Discrepancies in lacustrine bacterial lipid temperature reconstructions explained by microbial ecology

Abstract

Bacterial lipid branched glycerol dialkyl glycerol tetraethers (brGDGTs) are a valuable tool for reconstructing past temperatures. However, a gap remains regarding the influence of bacterial communities on brGDGT profiles. Here, we identified two distinct patterns of brGDGTs from the surface sediments of 38 Tibetan Plateau lakes using an unsupervised clustering technique. Further investigation revealed that salinity and pH significantly change bacterial community composition, affecting brGDGT profiles and causing brGDGT-based temperatures to be overestimated by up to 2.7 ± 0.7 °C in haloalkaline environments. We subsequently used the trained clustering model to examine the patterns of bacterial assemblages in the global lacustrine brGDGT dataset, confirming the global applicability of our approach. We finally applied our approach to Holocene brGDGT records from the Tibetan Plateau, showing that shifts in bacterial clusters amplified temperature variations over timescales. Our findings demonstrate that microbial ecology can robustly diagnose and constrain site-specific discrepancies in temperature reconstruction. Salinity and pH-driven variations in microbial communities significantly affect bacterial lipid compositions, leading to temperature overestimations in reconstructions, based on machine learning clustering of lake sediment data and bacterial analysis.