• English 
  • Spanish 

A critical reassessment of statistical downscaling techniques under climate change conditions in Spain: Accuracy, distributional similarity and robustness

The performance of a wide range of statistical methods to downscale temperature (both maxima and minima) and precipitation in Spain is critically reassessed with respect to their robust applicability in climate change studies. For such purpose, we cross-validate the different techniques focusing on three distinct aspects: 1) accuracy (correlation, etc.), 2) distributional similarity (KS and PDF scores), and 3) stationarity/robustness under climate change conditions. For the latter, which is normally missed in state-of-the-art downscaling studies, we consider anomalous warm/cold (temperature) and dry/wet (precipitation) periods from the observational records - which serve as surrogates of possible future climate alterations - and test the performance of the methods, as compared to random periods. It is found that a near-perfect association between observed and downscaled series (e.g. high correlations) may hide important shortcomings in terms of distributional similarity and robustness. Therefore, a wide collection of downscaling configurations (domain-predictor-methodology) are evaluated in order to identify those being the most appropriate to deal with this issue.
This study is part of the work done in the project ESTCENA, a strategic action of Plan Nacional de I+D+i 2008-2011.