Skip to main content
Dryad

Fungal trait-environment relationships in wood-inhabiting communities of boreal forest patches

Cite this dataset

Dawson, Samantha et al. (2024). Fungal trait-environment relationships in wood-inhabiting communities of boreal forest patches [Dataset]. Dryad. https://doi.org/10.5061/dryad.pg4f4qrzc

Abstract

Fungal traits can provide a mechanistic understanding of how wood-inhabiting fungi interact with their environment and how that influences community assembly in deadwood. However, fungal trait exploration is relatively new and almost no studies measure fungal traits in their environment. In this study we tested species- and trait-environment relationships in reproducing fungal communities inhabiting 571 Norway spruce (Picea abies) logs in 55 isolated forest patches (0.1-9.9 ha) of different naturalness types, located in Northern boreal Sweden. The studied patches were (1) semi-natural set-aside patches within highly managed landscapes, or (2) old-growth natural patches located in an unmanaged landscape. We tested species and trait relationships to deadwood substrate and forest patch variables. We measured mean fruit body size and density for each of the 19 species within communities. Traits assembled in relation to log decay stage and forest patch naturalness, illustrating the important role of deterministic environmental filtering in shaping reproducing wood-inhabiting fungal communities. Early decay stage communities had larger, less dense, annual fruiting bodies of half-resupinate type and were more often white-rot fungi. Species rich mid decay stage communities had mixed trait assemblages with more long lived perennial fruit bodies of intermediate size, and both brown- and white-rot fungi equally represented. Finally, late decay stage communities had smaller, denser and perennial fruit bodies, more often of the brown-rot type. The relationships between the studied traits and decay stages were similar in both set-aside and natural patches. However, set-aside semi-natural patches in highly managed landscapes more frequently supported species with smaller, perennial, and resupinate fruit bodies compared to natural patches in an unmanaged landscape.

Synthesis

We found that log decay stage was the primary driver of fungal community assembly of species and traits in isolated forest patches. Our results suggest that decay stage filters four reproduction traits (fruit body density, size, lifespan and type) and one resource-use trait (white or brown rot). Our results highlights, for the first time, that communities with diverse fungal reproductive traits are maintained foremost across all deadwood decay stages under different forest naturalness conditions.

README: Fungal trait-environment relationships in wood-inhabiting communities of boreal forest patches

https://doi.org/10.5061/dryad.pg4f4qrzc

This file contain four datasets: (1) occurrences of 19 reproducing wood fungi on 571 Norway spruce (Picea abies) logs (Occurrences2017_loglevel.csv), (2) Environmental metadata associated with the 571 logs and their location in 55 forest patches in Norrbotten County in northern Sweden in 2017 (Environmental_data_2017_loglevel.csv), (3) mean trait data for the 19 species (TraitMatrix.csv), and (4) a table for taxonomic relatedness of the 19 species (Spp_taxon_traitmatch.csv).

Description of the data and file structure

Occurrences2017_loglevel.csv

  • Site =The ID of the forest patch where the log is located in a 20 m radius circular permanently marked plot.
  • SiteLogYear = The ID of the forest patch (Site) and log (Log), as well as survey year (2017) of the recorded occurrence. The individual 571 Norway spruce logs and 55 forest patches can be linked by these site IDs (Site/Si) and log IDs (Log/L) in the different data sets.
  • 19 abbreviations of species names = Occurrences (1= presence and 0 =absence) of  the studied 19 wood fungal species in 2017. Species abbreviations and their respective former and (if applicable) currently new valid scientific names:
    • amylap = Amylocystis lapponica
    • antser = Antrodia serialis; new name Neonatrodia serialis
    • antsin = Antrodia sinuosa
    • astfer = Asterodon ferruginosus
    • fompin = Fomitopsis pinicola
    • fomros = Fomitopsis rosea; new name Rhodofomes roseus
    • glosep = Gleophyllum sepiarium
    • junlut = Junghunia luteoalba; new name Butyrea luteoalba
    • lepmol = Leptoporus mollis
    • mertax = Meruliopsis taxicola
    • phechr =  Phellinus crysolom*a; new name *Porodaedalea chrysoloma
    • phefer = Phellinus ferrugineofuscus, new name Phellinidium ferrugineofuscum
    • phevit = Phellinus viticola; new name Fuscoporia viticola
    • skeste = Skeletocutis stellae
    • stesan = Stereum sanguinolentum
    • triabi = Trichaptum abietinum
    • trifus = Trichaptum fuscoiolaceum
    • trilar = Trichaptum laricinum
    • velabi = Veluticeps abietina

