Steppe-tundra composition and deglacial floristic turnover in interior Alaska revealed by sedimentary ancient DNA ( sed aDNA)

When tracing vegetation dynamics over long timescales, obtaining enough floristic information to gain a detailed understanding of past communities and their transitions can be challenging. The first high-resolution sedimentary DNA ( sed aDNA) metabarcoding record from lake sediments in Alaska — reported here — covers nearly 15,000 years of change. It shows in unprecedented detail the composition of late-Pleistocene “ steppe-tundra ” vegetation of ice-free Alaska, part of an intriguing late-Quaternary “ no-analogue ” biome


Introduction
Interest in the full-glacial (ca.25,000-15,000 cal yr BP) landscape and vegetation of far-northern, largely unglaciated regions, which were open and herb-dominated, has inspired palaeoecological studies since the 1960's, particularly in eastern Beringia (e.g., Colinvaux, 1964;Ager, 1975;Cwynar, 1982;Anderson et al., 2004;Zazula et al., 2007;Mann et al., 2019;Monteath et al., 2023).At a time when most other northern high-latitude regions were covered by extensive ice sheets, greater Beringia comprised an unglaciated area extending from west of the Kolyma River eastward across the exposed Bering Land Bridge into Alaska-Yukon to the limit of the Cordilleran-Laurentide ice sheet complex in Canada (Hopkins, 1982;Dyke et al., 2002;Clark et al., 2009;Hughes et al., 2016).This region provided both a refuge and a centre of diversity for the high-latitude flora (Hultén, 1937;Murray et al., 1983;Ritchie, 1984), particularly arctic-alpine taxa, but also steppe and boreal taxa (Swanson, 2002;Abbott and Brochmann, 2003;Edwards et al., 2018).How the ecosystem responded to end-Pleistocene climate warming is of particular interest, given current pressures on northern ecosystems due to rapid environmental change (Myers-Smith et al., 2011;Monteath et al., 2021).
The nature of the full-glacial vegetation and the compositional changes that occurred during the transitional period from Pleistocene to Holocene also form a critical environmental context in the story of the megafauna; the period ca.15,000-10,000 cal yr BP saw both turnover and extinction in the mammalian fauna (Guthrie, 1990(Guthrie, , 2006;;Mann et al., 2013;Monteath et al., 2021 and references therein).Ecosystem changes, such as a shift to greater cover by woody taxa, plus changing populations of megafauna taxa, would also have affected resource use of early human populations, which entered the region at a transition from open to more woody-dominated landscapes shortly before 14,000 cal yr BP (Potter et al., 2017).In pollen records, the expansions, over a one-or two-century period, of Betula (ca.14,000 years ago) and Picea (ca. 10, 000) years ago, which represent successive transformations of Beringian landscapes, both suggest rapid vegetation adjustments, probably to major climate shifts.
Knowledge of past flora and vegetation has been limited by available techniques.The abundant (anemophilous) pollen of dominant woody taxa and variable pollen-taxonomic resolution in most records of past vegetation masks the full extent of taxonomic turnover in the nonwoody flora, and speciose plant macrofossil localities are, unfortunately, rare.This problem can be potentially addressed by a reconstruction technique that samples a relatively large, but spatially defined area-a hydrologic catchment-and resolves floristic detail, namely, sedimentary ancient DNA (sedaDNA) metabarcoding of lake sediments (see below).To this end, we analysed sedaDNA from Chisholm (also called Lost) Lake in the Tanana River Valley (Fig. 1) for the past ca.15,300 cal yr BP with the following aims: i) obtain a detailed floristic record of late-glacial herbaceous vegetation and interpret this in terms of palaeo-vegetation communities, ii) examine in more detail than previously possible floristic and functional plant turnover from late-glacial to Holocene and its relation to environmental changes, and iii) place Chisholm Lake (Lost Lake in Tinner et al., 2006); southern tip of Birch Lake to north; Tanana River floodplain (blue wavy lines) to west; Richardson Highway shown in red.Contour interval is 20 feet.c) Looking west from the eastern shore to the location of the intermittent outlet.d) Bathymetric map of the lake basin with 2-m contours, from Alaska Department of Fish and Game (http://www.adfg.alaska.gov).The dot marks the approximate Chis 17A-B coring site (see below).
C.L. Clarke et al. findings in the context of the landscapes encountered by humans and megafauna at the end of the Pleistocene.

Steppe, tundra, or steppe-tundra?
The nature of Beringian cold-stage treeless vegetation has been widely discussed (e.g., Colinvaux, 1964;Guthrie, 1982;Cwynar and Ritchie, 1980;Ritchie, 1984;Anderson et al., 2004).Full-and late-glacial vegetation has no strong modern pollen analogue (Anderson et al., 1989), but fossil records indicate it most likely comprised floristic and structural elements of both steppe and herbaceous tundra.For example, based on modern Alaskan distributions, prostrate Salix and the dwarf shrub Arctous indicate tundra, as do forbs such as Draba, Papaver and Pedicularis.Speciose grass genera, such as Festuca, Poa, and Agrostis, and other forb genera (e.g., Plantago, Galium) have representatives in both steppe and tundra, Given cold-stage vegetation covered a topographically complex region with substantial climatological gradients (Bartlein et al., 2015), several authors have argued, based on diverse pollen and macrofossil records, that there was local and regional habitat heterogeneity, with vegetation ranging from fellfield through forms of dry tundra to boreal steppe (e.g., Ritchie and Cwynar, 1982;Murray et al., 1983;Lloyd et al., 1994;Goetcheus and Birks, 2001;Kienast et al., 2001Kienast et al., , 2005;;Zazula et al., 2006a;Monteath et al., 2023).Today in interior Alaska and Yukon, elements of this past treeless mosaic are reflected in contemporary azonal facies of the vegetation, ranging from south-facing hill slopes and river bluffs in interior Alaska and the Yukon that support boreal steppe (Murray et al., 1983;Lloyd et al., 1994;Walker et al., 2001;Zazula et al., 2006a), to a rare steppe-tundra ecotone that occurs at mid elevation on south-facing mountains (Edwards and Armbruster, 1989;Lloyd et al., 1994), to alpine tundra and low-productivity fellfields on exposed uplands (Cwynar, 1982).
The term "mammoth steppe", initially coined by Guthrie (1982) to describe a no-analogue ecosystem with large grazers, is now widely used."Steppe" is generally taken to imply (relatively) high vegetation productivity, whereas low pollen productivity is indicated by pollen accumulation rates in sites in eastern Beringia (e.g., Bigelow and Edwards, 2001).This 'productivity paradox' (Hopkins et al., 1982;Yurtsev, 2001) remains largely unresolved, as biomass (both mammalian and plant) is hard to estimate from fossil data.An ecological assessment of detailed past floristic records such as that of Ritchie and Cwynar (1982), which can now be done using molecular techniques, may go some way to constraining biomass estimates.

