Assessment of Climate Model Performance over the Australian Region

Climate modeling simulations from 5 models are assessed over the Australian region, by comparing them with the NCEP-NCAR reanalysis data sets for the same region. The motivation for the study is the need to quantify the performance of climate models in simulating the relatively well-observed 30-year period from 1970-1999. This is a critical first step in using a much longer climate model output, such as sea level change. The future climate will cover the same region, but extend out to 2100. The longer simulation also will examine a range of other climate parameters such as sea-level change. If the model performance is poor over the present period, the future climate simulations are open to obvious doubt about how well they are representing future climate, especially in marginal areas that might experience major climate shifts from, for example, semi-arid to arid. The models are the Hadcm3, Echam5, MIROC_3.2, the new CSIRO model and PCM1. The focus of the 30-year simulations is on the accuracy of the key variables, precipitation and near-surface air temperature. If such variables are not well simulated, there is greatly reduced confidence in the accuracy of future climate model simulations of these and other variables. Differences between each of the 5 models and the long-term NCEP-NCAR climatology for that period are calculated; as is the difference of an ensemble mean of the models from the NCEP-NCAR reanalysis. For both near-surface air temperature and precipitation, the models are relatively consistent with some notable differences that will be described in the presentation. In particular, the trends track closely some marginal areas that are in danger of experiencing genuine climate change by exceeding the bounds of natural variability in the observed data sets. Examples will be shown of such cases. In general, the air temperatures from the models increase at a rate that exceeds the observed over much of Australia, and rainfall is more evenly spread over the year than observed among regions that have distinct wet and dry seasons, resulting in a blurring of these seasons in the climate models. Finally, ensemble averaging the 5 climate models performs best but still retains the biases mentioned above present in most of the individual models.