Résumé
Palaeoclimates across Europe for 6000 y BP were estimated from pollen data using the modern pollen analogue technique constrained with lake-level data. The constraint consists of restricting the set of modern pollen samples considered as analogues of the fossil samples to those locations where the implied change in annual precipitation minus evapotranspiration (P-E) is consistent with the regional change in moisture balance as indicated by lakes. An artificial neural network was used for the spatial interpolation of lake-level changes to the pollen sites, and for mapping palaeoclimate anomalies. The climate variables reconstructed were mean temperature of the coldest month (T-c), growing degree days above 5 degrees C (GDD), moisture availability expressed as the ratio of actual to equilibrium evapotranspiration (alpha), and P-E. The constraint improved the spatial coherency of the reconstructed palaeoclimate anomalies, especially for P-E. The reconstructions indicate clear spatial and seasonal patterns of Holocene climate change, which can provide a quantitative benchmark for the evaluation of palaeoclimate model simulations. Winter temperatures (T-c) were 1-3 K greater than present in the far N and NE of Europe, but 2-4 K less than present in the Mediterranean region. Summer warmth (GDD) was greater than present in NW Europe (by 400-800 K day at the highest elevations) and in the Alps, but >400 K day less than present at lower elevations in S Europe. P-E was 50-250 mm less than present in NW Europe and the Alps, but alpha was 10-15% greater than present in S Europe and P-E was 50-200 mm greater than present in S and E Europe.