Comparison of temperature variability in observations and sixteen climate model simulations

Understanding how much, if any, of observed climate changes are anthropogenic depends upon understanding the magnitude and spatial patterns of natural climate variability. We have compared simulated surface air temperature (SAT) variability in 16 coupled ocean-atmosphere-sea ice climate model simulations to observed temperature variability. The majority of the simulations exhibit excessive air temperature variability over land while simulated temperature variability over oceans is generally too low. The ratio of variability over land to over oceans is too high in all the simulations, relative to observations. We have identified several factors which may contribute to the differences in temperature variability. In particular, many of the models use "bucket" land surface schemes which produce greater temperature variability over land, due to lower levels of soil moisture, than more realistic land surface schemes produce.