• English 
  • Spanish 

Assessing the role of regional climate models as intermediaries between global models and bias correction methods

Conference: Avances en la detección y proyecciones del cambio climático en España a la luz del 5º informe del IPCC
Year: 2015
Contribution type: Oral
PDF file: 2015_Casanueva_climate-es.pdf
Poster: 2015_Casanueva_CLIMATE-ES.pdf

Recently, the amount of publicly available future climate simulations has greatly increased, allowing for more robust climate change uncertainty assessments. However, the number of available experiments is strongly reduced when considering only computationally costly regional climate model (RCM) runs. These simulations are used as input for many impact studies, but usually after a statistical post-processing which takes into account the observations at station level (or very high resolution grids) and is commonly known as bias correction. These methods have been primarily developed for removing model biases, although they are also implicitly used to bridge the scale mismatch between model grids and local observations.

Given that bias correction can be directly applied to Global Circulation Models (GCMs), we assess the role of RCMs as intermediaries between GCMs and bias correction methods. We apply different bias correction methods (from the simple scaling to different versions of the quantile-quantile mapping) directly to daily precipitation from a set of coarse resolution CMIP5 GCMs and compare them with bias-corrected EURO-CORDEX RCMs (namely COSMO-CLM and RCA at 12km horizontal resolution) for historical and future periods. We use a subset of European stations from the ECA&D dataset as observational reference for the bias correction. Additionally, a refined analysis in the Ebro catchment illustrates the effect of bias correction on spatial climate variability.

The evaluation experiment of the historical simulations in a cross-validation framework shows that bias corrected RCMs and GCMs typically present similar biases, but an added value of RCMs is found for the spatial analysis. Second, the basic features and the overall magnitude of the climate change signal are preserved after applying bias correction, although in some cases the uncertainty in the climate change signal is amplified by bias correction.