The potential of a sedaDNA approach
Ancient DNA extracted from sediments provides an important source of information on floristic composition and diversity; data have been shown to be consistent in relation to pollen and macrofossil records (see, for example, Alsos et al., 2016;Pedersen et al., 2016;Zimmermann et al., 2017;Clarke et al., 2020).Lake sedaDNA is almost entirely derived from the local hydrologic catchment and therefore avoids uncertainties of long-distance origin by wind transport, as is commonly found in pollen studies (Alsos et al., 2018, Sjögren et al., 2017).The overall higher taxonomic resolution (Parducci et al., 2017;Clarke et al., 2020) can highlight floristic patterns not discerned via pollen.
In two related sedaDNA studies of 242 and 535 samples, respectively, from across the circumpolar Arctic, (Willerslev et al. 2014;metabarcoding) and Wang et al. (2021; shotgun metagenomics) presented a large-scale composite of full-glacial Arctic vegetation dominated by forbs and graminoids.The samples in these studies derive largely from terrestrial settings, they reflect different sedimentary environments and temporal periods, and many represent short snapshots in time with limited spatial coverage, meaning there is, inevitably, constraint on insight into plant ecology and regional floristic/vegetation dynamics; however, the studies clearly demonstrate the power of sedaDNA to provide floristically rich records and to combine them with key information on mammalian dynamics.In contrast, long, continuous sedaDNA records from key single sites reveal geographically precise details of temporal change.In a ca.25,000-year record from the Polar Urals, Clarke et al. (2019) showed that taxonomic diversity increased, particularly over the deglacial period (ca.17,000-9000 cal yr BP), and also that species were not lost from the extensive, altitudinally varied catchment through time as vegetation became more complex and heterogeneous.Rijal et al. (2021), combined multiple sites to show a similar trend of increased diversity across northern Fennoscandia after regional deglaciation.
In eastern Beringia, at two locations, sedaDNA studies of lake sediments both used a shotgun metagenomics approach.Graham et al. (2016) and Wang et al. (2017) traced the demise of the isolated Holocene population of woolly mammoths (Mammuthus primigenius) on St. Paul Island and associated environmental changes.Murchie et al. (2021a, b) used a targeted-capture technique to retrieve sedaDNA from multiple permafrost sites in Klondike, Yukon, to form a composite sedaDNA record spanning the last 30,000 years.It showed steppe-tundra vegetation giving way to woody shrub-tundra taxa between 13,500-13, 200 cal yr BP, coincident with change in associated mammalian and insect faunas (Monteath et al., 2023).Nevertheless, the various applications of sedaDNA in the region to date lack the specific combination of landscape-scale spatial and fine temporal and taxonomic resolution provided by a high-resolution metabarcoding record from lake sediment.

Interior Alaska in the late quaternary
Today, Eastern Beringia comprises Alaska and unglaciated northwest Canada.Between 18,000 and 15,000 cal yr BP, the climate of interior Alaska was probably cooler overall (Bartlein et al., 1991(Bartlein et al., , 2014) ) and certainly drier than present, with precipitation possibly half that of modern values (Barber and Finney, 2000).Lake levels in the Tanana Valley were extremely low; some basins may have been dry at times (Nakao and Ager, 1985;Abbott et al., 2000;Finkenbinder et al., 2014).Regionally, glaciers were restricted to the Alaska Range, plus a few, small glaciers in the Yukon-Tanana Uplands (Coulter et al., 1965;Kaufman and Manley, 2004).Deglaciation of the northern Alaska Range was early, with glaciers well up-valley by 14,600 cal yr BP (Briner et al., 2017).
With hemispheric deglaciation, a circulation switch from an anticyclone over the Laurentide ice sheet that maintained dry conditions to a more zonal, westerly flow brought higher levels of moisture (Bartlein et al., 2014) and rising lake levels (Abbott et al., 2000;Barber and Finney, 2000).Summer insolation and July temperatures peaked in the early Holocene (Kaufman et al., 2004), but the overall growing season lengthened in subsequent millennia (Bartlein et al., 2015).A transitional sequence from open to wooded vegetation ca.14,000-8,000 cal yr BP began with expansion of structurally diverse woody taxa such as Betula and Salix, followed by Populus, Picea and Alnus (Anderson et al., 2004).

Modern setting
Chisholm Lake (64.30229N, − 146.68623W) lies in the Tanana River valley at 256 m a.s.l.(Fig. 1a-d).The lake is accessible by road, and several houses and a camp are located near to the shore.Its surface area is ca.42 ha, and the bedrock catchment area is ca.780 ha.It is downstream of the much larger Birch Lake, in which water level is controlled by a weir.Currently, Chisholm Lake has a maximum depth of ca.11.4 m (lake data from Alaska Department Fish & Game (ADFG), 2023; Fig. 1d).The pH is slightly greater than 7.0, the water clear, but brown (MEE pers. obs. 2022;ADFG, date unknown).The west end of the lake, where there is an intermittent outlet, is formed by a dam of outwash sand and gravel deposited by the Tanana River during the late C.L. Clarke et al.Pleistocene (Ager, 1975).Slow-flowing inlets drain today from the east and from the direction of Birch Lake, and the eastern drainage extends the catchment area considerably.The size of the lake and the nature of its catchment affect the taphonomy of sedaDNA (Giguet-Covex et al., 2019, 2023).In Norway, Alsos et al. (2018) showed that for small lakes only a few ha in extent with limited inflowing streams, the plant sedDNA signal mainly derives from within 50 m of the lake shore and may be dominated by within-lake biomass.Chisholm is far larger, and has an inflowing stream (see above), in which case we may expect some components of the terrestrial sedaDNA flux to be derived from both nearshore locations and more distant points via slope-wash and stream transport (Giguet-Covex et al., 2023).
Extensive Quaternary sand and loess deposits occur throughout the Tanana valley (Péwé, 1975).Loess drapes the bedrock slopes to the north and the higher, more extensive hills to the south (ca.130 m above lake level), and the region is underlain by discontinuous permafrost (Jorgenson et al., 2008).Present-day climate conditions are characterised as boreal and continental, with mean July and January temperatures of 15.6 • C and − 20 • C, respectively, recorded at Big Delta (305 m a.s.l., ca.55 km SE of Chisholm Lake; Fig. 1a) for the period 1961-1990(WorldClimate, 2012)).The mean annual temperature is − 2.3 • C, and mean annual precipitation 304 mm.The catchment vegetation is boreal forest, dominated by Picea glauca (white spruce) with Betula neoalaskana (tree birch), Populus tremuloides (aspen) and a few isolated Larix laricina (tamarack) trees.In the flats south of the Tanana River, where permafrost is usually close to the soil surface, Picea mariana (black spruce) and Larix laricina dominate.Fire is a major driver of disturbance (Viereck et al., 1992).
A pollen and plant-macrofossil record from the lake (Tinner et al., 2006) documents low pollen concentrations between ca.15,000 and 13, 500 cal yr BP, prior to the establishment of Salix and Betula shrub-tundra.Around 15 terrestrial plant taxa, many at a low concentration, occur within the lowermost pollen zones; Poaceae, Cyperaceae and Artemisia dominate, with occasional forbs such as Potentilla, Bupleurum, Plantago canescens and Chenopodiaceae.The pollen record from this site provides a useful baseline for assessing the floristic accuracy of the sedaDNA metabarcoding approach.

Sediment retrieval and sub-sampling
Lake depth at the coring site (WGS84, 64.30229 N, − 146.68623W) was 10.5 m, as measured by a single-beam echo-sounder (Echotest II Plastimo).A 455-cm long sediment core (Chis17A-B) was retrieved in April 2017 from the winter ice surface using a Bolivia corer with 7.5-cmdiameter polycarbonate tubes for the uppermost sediments and 7.5-cmdiameter aluminium tubes for deeper sediments.For the basal sediments, a 5-cm-diameter modified Livingstone piston corer (Wright, 1967) was used.Due to logistic constraints, contiguous non-overlapping cores were taken.As this raises the possibility that material might be missing or that out-of-place material might find its way into the sediment at a core break, we avoided sampling near core breaks.Coring stopped when gravel was encountered; the lowermost retrieved sections were <10-cm long.Sediments were extruded, immediately wrapped, and placed in acrylonitrile-butadiene-styrene (ABS) pipe halves.
An exotic DNA tracer, comprising DNA extracted from a modern specimen of Schlumbergera truncata (Christmas cactus, family (Cactaceae)) mixed with DNA-free water (see Rijal et al., 2021), was sprayed on the surface of all coring equipment and core tubes.After retrieval of the sediments, all tubes and core sections were sealed immediately to minimise the risk of contamination by airborne or other modern environmental DNA.We did not detect tracer DNA in any of the samples analysed and conclude there was no detectable contamination from any of the equipment or core tubes used.
The core sections were transported and stored at 4 • C in the cold storage facility at the Alaska Quaternary Center, University of Alaska Fairbanks (UAF).Core sections were opened by longitudinal splitting in the summer of 2018.One half was used for subsampling, and the other half kept for archival purposes.The 4.5-m core was sub-sampled at 4-cm resolution, and 2-cm resolution within key sections of interest, within a dedicated ancient DNA clean-room facility at the University of Alaska Museum of the North (UAMN).Sterile tools, a full bodysuit, face mask and gloves were used.Following the protocol described by Parducci et al. (2017), the outer 10 mm of sediment was avoided or discarded, and a ca.2-g subsample retrieved from inside the freshly exposed centre.

