1. Plants vary widely in how common or rare they are, but whether commonness of species is associated with functional traits is still debated. This might partly be because commonness can be measured at different spatial scales, and because most studies focus solely on aboveground functional traits.

2. We measured five root traits and seed mass on 241 Central European grassland species, and extracted their specific leaf area, height, mycorrhizal status and bud-bank size from databases. Then we tested if trait values are associated with commonness at seven spatial scales, ranging from abundance in 16-m² grassland plots, via regional and European-wide occurrence frequencies, to worldwide naturalization success.

3. At every spatial scale, commonness was associated with at least three traits. The traits explained the greatest proportions of variance for abundance in grassland plots (42%) and naturalization success (41%), and the least for occurrence frequencies in Europe and the Mediterranean (2%). Low root tissue density characterized common species at every scale, whereas other traits showed directional changes depending on the scale. We also found that many of the effects had significant non-linear effects, in most cases with the highest commonness-metric value at intermediate trait values. Across scales, belowground traits explained overall more variance in species commonness (19.4%) than aboveground traits (12.6%).

4. The changes we found in the relationships between traits and commonness, when going from one spatial scale to another, could at least partly explain the maintenance of trait variation in nature. Most importantly, our study shows that within grasslands, belowground traits are at least as important as aboveground traits for species commonness. Therefore, belowground traits should be more frequently considered in studies on plant functional ecology.

Species traits

Species selection, seed material and precultivation

The species used are herbaceous angiosperms occurring in the grassland plots of the German “Biodiversity Exploratories” (Fischer et al. 2010; Socher et al. 2012). In each of three regions of Germany, the Schwäbische-Alb (south-western Germany), Hainich-Dün (central Germany), and Schorfheide-Chorin (north-eastern Germany), 50 plots (4 m × 4 m) were selected in grassland habitats covering a wide range of land-use intensities. From 2008 to 2016, the vegetation composition of each of the 150 plots was assessed annually in late spring by estimating the cover of each species. We standardized the species names according to the accepted names in  http://www.theplantlist.org, accessed on 15^th June 2019, using the Taxonstand package (Cayuela et al. 2012) to allow us to align the species names between different distribution and trait datasets (see below). In total, 363 vascular plant species have been identified in the plots of the “Biodiversity Exploratories”. For 311 of those species, we were able to obtain seeds from commercial seed suppliers or botanical gardens for our experiments (see Appendix S1 in Supporting Information). For Alchemilla vulgaris agg., which also includes taxa that are difficult to distinguish from A. vulgaris, we used seeds from A. vulgaris. For Leucanthemum vulgare agg., which includes both L. vulgare and L. ircutianum, we used seeds from both species and the trait values were averaged.

In two experiments, we measured functional traits on those species. Before the first experiment, we individually weighed 10 seeds, randomly chosen from the supplier’s bag, for each of the 311 species. Then we did an indoor pot experiment to determine root morphology of the species, and an outdoor pot experiment to determine rooting depth. For both experiments, seeds were first sown in plastic pots (7 cm × 7 cm × 6.5 cm) filled with peat soil. The pots were then placed in a growth chamber for two to three weeks (night/day 9/15 h; 18/21 ± 1.5°C; relative humidity 90 ± 5%) before transplanting the seedlings into the pots used for the experiments (for cultivation times, see Appendix S1). In addition to the traits measured in the experiments, we obtained data on aboveground traits (specific leaf area, height), bud-bank size and mycorrhizal status from databases (see the section “Traits from databases and data imputation” below).

Experiment on root-system morphology

From May 1 to October 6, 2017, we performed a glasshouse experiment to measure root-system morphological traits of the study species. As root morphology might depend on nutrient availability, we grew half of the plants per species at intermediate nutrient levels and the other half at high nutrient levels, after which we averaged the trait values per species. Because of the large number of species and the time-consuming measurements, we grew the plants in four temporally shifted (4-6 weeks) batches. We aimed to have each species represented in each batch, and to have a total of seven replicates per species and nutrient level across all batches (Appendix S1). The seedlings of the species that had germinated (N=233) were transplanted individually into plastic pots (1.3 L) filled with a mixture of sand and vermiculite (1:1 volume ratio). The pots were then randomly allocated to positions in two glasshouse compartments, and allowed to grow for four weeks (night/day 10/14 h; 22/28 ± 1.5°C; relative humidity 80 ± 15%). Plants were fertilized three times a week with either an intermediate nutrient solution (40 ml with 1500 µM KNO₃) or a high nutrient solution (40 ml with 12000 µM KNO₃). The fertilizer was a modified version of the Hoagland recipe (see Appendix S2).

