Reconstruction of climate anomalies for the Last Glacial Maximum (LGM, c.a. 21,000 years ago), made by combining pollen based reconstructions (from Bartlein et al. 2011) and averaged outputs of LGM simulations from the 3rd round of the Palaeoclimate Model Intercomparison Project (PMIP, Braconnot et al. 2011), under a variational data assimilation technique. This reconstruction is designed to be used for data-model comparison, specifically against the results of PMIP4. The dataset consists of 6 variables: moisture index (the ratio precipitation and equilibrium evapotranspiration), mean annual precipitation (mm), mean annual temperature (degrees C), mean temperature of the coldest month (degrees C), mean temperature of the warmest month (degrees C), growing degree days above 5 degrees C (day degrees C). The standard deviation of these variables is also given.

This version of the dataset has been withdrawn because of an error discovered in the data file. The method of accounting for CO2 change was being applied to all pollen observations, whereas it should not be applied to observations generated from model inversion. The corrected version of the dataset is published at http://dx.doi.org/10.17864/1947.206.