Lithology, radiocarbon dating and age-depth modelling
Upon longitudinal splitting of the core sections (see above), a basic lithostratigraphy was created from visual inspection.Subsamples of 2 cm 3 were taken for loss-on-ignition (LOI) analyses at contiguous 1-cm intervals throughout the core using a volumetric sampler.Samples were weighed in crucibles, dried overnight at 100 • C and reweighed for dry weight determination, then ignited at 550 • C for 2 h, placed in a desiccator to cool to room temperature and reweighed.Total LOI was calculated as the percentage loss of dry weight after ignition (Heiri et al., 2001).Magnetic susceptibility was measured at 1-cm intervals on the archival core halves using a Bartington point sensor on a Geotek Ltd.Multi-Sensor Core Logger using a 10-s exposure time.
The preservation and abundance of plant macrofossil remains in the sediment core is poor.The individual plant remains picked for radiocarbon dating were too small for secure identification (most had a diameter of ca.1.0 mm) and had to be combined to meet the minimum sample weight for Accelerator Mass Spectrometry (AMS) radiocarbon dating.Some were likely of aquatic origin, which introduces extra uncertainty to the age-depth model, as the radiocarbon age of aquatic macrofossils may be affected by old carbon from the catchment (Abbott and Stafford, 1996).Six samples were analysed for radiocarbon dating at the Keck-Carbon Cycle AMS Facility, University of California, Irvine in the spring of 2019.A further five samples were analysed in the summer of 2020 at the Poznań Radiocarbon Laboratory of the Adam Mickiewicz University, Poland.
We developed an age-depth model for core Chis17A-B using the northern hemisphere IntCal20 calibration curve (Reimer et al., 2020) and the OxCal v.4.4.Bayesian statistical program (Bronk Ramsey, 2009a).A P_Sequence depositional model was constructed using a General_Outlier model and a variable k parameter (Bronk Ramsey, 2008, 2009b;Bronk Ramsey and Lee, 2013).One radiocarbon date (UCIAMS-219310) was removed from the age-depth model as it caused a substantial age reversal and thus was considered unreliable (Table A1).
To reduce the chronological uncertainty caused by the limited availability of reliable macrofossils for radiocarbon dating, we produced a second, integrated age-depth model using similar methods, but multiple records.We used sharp pollen transitions (Betula rise and Picea rise) to link the Chis17A-B age-depth model with other AMS-dated records in the Tanana River Valley.New Bayesian age models were developed for Birch Lake (Bigelow, 1997), Jan Lake (Carlson and Finney, 2004) and Chisholm Lake (Tinner et al., 2006) and linked with the Chis17A-B age-depth model using the ' = ' operator (Bronk Ramsey, 2009a).Combining the chronological data allows the Bayesian models to inform each other.The timing of these rapid changes in the pollen records can reasonably be assumed consistent among sites, and any differences in timing are likely to be less than the two-sigma precision of the age-depth model (Figs.A1.1a and b).
Age and uncertainties are reported at two sigma (95.4%) uncertainty or as the median age throughout the manuscript using results from the integrated model.A further explanation of the age-depth models is available in Appendix A1.For the full OxCal code, see Supplementary File 1.

DNA extraction, amplification, library preparation and sequencing
DNA extraction followed a modified version of the Qiagen DNeasy PowerSoil® DNA Isolation Kit that included the addition of Proteinase-K and Dithiothreitol (DTT) to the cell lysis mix with Solution C1 and overnight incubation (Rijal et al., 2021).DNA was extracted from 120 sediment subsamples and 12 negative extraction controls, which contained no sediment and were used to monitor for contamination, within a dedicated ancient DNA clean-room facility in the Palaeoecology Laboratory at the University of Southampton (PLUS).Aliquots of DNA extract were then transferred to the ancient DNA facility at The Aurora Centre for Arctic Ecosystem Genomics, The Arctic University Museum of Norway, Tromsø, Norway, for sedaDNA metabarcoding data generation.Metabarcoding PCR amplification, PCR product pooling and purification, and sequencing protocols follow Rijal et al. (2021), unless otherwise stated.Each DNA extract and negative extraction control was independently amplified using uniquely tagged "gh" primers that amplify the trnL P6 loop of the plant chloroplast genome (Taberlet et al., 2007).Nine negative PCR controls, which contained no DNA template, and three positive ones (Rijal et al., 2021) were run alongside the sediment samples and negative extraction controls.Every sample and all negative controls underwent eight individual PCR replicates.Pooled and cleaned PCR products were then converted to three Illumina-compatible amplicon libraries using a modified version of the Illumina TruSeq PCR-free kit with unique dual indexes (Rijal et al., 2021).Each library was then sequenced on ~10% of a mid-output flow cell for 2 x 150 cycles on the Illumina NextSeq-550 platform at the Genomics Support Centre Tromsø (GSCT) at The Arctic University of Norway.

Sequence analysis and taxonomic assignments
Next-generation sequence data were analysed and identified using the OBITools software (Boyer et al., 2016; http://metabarcoding. org/obitools/doc/index.html) and further filtered with R v. 3.2.4.(R Development Core Team, 2016).Briefly, the paired-end reads were adapter-trimmed and merged with SeqPrep (https://github.com/jstjohn/SeqPrep/releases, v1.2).The merged data were demultiplexed according to the unique PCR tags with ngsfilter using a primer tag-sample lookup file (Supplementary File 2), and identical sequences were collapsed with obiuniq.Rare sequences, represented by two reads or less, and PCR errors were identified and removed with obigrep and obiclean, respectively.
Taxonomic assignments were performed using the ecotag program (Boyer et al., 2016) by matching sequences against a local taxonomic reference library (ArcBorBryo) comprising 815 arctic and 835 boreal vascular plant taxa, and 455 bryophytes (Sønstebo et al., 2010;Willerslev et al., 2014;Soininen et al., 2015).Sequences were then matched to a second reference library generated after running ecopcr (Ficetola et al., 2010) on the global EMBL database (release r143).To minimise any erroneous taxonomic assignments, only taxa with a match of 100% to a sequence in at least one of the reference libraries were retained.We further considered a taxon to be undetected in a PCR replicate if it was represented by fewer than 10 DNA reads in the entire dataset.Identified taxa were checked against the flora of Alaska (Hultén, 1968) and the Pan Arctic Flora (PAF) checklist (http://panarcticflora.org/).If a sequence identification was inconsistent across reference libraries, such sequences were checked against the NCBI BLAST database (http://www.ncbi.nlm.nih.gov/blast/) for alternative taxonomic assignments, and a biogeographically plausible taxon assigned where possible.In some cases, this meant a taxon at a higher taxonomic level was assigned.
The 0-8 PCR replicate dataset suppresses large variations in sequence read counts among samples but should nevertheless reflect the relative success of each sample.We assessed the success with which we recovered sequences of interest via the metabarcoding analytical quality (MAQ) score as described by Rijal et al. (2021).The 10 identified sequences with most reads were checked for the number of replicates in which they occurred (i.e., a top score of 1.0 reflects all 10 sequences occurred in all 8 replicates; lower scores reflect taxa missing from some replicates).We did not use the post-PCR lab controls for any data quality comparisons.Strong read-replicate concordance suggests that replicate number is linked to template abundance (Rijal et al., 2021).
Constrained clustering is typically used for objective zonation of biostratigraphic diagrams.We used a stratigraphically constrained cluster analysis (CONISS; Grimm, 1987) in TILIA (Grimm, 1992) v 3.0.1 (available at https://www.neotomadb.org/apps/tilia) with all 181 taxa (see below).To assess consistency of clustering among structural-functional groups, we also re-zoned the data for the following groups: woody, graminoids, forbs, and aquatics.The sedaDNA figures were made using TILIA and resemble conventional pollen diagrams, but abundance is based on 0-8 replicates per sample.Separate rarefaction of terrestrial and aquatic taxa was done on the dataset using PAST v 4.03 using default settings (Hammer et al., 2001).Original and rarefied data were used to describe changes and trends in floristic richness.The 0-8 dataset was also subject to detrended correspondence analysis in PAST to assess trends in compositional turnover (an indication of β diversity) across the record.

