The FunDivEUROPE project, short for "Functional Significance of Forest Biodiversity in Europe," aimed at exploring the intricate relationships between forest biodiversity and ecosystem functionality, focusing specifically on European forests. The project was a collaborative effort involving scientists from multiple disciplines and institutions. It entailed a comprehensive, large-scale assessment of forest biodiversity and its impact on ecosystem functions in a network of observational plots spanning the European continent. This extensive network enabled us to systematically examine how variations in tree species diversity and functional traits influenced key ecosystem functions.

In total, 209 mature forest plots measuring 30 x 30 meters were located in six European countries, ranging from boreal to Mediterranean zones, with each representing a major European forest type: Finland (boreal forest), Poland (hemiboreal forest), Germany (temperate deciduous forest), Romania (mountainous deciduous forest), Italy (thermophilous deciduous forest), and Spain Mediterranean mixed forest). Richness levels of one, two, three, four, and five target species were replicated within and across regions.

A major strength of the FunDivEUROPE project was the general philosophy to measure all ecosystem functions in all plots, following the same protocol by the same observers across the six forest types. In each of the 209 plots, 27 ecosystem functions were measured.

Here, we present data on a high number of basic data for each of the 209 plots, describing geographic and geomorphological, as well as soil and bedrock characteristics, climate variables, and several measures of tree diversity. We further show data of the 27 ecosystem functions, which were classified into six groups reflecting basic ecological processes, and which have established links to supporting, provisioning, regulating or cultural ecosystem services. Details about the measurement protocols are provided.

Major results from the FunDivEUROPE project shed light on the fundamental importance of biodiversity in European forests. The project revealed that diverse forests tend to be more resilient to disturbances, sequester more carbon, and provide enhanced diversity of forest-associated taxa. Moreover, the study highlighted the crucial role of particular tree species and functional traits in shaping ecosystem services and functions. The findings of FunDivEUROPE thus offer insights for forest management and conservation practices, advocating for the preservation and restoration of diverse forest ecosystems.

General design

The FunDivEUROPE project, short for "Functional Significance of Forest Biodiversity in Europe," aimed at exploring the intricate relationships between forest biodiversity and ecosystem functioning, focusing specifically on European forests (Baeten et al., 2019; Baeten et al., 2013; Ratcliffe et al., 2017; van der Plas et al., 2016a; van der Plas et al., 2016b; van der Plas et al., 2018). In total, 209 mature forest plots measuring 30 x 30 meters were located in six European countries, ranging from boreal to Mediterranean zones, and with each representing a major European forest type: Finland (28 plots, boreal forest), Poland (43 plots, hemiboreal forest), Germany (38 plots, temperate deciduous forest), Romania (28 plots, mountainous deciduous forest), Italy (36 plots, thermophilous deciduous forest), and Spain (36 plots, Mediterranean mixed forest). These plots were primarily established to investigate the role of the richness of regionally common and economically important ‘target’ species on ecosystem functioning and were hence selected to differ as much as possible in the richness of these. Plot selection was aimed at mimicking the design of a biodiversity experiment, in which variation in environment is minimized and diversity is not confounded with composition, as in most observational studies of diversity. Hence, plots were carefully selected so that correlations between tree species richness and community composition, topography (slope, altitude), and potentially confounding soil factors (texture, depth, pH) were minimized, thus ensuring robust tests of diversity-ecosystem function relationships (comparative study design). Most forest plots were historically used for timber production but are now managed by low-frequency thinning or with minimal intervention. Hence, species compositions and diversity patterns in forests are predominantly management-driven and/or are the result of random species assembly, from the regional species pool. All sites are considered as mature forests.

In total, there were 15 target species across all 209 plots, and plots were selected so that almost all possible combinations of these target species were realized. Target species contributed to more than 90% of the tree biomass in the plots and therefore we expected them to be most important for ecosystem functioning. Richness levels of one, two, three, four, and five target species were replicated 56, 67, 54, 29, and 3 times, respectively, across countries, and most possible target species compositions were realized. For the majority of species combinations, we included two or more “realizations” (not strict replicates, because species abundances differ), which allows for comparing the importance of species diversity with that of species composition for this subset of plots. At each richness level, each target tree species was present in at least one plot, allowing us to statistically test for the effects of presence/absence of species on ecosystem functioning. Since species evenness might also affect ecosystem functioning, all plots were selected to have target species with similar abundances (with Pielou’s evenness values above 0.6 in > 91% of the plots). To reach this goal, we a priori decided to exclude locally rare target species (<2 individuals per plot) in richness measures. To describe community composition and to estimate biomass values of each tree in each plot, we identified all stems ≥7.5 cm in diameter to species and permanently marked them (12,939 stems in total). More details about the design of the FunDivEUROPE plot network can be found in Baeten et al. (2013).

We determined a high number of basic data for each of the 209 plots, describing geographic and geomorphological, as well as soil and bedrock characteristics, see also Ratcliffe et al (2017). Soil pH was determined in the same samples used for C and N determination (see below) with a 0.01M CaCl₂ solution at a ratio of 1:2.5 using a 827 pH labs Metrohm AG, Herisau, Switzerland; see details in Dawud et al. (2017). For each plot, we extracted mean annual temperature, temperature seasonality (standard deviation of mean monthly temperatures), annual precipitation, and precipitation seasonality (standard deviation of mean monthly precipitation) from the WorldClim dataset (interpolated from measurements taken between 1960 and to 1990 and at a spatial resolution of one square kilometer) and the slope from the GTOPO30—digital elevation model with a spatial resolution of one square kilometer (data available from the U.S. Geological Survey); see details in Kambach et al. (2019). We further quantified several measures of tree diversity, based on the initial inventory made in each plot, see Baeten et al. (2013). Short description of all these variables are available in the “Metadata” sheet of the data file.

