Stochastic modelling of (not-so) long-term hydrological data: Current status and future research

Abstract

Stochastic modelling of long-tem hydrological data is one of the tools used to evaluate water supply system security. This paper summarizes the findings of recent research on stochastic modelling of annual hydrological data which has a significant impact on its practical application. Firstly, in the single-site context the relatively short lengths of observed records induce significant parameter uncertainty which severely hinders model identification. Secondly, a generalised framework has been developed which can evaluate model and parameter uncertainty for multi-site models of annual hydrological data. Two simplified reservoir simulation case studies demonstrated that estimates of drought risk were significantly increased when parameter uncertainty was incorporated. This indicates current approaches underestimate drought risk. Several issues are highlighted which need to be resolved prior to the application of this general framework to more realistic case studies. These are currently the subject of further research. The aim is that in the near future a general framework for calibrating multi-site stochastic models of hydrological data will be available for water supply authorities to improve their estimation of long-term drought risks.