The critical role of tree species and human disturbance in determining the macrofungal diversity in Europe

Aim: Knowledge concerning species distribution is important for biodiversity conservation and environmental management. Fungi form a large and diverse group of species and play a key role in nutrient cycling and carbon storage. However, our understanding of fungal diversity and distribution remains limited, particularly at large spatial scales. Here, we predicted the diversity and distribution of ectomycorrhizal


| INTRODUC TI ON
The measurement and understanding of species diversity and distribution forms a central topic in ecology and biogeography (Gaston, 2000). Knowledge of the distribution of species and explaining the factors that drive this spatial distribution at different spatial scales are essential for biodiversity conservation, especially in the context of global change (Bellard et al., 2012;Purvis & Hector, 2000). However, compared with animals and plants, the fungal kingdom is little studied, especially at the continental level (Willis, 2018).
Fungi are one of the most diverse groups of organisms on Earth (Blackwell, 2011;Tedersoo et al., 2014). They are essential organisms that interact with the abiotic and biotic environment, regulating nutrient and carbon cycles. Macrofungi are those fungi that produce visible fruiting bodies. They form an important part of the fungal kingdom, which is estimated to comprise 53,000-110,000 species world-wide (Mueller et al., 2007). Ecologically, macrofungi can be classified as parasites, saprotrophs or symbiotic (mycorrhizal) species. The two main functional groups of macrofungi associated with global nutrient and carbon cycles are ectomycorrhizal fungi (mutualistic fungi) and saprotrophic fungi (Dighton, 2016). Ectomycorrhizal fungi live mutualistically with tree species and provide some benefit to their host. As symbionts, ectomycorrhizal fungi are involved in various mutualistic associations, contributing to the uptake of water, nitrogen, phosphorus or other minerals from the soil to their host; hence, they underpin primary production in natural ecosystems (Dighton, 2016;Read, 1991). Saprotrophic fungi are fungi that use non-living organic matter, such as dead plants and animals; they are decomposers in the ecosystem. These saprotrophic fungi can release nutrients and energy stored in dead organisms, supporting the crucial process of nutrient cycling and partaking in the global carbon cycle (Dighton, 2016). They form the basis of soil food chains and play an indispensable role in the quest for sustainable development (Palm & Chapela, 1997). Many macrofungal species also form a food source for humans and wild animals, contributing to human welfare directly and indirectly (Boa, 2004). Despite their essential roles, the distribution pattern of macrofungi has rarely been explored in macroecology, and our knowledge of factors determining their distribution remains limited.
For ectomycorrhizal fungi, multiple local-scale studies have shown that soil nutrients and host tree species form the key factors influencing their diversity and distribution (Bahram et al., 2012;Ishida et al., 2007;Toljander et al., 2006). At the continental scale, Andrew, Halvorsen, et al. (2018) showed that climate (e.g., mean annual temperature) was strongly correlated with the ectomycorrhizal macrofungal assemblage in Europe. In contrast, using a roottip sampling method based on Sanger sequencing technology, van der Linde et al. (2018) found that host-related variables contributed more to explaining the continental-scale diversity pattern of ectomycorrhizal fungi than did climatic variables in Europe. The role of tree species and the relative importance of tree species versus climatic variables in determining the diversity and distribution of ectomycorrhizal fungi at a large spatial scale remain largely unexplored.
Regarding saprotrophic fungi, some local-scale studies showed that microclimate (temperature and humidity), resource availability (such as the amount of available dead wood) and host tree species were the most important factors determining their diversity (Bässler et al., 2010;Rayner & Boddy, 1988;Wollan et al., 2008). Some large-scale studies across Europe have shown that climate was the most crucial variable influencing their diversity (Andrew, Halvorsen, et al., 2018) and that there was no significant relationship between tree species and the diversity of saprotrophic macrofungi (Andrew, Büntgen, et al., 2019;Andrew, Halvorsen, et al., 2018). However, other studies provided evidence that saprotrophic fungal diversity was correlated with tree species (Baber et al., 2016;Purahong et al., 2018). Further studies need to be carried out to improve our understanding of the role of tree species in determining the diversity and distribution of saprotrophic macrofungi at a large spatial scale.
Macrofungi are far more challenging to collect than plants and animals, because the presence of fruiting bodies is usually seasonal and ephemeral, and the sampling of cryptic macrofungi can be time consuming and laborious (Mueller et al., 2004). For a long time, attempts to study fungi at a large spatial scale were hindered by the lack of data compilation and integration (Willis, 2018). Thanks to the development of DNA sequencing technology, scientists can rely on the indirect method, involving the isolation of spores and/or mycelia in the environment from soil samples, to study the fungal diversity at continental and even global scales (van der Linde et al., 2018;Tedersoo et al., 2014). However, some researchers suggest that such studies might need to be verified using various data sources and more comprehensive spatial gradients (Andrew, Büntgen, et al., 2019). For example, Porter et al. (2008) showed that the assemblage of fungi at the species level detected from fruiting bodies and soil samples is substantially different. An ideal complement to molecular data might be fungal data from open-source biodiversity data. These freely accessible databases offer access to biodiversity information in the form of species occurrences (Hochmair et al., 2020), in which observations of fungal species are becoming increasingly available worldwide (Andrew, Diez, et al., 2019). Macrofungal occurrence data have fungi are important in driving ecosystem processes, such as nutrient and carbon cycling.