Ecosystem functions methodology

A major strength of the FunDivEUROPE project was the general philosophy to measure all ecosystem functions in all plots, following the same protocol by the same observers across the six forest types. Measurements are thus directly comparable across plots and show high coverage.

In each of the 209 plots, 27 ecosystem functions were measured. The functions were a priori classified into six groups reflecting basic ecological processes (groups 1 to 5 below), and which have established links to supporting, provisioning, regulating, or cultural ecosystem services. These functions were also used in Chao et al. (in press): Hill-Chao numbers allow decomposing gamma-multifunctionality into alpha and beta components. Ecology Letters. In addition, we quantified timber quality as an additional ecosystem service.  

In the following, we describe the methodology for each measured ecosystem function/service. (For more details, see also Baeten et al., 2019; Ratcliffe et al., 2017; van der Plas et al., 2016a; van der Plas et al., 2016b; van der Plas et al., 2018), and other FunDivEUROPE publications that focus on specific ecosystem properties and functions. Additional datasets are stored in the FunDivEUROPE data portal (https://data.botanik.uni-halle.de/fundiveurope/, logon required to view most data; all metadata is publicly available).

1. Nutrient and carbon cycling-related drivers (header in the data table in parentheses):

a.       Earthworm biomass:  Biomass of all earthworms [g m^-2] (earthworm_biomass)

Earthworm sampling was carried out in spring 2012 in Italy, Germany, and Finland, and in autumn 2012 in Poland, Romania, and Spain. Plots were divided in nine (10 x 10) m subplots. One sample per plot was taken in the center subplot. Sampling close to tree stems was avoided and whenever possible performed, in between multiple, different tree species. At each sampling point, earthworms were sampled by means of a combined method. First litter was handsorted over an area of (25 x 25) cm². After litter removal over an enlarged area of 0.5 m², ethological extraction using a mustard suspension was applied. Finally, hand sorting of a soil sample of (25 × 25) cm² and 20 cm depth was performed in the middle of the 0.5 m² area. Earthworms were preserved in ethanol (70%) for two weeks, and transferred to a 5% formaldehyde solution for fixation (until constant weight), after which they were transferred to ethanol (70%) again for further preservation and identification. All worms were individually weighed, including gut content, and identified to species level.  Results per unit area of the three sampling techniques were summed to determine the total earthworm biomass per m². For details on earthworm biomass measurements, we refer to De Wandeler et al. (2018; 2016).

b.      Fine woody debris: Number of snags and standing dead trees shorter than 1.3 m and thinner than 5 cm DBH, and all stumps and other dead wood pieces lying on the forest floor (fine_woody_debris)

Fine woody debris (FWD) was measured in two circular subplots (radius of 7 m) located in the opposite corners of each plot. All standing dead trees thinner than 5 cm diameter at breast height and snags shorter than 1.3 m, and all stumps and other dead wood pieces lying on the forest floor, were surveyed. In this study, we used the number of FWD pieces in each plot.

c.       Microbial biomass: Mineral soil (0–5cm layer) microbial biomass carbon [mg C kg^-1] (microbial_biomass_mineral)

For soil sampling, each of the 209 plots was divided into nine 10x10m subplots. A soil sample was taken from five of the nine subplots and mixed to obtain one representative composite sample from each plot. Forest floor and mineral soil horizons (0-5 cm) were sampled separately. Soils were sieved fresh (4mm), stored at 4°C and analyzed within two weeks. Sampling was performed in spring 2012 in Italy, Germany, and Finland, and in autumn 2012 in Poland, Romania, and Spain. No forest floor was collected from the plots in Germany.

Soil microbial biomass C was determined by the chloroform fumigation extraction method, of 10g and 15g (organic and mineral soil, respectively) soil, followed by 0.5 M K₂SO₄ extraction of both fumigated and unfumigated soils (soil:solution ratio, 1:5). Fumigations were carried out for three days in vacuum desiccators with alcohol-free chloroform. Extracts were filtered (Whatman n° 42), and dissolved organic carbon in fumigated and unfumigated extracts was measured with a Total Organic Carbon analyser (Labtoc, Pollution and Process Monitoring Limited, UK). Soil microbial biomass C was calculated by dividing the difference of total extract between fumigated and unfumigated samples with a kEC (extractable part of microbial biomass C after fumigation) of 0.45 for biomass C (Joergensen and Mueller, 1996).

d.      Soil carbon stocks:  Total soil carbon stock in forest floor and 0–10 cm mineral soil layer combined [Mg ha^-1] (soil_c_ff_10)

