Plant functional and taxonomic diversity in European grasslands along climatic gradients

J Veg Sci. 2021;32:e13027.   | 1 of 12 https://doi.org/10.1111/jvs.13027 Journal of Vegetation Science wileyonlinelibrary.com/journal/jvs Received:10September2020 | Revised:11April2021 | Accepted:13April2021 DOI: 10.1111/jvs.13027 SPECIAL FEATURE: MACROECOLOGY OF VEGETATION Plant functional and taxonomic diversity in European grasslands along climatic gradients Coline C. F. Boonman1 | Luca Santini2,3 | Bjorn J. M. Robroek4 | Selwyn Hoeks1 | Steven Kelderman1 | Jürgen Dengler5,6,7 | Ariel Bergamini8 | Idoia Biurrun9 | Maria Laura Carranza10 | Bruno E. L. Cerabolini11 | Milan Chytrý12 | Ute Jandt13,14 | Tatiana Lysenko15,16 | Angela Stanisci17 | Irina Tatarenko18 | Solvita Rūsiņa19 | Mark A. J. Huijbregts1 1Department of Environmental Science, Institute for Water and Wetland Research, Radboud University, Nijmegen, The Netherlands 2Department of Biology and Biotechnologies “Charles Darwin”, Sapienza Università di Roma, Rome, Italy 3National Research Council, Institute of Research on Terrestrial Ecosystems (CNRIRET), Monterotondo, Italy 4DepartmentofAquaticEcologyandEnvironmentalBiology,InstituteforWaterandWetlandResearch,RadboudUniversity,Nijmegen,TheNetherlands 5VegetationEcologyGroup,InstituteofNaturalResourceSciences(IUNR),ZurichUniversityofAppliedSciences(ZHAW),Wädenswil,Switzerland 6Plant Ecology, Bayreuth Center of Ecology and Environmental Research (BayCEER), Bayreuth, Germany 7GermanCentreforIntegrativeBiodiversityResearch(iDiv)Halle-Jena-Leipzig,Leipzig,Germany 8WSLSwissFederalResearchInstitute,Birmensdorf,Switzerland 9PlantBiologyandEcology,FacultyofScienceandTechnology,UniversityoftheBasqueCountryUPV/EHU,Bilbao,Spain 10EnvixLab,UniversityofMolise,Pesche(IS),Italy 11DepartmentofBiotechnologiesandLifeSciences(DBSV),UniversityofInsubria,Varese,Italy 12DepartmentofBotanyandZoology,FacultyofScience,MasarykUniversity,Brno,CzechRepublic 13InstituteofBiology,MartinLutherUniversityHalle-Wittenberg,Halle(Saale),Germany 14GermanCentreforIntegrativeBiodiversityResearch(iDiv)Halle-JenaLeipzig,Leipzig,Germany 15GeneralVegetationLaboratory,KomarovBotanicalInstituteRAS,Saint-Petersburg,Russia 16PhytodiversityProblemsLaboratory,SamaraFederalResearchCenterRAS,InstituteoftheEcologyoftheVolgaBasinRAS,Togliatti,Russia 17EnvixLab,UniversityofMolise,Termoli,Italy 18SchoolofEnvironment,EarthandEcosystemSciences,FacultyofScience,Technology,EngineeringandMathematics,TheOpenUniversity,MiltonKeynes, UK 19FacultyofGeographyandEarthSciences,UniversityofLatvia,Rīga,Latvia ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium, providedtheoriginalworkisproperlycited. ©2021TheAuthors.Journal of Vegetation SciencepublishedbyJohnWiley&SonsLtdonbehalfofInternationalAssociationforVegetationScience ThisarticleisapartoftheSpecialFeatureMacroecologyofVegetation,editedbyMeelisPärtel,FrancescoMariaSabatini,NaiaMorueta-Holme,HolgerKreftandJürgenDengler. Correspondence ColineC.F.Boonman,Departmentof Environmental Science, Institute for Water and Wetland Research, Radboud University, POBox9010,NL-6500GL,Nijmegen,The Netherlands. Email: [email protected] Abstract Aim: European grassland communities are highly diverse, but patterns and drivers of their continentalscale diversity remain elusive. This study analyses taxonomic and functional richness in European grasslands along continentalscale temperature and precipitation gradients. Location: Europe. 2 of 12 | Journal of Vegetation Science BOONMAN et Al. 1 | INTRODUCTION Grasslands are among the most diverse ecosystems in Europe (Wilson et al., 2012; Dengler et al., 2020). They have been extensively studied for a long time and with longterm monitoring schemes (Scholz, 1975; Willems, 1983; Tilman et al., 2006). Various assembly processes have been put forward that may explain the origin and maintenance of European grassland diversity, e.g., competitive hierarchy and niche partitioning (Mori et al., 2018). Yet, over the last 50 years, European grassland diversity has seen a dramatic decrease, often being attributed to increased nutrient availability (Wesche et al., 2012), overgrazing (Dengler et al., 2020) or drought (Carmona et al., 2012; Nogueira et al., 2018). Insight into patterns of plant diversity over environmental gradients is needed to aid deeper understanding of the effects of global change on biodiversity (e.g., Mooney et al., 2009;Cardinaleetal.,2012;Funketal.,2017)andmayalsoimprove our understanding of the mechanisms that underlie community assembly(MacArthur&Levins,1967;McGilletal.,2006). In the context of macroecology, diversity patterns of vegetation are typically discussed via various filtering mechanisms (e.g., Weiheretal.,2011).First,thedispersalfilterdeterminestheability of a species to be present in a specific location, and hence the regionalplantspeciespoolandplanttraitpool(Cadotte&Tucker, 2017). Second, environmental conditions act as an additional filter on plant communities, sorting those that fulfill local (fundamental) nicherequirementsconstitutedbyphysiologicalconstraints(Leibold et al., 2004; Tingley et al., 2014). Under the favorability hypothesis, the more extreme or unfavorable environmental conditions are, the moreselectiveenvironmentalfiltersare(Fischer,1960).Thissuggests that only plants with trait values well