K E Y W O R D S
abiotic factors, biotic factors, climatic variables, ectomycorrhizal, functional group, saprotrophic, species distribution model been collected by amateur mycologists, naturalists and researchers for several hundred years, especially in Europe, which has more data available than any other region (Andrew, Diez, et al., 2019;Mueller et al., 2007). These data form a prime source of knowledge regarding fungal diversity and distribution (Andrew, Diez, et al., 2019).
Knowledge of species diversity and distribution has greatly improved as a result of the increasing availability of open biodiversity data (Maldonado et al., 2015) and the emergence of new techniques to analyse such information (Guisan & Zimmermann, 2000;Wollan et al., 2008). Species distribution models (SDMs; also known as ecological niche models, bioclimatic envelope models and habitat suitability models) are numerical tools that combine species occurrence data with environmental estimates. They are popular tools in quantitative ecology and are used to gain ecological and evolutionary insights and predict the distribution of species across landscapes . Given the increasing availability of extensive spatial environmental data and the growing sophistication of modelling algorithms, SDMs are widely used for various biological groups in the terrestrial, marine and freshwater realms (Cacciapaglia & van Woesik, 2015;Mateo et al., 2016;Niittynen et al., 2020;Oberdorff et al., 2019;Thuiller et al., 2014;Williams et al., 2009). Although this technique is commonly applied to terrestrial vascular plants and animals, its application to the fungal kingdom remains relatively rare Větrovský et al., 2019;Wollan et al., 2008).
In this study, by taking advantage of openly available biodiversity data, we aimed to determine the relative importance of biotic, abiotic and anthropogenic factors at a fine spatial resolution in predicting the diversity and distribution of macrofungi on a continental scale.
Specifically, our aims were as follows: (1) to predict the distribution of two different functional groups of macrofungi (i.e., ectomycorrhizal and saprotrophic macrofungi) at a 5 km spatial resolution from southern Spain to northern Norway, based on a comprehensive set of environmental variables using presence-only SDMs; and (2) to examine the importance of variables that affect the distribution of two different functional groups of macrofungi at the European scale.

