Work

Which uncertainties matter? Global sensitivity analysis of model-based and data-driven leakage detection methods via sparse-grid stochastic collocation

Year: 2026

Type: article

Abstract: Leakages in water distribution networks cause water loss globally, jeopardize water quality, and increase energy consumption for water supply. While numerous software-based detection methods to enable prompt leakage detection have been proposed in the literature, most are developed using synthetic data that fails to represent the range of errors and operating conditions encountered in practice. To address this gap, we conduct a variance-based global sensitivity analysis of the detection performance of two state-of-the-art leakage detection methods, data-driven and model-based, under uncertainties in the number of pressure sensors, sensor accuracy, hydraulic model parameters, and a method hyperparameter. To keep the analysis tractable, we use sparse-grid stochastic collocation, reducing the number of model evaluations by orders of magnitude relative to Monte Carlo. We assess convergence, robustness, and credibility of the estimates. In this study, sensor number and measurement accuracy dominate detection performance for both methods, whereas hydraulic parameters and the detection threshold are secondary; finer rankings among the remaining inputs differ by leakage type and method but often fall within the estimated uncertainty. The two methods fail in different regimes, motivating hybrid strategies that combine their strengths

Source: Water Research

DOI: 10.1016/j.watres.2026.127199

Authors: Ella Steins, Johannes Koslowski, Andrea Cominola

Institutions: Einstein Center Digital Future, Technische Universität Berlin, Berlin Centre of Competence for Water

Language: English

Topic: Water Systems and Optimization

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Clean water and sanitation

Open Access status: hybrid

Estimated APC paid: $4,660