Journal of Water Resources Planning and Management

This study utilizes a rich UK data set of smart demand metering data, household characteristics, and weather data to develop a demand forecasting methodology that combines the high accuracy of machine learning models with the interpretability of statistical methods. For this reason, a random forest model is used to predict daily demands 1 day ahead for groups of properties (mean of $3.8\mathrm{households}/\mathrm{group}$) with homogenous characteristics. A variety of interpretable machine learning techniques [variable permutation, accumulated local effects (ALE) plots, and individual conditional expectation (ICE) curves] are used to quantify the influence of these predictors (temporal, weather, and household characteristics) on water consumption. Results show that when past consumption data are available, they are the most important explanatory factor. However, when they are not, a combination of household and temporal characteristics can be used to produce a credible model with similar forecasting accuracy. Weather input has overall a mild to no effect on the model’s output, although this effect can become significant under certain conditions.

To link water and power systems on a regional scale, equivalent reservoir models—an aggregated representation of a multireservoir system—are commonly used because conventional river-basin scale optimization models become computationally expensive with increasing dimensionality. Although equivalent reservoir models are widely applied in power system operation, analyses comparing the performance of equivalent reservoir models with multireservoir cascade models are limited. To this end, this study systematically compares two equivalent reservoir models, an aggregated water balance and an energy balance representation, with a multireservoir cascade representation for a system of three reservoirs in series in Savannah, South Carolina, in terms of the total end-of-period release, hydropower and storage based on simulation, simulation optimization, and analytically over a 30-year period. Findings from the pilot basin are generalized by altering the storage-to-demand ratio (SDR) to understand the effect of different system characteristics on the equivalent reservoir representation under observed and predicted inflows of different skills. Equivalent reservoir models perform similarly to the cascade model for systems with large SDRs, but for systems with smaller SDRs, equivalent reservoir models perform poorly because spill and other losses from individual reservoirs cannot be effectively represented in the aggregated approach.

This study presents a heuristic multiobjective approach for segmenting and operating water distribution systems (WDS). The methodology employs a two-pronged strategy: the first is a heuristic method for dividing the network into clusters (i.e., district metering areas) based on connectivity analysis. The second is the application of the evolutionary multiobjective optimization method non-dominated sorting genetic algorithm (NSGA)-II for trading off the operational cost, excess pressure (serving as a proxy to leakage reduction), and water age (acting as a surrogate to water quality) in the WDS. Three example applications of increasing complexities with various cluster partitioning are explored, showing a clear trade-off among the objectives. This study introduces an unprecedented heuristic approach for jointly solving the multiobjective problem under a given system partitioning. However, by enforcing a priori clustering formation (rather than including it in the optimization), optimality, completeness, and precision are compromised in favor of computational speed and effort. Thus, additional sensitivities need to be conducted outside of the optimization for the clusters’ impact. Challenges of extending this study are in embedding the clusters’ formations in the optimization considering other objectives such as residual capacity, developments of other optimization frameworks outside of the generic link of simulation-optimization, and uncertainty inclusion (e.g., in demands). All data and codes are included for allowing full replications and comparisons.

Understanding residential water preferences allows utility and municipal water managers to improve residents’ welfare through informed pricing and investment decisions. This paper employs a discrete choice experiment (DCE) to elicit residents’ stated preferences for water sources, storage methods, and conservation policies across the inland Pacific Northwest. Generally, results suggest that—conditional on needing an additional municipal water supply—individuals have a strong aversion to creek water, and are relatively indifferent to reclaimed wastewater or additional diversions from large rivers. Residents also have a strong preference for additional local water storage, either in the form of aquifer injection or reservoirs. DCE results suggest strong heterogeneity in preferences correlated with gender and issue awareness, but few other seemingly relevant respondent characteristics are significant in explaining this heterogeneity. In particular, males are significantly more price-driven than females, while females more strongly oppose lawn watering restrictions and are more strongly favor local storage options. Respondents familiar with the water supply issues in the region favor using reclaimed water over all other available water sources.

