Received: 6 November 2018 | Revised: 17 February 2020 | Accepted: 18 February 2020 DOI: 10.1111/geb.13086 RE SE ARCH PAPER Assessing the reliability of predicted plant trait distributions at the global scale Coline C. F. Boonman1 | Ana Benítez-López1,2 | Aafke M. Schipper1,3 | 4 5 6 Wilfried Thuiller | Madhur Anand | Bruno E. L. Cerabolini | 7 8 Johannes H. C. Cornelissen | Andres Gonzalez-Melo | Wesley N. Hattingh9 | Pedro Higuchi10 | Daniel C. Laughlin11 | Vladimir G. Onipchenko12 | 13,14 15 16 Josep Peñuelas | Lourens Poorter | Nadejda A. Soudzilovskaia | 1 1,17 Mark A. J. Huijbregts | Luca Santini 1 Department of Environmental Science, Institute for Water and Wetland Research, Abstract Radboud University, Nijmegen, Aim: Predictions of plant traits over space and time are increasingly used to improve the Netherlands 2 our understanding of plant community responses to global environmental change. A Integrative Ecology Group, Estación Biológica de Doñana (EBD-CSIC), Sevilla, necessary step forward is to assess the reliability of global trait predictions. In this Spain study, we predict community mean plant traits at the global scale and present a sys- 3 PBL Netherlands Environmental Assessment Agency, The Hague, tematic evaluation of their reliability in terms of the accuracy of the models, ecologi- the Netherlands cal realism and various sources of uncertainty. 4 Université Grenoble Alpes, CNRS, University Location: Global. of Savoie Mont Blanc, LECA, Laboratoire d’Écologie Alpine, Grenoble, France Time period: Present. 5 School of Environmental Sciences, Major taxa studied: Vascular plants. University of Guelph, Guelph, Ontario, Methods: We predicted global distributions of community mean specific leaf area, Canada 6 Department of Theoretical and Applied leaf nitrogen concentration, plant height and wood density with an ensemble model- Science, University of Insubria, Varese, Italy ling approach based on georeferenced, locally measured trait data representative of 7 Systems Ecology, Department of Ecological the plant community. We assessed the predictive performance of the models, the Science, Vrije Universiteit, Amsterdam, the Netherlands plausibility of predicted trait combinations, the influence of data quality, and the un- 8 Facultad de Ciencias Naturales y certainty across geographical space attributed to spatial extrapolation and diverging Matemáticas, Universidad del Rosario, Bogota, Colombia model predictions. 9 School of Animal, Plant and Environmental Results: Ensemble predictions of community mean plant height, specific leaf area Sciences, University of the Witwatersrand, and wood density resulted in ecologically plausible trait–environment relationships Johannesburg, South Africa 10 and trait–trait combinations. Leaf nitrogen concentration, however, could not be Forestry Department, Santa Catarina State University, Lages, Brazil predicted reliably. The ensemble approach was better at predicting community trait 11 Department of Botany, University of means than any of the individual modelling techniques, which varied greatly in pre- Wyoming, Laramie, WY, USA 12 dictive performance and led to divergent predictions, mostly in African deserts and Department of Geobotany, Moscow Lomonosov State University, Moscow, the Arctic, where predictions were also extrapolated. High data quality (i.e., including Russia 13 CREAF, Vallès, Catalonia, Spain 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. © 2020 The Authors. Global Ecology and Biogeography published by John Wiley & Sons Ltd 1034 |  wileyonlinelibrary.com/journal/geb Global Ecol Biogeogr. 2020;29:1034–1051. BOONMAN et al. | 1035 14 CSIC, Global Ecology Unit CREAF-CEAB- UAB, Catalonia, Spain intraspecific variability and a representative species sample) increased model perfor- 15 Forest Ecology and Forest Management mance by 28%. Group, Wageningen University and Main conclusions: Plant community traits can be predicted reliably at the global scale Research, Wageningen, the Netherlands 16 Environmental Biology Department, when using an ensemble approach and high-quality data for traits that mostly re- Institute of Environmental Sciences, Leiden spond to large-scale environmental factors. We recommend applying ensemble fore- University, Leiden, the Netherlands 17 casting to account for model uncertainty, using representative trait data, and more National Research Council, Institute of Research on Terrestrial Ecosystems routinely assessing the reliability of trait predictions. (CNR-IRET), Monterotondo, Italy KEYWORDS Correspondence ensemble forecasting, environmental filtering, intraspecific trait variation, leaf nitrogen Coline C. F. Boonman, Department of Environmental Science, Institute for Water concentration, plant height, specific leaf area, trait–environment relationships, trait model, and Wetland Research, Radboud University, wood density PO Box 9010, NL-6500 GL Nijmegen, The Netherlands. Email: c.boonman@science.ru.nl Funding information H2020 European Research Council, Grant/ Award Number: 647224 Editor: Brian McGill 1 | I NTRO D U C TI O N the selection of environmental predictors; and (d) the model fitting techniques. The first difference translates to the use of global spe- Global trait-based models have proliferated in recent years owing to cies trait averages, which may cause a potential mismatch of trait the increasing availability of plant trait data (Kattge, 2019). Fitting and environmental data. The second and third differences may re- and projecting trait–environment relationships over large spatial sult in different spatial patterns of uncertainty owing to extrapola- scales is becoming increasingly common to study trait trade-offs tions of trait–environment relationships (Thuiller, Brotons, Araujo, (e.g., Díaz et al., 2016; Wright et al., 2004), to relate traits to envi- & Lavorel, 2004). The fourth difference may render considerable ronmental gradients (e.g., Moles et al., 2009; Wright et al., 2005) variation among predictions (Thuiller, Guéguen, Renaud, Karger, and to describe geographical