Evaluation of climate models using palaeoclimatic data

1 The fifth phase of the Coupled Model Intercomparison Project (CMIP5) is at present running simulations using state-of-theart models to provide information about the likely evolution of climate over the twenty-first century, with additional experiments to analyse the uncertainties inherent in these projections1. Models that perform equally well for present-day climate produce very different responses to anthropogenic forcing2,3. Palaeoclimate simulations of the Last Glacial Maximum, the mid-Holocene and the Last Millennium are to be included for the first time in the suite of CMIP5 simulations4. Simulations of times in the past when the change in natural forcing was of a similar magnitude to that projected for the next century — and for which abundant palaeoenvironmental data are available for evaluation — provide a unique opportunity to assess model performance outside the climate range for which current models have been developed and, in particular, beyond the calibration range for parameterization that models still necessarily incorporate5,6. The inclusion of the Last Glacial Maximum (LGM; around 21,000 years (21 kyr) ago) and mid-Holocene (around 6 kyr ago) simulations in the CMIP5 experimental design is a reflection of these periods having been major foci for the Palaeoclimate Modelling Intercomparison Project (PMIP; Box 1). This is because both the LGM and the mid-Holocene represent radically different climate states from the present-day and from each other, and have large natural forcings that are relatively well known (Supplementary Table S1). These simulations provide an opportunity to quantify the feedbacks associated with, for example, low carbon dioxide levels, large ice sheets (at the LGM) or vegetation changes (in both the mid-Holocene and LGM). Evaluation of model simulations against palaeodata shows that models can reproduce the observed direction and large-scale patterns of changes in climate at the LGM and mid-Holocene. This would not be possible unless the models incorporated both the Evaluation of climate models using palaeoclimatic data