Combining model results and paleodata using data assimilation

Observations and model results provide two complementary sources of information on past climate variations. The paleo -or proxy - data characterize the changes that occurred at a particular time while models can be used to infer the mechanisms responsible for those changes. Observations and model results are thus ideally used jointly and the first step before applying models to study past and future climate is to evaluate their ability to reproduce the signal recorded in the data. Paleodata and model results can also be combined more profoundly using data assimilation. While classical model-data comparison can only assess the skills of the model, data assimilation can guide the model results to have a better agreement with the real observed changes and thus to reproduce more accurately the processes at their origin.

The main challenges in paleo data assimilation will be reviewed here. Issues related to the generation of the model results, the model data comparison and the data assimilation technique itself will be addressed. Some recent achievements and perspectives will then be presented, first for spatial reconstructions based on data assimilation and secondly for the quantification of internal and forced climate variability as well as the role of atmospheric and oceanic circulation in past climate changes.