Soil sampling was carried out from May 2012 to October 2012 (i.e. Poland in May 2012, Spain in June 2012, Finland and Germany in August 2012, Romania in September 2012 and Italy in October 2012). Nine forest floor samples and nine cores of mineral soil were collected from each plot and these were subsequently pooled into one sample per plot by each soil layer, i.e. forest floor, 0–10cm and 10–20cm depths for samples from Germany, Finland, Italy, and Romania. For Poland, the fixed depth was extended to 20–30cm and 30–40 cm whereas for Spain it was only possible to sample up to the 0–10cm layer due to the stoniness of the site. We oven-dried the samples at 55°C to constant weight, sorted out stones and other materials, ground the forest floor first with a heavy-duty SM 2000-Retsch cutting mill, and we then took subsamples and ground it further into finer particles with a planetary ball mill (PM 400-Retsch) for six minutes at 280rpm. The mineral soil samples were sieved through 2mm diameter mesh. We carried out carbonate removal treatments for those soil samples whose pH value exceeded the threshold point and proved presence of carbonates when tested with a 4N HCl fizz test. We used 6% (w/v) H₂SO₃ solution and followed the carbonate removal procedure described by (Skjemstad and Baldock, 2007). We took subsamples and further ground it into finer particles with a planetary ball mill (PM 400-Retsch) for six minutes at 280 rpm before analyzing soil organic carbon (SOC) with a Thermo Scientific FLASH 2000 soil CN analyzer. Soil organic C stocks were estimated by multiplying the SOC concentrations with soil bulk density, relative root volume and relative stone volume using the formula described in Vesterdal et al., (2008). We also determined the moisture content of the soil samples by oven-dried subsamples at 105°C and the reported SOC stock is thus on 105°C dry weight basis. 

2. Nutrient cycling related processes

a.       Litter decomposition: Decomposition of leaf litter using the litterbag methodology [% daily rate] (litter_decomp_day)

Litter collection and litterbag construction

Leaf litter from all target tree species of the cross-region exploratory platform was collected at tree species-specific peak leaf litter fall between October 2011 and November 2012. Except for the Finnish forests, where freshly fallen leaf litter was collected from the forest floor, litter was collected using suspended litter traps, which were regularly harvested at one to two-week intervals. In all cases, litter was collected nearby, but not within the experimental plots. Litter was then air-dried and stored until the preparation of the litterbags.

Litterbags (15 x 15 cm) were constructed using polyethylene fabrics of two different mesh sizes. For the bottom side of the litterbags, we used a small mesh width of 0.5 x 0.5 mm in order to minimize losses of litter fragments, while for the upper side, we used a large mesh width of 5 x 8 mm to allow soil macrofauna access to the litter within bags. Litterbags were filled with 10 g of litter. For litter mixtures, litterbags were filled with equivalent proportion of each litter species. Subsamples of all litter species were weighed, dried at 65°C for 48 h and reweighed to get a 65°C dry mass correction factor.

Litterbag incubation

Within each experimental plot, three litterbags with the plot-specific litter type (either single litter species or specific mixtures) were placed on bare soil after the natural litter layer had been removed, and fixed to the soil by placing chicken wire on top of it. The litterbags were removed from the field when 50–60% of the initial litter mass of the region’s fastest decomposing species was remaining (evaluated with an extra set of litterbags that were harvested regularly). As a consequence, the duration of litter decomposition varied among regions. This procedure ensured that litter was sampled at similar decomposition stages across all sites, facilitating meaningful comparisons of litter diversity effects.

Litter processing

Harvested litterbags were sent to Montpellier where they were dried at 65°C. Litter was cleaned of pieces of wood, stones or other foreign material that occasionally got into the litterbags. Litter was then weighed, ground to a particle size of 1 mm with a Cyclotec Sample Mill (Tecator, Höganäs, Sweden). To correct for potential soil contamination during decomposition in the field, we determined the ash content of initial and final litter material on all samples and expressed litter mass loss on ash-free litter mass. 

Litter mass loss was expressed as the percentage of mass lost from each litterbag, calculated as followed: Mass Loss = 100 x (Initial (ash free) mass – Final (ash free) mass)/Initial (ash free) mass. For details on litter decomposition measurements, we refer to Joly et al. (2017; 2023).

b.      Nitrogen resorption efficiency: Difference in N content between green and senescent leaves divided by N content of green leaves [%] (nutrient_resorption_efficiency)

In each plot, fresh leaf and needle samples were collected from the south-exposed sun crown of all dominant tree species during the growing season (June to August) of 2012 and 2013. Twigs with leaves and needles were cut down from six trees per species in the monocultures and from three trees per species in the mixtures. Depending on the local conditions, tree loppers, tree climbers, or ruffles were used for this purpose. The selected material was placed in paper bags and was either oven-dried or air-dried, depending on the facilities available. Furthermore, collection of leaves from the litter traps, as representative of senescent leaves, has been conducted at periods of maximum litterfall during 2012 and 2013. For this purpose, five litter traps per plot were established and the collected litter was separated into the different species it originated from (see “Litter production” below). All samples were ground and analysed for nitrogen and calcium content by means of Near Infra Red Spectroscopy (NIRS) as described in detail by Pollastrini et al. (2016a). For the calibration of the NIRS spectra for the Ca analysis, a subset of samples was analysed with an atom absorption spectrometer (AAS, iCE 3000 series, ThermoScientific, China). Nitrogen resorption efficiency was calculated as follows, taking into account the N content of green and senescent leaves:

NRE(%) = 100 x ((N green leaves - N senescent leaves)/(N green leaves))

Furthermore, the estimated NRE was corrected in order to take into account the leaf mass loss occurring during senescence. Thus, NRE was corrected based on the Ca foliar concentration, since Ca is rather immobile and is not resorbed during senescence (Van Heerwaarden et al., 2003). To validate the correction of NRE based on Ca concentrations, the Mass Loss Correction Factors (MLCF) suggested by Vergutz et al. (2012) have also been used.

