Partitioning model uncertainty in multi-model ensemble river flow projections

Abstract

AbstractFloods are the largest natural disaster currently facing the UK, whilst the incidents of droughts have increased in recent years. Floods and droughts can have devastating consequences on society, resulting in significant financial damage to the economy. Climate models suggest that precipitation and temperature changes will exacerbate future hydrological extremes (i.e., floods and droughts). Such events are likely to become more frequent and intense in the future; thus to develop adaptation plans climate model projections feed hydrological models to provide future water resource projections. ‘eFLaG’ is one set of future river flow projections produced for the UK driven by UKCP18 climate projections from the UK Met Office. The UKCP18-derived eFLaG dataset provides state-of-the-art projections for a single GCM driven by RCP 8.5 across the entire UK. A QE-ANOVA approach has been used to partition contributing sources of uncertainty for two flow quantiles (Q5 high flows and Q95 low flows), at near and far future time scales, for each of the 186 GB catchments in the eFLaG dataset. Results suggest a larger hydrological model uncertainty associated with low flows and greater regional climate model uncertainty for high flows which remains stationary between flow indicators. Total uncertainty increases from near to far future and highly uncertain catchments have been identified with a high concentration in South-East England.