Environmental_data_2017_loglevel.csv

  • Site =The ID of the forest patch where the log is located in a 20 m radius permanently marked plot.
  • SurveyYear = The year of the survey (2017).
  • SiteYear = The ID of the forest patch (Site) and survey year (Yr2017).
  • SiteLogYear = The ID of the forest patch (Si), log ID (L) and survey year (Yr2017) of recorded fungal occurrence. The individual 571 Norway spruce logs and 55 forest patches can be linked by these site IDs (Site/Si) and log IDs (Log/L) in the different data sets.
  • Log =  The ID of the log within each forest patch (site).
  • Decay_class = Log decay class categorised into early = deadwood hard and >50% bark remaining, mid = deadwood hard or starts to soften with smooth texture and <50% bark remaining, or late = deadwood soft with small crevices, pieces or wood fragments lost and the outline of the log can have started to become deformed, based on McCullough (1948).  
  • Log_vol_Con_para_m3 =  The individual log volume (m3) calculated using the Fraver et al. (2007) conic-paraboloid formula.
  • VolperHa = For individual 0.13-ha plots, the summed log volume per hectare (m3ha-1).
  • SpruceLiving = The number of living Norway spruce trees with a diameter at breast height (1.3 m) >= 25 cm, within 20 m radius circular plots (not analysed in the associated paper in Functional Ecology).
  • WKHvGL = Categorial variable if a set-aside semi-natural Woodland Key Habitat (WKH) in the managed forest landscape or natural old-growth forest patch in Granlandet Nature Reserve (GL) unmanaged forest landscape. 
  • AreaAtSurvey = The forest patch area at survey in 2017 (m2).
  • CutsW1km = Proportion clear cutting in the immediate surrounding land within 1 km of the forest patch that experienced clear cutting in the 10 years prior to the survey in 2017, calculated using desktop geospatial analyses.

TraitMatrix.csv

  • Species = Abbreviations for the 19 wood fungal scientific species names.
  • FBDryWei.mg.mg3 = Mean fruit body dry density (mg/mm3) per species measured in the field and laboratory (Dawson et al. 2019, 2020). Measurements and samples were taken from one randomly selected log and up to three random mature fruit bodies of each species growing on each log. A total of 164 fruit bodies were measured, with a mean of nine fruit bodies measured per species. Young undeveloped and overmature decaying fruit bodies were not measured. Data were collected from each species in every set-aside patch and from 11 out of the 29 natural patches. Fruit bodies were randomly selected by numbering the fruit bodies from one end of a log to the other and randomly selecting numbers and corresponding fruit bodies.
  • FBSize.mm3 = Mean fruit body size (mm3) per species measured in the field (Dawson et al. 2019, 2020). Measurements and samples were taken from one randomly selected log and up to three random mature fruit bodies of each species growing on each log. A total of 164 fruit bodies were measured, with a mean of nine fruit bodies measured per species. Young undeveloped and overmature decaying fruit bodies were not measured. Data were collected from each species in every set-aside patch and from 11 out of the 29 natural patches. Fruit bodies were randomly selected by numbering the fruit bodies from one end of a log to the other and randomly selecting numbers and corresponding fruit bodies.
  • SporeVol = Mean spore volume (µm3) were sourced from the literature (Nordén et al., 2013).
  • RedList = Red-listed status, least concern (LC) or red-listed (RL) following SLU Artdatabanken (2020).
  • Decay = The resource-use (rot type), white vs brown rot, sourced from the literature (Nordén et al., 2013).
  • FBlife = The fruit body life span, annual (a) vs perennial (p), sourced from the literature (Nordén et al., 2013).
  • FBtypen =The fruit body type, half resupinate (half-resu), pileate (p), and resupinate (res), sourced from the literature (Nordén et al., 2013).