c.       Soil C/N ratio: Soil C/N ratio in forest floor and 0–10 cm mineral soil layer combined (soil_cn_ff_10)

Soil sampling was carried out between May 2012 and October 2012 in all the regions. Nine forest floor samples were collected using a 25 x 25 cm wooden frame, and the mineral soil (0-10 cm layer) was sampled, after forest floor removal, using a cylindrical metal corer. Total soil carbon and nitrogen concentrations were measured with a Thermo Scientific FLASH 2000 soil CN analyser on the forest floor and 0-10 cm layer samples. For full details on soil carbon and nitrogen methodology see Dawud et al. (2017).

d.      Wood decomposition: Decomposition of flat wooden sticks placed on forest floor [% daily rate] (wood_decomp_day)

Flat wooden sticks (wooden tongue depressors made of Betula pendula wood) were placed to decompose at each plot of the exploratory platform. Each wooden stick was initially weighed (average of 2.5 g). As the weighing was done on air-dry sticks, subsamples were weighed, dried at 65°C for 48 h and reweighed to get a 65°C dry mass correction factor.

Within each plot, three wooden sticks were placed on the bare soil after the natural litter layer had been locally removed, and fixed to the soil by placing chicken wire on top of it. The wooden sticks stayed in the field for different durations among regions depending on the mass loss of the region’s fastest decomposing litter species (target of 50 to 60 % mass remaining), that was placed in the field at the same time as the wooden sticks. 

After field exposure wooden sticks were harvested, dried at 65°C, and weighed. Mass loss of wooden sticks was expressed as the percentage of initial mass lost, calculated as followed: Mass Loss = 100 x (Initial mass – Final mass)/Initial mass. For details on wood decomposition measurements, we refer to Joly et al. (2017; 2023).

3. Primary production

a.       Fine root biomass: Total biomass of living fine roots in forest floor and 0-10 mineral soil layer combined [g m^-2] (root_biomass)

On each plot for determining fine root biomass, nine soil samples were taken from a predefined grid. The sampling was done in the six countries during May-October 2012. The forest floor was sampled using a wooden frame of size 25 cm x 25 cm, and thereafter the mineral soil was sampled using a cylindrical metal corer with 36 mm of inside diameter. The mineral soil was sampled down to 20 cm, except for the plots in Poland (down to 40 cm) and in Spain (down to 10 cm). Samples were pooled by layer and plot into one sample. Living fine roots (diameter ≤ 2 mm) were separated from the soil samples by hand to two categories, tree roots and ground vegetation roots. After separation, the roots were washed with water to remove adhering soil. Subsequently, the roots were dried at 40°C until constant mass and weighed for biomass. The root biomass was corrected with a correction factor for soil stoniness (CFstones= 100-(% stones)/100), where the respective volumetric stoniness was estimated with the metal rod method (Tamminen and Starr, 1994) on each plot. For this study, total tree fine root biomass for each plot was calculated (g m^-2) for the sampled soil layer (forest floor + sampled mineral soil). For further details, see also Finér et al. (2017).

b.      Leaf mass: Leaf Area Index (lai)

As a proxy for the leaf mass of each plot, we used the Leaf Area Index (LAI), which is the projected leaf area per unit of ground area. Five measurements of LAI in each plot were carried out at two time points, either early in the morning (shortly before sunrise) or late in the evening (shortly after sunset) in order to work in the presence of diffuse solar radiation and thus reduce the effect of scattered blue light in the canopy. LAI measurements were carried out in early September 2012, before the beginning of leaf shedding, using a Plant Canopy Analyzer LAI-2000 (LI-Cor Inc., Nebraska). With the LAI-2000, the incident light above the canopy and the light transmission below the canopy were measured using one sensor with five fisheye light sensors (lenses), with central zenith angle of 7°,23°, 38°, 53° and 68° (LAI-2000 manual, Li-Cor). The protocol used in each plot consisted of five measurements within the plots (light transmission below the canopy), and five measurements outside the forest (as proxy of the light incidence above the canopy), in an open space that was in close proximity of the sampled plots. LAI data were processed using Li-Cor’s FV2200 software (LI-COR Biogeosciences, Inc. 2010). The light transmittance measurements of the fifth ring were removed to minimise the boundary effects on LAI. The LAI value per plot was the mean value of the five measurements for each plot.  For full details of the LAI measurement, see Pollastrini et al., (2016a)

c.       Litter production: Annual production of foliar litter dry mass [g] (leaf_litter_production)

In each of the 209 plots, five geodetic litter traps of 0.5m² collection surface were installed in a regular grid. The sampling period covered a whole year and litters were collected several times. Sampling frequency was irregular and depended on working capacity within a region and seasonality of litter production. The litter was pooled per plot, and stored in plastic bags for transportation from the field site to the local laboratories. After air-drying, litter samples were sorted by species and by different fractions for dry weighing and chemical analysis. The following fractions were used: foliar litter (leaves or needles), woody litter (twigs, branches, bark parts), reproductive litter (flowers, cones, fruits, seeds, fruit capsules, etc.), other (e.g. bud scales, indefinable or small parts). Here, only the foliar litter is reported. A subsample of all litter types per species and region was dried at 65°C to constant weight to determine the conversion factor from air-dried to oven-dried values of litter dry mass (g).

