Reconstructing 15 000 years of southern France temperatures from coupled pollen and molecular (branched glycerol dialkyl glycerol tetraether) markers (Canroute, Massif Central)

Abstract. Climatic changes in southern Europe during the Holocene are characterized by a strong spatial and temporal heterogeneity whose patterns are still poorly understood, notably the presence or not of a Holocene thermal maximum (HTM; 10 000–6000 cal BP). The climatic patterns also differ according to the proxies used (e.g. pollen, chironomid) and the latitude of the record. Here, a multi-proxy approach combining pollen and lipid biomarkers (branched glycerol dialkyl glycerol tetraethers, brGDGTs) is applied to the Canroute sedimentological sequence (Massif Central, France) to reconstruct the climatic variation over the last 15 000 years in southern Europe. This area is poorly documented in terms of vegetation and climate change. To provide reliable climate reconstructions, we have (1) performed a multi-method approach applied to pollen (modern analogue technique, MAT; weighted averaging partial least squares regression, WA-PLS; boosted regression trees, BRT; and random forest, RF) and molecular biomarkers brGDGTs (five calibrations) and (2) investigated the role of modern databases and calibrations in climate reconstructions. Three different databases were tested for pollen data: one global database based on a Eurasian pollen database and two regional databases corresponding to Mediterranean–Temperate Europe and Temperate Europe–Scandinavian databases respectively. Five global calibrations were tested for lipid biomarkers including four for soil and one for peat. Results show that the use of different modern databases highlights the importance of considering environmental and ecological constraints when using transfer functions on pollen sequences. Pollen- and brGDGT-inferred climate trends are consistent, notably for the Late Glacial and the Early and Late Holocene. However, the reconstructions notably differ concerning the presence of a Holocene thermal maximum with the MAT pollen-based method, but no difference is apparent with the BRT pollen method nor brGDGT. The temperature reconstructions estimated from the two independent pollen and lipid proxies are then compared to regional climate signals (chironomids, pollen, molecular biomarkers) to better understand global regional climatic patterns in southern Europe. Altogether, our results from the Canroute sequence and those already available in southern Europe reveal that for the Late Glacial and Early Holocene, the regional climate trends are consistent between sites and proxies, supporting the reliability of their reconstructions despite some discrepancies. During the Holocene, the temperature signal of Canroute does not indicate the clear presence of a pronounced HTM, but rather stable temperatures