adapted to the extreme Funding information CCFB,MAJHandSHweresupportedby the ERC project (CoG SIZE 647224), IB by theBasqueGovernment(IT936-16),MCby theCzechScienceFoundation(19-28491X), andSRbytheUniversityofLatviagrant (AAp2016/B041//Zd2016/AZ03). Co-ordinating Editor:MeelisPärtel Methods: We quantified functional and taxonomic richness of 55,748 vegetation plots.Sixplanttraits,relatedtoresourceacquisitionandconservation,wereanalysed to describe plant community functional composition. Using a nullmodel approach we derived functional richness effect sizes that indicate higher or lower diversity than expected given the taxonomic richness. We assessed the variation in absolute functional and taxonomic richness and in functional richness effect sizes along gradients of minimum temperature, temperature range, annual precipitation, and precipitation seasonality using a multiple general additive modelling approach. Results: Functional and taxonomic richness was high at intermediate minimum temperatures and wide temperature ranges. Functional and taxonomic richness was low in correspondence with low minimum temperatures or narrow temperature ranges.Functionalrichnessincreasedandtaxonomicrichnessdecreasedathigher minimum temperatures and wide annual temperature ranges. Both functional and taxonomic richness decreased with increasing precipitation seasonality and showed a small increase at intermediate annual precipitation. Overall, effect sizes of functional richnessweresmall.However,effectsizesindicatedtraitdivergenceatextremelylow minimum temperatures and at low annual precipitation with extreme precipitation seasonality. Conclusions: FunctionalandtaxonomicrichnessofEuropeangrasslandcommunities vary considerably over temperature and precipitation gradients. Overall, they follow similar patterns over the climate gradients, except at high minimum temperatures and wide temperature ranges, where functional richness increases and taxonomic richness decreases. This contrasting pattern may trigger new ideas for studies that targetspecifichypothesesfocusedoncommunityassemblyprocesses.Andthough effect sizes were small, they indicate that it may be important to consider climate seasonality in plant diversity studies. KEYWORDS environmental filtering, favourability hypothesis, functional richness, grassland diversity, limiting similarity, null model, plant trait diversity, precipitation gradient, seasonality, taxonomic richness, temperature gradient, traitenvironment relationship | 3 of 12 Journal of Vegetation Science BOONMAN et Al. conditions persist, resulting in a reduction in community functional diversity in extremeconditions(de Belloet al., 2009; Mayfield& Levine,2010;Shenetal.,2016).Athirdfiltermaybebioticinteractions, representing a countergradient with competitive interspecific interactions being the main driver of plant community composition at low abiotic stress (the stress gradient hypothesis; Bertness &Callaway,1994).Whilethiscompetitionmayleadtoexclusionof speciesandreducedtraitvariation(Grime,2006;Mayfield&Levine, 2010;Kunstleretal.,2012),facilitativeinteractionshavebeensuggestedtodominatewhenabioticstressishigh(Bertness&Callaway, 1994;Brooker&Callaghan,1998),leadingtohighertraitdivergence andincreasedfunctionaldiversity(Valiente-Banuet&Verdú,2007; McIntire&Fajardo,2014).Finally,theoccurrenceandabundanceof a plant may also be influenced by temporal variation in climate or otherstressors(Díaz&Cabido,1997;González-Morenoetal.,2015; Fischeretal.,2020).Plantscanoccurinaplaceaslongasspecies’ nicherequirementsaremetatsometimeoftheyear,i.e.,exploiting temporarily empty niches (Godoy et al., 2009). This expansion of the realized niche can be especially important for the diversity of plant communitiesintemperateregions(Scheiner&Rey-Benayas,1994; Breitschwerdt et al., 2018). When describing diversity patterns, the focus should not only be on plant species identity (i.e., their taxonomy) but also on plant traits.Astraitsdescribeamoredirectlinkbetweentheperformance of an organism, local environmental conditions, and a plant's functioninginthecommunity(Keddy,1992;Funketal.,2017),theyplay a critical role in determining local plant community diversity (Tilman et al., 1997; Weiher et al., 1998). Especially when nonrandom processes determine community assemblages, species diversity is notanadequatesurrogateforfunctionaldiversity(Díaz&Cabido, 2001).However,bothspeciesandtraitcompositionvaryalongenvironmentalgradients(e.g.,Raymundoetal.,2018),makingtheseparationoftaxonomicandfunctionaldiversitydifficult(e.g.,Hooper et al., 2002; Petchey & Gaston, 2002). To control for potentially coincidental similarity in diversity– environment trends, specifically designed field experiments are needed or one can use null models thatcheckforfunctionaldiversitypatternsdifferentfromrandom samplingofspecies(Gotelli&Graves,1996;Swensonet al.,2012). Takentogether,thenetinfluenceofthevariousfilteringmechanisms on taxonomic and functional plant richness along large environmental stress gradients remains elusive. Part of this ambiguous questionisthevalueofconsideringplanttraitdiversityinaddition totaxonomicdiversity.Here,weanalyzedfunctionalandtaxonomic richness in European grasslands along temperature and precipitationgradients.Wequantifiedtheabsolutefunctionalandtaxonomic richness directly derived from the community data, as well as the effect size of functional richness. The effect size indicates higher or lower functional richness than expected given the observed taxonomicrichness(e.g.,Harveyetal.,1983;Götzenbergeretal.,2016). First,wefocusedonannualminimumtemperature,asextremetemperatures limit metabolic activity, growth and constrain leaf size, and thusarelikelytofiltertraitcomposition(Went,1953).Second,we included annual precipitation, as drought constrains leaf longevity, photosyntheticefficiency,andseedmass(Sandeletal.,2010).Last, we included annual temperature range and precipitation seasonality to be able to assess the influence of seasonality on plant diversity at thecontinentalscale(Scheiner&Rey-Benayas,1994;Godoyetal., 2009). 