Artificial intelligence has been extensively used to predict surface water quality to assess the health of aquatic ecosystems proactively. However, water quality prediction in data-scarce conditions is a challenge, especially with heterogeneous data from monitoring sites that lack similarity in water quality, hindering the information transfer. A deep learning model is proposed that utilizes representation learning to capture knowledge from source river basins during the pre-training stage, and incorporates meteorological data to accurately predict water quality. This model is successfully implemented and validated using data from 149 monitoring sites across inland China. The results show that the model has outstanding prediction accuracy across all sites, with a mean Nash-Sutcliffe efficiency of 0.80, and has a significant advantage in multi-indicator prediction. The model maintains its excellent performance even when trained with only half of the data. This can be attributed to the representation learning used in the pre-training stage, which enables extensive and accurate prediction under data-scarce conditions. The developed model holds significant potential for cross-basin water quality prediction, which could substantially advance the development of water environment system management.

Hydraulic models have been widely used to facilitate the design, management, and operation of water distribution systems (WDS). However, model parameters including pipe roughness coefficients (PRCs) and nodal demands (NDs) must be calibrated before practical applications, and the calibration accuracy and efficiency is important to enable their wide uptake in practice. This therefore motivates intensive studies to develop approaches for model calibration focusing on either optimization algorithms or numeric methods. Although these previous studies have advanced this research area, the calibration efficiency and effectiveness need to be further improved, especially when dealing with real-world problems. To attain this objective, this paper proposes an efficient numerical approach to simultaneously calibrate PRCs and NDs for WDS. Within this method, the calibration objective function is formulated as a least-square equation, followed by the derivation of iterative formulas according to a local sensitivity analysis. These formulas allow PRCs and NDs to be iteratively updated in an efficient manner until the given convergence criteria are met. To further improve the efficiency, the pipes and nodes are grouped according to their physical properties, which allows the calibration dimensions to be significantly reduced. Additionally, within the calibration process, values of PRCs and NDs are restricted into specified ranges to ensure the calibrated results are practically meaningful. Two case studies are used to demonstrate the utility of the proposed method, and results show that this approach is able to simultaneously calibrate PRCs and NDs efficiently, with the modeled hydraulic parameter values (pipe flows and nodal pressures) matching well with observations.

Leakage or pipe burst detection, often carried out by using the pressure sensor systems (PSSs) within a water distribution network (WDN), is critical to enable such networks to operate in a safe manner. The majority of previous studies have focused on either the advancement of detection equipment or the development of detection algorithms (leakage localization or sensor deployment). In contrast, this paper proposes a methodology to investigate the underlying capacity of existing PSSs for pipe burst detection using a set of quantitative metrics. These metrics focus on the identification of (1) nodes where bursts cannot be detected by the PSS (undetectable nodes); (2) undetectable demands, derived from the demands of undetectable nodes; (3) detection dimension, representing the number of sensors that can simultaneously detect a burst at a node; (4) the spatial region where bursts within each subregion can be detected by a particular sensor with a minimum outflow; and (5) the detectable threshold, which indicates the minimum burst outflow of each node triggering its associated pressure sensor. The proposed methodology has been applied to two real-world WDNs with varying sizes and configurations. Results show that the former two metrics can indicate the spatial distribution of the nodes at which bursts can be detected, and the latter three metrics can successfully reveal the PSS’s detection dimension and spatial regions, as well as the detectable threshold of each node. Such improved understanding offers guidance to develop effective burst-management strategies and to facilitate decision-making processes regarding sensor placement and burst localization.

Erosion of a saturated sand bed by a vertically upward water jet was studied experimentally. Direct observations on the behavior of sand-bed erosion were obtained by increasing the flow rate at a small increment until fluidization developed to the surface of the sand layer. With the increase in flow rate, the sand bed underwent different stages: no cavity, a stable cavity above the jet, an unstable cavity propagating upward, and full fluidization. The mechanism for the cavity evolution was investigated, and an analytical model was developed to predict the critical flow rate for full fluidization. The model predictions compared well with the measurements in the experiments with different sand-layer thicknesses.

Leak detection is nowadays an important task for water utilities as leakages in water distribution systems (WDS) increase economic costs significantly and create water resource shortages. Monitoring data such as pressure and flow rate of WDS fluctuate with time. Diagnosis based on time series monitoring data is thought to be more convincing than one-time point data. In this paper, a threshold selection method for the correlation coefficient based on time series data is proposed based on leak scenario falsification, to explore the advantages of data interpretation based on time series for leak detection. The approach utilizes temporal varying correlation between data from multiple pressure sensors, updates the threshold values over time, and scans multiple times for a scanning time window. The effect of scanning time window length on threshold selection is also tested. The performance of the proposed method is tested on a real, full-scale water distribution network using synthetic data, considering the uncertainty of demand and leak flow rates, sensor noise, and so forth. The case study shows that the scanning time window length of 3–6 achieves better performance; the potential of the method for leak detection performance improvement is confirmed, though affected by many factors such as modeling and measurement uncertainties.