Anthropogenic changes in temperature and precipitation extremes over land have been analyzed from simulations with two different atmosphere-ocean general circulation models, CCCma and HadCM3. The goal of this analysis is to determine indices for changes in extremes that are best suitable for an early detection of climate change and that are moderately robust between models. Extremes are analyzed including the seasonal cycle, which emphasizes changes that should have the strongest impacts on society.

In various regions changes in temperature extremes are significantly different from changes in seasonal means, and the difference between both is often robust between models. This indicates that a detection of mean changes cannot be substituted for detection of changes in extremes. Similarly, extreme precipitation seems to generally increase stronger than mean precipitation. Correlations between climate change patterns for precipitation are generally low between models. They increase if only areas with significant changes are compared, and increase for changes in extreme rather than mean precipitation. We have estimated the signal-to-noise ratio of changes in mean and extreme precipitation between the end of the 20th century and the time of CO2 doubling. Results suggest that changes in moderately extreme precipitation should be slighly better detectable than changes in mean precipitation and also more robust between models.