Spp_taxon_traitmatch.csv

  • shorthand = Abbreviations for the 19 wood fungal scientific species names.
  • Kingdom = Second highest taxonomic rank recognised by the nomenclature codes (Fungi).
  • Phylum = Taxonomic rank recognised by the nomenclature codes (Basidiomycota).
  • Class = A taxonomic rank recognised by the nomenclature codes (Agaricomycetes).
  • Order = A taxonomic rank recognised by the nomenclature codes.
  • Family = A taxonomic rank recognised by the nomenclature codes.
  • Genus = A taxonomic rank recognised by the nomenclature codes.
  • Species = A taxonomic rank recognised by the nomenclature codes as the basic unit, the scientific species name.

References

SLU Artdatabanken (2020). Red-listed species in Sweden 2020. SLU, Uppsala

Dawson, S. K., Boddy, L., Halbwachs, H., Bässler, C., Andrew, C., Crowther, T. W., … & Jönsson, M. (2019). Handbook for the measurement of macrofungal functional traits; a start with basidiomycete wood fungi. Functional Ecology, 33, 1365–2435. https://doi.org/10.1111/1365-2435.13239

Dawson, S. K., Berglund, H., Ovaskainen, O., Snäll, T., Jonsson, B. G., & Jönsson, M. (2020). Convergence of fungal traits over time in natural and forestry-fragmented patches. Biological Conservation, 251, 108789. https://doi.org/10.1016/j.biocon.2020.108789

Fraver, S., Ringvall, A., & Jonsson, B. G. (2007). Refining volume estimates of down woody debris. Canadian Journal of Forest Research, 37, 627–633. https://doi.org/10.1139/X06-269

McCullough, H. A. (1948). Plant Succession on Fallen Logs in a Virgin Spruce-Fir Forest. Ecology, 29, 508–513. https://doi.org/10.2307/1932645

Nordén, J., Penttilä, R., Siitonen, J., Tomppo, E., & Ovaskainen, O. (2013). Specialist species of wood-inhabiting fungi struggle while generalists thrive in fragmented boreal forests. Journal of Ecology, 101, 701–712. https://doi.org/10.1111/1365-2745.12085

Code/Software

We analysed the data with Hierarchical Modelling of Species Communities (Ovaskainen et al., 2017), fitted with the Hmsc R package (Tikhonov et al., 2020). 

Ovaskainen, O., Tikhonov, G., Norberg, A., Blanchet, G.F., Duan, L., Dunson, D., Roslin, T. & Abrego, N. (2017). How to make more out of community data? A conceptual framework and its implementation as models and software. Ecology Letters, 20, 561–576. https://doi.org/10.1111/ele.12757

Tikhonov, G., Opedal, Ø. H., Abrego, N., Lehikoinen, A., Jonge, M. M. J., Oksanen, J., & Ovaskainen, O. (2020). Joint species distribution modelling with the r-package Hmsc. Methods in Ecology and Evolution, 11, 442–447. https://doi.org/10.1111/2041-210X.13345

