Elevated graminoid cover co-occurs with Ascomycota-dominated soils in Longyearbyen, Svalbard
Lena Bakker, Annina Maier, Moritz Mainka, Jana Ruethers, Aline Frossard, Jamila Gisler, Elias Meier, Dario Barillà, Simone Fior, Kristine Bakke Westergaard, Jake Alexander, Sebastian Doetterl, Cara Magnabosco
TL;DR
Arctic vegetation changes, especially graminoid dominance, are linked to shifts in soil fungi, which may affect carbon cycling in a warming climate.
Contribution
The study reveals that graminoid-rich vegetation is associated with Ascomycota-dominated fungal communities and altered soil processes in the Arctic.
Findings
Graminoid-rich hotspots correlate with elevated soil fertility and high CO2 fluxes.
Fungal communities shift from heterogeneous to Ascomycota-dominated with graminoid vegetation expansion.
Soil fungi show greater sensitivity to environmental changes compared to prokaryotes and plants.
Abstract
Arctic warming has coincided with dramatic changes in plant cover, but the impact that aboveground biomass shifts have on soil microbial communities and processes remains poorly understood. To address this, we investigated spatial patterns of soil microbes in relation to vegetation changes using a space-for-time approach in the high Arctic region of Longyearbyen, Svalbard. We collected and characterized 31 topsoil samples from three sites that differed in nutrient input, CO2 flux, soil chemistry, and plant cover. Pronounced vegetation differences were observed at fine spatial scales, including a highly localized graminoid-dominated hotspot within areas of mixed plant communities. This graminoid-rich hotspot coincided with locally elevated soil fertility and exhibited particularly high CO2 fluxes. In areas that transitioned from dwarf shrub- to graminoid-dominated vegetation, we observed…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6| Core plots | |||
|---|---|---|---|
| Adventdalen | Bjørndalen | Old Town | |
|
|
|
| |
| CO | 11 ± 3 | 39 ± 5 | 29 ± 6 |
| CO | 16 ± 3 | 37 ± 5 | 46 ± 21 |
| Graminoid [%] | 3.50 ± 1.57 | 30.00 ± 19.75 | 85.00 ± 8.26 |
| Forbs [%] | 1.38 ± 0.68 | 57.73 ± 31.57 | 15.00 ± 6.40 |
| Bryophytes [%] | 32.50 ± 24.07 | 70.45 ± 27.06 | 53.75 ± 30.31 |
| Dwarf Shrubs [%] | 58.75 ± 15.39 | 5.45 ± 4.87 | 0.00 ± 0.00 |
| Temperature [ | 6.9 ± 0.6 | 4.0 ± 1.3 | 6.7 ± 1.1 |
| Water Content [%] | 22.9 ± 5.9 | 25.6 ± 3.3 | 43.8 ± 6.7 |
| pH | 6.0 ± 0.1 | 5.8 ± 0.03 | 6.3 ± 0.1 |
| Conductivity [mS cm-1] | 0.04 ± 0.04 | 0.06 ± 0.06 | 0.32 ± 0.09 |
| CEC [cmolc kg-1] | 6.39 ± 0.44 | 5.72 ± 1.04 | 9.07 ± 0.15 |
| TC [%] | 10.54 ± 1.02 | 8.27 ± 0.80 | 16.38 ± 0.27 |
| TN [%] | 0.87 ± 0.06 | 0.67 ± 0.07 | 1.21 ± 0.05 |
| Soil Depth | 11.18 ± 2.91 | 26.25 ± 4.42 | 19.48 ± 9.87 |
| qPCR 16 s rRNA [gene copies g-1 soil] | 5.76 × 109 ± 8.57 × 108 | 4.98 × 109 ± 1.46 × 109 | 8.07 × 109 ± 3.08 × 109 |
| qPCR ITS [gene copies g-1 soil] | 1.35 × 109 ± 4.63 × 109 | 2.43 × 109± 1.85 × 109 | 3.20 × 109 ± 1.76 × 109 |
| Gene copy ratios ITS/16S | 0.24 ± 0.07 | 0.57 ± 0.54 | 0.43 ± 0.29 |
| Microbe Shannon | 7.39 ± 0.17 | 7.61 ± 0.06 | 6.61 ± 0.24 |
| Fungi Shannon | 3.30 ± 0.52 | 4.19 ± 0.28 | 3.68 ± 0.58 |
| Plant Shannon | 1.25 ± 0.10 | 1.27 ± 0.08 | 1.01 ± 0.11 |
- —ETH Zurich10.13039/501100003006
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsPolar Research and Ecology · Biocrusts and Microbial Ecology · Climate change and permafrost
Introduction
The Arctic is warming 3–4 times faster than the rest of the world due to anthropogenic climate change and the phenomenon of Arctic Amplification (Dai et al. 2019). This rapid warming is having profound impacts on the vegetation of the Arctic and large scale changes are observable through satellite imagery collected over the past 20+ years (Berner et al. 2020, Myers-Smith et al. 2020, Phoenix and Treharne 2022). These observations reveal that longer growth periods, shifts in plant abundance, as well as the expansion of lower-latitude flora into Arctic regions are creating greener landscapes—Arctic greening; however, this greening is not occurring uniformly across the Arctic. In many areas, vegetation cover has remained stable or even declined—Arctic browning. These differences in vegetation trajectories throughout the Arctic are partially explained by warmer temperatures, water regime changes and land use changes and are also expected to have an effect on the soil ecosystem (Berner et al. 2020, Myers-Smith et al. 2020, Phoenix and Treharne 2022). Yet despite these insights, the feedback between changes in Arctic aboveground vegetation and soil ecosystem functions remain poorly understood (Doetterl et al. 2022).