patterns of traits (e.g., Madani et al., & Zimmerman, 2019). Furthermore, the ecological realism of the 2018; Yang et al., 2016). Trait-based models can not only increase combination of predicted plant traits needs to be tested against our understanding of trait–environment relationships, but can also observed trait combinations, which has, to our knowledge, not yet allow us to estimate how plant traits might respond to global en- been done at the global scale. Additionally, although all studies re- vironmental change (Bjorkman et al., 2018). This, in turn, is consid- ported the variance explained by the trait-based models, the pre- ered particularly useful for predicting the impact of environmental dictability of independent samples (i.e., data not used to train the change on vegetation (Webb, Hoeting, Ames, Pyne, & LeRoy Poff, models) has not yet been assessed thoroughly. A necessary step 2010), because trait-based models allow to directly link plant fitness forward to increase macroecological insights in global trait–environ- to environmental filters, including climate, soil properties and distur- ment relationships with potential application in ecological impact bance (Keddy, 1992). or conservation assessments (e.g., Lavorel & Garnier, 2002; Madani Recent attempts to model global plant trait distributions as a et al., 2018) is to perform a thorough assessment of the reliability of function of environmental conditions have yielded different trait global plant trait predictions. patterns (e.g., Butler et al., 2017; Madani et al., 2018; Moreno- Here, we predict community mean plant traits at the global Martínez et al., 2018; Van Bodegom, Douma, & Verheijen, 2014). scale and present a systematic evaluation of the reliability of the For example, Butler et al. (2017) and Moreno-Martínez et al. (2018) predictions in terms of the models' accuracy, ecological realism predicted specific leaf area to be low in western Canada and high and various sources of uncertainty. We systematically selected in northern Russia and eastern Brazil, whereas Van Bodegom et al. locally measured, representative data focusing on four widely (2014) and Madani et al. (2018) predicted the opposite. Such dis- studied plant traits (specific leaf area, leaf nitrogen concentration, crepancies among existing global trait models might arise from dif- height and wood density). These traits reflect the global spectra in ferences in: (a) the sources of the underlying trait data (floras and plant form and function, are responsive to the abiotic environment measurements in natural areas, manipulated field experiments, and show physical trade-offs with other traits (Table 1; Díaz et al., botanical gardens and greenhouses); (b) the representativeness 2016; Lavorel & Garnier, 2002). For each of the traits, we calcu- of the sampled species for the entire natural plant community; (c) lated community mean trait values and predicted global patterns at 1036 | BOONMAN et al. TA B L E 1 Expectations for community mean trait–environment relationships Predictor Rationale Impact and adaptations Expectations Minimum temperature of the Temperature-induced leaf damage, at low Lower damage risk with increased ˄ SLA coldest month (Tmin) and high Tmin, affects plant performance1 tolerance via decreasing SLA 2 Enhanced metabolic activity and frost ↓ LNC4 tolerance with increased LNC1,3 Lower frost-induced mortality risk ↑ Plant height4 with increasing Tmin ↑ Wood density4 Humidity index (HumInd) Increased drought tolerance with low ↑ SLA 2 transpiration rates via decreased SLA ↑ LNC2 Lower drought-induced mortality Cavitation risk increases with height 5 ↑ Plant height risk with increasing HumInd Increasing wood density increases plant ↓ Wood density performance at low HumInd6 Precipitation in the driest ᴑ SLA quarter of the year (PrecDryQ) ᴑ LNC Lower drought-induced mortality Cavitation risk increases with height5 ↑ Plant height risk with increasing PrecDryQ ᴑ Wood density Precipitation seasonality Moisture availability affects Changing leaf habit (deciduousness) not ‒ SLA (PrecSeas) plant survival traits7 ‒ LNC4 ↓ Plant height4 ↓ Wood density4 Soil cation exchange capacity Higher CEC indicates a greater capacity to Less durable structures can be ↑ SLA (CEC) retain easily attainable nutrients maintained9,10 ‒ LNC (i.e., higher soil fertility) 8 ᴑ Plant height ↓ Wood density Soil pH Higher pH increases the available Less durable structures can be ↑ SLA phosphorus and nitrogen in the maintained9,10 ‒ LNC soil (i.e., higher soil fertility)11 ᴑ Plant height ↓ Wood density Note: Hypotheses are indicated as follows: ↑ = positive relationship; ↓ = negative relationship; ‒ = flat relationship; ˄ = unimodal response. Hypotheses are based on theory (no reference) or on significant, biologically relevant trends found at the community level in literature (reference included). ᴑ indicates relationships between trait and predictor that have not been discussed at the community level in literature. Abbreviations: LNC = leaf nitrogen concentration; SLA = specific leaf area. References in table: 1Went (1953); 2Wright et al. (2005); 3Reich, Oleksyn, and Tjoelker (1996); 4Swenson and Weiser (2010); 5Tyree and Sperry (1989); 6 Reich (2014); 7Borchert (1998); 8Ross and Ketterings (1995); 9Chave et al. (2009); 10Wright et al. (2004); 11Maire et al. (2015). a 0.5° resolution (c. 55 km × 55 km at the equator) using an ensem- space owing to discrepancies among the predictions of the four ble modelling approach based on two regression and two machine modelling techniques. We build upon this assessment to provide learning techniques. Subsequently, we evaluated the predictive guidelines for the further development, interpretability and us- performance of the models and assessed their ecological plausibil- ability of global trait-based models. ity in terms of the trait–environment relationships and the correla- tions and combinations of individually predicted community mean trait values (Díaz et al., 2016; Lavorel & Garnier, 2002). Finally, 2 | M E TH O DS we evaluated