Methods

Study sites

We studied deadwood fungal communities in boreal forest patches that had been isolated through either natural or anthropogenic processes in Norrbotten County, in Northern Sweden (Berglund and Jonsson, 2003; Berglund and Jonsson, 2005). Natural patches consisted of old-growth forest patches surrounded by open mire randomly selected within the Granlandet Nature Reserve (ca 27 000 ha), a landscape largely unaffected by forest management operations (Lövgren, 1986). Granlandet is primarily protected as a reference landscape for studying Norway spruce old-growth forest dynamics and fragmentation. It has no recreational access roads, but limited hunting is permitted within the reserve. The patches have no signs of forest fires or forest management activities (two cut stumps have been recorded in one of all 29 forest patch plots, likely cut by local hunters for fire wood), and patches have likely been in a natural old-growth state for many centuries with spruces attaining ages up to 300 years (Lövgren, 1986; Berglund and Jonsson, 2005). The mean deadwood volume was 30.7 ± standard deviation (SD) 18.5 m3ha-1 in natural patches. We assume that extinction and colonization have reached a dynamic equilibrium in these patches (Berglund and Jonsson, 2005). Consequently, the potential long-term impact of forest structure and patch area on species distribution and community assembly should be detectable. Anthropogenic set-aside patches consisted of randomly selected semi-natural forest stand patches located within the managed forest landscape surrounding the natural Granlandet Nature Reserve landscape. The set-aside patches were identified as having high biodiversity and/or conservation value, due to presence of old-growth forest indicator species and/or the forest structure and set-aside from management in the late 1990s (Berglund and Jonsson, 2005). These set-aside patches have been isolated from natural forests at different time points since clear cutting and intensive forest management began in the region in the 1950s and up to the 1990s (Berglund and Jonsson, 2005). The set-aside patches have to different degrees been affected by historical selective logging, but never clear cut (mean 5.6 ± SD 9.1 cut stumps recorded per 0.13 plot). The mean deadwood volume was 19.2 ± SD 17.3 m3ha-1 in set-aside patches.  All patches were dominated by Norway spruce (Picea abies), with bilberry dominating the understory layer, and with moist to mesic ground conditions on moraine soils. Surveying permit for Granlandet Nature Reserve (525-8301-17) was issued by the Swedish County Administrative Board of Norrbotten and survey permit for one set-aside patch that was given protected status (Dnr 2018:906) was issued by Jokkmokks Municipality.

Fungal survey

In the autumn of 2017 we surveyed wood-inhabiting fungal fruit bodies present on Norway spruce logs originating within circular 0.13 ha plots at the centre of 26 set-aside patches (0.08–6.7 ha) and 29 natural patches (0.17–9.9 ha) (Dawson et al., 2020a). All logs with ≥10 cm diameter and ≥1 m length originating within sample plots were surveyed for fruit bodies. We recorded all polypore species along with six corticoid species considered important indicator and decomposer species in old-growth forest communities (Asterodon ferruginosus, Cystostereum murrai, Laurilia sulcata, Phlebia centrifuga, Stereum sanguinolentum and Veluticeps abietina). Polypores and corticoids, as pileate and resupinate fruit body types, are especially suitable for studying forest naturalness responses in communities inhabiting Norway spruce (Purhonen et al., 2021) and also provide more reliable results for community composition comparison from single surveys (Abrego et al., 2016). Earlier fruit-body surveys combined with molecular inventories of fungal mycelial abundance in the deadwood show that the most abundant mycelium form fruit bodies (Kubartová et al., 2012; Ovaskainen et al., 2013). It is therefore reasonable that the fruit body size and density of the dominant fruiting species studied here, to some extent also reflect the species mycelial size and resource acquisition in the deadwood. Whilst we acknowledge that the presence of fungal fruit bodies does not give complete information about the fungal species succession, it provides meaningful information from a trait and species association perspective (Saine et al., 2020). The species nomenclature follow the Swedish taxonomic database Dyntaxa (www.artfakta.se).

Environmental variables

We used six environmental variables in our analysis: log decay stage, log volume, summed log volume per hectare, forest patch area, natural patch versus set-aside patch, and clear cutting in the immediate surrounding 1 km. For individual downed logs sampled within plots, decay stage was categorised into early (deadwood hard and >50% bark remaining), mid (deadwood hard or starts to soften with smooth texture and <50% bark remaining) or late (deadwood soft with small crevices, pieces or wood fragments lost and the outline of the log can have started to become deformed), based on McCullough (1948).  The individual log volume (m3) was calculated using the Fraver, et al. (2007) conic-paraboloid formula. For individual 0.13-ha plots, the summed log volume per hectare (m3ha-1) was calculated. Patch area (ha) and the proportion of surrounding land within 1 km of the patch that experienced clear cutting in the 10 years prior to the survey was calculated using desktop geospatial analyses (Dawson et al., 2020a). No environmental variables were strongly correlated (r ≤ 0.6).

