Glycerol dialkyl glycerol tetraethers (GDGTs), including both the archaeal isoprenoid GDGTs (isoGDGTs) and the bacterial branched GDGTs (brGDGTs), have been used in paleoclimate studies to reconstruct temperature in marine and terrestrial archives. However, GDGTs are present in many different types of environments, with relative abundances that strongly depend on the depositional setting. This suggests that GDGT distributions can be used more broadly to infer paleoenvironments in the geological past. In this study, we analyzed 1153 samples from a variety of modern sedimentary settings for both isoGDGT and brGDGTs. We used machine learning on the GDGT relative abundances from this dataset to relate the lipid distributions to the physical and chemical characteristics of the depositional settings. We observe a robust relationship between the depositional environment and the lipid distribution profiles of our samples. This dataset was used to train and test the Branched and Isoprenoid GDGT Machine learning Classification algorithm (BIGMaC), which identifies the environment a sample comes from based on the distribution of GDGTs with high accuracy. We tested the model on the sedimentary record from the Giraffe kimberlite pipe, an Eocene maar in subantarctic Canada, and found that the BIGMaC reconstruction agrees with independent stratigraphic information, provides new information about the paleoenvironment of this site, and helps improve paleotemperature reconstruction. In cases where paleoenvironments are unknown or are changing, BIGMaC can be applied in concert with other proxies to generate more refined paleoclimatic records.

not-yet-known not-yet-known not-yet-known unknown Ecotones are expected to be sensitive to climate change, but supporting evidence is scarce. Dense paleoecological data networks are ideal for studying past ecotonal dynamics because of their broad spatiotemporal coverage. Here we test the ecotonal sensitivity hypothesis by reconstructing movements of a major temperature-governed forest ecotone over the last 11,000 years in lower Michigan and comparing the variability of ecotone-proximal and distal vegetation. NMDS ordination indicates four meaningful axes of variation in Holocene Michigan vegetation, with axis 2 capturing the latitudinal ecotone and axis 4 differentiating by moisture sensitivity. The ecotone advanced northwards by 150 km from 11 to 6 ka, during a 3.5⁰CTJuly warming, then retreated. Ecotone-proximal sites have higher rates of compositional change than distal sites. These findings support the ecotonal sensitivity hypothesis, constrain estimates of climate-driven forest ecotonal movement, and reinforce concerns about the sensitivity of ecotonal ecosystems to anthropogenic global warming.

Changes in vegetation in North America indicate Holocene shifts in the latitudinal temperature gradient along the western margin of the North Atlantic. The dynamics of tree taxa such as oak (Quercus) and hickory (Carya) showed opposing directions of change across different latitudes, consistent with changes in temperature gradients. Pollen-inferred temperatures from 34 sites quantify the changes and reconstruct a long-term southward shift of the sharpest temperature gradient in winter and a northward shift in summer. During the mid-Holocene, however, an oscillation in tree distributions interrupted the trends indicating that the steepest portion of the seasonal temperature gradients migrated rapidly northward at 5.8-3.2 ka. The shift produced an unusually late summer thermal maxima at 42-43.5°N where oak abundance peaked both in the early and mid-Holocene. The changes appear consistent with orbital and ice sheet forcing as well as millennial variability in the North Atlantic pressure field during the mid-Holocene.