Methodology:

A series of peat cores were extracted from Site PB1 (Puise Bog site 1) on Marion Island, on 2016-12-26.

ITRAX XRF core scanning (XRFCS) was used to track inputs of minerogenic aerosols into the peat, focusing on Ti as the most widely used indicator of allochthonous inputs.

All cores were measured on the ITRAXTM XRF core scanner at Aberystwyth University using the same settings and type/make of X-Ray Mo-tube. Contiguous and non-destructive downcore XRF-CS was then undertaken using the same core scanner and a Molybdenum (Mo) anode X-ray tube (settings: 30 kV, 50 mA, count time 10 seconds per 1 mm distance travelled by the detector). Raw count per second (cps) XRF-CS data were produced using Cox Analytical Q-spec software v8.6.0, aiming for as low mean square errors (MSE) as possible, which is indicative of an optimal fit between 'as measured' and modelled spectra.

XRF-CS measurements were made over a period of four years between 2015 and 2019, shortly after cores had returned from each of the field sites. In that time, a new detector with higher count rates was fitted to the Aberystwyth ITRAX core scanner in 2016. To account for this, and other internal differences in cps caused by downcore variations in count rates, tube ageing, and variable matrix density, water and organic content, raw cps data should be normalised by incoherent scatter, calculated as percentages of total element and scatter cps sum (percentage cps) (Saunders et al., 2018), and/or centred log ratios (clr) - these datasets have been constructed and are available and code to transform the cps dataset as described above can be found in the following link: https://github.com/stever60/ACE_peat_calibration. Incoherent and coherent scatter ratios can be used as proxies for changes in dry mass (coh./inc.; following Boyle et al., 2015) and have been linked to changes in water (and organic) content (inc./coh) (Davies et al., 2015; Bertrand et al., 2024).

For GEOTEK MSCL data, intact half cores were slowly defrosted while vacuum-packed to minimise shrinkage and then analysed with a non-destructive Geotek® multi-sensor core logger (MSCL) (Gunn and Best, 1998) to obtain gamma-ray wet density (gamma-density; GRD),resistivity and magnetic susceptibility (MSK; SIx10-5) data (Bartington Instruments: MS2E point sensor, 2 or 5 mm interval; 10 seconds measurement time) and volume specific (density-corrected) MSx (K/P; kg m-3) using Geotek Software. Age models were constructed using the RBacon and the 2020 radiocarbon calibration curve using the R package BACON v. 2.5 for Bayesian age-depth modelling (Blaauw and Christen, 2011). Data for the age-depth models is published elsewhere, only outputs from specified depths are included in this dataset.

Following XRF-CS and MSCL analysis, cores were frozen and sliced at ~1 cm intervals using a stainless-steel band saw, divided into three sub-samples (for density/geochemistry, dating/macrofossil and archive) using ceramic knives and Teflon boards, and stored frozen in plastic Ziplock bags. Water content and dry bulk density were calculated by weighing samples before and after freeze-drying, and by measuring the dimensions of each fresh density/geochemistry sample using a vernier calliper30.

Subsamples for ICP-MS were prepared following the method in Le Roux and De Vleeschouwer (2010). Dry samples were homogenised in 15-50 ml falcon digitubes with eight glass beads of 4 mm using a FastPrep-24 (3x20 s at 6 m s-1) and then digested in a class 100 clean room at CRBE (Toulouse, France) following an established protocol whereby 100 mg of bulk sample was digested in Teflon vials on hot plates using an ultrapure HNO3-HF mixture (Optima ThermoFisher) at 110degreesC for 48 h followed by evaporation steps (50degreesC) and re-digestion using an HNO3-HCl mixture if required. Samples were then evaporated, 2 ml of HNO3 were added and following proper ...(6)

Data collection:

XRF-CS measurements were made shortly after cores had returned from each of the field sites. Contiguous and non-destructive downcore XRF-CS was then undertaken using the same core scanner and a Molybdenum (Mo) anode X-ray tube (settings: 30 kV, 50 mA, count time 10 seconds per 1 mm distance travelled by the detector). Raw count per second (cps) XRF-CS data were produced using Cox Analytical Q-spec software v8.6.0, aiming for as low mean square errors (MSE) as possible, which is indicative of an optimal fit between 'as measured' and modelled spectra.

For ICP-MS, measurements were either done on a quadrupole ICP-MS equipped with a collision cell (Agilent Technologies 7500ce, Toulouse, France) or a Triple Quadrupole ICP-MS (iCap TQ, Thermoscientific) at Observatoire Midi Pyrenees (Toulouse, France). The ICP-MS instruments were calibrated using a synthetic multi-element standard while an InRe solution was used as an internal standard. Procedural blanks were regularly measured and showed negligible values.

Data quality:

ITRAX XRF-CS scanner performance checks were undertaken using a synthetic glass standard and (informally) using XRF fused glass discs from the Ardley Lake and Yanou Lake sediment cores from the maritime Antarctic South Shetland Islands, as these contain highly organic units (Roberts et al., 2017).

For ICP-MS, blanks and several organic Certified Reference Materials (CRM) were run to assess accuracy and reproducibility for each element. The CRM were selected to match the composition and concentration range from pure organic material towards a mix between organic and mineral matrix: NIST-1547a (peach leaves), NIST-1515 (apple leaves), GBW-07603 (bush branches and leaves), IPE-176 (Phragmites communis) and NJV-94-1 (peat).

Data resolution

1 mm XRF-CS data; ~1 cm.

ICP-MS and subsample data; 5 mm.

GEOTEK MSCL data; subsample data is ~1 cm as detailed in top and bottom depth measurements.