Journal of Water Resources Planning and Management

A growing population and urbanization place increased demands on water supply and distribution networks. Pipelines are one of the most critical components of water supply systems. It is, therefore, necessary to identify the relevant factors that affect the deterioration of water distribution pipelines. This will help decision makers in future planning and prioritization of the required maintenance. In this study, a systematic review is performed to identify critical factors that affect the failure of water pipelines. A meta-analysis is conducted to determine the relative importance of each factor that contributes to pipe failure. In addition, the source of contradictory results across studies is investigated. The results show that climatic factors, such as air temperature, minimum antecedent precipitation index, and net evaporation, contribute to water pipe failure. Additionally, the results of subgroup meta-analyses show that primary sources, such as pipe material and water pipe size, can lead to high heterogeneity across studies. This study is expected to help water utility owners to collect relevant data and make timely renewal decisions.

The present study proposes a knowledge-based system for optimizing the operation of multipurpose reservoirs in which the machine learning models, fuzzy inference systems, and particle swarm optimization are linked. The purposes were defined based on the key responsibilities of a reservoir and environmental requirements including (1) maximizing water supply, (2) maximizing hydropower production, (3) minimizing flood damage, and (4) mitigating the downstream environmental impacts, consisting of water quality and water quantity impacts on the aquatic habitats. The total loss was assessed by a fuzzy inference system in which flood damage, environmental loss, water supply loss, and hydropower production loss were the inputs of the system. Moreover, the fuzzy inference system was applied to assess the environmental loss of the reservoir. Two machine learning models were utilized to simulate water temperature and dissolved oxygen concentration. Moreover, the results of the knowledge-based system were compared with a conventional multiobjective optimization of reservoir operation in the case study. According to the results, the developed simulation-optimization method is able to optimize the release from the reservoir based on the defined purposes. However, it is not able to maximize benefits and mitigate environmental impacts perfectly. The proposed method can reduce the possible flood damage by more than 70%. The total loss is 60%, which corroborates that the performance of the proposed method is acceptable for balancing the conflicts of interest in the reservoir operation. Furthermore, results indicated that the knowledge-based system is able to reduce environmental impacts compared with the conventional multiobjective model in the case study.

An adequately calibrated hydraulic model is critically to the water distribution digital twin, which requires accurate representation of a water distribution network (WDN) in near real-time. It is thus imperative to construct an extended-period simulation model that captures the network’s baseline normal and abnormal diurnal demand patterns, which are usually derived using monitored flow data. However, from a practical field aspect, it remains challenging to calibrate multiple demand groups of varying diurnal patterns in large-scale WDNs, as flow data are typically collected sparsely due to cost considerations, while pressure sensors are commonly deployed throughout the network. In this paper, we propose a new pressure-based demand aggregation and pattern calibration method that leverages on monitoring pressure data to aggregate demands, identifying abnormal consumptions, and calibrating the diurnal patterns of various demand groups. The new method is integrated with previously developed model calibration framework and applied to a large-scale WDN system having more than 330 km of underground water pipelines with weekly averaged pressure and flow data, as derived from a maximum historical period of nine months. Key findings from our case study analysis for the seven averaged days (Monday to Sunday) include: (1) calibrating the system’s flow balance to within 99% average accuracy by identifying and calibrating five unique demand patterns, inclusive of those associated with abnormal consumptions, via grouping 34 available pressure sensors; (2) calibrating the system’s energy balance to within 95% average accuracy by iterating the simulated pressures against representative monitored pressure profiles of the different demand groups during the flow calibration process; and (3) achieving an average 2.5% accuracy improvement for the overall energy calibration, relative to that of the previous calibration approach. Throughout the solution process, significant engineering judgment is adopted, coupled with optimization analyses, to calibrate the system’s flow and energy balances while meeting the model constraints and data availability.

Abstract

Discharges from combined sewer overflows (CSO) are unacceptable, particularly when they are not linked to wet weather. This paper presents an evaluation of an online artificial-intelligence-based analytics system to give early warning of such overflows due to system degradation. It integrates a cloud-based data-driven system using artificial neural networks and fuzzy logic with near real-time communications, taking advantage of the increasingly available real-time monitoring of water depths in CSO chambers. The data-driven system has been developed to be applicable to the vast majority of CSO and requiring a minimum period of data for training. Results are presented for a live assessment of 50 CSO assets over a six-month period, demonstrating continuous assessment of performance and reduction of CSO discharges. The system achieved a high true positive rate (86.7% on confirmed positives) and low false positive rate (3.4%). Such early warnings of CSO performance degradation are vital to proactively manage our aging water infrastructure and to achieve acceptable environmental, regulatory, and reputational performance. The system enables improved performance from legacy infrastructure without gross capital investment.

