Simulation of climate change over Europe using a nested regional-climate model .2. Comparison of driving and regional model responses to a doubling of carbon dioxide

Results are assessed from a 10-year simulation of the equilibrium response to doubled carbon dioxide (CO2) over Europe made with a nested high-resolution regional-climate model (RCM). The simulated changes are compared against those produced by the driving global general-circulation model (GCM). The domain-averaged increases in temperature and moisture content are similar in both models. Because of a stronger hydrological cycle the increases in precipitation and evaporation are larger in the RCM than in the GCM, whereas the reductions in lower and middle tropospheric relative humidity and cloud cover are smaller. The frequency of intense precipitation events increases substantially in both models; however, the fractional changes are significantly smaller in the RCM. The proportion of precipitation associated with convection also increases in both models; however, much of the increase in intense events is explained simply by increased atmospheric moisture concentrations, especially in the RCM. The time-averaged precipitation changes in the RCM contain a substantial mesoscale component on scales not resolved by the GCM. Attempts are made to reconstruct this component from the changes in the large-scale atmospheric circulation using empirical relationships calibrated from the 1 x CO2 integration. These are largely unsuccessful, indicating that simple downscaling schemes to generate high-resolution scenarios of climate change from coarse-grid GCM data may be of only limited applicability. Further statistical calculations suggest that longer integrations (similar to 30 years) are needed to reduce the sampling uncertainty associated with the simulated mesoscale component to an acceptable level. The large-scale patterns of change of surface temperature and precipitation reveal significant regional contrasts which are influenced both by changes in atmospheric circulation and regional physical feedbacks. The RCM changes are similar to those of the driving GCM except in summer. The differences in the summer changes are traced to differences between the 1xCO(2) integrations, in which the influence of the lateral boundary forcing on the RCM simulation is found to be anomalously weak. It is argued that any development of significant divergence between the RCM and GCM solutions on scares resolved by the latter may imply the need to refine or replace the one-way nesting technique currently used in many regional modelling experiments.