d.      Photosynthetic efficiency: Chlorophyll fluorescence methodology [ChlF] (photo_eff_tot)

Photosynthetic efficiency was measured using chlorophyll fluorescence (ChlF). ChlF measurements were replicated on eight randomly chosen leaves per tree from both the top and the bottom of the crown. The measurements were done on the twigs after the dark adaptation (i.e. after a minimum of 4 hours in a black plastic bag, at ambient temperature). In evergreen conifers, chlorophyll fluorescence measurements were taken in the current year’s needles (i.e. needles sprouted in 2012). For full details of the ChlF measurement see Pollastrini et al. (2016b).

e.       Tree productivity: Annual aboveground wood production [Mg C ha^-1 yr^-1] (tree_growth)

Wood cores

Tree ring data were used to reconstruct the past annually resolved wood production. Between March and October of 2012, bark-to-pith increment cores (5 mm in diameter) were collected for a subset of trees in each plot following a size-stratified random sampling approach (Jucker et al., 2014a). We cored 12 trees per plot in monocultures and six trees per species in mixtures (except in Poland, where only five cores per species were taken in all plots due to restrictions imposed by park authorities), for a total of 3138 cored trees. Short of coring all trees within a stand, this approach has been shown to provide the most reliable estimates of plot-level productivity when using tree ring data, as it ensured that the size distribution of each plot is adequately represented by the subsample. Wood cores were stored in polycarbonate sheeting and allowed to air dry before being mounted on wooden boards and sanded with progressively finer grit sizes. A high-resolution flatbed scanner (2400 dpi optical resolution) was then used to image the cores.

 From tree rings to aboveground wood production

We followed a four-step approach (i–iv) to estimate temporal trends in aboveground wood production (AWP, in MgC ha^-1 yr^-1) from tree ring data (Jucker et al., 2014a).

i.        Measuring growth increments from wood cores

We measured yearly radial growth increments (mm yr^-1) for each cored tree from the scanned images. To minimize measurement errors associated with incorrectly placed ring boundaries, we crossdated each sample against a species-level reference curve obtained by averaging all ring-width chronologies belonging to a given species from a given site. In this process, 188 cores which showed poor agreement with reference curves were excluded from further analysis, giving a final total of 2950 tree ring chronologies. Both radial growth measurements and crossdating were performed using CDendro (Cybis Elektronik & Data, Saltsjöbaden, Sweden). Here we report data from the five-year period between 2007 and 2011.

ii. Converting diameter increments into biomass growth

We combined radial increments and allometric functions to express the growth rate of individual trees in units of biomass. We calculated the average yearly biomass growth between 2007–2011 (G, kgC yr^-1) of cored trees as G = (AGBt₂ – AGBt₁)/ Δt, where AGBt₂ is the tree’s biomass, estimated with equations presented in Jucker et al. (2014b) in the most recent time period (i.e., end of 2011) and AGBt₁ is its biomass at the previous time step (i.e., end of 2006), Δt and is the elapsed time (i.e., five years). AGBt₁ was estimated by replacing current diameter and height measurements used to fit biomass equations with past values. Past diameters were reconstructed directly from wood core samples by progressively subtracting each year’s diameter increment. Height growth was estimated by using height-diameter functions to predict the past height of a tree based on its past diameter.

iii. Modelling individual tree biomass growth

We modelled the biomass growth of each species as a function of tree size, competition for light, species richness, and a random plot effect:

log(G_i) = α_j[i] + β₁ x log(D_i) + β₂ x CI_i + β₃ x SR_j + ε_i 

where G_i, D_i and CI_i are, respectively, the biomass growth, stem diameter and crown illumination index of tree i growing in plot j; SR_j is the species richness of plot j; α_j is a species’ intrinsic growth rate for a tree growing in plot j; β_1-3 are, respectively, a species’ growth response to size, light availability and species richness; and ε_i is the residual error. The structure of the growth model is adapted from Jucker et al. (2014b) and was fitted using the lmer function in R. Model robustness was assessed both visually, by comparing plots of predicted vs observed growth, and through a combination of model selection and goodness-of-fit tests (AIC model comparison and R²). Across all species, individual growth models explained much of the variation in growth among trees (Jucker et al., 2014a).

iv. Scaling up to plot-level AWP

To quantify AWP at the plot level, we used the fitted growth models to estimate the biomass growth of all trees that had not been cored. For each plot, we then summed the biomass growth of all standing trees to obtain an estimate of AWP. Growth estimates were generated using the predict.lmer function in R.

f.        Tree biomass: Aboveground biomass of all trees [Mg C ha^-1] (tree_biomass)

In each plot, the aboveground biomass (AGB, Mg C ha^-1) of all the individual trees was estimated using tree diameter and height measurements in combination with species-specific biomass functions (see above). Biomass estimates of the individual trees were then summed to quantify the plot-level tree biomass.

g.      Understorey biomass: Dry weight of all understorey vegetation in a quadrant [g] (total_understorey_weight)

In three subplots in each plot (upper right, central, lower left), a quadrant of 5 m x 5 m was marked for identification and estimation of cover of understorey vascular plant species (both woody and non-woody). Within each quadrant, all understorey vegetation was identified to species and afterwards clipped in a zone of 0.5 m x 0.5 m, where vegetation was relatively abundant and the composition was representative of the whole quadrant. The biomass samples (g) were dried for 48 h at 70°C before weighing.

