What impacts long-term changes in biomarker-derived temperature in Holocene lake sediments: lake hydrology or watershed changes?

Branched glycerol dialkyl glycerol tetraethers (brGDGTs) are bacterial lipids that are used as paleotemperature and paleo-pH proxies. Developed originally for application in paleosoils, and geo-archives dominated by soil organic matter, they are also used for temperature reconstructions in lake sediments. For this, brGDGT ratios such as the MBT'_5ME are translated into mean annual air temperatures using linear transfer functions. However, water depth (Yao et al., 2019; Stefanescu et al., 2021) has recently been shown to influence the MBT'_5ME values in freshwater lakes. In addition, variable inputs of soil-derived GDGTs can skew the MBT'_5ME ratio values encountered in lake sediments. Unfortunately, the diagnostic ratios used to recognize either changes in water depth (HP5) or soil input (ΣIII/ΣII), are based on the relative abundance of the same 2 compounds (IIa and IIIa).

Currently, most of the work on environmental drivers of brGDGT lipids has been done on modern lake sediments. A view on the paleo-variability, i.e. the variability on brGDGTs in lacustrine archives, was still lacking. In this contribution, we will revisit several published and unpublished brGDGT records in last glacial and/or Holocene lake sediments that report a change in soil input and/or water level with time (e.g. Cao et al., 2021; Robles et al., 2022, Ramos-Román et al., submitted). We will use a compilation of these records to highlight how changes in hydrology and soil input influence brGDGT compositions. To distinguish between soil and lake-derived GDGTs, we will employ a novel machine learning approach (linear discriminant analysis). This method allows to identify soil and lake brGDGT distributions in modern soils and sediments (85% accuracy), and is now tested for the first time in paleolacustrine settings.

We show that natural or anthropogenic changes in the landscape can impact the diagnostic GDGT ratios for soil input and the MBT'_5ME ratio. The machine learning approach also allows to identify those depths where soil input is significant. This exercise is a first step in investigating the paleo-variability of brGDGTs with a machine learning approach, to determine variables that impact their downcore variability.

 

Cao J., et al. 2021. Lake-level records support a mid-Holocene maximum precipitation in northern China. Science China Earth Sciences 64, 2161–2171.

Ramos-Román, M. J., et al. Lipid biomarker (brGDGT)- and pollen-based reconstruction of temperature change during the Middle to Late Holocene transition in the Carpathians. Submitted to Global and Planetary Change.

Robles, M., et al. 2022. Impact of climate changes on vegetation and human societies during the Holocene in the South Caucasus (Vanevan, Armenia): A multiproxy approach including pollen, NPPs and brGDGTs. Quaternary Science Reviews 277, 107297.

Stefanescu, I.C., et al. 2021. Temperature and water depth effects on brGDGT distributions in sub-alpine lakes of mid-latitude North America. Organic Geochemistry 152, 104174.

Yao, Y., et al. 2020. Correlation between the ratio of 5-methyl hexamethylated to pentamethylated branched GDGTs (HP5) and water depth reflects redox variations in stratified lakes. Organic Geochemistry 147.