Search

Urban drainage models (UDMs) are often used to manage urban flooding. However, these models generally involve many parameters to represent the underlying complex hydrodynamic processes. This results in significant challenges to achieving effective and robust model calibration especially with frequently limited observations, leading to unreliable model predictions. This paper makes the first attempt at UDM calibration using the Bayesian‐based Ensemble Smoother (ES) method. Three ES variants are considered, that is, the primary ES, the versions with multiple data assimilation (ES‐MDA) and iterative local update (ES‐ILU). Two synthetic cases and one real‐world application with up to 5,236 calibration parameters are tested. Results obtained show that: (a) both ES‐MDA and ES‐ILU can produce effective model calibration with ES‐ILU outperforming ES‐MDA in terms of both accuracy and uncertainty while ES exhibits limited performance; (b) for the real‐world case, both the ES‐MDA and ES‐ILU methods provide better calibration results than the best‐known solution manually obtained, (c) a minimum number of observations are required to enable an overall accurate model calibration (e.g., four and ten more monitoring sites are needed in the two synthetic cases); and (d) the model calibrated using an intense rainfall event is generally robust to make reliable predictions across different rainfall events while the model calibrated using less intense rainfall event does not perform well for more intense rainfall events. It was also found that ubiquitous parameter equifinality significantly hinders unique parameter identification even when overall accurate state estimates are obtained. This should be clearly understood in practical applications. Plain Language Summary: Urban floods have been a serious disaster worldwide. Urban drainage models (UDMs) have been widely used to facilitate flooding prevention and mitigation. However, a challenge associated with the UDMs is that a large number of parameters need to be specified and calibrated. While some optimization‐based or manual calibration methods are available, their efficiency or accuracy are often unsatisfactory. This can lead to unreliable predictions of the rainfall‐runoff process. To this end, this paper proposes the ensemble smoother (ES) methods to calibrate the UDMs. Benefits of these ES methods include the great efficiency, high accuracy and the identification of the uncertainty when calibrating model parameters. These conclusions are based on results of two synthetic cases and one real‐world UDM. Key Points: Ensemble smoothers are promisingly effective and robust methods to calibrate urban drainage modelsUbiquitous parameter equifinality hinders unique parameter identification, which is a concern for practical applicationsEnsemble smoothers are validated on a real‐world case, providing insights into how improved model performance can be achieved [ABSTRACT FROM AUTHOR]

Correctly calibrated water distribution network models are valuable assets for water utilities. Among possible uses of hydraulic modelling, the detection and location of leakage hotspots are important operational considerations, with companies often spending large sums of money finding leaks, but many remaining undetected. For a more reliable modelling and calibration process, water utilities need to ensure that asset state and status is accurate. The paper considers a new graph-theory based technique, called pipe tree analysis, for clustering water distribution networks. The aim is to reduce the calibration problem size for leakage hotspot detection and to establish a foundation for improved model quality assurance. The pipe tree topological analysis is applied to divide the “Anytown” network from literature, into different pipe trees and combined with model pre-processing, to reduce the solution search space. A Genetic Algorithm is, then, used to solve the optimization problem of searching for calibration parameters values, while minimizing the differences between observations and model predictions. The new modelling method highlighted important calibration parameters and contributed to successful detection of model anomalies, such as unknown closed valves and leakage hotspots, providing additional benefits to optimisation-based calibration. [ABSTRACT FROM AUTHOR]

