Abstract:
The terrestrial high-Arctic has, so far, escaped the worst impacts of non-native plant establishment.
However, increasing human activity and changing climate raise the risk of introductions and establishment,
respectively. The lack of biosecurity in the terrestrial Arctic is thus of concern. To facilitate
the development of biosecurity measures on the rapidly warming and highly trafficked archipelago
of Svalbard, we generated ecological niche models to map the bioclimatic niche potential of 27
non-native established or door-knocker vascular plant species across Svalbard, identify species with
a high risk of widespread occupancy, and locate hotspots of potential current and future invasions.
Under the current climate the three species with the highest threat in terms of broad potential area of
occupancy and known invasion potential were Deschampsia cespitosa, Ranunculus subborealis subsp.
villosus and Saussurea alpina. However, under future climate, most of the considered species have
potentially wide distributions across the archipelago. Remote eastern islands were a hotspot region
for broader potential establishment of non-native species under the current climate. Our results suggest
that many non-native plant species have a broader macroclimatic niche on Svalbard than they
currently occupy, and that other factors probably limit both dispersal and establishment outside their
current localised distributions. Environmental management on Svalbard has a limited window of opportunity
to act early in containing and preventing the spread of non-native plant species beyond the
few settlements where they currently exist. Moreover, preventing introductions and establishments
on the remote and rarely visited islands of Edgeøya, Barentsøya and Bjørnøya could be also a priority
action to safeguard sanctuaries of the archipelago’s natural ecosystems.
Description:
SUPPLEMENTARY MATERIAL : Explanation note: TABLE S1. References with species-specific strings and permanent links for each of the studied 27 non-native plant species’ ecological impact assessments. TABLE S2. The total area of potential occupancy (km2) of the 27 non-native species across the Svalbard archipelago under current and future (SSP 2-45 and 5-85) climate scenarios (data from Fig. 2, also see figs S8, S11 and S12 for the corresponding maps). FIGURE S1. GBIF spatial occurrence records downloaded for the plant species used in the study (GBIF.org, 2023a). FIGURE S2. The number of GBIF spatial occurrence records across the world for vascular plants (Tracheophyta) per 2.5’ grid cell, used as a bias weighting in sampling pseudo-absences (GBIF.org, 2023b). FIGURE S3. Pearson correlation matrix used to evaluate the colinearity of the 19 global bioclimate variables and select the five representative macroclimate variables used for the distribution modelling. FIGURE S4. Evaluation statistics of the models predictive performance across all species. FIGURE S5. Evaluation statistics (TSS and ROC) of the models predictive performance plotted against the number of occurrence records per species. FIGURE S6. Variable importances of the macroenvironmental predictors estimated across modelling methods and replicates for each species and bioclimatic variable. FIGURE S7. Response curves to the five macroenvironmental predictors per species and model algorithm. FIGURE S8. Thresholded binary species predictions of potential distribution under the current climate conditions. FIGURE S9. Model predictions for each species potential distribution under future climate (SSP2-45, 2061-2080). FIGURE S10. Model predictions for each species potential distribution under future climate (SSP5-85, 2061-2080). FIGURE S11. Thresholded model predictions for each species local potential distribution under future climate (SSP5-45, 2061-2080). FIGURE S12. Thresholded model predictions for each species local potential distribution under future climate (SSP5-85, 2061-2080). FIGURE S13. Species associations within the assemblage clusters under current and future scenarios (the clustering was independent within each time period so the cluster number is not comparable between scenarios).