Holocene plant diversity revealed by ancient DNA from 10 lakes in northern Fennoscandia

It is crucial to understand how climate warming and other environmental factors affect biodiversity, especially in the rapidly changing northern latitudes. We use sedimentary ancient DNA (sedaDNA) metabarcoding to estimate taxonomic richness, and local and regional species pools of terrestrial plants for 10 lakes in northern Fennoscandia over the Holocene. In total, 288 taxa were found in the 316 samples analysed, with local species pools of 89-200 and mean taxonomic richness of 21-65 per catchment. Quality control showed that sedaDNA is a reliable estimate of richness. Local and regional species pools showed a steep increase in the Early Holocene, when the highest rate of warming took place, and continued to increase through the Middle and into the Late Holocene, although temperature decreased over these periods. Only the regional species pool levels off during the last two millennia. Richness and local species pools were always higher in catchments with higher bedrock nutrient availability. We find sedaDNA to be a good proxy for diversity, opening avenues to detect patterns hereto unknown, and we provide a robust methodological approach to its application. Our findings suggest we can expect time lags and environmental factors to affect species richness also of the following global warming.

richness based on the lowest number of reads assigned to a sample within a lake, and 2 4 1 calculated its correlation with Hill N0 (Table S8). We used generalized additive models 2 4 2 (GAMs) (Wood, 2017) to evaluate temporal biodiversity changes during the Holocene 2 4 3 (Methods S9). We treated Hill N0 and N1 as the response, and median calibrated age of the 2 4 4 samples as predictor variables, and used the "poisson" family with log link. To account for 2 4 5 residual temporal autocorrelation between samples, we also included a continuous time first-2 4 6 order autoregressive process (CAR(1)) in generalized additive mixed models (GAMM; 2 4 7 (Simpson, 2018)). We found near identical results for taxonomic richness trends between Nesservatnet and Sierrvannet, GAMM(CAR(1)) provided a reasonable fit to the data, and 2 5 0 hence was included in the main results. We evaluated how local and regional species pools affected richness estimates at their 2 5 2 respective scales. In our case, the regional species pool is the total number of taxa found 2 5 3 across all samples. In addition, we also generated a regional species pool for each 500-year 2 5 4 time bin. We define a local species pool as the number of taxa recorded within a lake. First, GAM to highlight the regional trend in taxonomic richness through time. We then performed to test whether observed richness is correlated with the species pools of respective scales. To examine the relationship of climate and diversity estimates, we have used oxygen isotope Holocene to the Early Holocene using a linear mixed model with taxonomic richness as the 2 6 7 response and an interaction between δ 18 O and the Holocene period as predictor along with 2 6 8 lakes as the random variable. We used a new semi-quantitative nutrient index derived from the sum of the phosphorus, The dating of the individual DNA samples was dependent on the age-depth models for each 2 7 7 lake. Since the cores were all central or near central lake locations and the lakes were 2 7 8 medium-small with in most cases only one depositional basin, the age-depth curves were approximately linear or curvilinear with three exceptions (Fig. S1). Sandfjorddalen had a step 2 8 0 in the sedimentation rate with possible hiatus in the Early Holocene (11 000-8000 cal BP) and distinct reduction in sedimentation rate from around 4000 cal BP (Fig. S1i). Some age-depth 2 8 5 models showed minor concavities and convexities, which was common, with the concavities usable Nesservatnet record was reduced to the Late Holocene after removal of low quality 2 9 2 sedaDNA samples (see below).

9 3
Across our 10 lake sediment records, we generated 91.6 million raw sequence reads from 387 2 9 4 sediment samples and 90 control samples. We retained 316 samples after removing duplicates 2 9 5 and applying our QC thresholds (Fig. 2), with 12-55 samples retained per record (Tables S4,   2 9 6 S10; Fig. S4). Based on our measures of sedaDNA data quality, we found that the MTQ and 2 9 7 MAQ score QC thresholds removed the worst performing samples. The records with the best 2 9 8 sedaDNA quality are Gauptjern, Horntjernet, Nordvivatnet, Sandfjorddalen, and Sierravannet. Samples from the Early Holocene should be treated with caution from the Eaštorjávri South,  We retained 402 barcodes, which were collapsed to 346 taxa with between 89-200 taxa 3 0 2 recorded from each lake record (Table S10). Of these, 50% could be assigned to the species level (Tables S6, S11). As our focus was on the terrestrial plant diversity, we excluded 13   Data in black, samples that passed quality control (QC); blue, samples that failed QC; red,  Species pool and richness within the 10 catchments The local species pools increased over time for all catchments with the highest numbers glacier in its upper reaches (Fig. 3). Rich species pools were also found at Gauptjern, which is at the border between pine and birch forest, and at Nordvivatnet and Langfjordvannet, which Sierravannet, a site with birch forest, and pine and larch plantations. The two shrub-tundra sites, Eaštorjávri South and Sandfjorddalen, had smaller species pools, similar to Nesservatnet, which is surrounded by heathland/mires (93 taxa) and located on the small  There were clear differences among lakes both in the overall levels of richness and in the 3 2 3 change in richness over the period (Fig. 3). The mean taxonomic richness (Hill N0) ranged 3 2 4 from 20.6 (± 6.4) at Horntjernet to 65.5 (± 24.5) at Jøkelvatnet, whereas Hill N1 ranged from 14.9 (± 7.8) at Eaštorjávri South to 52.4 (± 20.5) at Jøkelvatnet ( of observed taxonomic richness for all the lakes except Sierravannet (Fig. 3). are indicated by dotted vertical lines. Note difference in scale on the y-axes.
We observed a significant effect of the age of samples on taxonomic richness as indicated by 3 4 1 statistically significant smooth terms in GAMM models (Table 2), except for Sierravannet, complex pattern of increase in richness (edf=5.93, Table 2). The steepest increase was seen in and Gauptjern, did richness reach plateau during the Late Holocene; for most lakes no nearly four taxa (regression slope=0.36) when 10 taxa were added in the local species pool.

