• English 
  • Spanish 

On the Credibility of Regional Climate Models to Reproduce Extreme Precipitation

Regional Climate Models (RCMs) are key tools in the projection of future regional climate, by numerically solving the governing equations of the atmosphere nested into Global Climate Models (GCMs), forced with future emission scenarios. Despite being physically based, RCMs include many non-observable parameters, which need to be adjusted to represent the observed climate as closely as possible. This tuning process is usually not well documented, but central statistics of well-observed variables, such as temperature or precipitation, have greatly reduced their bias in state-of-the-art RCMs as compared to earlier versions. The availability of new gridded observational products has also contributed to this advance.

Extreme events in climate science follow essentially two approaches: (1) a pragmatic approach, based on sample high-order percentiles (e.g. 95 th percentile) and (2) a proper characterization of extremes, by fitting a theoretical model (GEV, POT) to data and considering measures such as return values or distribution parameters. The percentile approach has shown that, due to the nature of precipitation (reasonably fitted by a single-parameter exponential distribution), a bias correction of the mean also reduces biases in tail percentiles. In this work, we compared this result with a proper characterization of extremes, showing that the bias correction of central statistics also improves return values of precipitation. This has implications for the credibility of RCMs in an altered climate, when they have been tuned to fit the current observed climate.

Palabras clave: