reduced geological uncertainty. This research highlights the potential of combining deep learning techniques with data assimilation models for efficient forward modeling and uncertainty quantification in GCS projects.
