Statistical Downscaling of General Circulation Models (GCMs); History, Principles, and Methods

Abstract

The climate system is very complex and has made the modeling and predicting/projecting face many challenges. Although climate variability may be detected and identified through a time series of observations, it cannot express the interaction of various components of the Earth's climate system. General circulation models (GCMs) are essential for simulating the physical processes governing the atmosphere and the interaction of the components involved in the Earth's climate system. Statistical downscaling extracts empirical relationships between small-scale observational variables (often at the station level) and the direct GCM output by applying three approaches: Perfect Prognosis (PP), Model Output Statistics (MOS), and Weather Generators (WGs). Bias correction, widely used in climate change studies, is the MOS statistical downscaling approach. To clarify the role of using the inappropriate method and software in increasing uncertainty, two scaling methods from the model output statistics (MOS) approach are compared to correct the bias of the minimum and maximum temperatures. In this research, the outputs of R and CMhyd software are compared to check the uncertainty caused by using inappropriate software. The output of the EC-Earth3-CC model for two variables of the minimum and maximum temperatures was examined using CMhyd and R software. Examining the results showed that the CMhyd software has a significant error in both extracting the direct model output and the bias correction method. For example, the PBIAS of direct output of maximum temperature in Abadan was 2.10%, while CMhyd software gives 5.10%. The result of this research shows the need to use the correct methods and software for processing the output of GCMs.

Keywords

• statistical downscaling
• model output statistics (mos)
• bias correction
• cmhyd software