Non-Revenue Water (NRW) and Leakage Management: Global KPIs and Modelling Challenges

The water distribution system is a critical infrastructure for achieving Sustainable Development Goal 6, yet global inefficiencies remain due to high levels of Non-Revenue Water (NRW). This term refers to water that is lost before it reaches the consumer, primarily due to physical leaks, metering inaccuracies, or unauthorised consumption. Global NRW averages 29.5%, with extremes ranging from 4% in Singapore to 83% in Armenia (Liemberger & Wyatt, 2019). These losses equate to over 126 billion cubic metres annually, undermining both economic performance and water security in the face of climate change.

This study conducts a global assessment of NRW using two key performance indicators: the NRW percentage and the leak flow index (L/s/km). Drawing from international databases and reports (Lambert, 2001; Pokhrel et al., 2023), the analysis categorises countries and continents according to infrastructure conditions and efficiency thresholds. The review also evaluates conventional hydraulic models and compares them with dynamic approaches such as the Rigid Water Column Model (RWCM), which accounts for transient pressure effects during operations (Coronado-Hernández et al., 2024; Fuertes-Miquel et al., 2024).

Our findings show that 76.6% of countries exceed the 25% NRW threshold that is deemed economically inefficient, and 25% report values above 40%. Leak indices range widely, from 0.002 to 4.93 L/s/km, and correlate strongly with ageing infrastructure and inadequate pressure control (Thornton et al., 2008; Giustolisi et al., 2024). RWCM-based simulations reveal that standard Extended Period Simulation (EPS) models may underestimate leakage by 18–55% due to their inability to capture inertial effects.

Reducing NRW by one-third could supply 800 million people annually and generate savings of up to USD 13 billion. The proposed integration of RWCM with machine learning involves using time series generated by dynamic hydraulic simulations as training data for predictive models that estimate leakage behaviour under varying operational conditions. This hybrid approach offers a promising pathway to design smarter interventions, optimise pressure management, and enhance the resilience of water distribution networks.