Selected functional traits

We assessed species traits that we considered functionally important for dispersal (spore volume; µm3), resource-use (rot type), and reproduction (fruit body size, density, lifespan and type). Mean fruit body size (mm3) and dry density (mg/mm3) per species were measured in the field and laboratory. Measurements and samples were taken from one randomly selected log and up to three random mature fruit bodies of each species growing on each log. A total of 164 fruit bodies were measured, with a mean of nine fruit bodies measured per species. Young undeveloped and overmature decaying fruit bodies were not measured. Data were collected from each species in every set-aside patch and from 11 out of the 29 natural patches  (Dawson et al., 2019; Dawson et al., 2020a). Fruit bodies were randomly selected by numbering the fruit bodies from one end of a log to the other and randomly selecting numbers and corresponding fruit bodies. Mean spore volume (µm3) and categorical traits such as rot type, fruit body lifespan and type were sourced from the literature (Nordén et al., 2013). No trait or attribute were strongly correlated (r < 0.5). A conservation attribute, red-listed status of a species, was also included as it is relevant for management and conservation, and these species are often resource specialists in terms of deadwood amount and quality (SLU Artdatabanken, 2020).

Preparation of data and statistical analyses

We included species that occurred in three or more patches (28 of 45 species) for which we had complete trait information (19 of 28 species, see supporting information Table S2 for species omitted due to rarity and/or lack of trait measurements). These 19 species contained 93% of all occurrences (i.e. defined as species found fruiting on individual logs), which is above the 80% recommended for trait analyses (Pérez-Harguindeguy et al., 2013). In total, there were 745 fruit-body occurrences across 571 logs in the dataset that was analysed statistically (supporting information Table S3).

We analysed the data with Hierarchical Modelling of Species Communities (HMSC; Ovaskainen et al., 2017), which is a joint species distribution modelling (JSDM)  framework (Warton et al., 2015) that enables the integration of data on species occurrences or abundances, environmental variables, species traits and phylogenetic relationships (Ovaskainen et al., 2017). In the HMSC analyses, the  response matrix Y consisted of presence-absences of the  species observed in the  surveyed logs. We modelled Y using a probit link function and included a matrix X of environmental variables at the scale of the log and the patch (i.e. sample plot), respectively.  At the log scale, we included the log transformed log volume and the decay stage classified as early, mid, or late. At the patch scale, we included the log transformed summed deadwood volume per hectare, the log transformed area of the patch, the proportion of land immediately surrounding patches within 1 km that had been clear cut within the 10 years prior to the survey (clear cutting), and patch type classified as natural or set-aside patch. Further, a latent variable approach was used to have a patch-scale random effect to account for the grouping of logs within patches (Ovaskainen et al., 2017). All continuous explanatory environmental variables were scaled (subtracting the mean and dividing by the standard deviation) to improve model performance and facilitate interpretability of model outputs.

Similar to Dawson et al. (2020a), species traits and phylogenetic relationships were used for modelling the relationships between X and Y (Abrego, Norberg, & Ovaskainen, 2017; Ovaskainen et al., 2017). Traits used in the T matrix are described above. These included log transformed fruit-body size, fruit-body density, spore volume, decay type (brown or white rot), fruit-body lifespan (annual or perennial), fruit-body type (pileate, resupinate, or half-resupinate), and red-listed status (least concern or red-listed following SLU Artdatabanken, 2020). Following Abrego et al. (2017) and Dawson et al. (2020a), as no quantitative phylogeny existed, we used as a proxy for the phylogenetic correlation matrix C a taxonomical correlation matrix, constructed from the five levels of class, order, family, genus and species, and assuming equal branch length for each level.  

We fitted the model using the Hmsc R package (Tikhonov et al., 2020) with default prior distributions. The sampled posterior distribution consisted of 150,000 MCMC iterations, of which 50,000 iterations were discarded as burn-in. We ran two MCMC chains and thinned by 100, giving a posterior distribution sample size of 1,000 per chain. We assessed the convergence of the two MCMC chains visually.

