Simulating multimodal seasonality in extreme daily precipitation occurrence

Abstract

Floods pose multi-dimensional hazards to critical infrastructure and society and these hazards may increase under climate change. While flood conditions are dependent on catchment type and soil conditions, seasonal precipitation extremes also play an important role. The extreme precipitation events driving flood occurrence may arrive non-uniformly in time. In addition, their seasonal and inter-annual patterns may also cause sequences of several events and enhance likely flood responses. Spatial and temporal patterns of extreme daily precipitation occurrence are characterized across the UK. Extreme and very heavy daily precipitation is not uniformly distributed throughout the year, but exhibits spatial differences, arising from the relative proximity to the North Atlantic Ocean or North Sea. Periods of weeks or months are identified during which extreme daily precipitation occurrences are most likely to occur, with some regions of the UK displaying multimodal seasonality. A Generalized Additive Model is employed to simulate extreme daily precipitation occurrences over the UK from 1901-2010 and to allow robust statistical testing of temporal changes in the seasonal distribution. Simulations show that seasonality has the strongest correlation with intra-annual variations in extreme event occurrence, while Sea Surface Temperature (SST) and Mean Sea Level Pressure (MSLP) have the strongest correlation with inter-annual variations. The north and west of the UK are dominated by MSLP in the mid-North Atlantic and the south and east are dominated by local SST. All regions now have a higher likelihood of autumnal extreme daily precipitation than earlier in the twentieth century. This equates to extreme daily precipitation occurring earlier in the autumn in the north and west, and later in the autumn 41 in the south and east. The change in timing is accompanied by increases in the probability of extreme daily precipitation occurrences during the autumn, and in the number of days with a very high probability of an extreme event. These results indicate a higher probability of several extreme occurrences in succession and a potential increase in flooding