United Utilities in the United Kingdom requested cutting-edge water resources modeling to more fully understand its water supply, especially as it relates to improving drought resilience. Lynker Technologies deployed a two-part stochastic simulation technique to model both the inter-annual flow sequences using a non-homogeneous Markov Chain (NHMC) model and the intra-annual variability using a k-nearest neighbor (KNN) nonparametric method. The flow simulation was performed simultaneously at four gauge locations so that the water system could be analyzed together as it is operated by the client.
The final output was 5,000 years of simulated streamflow at a daily timestep, which preserved spatial correlation between the sites, decadal variability of the streamflow, and seasonal trends while using empirically-based methods. The simulated streamflow data greatly expanded upon the 50-year observed record, which provides improved estimates of for instance the 100-year drought recurrence interval.