We evaluated the explanatory powers of the occurrence probit model through species-specific Area Under the Curve (AUC) values, which were then averaged across the species to obtain the model-specific metric. To compute explanatory power, we made model predictions based on the model fitted to all data. The fitted joint species distribution model was also used to assess the scale dependency in the trait-level responses of the species to the environmental variables (Abrego, Norberg, & Ovaskainen, 2017). We utilized a variance partitioning technique to determine the proportion of variability in the occurrences of species that could be ascribed to the environmental variables at the log- and patch-scales in contrast to that explained by the random effects, following the methodology proposed by Ovaskainen et al. (2017). Next, we calculated the proportion of variation in the species' response to environmental variables that could be accounted for by the variation in their traits (Abrego, Norberg, & Ovaskainen, 2017).

With our analysis, we aimed to answer if traits, as well as species richness and occurrence probabilities of red-listed species, differed with log decay stage and between natural and set-aside patches. To answer these questions, we predicted species communities on logs within patches based on 2000 draws (1000 draws from each chain) from the joint posterior distribution of the fitted model, with additional non-focal environmental variables held constant at their observed mean values (Dawson et al., 2020a). For example, when making the hypothesis for set-aside patches, we set the log volume, summed log volume, patch area, and surrounding clear-cutting to their mean values in this subset of the data. We compared predictions for species richness, red-listed species occurrence probability, as well as trait values (i.e., community weighted mean values of continuous traits and occurrence probabilities of categorical traits). We considered an observed difference to have moderate or strong statistical support if the parameter of interest was greater in one case (i.e. log decay stage or naturalness patch type) than in another with at least 90 or 95% posterior probability, respectively.

Study sites

We studied deadwood fungal communities in boreal forest patches that had been isolated through either natural or anthropogenic processes in Norrbotten County, in Northern Sweden (Berglund and Jonsson, 2003; Berglund and Jonsson, 2005). Natural patches consisted of old-growth forest patches surrounded by open mire randomly selected within the Granlandet Nature Reserve (ca 27 000 ha), a landscape largely unaffected by forest management operations (Lövgren, 1986). Granlandet is primarily protected as a reference landscape for studying Norway spruce old-growth forest dynamics and fragmentation. It has no recreational access roads, but limited hunting is permitted within the reserve. The patches have no signs of forest fires or forest management activities (two cut stumps have been recorded in one of all 29 forest patch plots, likely cut by local hunters for fire wood), and patches have likely been in a natural old-growth state for many centuries with spruces attaining ages up to 300 years (Lövgren, 1986; Berglund and Jonsson, 2005). The mean deadwood volume was 30.7 ± standard deviation (SD) 18.5 m3ha-1 in natural patches. We assume that extinction and colonization have reached a dynamic equilibrium in these patches (Berglund and Jonsson, 2005). Consequently, the potential long-term impact of forest structure and patch area on species distribution and community assembly should be detectable. Anthropogenic set-aside patches consisted of randomly selected semi-natural forest stand patches located within the managed forest landscape surrounding the natural Granlandet Nature Reserve landscape. The set-aside patches were identified as having high biodiversity and/or conservation value, due to presence of old-growth forest indicator species and/or the forest structure and set-aside from management in the late 1990s (Berglund and Jonsson, 2005). These set-aside patches have been isolated from natural forests at different time points since clear cutting and intensive forest management began in the region in the 1950s and up to the 1990s (Berglund and Jonsson, 2005). The set-aside patches have to different degrees been affected by historical selective logging, but never clear cut (mean 5.6 ± SD 9.1 cut stumps recorded per 0.13 plot). The mean deadwood volume was 19.2 ± SD 17.3 m3ha-1 in set-aside patches.  All patches were dominated by Norway spruce (Picea abies), with bilberry dominating the understory layer, and with moist to mesic ground conditions on moraine soils. Surveying permit for Granlandet Nature Reserve (525-8301-17) was issued by the Swedish County Administrative Board of Norrbotten and survey permit for one set-aside patch that was given protected status (Dnr 2018:906) was issued by Jokkmokks Municipality.

