Editorial

Hydraulic and water quality simulation models are widely used nowadays by planners, water utility personnel, consultants and many others involved in analysis, design, operation or maintenance of water distribution systems (WDS). As with all mathematical models, WDS models need to be calibrated before useful results may be obtained. This special issue contains a number of papers dealing with important topics related to WDS simulation model calibration.

The paper by Savic, Kapelan and Jonkergouw reviews the current state-of the-art in the field of WDS model calibration. The review covers a wide range of past and current approaches to the calibration of both water quantity (i.e. hydraulic) and quality models. The review also covers related sampling design methods for the calibration of WDS models. The authors note that the required calibration accuracy of a model depends mainly on the intended use of the model and that it is important to address this issue before starting a model calibration. The intended use of a model, in turn, dictates the quality and quantity of calibration data that are required as well as the calibration approach that should be used. The authors also address the issues of individual vs. simultaneous calibration of WDS quantity and quality models. They conclude that the simultaneous calibration of hydraulic and water quality parameters with both hydraulic and water quality observed data should be performed to decrease overall uncertainty and provide superior results to independent calibration.

The paper by Speight and Khanal discusses the typical model calibration approaches currently carried out in the USA. The authors note that the reality of modelling for real distribution system operators is that often an ideal calibration cannot be achieved due to the lack of resources, data scarcity and time constraints. They also observe that despite the advances that have been made in research on model calibration, including numerical techniques for optimisation, the actual practice for water utilities in the USA falls behind the academic literature. It turns out that, despite the inclusion of optimisation routines for calibration in commercial software, relatively few utilities are using these tools on a regular basis. The authors also remark that many of the large water utilities have chosen to invest first in GIS and SCADA to develop good input data for a model, with plans to update the hydraulic model of the distribution system once those data sources are ready. The reason for this seems to be that the calibration procedure is often seen more as a way to find and correct errors in input data than to adjust model parameters to a fine degree of precision.

An example of the practical use of a calibration approach in the Las Vegas Valley Water District Planning Division (USA) can be found in the paper by Jacobsen and Kamojjala. Unlike many others, this division conducts annual and daily calibration of the hydraulic model of its water distribution system. Once calibrated, the model is used to estimate the hydraulic characteristics of the water distribution system at locations where measured data are unavailable or unknown, spatially and temporally. The hydraulic model calibration process involves a comparison of predicted model results against actual field data and supervisory control and data acquisition (SCADA) system measurements and adjusting the attribute values and model parameters, if necessary.

The paper by Giustolisi and Todini addresses the issue of modelling demands in the context of WDS model calibration. The authors argue that the conventional approach of modelling demands represented as lumped water consumption at network nodes (instead of being distributed along pipes) is introducing an approximation that may lead to significant headloss errors. These errors, in turn, may affect the calibration of a WDS hydraulic model which is typically based on observed pressures. To correct this, the authors suggest an extension of the global gradient algorithm for network analysis by Todini and Pilati (1989) where a corrective term for each pipe's head loss is introduced. The purpose of this corrective term is to allow the current practice of using lumped nodal demands, but without losing the physically correct representation of head losses. This should lead to more accurate WDS hydraulic models that, in turn, should enable better calibration of these models.

The paper by Wu utilises the calibration methodology to detect leaks in the WDS. Leakage is represented as pressure-dependent emitter flow at WDS nodes. Leakage detection is formulated as a nonlinear parameter identification problem to search for the most likely emitter node locations and the emitter coefficients while minimising the difference between the field-observed and the model-simulated flows and pressures. This way, the leakage detection optimisation is developed as an integral component of the unified model parameter optimisation framework. The methodology is applied to a water system in the UK. It illustrates the effectiveness of the unified approach for both detection of leakage hotspots and extended-period simulation model calibration using the real field data.

Finally, the paper by Branisavljević, Prodanović and Ivetić deals with the issue of nodal demand uncertainty and its propagation to uncertain, calibrated WDS model predictions (nodal pressure and pipe flows). More specifically, the paper explores how the information on observed system inflow (which should equate to the total demand in the system at any time) can be used to reduce the uncertainty in predicted model outputs. The uncertainty quantification methodology shown is based on the fuzzy α cut method and the aforementioned uncertainty reduction is calculated by quantifying the prediction uncertainty with and without considering the observed system inflow as an additional constraint. The results obtained on a real-life case study are verified by using the well known Monte Carlo Simulation method.

Enjoy reading the papers!