Evolution of mean, variance and extremes in 21st century temperatures

Warming of the climate system can result in very large corresponding changes in the occurrence of climate extremes. Temperature extremes may occur due to a shift in the whole distribution, where there is an increase in the entire temperature probability distribution, or to changes in the shape of the distribution, such as an increase in variability causing a widening of the distribution. Understanding the precise characteristics of changes in temperature distributions in response to background warming is an important aspect of fully understanding changes in heat extremes and their associated impacts on human and ecosystem health. This study investigates projected 21st century changes in the characteristics of mean, maximum and minimum temperature on daily- and annual- timescales for various regions (Australia, Asia, Europe and North America) using data from seven models participating in the Coupled Model Intercomparison Phase 5 (CMIP5). Using the RCP8.5 experiment we show that an increase in mean temperature throughout the 21st century is a consistent feature of all models for each region. Changes in the variance of simulated temperatures are equivocal, with the sign and magnitude of variance changes in the 21st century varying in different models and regions. A quantile regression analysis demonstrates differences in upper and lower quantile slopes, relative to the mean, including a consistent skew in daily temperatures towards hot extremes. These potentially complex characteristics of temperature changes should not be overlooked, as temperature extremes are potentially more sensitive to changes in the variance and higher order moments than in the mean. Furthermore, a wider range of extreme temperature behaviour may have important consequences for various stakeholders, due to impacts on public health, agriculture and ecological systems.