Monitor selection and data assimilation to improve PHC in Severn Trent (2603)

Project Description

Severn Trent Water (STW) have been using a small area monitor (SAM) to estimate unmeasured per household consumption (uPHC) for several years. However, in time some doubts about the representativeness of the SAMs started to emerge.
In 2022 Artesia supported STW augmenting the SAM-based estimation through a hybrid approach, selecting a monitor from an existing cohort of unmeasured properties fitted with an AMR meter. The monitor selection was based on a stratified random sampling approach, that selects a representative sample based on demographic information: occupancy, water usage groups, property type, and population density.


The hybrid methodology aimed at merging the existing SAM-based estimation of uPHC with the monitor-based estimation of uPHC using a Data Assimilation (DA) approach. In simple terms, this is a weighted average, where the weights are inversely proportional to the uncertainty. To take advantage of the better representativeness of the new monitor for specific strata of the unmeasured population, the DA approach was applied with stratification.

The monitor-based estimation of uPHC considered some corrections and uncertainty estimations due to known biases. Four sources of biases were identified and accounted for: geographical bias (as properties are not geographically representative), monitor bias (as knowing they are part of a monitor, the unmeasured properties may behave differently), Meter Under-Registration and Underground Supply Pipe Leakage.

The new hybrid methodology proved very successful and was adopted for both regions of Severn Trent. Compared to the SAM-based method, the uncertainty on the uPHC is significantly reduced using the new Data Assimilation approach.

See more case studies