2 | METHODS 2.1 | Data collection and cleaning 2.1.1 | Vegetationplots Plant community data (i.e., vegetationplot records or relevés) were obtained from the European Vegetation Archive (EVA; Chytrý et al., 2016) on 2 March 2020. Relevés were classified by the expert system EUNISESy to the habitat types of EUNIS (European Nature InformationSystem)HabitatClassification(Chytrýetal.,2020).All grassland habitats (EUNIS group R) and coastal dune habitats (EUNIS habitats N15, N16 and N17) were selected. Plots not classified to anyofthesehabitattypesbutassignedintheEVAdatabasetothe phytosociological class MolinioArrhenatheretea (Mucina et al., 2016) were added to the data set. Data covered the whole of Europe. The islands of Macaronesia were excluded to remove any effects of isolated oceanic islands withsignificantwithin-islandspeciationandendemism(Humphries, 1979). We selected plots recorded between 1979 and 2013 to create an optimal match with climatic data. In addition, we selected plotswithageoreferencinguncertaintysmallerthan1km,which further improved the match with highresolution climatic data. Such adirectlinkbetweenplantcommunityandclimatedatareducesthe effectsofconfoundingfactorslikehabitatheterogeneity(Szilágyi& Meszéna, 2009). We only included flowering plants as the available traits for this group are consistent within the data set. By doing so, we excluded ferns, bryophytes, lichens and fungi. These selection criteria resulted in a georeferenced occurrence data set with presences of all selected plant species in all selected plots.From thisinitialdataset, we excludedplotswithparticular ecologicalconditions(Münkemülleretal.,2020)usingthefollowing criteria: (a) plots influenced by high salinity, identified by the presenceofspeciesknowntofavorextremesaltyconditions(e.g., Suaeda maritima)wereexcluded,astheircompositionislikelydriven by salinity rather than temperature or precipitation; and (b) plots with a cover of tree or tall shrub species, identified as all plant species with an average height of more than 3 m (obtained from the TRYdatabase,seebelow),thatexceeded5%incover.Allsourcesof vegetation-plotdataarelistedinAppendixS1. 2.1.2 | Traitdata WeobtainedtraitdatafromtheTRYdatabaseversion5(Kattgeetal., 2020), containing both public and restricted data sets. We retrieved 4 of 12 | Journal of Vegetation Science BOONMAN et Al. roottraitdatafromtheFine-RootEcologyDatabase(FRED;Iversen etal.,2018).AlldatasourcesarelistedinAppendixS2.Afterstandardization of trait units, we merged trait data from the two databases and calculated species averages. While averaging excludes intraspecific trait variation and trait plasticity, which determines a largepartofcommunitydiversity(Albertetal.,2010;Jungetal., 2010; Ross et al., 2017; Barbour et al., 2019; Niu et al., 2020), this step was necessary as trait data were not available for most species across the different plots. Functionalrichnessdepends ontheselectedsetoftraits. We usedasequenceofselectioncriterialeadingtoourfinaltraitdata set.First,weselectedtraitsbasedontheirimportanceforvegetative growth and reproduction (Wright et al., 2004; Díaz et al., 2016). Then, we selected both aboveand belowground traits with good data coverage. Third, we removed traits of the same plant organ with high covariance (e.g., specific leaf area and leaf dry matter content) inordertoequalizeorganandtraitimportanceinthefunctionaldiversitycalculation,keepingthetraitwiththehighestdatacoverage. Fourth,weselectednumerictraitsonlytominimizetheeffectsofdimension choice to calculate functional diversity (Maire et al., 2015). Our final trait data set contains six plant traits (Table 1). Asfunctionalrichnesscalculations(seebelow)aresensitiveto thecompletenessofthespecieslistperplot(Pakeman,2014)and thusrequireacompletetraitdataset,wefilledthegaps(14.7%of overallmissingdata)inourtraitdatabase(Table1;AppendixS3). Imputing missing data is considered a better alternative to removing incomplete plots, which can potentially result in systematic biases(Nakagawa&Freckleton,2008).Weusedthemicefunctionof the micepackageinR-3.6.1,whichreliesonbetween-traitcorrelationsusingmulti-variateimputationwithchainedequationstofill gaps. This method has been shown to have lower imputation error and bias compared to other multiple imputation approaches, especially when the percentage of missing data is high (Penone et al., 2014).Althoughgap-fillingtechniquesrequireonlyonetraitvalue per species, we only retained species with at least two trait values to achieve a more accurate imputation. This resulted in the removal