Under the influence of climate change and human activities, the intensification of salinity intrusion in the Modaomen (MDM) estuary poses a significant threat to the water supply security of the Greater Bay Area of Guangdong, Hong Kong, and Macao. Based on the daily exceedance time data from six stations in the MDM waterway for the years 2016-2020, this study conducted Empirical Orthogonal Function (EOF) and decision tree analyses with runoff, maximum tidal range, and wind. It investigated the variation characteristics and key factors influencing salinity intrusion. Additionally, Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN) were employed to predict the severity of salinity intrusion. The results indicated that: (1) the first mode (PC1) obtained from EOF decomposition explained 89% of the variation in daily chlorine exceedance time, effectively reflecting the temporal changes in salinity intrusion; (2) the largest contributor to salinity intrusion was runoff (40%), followed by maximum tidal range, wind speed, and wind direction, contributing 25%, 20%, and 15%, respectively. Salinity intrusion lagged behind runoff by 1-day, tidal range by 3 days, and wind by 2 days; North Pacific Index (NPI) has the strongest positive correlation with saltwater intrusion among the 9 atmospheric circulation factors. (3) LSTM achieved the highest accuracy with an R² of 0.89 for a horizon of 1 day. For horizons of 2 days and 3 days, CNN exhibited the highest accuracy with R² values of 0.73 and 0.68, respectively. This study provides theoretical support for basin scheduling and salinity intrusion prediction and serves as a reference for ensuring water supply security in coastal areas.

Storm surges, a significant coastal hazard, cause substantial damage to both property and lives. Precise and efficient storm surge models are crucial for long-term risk assessment and guiding emergency management decisions. While high-fidelity dynamic models offer accurate predictions, their computational costs are substantial. Hence, recent efforts focus on developing data-driven storm surge surrogate models. This study focuses on the Pearl River Estuary in Guangdong Province. Initially, the dynamic ADvanced CIRCulation (ADCIRC) model was utilized to construct storm surge data for 16 historical typhoons, serving as training, validation, and testing data for data-driven models. Subsequently, Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Informer deep learning (DL) models were employed for forecasting of storm surge over the next 1h, 3h, 6h, 12h, and 18h. Finally, Shapley Additive exPlanations (SHAP) values were used for interpretability analysis of the input factors across different models. Results indicated that the proposed DL storm surge prediction model can effectively replicate the dynamic model’s simulation results in short-term forecasts, significantly reducing computational costs. This model offers valuable scientific assistance for future coastal storm surge forecasts in the Greater Bay Area.

The Chushandian Reservoir, a key control project on the Huaihe River, is vital for flood control, water allocation, and maintaining ecological baseflow. This study analyzes runoff evolution and provides predictive insights for sustainable water management. Methods employed include Extremum Symmetric Mode Decomposition (ESMD) for decomposing complex signals, a mutation detection algorithm to identify significant changes in time-series data, and cross-wavelet transform to examine correlations and phase relationships between time series across frequencies. Additionally, the hybrid models GM-BP and CNN-LSTM were used for runoff forecasting. Results show cyclical fluctuations in annual runoff every 2.3, 5.3, and 14.5 years, with a significant decrease observed in 2010. Among climate factors, the Atlantic Multidecadal Oscillation (AMO) had the strongest correlation with runoff variability, while ENSO and PDO showed more localized impacts. Model evaluations indicated strong predictive performance, with Nash–Sutcliffe Efficiency (NSE) scores of 0.884 for GM-BP and 0.909 for CNN-LSTM. These findings clarify the climatic drivers of runoff variability and provide valuable tools for water resource management at the Chushandian Reservoir under future climate uncertainties.

Lakes, as key freshwater reserves and ecosystem cores, supply human water, regulate climate, sustain biodiversity, and are vital for global ecological balance and human sustainability. Lake Chaohu, as a crucial ecological barrier in the middle and lower reaches of the Yangtze River, faces significant environmental challenges to regional sustainable development due to water quality deterioration and consequent eutrophication issues. To address the limitations of conventional monitoring techniques, including insufficient spatiotemporal coverage and high operational costs in lake water quality assessment, this study proposes an enhanced Informer model optimized by the Whale Optimization Algorithm (WOA) for predictive analysis of concentration trends of key water quality parameters—dissolved oxygen (DO), permanganate index (CODMn), total phosphorus (TP), and total nitrogen (TN)—across multiple time horizons (4 h, 12 h, 24 h, 48 h, and 72 h). The results demonstrate that the WOA-optimized Informer model (WOA-Informer) significantly improves long-term water quality prediction performance. Comparative evaluation shows that the WOA-Informer model achieves average reductions of 9.45%, 8.76%, 7.79%, 8.54%, and 11.80% in RMSE metrics for 4 h, 12 h, 24 h, 48 h, and 72 h prediction windows, respectively, along with average improvements of 3.80%, 5.99%, 11.23%, 17.37%, and 23.26% in R² values. The performance advantages become increasingly pronounced with extended prediction durations, conclusively validating the model’s superior capability in mitigating error accumulation effects and enhancing long-term prediction stability. Spatial visualization through Kriging interpolation confirms strong consistency between predicted and measured values for all parameters (DO, CODMn, TP, and TN) across all time horizons, both in concentration levels and spatial distribution patterns, thereby verifying the accuracy and reliability of the WOA-Informer model. This study successfully enhances water quality prediction precision through model optimization, providing robust technical support for water environment management and decision-making processes.

