A simple statistical-dynamical downscaling scheme based on weather types and conditional resampling

Abstract

A multivariate statistical downscaling methodology is implemented to generate local precipitation and temperature series at different sites based on the results from a variable resolution general circulation model. It starts from regional climate properties to establish discriminating weather types for the chosen local variable, precipitation in this case. Intratype variations of the relevant forcing parameters are then taken into account by multivariate regression using the distances of a given day to the different weather types as predictors. The final step consists of conditional resampling. The methodology is evaluated in the Seine basin in France. Using reanalysis fields as predictors, satisfying results are obtained at daily timescale and concerning low-frequency variations, both for temperature and precipitation. The use of model results as predictors gives a realistic representation of regional climate properties. Nevertheless, as the validation of a statistical downscaling algorithm for present day climate conditions does not necessarily imply the validity of its climate change projections, the plausibility of the downscaled climate projections is assessed by verifying the consistency between spatially averaged downscaled results and direct model outputs for two climate change scenarios. Despite some discrepancies for precipitation with the more extreme scenario, the consistency is good for both local variables. This result reinforces the confidence in the use of the downscaling scheme in altered climates. Finally, it is shown that the intertype variations of the atmospheric circulation represent only a fraction of the climate change signal for the local variables. Thus a downscaling methodology based on weather typing should incorporate information concerning intratype modifications. Copyright 2006 by the American Geophysical Union.