Lithostratigraphy and chronology
The basal 5 cm of the 455-cm core were composed of sand, a 2-cm diameter pebble, and terrestrial gastropod shells (family Succineidae), suggesting fluvial or possibly colluvial material the base of the lacustrine sequence.The lower part of the core, to 225 cm, comprises minerogenic sediments differentiated into the lowermost three units (Fig. 2).Grey silty sand grades upward to silt with some laminae, both with low LOI (2-3%) and high magnetic susceptibility.Unit 3 contains slightly more organic silt.At ca. 225 cm (9200 cal yr BP) there is an abrupt transition to silty gyttja with LOI values increasing upward (up to 27%) and lower but fluctuating MS values (Unit 4).Upward to the surface, sediment properties remain relatively stable.No visible tephra deposits were observed.
The composite and independent age-depth models differ subtly during the early Holocene around the pollen-rise linkages (Fig. 2, A1.1, A1.2).The Betula linkage results in a younger modelled age for this depth: ca.14,470 cal yr BP in the composite model, compared with 14,810 cal yr BP in the independent model.In contrast, the Picea linkage has little effect on the modelled age of the Picea rise itself.It does, however, result in older modelled ages (up to 470 years) for the sediments immediately above the linkage, which are loosely dated.This substantially reduces the model uncertainty around the sediment transition at 225 cm.The consistency between the composite age-depth model and the independent age-depth model at the Picea linkage suggests that Holocene portion of the Chisholm Lake radiocarbon chronology is not strongly affected by old carbon.

SedaDNA data quality
We obtained around 47 million raw DNA sequencing reads from sediment samples, of which 32 million could be assigned to a sample (Supplementary File 3).Following post-identification filtering, million reads remained, representing 167 vascular plant and 14 bryophyte taxa (Supplementary File 4).Of these 181 taxa, 45% were identified to the species level, 39% to genus and 16% to higher taxonomic levels.Most samples contained 100-200 total PCR replicates.All MAQ values were high, except for 13 problematic samples (Table A2, Supplementary File 3, and see below).Aquatic macrophytes account for up to 93% of all reads across samples and were thus treated separately for rarefied richness estimates.
Between 141 cm and 56 cm depth, 11 sediment samples failed (i.e., no plant taxa were detected by sedaDNA).Failed samples were Fig. 2. Left: Lithology, loss-on-ignition, and magnetic susceptibility of the Chisholm Lake core.Middle: Age model incorporating sequences of dates tied by Picea and Betula pollen rises, from Chisholm (Lost) Lake (Tinner et al., 2006), Birch Lake (Bigelow, 1997) and Jan Lake (Carlson and Finney, 2004).Right: Age model based only on this dated core.processed in two different DNA extraction batches that each also contained successfully amplified samples.There is no obvious explanation arising from sediment properties, data generation, or bioinformatic issues.Possibly, there was a problem with DNA extraction (e.g., inhibition that reached a lethal threshold in some samples, but not others, at this depth window).These samples were omitted from the datasets used for plotting and statistics.The continuity of the record for ca.6000-1500 cal yr BP is somewhat compromised by these failed samples.In addition, samples at 101 cm and 350 cm had extremely low detections across all eight PCR replicates (13, 7 respectively).They were retained in plots as they nevertheless recorded the presence of multiple taxa (11, 5 respectively), but they were excluded from statistical analyses.We note that the MAQ scores for these two samples (MAQ of 0.10-0.15)are still substantially higher than the negative controls (all with MAQ of <0.07) (Supplementary File 3).
The full list of molecular taxa reported in Supplementary File 4 includes likely contaminants (e.g., common food species) and out-of-range taxa (according to Hultén, 1968).Due to sequence ambiguities, Hulteniella was placed in Anthemidae-2 (which includes Achillea and Tanacetum) and Vaccinium ovalifolium reclassified as the similar and biogeographically reasonable V. uliginosum.

Temporal record of floristic change
Key compositional/ecological changes for terrestrial and aquatic taxa from ca. 15,000 cal yr BP to present are summarized by a suite of key taxa (Fig. 3; Figs.A3a-d show records of all terrestrial taxa).For the sample-taxon two-way table see Supplementary File 5. Zone boundaries were defined using all taxa; boundaries defined for forbs and for aquatics separately tended to differ slightly from the all-taxon boundaries, which reflected change in tree and shrub taxa.Diagrams are plotted against depth to facilitate viewing of the early record.
Forbs and graminoids dominate the earliest period.Of all taxa identified in these groups, 54% and 78%, respectively, are detected in this zone.High values (4-8 repeats) are shown by the grass taxa Agrostidinae, Bromus, Festuca, and Puccinellia.Saliceae (comprising Salix and Populus), Dryas and Equisetum are abundant and present in all samples.Potamogeton, Myriophyllum sibiricum, Stuckenia and Zannichellia palustris form a distinct aquatic assemblage (Fig. 4).
Betula and several new aquatic taxa are temporarily abundant in several samples at ca. 15,100 cal yr BP (these samples are not near a core break, but they resemble coring contamination; Fig. 3).Picea and Alnus alnobetula (=viridis) are sporadically detected (1-2 PCR replicates).4. 14,000 cal yr BP) Distinct and abrupt increases in Betula (from this point a sustained presence), Arctous and Arctostaphylos uva-ursi occur at ca. 14,470 cal yr BP.Saliceae and Dryas remain consistently present throughout, while Puccinellia and Juncus decrease.Previously common forbs (e.g., Anthemideae, Anemone patens, Asteraceae, Potentilla, Pyrola sp. and Castilleja) continue.Lycopods and some bryophytes appear and/or increase, including Sphagnum russowii.Aquatic taxon richness increases overall, although abundances of new taxa fluctuate, or taxa disappear temporarily.
The main expansion of Picea occurs ca.9600-9500 cal yr BP near the change to silty gyttja in the lithostratigraphy and roughly synchronous with the almost compete disappearance of Arctous and Arctostaphylos uva-ursi.Several forb taxa continuing from previous zones also decline or disappear coincident with the Picea rise (e.g., Castilleja, Pedicularis, Pyrola).Sphagnum russowii values increase coincident with the Picea rise.Aquatic taxa (e.g., Nuphar polysepala, Isoetes, Callitriche sp., Nymphaceae, Sparganium) appear or increase and become common within this zone (Fig. 4).At 246 cm, Picea and Alnus alnobetula (=viridis) show high values.This sample is adjacent to a core break and the transient presence of these two taxa here probably reflects contamination (Fig. 3).

Zone CL4 (215-0 cm; 8380 cal yr BP to present)
The zone features a rise in Alnus alnobetula and Cornus spp., and Ribes and Myrica are consistently present.Sphagnum russowii (which may include other Sphagnum taxa) becomes abundant.Many taxa present in previous zones become rare or disappear (e.g., the dwarf shrubs, Arctous and Arctostaphylos uva-ursi, and forbs such as Lupinus, Pedicularis, and several Asteraceae groups), whereas new forb taxa appear (e.g., Mertensia, Ranunculus, Araceae and Linnaea borealis).Assemblages representing the period ca.4500-1500 cal yr BP are more widely separated temporally than in other parts of the record (this section contained the failed samples).At ca. 1500 cal yr BP, Larix is last arboreal taxon to enter the record, but otherwise terrestrial assemblages remain compositionally similar.This zone is marked by the disappearance of the aquatics Stuckenia and Myriophyllum sibiricum (Fig. 4).

