Transient changes in flood frequency and timing in Britain under potential projections of climate change

Climate change could have dramatic consequences for the earth's environment, especially its hydrology, yet the noise of natural climate variability can mask the impacts of climate change on shorter time scales, and can act (in an unpredictable way) to enhance or reduce its effect in any given period in the future. Thus, impact studies based on time slices, which look at modelled differences between baseline (e.g. 1961-1990) and future (e.g. 2070-2099) periods, can be misleading. This paper makes use of three new transient climate projections, from a perturbed parameter ensemble of a regional climate model (RCM) covering the period 1950-2099, to investigate transient changes in flood frequency and timing for two example catchments in England. Annual maximum (AM) time series are extracted from modelled flow timeseries (for hourly, daily mean and running 30-day mean flows). The AM series are analysed in terms of flood frequency (using a fitted generalised logistic distribution) and timing in a 30-year moving window. A non-linear trend analysis is performed on the derived time series, with permutation testing to estimate statistical significance. The results show that changes over the period are often non-linear, and vary considerably in size and statistical significance according to catchment, flow time step and RCM ensemble member. The relative effects of the three RCM ensemble members are consistent with their relative climate sensitivities. A nationwide analysis, using daily mean flows from a grid-based runoff and routing model for the UK with one of the RCM ensemble members, was consistent with the catchment results in terms of direction of trends, but generally gave trends of a higher magnitude, indicating the presence of some hydrological model structure uncertainty. The nationwide results suggest increased flood risk across much of the country. Copyright (C) 2010 Royal Meteorological Society