Pollen preserved in lacustrine sediments form a crucial archive for reconstructing past climate and terrestrial vegetation change. Currently, climate reconstructions are based on statistical models describing the link between climate and pollen. These models raise methodological problems because they are all based on the hypothesis that climate-pollen relationships are constant over times; implying that non-climatic parameters driving the relation have weak influence. This is not in agreement with recent developments in ecology and ecophysiology. That is why, in this work, we develop an approach integrating a dynamical vegetation model and major processes linking vegetation and pollen trapped in lakes. The Bayesian framework provides us wit...