Temperature and Precipitation Scaling in CMIP5 Simulations and Paleoclimate Records of the Last Millennium

Accurate assessments of future climate impacts require realistic simulation of decadal to centennial-scale temperature and precipitation variability, yet there is no clear consensus regarding the global nature of this variability across these timescales. We compare the magnitude and spatial consistency of surface temperature and precipitation variance distributions in the latest generation of Earth system models. We find that modeled temperature variance generally increases with timescale in patterns that are consistent across space, especially over most of the mid- and high-latitude oceans. However, there is little spatial agreement in precipitation scaling, even between different runs of the same model. Unforced temperature and precipitation scaling in Earth system models is difficult to distinguish from scaling generated by a simple autocorrelated process, but the magnitude of paleoclimate temperature reconstruction scaling falls well outside an autocorrelated process. Paleoclimate records of the last millennium indicate that Earth system models likely overestimate interannual variability and underestimate multi-decadal to centennial-scale climate variability, especially in the central and eastern tropical Pacific.