| Occurrence data of macrofungal species
For most regions, systematic biological survey data with both presence and absence information for species are often scarce and/or limited in coverage. The Global Biodiversity Information Facility (GBIF; http://data.gbif.org) is an international network devoted to making the world's biodiversity data publicly available. It is currently the largest database of species occurrence observations, gathering species records from different sources (e.g., museums, herbaria, survey projects, collections of individual researchers, citizen science) and other biodiversity platforms world-wide (Chandler et al., 2017;Hochmair et al., 2020). These databases embody the long-term public and private investment in biological science and are hugely important sources of species occurrence data (Elith et al., 2011). The GBIF defines standards for publishing data and metadata, and all occurrence records indexed on the GBIF must satisfy certain data-quality criteria before being included in the platform (GBIF.org, 2020).
According to data-quality requirements, data are published initially as primary occurrence records on GBIF in a standardized format.
Then, GBIF performs additional checks to verify data formats, data quality and their suitability for use. To date, the GBIF network has collected > 21 million occurrence records of fungi across the globe.
For our analysis, we downloaded georeferenced fungal occurrence records from eight European countries (i.e., Finland, France, Spain, Germany, Norway, The Netherlands, Denmark and Sweden) from GBIF that contained a relatively large amount of fungal occurrence data, which were provided mainly by comparatively authoritarian organizations (e.g., museum, university project or biodiversity centre; see Supporting Information Appendix S1). To include as many species as possible in our study and match the corresponding environmental data layer, we extracted the records collected between 1990 and 2018, a period during which there was a relatively stable number of fungal species (Supporting Information Appendix S2: Figure S2.1).
The majority of the macromycetes are in taxa from the phylum Basidiomycota, subphylum Agaricomycotina (Porter et al., 2008). The  Figure S2.2). We therefore screened the species in these seven orders to select macrofungal species to use in our study.
The following procedures were implemented on this subset.
We removed records without a specific species name and/or flagged with taxonomic issues by GBIF; we disregarded fungal records with "coordinate uncertainty in meter" > 5 km. We rarefied records spatially by eliminating all but one point present per species within each 5 km × 5 km grid cell, to avoid double counting of the presence of each species and to reduce overfitting to sampling bias in SDMs (Boria et al., 2014;Kramer-Schadt et al., 2013). Furthermore, we removed species recorded in < 30 grid cells (of 5 km × 5 km) (Wisz et al., 2008) to safeguard the validity of the predictive performance of the SDMs and ensure access to a reasonable number of test data for modelling.
Given that our study focused on two nutritional modes (i.e., saprotrophic and ectomycorrhizal), we compiled checklists of nutritional mode at the species level based on the database published by , Sánchez-García et al. (2020) and Nguyen et al. (2016), in addition to genus-level databases published by Rinaldi et al. (2008) and Tedersoo and Smith (2013). Initially, we assigned a nutritional mode value to records according to the specieslevel checklist; then, for the rest of the species without species-level nutritional mode information, we assigned their nutritional mode according to the genus-level checklist. To ensure the nutritional mode of species with genus-level information, we checked their nutritional mode carefully. Most species within the same genera have a similar lifestyle. However, there are still exceptions, such as species with unsolved lifestyles in the genera of Entoloma, Volvariella, Clitopilus and Arrhenia. We removed species in these genera to ensure our assignment of the fungal nutritional mode at genus level.
As a result, a total of 1,845 species (i.e., 774 ectomycorrhizal macrofungal species and 1,071 saprotrophic macrofungal species) with 878,978 occurrence records were extracted for our modelling, of which > 50% of the species were found in at least seven countries in our study area (Supporting Information Appendix S2: Figure S2.2c). The detailed occurrence statistics are presented per country in Table 1

| Climate
We obtained 19 bioclimatic variables (v.2.0) from the WorldClim database (http://www.world clim.org/) at a resolution of c. 5 km (Supporting Information Appendix S3: Table S3.1). These bioclimatic variables were derived from the monthly precipitation and temperature values between 1970 and 2000 and processed further to generate more biologically meaningful variables, such as annual trends, seasonality and extreme factors determining the species niches (Fick & Hijmans, 2017).

| Soil
We chose eight soil properties from the topsoil layer (0-30 cm) that characterize the fertility, soil water and rooting conditions to repre-   (Brus et al., 2012), we calculated the richness and diversity of the selected tree species in each 5 km grid cell using the "vegan" package (Dixon, 2003)  and ancillary geospatial data (Simard et al., 2011). For each 5 km grid cell, we calculated the mean and the standard deviation of the forest canopy height using Zonal Statistics in Spatial Analyst toolboxes in arcGIS.
SPOT VEGETATION (VGT), a normalized difference vegetation index (NDVI) product derived from the Satellite Pour l'Observation de la Terre (SPOT) with a 1 km spatial resolution, was also used. We Initially, we first calculated one 12-month (January-December) time series NDVI by averaging the 12-year SPOT-VGT NDVI series. We then used an adaptive Savitzky-Golay smoothing filter to remove the noise in the averaged NDVI time series using the TIMESAT package F I G U R E 1 Location of the study area in Europe and the distribution of (a) saprotrophic macrofungi and (b) ectomycorrhizal macrofungi occurrence data used in this study (Jönsson & Eklundh, 2004). Finally, we calculated the annual mean, minimum, maximum and standard deviation value of NDVI.