the effect of various sources of uncertainty: (a) the effect of data quality in terms of representativeness of the sam- 2.1 | Plant functional traits pled species to the entire plant community and the use of global species trait averages versus local trait measurements; (b) the un- We selected four plant functional traits: specific leaf area (SLA; in certainty across geographical space attributed to extrapolation of square millimetres per milligram), leaf nitrogen concentration (LNC; traits outside the applicability domain (i.e., the geographical area in milligrams per gram), height (in metres) and wood density (in mil- with environmental variation covered by the environmental varia- ligrams per cubic millimetre). The SLA and LNC are both linked to tion of the trait data); and (c) the uncertainty across geographical photosynthetic capacity and nutrient investment, where a high SLA BOONMAN et al. | 1037 and high LNC represent a fast return on investment at the expense measurements conducted from 1980 onwards, in order to link up with of a shorter life span (Wright et al., 2004). Plant height is considered the environmental covariate data, which span from 1979 to 2013. We to be indicative of the ecological strategy of carbon distribution, excluded trait observations for which the listed criteria could not indirectly determining growth and reproduction and their initial re- be checked. The Supporting Information (Appendix S2) contains an sponse to climate change (Moles et al., 2009). Last, wood density is overview of the number of datasets excluded per criterion. In total, a measure of carbon investment, representing a trade-off between we removed 63.1% of all the datasets we originally received from TRY growth (e.g., to overcome light limitation) and strength (e.g., me- and 54.5% of all datasets in the TTT database. chanical support and drought tolerance), with higher wood density reflecting slower growth but increased strength at a similar stem diameter (Chave et al., 2009; Larjavaara & Muller-Landau, 2010). 2.3 | Data processing The selected traits are easily measured according to standardized measurement procedures (Perez-Harguindeguy et al., 2013), which fa- We checked and corrected for possible errors in our database, such cilitates the integration of data from multiple datasets and reduces trait as duplicates, coordinate and unit inaccuracies and outliers (see variation caused by seasonal changes (Bloomfield et al., 2018). To opti- Supporting Information Appendix S2). We then calculated location- mize between the number of observations and the consistency of the specific community means per trait and per study to represent the measurement methods, we included SLA measurements on both sun- mean response of the plant community to the local environment leaves and shade-leaves, wood density measurements on both heart- (Ackerly & Cornwell, 2007). Using unweighted community means wood and sapwood, and plant height measurements on both vegetative was the better compromise over abundance-weighted community and generative plant organs (see also Siefert et al., 2015). means, mainly because previous studies were not conclusive on the superiority of abundance-weighted means and because > 50% of our data did not include species abundances (for further discussion, see 2.2 | Data collection and selection procedure Supporting Information Appendix S3). We averaged the location- specific community means to 0.5° grid cells (with a median of 22 Our main source for plant trait data was the TRY database (Kattge communities per grid cell) in order to describe large-scale patterns in et al., 2011), from which we received 964,464 trait records on 85,437 traits. This is a common resolution used in global plant trait studies species from 168 datasets. We also obtained data from the Tundra (Butler et al., 2017; Van Bodegom et al., 2014) and a good compro- Trait Team (TTT) database (Bjorkman et al., 2018) and from various mise, accounting for the effect of climatic filtering on plant traits, other published and unpublished datasets (Supporting Information the coordinate uncertainty of the trait data, avoidance of pseudo- Appendix S1). A list of data sources is provided in the Appendix. All replicates and the uncertainty related to interpolated climatic vari- species names were standardized using The Plant List (2013). ables (Stoklosa, Daly, Foster, Ashcroft, & Warton, 2015). These We selected trait observations based on six criteria. First, we in- community means represent plant community average trait values, cluded only georeferenced observations in order to enable a mean- thereby making values not directly transferable to a specific species ingful link with environmental covariates. Second, we considered or a class of species. The final dataset used as input for model fitting only real measurements of plant traits (i.e., no species-level averages) included community mean trait values based on data from 76 stud- in order to include intraspecific trait variation, which can contribute ies and 8,955 species averaged to 486 grid cells, including 361 grid substantially to the trait variation within and between communities cells for SLA, 338 for LNC, 217 for height and 125 for wood density (Albert et al., 2010; Bloomfield et al., 2018; Siefert et al., 2015). Third, (Supporting Information Figure S4). we considered only measurements obtained from natural vegeta- tion, to minimize the influences of local management practices and legacy effects of historical land use (Perring et al., 2017). Fourth, 2.4 | Environmental data we included observations only from studies that measured all or the most abundant species present in the entire plant community or in We considered environmental variables that are expected to affect the dominant vegetation structure, in order to account for the repre- plant performance (e.g., Kimball, Gremer, Angert, Huxman, & Venable, sentativeness of the sampled species for the plant community. This 2012). This allowed us to check the ecological plausibility of the result- is in line with