Journal of Water Resources Planning and Management

Sediment delivery to reservoirs causes a reduction in storage capacity, leading to emergency water conditions and difficulties in managing water resources and dams, especially on dry land. This paper presents a methodology for calculating the time period of sedimentation-induced operational failures of a water supply dam, based on the reservoir’s daily water balance equation. The method applies a shot noise stochastic model to generate daily streamflow data calibrated with historical data, and accounts for sedimentation rates and the water volumes supplied to users. A Monte Carlo simulation was employed to generate different combinations of inflow and water supply release scenarios. The proposed methodology, which is completely generalizable, was applied in a case study of the Camastra reservoir in southern Italy. The results were in good agreement with the reservoir’s historical operational data.

The occurrence of bursts in water pipelines can not only prevent the system from functioning properly, but it can also produce significant water loss that disrupts activities in urban areas. Therefore, the detection and location of bursts in water distribution systems is a vital task for water utilities. Various techniques currently exist to detect the occurrence of these events, but there is a need for a permanent monitoring method that can detect and identify anomalous events quickly and accurately. This paper presents a new technique that uses artificial neural networks (ANNs) to detect and identify bursts in pipelines by interpreting the transient pressure waves that a burst causes along pipelines. The technique is divided into two stages: a model development stage and an application stage. The model development stage includes the generation of transient pressure traces and the training and testing of two different ANNs to (1) detect burst occurrence and (2) identify burst location and size. The application stage includes the processing of a potentially continuous transient pressure trace, analysis by the previously trained ANNs, and then the verification of the results using a transient flow forward numerical model. A numerical application demonstrates the principles of the technique and the potential for merging the use of fluid transient waves and ANNs. The technique has also been validated in the laboratory, indicating that the prediction of the location of the burst is very accurate while the prediction of the burst size requires an additional step to ensure its accuracy.

Both hydrodynamic factors and social factors have large impacts on the loss of life caused by dam failure. Relatively large uncertainty intervals of the influencing factors lead to changes in the potential loss of life. Based on an analysis of the formation mechanism of loss of life, the influencing factors were identified. Combined with interval theory, a method for calculating loss of life and determining the impacts of the influencing factors on loss of life was proposed. The intervals of the exposure rate of the population at risk and the mortality of the exposed population, which are impacted by the major influencing factors such as the flood severity, warning time, understanding of dam failure, and building vulnerability, were recommended. Furthermore, a range of correction coefficients caused by the minor influencing factors, such as the dam failure time, rescue ability, and age distribution, was analyzed. The proposed method was validated by analyzing the losses of life in 21 flooded regions after 10 dam failure events and 2 flash river floods, in which the intervals of the estimated results all contained the actual loss of life. In addition, the ratios of the upper bounds to the corresponding lower bounds of the intervals were all less than 10, which is in accordance with the characteristic that the results of different existing methods vary within an order of magnitude. This is the first work that pays careful attention to the uncertainty intervals of loss of life estimates, and the proposed method effectively determined the severity of the potential loss of life caused by dam failure.

Despite recent efforts to apply machine learning (ML) for water demand modeling, overcoming the black-box nature of these techniques to extract practical information remains a challenge, especially in developing countries. This study integrated random forest (RF), self-organizing map (SOM), and artificial neural network (ANN) techniques to assess water demand patterns and to develop a predictive model for the city of Fortaleza, Brazil. We performed the analysis at two spatial scales, with different level of information: census tracts (CTs) at the fine scale, and census blocks (CBs) at the coarse scale. At the CB scale, demand was modeled with socioeconomic, demographic, and household characteristics. The RF technique was applied to rank these variables, and the most relevant were used to cluster census blocks with SOMs. RFs and ANNs were used in an iterative approach to define the input variables for the predictive model with minimum redundancy. At the CT scale, demand was modeled using HDI and per capita income. Variables which assess the education level and economic aspects of households demonstrated a direct relationship with water demand. The analysis at the coarse scale provided more insight into the relationship between the variables; however, the predictive model performed better at the fine scale. This study demonstrates how data-driven models can be helpful for water management, especially in environments with strong socioeconomic inequalities, where urban planning decisions should be integrated and inclusive.