Statistical Modeling of the Global River Runoff Using GCMs: Comparison with the Observational Data and Reanalysis Results

Abstract

Specific methods are proposed to assimilate the results of the “historical” experiments on 28 climate models. The results of the analysis confirm the hypothesis regarding a stationary character of changes in the global river runoff during “instrumental” period (approximately 150 years). Part of the models (about one third) reproduces the non-stationary changes in the global runoff with respect to the mean. At the same time, the number of such models indicating increased runoff is exactly equal to the number of models that indicate a decrease in runoff. The models generally reproduce well the coefficient of variation of global river runoff in comparison with the observational data, as well as the small value of the coefficient of asymmetry. The model of the Gaussian white noise is optimal for the description of the majority of the annual time series of global river runoff generated by the GCMs.