Summer warming is driving a greening trend across the Arctic, with the potential for large-scale amplification of climate change due to vegetation-related feedbacks (Pearson et al., 2013). Because observational records are sparse and temporally limited, past episodes of Arctic warming can help elucidate the magnitude of vegetation response to temperature change. The Last Interglacial (LIG; 129,000–116,000 years ago) was the most recent episode of Arctic warming on par with predicted 21st century temperature change (Otto-Bliesner et al., 2013; Post et al., 2019). However, high-latitude terrestrial records from this period are rare, so LIG vegetation distributions are incompletely known. Pollen-based vegetation reconstructions can be biased by long-distance pollen transport, further obscuring the paleoenvironmental record. Here, we present the first LIG vegetation record based on ancient DNA in lake sediment and compare it with fossil pollen. Comprehensive plant community reconstructions through the last and current interglacial (the Holocene) on Baffin Island, Arctic Canada, reveal coherent climate-driven community shifts across both interglacials. Peak LIG warmth featured a ~400-km northward range shift of dwarf birch, a key woody shrub that is again expanding northward. Greening of the High Arctic—documented here by multiple proxies—likely represented a strong positive feedback on high-latitude LIG warming. Authenticated ancient DNA from this lake sediment also extends the useful preservation window for the technique and highlights the utility of combining traditional and molecular approaches for gleaning paleoenvironmental insights to better anticipate a warmer future.

We extracted and analyzed sedimentary ancient DNA (sedaDNA) from a lake sediment core from Baffin Island, Arctic Canada, that spans the Holocene and Last Interglacial. We primarily used a metabarcoding approach, targeting the P6 loop of the chloroplast trnL (UAA) intron to identify vascular plant DNA through time. For select samples, we also prepared and sequenced single-stranded shotgun libraries. See "Materials and Methods" section of the publication for methodological details. 

The metabarcoding summary table, "CrumpEtAl2021_MetabarcodingSummaryTable.csv", includes post-filtered taxonomic results based on the metabarcoding methods outlined in the publication. The zipped .fastq files are demultiplexed (one file per sample) sequences with primers and adapters removed, but no further processing steps performed. See "README_CrumpEtAl2021_DNAmetabarcodingDataset.txt" and "README_CrumpEtAl2021_ShotgunDataset.txt" for additional usage notes.