previous studies on this topic (e.g., Poorter et al., 2017) ing trait–environment relationships. Based on ecological relevance, we and in accordance with the biomass ratio hypothesis (Grime, 1998). selected bioclimatic variables from CHELSA v.1.2 (Karger et al., 2017). Fifth, we included observations only from studies that targeted all Given that water availability to plants is not determined by precipitation life stages and/or size classes and from studies that targeted only alone, we calculated the aridity index as the mean annual precipitation adults. Thus, we excluded observations from studies focusing only divided by the mean annual potential evapotranspiration, which in turn on early-successional plant communities and studies measuring only was calculated using the Penman–Monteith model (Zomer, Trabucco, seedlings or juveniles in a more established vegetation, in order to re- Bossio, & Verchot, 2008). However, given that lower values of the arid- duce confounding effects of ontogeny (e.g., Thomas & Bazzaz, 1999) ity index indicate higher aridity, to avoid confusion this predictor will and succession (e.g., Purschke et al., 2013). Sixth, we included only be referred to as the “humidity index” from now on. Furthermore, we 1038 | BOONMAN et al. selected soil characteristics from SoilGrids250m (Hengl et al., 2017) each of the four traits, following the procedure described above for and resampled them to a resolution of 0.5° to match the resolution our default dataset. We then evaluated each model separately, where of the plant community mean trait data. We averaged the soil data to the predictions are confronted with the observed community means a depth of 30 cm, which we considered most relevant for community of the full dataset (Supporting Information Appendix S7). composition via plant establishment and by influencing plants in later life stages, for example, through the potentially high nutrient availabil- ity (e.g., Vitousek & Sanford, 1986). 2.7 | Spatial predictions and their assessment To avoid collinearity, we reduced the number of predictors based on variance inflation factors (VIF < 4), while ensuring a combination To derive spatial predictions per trait, we used an ensemble fore- of different environmental factors (i.e., both climatic and soil charac- casting procedure, which averages the predictions of the four teristics; Franklin, Serra-Diaz, Syphard, & Regan, 2016). A complete algorithms weighted by their cross-validated pseudo-R 2 values list of the predictors considered and more details on our selection (Marmion, Parviainen, Luoto, Heikkinen, & Thuiller, 2009). Likewise, procedure can be found in the Supporting Information (Appendix we estimated ensemble variable importance and partial response S5). The final selection included six predictors: minimum tempera- curves. Furthermore, we calculated the accuracy of the ensemble ture of the coldest month (Tmin; in degrees Celsius), humidity index predictions by regressing the 20% test data of the cross-validations (HumInd; dimensionless), precipitation in the driest quarter of the against the observed trait values. We checked whether among-trait year (PrecDryQ; in millimetres), precipitation seasonality (PrecSeas, correlations observed in the data were retained in the predictions coefficient of variation based on monthly precipitation; as a per- by correlating the trait correlation matrix of the original dataset with centage), soil cation exchange capacity (CEC; centimoles of positive the trait correlation matrix of the predictions. Additionally, we as- charge per kilogram of soil) and soil pH. sessed whether predicted trait combinations were realistic by com- paring them with a hypervolume of all trait combinations existing in our dataset. We built the hypervolume with the “hypervolume_svm” 2.5 | Model fitting and validation function of the “hypervolume” R package using the original trait values before calculating community means or grid cell averages, For each relationship between the plant traits and environmental vari- because we wanted to check whether predicted trait combinations ables, we formulated expectations on the shape of the response based occurred naturally and because this ensured enough observations on existing literature (Table 1). To quantify the trait–environment rela- with all four traits measured to build the hypervolume (Blonder, tionships, we fitted four different models per trait: two statistical mod- Lamanna, Violle, & Enquist, 2014). Note that this test does not nec- els [generalized linear model (GLM) and generalized additive model essarily imply that predicted trait combinations do not or cannot (GAM)] and two machine learning models [random forest (RF) and exist, but merely that they did not occur in the input data and should boosted regression trees (BRT); for full details on model parameteriza- therefore be interpreted with caution. tion, see Supporting Information Appendix S6]. We validated all mod- To identify the applicability domain, we calculated and mapped els by running a 10-fold cross-validation using a split-sample procedure the multivariate environmental similarity surface (Elith, Kearney, (80%–20%) and evaluated the predictive performance (cross-validated & Phillips, 2010). This analysis quantifies, per grid cell, the differ- pseudo-R2) of the models by regressing the predicted and observed ence of the most extrapolated environmental predictor and the trait values from all repetitions of the cross-validation. environmental range of that predictor covered by locations in the To quantify the relative importance of each predictor in a consis- plant trait dataset, while considering the distribution of these data tent way across the models, we predicted traits using permuted val- within the global environmental range. Finally, to quantify model ues for the predictor of concern, correlated those predictions with uncertainty (i.e., the variability in predictions across models), the predictions of the model using the original data and quantified rela- coefficient of variation was calculated as the difference in the pre- tive variable importance as one minus the Spearman rank correlation dictions of each of the four individual models compared with the coefficient (Thuiller, Lafourcade, Engler, & Araújo, 2009). ensemble prediction weighted by the predictive performance of each of the models (further explanation in Supporting Information Appendix S8). 