Regional climate‐model predictions of extreme rainfall for a changing climate

Abstract Major floods occurred in the United Kingdom during autumn 2000. These were caused by a rapid sequence of heavy rainfall events that occurred over a period of many weeks leading to record‐breaking monthly‐to‐seasonal rainfall totals. The question was raised as to whether such rainfall events may be related to human‐induced climate change. Climate‐model predictions of future changes in mean precipitation behaviour are well established. However, to understand flooding requires an examination of predictions of extreme rainfall behaviour at a relatively small spatial scale. For three areas within the United Kingdom, output from a Hadley Centre regional climate model, ‘nested’ within one of its general‐circulation models, is compared with raingauge data averaged over these areas for the period 1961–1990. This shows that the modelling system is good at predicting the statistical likelihood of extreme rainfall events seen in historical data. This result holds for extreme rainfall totals over daily to monthly timescales. When the modelling system is used to predict changes in these extreme events resulting from atmospheric CO 2 concentrations that may be representative of the period 2080–2100, significant reductions in the return periods of such events are seen. For example, 30‐day rainfall totals, which happened in the recent past on average once in 20 years, are predicted to happen once in 3–5 years. An interpolation method based upon climate‐model output and incorporating raingauge data is used to estimate how rainfall extremes may have changed between the middle of the 19th century, and for a period centred on the year 2000. This also predicts that increased greenhouse gases have led to reduced return periods of extreme rainfall events for three sites of interest, though in this case the changes are not statistically significant. © Royal Meteorological Society, 2003. R. G. Jones's contribution is Crown copyright