of complete relevés if the trait values for at least one species in the communitydidnotmeetthisrequirement.Thisresultedinthedata set having 61,714 relevés containing a total of 2,884 species. Prior to the imputation, we logtransformed all trait data since some approaches could be sensitive to data with varying scales in the variables (Penone et al., 2014). Then, we used the predictive mean matching method for the imputation of all traits to preserve nonlinear trait– trait relationships. We ran 24 multiple imputations (van Buuren, 2012) where trait predictions were updated five times in thechainedequationstoimprovethequalityofthepredictions. Because multiple imputation performance benefits from the addition of other traits that may be related to those of interest, we also included plant longevity (categorical trait), as this trait was complete at the genus level. The gapfilling step resulted in 24 trait data sets, where the variation between them represents the uncertainty ofthegap-fillingtechnique.Qualitativelyconsistentresultsfrom different imputed data sets indicate that the conclusions on the directionality of the effects are not influenced by the gapfilling procedure. 2.1.3 | Climatedata We constructed four environmental gradients. The first was daily average minimum temperature (°C) of the coldest month of the year andthesecondwasannualprecipitation(mm).Forboth,environments were assumed to be harsher at the low end of the gradient (Went, 1953). Third, we considered precipitation seasonality (%), defined as the coefficient of variation based on monthly precipitation.Fourth,weconsideredannualtemperaturerange(°C),defined as the absolute difference between the daily average maximum temperature of the warmest month and the daily average minimum TABLE 1 Listoffunctionalplanttraitsused.Missingdataforspeciesrepresentthepercentageofspeciesthathavenodataforthattrait. Missing data overall represent the percentage of gaps in the overall data set for that trait Trait Unit Descriptiona Related function Missing data (%) Species Overall Specific leaf area mm2 mg1 Onesided area of a fresh leaf divided by its ovendried mass Photosynthetic rate, drought tolerance 16.4 1.2 Leafnitrogen concentration mg g1 Total amount of nitrogen per unit of dry leaf mass Photosynthetic rate, stress tolerance 46.9 13.3 Plant height m Shortest distance between the upper boundary of the main photosynthetic tissues on a plant and the ground level Lightcapture,above-ground competition, dispersal distance 3.0 0.7 Seed mass g Ovendried mass of an average seed of a species Dispersal distance, seedling competition 12.8 1.7 Specific root length mm mg1 Ratio of root length to dry mass of fine roots Resourceuptake,stresstolerance 85.3 35.3 Rooting depth m Maximum soil depth from which resources canbeacquired Resourceuptake,droughttolerance 76.1 35.9 aStandardizeddefinitionfromPérez-Harguindeguyetal.(2013). | 5 of 12 Journal of Vegetation Science BOONMAN et Al. temperature of the coldest month of the year. Other climatic variables were also used in this study as additional information (daily average warmest temperature of the warmest month [°C], mean temperatureinthewarmestquarteroftheyear[°C],meantemperatureinthecoldestquarteroftheyear[°C],annualmeantemperature [°C],temperatureseasonality).Allclimaticdataweredownloaded fromCHELSAversion1.2ataresolutionof0.01degree(~1kmgrid cells;Kargeretal.,2017). 2.2 | Diversity We used two diversity metrics, taxonomic richness and multivariate functional richness. Taxonomic richness quantifies the number of species present in a community or sampling plot, while functional richness expresses the minimum volume encompassing the most extreme trait values in an assemblage and is a commonly used metrictoquantifychangesinassemblyprocessesalongenvironmental gradients(Mason&deBello,2013;Kraftetal.,2015;Münkemüller et al., 2020). These metrics suit the goals of describing plant diversity along environmental gradients. To calculate functional richness, we first created a speciesspecific trait distance matrix based on scaled values of the six selected traits and using multivariate Euclidean distances without trait weighting. Second, we performed a Principal CoordinateAnalysis(PCoA)onthedistancematrix.Third,weused theresultingcoordinatesofthePCoAtobuildamulti-variatetrait spectrum of all traits within a plot. Each data point in the multivariate trait spectrum describes the traits of a species, and the distances between points represent the similarity or diversification in traits between species. The global trait functional richness was computed as the smallest sixdimensional convex hull enclosing all trait valuesinanassemblage.Foraspecificplot,thefunctionalrichnessis quantifiedasthevolumeoftheconvexhullthatenclosesallspecies of that individual plot (Villéger et al., 2008; Mouillot et al., 2013). In this functional richness calculation method, selecting the numberofaxesresultingfromthePCoAwasatrade-off.Thiscalculation needs a higher number of species per plot than the number of includedPCoAaxesandthusconvexhulldimensions.Wechoseto include all six axes to avoid excluding any trait variation. The removal of plots with six or fewer species did not create any bias as a low FIGURE 1 Geographic distribution of vegetation plots used in this study. Colours represent the minimum temperature (a) and annual temperature range (b) of each plot in °C, as well as the annual precipitation in mm (c) and precipitation seasonality in percentages (d). The environmentaldistributionofalllocationsisplottedinAppendixS4 6 of 12 | Journal of Vegetation Science BOONMAN et Al. number of species per plot occurred along the entire environmental gradient and followed the same pattern as plots with more than six species(AppendixS4).Removingthesecommunitiesresultedin24 final data sets, each with 55,748 plots and 2,830 different species (Figure1). 