Evaluation of simulated recent climate change in Australia

Abstract

The ability to reproduce recent observed climate change in climate models is a pertinent prerequisite for trust in climate projections. Also, information on the consistency of simulated and observed recent changes helps users to interpret near-term climate change projections. A comprehensive assessment of simulated regional trends, however, is often not available. Therefore, we evaluate daily maximum and minimum temperature trends and rainfall trends from 1956-2005 in Australia in simulations from the CMIP5 archive. For all variables and all models, we find significant (at the 10% level) differences between simulated and observed trends in some areas. Except for daily minimum temperature in spring and summer, however, the areas where we find significant differences are smaller than what we expect by chance. In a multivariate evaluation, simulated joint temperature and rainfall trends of all but one model, however, are found to be significantly (at the 10% level) different from the observed trends. Hence, multivariate evaluation provides a stricter test. We conclude that CMIP5 models share trend biases and regional projections therefore have to account for the presence of biases shared across models.