Success in the application of any model-based methodology (e.g., design, control, supervision) highly depends on the availability of a well-calibrated model. The calibration of water distribution networks needs to be performed online due to the continuous evolution of demands. During the calibration process, background leakages or bursts can be unintentionally incorporated to the demand model and treated as a system evolution (change in demands). This work proposes a leak-detection and localization approach to be coupled with a calibration methodology that identifies geographically distributed parameters. The approach proposed consists in comparing the calibrated parameters with their historical values to assess if changes in these parameters are caused by a system evolution or by the effect of leakage. The geographical distribution allows unexpected behavior of the calibrated parameters (e.g., abrupt changes, trends, etc.) to be associated with a specific zone in the network. The performance of the methodology proposed is tested on a real water distribution network using synthetic data. Tested scenarios include leaks occurring at different locations and ranging from 2.5 to 13% of the total consumption. Leakage is represented as pressure-dependent demand simulated as emitter flows at the network nodes. Results show that even considering a low number of sensors, leaks with an effect on parameters higher than the parameters' uncertainty can be correctly detected and located within 200 m. [ABSTRACT FROM AUTHOR]

Estimating model parameters is a difficult, yet critical step in the use of water distribution system models. Most of the optimization-based approaches developed so far concentrate primarily on efficient and effective ways of obtaining optimal calibration parameter values. At the same time, very little effort has been made to determine the uncertainties (i.e., errors) associated with those values (and related model predictions). So far, this has typically been done using the first-order second moment (FOSM) method. Even though reasonably computationally efficient, the FOSM approach relies on several restrictive assumptions and requires computationally demanding calculation of derivatives. To overcome these limitations, the recently developed shuffled complex evolution metropolis (SCEM-UA) global optimization algorithm is linked to the Epanet2 hydraulic model and used to solve a least-squares-type calibration problem. The methodology is tested and verified on the Anytown literature case study. The main advantage of the SCEM-UA algorithm over existing approaches is that both calibration parameter values and associated uncertainties can be determined in a single optimization model run. In addition, no model linearity or parameter normality assumptions have to be made nor any derivatives calculated. The main drawback of the SCEM-UA methodology is that it could, potentially, be computationally demanding, although this is not envisaged as a major problem with current computers. [ABSTRACT FROM AUTHOR]

Sampling design (SD) for water distribution systems (WDS) is an important issue, previously addressed by various researchers and practitioners. Generally, SD has one of several purposes. The aim of the methodologies developed and presented here is to find the optimal set of network locations for pressure loggers, which will be used to collect data for the calibration of a WDS model. First, existing SD approaches for WDS are reviewed. Then SD is formulated as a multiobjective optimization problem. Two SD models are developed to solve this problem, both using genetic algorithms (GA) as search engines. The first model is based on a single-objective GA (SOGA) approach in which two objectives are combined into one using appropriate weights. The second model uses a multiobjective GA (MOGA) approach based on Pareto ranking. Both SD models are applied to two case studies (literature and real-life problems). The results show several advantages and one disadvantage of the MOGA model when compared to SOGA. A comparison of the MOGA SD model solution to the results of several published SD models shows that the Pareto optimal front obtained using MOGA acts as an envelope to the Pareto fronts obtained using previously published SD models. [ABSTRACT FROM AUTHOR]

Sampling design (SD) for water distribution systems (WDS) is undoubtedly an important issue, and has been addressed in the past by a number of scientists and practitioners. The aim of the SD methodology developed here is to find a set of optimal network locations at which to place measurement devices. Optimal locations are determined with the aim of collecting data that will be used later on in the calibration of the analyzed WDS hydraulic model. First, existing calibration and SD approaches in the case of WDS are reviewed. After that, SD is formulated as a two-objective optimization problem. The objectives are maximization of the calibrated model accuracy by minimization of the relevant uncertainties, and minimization of total SD costs. The optimal SD problem is then solved using a multiobjective genetic algorithm based on Pareto ranking, niching, and restricted mating. The methodology developed is applied and verified on a case study. At the end, a summary is made and relevant conclusions are drawn. [ABSTRACT FROM AUTHOR]