2.3 | Null model We used a null model to test for potential effects of taxonomic richness on functional richness, revealing nonrandom patterns that indicate ecological processes as opposed to random patterns expected bychance(Gotelli&Graves,1996;Swensonet al.,2012;Chalmandrier etal.,2013).Hence,wecalculatedafunctionalrichnesseffectsizeusing results from a null model which randomly shuffles species trait values inthetraitdatasetbutretainsspecies’traitcombinations.Specifically, thespecies’PCoAscoreswererandomlyshuffledamongspecies(but not within species) from the entire species pool (de Bello et al., 2012). Thenumberofspeciespercommunitywaskeptequaltothenumber of species observed to ensure the assessment of functional richness trends, independent of differences in taxonomic richness (similarly to Swensonet al.,2012;Cravenetal.,2018). The effect sizes were calculated using probabilities since the distributionofthenullmodelwasnotsymmetric(AppendixS5;Ulrich& Gotelli,2010;Bernard-Verdieretal.,2012;Lhotskyetal.,2016).We usedtheprobitlinkfunction(Φ1) of the VGAMpackage: where Here, we estimated the one-tailed probability of the observed functional richness (obs) having a lower value than expected compared to the functional richness values resulting from the null model with1,000iterations(NULL)(Equation2).Wheneffectsizecalculations result in values close to zero (i.e., obs is close to the median of the null distribution), functional diversity is not different from the null expectationofnofilteringofplanttraits.Largerdeviationsofeffect size values from zero indicate larger differences between observed and expected (by chance, i.e., without any plant trait community assembly processes) functional richness. Since we randomized trait values across species, negative effect size values indicate less variation in trait range than would be expected based on the number of species in the community, i.e., functional convergence or functional underdispersion (clustering). Positive effect size values indicate more variation in trait range than would be expected based on the number of species in the community, i.e., functional divergence or functional overdispersion (BernardVerdier et al., 2012; Swenson et al., 2012; Lhotskyetal.,2016). 2.4 | Statistical analyses We used the average of the original functional richness values and effect size values in one grid cell (~1km2) if more than one plot was sampled to reduce possible effects of spatial autocorrelation. This resultedin19,179datapointsandgridcells(Figure1).Foreachofthe 24 imputed data sets, functional richness, taxonomic richness, and effect sizes were assessed along gradients of minimum temperature, temperature range, annual precipitation and precipitation seasonality in multiple generalized additive models with a Gaussian distribution using the gamm4 package. These models allowed for finding patterns in the data without a priori hypotheses on the shapeoftherelationship.However,werestrictedthecurvinessof thetrendsbysettingtheknots(k) to 4 to prevent overfitting. Since the difference between the models was small (Appendix S6), we averaged the trait data over the 24 imputed data sets, recalculated functional richness values, reran the null model, and used these data in our plots of functional richness and effect sizes so that confidence intervals could indicate model uncertainty instead of differences between imputed data sets. 3 | RESULTS The diversity of grassland communities varied considerably over temperatureandprecipitationgradients(Figure2;AppendixS7).We found functional and taxonomic richness to be lower in areas with low minimum temperatures and in areas with narrow temperature ranges(Figure2a,c).Thesecorrespondedtolocationsinmountain ranges or with higher latitude or longitude and locations on islands or along the west coast of France, Belgium and the Netherlands, respectively(AppendixS9).Thissimilarityintrendbetweenfunctional and taxonomic richness may be due to a substantial correlation (r =0.7;AppendixS8).However,atwideannualtemperature ranges, functional richness increased with increasing minimum temperatures(Figure2a),whiletaxonomicrichnessdecreasedwith increasing minimum temperatures after reaching an optimum at a minimum temperature of around −5 °C (Figure 2c). These plots were located in continental Europe, where the highest minimum temperatures (i.e., highest functional richness values) were found alongthecoastsofsouthernEurope(AppendixS9).Notethatfor all temperature variables (annual maximum, maximum of the warmestquarter,annualminimum,minimumofthecoldestquarter,and annual mean temperature), both richness metrics showed the same non-linearpatternasfortheminimumtemperature(AppendixS10). Likewise,temperaturevariationvariables(annualtemperaturerange and