Abstract
Paleoclimatologists have developed numerous indicators for estimating continental and oceanic surface temperatures. These paleothermometers are mainly based, on the one hand, on the distribution of plant species(e.g. pollen, diatoms) or animal species(e.g. foraminifera) whose habitat areas are restricted by temperatures, and on the other hand, on geochemical indicators (stable isotopes, trace elements and specific molecules) sensitive to temperature. Several decades of analyses of marine and lake sediments have made it possible to compile several hundred local postglacial records and map thermal variations for different latitude bands. These compilations clearly show an early postglacial warming in the Southern Hemisphere, contrasting with a complex evolution in the Northern Hemisphere, characterized by two cold transients centered around 16 000 and 12 000 years BP. On a global scale, mean global temperature rises significantly around 17 000 years BP, shows a transient reversal around 12 000 years BP, before stabilizing around 10 000 years BP.
Another approach is to use climate models perturbed by the climate forcings described in the first two lectures. These models simulate a glacial cooling of around 5 °C, followed by a unidirectional warming that continues throughout the Holocene. The absence of any marked global or regional transient suggests that conventional forcings are insufficient to explain the complexity of Tardiglacial climate change.
The separate consideration of climate forcings by models demonstrates the existence of systematic biases in the responses to these forcings. As discussed in previous lectures, the data-model mismatch in the early Holocene may be linked to a seasonality bias in paleoclimate indicators. Nevertheless, a recent reconsideration of modelling shows the sensitivity of simulated temperatures to the extent of pack ice in the polar regions.
An intermediate approach makes it possible to use both observed time series and models. This involves reanalysis by data assimilation, the principle of which is to correct the model's temporal simulation step by step, taking into account any discrepancies with trends observed during the previous time step. Logically, the first applications to deglaciation converge on an intermediate evolution compared to the independent series of models and data : a glacial cooling of around 6 °C on a global scale, followed by a marked warming from 17 000 years BP, a slight reversal around 12 000 and a stabilization around 8 000 years BP, without a Holocene optimum.