Arctic tundra ecosystems are shaped by freezing, drought, salinity, low nutrient concentrations and a wide range of pH (Malard and Pearce 2018, Robinson et al. 2022). They support above- and belowground organisms that are highly adapted to these conditions and especially vulnerable to environmental changes (Bardgett et al. 2008, Hagedorn et al. 2019, Pugnaire et al. 2019). Previous studies have observed that, in addition to shrubification (Mekonnen et al. 2021), increases in the productivity of graminoids are often linked to the greening of regions throughout the Arctic (Berner et al. 2020, Thomson et al. 2021). While only representing a small proportion of the total aboveground vegetation in high Arctic tundra, the plant functional group of graminoids tends to become dominant within moist regions with high nutrient input (Thomson et al. 2021).
At the ecosystem level, graminoid-rich areas often exhibit higher rates of gross primary productivity and ecosystem respiration, largely due to the production of abundant, high-quality litter that decomposes rapidly (for a review, see Virkkala et al. 2018). However, it remains unclear how these processes vary spatially and how soil microbial communities mediate carbon cycling in response to increasing litter inputs and ongoing environmental change, particularly in the high Arctic tundra (Carlson et al. 2017, Geml et al. 2021). With continued warming, Arctic soils are expected to remain a net source of CO_2_ to the atmosphere, primarily because of permafrost thaw (Hugelius et al. 2014). At the same time, aboveground biomass productivity and litter input are projected to increase, but whether this additional organic carbon—such as that derived from graminoids during greening—will persist in soils or be rapidly respired remains uncertain (Sistla et al. 2014). The main microbial groups driving heterotrophic respiration of soil carbon in the Arctic include bacterial taxa such as Pseudomonadota, Actinomycetota, and Acidobacteriota, along with fungal decomposers like Ascomycota and Basidiomycota (Wallenstein et al. 2007, Malard and Pearce 2018). Whether these dominant decomposers are similarly enriched in areas experiencing greening and whether this affects the carbon cycle has yet to be determined.
In addition to greening and increasing litter input, more frequent disturbances from extreme weather patterns and human activities are occurring across the Arctic (Cooper 2014, Landrum and Holland 2020, Bartlett et al. 2021, Overland 2024). These changes and disturbances may destabilize ecosystems, potentially impacting which soil microbial life strategies are most successful. Life strategies of microbial organisms can generally be categorized into r-strategists (also known as copiotrophs or opportunistic ruderals with fast growth rates, that are able to profit from large labile carbon pools and quickly re-mineralize carbon), and organisms with an opposite K-strategy (also known as oligotrophs, with slow growth rates and an efficient metabolism using smaller recalcitrant carbon pools). These life strategies are used to coarsely group functionalities and traits of micro-organisms (Fierer et al. 2007, Ho et al. 2016, Chen et al. 2022). With continued warming, greening and disturbances, an enrichment of ruderal, copiotrophic. and opportunistic micro-organisms is possible, which might also promote CO_2_ flux from the soil (Ho et al. 2016, Chen et al. 2022). Previous studies both in the low Arctic and in temperate regions have shown that fast-growing, opportunistic r-strategist micro-organisms such as Bacillota (formerly known as Firmicutes), Pseudomonadota, Actinomycetota, and Ascomycota, are enriched, and that overall alpha diversity decreases, with greater litter input and introduction of more labile carbon (Zhang et al. 2016, Yao et al. 2017, Adamczyk et al. 2020). It is not clear if this is also true in the rapidly changing and greening high Arctic tundra soils.
To address this, we evaluated whether the potential ecosystem-level feedback described above are detectable at the local scale in the high Arctic by characterizing three sites with contrasting nutrient inputs and greenness. Using a space-for-time approach around the settlement of Longyearbyen, Svalbard, we evaluate the relationship between plant cover, soil carbon and fertility, CO_2_ flux, and microbial community composition. By evaluating the associations between vegetation cover, environmental gradients, CO_2_ fluxes, and soil microbial community compositions and abundances, we aimed to better understand how graminoid expansion and nutrient input history can influence microbial spatial distribution and carbon cycling across the Longyearbyen landscape. Specifically, we ask: (1) Is greening accompanied by shifts in microbial abundance and community structure? (2) Which micro-organisms are most associated with elevated graminoid cover? (3) Which micro-organisms—bacteria or fungi—are more sensitive to environmental changes associated with greening, and how can the responsiveness of these organisms influence carbon cycling and overall ecosystem functioning in the high Arctic tundra?
Methods
Locations of study
Three sites (Adventdalen, Bjørndalen, and Old Town of Longyearbyen) within the high Arctic tundra of Svalbard’s Longyearbyen region were selected based on differences in vegetation cover, and serve as a space-for-time substitution for studying the process of greening (find plant species cover in Supplementary Table B.1). These sites are separated by less than 15 km and all overlay the ca. 100–120 Myr old sandstone, siltstone and shale bedrock of the Carolinefjellet Formation (Mikhailova et al. 2021, Fig. 1). The Adventdalen site is located within the Bolterdalen valley outside of the Longyearbyen settlement (coordinates 78°10′26″N and 15°53′11″E) and does not receive additional nutrient input from human or sea bird activity. It is a site with high Arctic prostrate dwarf-shrub and herb tundra vegetation, considered to be the least green site based on vegetation cover (total plant cover of 133.67 ± 28.17%), with Dryas octopetala as the most dominant species is in the core plots. The Bjørndalen site (coordinates 78°14′07″N and 15°19′57″E) receives sea bird input (droppings) from Isfjorden as it is situated on the outwash of a little auk (Alle alle) colony in the surrounding cliffs. It is notably greener than Adventdalen (total plant cover 168.67 ± 37.67%), representing a seabird-influenced, grass-dominated high Arctic vegetation; with Equisetum arvense ssp. alpestre being the dominant species in the core plots. The Old Town site (coordinates 78°13′01″N and 15°37′00″E) is located within the Old Town of Longyearbyen and was previously used as a stable and pasture for pigs and cows before it was abandoned in 1979 (Holm 1999). Although no longer used for animal husbandry, it is an anthropogenically disturbed grass-dominated vegetation and exhibits the most extreme greenness overall (total plant cover 185.00 ± 11.46%), the most dominant species is Alopecurus ovatus in the core plots.