temperature seasonality) showed the same nonlinear pattern (Appendix S10). Consideringthe variation of functional and taxonomic richness across precipitation gradients, we found a decrease inrichnesswithincreasingprecipitationseasonality(Figure2b,d; Appendix S9). In addition, functional and taxonomic richness appeared to have an optimum at average values of annual precipitation, (1) ES =Φ − 1(p) (2) p = number(NULL <obs) + number(NULL = obs) 2 1, 000 | 7 of 12 Journal of Vegetation Science BOONMAN et Al. though the variation in functional richness across this gradient was small(Figure2b,d).Overall,temperaturecouldexplainmoreofthe variation in functional and taxonomic richness than precipitation (AppendixS11). Assessingthevariationintraitdiversitywhilecontrollingfor the number of species in the assemblage, functional richness effect sizes showed an opposite trend from functional and taxonomicrichnessalongallenvironmentalgradients(Figure2e,f). Effect size values were positive when minimum temperatures droppedbelow−11°Corexceeded−1°C,representingsampled grasslands in the Alps, Russia and Scandinavia, and in southwesternEuroperespectively(AppendixS9).Largereffectsizes were found in locations with extremely low minimum temperature and at low annual precipitation with extreme precipitation seasonality. Slightly negative effect size values were found at intermediate minimum temperatures, where sampled grasslands aremostlylocatedincentralEurope(Figure2e;AppendixS9). In contrast, effect sizes did not show a clear pattern along the temperaturerangegradient(Figure2e),nordidtheyvarymuch acrossbothprecipitationgradients(Figure2f). 4 | DISCUSSION The results of our continentalscale study show that European grassland communities vary substantially in functional and taxonomic richness along temperature and precipitation gradients. Highestrichnessinbothtraitsandspecieswasfoundincommunities experiencing favorable, intermediate to warm climates. These locations contain more species with trait combinations that show FIGURE 2 Heatmapsdepictingfitted results of three multiple generalized additive models. Values for functional richness (a + b) indicate the logarithm of the original values, while values for taxonomic richness (c + d) and effect sizes of functional richness (e + f) represent the original values. In panels (a– d), lighter colours indicate lower functional richness valuesanddarkercoloursindicatehigher functional or taxonomic richness values. In panels (e) and (f), yellow to red colours are positive effect sizes, and blue to purple colours are negative effect sizes. White spaces are locations without data and/or represent combinations of climatic variables that do not exist anywhere in Europe 8 of 12 | Journal of Vegetation Science BOONMAN et Al. greater functional variation compared to other locations. This pattern is expected under the favorability hypothesis of lower diversity in harsh (low minimum temperatures and less precipitation) environments and higher diversity in more benign (warmer and more frequent precipitation) environments (Fischer, 1960). Additionally,thispatternisinlinewithtaxonomicandfunctional diversity patterns found in non-European grasslands (Moradi & Oldeland,2019).NotethatMoradi&Oldeland(2019)reportedthat precipitation limits plant diversity more than minimum temperatureindryareas,whichcontrastswithourresults(AppendixS11). This may suggest that, when drier areas would have been included in this study, taxonomic richness may decrease more toward more waterlimited sites and determine the diversity patterns more than temperature.Anotherexplanationofthepeakinfunctionaland taxonomic richness at average minimum temperature values is the mid-domaineffect(Colwell&Hurtt,1994;Colwell&Lees,2000). Here,thegeometriclimitsinrelationtothedistributionofspecies increases the overlap of species ranges and thus of species richnessinthemiddleofthegradient(Colwell&Lees,2000).We assume this effect to be of minor importance as grasslands extendfromsouthwesternEuropetoMiddleandCentralAsia,many grassland species have large geographical ranges (extending beyond the area included in this analysis), and many included species also occur in other habitats besides grasslands. Remarkably, functionalrichnessvaluesfurther increasedtoward higher minimum temperatures and wider temperature ranges, while taxonomic richness decreased. Environments with a wide temperature range may hold coexisting species with dissimilar trait values as they can exploit different temporal niches, thereby dominating in differentseasonswithoutoutcompetingtheotherspecies(Figure2e;Díaz et al., 1999). This may explain the positive effect sizes at high minimum temperaturesandwidetemperatureranges(Scheiner&Rey-Benayas, 1994;González-Morenoetal.,2015).Nevertheless,suchahypothesis should be investigated further by inspecting variation in species abundance in a community over time (Tilman, 1996; Pescador et al., 2015). Furthernotethatduetothedisregardofintraspecificvariationinthis study, we potentially overestimate functional richness in environmentally constricted areas as adaptive traits may be more similar between species in these locations due to environmental filtering under the favorability hypothesis (Grant & Abbott, 1980; Swenson & Enquist, 2009; Swenson et al., 2012). Conversely, we may