Practical Applications

Combined sewerage networks convey wastewater from residential and commercial properties as well as rainfall runoff from urban catchments. The CSO provides a relief valve when runoff from rainfall would overwhelm the downstream network and treatment works. Excess water is spilled into a nearby watercourse, ideally when the watercourse flow has increased to provide additional dilution and thus minimize impacts. If a blockage or other defect downstream of a CSO results in a decrease in discharge capacity, the CSO can spill earlier than it is designed to or even in dry weather. Prior to the deployment of level sensors, such premature spills could only be identified through a visible spill or water quality impact. Sensors allow water utilities to monitor depths in CSO chambers; however, each utility will have a large number of CSO, which means that an automated system is needed to identify premature spills. This paper discusses the development and validation results obtained from a pilot deployment of a data analytics solution to identify abnormal water depths in a CSO.

Abstract

Floods have claimed the lives of countless people and caused significant property damage in many countries, putting their livelihoods in the jeopardy. The Vembanad lake system (VLS) in Kerala, India, has faced adverse mishappening during 2018, 2019, and 2021 floods in the state due to torrential rainfall. The goal of this research is to construct effective decision tree–based machine learning models such as adaptive boosting (AdaBoost), random forest (RF), gradient boosting machines (GBMs), and extreme gradient boosting (XGBoost) for integrating data, processing, and generating flood susceptibility maps. There are 18 conditioning parameters considered, which include seven categories and 11 numerical data. These seven categorical data were converted to numerical data, bringing the total amount of input data to 61. The recursive feature elimination (RFE) was utilized as the feature selection technique, and a total of 22 layers were chosen to feed into the machine learning models to generate the flood susceptibility maps. The efficiencies of the models were evaluated using receiver operating characteristic (ROC)–area under the ROC curve (AUC), F1 score, accuracy, and kappa. According to the results, the performance of all four models demonstrated their practical application; however, XGBoost fared well in terms of the model’s metrics. For the testing data set, the ROC-AUC values of XGBoost, GBM, and AdaBoost are 0.90, whereas it was 0.89 for RF. The accuracy varied significantly among the four models, with XGBoost scoring 0.92, followed by GBM (0.88), RF (0.87), and AdaBoost (0.87). As a result, this map may be utilized for early mitigation actions during future floods, as well as for land-use planners and emergency managers, assisting in the reduction of flood risk in regions prone to this hazard.

Practical Applications

The parameters used in the current methodology can be used to develop a flood susceptibility model for rainfall-induced flooding. The machine learning models used are some of the most widely accepted models, and they perform well in terms of accuracy and model reliability. This study will be extremely useful to the government and nongovernmental organizations in terms of risk assessment and mitigation during times of crisis, as well as early people rescue and worthwhile preparation. The focus of risk evaluation and prevention efforts is no longer on controlling floods, but rather on local governments’ obligations to lessen flood impacts. Residents in flood-prone areas should be warned about the hazards and possibilities. Land-use planners and government entities are obligated to inform local communities about the most recent flood susceptibility evaluations and the rules prohibiting new projects in areas with a high risk of flooding. Using maps of flood-prone areas and the severity of the damage, a possible rescue route during difficult times can be planned.

Coastal regions are witnessing unsustainable aquifer overdrafts, leading to subsidence and increased flooding. Flooding introduces its own assortment of problems because managing substantial volumes of stormwater is often infeasible through surface storage in low-gradient areas. Aquifer recharge operations, such as aquifer storage and recovery (ASR), in conjunction with small-scale surface reservoirs, has the potential to supplement the function of flood retention basins and provide additional flood relief to local communities. The purpose of this study is to test the effectiveness of surface water capture and injection at the local watershed level in mitigating flood impacts. The study is based on using a hydrologic-hydraulic numerical modeling analysis for a pilot watershed that was found to be suitable for aquifer recharge operations in the coastal region of south central Louisiana in the US. The concept of capturing excess surface water and injecting into the subsurface was found most effective during storms with multiple flood peaks and was able to add more than 75% reduction in flood peaks compared with using a retention basin alone. A key finding is that diverting surface flows into the subsurface can be effective only when implemented in conjunction with intermittent storage. The results also showed that the benefit of such operations is realized in the case of storms that have multiple flood peaks. The results have implications on the use of subsurface injections via techniques such as ASR for flood mitigation, in addition to their original purpose for groundwater management and sustainability.

Policy and behavioral responses to the COVID-19 pandemic led to sizeable and abrupt changes in water use. This study investigates the financial implications of these changes for water providers using annual data from public water systems in Wisconsin. Fixed-effects panel regressions were estimated to evaluate how COVID-19 affected water sales, total revenue, operating cost, and the revenue-to-cost ratio. Results indicate that COVID-19 led to increased residential water sales and decreased nonresidential water sales that resulted, on average, in higher revenue, higher costs, and lower net operating income. These effects varied markedly by system size and proportion of residential to total customers, although these differences were not statistically significant. Findings further suggest that large and moderate-sized water providers were relatively resilient to the effects of COVID-19, while small providers ($<1,000$ customer accounts) experienced the largest declines in net operating income.