(a) Map locating the study area of Svalbard and (b) geological map representing the three sites discussed in this study. The native tundra in Adventdalen is shown as an orange square on the right of the geological map (the green areas are the Carolinefjellet formation and the yellow areas from light to darker are the Grumantbyen and Hollendardalen, the Basilika, and the Firkanten Formation), the site influenced by outwash of the bird cliff at Bjørndalen is shown in blue on the left and the site of anthropogenic disturbance at the old stable of Longyearbyen in the Old Town neighborhood is shown in green in the middle, overview pictures of the sites and close ups of typical vegetation are shown below the geological map. Map data Norwegian Polar Institute and OpenStreetMap contributors. (c) A stylistic representation of a site is shown in the bottom left. The core plots of the site situated within the large replicated 10 m2 plots are circled, while all plots outside of the circle are gradient plots. The sampling plots (0.25 m2) are shown as filled squares and the CO\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \end{document} flux measurements situated around the core plots of the site are shown as white dots.
Soil sampling and CO2 flux measurements
Sampling occurred during the peak growing season in the summer of 2022 between July 8^th^ and July 12^th^. Throughout the sampling period, the average air temperature was 11.9±2.5°C with minimal rainfall (<0.1 mm) and overcast cloud conditions (Supplementary Table B.2). At each site, three replicate plots of 10 m^2^ were established on relatively flat plateaus for the measurements of soil CO_2_ fluxes and physico-chemical properties. These plots were established on the areas of the nutrient amended sites Bjørndalen and Old Town with the highest plant cover and on the most representative areas for the vegetation of the Adventdalen site. Within each of these large plots, a small 0.25 m^2^ “core plot” was established. An additional set (n = 6–7) of 0.25 m^2^ “gradient plots” were established at each site to capture the environmental variability within the landscape. These plots were established within an approximately 150 × 150 m area around the core plots and were designed to capture differences in the plant cover (such as differences in plant functional group cover and dominant species) and topographical gradients (such as slopes and ditches) of the site. All plots were analyzed using the same methods with the exception of CO_2_ flux measurements only being performed on the core plots, not on the gradient plots.
CO_2_ flux measurements were collected in situ via a microportable greenhouse gas analyzer (ABB Los Gatos Research microportable gas analyzer GLA131-GGA; Quebec, Canada). Three PVC base collars of 10 cm diameter were installed, with 9 cm hammered into the soil and 1 cm of the collar protruding from the soil surface. An air-tight, transparent acrylic glass respiration chamber was placed on each PVC collar for flux measurements (Han et al. 2024). In brief, continuous CO_2_ flux measurements were taken under light and dark conditions. Ambient light flux measurements were performed for 5 min. By contrast, dark CO_2_ flux measurements were performed by covering the respiration chamber with a blacking-out fabric for 10 min prior to 5 min of continuous measurements. Gas fluxes were calculated using the ideal gas law and regression.
In each core and gradient plot, soil temperature, conductivity, water content, substrate depth, and plant functional covers of graminoids, forbs, dwarf shrubs, and bryophytes were measured (for details see Mainka et al. 2025). After the physical-, chemical, and vegetational characterization, the four corners of each plot were cored with an ethanol-sterilized, open-sided, 3 cm diameter soil corer. For each core, the uppermost 3 cm of soil was collected, quartered longitudinally, subsampled, and homogenized to obtain a representative sample for DNA analysis and spectroscopical soil assessment. The homogenized soil was subsampled into three cryovials, shock frozen in liquid nitrogen and stored at -80°C until further analysis at ETH Zurich. A parallel study combining soil wet chemistry measurements and spectral analysis provides total carbon (TC), total nitrogen (TN), effective cation exchange capacity (CEC) and pH measurements reported in this study (for details see Mainka et al. 2025).
DNA extraction, amplification, and sequencing
DNA was extracted from 0.25 g of homogenized sample material using the commercial soil DNA extraction kit DNeasy PowerSoil Pro Kit (Qiagen, Germany) with a bead beater FastPrep24^TM^ (MP Biomedicals, Germany) for mechanical lysis, otherwise following the manufacturers protocol. DNA concentrations were determined using Qubit fluorometric quantification with the high sensitivity kit (Invitrogen^TM^, USA). All DNA extracts were normalized to 2 ng/µl for further processing.
The amplification and sequencing of the ITS2 region was performed by Microsynth (Microsynth, Switzerland) using the ITS3 (5′-GCATCGATGAAGAACGCAGC-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) primers with an Illumina NovaSeq SP cartridge (Illumina, USA). 16S rRNA gene sequencing libraries were prepared at ETH Zurich following the Earth Microbiome Project protocol (Parada et al. 2016) and as adapted in (Rodriguez et al. 2025). Briefly, the extracted and normalized DNA was amplified by using a one-step PCR barcoding of the V4 and V5 region with the primers 515F (5′-GTGYCAGCMGCCGCGGTAA-3′) (Caporaso et al. 2012) and 926R (5′-CCGYCAATTYMTTTRAGTTT-3′) (Quince et al. 2011). Prior to sequencing, the amplified product was cleaned with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} 0.8\ \times \end{document} AMPure XP magnetic beads (Beckman Coulter, USA) and twice washed with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} 80{\mathrm{% }}\end{document} EtOH. Then the products were normalized to a concentration of 2.1 ng/µl and pooled for a final library concentration of 3.35 nM. The library was sequenced on an Illumina NextSeq2000 (Illumina, USA), on a P1 low output cell, set on PE 300 bp with a spike of 20% PhiX at the Functional Genomics Center (Zurich, Switzerland).
