A spatial reconstruction of Siberian Last Glacial Maximum climate from pollen data

The Last Glacial Maximum (LGM, around 21.000 years before present) was a period with significantly colder global mean temperature, large Northern Hemisphere ice sheets, and lower CO2 concentrations. Siberia was affected by a lower sea level which led to a closed Bering strait and a northward shift of the Arctic Ocean coastline. However, unlike other high-latitude areas, Siberia was not covered by a terrestrial ice sheet at that time. Climate simulations with LGM boundary conditions show large inter-model differences especially in Northern Siberia, which impede a direct analysis of Siberian LGM climate from simulations. Preserved pollen sample...  more

The Last Glacial Maximum (LGM, around 21.000 years before present) was a period with significantly colder global mean temperature, large Northern Hemisphere ice sheets, and lower CO2 concentrations. Siberia was affected by a lower sea level which led to a closed Bering strait and a northward shift of the Arctic Ocean coastline. However, unlike other high-latitude areas, Siberia was not covered by a terrestrial ice sheet at that time. Climate simulations with LGM boundary conditions show large inter-model differences especially in Northern Siberia, which impede a direct analysis of Siberian LGM climate from simulations. Preserved pollen samples provide information on the LGM vegetation and climate state.
We reconstruct mean temperature of the warmest month and mean annual precipitation from a network of pollen samples using weighted averaging partial least squares transfer functions. Combining these local reconstructions with a multi-model ensemble of PMIP3 climate simulations in a Bayesian framework, we obtain the first probabilistic spatial reconstruction of Siberian LGM climate. Our reconstruction shows less cold summer temperature anomalies along the coastlines than in Western Siberia, and a weaker precipitation reduction in Eastern Siberia compared to Western Siberia. Finally, we quantify the mismatch between each ensemble member and the proxy data. By examining common features among similarly performing simulations, we find that most simulations with a comparatively warm and wet Siberian climate score better.
Our data constrained state estimate is well-suited for future analyses of Siberian LGM climate. Potential applications include the investigation of reasons for the absence of a Siberian ice sheet during the LGM, the quantification of land-atmosphere carbon fluxes, and the estimation of past afforestation rates.