4 4. Regeneration

a.       Sapling growth: Growth of saplings up to 1.60 m tall [cm] (sapling_growth)

Sapling growth measurements (cm) were taken in 2012 on a total of 30 saplings per species wherever possible. Saplings (up to 1.60 m tall) of all tree species in the regional species pool were selected in a subplot of 4x4 m located in the central part of the main plot. Sapling growth was quantified as the distance between the bud scars (internodes) along the main stem of the last five years (i.e. from 2007 to 2011), without considering the shoot of the current growing season. For details on the methodology, see Bastias et al. (2019).

b.      Tree seedling regeneration: Number of saplings up to 1.60 m tall (regeneration_seedlings)

Field sampling for tree seedling regeneration was carried out at the same time and in the same subplot as the tree juvenile regeneration (see below). Tree seedling regeneration was quantified as the number of tree seedlings (i.e. less than a year old) of all tree species in the regional species pool. For details on the methodology, see Bastias et al. (2019).

c.       Tree juvenile regeneration: Number of tree seedlings less than a year old (regeneration_juveniles)

Field sampling to quantify regeneration was carried out in 2012, from April to late August, in a subplot of 4x4m (16m2) delimited in the central part of the main plot. Tree juvenile regeneration was quantified as the number of sapling trees of tree species in the regional species pool over one year old and up to 1.60 m tall. For details on the methodology, see Bastias et al. (2019).

5 5. Resistance to disturbance

a.       Resistance to drought: Difference in carbon isotope composition in wood cores between dry and wet years [‰] (wue)

For each plot, we randomly selected six trees among the 12 largest ones (i.e. largest diameter at breast height, DBH). For the mixed plots, three trees per species were randomly selected among the six largest trees of each species. This selection was conducted as to only select dominant and/or co-dominant trees in order to avoid confounding factors related to light interception. From each selected tree, a wood core was extracted at breast height during the summers of 2012 and 2013. For each site, we selected two years with contrasting climatic conditions during the growing season (dry vs. wet year) during the 1997-2010 period, see Grossiord et al. (2014) for full details. Latewood samples from these two years were carefully extracted from each wood core. The late wood sections from a given year and a given species in a given plot were bulked and analyzed for their carbon isotope composition (δ¹³C, ‰) with a mass spectrometer. By only selecting latewood sections, we characterized the functioning of the trees during the second part of the growing season and avoided potential effects related to the remobilization of stored carbohydrates from the previous growing season or to a favorable spring climate. Plot-level δ¹³C was calculated as the basal-area weighted average value of species-level δ¹³C measurements. Soil drought exposure in each forest stand was calculated as the stand-level increase in carbon isotope composition of late wood from the wet to the dry year (Δδ¹³CS). For more details on resistance to drought measurements, we refer to Grossiord et al. 2014 (2014).

b.      Resistance to insect damage: Foliage not damaged by insects [%] (resistance_insects)

As for fungal pathogens sampling (see below), we estimated insect herbivory on six trees per species in monocultures and three trees per focal species in mixed forests. The herbivory assessment was done once, from late spring to early summer (see periods on fungal pathogens protocol below). The insect herbivory protocol was derived from the ICP Forests manual. It was adapted to better account for total insect damage by observing the whole tree crown, instead of the “assessable crown” only. Damage on the crown exposed to sunlight and in the shade was recorded separately, as foliar loss may be also due to competition for light or natural pruning in the shaded part, particularly in heliophilous tree species. We considered damage as leaf area loss or shoot mortality i.e. defoliation. To estimate herbivore impact, we compared the sampled trees to a “reference tree”, i.e. a healthy tree with intact foliage in its vicinity. Using binoculars, we estimated the proportion of defoliation in the living crown (i.e. the crown excluding the dead branches) in both parts of the crown (sunlight-exposed PDL and in the shade PDS) and put the estimates in one out of seven percentage classes: 0%, 0.5-1%, 1-12.5%, 12.5-25%, 25-50%, 50-75% and > 75% damage. The assessment was done from at least two sides of the crown to account for all damage. When a different score was attributed from different sides to a focal tree, the mean of damage class median was used. The total percent of defoliation was calculated as the natural logarithm of the sum of PDL and PDS. For further details on the methodology, see Guyot et al. (2016).

c.       Resistance to mammal browsing: Twigs not damaged by browsers [%] (lack_browsing)

All plots were sampled using four 5m x 5m subplots located in the same areas of each plot.  Within each of the four 5x5m subplots each woody species individual was visually inspected for browsing damage (bitten twigs).  When browsing was found, the species was recorded, an estimation of the percentage of twigs browsed (between a height of 0.5–2 m) was made (biomass removed), and the stem diameter (at the base) and upper and lower limits of browsing were recorded. With these data, a plot-level average of the percentage of twigs browsed was calculated, and resistance to mammal browsing was defined as 100 - % of twigs browsed.

d.      Resistance to pathogen damage: Foliage not damaged by pathogens [%] (no_pathogen_damage)

Fungal pathogen damage was assessed over a two-week period at each plot during the growing period, over two years. Foliage was collected from Italy (June-July 2012), Germany (July 2012), Finland (August 2012), Spain (June 2013), Romania (July 2013), and Poland (July-August 2013).