Enhancing the performance of water distribution networks (WDNs) on a day-to-day basis or under extreme disturbances is an utmost priority for utilities. Previous research has characterized the structure of WDNs in the pipe–junction or segment–valve representation to gain insight on various aspects of their performance; however, the research on characterizing WDN structure in a hierarchical representation and its relationship with performance is lacking. Two key physical properties of WDNs are loops and pipe diameters that are organized in a hierarchical way. Novel indicators have been created to quantify the network hierarchy related to these key properties in other spatial flow distribution networks: loop nestedness and pipe diameter gradation along flow paths. The goal of this study is to adopt such indicators to characterize the hierarchy of WDNs and evaluate its relationship with WDN performance. This study applies a hierarchical decomposition process to model the relationships among loops as a tree network for quantifying loop nestedness. Flow paths of monotonically increasing and decreasing pipe diameters are traced to quantify pipe diameter gradation. Statistical distributions are approximated for these two indicators. Then, relationships between these network hierarchy indicators and two performance indicators (measuring path redundancy and power surplus) are identified. For 15 benchmark networks, this study finds the statistical distributions representing loop nestedness and pipe diameter gradation closely follow a power law. Results suggest that gradual pipe diameter gradation along flow paths and high loop nestedness increase WDN path redundancy, and gradual pipe diameter gradation increases WDN power surplus. The study demonstrates that the hierarchical analysis of WDNs can significantly supplement traditional topological analyses in explaining WDN performance.

This paper presents an assessment methodology that considers the impact of actual valve locations in a water distribution system in creating discrete isolated groups of pipes or segments when evaluating the performance of the network under a failure condition. The layout of the distribution network used as a case study and the location of the isolation valves are based on a survey of the real system instead of being artificially generated. In addition to evaluating the performance of the system under typical conditions with metrics based on loss of connectivity and the reduction in demand satisfaction, the assessment also includes a consideration for fire flow requirements that has not been widely used in combination with a segment-based reliability assessment.

Dams heavily impact the Lancang-Mekong River, leading to water-use conflicts for hydropower, food production, and ecosystem sustainment. Basin-scale cooperation has been shown to yield incremental benefits; however, understanding of the water–energy–food (WEF) nexus—critical for cooperation and benefit-sharing in transboundary basins—is insufficient. A WEF nexus model for the Lancang-Mekong River Basin (LMRB) was developed to investigate the impacts of value preferences from riparian countries, reservoir operation policies, and future dam construction. The analysis shows that the value preferences for hydropower and agriculture significantly influenced the WEF nexus under cooperation. If downstream countries preferred rice production (RP), there was a substantial reduction in hydropower generation (HP) to meet the irrigation demands of RP because reservoirs operate for basin-level optimization under cooperation. Ecological water demands in China and downstream countries are represented by environmental flow constraints and competed with potential HP and potential RP at the basin scale, whereas trade-off relationships among the WEF nexus depended on hydrology. HP and RP benefited from river damming in downstream countries, considering better dam regulation of streamflow. The complex relationships among the WEF nexus highlight the importance of flexible cooperation policies considering various sociohydro conditions in the LMRB; thus, this study’s results can offer real-world preferences.

Rainwater harvesting provides a potential source of supplemental water supply to meet increasing urban demand. Distributed solutions based on many small capture devices are often viewed as inefficient in terms of both cost and volume of water captured. This paper describes a framework for assessing the value of distributed rainwater capture as a substitute for a share of municipal water supplied to residential and commercial buildings. The analysis was applied to the Ballona Creek watershed in Los Angeles County, California, evaluating several cistern sizes and participation rates to estimate the amount and value of water, energy, and carbon savings offset by captured water. Results show that rainwater capture may be an efficient approach to augmenting water supply under certain conditions. From the perspective of the water supplier, capturing water for outdoor use is cost-effective, but including indoor use is not due to installation and maintenance costs. From the perspective of the homeowner, installing a cistern for irrigation would result in net cost savings over time, and under certain conditions, higher benefits are realized from larger cisterns.