| Topography
We used the digital elevation model (DEM) data at c. 1 km resolution from the United States Geological Survey (USGS) Earth Resources Observation and Science (EROS) Data Center (https://www.usgs. gov/cente rs/eros) to compute topographic variables, such as elevation, slope and terrain curvature, using the Spatial Analyst toolbox for arcGIS. For each 5 km grid cell, we calculated the standard deviations of elevation and slope, in addition to the curvature and range of slope using the Zonal Statistics tools from the Spatial Analyst toolbox in arcGIS.

| Anthropogenic variables
In this study, we considered three anthropogenic variables: landcover and land-use data, forest management approach and human footprint index.
We downloaded the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Cover Type (MCD12Q1) v.6 data from the NASA Land Processes Distributed Active Archive Center (https://lpdaac.usgs.gov/produ cts/mcd12 q1v00 6/). To ensure consistency of the data with other remote sensing-derived land-cover products, we followed the classification scheme used in the study by Tuanmu and Jetz (2014) and reclassified the MODIS land-cover data into their 12 categories.
We also obtained the European dominant forest management approach (FMA) data presented by Hengeveld et al. (2012). This dataset provided information about different stages in forest succession and could be used to imply forest disturbance and determine the services that the forest provided .  (Sanderson et al., 2002).
In total, we selected 47 potential predictor variables covering five categories (i.e., climate, soil, vegetation, topography and anthropogenic influence; Table 2; Supporting Information Appendix S3:

| Collinearity analysis
Multicollinearity is a common feature of ecological data. It inflates the variance of regression parameters, potentially leading to bias and uncertainty in identifying relevant predictors in a statistical model (Dormann et al., 2013). To avoid potential multicollinearity issues in model fitting and reduce the redundancy of input predictor variables, we performed a correlation analysis. We calculated the Pearson correlation coefficient and created a correlation matrix using the "corrplot" package (Wei & Simko, 2017) in R v.3.6.1 (R Core Team, 2013) (Supporting Information Appendix S3: Figure S3.1). We retained variables with low correlation (|r| < .7) (Dormann et al., 2013). For those variables that were highly correlated (|r| > .7), we selected the variables that were most ecologically relevant to fungi based on expert knowledge and scientific literature (Dormann et al., 2013;Moore et al., 2011;Wollan et al., 2008). Ultimately, 25 of the 47 variables were selected for use in our model (Table 2).

