Evaluating Climate Model Data
The initial evaluation was completed by comparing the annual rainfall and surface air temperature from the models with the observational data (see the case study report for results). Many models are within +/- 1 °C of the observations. Almost all models show a negative bias (too cold) relative to the observations (only ACCESS1-3 is slightly warmer). The range of simulated rainfall is wider, from less than 3 mm/day to more than 8 mm/day, with the observations of around 5.3 mm/day.
Comparison of the observations with maximum and minimum temperatures shows the maximum temperatures have a much larger cool bias, while the minimum temperatures show a warm bias.
The six regional climate model results show very tight clustering for air temperature. This is the result of the bias correction used in these simulations which corrected the mean monthly ocean temperatures for the historical period to match the observed climatological ocean temperatures. With similar climatological ocean temperatures and using the same atmospheric model, the resulting rainfall amounts are also similar (though not as similar as temperatures).