Role of statistics in the validation of general circulation models

The field of statistics should play a central role in the validation of general circulation models (GCMs). Both the perspectives of climate modelers and of climate impact researchers are addressed. Confusion over what statistics can or cannot be expected to achieve is described, and an overemphasis of statistical tests of significance is identified. A compelling case is made for assessing the performance of GCMs in terms of their ability to reproduce climate variability, not just average conditions. Unsatisfactory aspects of current statistical practice are pointed out, with the problem of multiple comparisons illustrated by an example. Prospects for the future are considered, emphasizing the need for factoring statistical issues into the design of GCM experiments, as well as the possible application of new statistical techniques. Indicative of ways in which existing techniques could be used more appropriately in the validation of GCM output, an example that involves a multiple confidence interval is presented. If the appropriate role of statistics were recognized, one obstacle currently hindering the improvement of GCMs and limiting their use In climate impact studies would be removed.