The presented analysis focuses on the problem of statistical downscaling of large-scale data, and suitability of different nonlinear methods of time series analysis for processing GCM predictions of the future climate.