Changes in lake status and aquatic taxa
Large late-Quaternary lake-level changes reported for neighbouring Birch Lake by Abbott et al. (2000) should also have occurred in Chisholm Lake, as the two are hydrologically connected.Major shifts of aquatic taxon composition at Chisholm occur at times of recorded lake-level rise at Birch (Fig. 4).At the first rise of lake-level (at 14,500 cal yr BP, or earlier) Zanichellia almost disappears and Hippuris and Utricularia vulgaris join the small group of early dominants (Potomageton, Myriophyllum sibiricum, Stuckenia).Most notably, the second major rise to near modern levels at the onset of zone CL-3 (ca.11,000 cal yr BP) sees the number of aquatic taxa increase further, and semi-aquatic taxa, such as Comarum and Cicuta, appear.

Relative importance of structural-functional groups
The relative number of read counts shows more clearly terrestrial taxa/functional groups that dominate the relative biomass signal or are possibly dominant for other reasons (aquatics are excluded as they represent >90% of all reads).Dominance changes over time (Fig. 5), with most groups showing a step-change in values ca.6000 cal yr BP.Saliceae reads dominate until this point and may over-estimate its biomass contribution (see below).After the early, herb-dominated period, graminoids decline while dwarf shrubs increase, and forbs remain relatively important.Notably, the combined contribution of graminoids and forbs to total DNA reads remains as high as 30% between ca.10,000 and 6000 cal yr BP, the early part of the forested period that is dominated by reads of trees and tall shrubs.

Diversity change
Terrestrial patterns of richness show high-frequency variability, but little overall trend.Lower values occur ca.13,500-12,000 cal yr BP, then increase to a maximum ca.11,000 cal yr BP, but the trends are slight (Fig. A4a).In contrast, aquatic taxa show two declines in the earlier part  of the record and peak richness ca.7000 cal yr BP (Fig. A4b; the same pattern is retained with and without facultative aquatics).While the DCA analysis of all taxa violated the normality requirement because of skewness, as a rough guide to turnover, DCA axis one, which accounts for 52.3% of the observed variation, spans over six standard deviations (>600 units) from the earliest to latest millennia, reflecting almost complete turnover in the flora from the start to the end of the record; ca.11,000-7000 cal yr BP was a period of particularly rapid change (Fig. A4c).

Lithostratigraphy
Relatively high and primarily minerogenic sediment input early in the history of the lake is likely to reflect several processes.First, treeless vegetation would have produced little organic detritus while enhancing runoff.Second, high loess deposition rates in the Tanana valley in the late glacial and early Holocene (Péwé, 1975;Muhs et al., 2003) would have contributed airfall loess and in-washed silt.Second, greater aridity than today would have meant far lower lake levels, a reduced lake area and a relatively larger catchment area (see below and Abbott et al., 2000).The increase in sediment organic content at 225 cm (ca.9300 cal yr BP) is likely due to rising within-lake productivity related to increasing lake volume, greater organic input linked to the spread of evergreen boreal forest (see also Bigelow, 1997) and a continuing trend of landscape-scale paludification (Mann et al., 2002;Reuther et al., 2016).That the whole section has a high in mineral component may partly explain the excellent recovery of sedaDNA, as mineral particles provide good binding sites for sedaDNA fragments (Giguet-Covex et al., 2023).

Quality of the sedaDNA record
Many vascular taxa were detected at Chisholm Lake throughout the record (Fig. 3 and Figs.A3a-d; Supplementary File 4).While Chisholm resembles some other long northern records in the total number of taxa identified and richness per sample (e.g., Zimmermann et al., 2017;Clarke et al., 2019), in zone 1 (herb zone), samples had high MAQ scores (0.7-1.0; Table A2), and 21-53 taxa were identified (Figs.A3a-c).This quality and richness contrasts with other records, in which the lowermost samples are often characterized by low diversity (Zimmermann et al., 2017;Alsos et al., 2021;Rijal et al., 2021).This may reflect the number of samples analysed (but all samples are rich in this zone) or unusually good preservation conditions-minerogenic sediments and a substantial catchment-lake area ratio (given a small lake) may have enhanced DNA capture and preservation (see above).It is also likely that the comparatively high per-sample richness reflects diversity inherent in eastern Beringia, which was part of a continental-scale glacial refugium (Hultén, 1937;Abbott andBrochmann, 2003, Stewart et al., 2016).
Saliceae (the poplar and willow tribe within Salicaceae) is detected in every sedaDNA sample and nearly every PCR replicate in the record, and it accounts for up to 95 % of total terrestrial DNA reads between ca.15,000-9000 cal yr BP.Its dominance is a feature common to sedaDNA records from other northern late-Pleistocene and Holocene sites (e.g., Clarke et al., 2019;Alsos et al., 2021;Rijal et al., 2021;Murchie et al., 2021b).As it is also dominant in DNA records based on shotgun metagenomics (Wang et al., 2021), its abundance is unlikely to be related to PCR or sequencing bias.The high proportion of Saliceae and its numerical effect on read proportions supports our choice of the 0-8 abundance scale to indicate the importance of the other taxa.

Comparison with Chisholm (Lost) Lake pollen record
This new vascular plant sedaDNA record and the previous pollen stratigraphy of Tinner et al. (2006) from Chisholm Lake together provide an excellent opportunity to assess how sedaDNA augments and/or provides additional perspectives on the changing taxonomic composition of the late-Quaternary vegetation in interior Alaska.The overall zonation is similar; however, sedaDNA identified 60 additional terrestrial plant taxa not found as pollen, and the DNA was better able to resolve the identity of graminoids to species or genus level (Fig. A3a).Many molecular taxa also occur far more consistently than their pollen counterparts.For the earliest (herbaceous) zone, the pollen spectra are dominated by herbaceous taxa such as Poaceae, Cyperaceae, Artemisia, Potentilla, Bupleurum, Plantago canescens, Chenopodiaceae and Equisetum, with some Dryas and Salix, (Tinner et al., 2006), taxa also seen in the sedaDNA (Figs.A3b and c), but the sedaDNA record is floristically far richer.SedaDNA detected 48 forb taxa, most of which were not detected by pollen, likely due to low pollen productivity and/or low overall pollen concentrations (i.e., pollen sum <200 grains; 23 terrestrial plant taxa detected in total, many at <5 % of terrestrial pollen sum).