We grew the plants for four weeks only to avoid roots becoming pot-bound, to be able to analyse the entire root systems, and to ensure that all the belowground biomass was formed by roots, excluding rhizomes. After washing off the substrate, the root system was cut below the collar and stored for <1 week in a plastic tube filled with distilled water at 4°C. Then, root systems were spread individually in a thin layer of water in transparent trays (11 cm × 11 cm) and scanned at 800 dpi with a flatbed scanner modified for root scanning (Epson Expression 10000 XL and 11000 XL). The images were analysed using the software WinRHIZO^TM 2017a (Regent Instruments, Quebec, Canada) to obtain the total root length and root volume. Root systems were then oven-dried for >48 hours at 65°C and weighed. We calculated specific root length by dividing the total root length by the belowground dry biomass, and root tissue density by dividing the belowground dry biomass by the sum of the root volumes according to Rose (2017). The diameter of fine roots (i.e. distal roots), thought to be the most important roots for nutrient uptake (Freschet and Roumet 2017), was determined by randomly sampling a distal root branch (or a portion of it) for each root system and calculating the mean of the external-internal links diameter obtained with the “Link analysis” function in WinRHIZO^TM. This subsampling avoided the inclusion of thicker transport roots and allowed us to obtain values that were representative for first order roots. We also dried and weighed the aboveground biomass of each plant, and calculated the root weight ratio (i.e. root biomass divided by total biomass).

Experiment on rooting depth

From the 15^th of May to the 10^th of October 2018, we performed an outdoor pot experiment to measure the maximum rooting depth of the species. Up to five seedlings of the species that had germinated (n=196; Appendix S1) were transplanted individually into 120 cm tall plastic tree shelter tubes (Tubex ® Standard Plus, http://www.tubex.com/products/tree-shelters/tubex-standard-treeshelters/specification.php), which are normally used in forestry to protect young trees against animals and the elements. We closed the bottoms of these tubes with thick pieces of cotton tissue to be able to use them as pots. The tubes were filled with a mixture of sand and vermiculite (1:1 volume ratio) up to a height of 115 cm. This substrate can be easily penetrated by the roots, and therefore allows each plant to reach its maximum rooting depth quickly. The tree shelter tubes were delivered in packages of five tubes stacked into each other, and they therefore came in five diameter classes (8.0, 8.4, 10.0, 10.8 and 12.0 cm). To avoid that tube diameter would be confounded with species identity, each of the five seedlings per species was planted in a different tube-diameter class. We placed the tubes upright in a randomized design in the Botanical Garden of the University of Konstanz (47°41'24.0"N 9°10'48.0"E; see Appendix S11 for pictures).

We planted 734 plants but, due to early mortality, we had to replace 126 of them within the next three weeks. The growth period therefore ranged from 16 to 19 weeks. The experiment took place during the summer of 2018 (mean temperature: 19.5°C, min/max 2.5/37.4°C; relative humidity: mean 74%, min/max 22.7/100%). All plants were fertilized once a week with 60 ml of a standard nutrient solution (1‰ Universol^® Blue, Nordhorn, Germany), and watered regularly from above. We harvested the plants in October 2018. Each tube was sliced open, and we measured the distance from the top of the substrate to the deepest root.

Traits from databases and data imputation

Data on the aboveground traits specific leaf area (230 species) and height (228 species) were obtained from the LEDA database (Kleyer et al. 2008). Data on bud-bank size (230 species) including stem and root-derived buds occurring belowground or at the soil surface was obtained from Klimešová et al. (2017). In addition, mycorrhizal status was extracted from the FungalRoot database (Soudzilovskaia et al. 2020). We assigned the corresponding genus-level mycorrhizal status for each of our species (241 species) included in the analysis. Though most of our species are considered either obligatorily arbuscular-mycorrhizal (167 species), facultatively arbuscular-mycorrhizal (57 species) or non-mycorrhizal (16 species), Helianthemum nummularium is considered ectomycorrhizal. Therefore, it was grouped with the obligatorily arbuscular-mycorrhizal species to form the obligate-mycorrhizal category (168 species).

Although for each of the traits we had data for 196 (rooting depth) to 311 (seed weight) species, the number of species with complete data for all traits was 170. Therefore, we did phylogenetically informed imputation of missing data for the 241 species that germinated and survived until trait measurement in at least one of our two experiments. Data imputation is a powerful but still underutilized tool that increases sample size —and thus statistical power— and reduces potential biases that might occur if the species with missing data are a non-random subset (Nakagawa 2015). Imputation can perform well with up to 50% of missing data (Graham 2009). In our case, 4.6% of the trait values were missing and had to be imputed (see Appendix S4 for details on the imputation procedure). We also ran all analyses for the subset of 170 species with complete data (i.e. without imputed data), and the results were largely similar to the analyses of the 241 species with partly imputed data (Appendix S9). Because the analyses with the imputed data allowed us to include more species (i.e. increase statistical power and generality), we present only those results in the main text. The phylogenetic tree of the species used, their standardized trait values and phylogenetic signal can be found in Appendix S5, S6 and S7.