underestimate functional richness in other areas as adaptive traits may be more dissimilar betweenspeciesundertheprincipleoflimitingsimilarity(Weiher& Keddy,1995;Stubbs&Wilson,2004;Mason&Wilson,2006;Violle etal.,2011).Futurefieldcampaignsmightinvestinlocaltraitmeasurements for all species to enable the inclusion of intraspecific trait variation(Albertetal.,2010;Niuetal.,2020). Functional and taxonomic richness decreased toward higher minimum temperatures and narrow temperature ranges. Under the stress gradient hypothesis, the reduced diversity may indicate strong competitiondue toa benign, warmclimate(Bertness&Callaway, 1994).However,temperature rangeis known toaffect functional richness(González-Morenoetal.,2015)andmightactasastressorin these locations explaining the reduction in functional and taxonomic diversity under the favorability hypothesis. Since positive effect sizes were found in these environments, it could be a more stressful environment where facilitation possibly causes trait divergence. This may follow the same explanation for the trait divergence found at extreme low minimum temperatures. In addition, facilitation was also observed in an alpine study where greater variation in temperature made for more stressful environments (Molenda et al., 2012). Nevertheless, these opposite effects may suggest that the locations with high minimum temperatures and narrow temperature ranges might be restricted by additional stressors that were not included in thisstudy,likesummertemperatures,wind,solarradiation,oreven differentherbivorylevels.Experimentsarerequiredtodisentangle the causal mechanisms underlying the functional and taxonomic richness patterns. They may also determine if and which environmental factors determine the diversity in these European grasslands, but our results indicate the importance of climate seasonality. Interpreting patterns in community diversity along climatic gradients is difficult, particularly because the variation in our data explained by the temperature and precipitation variables was relativelylow(8–10%;AppendixS10).Resultsshouldbeinterpretedwith cautionandexperimentsarerequiredtobackuppossibleexplanations of diversity patterns. The low explained variation indicates that other factors beside climate vary between locations and play a large role in determining the plant diversity of European grassland communities, such as variation in soil, herbivory, management, landscape history and various biogeographical influences (e.g., Willems, 1983; Bakkeretal.,2006;Daineseetal.,2015).Itshouldalsobeemphasized that functional diversity is always dependent on the traits that are included in the study, where results may change when different traits are selected (Villéger et al., 2008). This means that the conclusions regarding functional richness trends in this study depend on thespecificsetoftraitsweselected.Futureresearchcouldbenefit fromtheever-growingavailabilityoftraitdata(Kattgeetal.,2020) and vegetationplot data from other ecosystems (Bruelheide et al., 2019) to assess the generality of plant community diversity patterns along environmental gradients and to assess how community assembly may change in response to global warming (Mouillot et al., 2007; Mason et al., 2011; Catford et al., 2020). ACKNOWLEDGEMENTS The study was supported by the TRY initiative on plant traits (http:// www.try-db.org)andtheEuropeanVegetationArchive(EVA;http:// eurov eg.org/evadatabase). The TRY initiative and database are hosted, developed and maintained by Jens Kattge and Gerhard Bönisch(MaxPlanckInstituteforBiogeochemistry,Jena,Germany). TRY is currently supported by DIVERSITAS/Future Earth and the GermanCentreforIntegrativeBiodiversityResearch(iDiv)Halle– Jena–Leipzig. EVA is a database of the Working Group European VegetationSurveyoftheInternationalAssociationforVegetation Science.WethankSergeiN.Sheremet'evforthecontributionofThe GlobalLeafTraitsdatasettoTRYandIlonaKnollováfortechnical assistancewiththeEVAdata. | 9 of 12 Journal of Vegetation Science BOONMAN et Al. AUTHOR CONTRIBUTIONS CCFB conceived the research idea, with significant contributions ofLS, BJMR and MAJH; CCFBgathered data andperformed the analyses. CCFB wrote the manuscript with major contributions fromLS,BJMR,SH,MCandMAJH;allauthorscommentedonthe manuscript. DATA AVAILABILITY STATEMENT The functional richness data and R scripts on the randomization and null model are available via DANS-EASY. The original trait datacanberequestedviaTRY.Vegetation-plotdatasetusedin thisprojectisstoredintheEuropeanVegetationArchiveunder project no. 97. ORCID Coline C. F. Boonman https://orcid.org/0000-0003-2417-1579 Luca Santini https://orcid.org/0000-0002-5418-3688 Bjorn J. M. Robroek https://orcid.org/0000-0002-6714-0652 Selwyn Hoeks https://orcid.org/0000-0001-5619-3233 Jürgen Dengler https://orcid.org/0000-0003-3221-660X Ariel Bergamini https://orcid.org/0000-0001-8816-1420 Idoia Biurrun https://orcid.org/0000-0002-1454-0433 Maria Laura Carranza https://orcid.org/0000-0001-5753-890X Bruno E. L. Cerabolini https://orcid.org/0000-0002-3793-0733 Milan Chytrý https://orcid.org/0000-0002-8122-3075 Ute Jandt https://orcid.org/0000-0002-3177-3669 Tatiana Lysenko https://orcid.org/0000-0001-6688-1590 Angela Stanisci https://orcid.org/0000-0002-5302-0932 Irina Tatarenko https://orcid.org/0000-0001-6835-2465 Solvita Rūsiņa https://orcid.org/0000-0002-9580-4110 Mark A. J. Huijbregts https://orcid.org/0000-0002-7037-680X REFERENCES Albert,C.H.,Thuiller,W.,Yoccoz,N.G.,Soudant,A.,Boucher,F.,Saccone, P. et al. (2010) Intraspecific functional variability: extent, structure and sources of variation. Journal of Ecology, 98, 604– 613. https://doi. org/10.1111/j.13652745.2010.01651.x Bakker, E.S., Ritchie, M.E., Olff, H., Milchunas, D.G. & Knops, J.M.H. (2006) Herbivore impact on grassland plant diversity depends on habitat productivity and herbivore size. Ecology Letters, 9, 780– 788. https://doi.org/10.1111/j.14610248.2006.00925.x Barbour,M.A.,Erlandson,S.,Peay,K.,Locke,B.,Jules,E.S.