ALBERT

Description: The dataset represents pollen and sedimentary charcoal counting data as well as XRF and grain size data of the lacustrine sediment core recovered from Lake Maudit in Montagne d'Ambre (northern Madagascar, 1,250 m asl). From the center of the Lake Maudit, accessible from a peat bog, two parallel sediment cores (LM1A and LM1B) with lengths of 10.5 m and 10.75 m were recovered in 2017 using a Russian peat corer in June 2017 by Vincent Montade, Laurent Bremond and Sandratrinirainy Ranarilalatiana. For pollen and extraction, 0.5 cm3 subsamples at an interval varying between 8 and 48 cm on LM1B were treated with hydrochloric acid, hydrofluoric acid, acetolysis mixture and stored in glycerol. A minimum sum of 300 terrestrial pollen grains was counted for each subsample using a light microscope at 400x magnification. Pollen and fern-spore percentages were calculated on the terrestrial pollen sum. For charcoal-particle extraction, 1 cm3 of sediment was sampled every cm along the core LM1B and soaked in a 3% NaP2O4 solution plus bleach for several hours to deflocculate sediments and oxidize organic matter. Samples were sieved through a 160 μm mesh and charcoal particles were counted using a stereomicroscope at x40 magnification coupled to a digital camera. Semi-quantitative measurements of inorganic chemical elements were conducted using an ITRAX (CS-8) X-Ray fluorescence (XRF) core scanner with a molybdenum (Mo) tube at the Geomorphological–Sedimentological Laboratory of the Geomorphology and Polar Research (GEOPOLAR), University of Bremen. XRF scanning was conducted at 55 kV and 30 mA with 10 s of exposure time at 0.2 mm resolution from both sediment cores, LM1A and LM1B. These elements were normalized to the counts of incoherent radiation (“Mo inc”) derived from the XRF scanner, to account for lithological changes and sediment matrix effects. Using the XRF data, a Principal Component Analysis (PCA) was carried out, using the selected elements (Si, K, Ca, Ti, Mn, Fe, and Ni) as input variables. Prior to this multivariate statistic, the normalized elements were transformed using standard “z-transformation”. The PCA carried out with XRF data resulted in three main axes, with Axis 1 representing 61.5%, Axis 2 24.4%, and Axis 3 9.4% of the total data variance. For the grain size analysis ~1 cm3 subsamples at an interval of 5 cm were taken on LM1A and measured after destroying carbonates and organic matter according to standard protocols with HCl and H2O2. Measurements were carried out with a laser diffraction particle-size analyzer (LS 13320 Beckman Coulter) in seven cycles of 60 s each. The first reproducible signal was considered as reliable and final distribution data were calculated using the Fraunhofer optical model. Based on the lithological description (marker layers), the XRF-element patterns, and digital and radiographic images, LM1A and LM1B were parallelized and combined to a composite core (depth 1). In addition, several event-related deposits (deposited within only hours or maybe days) originating from the catchment have been identified. These events disturbed the normal sedimentation process and were removed to correct the master core depth (depth 2) before establishing the age-depth model. Between 0 and 208 cm the sediment core corresponds to peat sediment and below 208 cm to lacustrine sediment. Age-depth model have been only established on the lacustrine sediment section.

Description: Marine Isotope Stage (MIS) 5e is the latest period when global atmospheric temperatures were warmer than present and global sea levels were higher than present. Environmental conditions during MIS 5e therefore represent an important 'process analogue' for understanding the climatic mechanisms and responses active under future anthropogenic warming. MIS 5e diatom assemblages were reconstructed for nine Southern Ocean marine sediment cores (ANTA91-8, ELT17-9, MD03-2603, NBP9802-04, PC509, TPC287, TPC288, TPC290 & U1361A). A modern analog technique (MAT) transfer function (Crosta et al. 1998) was applied to the diatom assemblages to reconstruct the MIS 5e September sea-ice concentration (SIC) and summer (January – March) sea-surface temperature (SSST) for each sample. The MAT compares the relative abundances of 33 diatom species in each MIS 5e sample to the abundances of the same species in a modern reference dataset composed of 257 surface sediment samples (modern analogs) from the SO. Modern conditions for each surface sediment sample are interpolated on a 1o x 1o grid, with SSSTs from the World Ocean Atlas 2013 (Locarnini et al., 2013) and September SIC from the numerical atlas of Schweitzer (1995). The MAT was implemented using the “bioindic” R-package (Guiot and de Vernal, 2011), with chord distance used to select the 5 most similar modern analogs to each MIS 5e assemblage. A cut-off threshold, above which any analogs are deemed too dissimilar to the MIS 5e sample, is fixed as the first quartile of random distances determined by a Monte Carlo simulation of the reference dataset (Simpson, 2007). The reconstructed SSSTs have a Root Mean Square Error of Prediction (RMSEP) of 1.09 oC and an R2 of 0.96, and the reconstructed September SICs have a RMSEP of 9 % and an R2 of 0.93.