Evaluation of the CMIP5 palaeo-simulations using palaeoclimate reconstructions

S. P. Harrison, P. J. Bartlein, S. Brewer, G. Li, K. Izumi

    Research output: Contribution to conferenceAbstract

    Abstract

    Past climates provide a test of how well state-of-the-art models predict climate change. We present a comprehensive evaluation of CMIP5 models against Last Glacial Maximum and mid-Holocene climates, using an updated global synthesis of quantitative reconstructions of seasonal land and ocean climates and multiple metrics to take account of uncertainties in the reconstructions and simulations. Several of the large-scale responses to changes in anthropogenic forcing are also found in simulations of past climates, including enhanced land-sea contrast and high-latitude amplification in temperature changes, covariation in temperature and precipitation changes, and differential sensitivity to annual forcing resulting in changes in seasonality. These large-scale responses are consistent with the palaeo-record, which suggests that they are features of the real climate that are being simulated successfully by the models. Models generally capture the large-scale gradients of climate change in the LGM and MH, but they have more limited ability to reproduce either the spatial patterns or the magnitude of variations in climate change. Models both over- and under-estimate seasonal climate changes. At a regional scale, over-estimation of one variable can be related to underestimation of another, making it possible to infer the underlying causes of model biases. In the MH simulations, for example, underestimation of spring and summer warming is responsible for underestimation of the increase in monsoon precipitation whereas in regions north of the monsoon belt underestimation of precipitation leads to an overestimate of sensible heat fluxes and hence unrealistic increases in summer temperatures. Despite common biases, some models perform better than others in reproducing specific features of regional climates. Differences in model performance are only weakly related to differences in climate sensitivity.
    Original languageEnglish
    Number of pages1
    Publication statusPublished - 2012
    EventAmerican Geophysical Union Fall Meeting (45th : 2012) - San Francisco, CA
    Duration: 3 Dec 20127 Dec 2012

    Conference

    ConferenceAmerican Geophysical Union Fall Meeting (45th : 2012)
    CitySan Francisco, CA
    Period3/12/127/12/12

    Keywords

    • Paleo studies
    • Paleoclimate

    Fingerprint

    Dive into the research topics of 'Evaluation of the CMIP5 palaeo-simulations using palaeoclimate reconstructions'. Together they form a unique fingerprint.

    Cite this