Prokaryotes (16S rRNA gene hypervariable region V4 and V5) and fungi (ITS gene region 2) were additionally quantified on a LightCycler 480 II (Roche, Switzerland) using the same 515F-926R and ITS3-ITS4 primers used for sequencing, respectively (for protocol, see Han et al. 2023). The copy numbers of 16S rRNA genes or ITS genes of the normalized samples were derived from the crossing point values (Cp values) and the transformation factor from standard curves obtained from designed gene constructs of known copy numbers (see standard curves and standard curve quality information in Supplementary Figure A.1, the full gene constructs in Supplementary Table B.3); for the detailed protocol, see Mainka et al. 2025.
Bioinformatic processing
DNA sequences were quality checked (average quality >38) with FastQC (version 0.12.1,S. Andrews 2010) and trimmed to a length of 230 bases with trimmomatic version 0.39, Bolger et al. 2014). Forward and reverse reads were joined using the fastx toolkit (minimum 20 base overlap) (version 0.0.14, Aronesty 2013). Reads were assigned to amplicon sequencing variants (ASVs) using DADA2 (single-end mode) (version 1.16.0, Callahan et al. 2016) and further processed to remove chimeras. The taxonomy of the 16S rRNA gene and ITS ASVs were assigned using the DADA2 pipeline and the Silva nr.99 v138.2 training set (Quast et al. 2013) and UNITE ITS database (Abarenkov et al. 2024), respectively.
Statistical analyses
Statistical analyses were performed using R (version 4.4.2). Sequencing data was rarefied to the shallowest read depth using phyloseq (version 1.42.0, McMurdie and Holmes 2013). Differences among sites in graminoid cover, CO₂ fluxes (dark and light), qPCR ITS gene copy numbers, and relative abundances of Ascomycota and Basidiomycota were assessed using non-parametric Kruskal–Wallis rank-sum tests. When significant effects were detected, post-hoc pairwise comparisons were conducted using Wilcoxon rank-sum tests with Benjamini–Hochberg correction to control for multiple testing. Vegan R package (version 2.6–4, Oksanen et al. 2012), was used to calculate distance matrices (Euclidian for environmental factors and Bray–Curtis dissimilarity for communities) and diversity metrics. Pearson correlation tests were performed using the cor.test function of the stats package (version 3.6.2). An enrichment analysis with multivariable associations using linear models of prokaryotic taxa on the class level with graminoids, TC, relative abundance of Ascomycota and Basidiomycota as predictor variables was performed with MaAsLin2 (version 1.12.0, Mallick et al. 2021) using a centered log ratio normalization for all predictor variables and a linear model was used. Constrained analysis of principal coordinates (CAP), was performed using microbial community Bray–Curtis dissimilarity matrices and scaled plant functional group or physico-chemical data using the capscale function (based on Legendre and Anderson 1999) from the vegan package. Within site variation for soil microbial community composition was tested using a NMDS (non-metric multidimensional scaling) ordination with the ordinate function of the phyloseq package with Bray–Curtis dissimilarity as a distance measure. Dispersion was analyzed with multivariate homogeneity of group dispersions with the betadisper function (Anderson 2006) of the vegan package and tested with Tukey honest significant differences with the TukeyHSD function of the stats package from base R. We assessed the predictive power of plant functional group cover, microbial community composition, and physico-chemical factors using multiple regression on distance matrices (MRM; using Bray–Curtis dissimilarity matrices for community composition and Euclidean distance matrices for physico-chemical factors) implemented in ecodist (version 2.1.3, Lichstein 2007). To assess the variable importance for the multiple regressions on distance matrices, we computed the full model R², removed the predictor variable in question, refit the model and calculated the difference in R^2^ to get ΔR^2^. To assess whether selection pressures or evolutionary processes are more associated with divergence in prokaryotic and fungal community composition, we performed per-site distance–decay analyses using linear regression of Bray–Curtis dissimilarities against geographic distance. Plots were generated with ggplot2 (version 3.5.1), except for upset plots, which were produced with UpSetR (version 1.4.0). All code is available on Github (https://github.com/ETH-Arctic-Greening/Microbiolgy_field_LYB).
Results
Plant cover, CO2 flux, and fungal phyla
We observed differences in vegetation cover and soil properties across the core plots of our three study sites. The Adventdalen core plots were dominated by dwarf shrubs (35%–75%), while the Bjørndalen and Old Town core plots were visibly greener—containing higher proportions of forbs (25%–90%) and graminoids (75%–90%), respectively (Table 1, Figs 1 and 2). The elevated graminoid and forb covers of Old Town and Bjørndalen were mostly confined to a small area (50 × 50 m^2^) of the greater landscape (approximately 150 × 150 m^2^; Supplementary Table B.4). This plot selection resulted in more biological variability and variability of environmental factors such as pH, water content, soil chemistry, and soil depth in the gradient plots than within the core plots (Fig. 4 and Supplementary Figure A.2).
