Construction of Winter Temperature Scenarios over Greece, Using an Alternative Statistical Downscaling Model Based on CCA

Abstract

In this study, an attempt is made to construct winter temperature scenarios over Greece, using an alternative statistical downscaling (Sd) model based on canonical correlation analysis. The proposed model derived, considering the long-term trends of the predictor variables (1,000–500 hPa thickness field geopotential heights) and the predictand variables (observed mean minimum winter temperatures over Greece). Both the trends of the above sets of variables were eliminated by using linear regression models based on generalized least square estimators. The Sd model is firstly calibrated for the period 1959–1978 and 1994–2000 and then it is validated for the intermediate years 1979–1993. Afterwards, the same model is applied in order to generate a present day scenario using the data from the General Circulation Model (GCM) HadAM3P, for the period 1961–1990 which is the control run period. In conclusion, a large scale output of IPCC – SRES is fed into these statistical models, in order to estimate the above temperatures in 2071–2100. The advantages of the proposed method to classical approach of CCA have been quantified in terms of a number of distinct performance criteria.