Inter-basin water transfer projects are core infrastructure for achieving sustainable water resource allocation and addressing regional water scarcity, and pumping station units, as their critical energy-consuming and operation-controlling components, are vital to the projects’ sustainable performance. With the growing complexity and scale of these projects, pumping station units have become more intricate, leading to a gradual rise in failure rates. However, existing fault diagnosis methods are relatively backward, failing to promptly detect potential faults—this not only threatens operational safety but also undermines sustainable development goals: equipment failures cause excessive energy consumption (violating energy efficiency requirements for sustainability), unplanned downtime disrupts stable water supply (impairing reliable water resource access), and even leads to water waste or environmental risks. To address this sustainability-oriented challenge, this paper focuses on the fault characteristics of pumping station units and proposes a comprehensive and accurate fault diagnosis model, aiming to enhance the sustainability of water transfer projects through technical optimization. The model utilizes advanced algorithms and data processing technologies to accurately identify fault types, thereby laying a technical foundation for the low-energy, reliable, and sustainable operation of pumping stations. Firstly, continuous wavelet transform (CWT) converts one-dimensional time-domain signals into two-dimensional time-frequency graphs, visually displaying dynamic signal characteristics to capture early fault features that may cause energy waste. Next, the multi-head attention mechanism (MHA) segments the time-frequency graphs and correlates feature-location information via independent self-attention layers, accurately capturing the temporal correlation of fault evolution—this enables early fault warning to avoid prolonged inefficient operation and energy loss. Finally, the improved convolutional neural network (CNN) layer integrates feature information and temporal correlation, outputting predefined fault probabilities for accurate fault determination. Experimental results show the model effectively solves the difficulty of feature extraction in pumping station fault diagnosis, considers fault evolution timeliness, and significantly improves prediction accuracy and anti-noise performance. Comparative experiments with three existing methods verify its superiority. Critically, this model strengthens sustainability in three key ways: (1) early fault detection reduces unplanned downtime, ensuring stable water supply (a core sustainable water resource goal); (2) accurate fault localization cuts unnecessary maintenance energy consumption, aligning with energy-saving requirements; (3) reduced equipment failure risks minimize water waste and environmental impacts. Thus, it not only provides a new method for pumping station fault diagnosis but also offers technical support for the sustainable operation of water conservancy infrastructure, contributing to global sustainable development goals (SDGs) related to water and energy.

To address the problems of noise sensitivity, insufficient modeling of long-term time-series dependence, and high cost of labeled data in the fault diagnosis of pumping station units, an intelligent diagnosis method integrating the improved Sparrow Search Algorithm (tSSA) and Informer model is proposed in this study. Firstly, an adaptive t-distribution strategy is introduced into the Sparrow Search Algorithm to dynamically adjust the degree of freedom parameters of the mutation operator, balance global search and local development capabilities, avoid the algorithm converging to the origin, and enhance optimization accuracy, with time complexity consistent with the original SSA. Secondly, by combining the sparse self-attention and self-attention distillation mechanisms of Informer, the model’s ability to extract key features of long sequences is optimized, and its hyperparameters are adaptively optimized via tSSA. Experiments were conducted based on 12 types of fault vibration data acquired from pumping station units. Outliers were removed using the interquartile range (IQR) method, and dimensionality reduction was achieved through kernel principal component analysis (KPCA). The results indicate that the average diagnostic accuracy of tSSA-Informer under noise-free conditions reaches 98.73%, which is significantly higher than that of models such as SSA-Informer and GA-Informer; under noise interference of SNR = −1 dB, it still maintains an accuracy of 87.47%, outperforming comparative methods like 1D-DCTN; when the labeled sample size is reduced to 10%, its accuracy is 61.32%, which is more than 40% higher than that of traditional models. These results verify the robustness and practicality of the proposed method in strong-noise and small-sample scenarios. This study provides an efficient solution for the intelligent fault diagnosis of complex industrial equipment.