The record of terrestrial change
The sedaDNA data provide an exceptional record of floristic composition from the onset of deglaciation through the Holocene, allowing a close examination of the glacial/late-glacial herbaceous flora as it was prior to the terminal declines of now regionally extinct megafauna species and arrival of humans, and of the subsequent changes affecting landscape, vegetation and lake status that would have impacted humans and other mammals alike.While key changes that shape the pollen record-driven by the expansion of a set of woody taxa after ca.14,470 cal yr BP-are mirrored in the sedaDNA, we can observe detail of the transitions and the turnover of numerous taxa well represented in sedaDNA but not commonly observed in pollen.
Overall, sedaDNA records a more complex mixture of plant communities than can be interpreted from pollen, in particular, the herbaceous communities of the late glacial, the early woodland communities and the aquatics.The high number of taxa provide a robust basis for examining turnover and richness patterns as communities were replaced through time (Figure A4a and A4c).The high turnover shows that there has been replacement of taxa over time, while a relatively stable level of richness has been maintained (especially clear in the rarified curve, Fig. A4a).This differs from the post-glacial richness patterns at Lake Bolshoye Shchuchye (Polar Urals), where the large elevation gradient in the catchment (>1000m) allowed arctic-alpine taxa to move upslope as climate ameliorated and new taxa occupied lower elevations; here, richness increased with time (Clarke et al., 2019).
Key patterns are evident in Fig. 3 (Figs.A3a-d show further detail).In the earliest period before ca.14,000 cal yr BP (zone CL-1), the lake catchment was occupied by forbs, grasses, and mat-forming plants, particularly Dryas and to a lesser extent, Arctous.Equisetum was abundant.While Saliceae includes Salix and Populus species, Saliceae reads are likely to represent shrub-Salix species, not trees.Prostrate forms of Salix can comprise a large proportion of root biomass in tundra, and their remains are abundant in a full-glacial buried soil on Seward Peninsula (Goetcheus and Birks, 2001).It is also possible that larger willow shrubs occurred in catchment watercourses.
Dominant forbs included several groups of the Asteraceae (Anthemideae, which includes Artemisia, Tanacetum, and Achillaea), Anemone patens, Astragalus, Bupleurum, Potentilla, Hedysarum, Bistorta vivipara, Lupinus arcticus, Plantago, Eritrichiumand Oxytropis; less common taxa were Papaver, Draba, and Suaeda(Figs.A3b and c).Only three of the 21 graminoid taxa identified were absent from this early community; grasses, such as Agrostidinae, Bromus pumpellianus, Festuca, and Puccinellia, several Carextaxa and Juncuswere particularly abundant.Generally, dry-to-mesic habitats are indicated by the flora.While some forb taxa have primary tundra affinities (e.g., Arctous, Papaver, Draba), others are found today in tundra and on disturbed slopes or in warm, dry boreal steppe communities (Anemone patens, B. pumpellianus).Other identified genera contain species that can occur in, but are not limited to, boreal steppe, e.g., Bupleurum triadiatum, Plantago cancescens, Hedysarum alpinum, Potentilla hookeriana, and P. pennsylvanica).Anticlea elegans is indicative of boreal steppe on south-facing river bluffs in interior Alaska today (Murray et al., 1983;Edwards and Armbruster, 1989;Lloyd et al., 1994).While the grass taxa were generally identified at too high a taxonomic level to narrow down ecological affinities, the presence of arid or saline conditions is not incompatible with the occurrence of several members of Anthemideae and Puccinellia (Kienast et al., 2005;Willerslev et al., 2014).The catchment likely supported a mosaic of open, dry tundra communities and more steppe-like communities on dry, even arid, soils, which were possibly saline along the lake shore.Given low regional moisture, trees and shrubs may have been excluded by aridity as much as cold.While the DNA reference library used for identification, ArcBorBryo, mostly contains present-day arctic and boreal plant taxa, the Chisholm assemblages notably contain more steppic elements than other records using the same reference library (e. g., Alsos et al., 2020;Norway;Zimmermann et al., 2017;NE Siberia).
Much of the debate about the nature of steppe-tundra disappears if one considers modern landscape-scale variation in growth conditions.The complex topography of the Chisholm Lake catchment (Fig. 1b) most likely underpinned a mosaic of plant habitats during the treeless late glacial, which would have been most apparent prior to the Betula expansion.A range of slope-aspect combinations would have created variation in incident radiation and soil moisture (Fig. A5a).South-facing slopes abutting the north side of the lake (extending several meters lower than today due to low lake-level) would have provided a hot, dry habitat in the growing season, not dissimilar to that of boreal steppe habitats today (Murray et al., 1983;Edwards andArmbruster, 1989, Chytrý et al., 2019).Indeed, incident summer radiation would have been higher than today's, given orbital changes (Berger and Loutre, 1991).In contrast, steep, north-facing slopes south of the lake would have been far cooler and more likely to be dominated by dry tundra taxa.An indication of how strongly these topoclimates can differ is given by their equivalent latitude (EL), which is calculated based on incident radiation (Lee, 1964).In the catchment, depending upon slope and aspect, the modern SW-facing equivalent latitude can be as low as 46 • N and N-facing EL > 80 • N (see A5).While the effects of insolation at the ground surface will have been mitigated by air temperature, taxa with greatly differing ecological preferences could have grown in the catchment, their locations determined by topography, and, possibly, how long any winter snow patches persisted in spring.Kienast et al. (2005) report a wide range of habitat preferences in macrofossil assemblages at a locality in northern Siberia that also samples a local landscape mosaic.Today, the plant communities at the extremes of an elevation gradient on arid slopes range from steppe to tundra (Edwards and Armbruster, 1989;Lloyd et al., 1994).
More generally, this study adds to increasing evidence that the northern herbaceous floras of the full-and late-glacial periods in the North were characterised by a mixture of grasses and diverse forb taxa, as observed by Willerslev et al. (2014), Murchie et al. (2021b) and Clarke et al. (2019).Chisholm Lake's sedaDNA assemblages contain a greater forb component than is suggested by many pollen records.In metabarcoding, there may be some representational bias towards forbs (Yoccoz et al., 2012), and specific polymerases can affect taxon recovery (Nichols et al., 2018), but a more recent shotgun-metagenomics study, which avoids the biases inherent in Willerslev et al. (2014) while being based on the same sample set, shows a dominance of forbs throughout the late glacial (Wang et al., 2021).It therefore seems reasonable to assume that the biomass distribution of forbs vs. graminoids within the lake catchment is better represented by sedaDNA than by either local or regional pollen, as pollen spectra are weighted against entomophilous forbs and towards anemophilous, widely dispersed graminoids (and Artemisia).A vegetation mosaic of arctic-alpine dwarf-shrub and forb communities on the one hand and grass-forb steppe on the other, plus aridity and summer growing-season temperatures lower than modern, all argue that "average" landscape productivity in the Tanana valley was patchy and climatically limited-neither as low modern dry tundra nor as high as steppe.
5.2.2.Regional expansion of woody taxa, ca.14,000 cal yr BP At the start of zone CL-2, the expansion and steady presence of Betula after ca.14,000 cal yr BP marks a shift towards woody vegetation cover that occurred across much of eastern Beringia (Anderson et al., 2004).Saliceae could reflect new Salix taxa with larger growth forms entering the vegetation, but generally a lack of large woody remains from early in this period suggests shrub-not tree-cover (both Betula and Salix are structurally highly variable; Edwards et al., 2005).Greater moisture availability was likely the main driver, as circulation patterns changed during deglaciation, restoring a zonal airmass flow that increased moisture transport to eastern Beringia (Bartlein et al., 1991(Bartlein et al., , 2014;;Lora et al., 2016).The dwarf shrubs Dryas, Arctostaphylos and Arctous were also abundant, while many of the dominant forb taxa also persisted.Pyrola, Pedicularis and Fragariinae (probably representing Fragaria virginiana) increased, while some other forbs (e.g., Suaeda, Eritrichium, Draba) disappeared, and grasses and sedges became far less abundant (Fig. 3; Figs.A3a-c).
While new species entered the record and others declined, this transition is marked by relatively low turnover in the terrestrial flora, a pattern driven particularly by the persistence of many of the forbs that were dominant in the preceding period (Fig. 3, Fig. A3b and c), but the decline of grasses (largely represented in fewer PCR replicates per sample) suggests the herb layer changed in structure and composition.In Alaska today, herbaceous taxa typical of tundra and/or boreal steppe require high light conditions to persist (Wesser and Armbruster, 1991), so continued presence of diverse forbs suggests that the landscape remained partially open.Thus, Betula-Salix scrub would not have occupied the whole catchment, an important feature not revealed in pollen records.This transition, as recorded by sedaDNA, features a subtle change of vegetation cover and some hindrance to the spread of shrubs, which may have been related to still low moisture levels compared with today (Barber and Finney, 2000), plus unique regional and local geologic disturbance regimes such as high aeolian activity.Kielhofer et al. (2020) and Reuther et al. (2016) point to periods of high aeolian activity and sediment source availability for aeolian transport and sediment accumulation as a driver for maintaining patches of herbaceous vegetation into the early Holocene in the Tanana Valley.Furthermore, there may have been continued disturbance by megafauna, which remained in the region during this period (Olofsson et al., 2009;Bakker et al., 2016;Monteath et al., 2021).

