A statistical assessment of tropical cyclone activity in atmospheric general circulation models

The properties of tropical cyclones in three low-resolution atmospheric general circulation models (AGCMs) in seven ocean basins are discussed. The models are forced by prescribed, observed sea surface temperatures over a period of 40 yr, and their simulations of tropical cyclone activity are compared with observations. The model cyclone characteristics considered include genesis position, number of cyclones per year, seasonality, accumulated cyclone energy, track locations, and number of storm days. Correlations between model and observed interannual variations of these characteristics are evaluated. The models are found able to reproduce the basic features of observed tropical cyclone behavior such as seasonality, general location and interannual variability, but with identifiable biases. A bias correction is applied to the tropical cyclone variables of the three models. The three AGCMs have different levels of realism in simulating different aspects of tropical cyclone activity in different ocean basins. Some strengths and weaknesses in simulating certain tropical cyclone activity variables are common to the three models, while others are unique to each model and/or basin. Although the overall skill of the models in reproducing observed interannual variability of tropical cyclone variables has not surpassed or often even equalled that of statistical models, there exists potential for higher future skills using improved versions of dynamical approaches.