The increasing number of Earth system model simulations that simulate the climate during the last deglaciation creates a demand for benchmarking against environmental proxy records. Syntheses of data from marine and terrestrial sediment records for the same time period are emerging. Comparing these two data sources over a period with changing background conditions requires new methods for model-data comparison that incorporate multiple sources of uncertainty.

We first present a method for the probabilistic, multivariate quantification of the deviation between temperature reconstructions and paleoclimate simulations. We demonstrate the accuracy and robustness of our method with pseudo-proxy experiments. Then, we show preliminary comparisons of simulations from the ongoing PalMod project against a global synthesis of sea surface temperature reconstructions.

In the second part of the talk, we present a network-based approach to evaluate vegetation and hydroclimate in transient simulations against pollen indices. This approach aims at assessing large-scale properties of simulations. In addition, our analysis reveals differences in the climate-vegetation relationship of dynamical vegetation models.

Our work aims for standardized model-data comparison techniques. The introduced methods can be extended subsequently with additional proxy data, new simulations, and improved representations of proxy uncertainties. By learning from past climate transitions, our results can be used to optimize Earth system models and thus improve projections of future climate states.