CR 72:101-110 (2017)  -  DOI: https://doi.org/10.3354/cr01456

ABSTRACT: Very few studies on the impact of climate change have been carried out in Kazakhstan, which is located in Central Asia. It is the largest landlocked country in the world and has a sensitive natural environment and a human society that is vulnerable to climate change. In this study, we evaluated a statistical model for downscaling the monthly mean temperature in the Kazakhstan area built from a linear regression model combined with a principal component analysis (PCA) as the preprocessing method for predictors. The air temperature, geopotential height and both components of the wind were selected as predictor variables. The result shows that the linear regression model was able to simulate monthly mean temperature averaged over the Kazakhstan region as a whole reasonably well, although there are a few mismatches with observations for some stations and in some months. A further analysis of the results of downscaling also reveals that the monthly mean temperature in summer is downscaled more accurately by this model than that in winter, with the R² value of 0.8 for summer being larger than that for winter of 0.7. Moreover, this statistical downscaling model shows poor performance in complex terrain areas compared to flat terrain areas, with R² values for the southeastern mountain station and the station by the Caspian Sea being smaller than those for other stations in Kazakhstan.

KEY WORDS: Kazakhstan · Statistical downscaling · Linear regression · Monthly mean temperature