| Species distribution modelling
Multiple SDMs have been proposed in the past two decades, with different methods depending on the type of species data (presence-absence or presence only). Maximum Entropy (MaxEnt) is a program based on a machine-learning algorithm for modelling species distributions from presence-only species records (Phillips et al., 2006). It can handle complex interactions between predictor and response variables efficiently to produce a continuous estimate of habitat suitability values ranging from zero (least suitable) to one (most suitable) (Elith et al., 2006).
MaxEnt is considered to be able to cope well with sparse, irregularly sampled data to produce robust results (Kramer-Schadt et al., 2013). In comparison to many other algorithms used in species distribution modelling, MaxEnt is much less sensitive to sample size and has the best predictive power across all sample sizes (Wisz et al., 2008). We ran all models with MaxEnt v.3.4.4 software (Steven et al., 2017) in our study.
Given that absence data are not available in the macrofungal database, MaxEnt requires additional data representing the range of environmental conditions in the modelled region. These data are known as background or pseudo-absence data and are usually generated randomly across the study area (Merow et al., 2013;Phillips et al., 2009).
However, occurrence data in natural history museums and herbaria are often thought to be spatially biased in surveys (e.g., toward easily accessible areas), and these differences between occurrence collection distribution and random distribution of background sampling might lead to inaccurate models (Phillips & Dudík, 2008;Phillips et al., 2009) For each species, we chose 10,000 biased background points in the study area according to the kernel density estimation layer to generate "pseudo-absence" data. We set 10-fold cross-validation in the "replicates" option for testing the model performance of each species and checked the built-in jackknife test to measure the variable importance (Phillips et al., 2006).
To evaluate model performance, we used both thresholdindependent and threshold-dependent methods in our study. With the threshold-independent method, we used the area under the receiveroperating characteristic curve (AUC), where higher values represented higher accuracy (Swets, 1988). Although AUC has been criticized, it is the standard method to assess prediction accuracy and has been used extensively in the species distribution modelling literature because of its threshold independence and the ease of interpreting its results (Elith et al., 2006;Massada et al., 2013;Phillips et al., 2006).
As a threshold-dependent method, we used true skill statistics (TSS; Allouche et al., 2006) to evaluate the performance of the models.

| Calculating distribution probability and diversity of macrofungi
Using the MaxEnt model, we predicted the distribution probability of each species across the study area. We calculated the average distribution probability of species having the same nutritional mode in each grid cell to produce the general distribution probability map for both saprotrophic macrofungi and ectomycorrhizal macrofungi. Max SSS can minimize the error rate of both commission and omission errors, which has been widely used in SDMs as a robust method for threshold selection when only presence data are available (Liu et al., 2013). The binary outcome of each species was summed per nutritional mode for each grid cell to yield the species richness of ectomycorrhizal and saprotrophic macrofungi separately. To show the latitudinal and longitudinal variation of species richness, we calculated the mean species richness of all grids at the same coordinate at an interval of 5 km along latitude and longitude, respectively, using the stacked binary map of the two functional groups.

| Model performance
Overall, the models for both saprotrophic macrofungi and ectomycorrhizal macrofungi performed well, with a mean AUC of .88 and a mean TSS of .69 for saprotrophic macrofungal species, and a mean AUC of .90 and a mean TSS of .72 for ectomycorrhizal macrofungal species (Figure 2). The standard deviation of the cross-validation is depicted in the Supporting Information (Appendix S3: Figure S3. 2) to show the predictive uncertainty of the model.

| Predicted distribution and diversity of macrofungi in Europe
Of the whole study area, eastern Denmark and southern Sweden

| Factors affecting the diversity and distribution of macrofungi in Europe
The top three variables that contributed most among the 25 environmental factors used in our model were dominant tree species, human footprint index and tree cover density ( Figure 5; Supporting Information Appendix S3: Figures S3.3 and S3.4). The distribution probability of both macrofungal groups varied for the 20 different dominant tree species, with a significantly (p = .05) higher distribution probability in areas dominated by Fagus spp. and Picea spp.
Saprotrophic macrofungi showed the highest distribution probability in areas dominated by Fagus spp. (Figure 6a), whereas ectomycorrhizal macrofungi showed the highest distribution probability in areas dominated by Picea spp. (Figure 6d). We found an increasing richness with increasing tree cover density, which stabilized when the tree cover density exceeded a value of c. 30% (Figure 6c,f). The human footprint index also provided an important contribution to the distribution of two functional groups of macrofungi ( Figure 5).
The response curve of the human footprint index showed that the distribution probability of both groups of macrofungi increased as the human footprint index rose (Figure 6b,e).

| D ISCUSS I ON
This is the first fine-resolution (i.e., 5 km) study of macrofungal diversity and distribution at the continental scale, covering a broad latitudinal range from northern to southern Europe, taking a comprehensive set of environmental variables into consideration. We found that tree species and human disturbance appeared to be more important than climatic variables in influencing the diversity and distribution of macrofungi in Europe.

| Relative importance of tree species
Our results showed that vegetation-related variables, such as dominant tree species, tree cover density and forest age, provided a larger contribution to ectomycorrhizal macrofungi than climate.