Fungal survey

In the autumn of 2017 we surveyed wood-inhabiting fungal fruit bodies present on Norway spruce logs originating within circular 0.13 ha plots at the centre of 26 set-aside patches (0.08–6.7 ha) and 29 natural patches (0.17–9.9 ha) (Dawson et al., 2020a). All logs with ≥10 cm diameter and ≥1 m length originating within sample plots were surveyed for fruit bodies. We recorded all polypore species along with six corticoid species considered important indicator and decomposer species in old-growth forest communities (Asterodon ferruginosus, Cystostereum murrai, Laurilia sulcata, Phlebia centrifuga, Stereum sanguinolentum and Veluticeps abietina). Polypores and corticoids, as pileate and resupinate fruit body types, are especially suitable for studying forest naturalness responses in communities inhabiting Norway spruce (Purhonen et al., 2021) and also provide more reliable results for community composition comparison from single surveys (Abrego et al., 2016). Earlier fruit-body surveys combined with molecular inventories of fungal mycelial abundance in the deadwood show that the most abundant mycelium form fruit bodies (Kubartová et al., 2012; Ovaskainen et al., 2013). It is therefore reasonable that the fruit body size and density of the dominant fruiting species studied here, to some extent also reflect the species mycelial size and resource acquisition in the deadwood. Whilst we acknowledge that the presence of fungal fruit bodies does not give complete information about the fungal species succession, it provides meaningful information from a trait and species association perspective (Saine et al., 2020). The species nomenclature follow the Swedish taxonomic database Dyntaxa (www.artfakta.se).

Environmental variables

We used six environmental variables in our analysis: log decay stage, log volume, summed log volume per hectare, forest patch area, natural patch versus set-aside patch, and clear cutting in the immediate surrounding 1 km. For individual downed logs sampled within plots, decay stage was categorised into early (deadwood hard and >50% bark remaining), mid (deadwood hard or starts to soften with smooth texture and <50% bark remaining) or late (deadwood soft with small crevices, pieces or wood fragments lost and the outline of the log can have started to become deformed), based on McCullough (1948).  The individual log volume (m3) was calculated using the Fraver, et al. (2007) conic-paraboloid formula. For individual 0.13-ha plots, the summed log volume per hectare (m3ha-1) was calculated. Patch area (ha) and the proportion of surrounding land within 1 km of the patch that experienced clear cutting in the 10 years prior to the survey was calculated using desktop geospatial analyses (Dawson et al., 2020a). No environmental variables were strongly correlated (r ≤ 0.6).

Selected functional traits

We assessed species traits that we considered functionally important for dispersal (spore volume; µm3), resource-use (rot type), and reproduction (fruit body size, density, lifespan and type). Mean fruit body size (mm3) and dry density (mg/mm3) per species were measured in the field and laboratory. Measurements and samples were taken from one randomly selected log and up to three random mature fruit bodies of each species growing on each log. A total of 164 fruit bodies were measured, with a mean of nine fruit bodies measured per species. Young undeveloped and overmature decaying fruit bodies were not measured. Data were collected from each species in every set-aside patch and from 11 out of the 29 natural patches  (Dawson et al., 2019; Dawson et al., 2020a). Fruit bodies were randomly selected by numbering the fruit bodies from one end of a log to the other and randomly selecting numbers and corresponding fruit bodies. Mean spore volume (µm3) and categorical traits such as rot type, fruit body lifespan and type were sourced from the literature (Nordén et al., 2013). No trait or attribute were strongly correlated (r < 0.5). A conservation attribute, red-listed status of a species, was also included as it is relevant for management and conservation, and these species are often resource specialists in terms of deadwood amount and quality (SLU Artdatabanken, 2020).

Preparation of data and statistical analyses

We included species that occurred in three or more patches (28 of 45 species) for which we had complete trait information (19 of 28 species, see supporting information Table S2 for species omitted due to rarity and/or lack of trait measurements). These 19 species contained 93% of all occurrences (i.e. defined as species found fruiting on individual logs), which is above the 80% recommended for trait analyses (Pérez-Harguindeguy et al., 2013). In total, there were 745 fruit-body occurrences across 571 logs in the dataset that was analysed statistically (supporting information Table S3).