This paper presents a novel multi-objective genetic algorithm (MOGA) based on the NSGA-II algorithm, which uses metamodels to determine optimal sampling locations for installing pressure loggers in a water distribution system (WDS) when parameter uncertainty is considered. The new algorithm combines the multi-objective genetic algorithm with adaptive neural networks (MOGA–ANN) to locate pressure loggers. The purpose of pressure logger installation is to collect data for hydraulic model calibration. Sampling design is formulated as a two-objective optimization problem in this study. The objectives are to maximize the calibrated model accuracy and to minimize the number of sampling devices as a surrogate of sampling design cost. Calibrated model accuracy is defined as the average of normalized traces of model prediction covariance matrices, each of which is constructed from a randomly generated sampling set of calibration parameter values. This method of calculating model accuracy is called the ‘full’ fitness model. Within the genetic algorithm search process, the full fitness model is progressively replaced with the periodically (re)trained adaptive neural network metamodel where (re)training is done using the data collected by calling the full model. The methodology was first tested on a hypothetical (benchmark) problem to configure the setting requirement. Then the model was applied to a real case study. The results show that significant computational savings can be achieved by using the MOGA–ANN when compared to the approach where MOGA is linked to the full fitness model. When applied to the real case study, optimal solutions identified by MOGA–ANN are obtained 25 times faster than those identified by the full model without significant decrease in the accuracy of the final solution. [Copyright &y& Elsevier]

Models play a pivotal role in advancing our understanding of Earth's physical nature and environmental systems, aiding in their efficient planning and management. The accuracy and reliability of these models heavily rely on data, which are generally partitioned into subsets for model development and evaluation. Surprisingly, how this partitioning is done is often not justified, even though it determines what model we end up with, how we assess its performance and what decisions we make based on the resulting model outputs. In this study, we shed light on the paramount importance of meticulously considering data partitioning in the model development and evaluation process, and its significant impact on model generalization. We identify flaws in existing data-splitting approaches and propose a forward-looking strategy to effectively confront the "elephant in the room", leading to improved model generalization capabilities. [Display omitted] •Data available for model development are often limited and have to be split into model development and evaluation subsets. •How data are split into subsets can result in biased models and misleading model evaluation. •Despite the significant impact of data-splitting, the way this is performed is very rarely justified. •We highlight the importance of data-splitting and problems with both traditional and state-of-the art data-splitting methods. •We chart a way forward so that data-splitting is not ignored and model development and evaluation practice is improved. [ABSTRACT FROM AUTHOR]

There are urban hydrological models, such as InfoWorks, which have been widely used in urban wastewater planning and management. However, such system models require calibration with real world data to achieve reliable and accurate results. One potential method of automating this process is to classify real-world satellite data to categorize image pixels into different classes of permeability within a catchment. This paper employs a two-stage urban wastewater approach with a combination of image classification techniques and modeling using InfoWorks ICM. In image classification, pixel-based and object-based image classification techniques for extracting three land-use categories (roofs, roads, and vegetation area) over six real-world satellite images are evaluated, and the performance of eight supervised machine learning classifiers are compared. Experimental results show that the Bagged Tree classification algorithm provides maximum overall accuracy results when combined with 10,000 objects using the SLIC segmentation method. Also, results show that the classification process requires approximately 12 times less computational time to run all the test images for object-based approach compared to a pixel-based one, demonstrating the power of the implemented combined approach. Pareto optimality, as an efficient analytical method, was used to establish the dominant classifiers. The results show that the knee point (error and time both to be minimum) always belong to object-based classification methods. This indicates that image segmentation produces superior performance, in terms of accuracy and runtime, compared to pixel-based approach. In the second section of this study, the classification results are used as an aid, with real-world InfoWorks model to improve the calibration of the wastewater network, by classifying the land-cover into pervious and impervious segments. By classifying such areas with an urban catchment, and by using the Wallingford PR Equation across all tests of Area 1 (roads), 2 (roofs) and 3 (permeable area), we explore the potential for a partially automated network system calibration process of a wastewater network. [ABSTRACT FROM AUTHOR]