F I G U R E 3
The average distribution probability per grid cell of (a) saprotrophic macrofungal species and (b) ectomycorrhizal macrofungal species Ectomycorrhizal macrofungi were shown to have a significantly higher distribution probability in areas dominated by Picea spp. and Fagus spp. than in areas dominated by other tree species. Local and regional studies have also found tree species to be a key factor affecting the diversity and distribution of ectomycorrhizal fungal (Bahram et al., 2012;Ishida et al., 2007;Toljander et al., 2006).
Spruce and beech are among the most important tree species of boreal and temperate forest, respectively, and support a very diverse community of fungal symbionts (Taylor et al., 2000). Many tree species might grow poorly or even fail to grow without the presence of ectomycorrhizas (Moore et al., 2011). However, at a continental and global level, climate rather than tree species was found to be the major factor driving the ectomycorrhizal macrofungal diversity (Andrew, Büntgen, et al., 2019;Tedersoo et al., 2014). Andrew, Büntgen, et al. (2019) found, using open-source macrofungal data, that climate, rather than tree species, was the prevailing driver shaping ectomycorrhizal fungal diversity patterns in central to northern Europe. However, their study was based on a grid size of 20 km × 20 km, which is much coarser than the grid used in our study. With such coarse spatial resolution, the detailed variation in biotic factors such as tree species composition might be blurred owing to the presence of a high proportion of small forest patches, probably enhancing the role of climate. For example, in Central-West Europe > 45% of the forests are < 10,000 ha (a grid size of 10 km × 10 km) (Köhl & Linser, 2020). The environmental variables driving species diversity can be different depending on the spatial scale, including both the extent and the resolution (grain size) over which species diversity is measured (Guisan & Thuiller, 2005;Willis & Whittaker, 2002). another study found that the assemblage of fungi at the species level detected using above-ground fruiting bodies and soil DNA sampling was significantly different (Porter et al., 2008). It is worth noting that these two approaches might not be directly comparable if the vast majority of fungi at a site do not produce fruiting bodies. However, we believe that the study based on different sampling methods (i.e., In compared to ectomycorrhizal fungi, saprotrophic fungi were thought to lack preference for specific tree species in temperate forests (Purahong et al., 2018), and some large-scale studies found no association between tree species and saprotrophic fungi (Andrew, Büntgen, et al., 2019;Tedersoo et al., 2014). In contrast, we found that the dominant tree species was an important factor influencing the diversity and distribution of saprotrophic macrofungi, which were shown to have a significantly higher distribution probability in areas dominated by Fagus spp. and Picea spp. than in areas dominated by of differing quantity and quality (Purahong et al., 2018), affecting the establishment of saprotrophic fungi. For instance, in deciduous forests, fallen trunks and branches constitute the greater part of dead wood, forming a key habitat for saprotrophic fungi (Boddy et al., 2007). Our results provide further evidence for tree species preference by saprotrophic macrofungi at a large spatial extent.
We note that European forests are dominated by Norway spruce (Picea abies), Scots pine (Pinus sylvestris) and European beech (Fagus sylvatica), together covering c. 50% of the European Union forest area (Brus et al., 2012). Theoretically, a dominant host is likely to have a stronger selective pressure on the overall species pool (Heilmann-Clausen, 2003). Therefore, the wide distribution of Picea spp. and Fagus spp., covering a broad range of environmental conditions, might provide high resource availability, providing ecological niches for macrofungi and explaining the importance of European beech and Norway spruce for fungal species richness at a continental level in Europe. and human activity has generally been found to have a negative impact on biodiversity world-wide (Pimm & Raven, 2000). However, widespread changes to species richness as a result of human niche creation are increasingly being discerned (Boivin et al., 2016). The anthropogenic alteration of the landscape often creates novel habitats (Sanderson et al., 2002), thereby increasing the abundance of some naturally rare habitats and benefitting certain groups of fungi (Boddy et al., 2007). For instance, the use of woodchips in horticulture has led to a strong increase in a number of saproxylic fungi (Shaw et al., 2004). Human movement (e.g., hunting or the extraction of resources from the ecosystem) might also provide a mechanism for species dispersal (Sanderson et al., 2002;Walsh et al., 2017) Interestingly, Pautasso (2007) and Pautasso and Chiarucci (2008) found that the positive relationship between human disturbance and species richness was scale dependent. They observed that there was usually a negative correlation between human disturbance and the species richness of plants and vertebrates at the local or landscape level. However, this correlation became positive as the study scale became coarser. Nevertheless, we cannot rule out the problems of whether there are potentially more observations in areas closer to cities or highly populated areas in our study. Therefore, the relationship between macrofungi and human disturbance at different spatial scales will require further study.