Establishment of deciduous and then evergreen boreal forest
One of the most prominent vegetation changes in the record occurred between ca.11,000 and 10,000 cal yr BP (start of zone CL-3; Fig. 3).Populus expanded and remained consistently present thereafter, its presence as an important Holocene boreal forest component far clearer in the sedaDNA than in pollen records.Both P. tremuloides and P. balsamifera are common today in interior Alaska, and P. tremuloides also occurs at the edge of boreal steppe communities on dry, steep slopes (Lloyd et al., 1994).Concurrently, a suite of understorey and/or open-area shrubs appeared, forming a community resembling that of warm slopes today that are wooded or partially wooded with P. tremuloides , Juniperus, Shepherdia, Elaeagnus, and Arctostaphylos uva-ursi (Viereck et al., 1992 pp 93-94).Maleae (probably Amelachier alnifolia), Rubus spp., and Viburnum spp.were also part of this community.Arctous persisted and was joined by Vaccinium uliginosum.Lycopods and mosses in the Polytrichaceae expanded.Many forbs already present in the catchment continued and were joined by Chamaenerion angustifolium, which is typical of warmer slopes and is seral after fire (Fig. 3 Figs.A3a-d).These changes in understory and field layers are far less clear in pollen records.
At about 9500 cal yr BP, a further vegetation change took place, not clearly identified by zoning, but nevertheless critical for both the terrestrial and aquatic ecosystems, namely, the expansion of Picea, which marks a shift to evergreen forest cover.The sedaDNA taxon Picea includes both P. glauca and P. mariana; the pollen record of Tinner et al. (2006) indicates that P. glauca originally dominated, and that P. mariana expanded ca.5000 cal yr BP (according to their age model).At ca. 8300 cal yr BP (zone boundary of CL-4), there was an expansion of Alnus, along with the virtual disappearance of Shepherdia, Eleagnus and Maleae (Amelanchier).Later in the zone, many forbs that had persisted from the earlier open and partially open communities declined or disappeared, but several others appeared, indicative of a Picea forest field layer or wet habitats: Mertensia paniculata, Geocaulon lividum, Linnaea borealis, Iris setosa, Caltha natans, Sium suave and Cicuta.After the Picea expansion, shrubs of more moist forest habitats, such as Chamaedaphne, Myrica gale and Spiraea, appeared sequentially, and Ranunculaceae expanded, all of which indicates further moistening of the catchment landscape, at least near the lake shore, and possibly paludification along the valley floors of inflow drainages.
The period ca.5000-1500 cal yr BP has low temporal resolution due to failed samples (see above), but samples that are present have high MAQ values (Table A2).Little change is apparent in the forest vegetation until the expansion of Larix (=L.laricina) in the catchment ca.1500 years ago.Today Larix grows near the lake close to the eastern inflow drainage.

Boreal tree refugium?
Ice-free northern regions may have supported refugial populations of boreal trees during the last glacial interval (e.g.Provan and Bennett, 2008;Parducci et al., 2012a, b; but see Alsos et al., 2020).Limited genetic evidence suggests that Picea may have persisted within lowland areas of interior Alaska and the Yukon Territory (Anderson et al., 2006(Anderson et al., , 2011;;Zazula et al., 2006) report Picea remains dating to ca. 27,000 cal yr BP; however, fossil evidence for local presence of Picea during the last glacial maximum (LGM) and late glacial within Alaska is relatively weak (Hopkins et al., 1981;Brubaker et al., 2005;Edwards et al., 2014).Sporadic occurrences of Picea sedaDNA early in the record (Fig. 3), prior to its establishment as a forest dominant, lead to three explanatory hypotheses: i) actual presence in the catchment, ii) derivation from pollen, and iii) lab contamination.
The sedaDNA patterns are mirrored in pollen records from Birch and Chisholm Lakes (Ager, 1975;Tinner et al., 2006).Furthermore, tree taxa occur in other sedaDNA records during periods when it is unlikely that they were present at the site (e.g., Alsos et al., 2020, Norway;Clarke et al., 2020;northern Ural Mountains).Pollen from gymnosperms can contain some chloroplast DNA; this, theoretically, could be introduced into the lake sediment matrix (Parducci et al., 2017).However, the extraction of cpDNA from fossil pollen grains has proved highly challenging (Bennett and Parducci, 2006), as it requires an additional lysis step during extraction (Kraaijeveld et al., 2015;Parducci et al., 2017).Current thinking is that DNA extracted from sediments does not derive from pollen grains (Jørgensen et al., 2012;Pedersen et al., 2016;Sjögren et al., 2017;Wang et al., 2017).
At Chisholm, Picea, while recorded unambiguously above 250 cm, is present sporadically in exceedingly low read abundance in PCR replicates below 250 cm (1.8%), whereas it occurs in negative controls at 2-3%.Furthermore, Alsos et al. (2020) caution that the number of controls in a single study is typically insufficient to detect low-frequency contamination with certainty.The parsimonious explanation for the sporadic occurrence of Picea is contamination, and so the sedaDNA provides no evidence of late glacial conifer persistence.

Changes in lake status and biota
Given the high read counts and proportions (>90%) of aquatic taxa, the richness of aquatic and semi-aquatic taxa (26 vascular taxa) is perhaps unsurprising.Nevertheless, the record provides an exceptional picture of changes in the aquatic macroflora (Fig. 4).Until ca.14,000 cal yr BP, a restricted group comprising Potamogeton, Myriophyllum sibiricum, Stuckenia and Zannichellia palustris may indicate an early, pioneering assemblage, probably in clear, shallow-and possibly brackish-water (e.g., Van Vierssen, 1982).(Note: the basal sediments of nearby Birch Lake are calcareous and record substantial authigenic carbonate production; Abbott et al., 2000.)After ca.14,000 cal yr BP, coincident with the initial Birch lake-level rise, there was some turnover (e.g., Hippuris and Potomageton praelongus appeared then later temporarily declined, while Zanichellia disappeared).
Taxon richness fluctuated over the next several millennia, during which lake levels fluctuated at Birch Lake.After the Picea rise (ca.9500 cal yr BP), when Birch Lake rose to overflow level (Fig. 4), taxon richness approximately doubled (Fig. A4b).Subsequently, after the establishment of Picea, the aquatics Myriophyllum sibiricum, Stuckenia, Persicaria amphibia and Ceratophyllum declined or disappeared, but other taxa appeared.Isoetes suggests deeper (Edwards et al., 2000) and browner water, consistent with an increase in humic input to the sediment from a Picea-dominated catchment.Nuphar and Nymphaceae (=Nymphaea), and Sparganium typically occupy shelves with water depths <1-~3 m, and their abundance indicates that the large area of the lake currently shallower than 4 m (Fig. 1d) was under water by this time.At this point, the lake increased considerably in size and physical complexity, and both factors are likely to have driven the marked increase in richness and diversity (Fig. 4).On a broad geographic scale, in Siberia, aquatic taxon diversity is correlated positively with July temperature and negatively with conductivity (Stoof-Leichsenring et al., 2022), which is consistent with an overall growing-season length increase into the middle Holocene in Alaska (Bartlein et al., 2015); also, salinity levels would have declined with increasing lake levels.The Chisholm Lake record demonstrates that aquatic sedaDNA record can be an effective recorder of responses of a lake's macroflora to changing regional palaeohydrology and water chemistry.