Barplots showing an overview of site differences as represented in the core plot sample means. The standard deviations are shown as black bars and Adventdalen is shown in orange on the left, Bjørndalen is shown in blue in the middle and Old Town is shown in green on the right. (a) Shows the graminoid plant functional group cover in % and (b) shows the CO\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \end{document} flux (the CO\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \end{document} flux in the dark is shown as empty bars and the CO\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \end{document} flux in the light is shown as dotted bars). (c) represents the absolute abundance of ITS copies and the relative abundance of the most abundant fungal phyla is shown in (d), the relative abundance Ascomycota is shown as empty bars and the relative abundance of Basidiomycota is shown with striped bars.
All core plot topsoils (0–3 cm below vegetational layer, comprising of the O and rarely the A layer) were slightly acidic (pH 4.7–6.7), cool (2.8°C–8.6°C) and exhibited low electrical conductivity (undetectable to 0.42 mS cm^-1^; Table 1). On the day of sampling, Bjørndalen core plot topsoils were slightly cooler than Adventdalen and Old Town core plot topsoils. Old Town core plot topsoils exhibited significantly higher water content, CEC, TC and TN than Adventdalen and Bjørndalen topsoils (Table 1). Across all sites, flux measurements indicated that soils acted as net CO_2_ sources under both light and dark conditions (Fig. 2). Soil respiration rates estimated by CO_2_ fluxes under dark conditions mirrored patterns of graminoid cover, the relative abundance of Ascomycota, and the abundance of soil fungi as estimated by internal transcribed spacer (ITS) copy numbers across sites (Fig. 2 a–d).
Graminoid cover differed significantly between sites (Kruskal–Wallis χ² = 30.7, P < 0.001), with median cover lowest in Adventdalen, intermediate in Bjørndalen, and highest at Old Town, and all pairwise comparisons were significant (Table 1). Dark and light CO_2_ fluxes also varied significantly between sites (dark: χ^2^ = 17.6, P < 0.001; light: χ² = 21.1, P < 0.001), with Adventdalen exhibiting consistently lower fluxes than Bjørndalen and Old Town (Table 1). ITS gene copy numbers differed between sites (χ^2^ = 11.9, P = 0.003), with significantly higher abundance at Old Town compared to Adventdalen and Bjørndalen. Relative abundances of both Ascomycota and Basidiomycota varied significantly between sites (Ascomycota: χ^2^ = 19.0, P < 0.001; Basidiomycota: χ^2^ = 16.5, P < 0.001), with Old Town characterized by higher Ascomycota and lower Basidiomycota relative abundance compared to the other sites (Table 1). This tendency for greater abundance and CO_2_ fluxes in Old Town relative to Bjørndalen and Adventdalen was not supported by significant associations. However, a significant correlation was observed between core plot TC and qPCR-estimated prokaryotic abundances (R^2^ = 0.55 and P-value = 0.02; Fig. 3 a). This correlation is maintained when TC and prokaryotic abundances from all (core and gradient) plots are considered (R^2^ = 0.46 and P-value << 0.01; Fig. 3 b). Contrary to the trends of TC and qPCR estimated prokaryotic abundance (highest in Old Town), the Shannon diversity of the prokaryotic community was highest in Bjørndalen and lowest in Old Town (Table 1).
Correlation plots of the relationship between prokaryotic abundance and total carbon (TC) in the soil are shown. (a) represents data from the core plots of the three sites, while (b) shows all samples both from the core plots and the gradients of the site. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \end{document} and P-values of the correlations are given in the upper part of the plots.
Topsoil microbial community composition and abundance
Overall, 19 fungal classes and 95 prokaryotic classes (the majority of all classes) were shared by all sites, showing that the soil microbial community at a coarse taxonomic level is maintained between sites (Supplementary Figures A.4 and A.5). Bjørndalen and Old Town contained the greatest number of unique fungal (n = 12) and prokaryotic (n = 15) classes, respectively. Within all plots, Ascomycota and Basidiomycota alternate as the dominant fungal group, with the greatest relative abundance of Ascomycota observed within the Old Town core plots and greatest relative abundance of Basidiomycota observed in the Adventdalen core plots (Fig. 5). The major classes within Ascomycota are the Sordariomycetes and Leotiomycetes, while Basidiomycota are dominated by Agaricomycetes (see class level bar plot in Supplementary Figure A.5). In contrast, inter-plot prokaryotic community compositions are less variable and are dominated by Alphaproteobacteria, Thermoleophilia (phylum Actinomycetota), and Actinomycetota in general (Fig. 5). An exception to this trend is an increased abundance of Clostridia (phylum Bacillota) within the Old Town core plots relative to the gradient plots. Across the full dataset, the relative abundance of Clostridia is negatively correlated with the relative abundance of Thermoleophilia (P-value \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} < <\end{document} 0.01, negative logarithmic association; see Supplementary Figure A.6).
Soil microbial community change, graminoid abundance, total carbon content, landscape history, and distance
Principal component analysis of plant cover and soil parameters showed that core plots from each site were distinct, whereas gradient plots from Adventdalen, Bjørndalen, and Old Town converged, exhibiting similar plant cover and soil characteristics (Supplementary Figure A.2). A constrained analysis of principal coordinates (CAP) was performed on prokaryotic and fungal ASV-level Bray–Curtis dissimilarities with vegetation or soil factor predictors (Fig. 4). For prokaryotes, the first two CAP axes explain 87% of the constrained variation associated with plant functional groups, while soil variables explain 72%. For fungi, the first two CAP axes explain 67% of the constrained variation attributed to plant functional groups and 41% to soil parameters. These analyses indicate that vegetation cover patterns are better at explaining the microbial community patterns than soil parameters. More specifically the arrow representing the correlation of graminoid cover with microbial community composition shows a clear association with the fungal and prokaryotic communities of Old Town core plots, while the occurrence of dwarf shrubs and bryophytes are correlated with Adventdalen and Bjørndalen microbial communities, respectively.