8 5
Similarly, we found a strong positive correlation between mean richness per 500-year time 3 8 6 period and total taxa available in the respective period, and 86% of the variation in richness was explained by the regional species pool (R 2 adj =0.86, p<0.001, df=21; Fig. 5b) where c.
23% to 39% of the taxa from the regional species pool were represented by the mean richness. The mean regional richness increased by more than two taxa (regression slope=0.23) with the 3 9 0 addition of 10 taxa in the regional species pool.
3 9 8 Effect of nutrient/bedrock on richness 3 9 9 We observed a positive correlation between nutrient index and taxonomic richness for all df=8; Fig. 6b; Table S13). The effect of nutrient index on taxonomic richness was strongest significant cause of site-to-site variation and sub-regional richness patterns was soil nutrient 4 0 6 availability which is dependent upon the bedrock and the rate of weathering. were not included in the analysis. See Table S12 for the summary statistics. Note difference in  Table S13 for summary statistics. The ability of sedaDNA to capture plant taxonomic richness 4 1 7 The mean observed richness (Hill N0) of terrestrial plants found per sample and site (~21-66) 4 1 8 is higher than that recovered for northern boreal sites based on pollen analyses (~20 taxa,  The temporal patterns evaluated here rely on the assumption that our ability to detect plant 4 3 3 taxa in sedaDNA is not impacted by differential preservation, due to sample age for example, crucial that data quality is scrutinized and, where possible, standardized to allow for In considering the positive association between nutrient index and mean taxonomic richness bedrock weathering, and the release of P, K and Ca, which acts as a surrogate for alkalinity.

5 0
During the Early Holocene, it is likely that nutrient release started immediately after good subset of the regional species pool), it is reasonable to consider nutrient index as an important driver for species pool development and hence regional richness. Indeed, it is the  The highest rate of increase in richness, and local and regional species pools, is observed in Nordvivatnet: 12 700, Sandfjorddalen: 12 500 cal BP; Fig. S1e,g,j), and they, as well as Especially during the rapid warming at 11 700-10 000 cal BP, we find a high increase in Our richness patterns show a continued strong increase after around 11 000 cal. BP, when the to swamping by trees than pollen analyses and therefore better reflect habitat complexity  Middle Holocene dispersal lags The moderate increase in local and regional species pools during the Middle Holocene (8300 - Holocene Thermal Maximum) then slight cooling during this period. This is in accordance show that two of four sites along a spruce-pine-birch tundra transect show stable levels of Sweden and Germany, whereas three sites in Central Sweden level off during this period pool for the Middle Holocene. As the climate was stable during this period, we infer the 5 2 5 increase to mostly be due to dispersal lags and/or establishment lags.

2 6
Late Holocene richness nears a plateau 5 2 7 The regional species pool clearly levelled off during the past few millennia suggesting that a  northern Fennoscandia is likely due to the near-saturation of the regional species pool and the 5 3 9 overall low impact of human land use within the catchments.

4 0
In contrast to the regional scale, our data suggest that the local species pools and richness are zones that also may include terrestrial taxa. Thus, a continued increase in richness and local 5 5 7 species pool may be a result of habitat diversification. By using standardized field and lab methods, age-depth models, and rigorously synchronized Holocene. Both the QC and statistical testing reveals that the resulting plant diversity data is 5 6 3 not biased by sample age or sequencing depth. The taxonomic precision and known source estimates of taxon richness, its spatial variation, and temporal patterns. The data reveal a steep that time and abundant vacant niches. However, richness, local and the regional species pool 5 6 8 continued to increase although at a slower rate throughout most of the period, suggesting that dispersal lags and habitat diversification had a major impact on diversity also through the 5 7 0 Middle and Late Holocene. This interpretation is strengthened by the strong correlation we observed between richness and the regional species pool. In addition, we found that local richness. Individual differences were observed among our sites, but our novel combined and standardised sedaDNA analyses of 10 sites provides a superior representation of the overall 5 7 5 2 5 regional patterns in plant taxonomic richness over the Holocene. Based on these patterns from 5 7 6 the past, we may expect time lags in species response to ongoing climate change. approach", which is financed by Research Council of Norway grant number 250963/F20.

8 9
Author Contributions  Figure S14. Regional accumulated taxonomic richness during the Holocene excluding two 9 2 9 sites.  Tables   9  3  1   Table S1. Geographic and site metadata for the ten lakes.  Table S2. Composite core construction and Bayesian age-depth modelling.  Table S3. Sample metadata, including depths, LOI values, dates, and modelled ages.  Table S4. Full sample metadata including QC and bioinformatic sequence processing.  Table S5. Primer tag to sample lookup, library preparation, and accession data.  Table S6. List of all identified barcodes, including those blacklisted, and their taxonomic 9 3 7 assignments and functional groups.   Table S8. Correlations between observed and rarefied taxonomic richness for each lake.  Table S9. Summary of generalized additive models (GAMs).  Table S10. Summary of all data used or generated in this study.