Early human interactions with megafauna in interior Alaska
Chisholm Lake lies close to several key archaeological sites (Fig. 1a), including Upward Sun River, which contains the earliest human remains in northern North America found to date (11,500 cal yr BP; Potter et al., 2011;2014, 2017), and other localities such as Swan Point and Mead, which record well-dated human presence by ca.14,000 cal yr BP (Holmes, 2001).Thus, early humans entered the region on the cusp of the major environmental changes described here.At the end of the Pleistocene, human settlement systems and desirable ecological habitats were largely centred on large river valley bottomlands and adjacent upland ecotones, and they overlapped with the ranges of megafauna, including bison (Bison sp.), wapiti (Cervus canadensis), horse (Equus ferus), and woolly mammoth (Mammuthus primigenius) (Potter et al., 2011;Rowe et al., 2024).While more evidence regarding human hunting of horse and mammoth in eastern Beringia is desirable to clarify the extent of this practice, evidence from Swan Point (Fig. 1a) suggests woolly mammoth hunting occurred in the middle Tanana Valley around 14,000 years ago (Rowe et al., 2024).After ca.13,500 cal yr BP, bison and wapiti became the preferred large-mammal subsistence resources as woolly mammoth and horse populations declined to extinction (Guthrie, 2006;Potter, 2008, Potter et al., 2011).
Topographic heterogeneity at the landscape scale likely maintained a mosaic of deciduous shrubland/woodland and open areas over a ca.5000-yr period (ca.14,500-9500 cal.yr.BP), during which key subsistence herbivore populations-particularly bison and wapiti-persisted in the Tanana Valley (Potter, 2008).By ca.9500 cal yr BP, a further increase in moisture availability, expansion of Picea glauca C.L. Clarke et al. (white spruce) forest, development of thicker organic soils, and a reduction in aeolian activity (Reuther et al., 2016), coincided with significant mammalian turnover, primarily the decline of bison and wapiti and persistence of moose (Guthrie, 2006;Monteath et al., 2021).Peat and muskeg formation began and accelerated over this period (Jones and Yu, 2010 ).By 6000-5000 cal yr BP, there was partial replacement of Picea glauca forests by P. mariana (black spruce) forests (Anderson et al., 2004), which led to a change in fire regime (Lynch et al., 2004).This coincided with marked changes in the archaeological record: a shift from the long-term conservative cultural practices of the Denali tradition (12,500-6000 cal yr BP) to a more broad-spectrum strategy based on caribou, small mammals, and later salmon and other fish-the Northern Archaic tradition (6000-1000 cal yr BP; Holmes, 2001Holmes, , 2008;;Potter, 2008;Potter et al., 2023).

Conclusions
The sedaDNA record from Chisholm Lake in interior Alaska resolves the composition of late Quaternary plant communities far more effectively than pollen, particularly from the Betula rise onward, when pollen records become dominated by a few anemophilous pollen producers.The late-glacial herbaceous plant community was floristically diverse, and the greater insight into minor taxa afforded by the sedaDNA record points to the flora comprising tundra taxa and more steppic elements across a topographic (topoclimatic) mosaic.The record shows a large, distinct and sustained floristic change ca.11,000 cal yr BP, when first Populus and then Picea became established, along with a number of other woody taxa and boreal herbs often not observed in pollen records.Additionally, sedaDNA affords a detailed picture of aquatic plant turnover, which in turn reflects lake-level dynamics and lake water chemistry.
The analysis of sedaDNA makes an important contribution towards resolving the composition of the full-glacial to late-glacial flora within ice-free northern regions.It indicates precise establishment dates of taxa in a catchment, avoiding the issue of long-distance pollen transport, and it highlights palynologically "silent" taxa, such as Larix, which currently have a largely undocumented history in Alaska.Furthermore, continuing developments in the field promise that expanded reconstructions of ecosystem dynamics, including other key biological groups such as vertebrates, invertebrates, and microorganisms, together with new proxies for the abiotic environment, can greatly enhance understanding of how northern ecosystems have responded to past climate change, and in turn, how these changes shaped the human environment and mammalian population dynamics.

Datasets
Raw sedaDNA sequence data are available at the European Nucleotide Archive (ENA) under BioProject accession code PRJEB74037.For the OxCal code for the age model see Supplementary File (SF)1.The primer tag-to-sample lookup file required to allow data demultiplexing is SF2.For outputs from the bioinformatic analyses, including sample metadata, raw read counts, negative control statistics, and quality control data (MAQ scores) see SF3.An annotated taxon list is available in SF4.The full sedaDNA sample/taxon two-way table (PCR replicate counts) is available in SF5.Fig. A1.1.Individual age models from (a) Chisholm Lake (this study), (b) Lost Lake (Tinner et al., 2006), (c) Birch Lake (Bigelow, 1997) and (d) Jan Lake (Carlson and Finney, 2004).Tie points are shown by dotted lines.A5.Topoclimates.Chisholm Lake lies in a complex topographic setting between steep SW-and N-facing slopes (Fig. A5a).The SW slope would have continued below the current waterline when the level of Chisholm Lake was lower.where α is azimuth from N, β is slope angle of inclination, and λ is latitude of the site, the equivalent latitude related to slope and aspect is estimated.

Fig. 1 .
Fig. 1. a) Location of Chisholm Lake within Alaska (inset) and in the Tanana River Valley of interior Alaska, and locations of the settlements of Fairbanks and Delta.Archaeological sites are shown as white circles and lakes mentioned in text as black circles.b) Topographic map (Big Delta quadrangle, 7.5-min series) showing Chisholm Lake (Lost Lake in Tinner et al., 2006); southern tip of Birch Lake to north; Tanana River floodplain (blue wavy lines) to west; Richardson Highway shown in red.Contour interval is 20 feet.c) Looking west from the eastern shore to the location of the intermittent outlet.d) Bathymetric map of the lake basin with 2-m contours, from Alaska Department of Fish and Game (http://www.adfg.alaska.gov).The dot marks the approximate Chis 17A-B coring site (see below).

Fig. 3 .
Fig. 3. Occurrence and abundance (as measured by number of PCR replicates out of eight) for key sedaDNA plant taxa at Chisholm Lake.Zones are determined by stratigraphically constrained cluster analysis (CONISS) based on abundance.Core breaks are shown to the far right.Zone CL1 (455-383 cm; ca.15,230-14,470 calyr BP; herb zone).

Fig. 4 .
Fig. 4. Occurrence and abundance (as measured by number of PCR replicates out of eight) for key aquatic plant sedaDNA at Chisholm Lake.Birch Lake lake-level stages are on the far right.Obligate aquatic taxa are shown in blue, facultative taxa in green.Apparently thick bars reflect two or more closely spaced samples.

Fig. 5 .
Fig. 5. Occurrence of main terrestrial structural-functional taxon groups over time at Chisholm Lake.Diamonds indicate the number of PCR replicates; bars show relative read abundance (%).Aquatic plants are not included in the calculation of relative read abundance as they comprise >90% of all reads.
C.L.Clarke et al.

Fig. A1. 2 .
Fig. A1.2.Probability density functions for the four lakes used in the integrated age-depth model.Individual distributions in colour (a,c), linked distributions in grey (b,d).

Fig. A3a .
Fig. A3a.Chisholm Lake, abundance of all trees, shrubs and graminoids shown as the proportion of PCR replicates (out of eight).

Fig. A3d .
Fig. A3d.Chisholm Lake, abundance of ferns, fern allies, bryophytes shown by the proportion of PCR replicates (out of eight).
C.L.Clarke et al.

Fig. A4a .
Fig.A4a.Total and rarefied richness for terrestrial plant taxa at Chisholm Lake.Zone boundaries defined by thin vertical lines.

Fig. A4b .
Fig. A4b.Rarefied and total richness for aquatic taxa plant at Chisholm Lake.Zone boundaries defined by thin vertical lines.

Fig. A4c .
Fig.A4c.Scores for DCA axis one (all taxa, x-axis, 52.3% of total variation) over time (y-axis); the four sedaDNA zones are defined by thin vertical lines.

Fig. A5a .
Fig. A5a.Part of Fig. 1 (main text) showing topography around Chisholm Lake.The red area denotes a steep SW-facing (warm) slope, the blue a steep N-facing (cold) slope.The lake (grey) shows contours above 8 m current depth, which is a conservative estimate of lake level during late-glacial aridity.Elevational transects measured using Google Earth Pro across the SW-facing and N-facing slopes surrounding Chisholm Lake (Fig.A5b) have a maximum slope of ca.20 • .Using the equation of modern equivalent latitude (EL) ofLee (1964): EL = sin -1 (sin β.cos α.cos λ + cos β.sin λ) The steepest measured angles were 21.8 • for the SW-facing slope (A-B, ca.220 • ) and 22.5 • for the N-facing slope (C-D, ca.355 • , represented by an absolute deviation from N of 5 • ).Site latitude is 64.3 • N.For ca. 20-degree slopes, the SW-facing slope EL is 45.5 • and the N-facing EL is 86.2 • .For less steep slopes of 10 • , the EL values are 49.0 • and 74.2 • , respectively.

Table A2
MAQ percent values for 120 samples (11 samples with zero repeats included) C.L.Clarke et al.A3.All terrestrial plant sedaDNA records for Chisholm Lake.C.L.Clarke et al.