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 p
ollen samples considered as analogues of the fossil samples to those l
ocations where the implied change in annual precipitation minus evapot
ranspiration (P-E) is consistent with the regional change in moisture
balance as indicated by lakes. An artificial neural network was used f
or the spatial interpolation of lake-level changes to the pollen sites
, and for mapping palaeoclimate anomalies. The climate variables recon
structed were mean temperature of the coldest month (T-c), growing deg
ree days above 5 degrees C (GDD), moisture availability expressed as t
he ratio of actual to equilibrium evapotranspiration (alpha), and P-E.
The constraint improved the spatial coherency of the reconstructed pa
laeoclimate anomalies, especially for P-E. The reconstructions indicat
e clear spatial and seasonal patterns of Holocene climate change, whic
h can provide a quantitative benchmark for the evaluation of palaeocli
mate model simulations. Winter temperatures (T-c) were 1-3 K greater t
han 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 pre
sent 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 Eu
rope. P-E was 50-250 mm less than present in NW Europe and the Alps, b
ut alpha was 10-15% greater than present in S Europe and P-E was 50-20
0 mm greater than present in S and E Europe.