Seasonal bias in temperature-sensitive biomarkers: a multi-proxy assessment of branched GDGT suitability for Holocene climate reconstruction

Understanding the evolution of Holocene climate is key for predicting what different futures may look like. However, global proxy and model-based climate reconstructions disagree on the general evolution of climate over the past 11.7 thousand years. Proxy-based reconstructions demonstrate a Holocene Climatic Optimum in the mid-Holocene, whilst model-based approaches show a trend of increasing temperatures throughout. This disagreement is largely believed to relate to seasonal biases within the proxy-based reconstructions, although model based-reconstructions are not without their flaws. Here we use a series of annually laminated (varved) lake sediment records from Europe (Diss Mere, United Kingdom; Meerfelder Maar, Germany; Lake Nautajärvi, Finland) to explore whether organic proxies are seasonally biased. To achieve this, we generate high-resolution (multi-decadal) branched glycerol dialkyl glycerol tetraether (brGDGT) lipid reconstructions of mean temperature of months above freezing (MAF; approximating MAAT in temperate locations) across the Holocene including the last 200-years and the mid-Holocene. We contrast our biomarker data with chironomid-inferred July summer temperature estimates from the same sample horizons within each lake to ascertain whether summer signals have an important imprint on the GDGT data. We show that brGDGTs are likely produced in situ within lake waters and that the varved nature of each lake does not impede brGDGT based climate reconstruction. We show that 1) GDGT-based temperatures record dominant climate variability at each site; 2) the mid-Holocene is warmer than present and pre-industrial mean annual temperatures; 3) biomarker and chironomid reconstructions from Diss Mere and Meerfelder Maar are more closely aligned than Nautajärvi suggesting location specific complexities; and 4) that biomarker and chironomid temperatures converge and diverge at various points in each record. Each of these results suggest seasonal biases exist within the GDGT-based climate reconstructions which may be non-stationary. Our data therefore reveals the need to generate multiple proxy-proxy assessments of climate from different archives to ascertain the influence of mean annual versus summer climate parameters.