Borealisation of Plant Communities in the Arctic Is Driven by Boreal-Tundra Species

Abstract

Following rapid climate change, tundra plant communities are experiencing extensive compositional shifts. A conservation concern is the potential encroachment of boreal species into the tundra ('borealisation'). Tundra borealisation has been sporadically reported, but not systematically quantified. Here, we synthesised data from across 32 study areas, spanning 1137 plots and 287 vascular plant species, resurveyed between 1981 and 2023. We (i) quantified tundra borealisation as the colonisation and increase in abundance of Boreal and Boreal-Tundra species, (ii) assessed biogeographical, climatic and local borealisation drivers and (iii) identified species contributing to borealisation and their associated traits. Half of the plots experienced borealisation, although borealisation rates were not different to random expectation. Borealisation was greater in Eurasia, closer to the treeline, at higher elevations, in warmer and wetter regions, where climate change was limited, and where initial boreal abundance was lower. Boreal coloniser species were generally short-statured, and more often shrubs and graminoids. Boreal species colonised around three times less frequently than Boreal-Tundra species. Hence, our findings indicate that tundra borealisation is mainly driven by the spread of already established boreal-low Arctic tundra species. These plant community composition changes could have cascading impacts on land-atmosphere interactions, trophic dynamics and Indigenous and local livelihoods.

Keywords: boreal forest; boreal‐tundra ecotone; climate change; plant borealisation; tundra; vascular plants.

FIGURE 1

The magnitude of boreal plant community colonisations (BCI) and abundance increases (BAI) varied across the tundra. (a) BCI estimated as the average of the plots within a study area that experienced colonisations of boreal species (BCI > 0), (b) BCI index of those plots within each study area, (c) BAI estimated as the average of the plots within a study area that experienced an increase in the abundance of boreal species (BAI > 0), (d) BAI index for those plots within each study area. Points in (a) and (c) are coloured according to the magnitude of increase (as BAI and BCI) as a study area average. Open circles in (b) and (d) indicate the mean value of the plot borealisation index at the study area level, which represent the same value as coloured points in (a) and (c). Study areas in (b) and (d) are arranged by longitude. Darker grey colours indicate overlap of multiple points. Note that this figure shows the magnitude for plots that experienced increases in boreal species; for an analysis that includes plots where boreal increases did not occur (BCI = 0 and BAI ≤ 0), see Figure S4.

FIGURE 2

Borealisation was associated with biogeographic, climatic and local variables across the tundra biome. Borealisation was estimated using boreal colonisations (BCI; a–g) and abundance change (BAI; h–l). As assessed using multivariate models, colonizations by boreal species were more likely to occur (i.e., ‘full‐range models’) at sites closer to the treeline (b), warmer and wetter sites (c, d) and at higher elevations (g), while greater magnitudes of boreal abundance increases (i.e., ‘positive‐only models’) occurred at sites in Eurasia (EA) versus Greenland‐Iceland (GI), while other regions overlapped (ENA = Eastern North America, WNA = Western North America; a), and at warmer sites (c) with the least amount of climate change (e, f). Multivariate models show that increases in abundance of boreal species were more likely (i.e., ‘full‐range models’) at higher elevations (k) and in plots with lower initial boreal status (l). The magnitudes of abundance increases (i.e., ‘positive‐only models’) were stronger in Eurasia (EA) than in Western North America (WNA; h), at sites closer to treeline (i), that had experienced the least increases in precipitation (j) and that had lower initial boreal abundance (l). The panel shows all variables that were significant in at least one of the two multivariate models (i.e., ‘positive‐only’ and ‘full‐range’ models). Lines and semi‐transparent ribbons represent the model estimate and 95% credible intervals, respectively, and are coloured according to model type (i.e., ‘positive‐only’ in green and ‘full‐range’ in grey). Solid lines indicate 95% credible intervals of fixed effects that did not overlap zero and dashed lines indicate 95% credible intervals that overlapped zero. Each point represents a plot, with darker colours indicating point overlap. Green points are included in both the ‘positive‐only’ dataset (i.e., positive values only, BCI: 598 plots, BAI: 488 plots) and in the ‘full‐range’ dataset (i.e., including also zeroes and negative values, BCI and BAI: 1137 plots). Grey points indicate plots only included in the ‘full‐range’ dataset (i.e., zeroes and negative values), in addition to the positive values. Asterisks indicate when two categorical variables differed significantly from each other, coloured according to model type.

FIGURE 3

Model estimates at the species level, with (a) total number of times colonising plots (model sample size = 220) and (b) mean annual abundance increases across all plots (model sample size = 129), as a function of class. Violin plots indicate the distribution of the raw values. Points indicate the mean model estimate for each class and error bars the 95% credible intervals. Sample sizes for categories in (a) are: Boreal = 9, Boreal‐Tundra = 113, Arctic = 9, Ubiquitous = 89 species. Sample sizes for categories in (b) are: Boreal = 5, Boreal‐Tundra = 77, Arctic = 7, Ubiquitous = 40 species.

FIGURE 4

Colonising boreal species were shorter and more likely to be shrubs or graminoids, though shrub species spanned the full range of height values. (a) Boreal species that were shorter colonised plots more often than taller species. Each point represents a species, coloured according to the functional group. The line and ribbon represent the model estimate and 95% credible intervals of the univariate model (to allow for illustration of all the available height values). (b) Boreal shrubs and graminoids colonised more often than forbs. Violin plots indicate the distribution of the raw values. Model outputs are represented as the mean estimate (points) and the 95% credible intervals (error bars). Sample sizes for each category in the model are: forb = 62, graminoid = 32, shrub = 28 species.