Beta diversity comparisons of prokaryotic (a and b) and fungal (c and d) communities in relation to ecological factors. The left panels (a and c) display constrained analysis of principal coordinates (CAP) using Bray–Curtis dissimilarity and plant functional cover, while the right panels (b and d) show Bray–Curtis dissimilarity in relation to soil parameters. Solid dots represent samples from core plots and crosses represent the gradient plots. Orange is used for Adventdalen, blue for Bjørndalen, and green for Old Town. Microbial and fungal community composition is shown as triplicates per plot. The abbreviations used are CEC for cation exchange capacity, TN for total nitrogen, and TC for total carbon. The amount of variance explained by each axis is indicated in brackets on the respective axes.
To further evaluate these associations, a series of enrichment tests were performed at the class level (see Supplementary Figure A.7). Across all plots, graminoid cover was associated with an enrichment of the classes Desulfobulbia, Desulfitobacteriia, multiple classes within the phylum Bacillota, and the class Coriobacteriia and a depletion of classes within the phyla Chloroflexota and Acidobacteriota, and the class Methylomirabilia. Soils with elevated TC are similarly enriched in members of the phylum Bacillota and the classes Desulfobulbia and Halanaerobiia. Ascomycota-dominated soils are also enriched in the classes Coriobacteriia and Desulfobulbia, members of the phylum Bacillota, and the class Halanaerobiia as well as the methanogenic Methanobacteria class and depleted in the class Methylomirabilia, classes within the phylum Nitrososphaeria, classes of Acidobacteriota, Entotheonellia, and the class P2.11E of the phylum Chloroflexota. Basidiomycota-dominated soils show an opposite trend; depleted in the class Coriobacteriia, members of the phylum Bacillota, the classes Methanobacteria, Halanaerobiia, and Rhodothermia and enriched in the class Methylomirabilia, the phylum Nitrososphaeria, and members of the phyla Acidobacteriota, Chloroflexota, and Actinomycetota.
Distance decay analyses reveal a correlation between fungal and prokaryotic beta-diversity and geographical distance (Supplementary Figures A.8 and A.9). Generally, the correlations between distance and 16S rRNA gene community similarity were stronger than for fungal community similarity. Exceptions to this trend are due to relatively low fungal community similarities in Old Town and relatively high and uniform prokaryotic community similarities in Bjørndalen (see also variation testing in Supplementary Figure A.10). Bjørndalen and Old Town contained the greatest number of unique fungal (n = 12) and prokaryotic (n = 15) classes, respectively (see Supplementary Figures A.3 and A.4).
Predictability of plant, fungal, and prokaryotic beta-diversity by ecological variables
To better understand the predictability of the associations described in the previous section, we performed multiple regression analyses of the per-plot dissimilarity matrices of prokaryotic, fungal and plant community composition (beta-diversity) and distance matrices of soil parameters (TC, pH, conductivity, bulk soil, and depth) and geographical distance. These analyses show that plant functional group beta-diversity is not significantly predicted by any change in prokaryotic or fungal beta diversity or environmental dissimilarity. On the other hand, fungal beta-diversity is predicted by plant functional group beta-diversity (coefficient of 0.024, P-value of 0.033 and variable importance of 0.024), prokaryotic beta-diversity (coefficient of 0.288, P-value of 0.001 and variable importance of 0.228), differences in soil TC (coefficient of −0.008, P-value of 0.005 and a variable importance of 0.005), differences in pH (coefficient of 0.006, P-value of 0.049 and variable importance of 0.009), and geographical distance (coefficient of 0.000001, P-value of 0.002, and variable importance of 0.019). Prokaryotic beta-diversity is predicted by fungal beta-diversity (coefficient of 1.92, P-value of 0.001 and a variable importance of 0.193) and differences in soil TC (coefficient of 0.003, P-value of 0.024, and variable importance of 0.117) but not by the plant functional groups (coefficient of −0.005, P-value of 0.888, and variable importance of 0.009). These relationships are illustrated in Fig. 6. All coefficients, P-values and variable importances can be seen in Supplementary Table B.5.
Discussion
In concordance with previous research on Arctic greening (e.g. Berner et al. 2020, Thomson et al. 2021, Frost et al. 2025), we found that the greenest areas within our sites were dominated by local patches of graminoids within a wider dwarf shrub-dominated tundra. This pattern was most apparent in Old Town—a site where manure from pigs and cows was deposited until 1979 (Holm 1999)—and Bjørndalen—a bird cliff vegetation site with outwash from an Alle alle (little auk) colony. On the other hand, the tundra site within our study with the least amount of external nutrient inputs (Adventdalen) exhibited a lower abundance of graminoids and greater abundance of dwarf shrubs (Table 1). Previous studies have shown that complex fertilization (e.g. with manure) will increase plant biomass and the abundance of competitive, fast growing, copiotrophic, or r-strategist taxa such as the graminoids, Ascomycota, and Clostridia we have observed by eliminating naturally-occurring nutrient limitation (Oksanen and Ranta 1992, Craine and Jackson 2010, Shen et al. 2010, Štýbnarová et al. 2014, Frouz et al. 2016, Ho et al. 2017). Additional ecological factors, such as human activity in mining towns and intense herbivory by animals have also been linked to transitions from dwarf shrub- to graminoid-dominated high Arctic tundra (Kruse et al. 2021, Huusko et al. 2024), highlighting the ability for animals (including humans) to serve as ecosystems engineers throughout the Arctic.