Species abundance and occurrence frequency

To quantify each species’ commonness from local scale abundance to global naturalization success, we used four different data sources.

The Biodiversity Exploratories

To obtain information on local abundance and occurrence frequency of our study species in German grasslands, we used data from the Biodiversity Exploratories grassland-composition surveys. In each of the three regions, c. 500 so-called grid plots (GPs) and a subset of those, the 50 so-called experimental plots (EPs), have been monitored for biodiversity measures. The plots are 50 m × 50 m, and in each of those there is a subplot of 4 m × 4 m, in which the relative abundance of each plant species has been determined. In the 1494 GPs, vegetation was sampled once from 25 May to 15 August 2007. In May 2009, 138 plots were re-assessed and earlier relevés were discarded, because they were considered unreliable as the vegetation had been recorded too late in the season of 2007 (Socher et al. 2013). Of our 241 study species, 213 were present in that census of the GPs (Appendix S1), and, when present in a plot, they covered on average 2.8% of the plot (min: 0.27%; median: 1.45%; max. 17.16%). For the 150 EPs, the vegetation data were collected annually between mid-May and mid-June from 2008 to 2016, and we averaged the data across years. Of our 241 study species, 239 were present in the EPs vegetation survey, and, when present in a plot, they covered on average 1.05% of the plot (min.: 0.01%; median: 0.34%; max.: 13.05%). Two study species, Spergula arvensis and Taraxacum campylodes, had been excluded because their names were included in an earlier version of the vegetation survey due to misidentification. While there are 10 times more GPs than EPs, the latter include data over a longer period. For both the GPs and EPs, we used two distribution metrics for each species: the occurrence frequency defined as the number of plots in which a species is present divided by the total number of plots, and the local abundance defined as the mean cover of a species across all the plots where it is present. Because it is based on the presence-absence only, the occurrence frequency estimates how frequent a species is within grasslands in Germany. The average abundance, on the other hand, which is calculated using abundance data for only those plots where the species occurs, estimates how dense the populations of the species are on average.

FloraWeb

For information on the occurrence frequency in all of Germany, irrespective of habitat type, we obtained data from the German plant distribution atlas of NetPhyD and BfN through the FloraWeb data portal (Bundesamt für Naturschutz 2013). For each species, we extracted the number of grid cells in which the species has been reported. Each grid cell is about 130 km², and there are 2995 grid cells in total. Of our 241 study species, 235 had grid-cell data available (Appendix S1).

Euro+Med PlantBase

To obtain information on the extent of the native distribution in all of Europe and its adjacent Mediterranean regions, we used Euro+Med PlantBase (PESI 2015). This on-line database provides information on the presence of vascular plant taxa in 117 regions (mostly countries) covering all of Europe and the Mediterranean regions of North Africa and the Near East. Of our 241 study species, 237 species were found in Euro+Med PlantBase, and for those we extracted the total number of regions with native occurrences (Appendix S1). The four remaining species, Cerastium nutans, Erigeron canadensis, Matricaria discoidea and Medicago × varia, are not native to the region.

GloNAF

As 237 of our 241 study species are native to Europe, we also assessed the extent of their global occurrence as naturalized alien species, using the Global Naturalized Alien Flora (GloNAF) database, version 1.2 (van Kleunen et al. 2019). GloNAF is a compendium of lists of naturalized alien plant species for 1029 regions covering >80% of the terrestrial ice-free surface. Of our 241 study species, 221 species had at least one record in GloNAF. For those species, we extracted the number of regions in which they are naturalized, and for the 20 species without GloNAF records, we set the number of GloNAF regions equal to zero.

Statistical analyses

All statistical analyses were performed in R version 3.6.1 (R Core Team 2019). To test whether more abundant and more widespread species have particular trait values, we used generalized linear models in which the response variables were the different measures of species commonness and the predictors were a selection of trait mean values. For number of occurrences in GloNAF regions, and in grassland GPs and EPs, we used negative binomial error distributions (with a log-link function). As the number of occurrences in FloraWeb grid cells and Euro+Med regions did not follow negative binomial or Poisson error distributions, we instead analysed the proportion of FloraWeb grid cells and Euro+Med regions in which a species had been recorded, with a binomial error distribution. To account for overdispersion, we used the ‘quasibinomial’ setting. For analyses of the mean local abundance (i.e. the cover proportion) of the species in the GPs and EPs, we used a gamma conditional distribution (with log-link function).