2.6 | Testing sensitivity to data quality To test the influence of data quality on the performance of the trait 3 | R E S U LT S models, we created three alternative datasets differing in terms of trait values (i.e., including intraspecific trait variation versus using species- 3.1 | Predicted global plant trait variation specific trait values) and in terms of species representativeness (i.e., a representative sample of species versus the random selection of one The variation explained by the environmental variables varied among species in the plant community; Supporting Information, Figure S7). traits (Figure 1). Model predictive performance was highest for For each alternative dataset, we fitted the four different models for plant height, followed by wood density, SLA and LNC (Figure 1). The BOONMAN et al. | 1039 F I G U R E 1 Model performance for different traits and different models presented as the cross-validated R 2 values. Per trait, the bars represent the predictive performance of GLM, GAM, RF, and BRT from left to right. The horizontal black lines represent the accuracy of the ensemble predictions. Abbreviations: LNC = leaf nitrogen concentration; SLA = specific leaf area [Colour figure can be viewed at wileyonlinelibrary.com] projected global trait maps revealed low SLA in arid areas and high SLA decreased overall with soil CEC (Table 1; Figure 4). Against our ex- in temperate climates (Figure 2a), whereas LNC was found to be low in pectations, wood density showed a flat response to PrecSeas and dry tropical regions (Figure 2c). We also found lower community plant soil pH (Table 1; Figure 4). Furthermore, we found a flat response height at higher latitudes, with the tallest vegetation in wet tropical of wood density to PrecDryQ (Figure 4). areas and shorter vegetation in temperate climates (Figure 2e). Height values, like the other predicted traits, represent community means and thus cannot be transferred directly to a specific species or a class 3.2.2 | Combinations of predicted traits of species. Finally, wood density, representing the potential average wood density for all woody vegetation in a location, was predicted to The predicted trait values largely preserved the global trait correla- be particularly low in areas with low PrecSeas (Figure 2g). tions present in the training data (r = .83). Furthermore, for 99.9% of all grid cells, the predicted trait combination of community mean traits was found within the existing trait combinations found in individual 3.2 | Ecological assessment plants or species. Locations with unrealistic predicted trait combina- tions are presented in the Supporting Information (Figure S9). 3.2.1 | Trait–environment relationships In our models, community mean SLA was mostly explained by 3.3 | Uncertainty assessment HumInd and Tmin (Figure 3a). As expected, SLA showed a uni- modal albeit mostly decreasing response to Tmin, an increase with 3.3.1 | Data quality HumInd and a flat response to PrecSeas. However, we did not find the expected response of SLA to soil CEC and soil pH (Table 1; The predictive accuracy decreased by 11% on average across all Figure 4). We also found a flat response of SLA to PrecDryQ traits and models when intraspecific trait variation was excluded (Figure 4). Community mean LNC was mostly explained by Tmin (Figure 5; Supporting Information Table S7). When species were (Figure 3b). We expected LNC to decrease with Tmin, to increase sampled randomly (i.e., an unrepresentative sample), the predictive with HumInd and to show a flat response to PrecSeas, but we accuracy of the models decreased on average by 19% (Figure 5). The found a unimodal response of LNC with Tmin around 0 °C, and combination of ignoring intraspecific trait variation and using a non- a decrease of LNC with HumInd and PrecSeas (Table 1; Figure 4). representative species sample amplified the reduction in accuracy to We also found LNC to increase with PrecDryQ and, as expected, 28% compared with the default models (Figure 5). to show a flat response to soil CEC and soil pH (Table 1; Figure 4). Community mean plant height was mostly explained by PrecDryQ (Figure 3c). As expected, height increased with increasing Tmin 3.3.2 | Applicability domain and PrecDryQ, whereas in contrast to our original expectation, we found height to decrease with HumInd and to increase with Despite data paucity in large areas of the world (e.g., India, Asian PrecSeas (Table 1; Figure 4). Height showed a flat response to soil Russia and Africa; Figure 6), the trait data covered a large part of the CEC and a unimodal response to soil pH. Community mean wood global environmental space (Figure 6; Supporting Information Figure density was mostly explained by Tmin (Figure 3d). As expected, S10). However, deserts, tropical islands and some parts of the Arctic wood density increased with Tmin, decreased with HumInd and were outside the environmental domain covered by the trait data. 1040 | BOONMAN et al. (a) (b) glm 0.20 gam0.21 ensemble rf 0.31 0.36 brt 0.25 (mm2.mg-1) (c) (d) glm 0.01 gam0.00 ensemble rf 0.25 0.24 brt 0.19 (mg.g-1) LNC  (e) (f) glm 0.67 gam0.60 ensemble rf 0.55 0.68 brt 0.47 (m) (g) (h) glm 0.27 gam0.28 ensemble rf 0.37 0.38 brt 0.34 (mg.mm-3)  F I G U R E 2 Global plant trait predictions. The left column presents the ensemble trait predictions, and the right column presents the ensemble coefficient of variation (CV) of different model predictions. The grey margin in each plot depicts the mean latitudinal value, and the horizontal dashed line indicates the Equator. Values in plots (a), (c), (e), (g) indicate the accuracy of the ensemble predictions (cross-validated R2). Values in plots (b), (d), (f), (h) indicate the predictive performance (cross-validated R2) of the different models, where glm indicates values of the GLMs, gam of GAMs, rf of RFs, and brt of BRT. These values indicate how well traits are predicted, while the CV maps only identify locations where predictions are divergent between models with high and/or similar predictive power. For extrapolation areas, see Figure 6. No predictions were made for Greenland (no soil data) , nor for the south of Egypt (no precipitation seasonality). All predictions represent potential trait values, as not all terrestrial world is vegetated or includes woody vegetation. Abbreviations: LNC = leaf nitrogen concentration; SLA = specific leaf area [Correction added on 16 April 2020, after first online publication: Figure 2 was previously incorrect and has been updated in this version.] [Colour figure can be viewed at wileyonlinelibrary.com] 3.3.3 | Model selection wood density, and GLM for height (Figure 1). The ensemble pre- dictions were always equal to or better than the individual mod- Predictive performance differed between models per trait, where els (Figure 1). The consistency in the predictions among the four RF showed the highest predictive performance for SLA, LNC and modelling techniques varied geographically (Figure 2b,d,f,h). The BOONMAN et al. | 1041 (a) SLA (b) LNC applicability domain for trait predictions, that including intraspecific trait variation and having a representative species sample improved Tmin the predictability of traits at the resolution and extent applied here, HumInd that different modelling techniques demonstrated different predic- PrecDryQ tive performances for different traits and that the ensemble ap- proach was better at predicting traits. Overall, these findings suggest PrecSeas that global predictions aimed at describing broad geographical plant CEC community mean trait patterns are reliable with the data currently pH available, but caution is needed for certain traits and areas. (c) Height (d) Wood density 4.1 | Global trait patterns Tmin The variance explained by our models predicting global community HumInd mean trait values along environmental gradients is comparable with PrecDryQ other global or large-scale trait-based studies (Butler et al., 2017; PrecSeas Madani et al., 2018; Van Bodegom et al., 2014; Yang et al., 2016). For some areas, predicted spatial patterns were consistent with CEC those predicted by previous studies. For example, we found lower pH SLA values in central Australia compared with the coast (Figure 2a), 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 demonstrating the adaptation of leaves to drier conditions (Butler et al., 2017; Madani et al., 2018; Moreno-Martínez et al., 2018; Van F I G U R E 3 Relative predictor importance, calculated as 1— Bodegom et al., 2014). Likewise, we found lower SLA values in the Spearman rank correlation coefficient (see Section 2.5), where high values indicate high importance. The bars represent the Amazon compared with the South American Cerrado (Figure 2a; ensemble relative predictor importance and the error bars show Butler et al., 2017; Moreno-Martínez et al., 2018). Furthermore, the ensemble coefficient of variation. Abbreviations: LNC = leaf we predicted vegetation in tropical climates to be taller than in nitrogen concentration; SLA = specific leaf area; Tmin = minimum temperate climates (Figure 2; Madani et al., 2018). temperature; HumInd = humidity index; PrecDryQ = precipitation On the contrary, our SLA, LNC and wood density predictions dif- in the driest quarter of the year; PrecSeas = precipitation seasonality and CEC for soil cation exchange capacity [Colour fered from the predictions of previous studies for some areas. For figure can be viewed at wileyonlinelibrary.com] example, we predicted higher SLA values for boreal forests com- pared with desert areas, similar to Van Bodegom et al. (2014) but coefficient of variation of SLA was mostly low, apart from some opposite to Butler et al. (2017). Additionally, predicted patterns for areas in Africa. LNC had an overall low coefficient of variation, ex- LNC and wood density did not match other maps (Figure 2; Butler cept for the Sahara. Plant height showed a shifting pattern of high et al., 2017; Moreno-Martínez et al., 2018; Van Bodegom et al., and low coefficients of variation over the globe. Wood density had 2014), but, given the poor predictive accuracy of the LNC models, an low coefficient of variation overall but was predicted with more we consider our LNC predictions unreliable (Figure 1). uncertainty in parts of Africa and most of the Arctic. Whenever trait predictions among different studies disagree, it is difficult to conclude which predictions are more reliable. One option is to validate global trait predictions against regional trait maps. For ex- 4 | D I S CU S S I O N ample, we found wood density to be higher in the east of the Amazon region compared with the north-west and south-west (Figure 2g), as In this study, we established global plant trait models based on locally was found by Baker et al. (2004). However, the paucity of regional measured trait data representative of natural vegetation and used trait maps makes this validation method impractical. Another option these models to make global trait predictions. We evaluated these is to consider model predictive performances and the applicability do- trait predictions based on the predictive performance of the models, main to infer reliability of predictions. Unfortunately, the lack of quan- their ecological realism, the influence of the quality of input data, tification and indication of these in previous studies prevents us from the applicability domain of the predictions and model choice uncer- drawing any conclusions about which predictions are more accurate. tainty. Our results showed that the models predicting community mean SLA, plant height and wood density were well able to explain trait variation. Although the trait–environment relationships were as 4.2 | Ecological evaluation of global trait