&Crutsinger, G.M. (2019) Trait plasticity is more important than genetic variation in determining species richness of associated communities. Journal of Ecology, 107, 350– 360. https://doi.org/10.1111/13652745.13014 Bernard-Verdier,M.,Navas,M.-L.,Vellend,M.,Violle,C.,Fayolle,A.& Garnier, E. (2012) Community assembly along a soil depth gradient: contrasting patterns of plant trait convergence and divergence in a Mediterranean rangeland. Journal of Ecology, 100, 1422– 1433. https://doi.org/10.1111/13652745.12003 Bertness,M.D.&Callaway,R.(1994)Positiveinteractionsincommunities. Trends in Ecology & Evolution, 9, 191– 193. Breitschwerdt,E.,Jandt,U.&Bruelheide,H.(2018)Usingco-occurrence information and trait composition to understand individual plant performance in grassland communities. Scientific Reports, 8, 9076. https://doi.org/10.1038/s4159 801827017 - 9 Brooker,R.W.&Callaghan,F.V.(1998)Thebalancebetweenpositiveand negative interactions and its relationship to environmental gradients: a model. Oikos, 81, 196– 201. Bruelheide,H.,Dengler,J.,Jiménez-Alfaro,B.,Purschke,O.,Hennekens, S.M.,Chytrý,M.et al.(2019)sPlot–Anewtoolforglobalvegetation analyses. Journal of Vegetation Science, 30, 161– 186. https://doi. org/10.1111/jvs.12710 Cadotte,M.W.&Tucker,C.M.(2017)Shouldenvironmentalfilteringbe abandoned? Trends in Ecology & Evolution, 32(6), 429– 437. https://doi. org/10.1016/j.tree.2017.03.004 Cardinale,B.J.,Duffy,E.,Gonzalez,A.,Hooper,D.U.,Perrings,C.,Venail, P. et al. (2012) Biodiversity loss and its impact on humanity. Nature, 486(7401), 59– 67. https://doi.org/10.1038/natur e11148 Carmona, C.P., Azcárate, F.M., de Bello, F., Ollero, H.S., Lepš, J. & Peco, B. (2012) Taxonomical and functional diversity turnover in Mediterranean grasslands: interactions between grazing, habitat type and rainfall. Journal of Applied Ecology, 49, 1084– 1093. https:// doi.org/10.1111/j.13652664.2012.02193.x Catford,J.A.,Dwyer,J.M.,Palma,E.,Cowles,J.M.&Tilman,D.(2020) Community diversity outweighs effect of warming on plant colonization. Global Change Biology, 26, 3079– 3090. https://doi.org/10.1111/ gcb.15017 Chalmandrier, L., Münkemüller, T., Gallien, L., de Bello, F., Mazel, F., Lavergne,S.et al.(2013)Afamilyofnullmodelstodistinguishbetween environmental filtering and biotic interactions in functional diversity patterns. Journal of Vegetation Science, 24, 853– 864. https:// doi.org/10.1111/jvs.12031 Chytrý,M.,Hennekens,S.M.,Jiménez-Alfaro,B.,Knollová,I.,Dengler, J.,Jansen,F.et al.(2016)EuropeanVegetationArchive(EVA):anintegrated database of European vegetation plots. Applied Vegetation Science, 19, 173– 180. https://doi.org/10.1111/avsc.12191 Chytrý, M., Tichý, L., Hennekens, S.M., Knollová, I., Janssen, J.A.M., Rodwell,J.S.et al.(2020)EUNISHabitatClassification:expertsystem, characteristic species combinations and distribution maps of European habitats. Applied Vegetation Science, 23, 648– 675. https:// doi.org/10.1111/avsc.12519 Colwell, R.K. & Hurtt, G.C. (1994) Nonbiological gradients in species richness and a spurious Rapoport effect. The American Naturalist, 144(4), 570– 595. Colwell, R.K. & Lees, D.C. (2000) The mid-domain effect: geometric constraints on the geography of species richness. Trends in Ecology & Evolution, 15(2), 70– 76. Craven,D.,Hall,J.S.,Berlyn,G.P.,Ashton,M.S.&vanBreugel,M.(2018) Environmental filtering limits functional diversity during succession in a seasonally wet tropical secondary forest. Journal of Vegetation Science, 29, 511– 520. https://doi.org/10.1111/jvs.12632 Dainese,M.,Lepš,J.&deBello,F.(2015)Differenteffectsofelevation, habitat fragmentation and grazing management on the functional, phylogenetic and taxonomic structure of mountain grasslands. Perspectives in Plant Ecology, Evolution and Systematics, 17, 44– 53. https://doi.org/10.1016/j.ppees.2014.09.002 deBello,F.,Price,J.N.,Münkemüller,T.,Liira,J.,Zobel,M.,Thuiller,W. et al.(2012)Functionalspeciespoolframeworktotestforbioticeffects on community assembly. Ecology, 93, 2263– 2273. https://doi. org/10.1890/111394.1 deBello,F.,Thuiller,W.,Lepš,J.,Choler,P.,Clément,J.-C.,Macek,P.et al. (2009) Partitioning of functional diversity reveals the scale and extent of trait convergence and divergence. Journal of Vegetation Science, 20, 475– 486. https://doi.org/10.1111/j.16541103.2009.01042.x Dengler,J.,Biurrun,I.,Boch,S.,Dembicz,I.&Török,P.(2020)Grasslands of the Palaearctic biogeographic realm: introduction and synthesis. In:Goldstein,M.I.,DellaSala,D.A.&DiPaolo,D.A.(Eds.)Encyclopedia of the World’s Biomes. Volume 3: Forests – Trees of Life. Grasslands and Shrublands – Sea of Plants.Amsterdam,NL:Elsevier,pp.617–637.