In each plot, the six trees with the largest DBH per species were selected for trees within monoculture plots, and three trees with the largest DBH per species for trees within mixture plots. Foliage (leaves and shoots) samples were collected from branches from two levels of the tree canopy (25-60 leaves and 10 current-year shoots per branch) for each focal tree species. The number of leaves sampled from each focal tree and the number of plots within each tree species richness levels are enumerated in Table S8 in van der Plas et al. (2016a).  Visual assessments for fungal pathogen damages were conducted on fresh leaves within one day of sampling. Leaves and shoots were assessed for four classes of fungal damages: oak powdery mildew and leaf spots for the broadleaved tree species, and rust and needle cast for the conifer species. The number of leaves or shoots with the respective damages per tree was recorded, as well as the number of leaves and shoots free from fungal pathogen damage, i.e. healthy foliage. To obtain a value of healthy foliage at the plot level, the sum of all healthy foliage for all trees within the plot was calculated and this was divided by the total number of foliage replicates to acquire a plot-level proportion of healthy foliage. All assessments were conducted by one person to avoid observer bias. For details on the sampling effort, we refer to Nguyen et al.(2016).

e.       Tree growth recovery: Ratio between post-drought growth and growth during the respective drought period (tree_growth_recovery)

Following Lloret et al. (Lloret et al., 2011), growth recovery was defined as the ability to recover growth rates (see tree productivity section) after a decline in growth experienced during the low-growth period (see growth resistance section). It corresponds to the ratio between the average post-drought growth in the five years after a drought year and the growth during the respective low-growth year. Values less than 1 indicate a decline in growth after the drought year, while values greater than one indicate (partial) recovery.

f.        Tree growth resilience: Ratio between growth after and before the drought period (tree_growth_resilience)

Following Lloret et al. (Lloret et al., 2011), growth resilience was defined as the capacity of the forest stand to return to pre-drought growth (see tree productivity section) levels after a drought and is estimated as the ratio between average growth in the five years after and before the low-growth period (see growth resistance section).

g.      Tree growth resistance: Ratio of tree growth during a drought period and growth during the previous five-year high-growth period (tree_growth_resistance)

Following Lloret et al. (Lloret et al., 2011), growth resistance was quantified by comparing tree growth in a low-growth year to the mean growth in the preceding five years. The year with the lowest growth across the regions was 2003, with the exception of Germany and Spain, where the lowest growth was in 1998 and 2005, respectively. 1998 and 2003 were known as drought years across Europe, with the exception of Spain where 2005 was even drier. Growth resistance was defined as the reversal of the reduction in growth (methodology described in the tree productivity section) during the drought: as the ratio of growth during the low-growth year and the growth during the previous five-year high-growth period. The larger the value, the greater the resistance of tree growth to drought.

h.      Tree growth stability: Mean annual tree growth divided by standard deviation in annual tree growth between 1992 and 2011 (tree_growth_stability)

Using the annual aboveground wood production (AWP, see tree productivity section above), for each plot the growth stability was calculated as:

mean(AWP) / sd(AWP)

where mean(AWP) is the temporal mean AWP and sd(AWP) is the standard deviation in AWP between 1992 and 2011. See Jucker et al. (2014) for more details.

6 6. Timber quality

a.       Stem quality: Mean plot silvicultural quality assessment based on stem characteristics (timber_quality)

For timber quality measurements, in each plot, dendrometric data and externally visible stem characteristics were recorded. The silvicultural quality assessment was based on stem characteristics that can be measured and evaluated non-destructively and rapidly along with a measurement of potentially influencing factors at the tree- and stand-level. For each tree within a plot, total height, height of the crown base, height of the lowest dead branch (> 1 cm diameter), and type of fork (or steeply angled branch) were measured. In addition, the presence of the following stem quality parameters was recorded: curving, stem lean, epicormic branching, coppicing, pathogenic, and other defects. Due to the multiple factors constituting stem quality and wood quality, a four-class stem quality grading scheme was used to aggregate all stem quality parameters collected for each tree into an appropriate stem quality score, allowing for the analysis of a single response variable across all regions, species diversity levels and compositions; see Table 1 in Benneter et al (2018), with quality class D=1 being the lowest, and class A=4 being the highest quality class. The assessment of stem quality parameters was limited to the butt log of the tree, which represented the lowest 5 meters of the stem for broadleaved tree species and a maximum of 10 meters from the stem base for conifers. Multiples of the 5-meter section were only considered if the second log showed at least quality class C=2, but only if the green crown base was above the section considered. It has been estimated that for most commercial species in Europe, these butt logs comprise up to 50-70 % (softwood) and 80-95 % (hardwood) of the total commercial tree value. Plot-level timber quality was then calculated as the average timber quality of all the individual trees. For further details, see Benneter et al. (2018).

We further quantified the diversity of several forest-associated taxonomic groups (bats, birds, spiders, insects, earthworms, fungal pathogens, soil microbes, understorey plants, and their multi-diversity and multi-abundance/-activity indices) and many aspects of habitat quality (tree functional and structural diversity), in each plot; the respective data can be found here:

Allan, E. et al. (2019). Tree diversity is key for promoting the diversity and abundance of forest‐associated taxa in Europe [Dataset]. Dryad. https://doi.org/10.5061/dryad.sf7m0cg22. See also: Ampoorter, E. et al. (2020) Tree diversity is key for promoting the diversity and abundance of forest-associated taxa in Europe. Oikos 129, 133-146.

In addition, detailed measurements on soil fauna, properties, and functions have been quantified within the SoilForEUROPE project, see https://websie.cefe.cnrs.fr/soilforeurope/.