predictions expected only in part, our trait predictions preserved the observed among-trait correlations and combinations. Furthermore, we showed In general, we found community mean SLA, plant height and that most of the global environmental variation was within the wood density to vary more with climatic factors than with soil 1042 | BOONMAN et al. F I G U R E 4 Ensemble partial trait responses. Shaded areas are the ensemble coefficients of variation. Predictors are minimum temperature (Tmin; °C), humidity index (HumInd; unitless), precipitation in the driest quarter of the year (PrecDryQ; mm), precipitation seasonality (PrecSeas; %), soil cation exchange capacity (CEC; cmol+ kg-1), and soil pH. Extrapolated trait responses are shown with a dashed line. The symbols at the bottom left of each panel represent the expectation reported in Table 1: ↑ = positive relationship; ↓ = negative relationship; ‒ = flat relationship; ˄ = unimodal response; ᴑ = the relationship has not been discussed at the community level in literature. Other Abbreviations: LNC = leaf nitrogen concentration; SLA = specific leaf area [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 5 Performances of models based on datasets of different quality. The bars represent the average cross-validated R 2 values over the four different models, where the errorbars indicate the minimum and maximum cross-validated R 2 value. Datasets including and excluding intraspecific trait variation are indicated with intra and inter, respectively. Species representativeness is indicated with the shade of the bar, where darker shades represent datasets with a species sample representative of the community and lighter bars represent datasets with a non-representative species sample (Appendix S7). Thus, per trait, the bars from left to right represent plot 1, 2, 3, and 4 of Appendix Figure S7. Abbreviations: LNC = leaf nitrogen concentration; SLA = specific leaf area [Colour figure can be viewed at wileyonlinelibrary.com] characteristics (Figure 3). This might reflect the greater variation in Our results confirmed most of the expected relationships between soil characteristics at smaller scales compared with climatic variables SLA and the environmental variables (Table 1; Figure 4). However, we (Supporting Information Figure S11). Furthermore, all predicted trait found SLA to decrease with soil CEC and to show a flat response to soil combinations and correlations were realistic. pH, whereas we expected SLA to increase with soil fertility because BOONMAN et al. | 1043 F I G U R E 6 Locations of trait observations (black dots) and the environmental coverage of trait observations, that is, the multivariate environmental similarity surface (Elith et al., 2010), where blue represents interpolation and red represents extrapolation. More intense shades indicate greater similarities (blue) or differences (red) in environmental conditions of the location compared to the environmental conditions covered by the training dataset. Though values indicate percentages, each griddcell might be restrained by a different environmental variable, which is why the actual values are difficult to interpret. Abbreviations: LNC = leaf nitrogen concentration; SLA = specific leaf area [Colour figure can be viewed at wileyonlinelibrary.com] less durable structures are thought to be maintained with higher soil Figure 4) can be explained by the fact that many species in areas fertility (Table 1; Figure 4). The flat response of SLA with soil pH might with seasonal rainfall can shed leaves, which makes the deciduous be explained by a high nutrient turnover rate in areas of low soil fer- vegetation less vulnerable to cavitation compared with evergreen tility (Vitousek & Sanford, 1986), although it might also be simply that vegetation, allowing it to grow taller than expected. Furthermore, SLA does not respond to changes in soil pH (Firn et al., 2019). taller plants are better able to reach soil water reserves owing to The expected responses of LNC to the environmental variables deeper root systems (Borchert, 1998; Canadell et al., 1996). were not found (Table 1; Figure 4). The partial responses of LNC Our results confirmed the expected relationships between wood showed great variation over small environmental ranges, possi- density and Tmin, HumInd and soil CEC (Table 1; Figure 4). Both the bly owing to the tendency of machine learning models to over-fit. increase in wood density with increasing Tmin and the unexpected Together with the low predictive performance of the models, this flat response with PrecSeas might reflect that colder areas with less might reflect that LNC responds primarily to small-scale environmen- seasonal precipitation are dominated by soft-wooded gymnosperms, tal variation, whereas our models make predictions at a coarser resolu- whereas warmer areas with higher seasonal precipitation are dom- tion (55 km × 55 km). Our results thus support the deviating responses inated by angiosperms with generally denser wood (Swenson & of LNC to environmental variables at the species level (Maire et al., Enquist, 2007), although xylem vulnerability for frost-induced cav- 2015; Ordoñez et al., 2009; Reich & Oleksyn, 2004; Reich et al., 1996). itation decreases with wood density (Reich, 2014). We expected a Additionally, LNC might be highly variable in relationship to multiple decrease in wood density with soil CEC and soil pH because higher environmental factors, leading to non-universal adaptations to the soil fertility is expected to sustain higher growth rates. Although the predictors in our model (Bloomfield et al., 2018; Reich & Oleksyn, overall trend confirmed our expectation, the response to pH was 2004). This variability makes it difficult to interpret LNC responses to very limited, and at the lower end of the CEC gradient, wood den- large-scale environmental gradients biologically, especially given that sity showed a small increase. This might indicate that extreme lim- they will vary depending on whether intraspecific variation is consid- itation