Streamflow forecasts play an important role in water resources operation and management, and skillful seasonal forecasts can significantly facilitate the decision-making process. Streamflow variability is often associated with various large-scale, slowly evolving climate phenomena (e.g., El Nino Southern Oscillation), promoting the value of climate indices for streamflow forecast development, and has been investigated extensively. Separately, global climate models (GCMs), which provide climate forecasts out to 12 months globally, have been demonstrated to enhance seasonal streamflow predictability. However, neither the combination nor the interaction of these two sources of predictability has been given much attention. In this work, we propose a framework that can simultaneously account for antecedent climate indices and GCM forecasts in statistical streamflow forecasts and address how their interactions affect predictability. More specifically, we build a streamflow forecast framework combining statistical forecast models conditioned on climate indices with dynamical North American Multi-Model Ensemble (NMME) precipitation forecasts to generate streamflow forecast ensembles, and merge ensemble members with a Bayesian model averaging with bootstrap aggregating (BMA-bagging) approach. The framework is applied to streamflow on the Blue Nile River upstream of the Grand Ethiopian Renaissance Dam (GERD). Most NMME models tend to improve one-month-ahead GERD inflow forecasts; however, longer lead times prove challenging, and require specific NMME model selection. In contrast, although the value of climate indices varies with forecast lead time, their potential contribution grows at multimonth leads. The GERD inflow forecasts including NMME models and climate indices can simultaneously take advantage of both sources of predictability and prove superior across lead times as compared to including either source individually.

The combination of rapid urbanization, population growth, and the hydric stress due to climate change effects demand innovative, optimized approaches to the operation and supervision of urban water services. In Chile, the Superintendency of Sanitary Services has underscored the need for automatized data integration and analysis tools that foster an evidence-based, preventive approach to the supervision of urban water systems. This motivates the development of a pilot supervision and early-warning system, conceived as a cybernetic entity whose objective is to enable efficient access, analysis, and predictive modeling of the data provided by water service companies, so as to identify risks and inefficiencies in water services, monitor their evolution, and anticipate possible failures. This article discusses the development and implementation of a prototype system that provides tools for visualization, statistical and temporal analysis of georeferenced data on water pressures, network disruptions, and client complaints and deploys machine learning capabilities for predicting the quality of service indicators at different locations. The initial implementation in one region of Chile has been shown to expedite the exploitation of data on urban water services, reduce time lags in the detection of service disruptions, and generate evidence for the planning and execution of supervision activities. Based on this successful pilot, a roadmap for geographical and technological expansion is formulated, including other technological, organizational, and regulatory gaps that must be addressed to establish a data-driven framework for the supervision of urban water services.

Existing nonlinear time series methods are mostly used to capture the conditional variance or volatility of hydrometeorological factors. Insufficient attention has been paid to the time-varying variance-covariance and dynamic conditional correlation among multivariate hydrometeorological time series. This paper introduces a type of multivariate generalized autoregressive conditional heteroscedasticity (MGARCH) model, named the dynamic conditional correlation (DCC) model, to characterize the time-varying relationship between rainfall (P), temperature (T), and streamflow (Q). For this purpose, the Yellow River basin is used as an example, and three association models are established, including the bivariate relationships of P-Q, T-Q, and the three-variable relationship of P&T-Q (rainfall and temperature jointly on streamflow). The results of the univariate generalized autoregressive conditional heteroscedasticity (GARCH) model show that the volatility of the temperature series is weaker than that of rainfall and streamflow. The conditional variance of the streamflow and temperature series has long-term persistence, while that of rainfall has short-term persistence, but the temperature has the lowest degree of variance memory. The bivariate DCC model shows that the streamflow is more sensitive to precipitation changes than temperature, and both rainfall-streamflow (P-Q) and temperature-streamflow (T-Q) associations have stronger long-term persistence and covariance memory. The P&T-Q correlation shows that the covariance of rainfall and temperature has stronger long-term persistence than streamflow fluctuations. In addition, the effectiveness of the P&T-Q relationship model is also stronger than that of the P-Q and T-Q relationship models. DCC, considering the combined effects of rainfall and temperature, can better represent the impact of environmental changes on streamflow than a single variable.

The unprecedented changes in human mobility in the United States during the novel coronavirus (COVID-19) pandemic provided a unique opportunity to relate changes in urban stream water quality to the intensity of human activity on six types of land use. This study evaluated the baseflow-dominated stream response of pH, total nitrogen (TN), and nitrate-nitrogen between five in-stream sampling sites located in four watersheds with different types and degrees of development in Delaware County, Pennsylvania before and during the COVID-19 pandemic (2018 to 2021). This investigation found that significant water quality changes coincided with dramatic declines in human mobility during the pandemic. Human mobility was quantified using county-level volunteer reports of the use of Google apps on electronic devices. A two-way analysis of variance identified significant changes in pH, TN, and nitrate-nitrogen at all four sampling locations during monthly baseflow-dominated stream flow sampling before and during the pandemic to the decline in human movement and activity. However, the limited pre-pandemic sampling design did not observe overall significant water quality trends for the interaction of water quality between locations before and during the pandemic. The lack of significant interaction could suggest that changes in human mobility due to COVID-19 restrictions impacted stream water quality across sites at least in a similar fashion. This study provided definitive evidence that cell phone tracking can identify some of the impact of urban nonpoint sources spatially and dynamically, highlighting the role of political and socio-economic factors on the intensity of human activities that might be reflected on stream water quality changes.