Using statistical techniques to deduce local climate distributions. An application for model validation

✍ Scribed by M. Noguer

• Publisher: John Wiley and Sons
• Year: 2007
• Tongue: English
• Weight: 868 KB
• Volume: 1
• Category: Article
• ISSN: 1350-4827
• DOI: 10.1002/met.5060010309

✦ Synopsis

Due to increasing interest in the prediction of detailed regional climate, a method to deduce local climate distributions from large-scale variables is proposed. Since surface parameters are dqficult to reproduce with the current resolution of General Circulation Models (GCMs), a statistical approach is used to fill the gap between G C M outputs and regional climate. The idea is to find relationships between local climate variables (the predictands) and predictors based on the large-scale flow using Canonical Correlation Analysis, and then to find a regression model to relate the important canonical variables to each of the local surface variables. Predictor variables are provided by the monthly average mean sea-level pressure field simulated by the Hadley Centre coupled model over the North Atlantic region. The predictands are the simultaneous monthly average precipitation simulated over the Iberian Peninsula at four grid points, where the grid-point values are intevreted as a proxy for station data. The results suggest that the technique can reproduce the major variations in regional climate on the monthly time scale.