For each commonness measure, we used a multivariate model with ten traits as predictors. We a priori chose traits that represent different aspects of plant functioning and that had relatively low pairwise correlations (all r ≤ |0.49|, Appendix S3) to minimize multicolinearity (the maximum generalized variance-inflation factor of a model was 3.32). We used the following traits: individual seed weight (measured on seeds ordered for the experiments), specific root length, root tissue density and fine roots diameter (measured in the root-system morphology experiment), maximum rooting depth (measured in the rooting-depth experiment), and bud-bank size, height, specific leaf area and mycorrhizal status (from trait databases). Seed weight was log transformed. To facilitate interpretation of and comparison between model coefficients, each trait was scaled to a mean of zero and a standard deviation of one (Schielzeth 2010). To test for potential non-linear effects of traits, orthogonal polynomial terms of second degree (i.e. quadratic terms) were also included for each trait, using the poly function. To estimate the proportion of variance explained by the models, we calculated delta R² values, applicable to all distributions and link functions, according to Nakagawa et al. (2017) using the package MuMIn (Barton and Barton 2015). To assess whether belowground traits explained more variance in commonness measures than aboveground traits, we also extracted delta R² values for models using only the three aboveground predictors and for models using only the three belowground predictors with the highest standardized coefficients. To account for phylogenetic non-independence of the study species, the models were also run using phylogenetic relatedness of species as a variance-covariance matrix (for details, see Appendix S4). Although the significances of the trait effects differed in some instances between the non-phylogenetic models and the phylogenetic ones, the directions of the effects were largely the same in both types of models (compare Fig. 1 and Appendix S8). Therefore, we only present the results of the non-phylogenetic analyses in the main text.

 

References

Barton, K. and Barton, M. K. 2015. Package ‘MuMIn’, https://CRAN.R-project.org/package=MuMIn.

Bundesamt für Naturschutz 2013. FloraWeb, http://www.floraweb.de accessed 22 June 2017.

Cayuela, L. et al. 2012. taxonstand: An r package for species names standardisation in vegetation databases. - Methods Ecol. Evol. 3: 1078–1083.

Fischer, M. et al. 2010. Implementing large-scale and long-term functional biodiversity research: The Biodiversity Exploratories. - Basic Appl. Ecol. 11: 473–485.

Freschet, G. T. and Roumet, C. 2017. Sampling roots to capture plant and soil functions. - Funct. Ecol. 31: 1506–1518.

Graham, J. W. 2009. Missing data analysis: making it work in the real world. - Annu. Rev. Psych 60: 549–576.

Kleyer, M. et al. 2008. The LEDA Traitbase: a database of life-history traits of the Northwest European flora. - J. Ecol. 96: 1266–1274.

Klimešová, J. et al. 2017. CLO-PLA: a database of clonal and bud-bank traits of the Central European flora. - Ecology 98: 1179.

Nakagawa, S. 2015. Missing data: mechanisms, methods, and messages - In: Fox, G. A. et al. (eds.), Ecological statistics. Contemporary theory and application. Oxford University Press.

Nakagawa, S. et al. 2017. The coefficient of determination R² and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded. - J. R. Soc. Interface 14: 20170213.

PESI 2015. Euro+Med PlantBase, https://www.emplantbase.org/home.html accessed 1 June 2019.

R Core Team 2019. R: A language and environment for statistical computing. - R Foundation for Statistical Computing, https://www.R-project.org.

Rose, L. 2017. Pitfalls in root trait calculations: how ignoring diameter heterogeneity can lead to overestimation of functional traits. - Front. Plant. Sci. 8: 898.

Schielzeth, H. 2010. Simple means to improve the interpretability of regression coefficients. - Methods Ecol. Evol. 1: 103–113.

Socher, S. A. et al. 2012. Direct and productivity-mediated indirect effects of fertilization, mowing and grazing on grassland species richness. - J. Ecol. 100: 1391–1399.

Socher, S. A. et al. 2013. Interacting effects of fertilization, mowing and grazing on plant species diversity of 1500 grasslands in Germany differ between regions. - Basic Appl. Ecol. 14: 126–136.

Soudzilovskaia, N. A. et al. 2020. FungalRoot: global online database of plant mycorrhizal associations. - New Phytol. 227: 955–966.

van Kleunen, M. et al. 2019. The Global Naturalized Alien Flora (GloNAF) database. - Ecology 100: e02542.

Please find complementarity information about the dataset in the Supplementary Information of the corresponding article.

For any questions, please contact the corresponding author.