We analysed the data with Hierarchical Modelling of Species Communities (HMSC; Ovaskainen et al., 2017), which is a joint species distribution modelling (JSDM)  framework (Warton et al., 2015) that enables the integration of data on species occurrences or abundances, environmental variables, species traits and phylogenetic relationships (Ovaskainen et al., 2017). In the HMSC analyses, the  response matrix Y consisted of presence-absences of the  species observed in the  surveyed logs. We modelled Y using a probit link function and included a matrix X of environmental variables at the scale of the log and the patch (i.e. sample plot), respectively.  At the log scale, we included the log transformed log volume and the decay stage classified as early, mid, or late. At the patch scale, we included the log transformed summed deadwood volume per hectare, the log transformed area of the patch, the proportion of land immediately surrounding patches within 1 km that had been clear cut within the 10 years prior to the survey (clear cutting), and patch type classified as natural or set-aside patch. Further, a latent variable approach was used to have a patch-scale random effect to account for the grouping of logs within patches (Ovaskainen et al., 2017). All continuous explanatory environmental variables were scaled (subtracting the mean and dividing by the standard deviation) to improve model performance and facilitate interpretability of model outputs.

Similar to Dawson et al. (2020a), species traits and phylogenetic relationships were used for modelling the relationships between X and Y (Abrego, Norberg, & Ovaskainen, 2017; Ovaskainen et al., 2017). Traits used in the T matrix are described above. These included log transformed fruit-body size, fruit-body density, spore volume, decay type (brown or white rot), fruit-body lifespan (annual or perennial), fruit-body type (pileate, resupinate, or half-resupinate), and red-listed status (least concern or red-listed following SLU Artdatabanken, 2020). Following Abrego et al. (2017) and Dawson et al. (2020a), as no quantitative phylogeny existed, we used as a proxy for the phylogenetic correlation matrix C a taxonomical correlation matrix, constructed from the five levels of class, order, family, genus and species, and assuming equal branch length for each level.  

We fitted the model using the Hmsc R package (Tikhonov et al., 2020) with default prior distributions. The sampled posterior distribution consisted of 150,000 MCMC iterations, of which 50,000 iterations were discarded as burn-in. We ran two MCMC chains and thinned by 100, giving a posterior distribution sample size of 1,000 per chain. We assessed the convergence of the two MCMC chains visually.

We evaluated the explanatory powers of the occurrence probit model through species-specific Area Under the Curve (AUC) values, which were then averaged across the species to obtain the model-specific metric. To compute explanatory power, we made model predictions based on the model fitted to all data. The fitted joint species distribution model was also used to assess the scale dependency in the trait-level responses of the species to the environmental variables (Abrego, Norberg, & Ovaskainen, 2017). We utilized a variance partitioning technique to determine the proportion of variability in the occurrences of species that could be ascribed to the environmental variables at the log- and patch-scales in contrast to that explained by the random effects, following the methodology proposed by Ovaskainen et al. (2017). Next, we calculated the proportion of variation in the species' response to environmental variables that could be accounted for by the variation in their traits (Abrego, Norberg, & Ovaskainen, 2017).

With our analysis, we aimed to answer if traits, as well as species richness and occurrence probabilities of red-listed species, differed with log decay stage and between natural and set-aside patches. To answer these questions, we predicted species communities on logs within patches based on 2000 draws (1000 draws from each chain) from the joint posterior distribution of the fitted model, with additional non-focal environmental variables held constant at their observed mean values (Dawson et al., 2020a). For example, when making the hypothesis for set-aside patches, we set the log volume, summed log volume, patch area, and surrounding clear-cutting to their mean values in this subset of the data. We compared predictions for species richness, red-listed species occurrence probability, as well as trait values (i.e., community weighted mean values of continuous traits and occurrence probabilities of categorical traits). We considered an observed difference to have moderate or strong statistical support if the parameter of interest was greater in one case (i.e. log decay stage or naturalness patch type) than in another with at least 90 or 95% posterior probability, respectively.

Funding

Academy of Finland, Award: 336212

Academy of Finland, Award: 345110

Swedish Research Council for Environment Agricultural Sciences and Spatial Planning, Award: 2016-00461

European Research Council, Award: 856506, European Union’s Horizon 2020 research and innovation programme

Swedish University of Agricultural Sciences