Late Quaternary Climatic Change in France Estimated from Multivariate Pollen Time Series

Abstract

In regions like southern France, where usual analytical techniques are unsatisfactory due to the heavy influence of human activity and the existence of a complex climate, the quantitative reconstruction of climate from pollen is particularly difficult. That is why an original method has been performed. The first step of the method, based on best analogs estimation and multiple regression, is to calculate a relationship between climate (monthly temperature and precipitation) and modern pollen spectra for data from 182 sites. The result is called the “analog climate.” The second step of the method, an original combination of canonical correlation and principal components analyses, extracts the common information from several fossil pollen sequences to produce a so-called “paleobioclimate.” This step removes the effects of local differences in the vegetation among nearby sites and also reduces the effects of human disturbance. In the third step, the signals obtained from the first two steps are merged, by Kalman filter, into a final reconstruction where the noise is reduced by around 37%. The results, presented with a 95% confidence level, suggest that the Pleni-Würm was 7° to 13°C colder than present and had 20 to 60% less precipitation than today (this large confidence takes into account the climatic diversity of the region). At 13,000 yr B.P., the climate reached the modern level of precipitation but the temperature was just less than present. The subsequent decrease of precipitation is less important than that of temperature. This fact may explain the advance of alpine glaciers during the Younger Dryas. Preboreal warming was abrupt (3° to 4°C per 500 yr) and precipitation increased more slowly.