Although limited to a small region of the high Arctic, our study underscores a variety of complex interactions between plants, animals, micro-organisms, and soils in Longyearbyen. In Old Town, the historical fertilization and disturbance effects have likely contributed to the extreme dominance of r-strategists within its core plots and the dampening of this group within the gradient plots in the surrounding landscape (Figs 2 and 5; Supplementary Figure A.7). It is important to note that the assignments of life strategies are taxonomy based and only putative. Today, large populations of barnacle geese (Branta leucopsis) graze on the graminoid-dominated soils of Old Town and may help perpetuate this fertilization effect by increasing the topsoil’s water and nutrient content (Beaulieu et al. 1996, Sjögersten et al. 2010, Foley et al. 2022, Deschamps et al. 2023). These geese may additionally influence the belowground community, as a previous study observed greater microbial diversity in ungrazed versus grazed Arctic soils (Foley et al. 2022). While we were unable to quantify grazing intensity or bird fecal deposition in this study, we similarly observed lower microbial richness in the soil plots of Old Town relative to the other sites of our study (see Table 1). Sea birds, like the little auks found in Bjørndalen, are also known to fertilize the Bjørndalen’s topsoils (Szymański et al. 2024) but a fertilization-driven skew toward microbial dominance in graminoid and forb rich patches was not observed at this site (Figs 2 and 5). Consistent with the observational lack of intense grazing around its plots, Bjørndalen harbored the largest set of site-specific fungal species (Supplementary Figure A.3), and the highest Shannon diversity for plant, prokaryote and fungal communities (see Table 1). Similar observations of elevated plant and microbial richness in soils surrounding little auk colonies in Svalbard have also been reported previously (Zwolicki et al. 2023). They have been attributed to more stable soil water regimes (Van Der Knaap 1988) and the generation of soil particles with elevated SOC (Jílková et al. 2021) that form microniches capable of supporting more diverse microbial communities than non-bird-cliff soils (Ling et al. 2014, Dubey et al. 2021). Favorable hydrological conditions may also contribute to Bjørndalen’s elevated soil microbial diversity as the area is known to contain soils with a high water holding capacity (Zmudczyńska-Skarbek et al. 2017) and receives input from a stream that crosses the bird colonies before reaching our soils of study. Together, these results provide new examples of how hydrological anthropogenic, and animal influences may affect belowground microbiological communities in the high Arctic.
Relative abundance of fungal phyla (a) and prokaryotic classes (b) per plot. Core plots are shown with a black horizontal line below the x-axis, the different sites are shown with an orange line for Adventdalen, a blue line for Bjørndalen and a green line for Old Town. The colors of the bars represent the assigned phyla and classes, and ASV sequences without any assignment to a phylum or class are transparent, explaining why most bars do not reach 1.
To further investigate the interrelatedness of the ecological factors influencing our observed diversity patterns, we performed multiple regression on ecological distance matrices (MRM). This analysis showed that the beta-diversity of other taxonomic groups was a better predictor of a distinct taxonomic group’s beta-diversity than the differences in environmental factors we observed in this study (Fig. 6, Supplementary Table B.5). Significant beta-diversity predictions contained variables with positive coefficients, but that had low variable importance such as TC, pH, and geographical distance for fungi and pH and geographical distance for prokaryotes. The lone exception to this trend is a small negative coefficient for TC in the prediction of fungal beta-diversity. As a significant negative correlation between the TC environmental distance matrix and fungal beta-diversity was not observed, this negative coefficient likely stems from TC’s co-correlation with other variables within our multiple regression analysis (Mason and Perreault 1991), see Fig. 4. By representing these beta-diversity predictions as a network of ecological variables, we found that fungi were the most central biological component of our network (Fig. 6). This centrality implies that fungi may be especially responsive to changes in the Longyearbyen soil ecosystem and, indeed, other studies have shown that fungi are sensitive to environmental changes such as grazing, soil carbon, soil nutrient levels, soil pH, and climate-related factors such as water regime, temperature, and vegetation cover in the Arctic (Zhang et al. 2016, Voříšková et al. 2019, Foley et al. 2022). Notably, one study showed a reduction of Basidiomycota in nutrient-amended Arctic wetland soils (Foley et al. 2022)—a trend that is similar to our observation of a transition from Ascomycota- to *Basidiomycota-*dominated soils along nutritional gradients, most notably in Old Town. These changes in fungal soil community composition may also directly influence the fate of high Arctic soil carbon and plant litter.
The prediction network for the biological members of the high Arctic tundra ecosystem (indicated with red boxes). Shown are significant results of multiple regressions of Bray–Curtis dissimilarity matrices of each biological group (plant functional group cover, prokaryotic community composition, or fungal community composition) and Euclidian dissimilarity matrices of soil parameters (TC, pH, conductivity, and depth) geographical distance. The size of the arrows is relative to the coefficients and the variable importance (as a ΔR2 of variable removal from the model) is given above the respective arrow.
Fungi are known to be significant contributors to the CO_2_ flux of soil ecosystems due to their activity as decomposers, especially of plant litter, but it remains unclear how Arctic greening will influence this contribution to the carbon cycle (Wallenstein et al. 2007, Christiansen et al. 2017). In our study, the lowest soil respiration rates were observed in the dwarf shrub- and Basidomycota-dominated Adventdalen, while the highest soil respiration rates were observed in the graminoid- and Ascomycota-dominated areas of Old Town (Fig. 2 and Table 1). Previous studies have additionally shown that Ascomycota-dominated communities frequently contain many cellulose and hemicellulose degrading taxa that have a tendency to exhibit higher respiration rates than the predominantly lignin-decomposing members of Basidiomycota (Riley et al. 2014, Auffret et al. 2016, Manici et al. 2024). As graminoids and dwarf shrubs will provide more hemicellulose- and lignin-rich litter, respectively (Chapin et al. 1986, Chapin III et al. 1996, Olofsson and Oksanen 2002), these observations highlight the tight connection between Longyearbyen’s above- and belowground biospheres and, potentially, future soil CO_2_ fluxes in the Arctic. Indeed, similar vegetation and soil respiration patterns have also been observed in Greenland but the contribution and composition of belowground biomass was not investigated (Bradley-Cook and Virginia 2018).