| Macrofungal diversity and distribution patterns in Europe
Open-access databases provided a large volume of macrofungal observations across an extensive area over a long period. Combined with robust species distribution modelling techniques, presence-only data are adequate to predict habitat suitability and indicate where the species is most likely to occur (Merow et al., 2013). Although these datasets offer the possibility of using predictive models to develop distribution maps (Steen et al., 2019), such data might still be suboptimal compared with carefully designed surveys. For example, there were large differences in the number of macrofungal records held across the eight countries in our study (Supporting Information Appendix S2: Table S2.1); hence, there might be bias in the database owing to differences in sampling level. However, careful data culling provided a potential mechanism for improving predictions from open-access data sources (Steen et al., 2019). Indeed, spatial thinning of species occurrence records and the biased background data sampling method we used in this study will mitigate the possible sampling bias further. There is emerging evidence that SDMs developed using open-source but relatively poorly structured data have the potential to perform as well as those using systematic data (Steen et al., 2019). In a general context, our study using open-access datasets analysed in conjunction with robust SDMs appears useful, especially when alternative data are unavailable.
Our results revealed that eastern Denmark and southern Sweden are likely biodiversity hotspots for both functional groups of macrofungal species. Southern Sweden covers both nemoral and boreonemoral zones (Rydin et al., 1997). The forest in the nemoral zone mainly consists of deciduous broadleaf trees, such as beech and oak, whereas the boreonemoral zone is dominated by either coniferous or broadleaf trees, providing ideal habitats for numerous species, including macrofungi (Nilsson et al., 2001;Nordén et al., 2007).
Denmark has also been found to have very high fungal species richness in its beech forest compared with some other European countries (Ódor et al., 2006). There is a relatively large area of oldgrowth deciduous broadleaf forest in eastern Denmark (Sabatini et al., 2018), which is of key importance to the macrofungal diversity in Europe (Dvořák et al., 2017), because numerous forest fungi have been confined to these old-growth forests according to field observations and red lists (Senn-Irlet et al., 2007). Fagus spp. and Picea spp., which are important hosts for macrofungi (Küffer et al., 2008;Moore et al., 2011), are widely distributed across eastern Denmark and southern Sweden (Supporting Information Appendix S3: Figure   S3.5). Ideal habitat conditions, plus the wide distribution of host tree species and old-growth forest, are likely to be important reasons for the high diversity of macrofungi in these areas.

| Conclusions
By taking a group of relatively conspicuous macrofungal species, we investigated the diversity and distribution of two different functional groups of macrofungi at a 5 km spatial resolution at a continental level across eight European countries. We analysed the importance of variables that affect the distribution of macrofungi.
Overall, our study demonstrated that tree species and human disturbance played more important roles than climatic factors in determining the diversity and distribution of macrofungi at a continental scale in Europe. Our study indicates that fungal diversity and distribution might change considerably if the strongest predictors (i.e., tree species) are altered by climate change and/or human activity.
With changes in fungal diversity, other ecological processes might also be transformed, because fungi are important drivers of ecosystem processes, such as nutrient and carbon cycling.

ACK N OWLED G M ENTS
We are grateful to the many mycologists and researchers who contributed and shared their fungal data via GBIF. This work was sup-

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analysed during the present study are available through the GBIF portal (https://www.gbif.org/). Download links for specific datasets can be found in

R E FE R E N C E S
Allouche, O., Tsoar, A., & Kadmon, R. (2006). Assessing the accuracy of species distribution models: Prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology, 43 (6)