Overview
Developed in partnership with Stream, V-KEMS is running a three-day virtual study group that aims to bring together mathematical sciences and other disciplines to address challenges involved in the water sector.
Background
Water companies aim to improve the accuracy and effectiveness of their leakage management strategies, yet current forecasting and operational approaches fall short in capturing the true drivers and behaviours of leakage across their networks. Leakage forecasts are typically generated by averaging historical weekly changes, a method that overlooks contextual factors such as weather patterns, ground conditions, and rapid leakage “breakouts.” This leads to substantial gaps between predicted and actual leakage levels, affecting workforce planning, resource deployment, and operational efficiency.
At the same time, District Metered Areas (DMAs)—the foundational units for monitoring and managing distribution networks—vary significantly in their physical and environmental characteristics. Without a robust, data‑driven way to classify these DMAs, it is difficult to apply tailored leakage management strategies that reflect each area’s underlying risk profile and operational needs.
The overarching challenge is therefore twofold:
- To enhance leakage forecasting models by incorporating additional contextual and historical data, moving beyond simple historical averaging to reduce prediction error.
- To develop a systematic method for characterising DMAs using available asset, environmental, and operational data, enabling more targeted, efficient, and effective leakage management interventions.
Together, these challenges highlight the need for improved analytical methods and models that help water companies optimise resource use, reduce leakage, and deliver better outcomes for customers, society, and the environment.
About the Partners
Stream is a collaboration between the majority of UK water companies, supported by industry and civil society partners, with a shared vision to unlock water data to deliver value to customers, society and the environment.
The Virtual Forum for Knowledge Exchange (V-KEMS) is an ongoing collaboration between The International Centre for Mathematical Sciences (ICMS), Isaac Newton Institute (INI), Newton Gateway to Mathematics and Innovate UK Business Connect and various representatives from the mathematical sciences community. The main aim is to identify a range of virtual approaches that will help address challenges from business and industry, the third sector, and other groups outside academia. These challenges may be long-standing or may have arisen directly as a consequence of the present disruption to UK society (such as the pandemic).
The collaborative partnership of V-KEMS means that your data will be shared with the organisation hosting the virtual study group, which for the March 2026 study group is the ICMS.
The challenges in-depth
Problem statement 1 – Forecasting leakage
The leakage team at NWL create a forecast throughout the year on how much leakage they expect to happen week to week. This is used to support decision making for workforce planning and deployment (faster mobilisation to fix leaks) and ultimately be more efficient (less wasted resource in terms of water, people and money). These forecasts are created by averaging historical (last 5 years) leakage changes on a weekly basis.
Whilst this approach is nice and simple to run, it is not particularly accurate and there is often a gap between forecast and actual. Averaging doesn’t deal very well with situations that are known as leakage breakouts (where lots of leaks happen all at once across an area – usually following a protracted cold or dry spell which causes the ground to move and fracture pipes). A breakout will create a spike in leakage but the level of leakage generally comes down quicker than the model predicts leading to inaccurate forecasts.
The challenge is to find ways to refine the forecast to narrow down the margin of error by looking at how other factors have played a part, such as preceding (say, 3 months) weather (particularly protracted dry and cold spells). At present the forecast only relies upon historic levels, with no understanding of the context at the time. Up to 20 years of daily leakage levels are available for each company (Essex and Suffolk Water and Northumbrian Water – so 1 data point per day per company) and leak repair data is also available going back about 5 years for both areas.
Problem statement 2 – Characterising district meter areas so that different strategic approaches to managing leakage can be applied appropriately
A district metered area is a discrete area of a water network that has a known boundary and flow meters that monitor the volume of water going into that area every day. On average they tend to feed between 1,000 and 2,000 properties and cover about 11km of distribution network, although there are some outliers on either side of these ranges. The DMA will have certain physical (and non-physical) characteristics that may individually, or in concert, determine whether that DMA is more or less likely to experience leakage under different sets of circumstances. Factors may include pipe material, age, ground conditions, property type, pressure (along with many more). How might we use available data to determine the character of a DMA so that different leakage management strategies can be applied appropriately? Being able to do this enables focus and intervention to be more fine-tuned, leading to better leakage performance and ultimately positive outcomes for customers, society, the environment and the water companies themselves.