These findings suggest that with grazing and increased graminoid abundance, which might increase the delivery of more easily degradable litter to the soil, changing fungal activity could lead to more CO_2_ emissions in the future. However, further studies are needed to evaluate generalizability of our findings and the ecological mechanisms underlying the associations we observed. For soil respiration, it is known that the actively respiring community often represents only a small fraction of the full community as captured by DNA sequencing (McMahon et al. 2011, Baldrian et al. 2012, Sannino et al. 2023). Therefore, the dominant microbial groups within our DNA amplicon-based study may not be responsible for the majority of the CO_2_ emissions we observed. Additionally, we only sampled the bulk soil and not the rhizosphere, an important component of high Arctic soil ecosystems. This is an important consideration as graminoid arbuscular mycorrhizal fungal symbionts have been associated with elevated soil respiration rates (Gui et al. 2018) and, while not directly addressed by our study, could additionally contribute to the elevated soil CO_2_ fluxes we observed in graminoid-dominated plots. Previous studies have shown that the dominant graminoid of our study, Alopecurus ovatus, is known to harbor arbuscular mycorrhizal fungi (Newsham et al. 2017) and a dwarf shrub of our study, Dryas octopetala, is known to harbor ectomycorrhizal fungi (Botnen et al. 2014, 2020, Marian et al. 2022); however, within our study design, we cannot relate these symbionts to any changes in the ecosystem. We therefore encourage future studies to directly investigate the plant-fungal interactions of the rhizosphere in association with the aforementioned ecological factors we have highlighted throughout this study. To evaluate whether the trends we observed are representative of other regions in Svalbard and the rest of the high Arctic, additional studies should evaluate these ecological feedback within the context of seasonality and across a larger geographic region as geology, climate, and site histories may differ. Unfortunately, our decision to perform a space-for-time analysis within the peak growing season meant that we were unable to observe important temporal changes that can affect below ground ecosystems. For example, a multi-month soil observational study in Svalbard found that prokaryotes are more responsive to freeze–thaw cycles than to other ecological factors (Perez-Mon et al. 2020). Despite these limitations, the aforementioned biotic interactions and corresponding CO_2_ flux patterns are not limited to the Arctic and have also been reported in temperate grasslands and alpine settings (Bowman et al. 1995, Segal and Sullivan 2014, Magnani et al. 2022) and, taken together with our observations, highlight the need for further research on high Arctic soil fungal communities, especially in the context of vegetation change and potential impacts on the carbon cycle.
Conclusion
The analysis of three high Arctic tundra sites around the town of Longyearbyen, Svalbard, during the 2022 summer growing season, with differing vegetation cover and nutrient input has demonstrated that areas with greater greenness and greater nutrient availability exhibited higher rates of CO_2_ emissions and larger populations of graminoids and Ascomycota fungi. Changes in soil fungal community composition were associated with the greenness gradient of increasing graminoid cover, while prokaryotic community composition was predicted by fungal communities and total carbon but not directly by plant community composition. These findings suggest that greening is tightly associated with the bulk soil fungal community composition. Although not directly assessed in this study, we propose that beyond the well-documented effects of animals and anthropogenic activities on high Arctic vegetation diversity, they may also exert a substantial influence on belowground biodiversity. Such findings have important implications for carbon cycling in Arctic ecosystems, as shifts in microbial composition may alter decomposition rates and, ultimately, long-term carbon storage. While our results are limited to a small number of sites, they highlight the need to consider both biotic and abiotic feedback—including the role of animals—in models of Arctic ecosystem change. Future research should aim to resolve the functional consequences of these shifts and their future trajectories.
Supplementary Material
fiag019_Supplemental_File
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Abarenkov K, Zirk A, Piirmann T et al. UNITE general FASTA release for eukaryotes. 2024. 10.15156/BIO/2959334. · doi ↗
- 2Adamczyk M, Perez-Mon C, Gunz S et al. Strong shifts in microbial community structure are associated with increased litter input rather than temperature in High Arctic soils. Soil Biol Biochem. 2020;151:108054. 10.1016/j.soilbio.2020.108054. · doi ↗
- 3Anderson MJ . Distance-based tests for homogeneity of multivariate dispersions. Biometrics. 2006;62:245–53. 10.1111/j.1541-0420.2005.00440.x.16542252 · doi ↗ · pubmed ↗
- 4Andrews S . Fast QC: a quality control tool for high throughput sequence data. 2010.
- 5Aronesty E . Comparison of sequencing utility programs. Open Bioinform J. 2013;7:1–8. 10.2174/1875036201307010001. · doi ↗
- 6Auffret MD, Karhu K, Khachane A et al. The role of microbial community composition in controlling soil respiration responses to temperature. P Lo S One. 2016;11:e 0165448. 10.1371/journal.pone.0165448.27798702 PMC 5087920 · doi ↗ · pubmed ↗
- 7Baldrian P, Kolařík M, Štursová M et al. Active and total microbial communities in forest soil are largely different and highly stratified during decomposition. ISME J. 2012;6:248–58. 10.1038/ismej.2011.95.21776033 PMC 3260513 · doi ↗ · pubmed ↗
- 8Bardgett RD, Freeman C, Ostle NJ. Microbial contributions to climate change through carbon cycle feedbacks. ISME J. 2008;2:805–14. 10.1038/ismej.2008.58.18615117 · doi ↗ · pubmed ↗