This paper presents a practical approach to adaptive management of dam risk based on robust decision-making strategies coupled with estimation of climate scenario probabilities. The proposed methodology, called multi-prior weighted scenarios ranking, consists of a series of steps from risk estimation for current and future situations through definition of the consensus sequence of risk reduction measures to be implemented. This represents a supporting tool for dam owners and safety practitioners in making decisions for managing dams or prioritizing long-term investments using a cost-benefit approach. This methodology is applied to the case study of a Spanish dam under the effects of climate change. Several risk reduction measures are proposed and their impacts are analyzed. The application of the methodology allows for identifying the optimal sequence of implementation measures that overcomes uncertainty from the diversity of available climate scenarios by prioritizing measures that reduce future accumulated risks at lower costs. This work proves that such a methodology helps address uncertainty that arises from multiple climate scenarios while adopting a cost-benefit approach that optimizes economic resources in dam risk management.

Rainwater harvesting (RWH) through household rainwater tanks has been increasingly recognized as a supplementary source of water in most urban cities. Previous studies indicated wide variations in annual water saving on residential water consumption. This study aimed to examine the factors causing these variations on the calculated savings of residential water use. The study involved the analysis of water consumption data collected by Yarra Valley Water (YVW) for 50 households with rainwater tanks and 50 without in Greater Melbourne, Australia. The results initially indicated an average annual saving of more than 4,600 L per year for an average household of 3.1 people; however, after considering seasonal variations, the estimated average potable water saving was more than 9,700 L per year. Further analysis revealed that properties without garden watering received more water saving than properties with garden watering (34,400–5,600 L per year, respectively). The results of this study are envisioned to lead to a better understanding of the effect of seasonal variations and local site conditions on the potential of RWH in integrated urban water management.

Increasing the power generation and reducing the sediment deposition in a reservoir are two contradictory and incommensurable objectives. Hence, these two operational objectives rarely are considered simultaneously in reservoir optimization. This study proposes a method to balance sediment desilting and power generation in reservoirs, and to study their quantitative relationship. First, a formula to calculate the suspended sediment transport rate for reservoirs was established. Second, based on that formula, a multiobjective model was built that considers simultaneously, the sediment transporting efficiency and the power generation efficiency. The nondominated sorting genetic algorithm (NSGA-II) along with Pareto optimality was used to find the Pareto front of the model. The proposed method was applied to the Three Gorges project in China. The research results not only provide the Pareto front of the reservoir operation scheme, but also imply that the maximum annual sediment runoff was improved by 6% at the expense of 9% of the annual power generation.

It has been recently observed that traditional probability-based design of urban drainage systems often fails because of frequent heavy rainfall events due to global climate change. Therefore, it is critical to consider a system’s ability to prepare, react, and recover from a failure (i.e., resilience) in urban drainage network design. This study proposes a resilience-constrained optimal design model of urban drainage networks that minimizes total system cost while satisfying predefined levels of failure depth and duration. Failure (e.g., flooding) depth refers to the level of system performance degradation, whereas failure duration is the time taken for a system’s recovery to its normal state. Optimal layout and pipe sizes are identified by the proposed model comprising the harmony search algorithm for optimization and the storm water management model (SWMM) for dynamic hydrology–hydraulic simulations. The proposed model is demonstrated through the design of two grid networks and an A-city drainage network. The obtained resilience-based design is compared to the least-cost design obtained with no resilience consideration according to optimized layout and pipe sizes and the resulting topological characteristics.

In water distribution systems, premature decay of disinfectants inside water storage tanks can increase the costs of disinfectant input, depending on the mixing type that takes place in the tank. In this paper, the influence of the flow regime of a water storage tank upon the disinfection cost was assessed. Equations relating disinfectant cost to the disinfectant’s decay rate and the mean residence time are presented and compared for three flow regimes: ideal mixed flow, ideal plug flow, and disperse flow. As the decay rate or the mean residence time increased, the cost went to the order of 100 times for plug flow, when compared to mixed flow. The dispersive flow regime presented disinfection costs that were up to 10 times the cost for the mixed flow regime. Given these differences, some general guidelines concerning baffling, inlet characteristics, and water depth-to-diameter ratio were presented in order to obtain a flow closer to the plug regime and, consequently, diminish the chlorination costs. Finally, a case study was presented, based upon the guidelines, showing that a change in the inlet diameter or inlet orientation of a water storage tank can reduce the chlorination cost.