References

Baeten, L. et al., 2019. Identifying the tree species compositions that maximize ecosystem functioning in European forests. Journal of Applied Ecology, 56(3): 733-744.

Baeten, L. et al., 2013. A novel comparative research platform designed to determine the functional significance of tree species diversity in European forests. Perspect Plant Ecol, 15: 281-291.

Bastias, C.C., Morán-López, T., Valladares, F. and Benavides, R., 2019. Seed size underlies the uncoupling in species composition between canopy and recruitment layers in European forests. Forest Ecol Manag, 449: 117471.

Benneter, A., Forrester, D.I., Bouriaud, O., Dormann, C.F. and Bauhus, J., 2018. Tree species diversity does not compromise stem quality in major European forest types. Forest Ecol Manag, 422: 323-337.

Dawud, S.M. et al., 2017. Tree species functional group is a more important driver of soil properties than tree species diversity across major European forest types. Functional Ecology, 31: 1153-1162.

De Wandeler, H. et al., 2018. Tree identity rather than tree diversity drives earthworm communities in European forests. Pedobiologia, 67: 16-25.

De Wandeler, H. et al., 2016. Drivers of earthworm incidence and abundance across European forests. Soil Biology and Biochemistry, 99: 167-178.

Finér, L. et al., 2017. Conifer proportion explains fine root biomass more than tree species diversity and site factors in major European forest types. Forest Ecol Manag, 406(Supplement C): 330-350.

Grossiord, C. et al., 2014. Tree diversity does not always improve resistance of forest ecosystems to drought. Proceedings of the National Academy of Sciences, 111(41): 14812-14815.

Guyot, V., Castagneyrol, B., Vialatte, A., Deconchat, M. and Jactel, H., 2016. Tree diversity reduces pest damage in mature forests across Europe. Biology Letters, 12(4): 20151037.

Joergensen, R.G. and Mueller, T., 1996. The fumigation-extraction method to estimate soil microbial biomass: Calibration of the kEN value. Soil Biology and Biochemistry, 28(1): 33-37.

Joly, F.-X. et al., 2017. Tree species diversity affects decomposition through modified micro-environmental conditions across European forests. New Phytologist, 214: 1281-1293.

Joly, F.-X., Scherer-Lorenzen, M. and Hättenschwiler, S., 2023. Resolving the intricate role of climate in litter decomposition. Nature Ecology & Evolution, 7(2): 214-223.

Jucker, T., Bouriaud, O., Avacaritei, D. and Coomes, D.A., 2014a. Stabilizing effects of diversity on aboveground wood production in forest ecosystems: linking patterns and processes. Ecol Lett, 17(12): 1560–1569.

Jucker, T. et al., 2014b. Competition for light and water play contrasting roles in driving diversity–productivity relationships in Iberian forests. J Ecol, 102: 1202–1213.

Kambach, S. et al., 2019. How do trees respond to species mixing in experimental compared to observational studies? Ecology and Evolution, 9(19): 11254-11265.

Lloret, F., Keeling, E.G. and Sala, A., 2011. Components of tree resilience: Effects of successive low-growth episodes in old ponderosa pine forests. Oikos 120: 1909–1920.

Nguyen, D. et al., 2016. Fungal disease incidence along tree diversity gradients depends on latitude in European forests. Ecology and Evolution, 6(8): 2426-2438.

Pollastrini, M. et al., 2016a. Physiological significance of forest tree defoliation: results from a survey in a mixed forest in Tuscany (central Italy). Forest Ecology and Management 361: 170-178.

Pollastrini, M. et al., 2016b. Taxonomic and ecological relevance of the chlorophyll a fluorescence signature of tree species in mixed European forests. New Phytologist, 212(1): 51-65.

Ratcliffe, S. et al., 2017. Biodiversity and ecosystem functioning relations in European forests depend on environmental context. Ecol Lett, 20: 1414-1426.

Skjemstad, J.O. and Baldock, J.A., 2007. Total and organic carbon. Soil sampling and methods of analysis. CRC Press, Boca Raton, FL.

Tamminen, P. and Starr, M., 1994. Bulk density of forested mineral soils. Silva Fennica 28 (1): article id 5528.

van der Plas, F. et al., 2016a. Jack-of-all-trades effects drive biodiversity-ecosystem multifunctionality relationships in European forests. Nature Communications, 7: 11109.

van der Plas, F. et al., 2016b. Biotic homogenization can decrease landscape-scale forest multifunctionality. Proceedings of the National Academy of Sciences, 113(13): 3557-3562.

van der Plas, F. et al., 2018. Continental mapping of forest ecosystem functions reveals a high but unrealised potential for forest multifunctionality. Ecol Lett, 21(1): 31-42.

Van Heerwaarden, L.M., Toet, S. and Aerts, R., 2003. Current measures of nutrient resorption efficiency lead to a substantial underestimation of real resorption efficiency: facts and solutions. Oikos 101: 664-669.

Vergutz, L., Manzoni, S., Porporato, A., Novais, R.F. and Jackson, R.B., 2012. Global resorption efficiencies and concentrations of carbon and nutrients in leaves of terrestrial plants. Ecological Monographs 82: 205-220.

Vesterdal, L., Schmidt, I.K., Callesen, I., Nilsson, L.O. and Gundersen, P., 2008. Carbon and nitrogen in forest floor and mineral soil under six common European tree species. Forest Ecol Manag, 255(1): 35-48.