Searching for extremes: A framework for decision-relevant stress tests using weather types

Climate change is expected to alter the frequency, duration, timing and severity of droughts. Traditional top-down approaches to assessing the performance of water supply systems under different climatic conditions can miss drought vulnerabilities, particularly where drought characteristics may be altered under climate change. Stress-tests under a bottom-up framework offer a way of identifying water supply system vulnerabilities to droughts more severe and extreme, and with different characteristics, to those experienced historically.

An inverse stochastic approach was developed to elicit weather type transitions that cause severe and extreme droughts for a water supply system in mid-west Wales, United Kingdom. Droughts are defined at the start of the process by the end-user, focussing on drought characteristics where consequences are decision-relevant. The inverse stochastic approach (using a Markov Chain stochastic model trained on historical synoptic weather types) then perturbs the likelihood of weather type transitions to produce the user-specified droughts.

The approach provides actionable insights for water managers by identifying water supply system vulnerabilities to different drought dynamics, as well as indicating how implausible the droughts would need to be in order to reach the targeted drought definitions. The impacts of climate change can be included by incorporating changes in weather types from validated climate models, using a range of methods from simple changes in future occurrence, to more complex stochastic models trained on future weather types.