to accessible soil nutrients cannot sustain high wood density, ered or not (Albert et al., 2010). We conclude that it is not possible to whereas low fertility but not nutrient-limited environments promote predict LNC distributions reliably at this extent and resolution. slow growth, resulting in high wood density (Table 1). Given that no Our results confirmed the expected relationships between aver- global trait–environment relationship has been described for wood age community height and Tmin and PrecDryQ (Table 1; Figure 4). density (Moles, 2018), these patterns should be investigated further. However, we found mean plant height to decrease with HumInd Literature reports that some relationships between traits and en- (Table 1; Figure 4). This might indicate that shorter vegetation (e.g., vironmental variables may vary with leaf habit, plant growth forms grasses, herbs and shrubs) is more abundant in more humid environ- and photosynthetic pathway at the local scale (e.g., Šímová et al., ments (i.e., higher annual rainfall and/or lower potential evapotrans- 2018). Additionally, trait–environment relationships are affected by piration). The unexpected increase in height with PrecSeas (Table 1; factors other than climate and soil properties, such as land-use type 1044 | BOONMAN et al. and disturbance (e.g., Chen et al., 2018). However, we expect that 4.3.3 | Model selection these factors are of limited relevance at the biological scale (com- munities instead of species), spatial resolution and extent consid- The selection of a specific modelling technique greatly affected the ered, where environmental filtering effects are expected to be much ability to predict traits, and no single “best” model could be identified less confounded by biotic interactions and fine-grain disturbances (Figure 1). Nevertheless, the ensemble forecasting approach as used (Pearson & Dawson, 2003). here is equal to or better at predicting each trait than any individual model. Additionally, this ensemble approach can be used to retrieve a single prediction relying mostly on the best-performing models, while 4.3 | Evaluation of uncertainty considering the variation in predictions between models with similar support. Furthermore, large variation in values predicted by different 4.3.1 | Data quality models, such as for plant height in the Sahara and the North of India, can be used as an indication of uncertainty. Thus, ensemble forecast- We found that including intraspecific trait variation contributed to ing is not only better in predicting traits compared with single mod- improve the predictability of traits (Figure 5; Supporting Information els, it also reduces the uncertainty attributable to subjective model Table S7). This improvement in trait predictions at the global scale technique selection and it enables the quantification and mapping of highlights the importance of considering intraspecific trait variation the uncertainty attributable to divergent model predictions. Finally, over and above the conclusion from small-scale studies that intraspe- it allows for one general modelling approach for multiple traits. cific trait variation contributes greatly to community trait variation (Albert et al., 2010; Bloomfield et al., 2018; Poorter, Castilho, Schietti, Oliveira, & Costa, 2018; Siefert et al., 2015). This indicates that wide- 4.4 | Reliability of global plant trait predictions spread species are likely to show adaptability in their traits, changing them in order to optimize performance for different environments. Our results suggest that plant community traits can be predicted Additionally, our results emphasize the need to build community trait reliably at the global scale when using an ensemble approach with models on a representative sample of species (Figure 5; Supporting high-quality data (i.e., including intraspecific trait variation and Information Table S7). Moreover, high-quality data not only improve a representative species sample). We show that intraspecific trait models statistically, but also theoretically lead to different results variation and the representativeness of species considered in a com- (Poorter et al., 2018). Thus, the inclusion of intraspecific trait varia- munity are important factors to consider, even at the global scale, tion of species representative of the local vegetation should be pre- and that an ensemble forecasting approach helps to deal with and ferred when the aim is to predict plant community mean trait values. quantify multiple types of uncertainty. Based on these results, we recommend the systematic and careful selection of data and model- ling techniques for trait–environment models, and more routine as- 4.3.2 | Applicability domain sessment of their reliability based on model predictive performance, applicability domain, model uncertainty and realism of predicted The strict selection criteria we set greatly reduced the amount of trait combinations. Such systematic presentation of validation re- available community trait data (Supporting Information Table S2). sults and applicability domain in studies presenting predictions of Nevertheless, our dataset covered the major part of the global ter- spatial patterns of community mean traits will enhance our ability restrial environmental space, indicating a wide applicability domain to build upon previous modelling attempts and improve our under- of our models (Figure 6). However, trait predictions should be inter- standing of trait–environment relationships. preted carefully for deserts, the Arctic and tropical islands because Our approach also led to new insights, such as the unexpected of the high variation in predictions between models, and for moun- increase of community height with the seasonality of precipitation, tainous areas because of the high environmental variation within a but the lack of proper model assessment by previous studies limits grid cell. Furthermore, predictions for wood density are extrapo- our ability to draw any objective conclusions on the observed dif- lated to a larger extent in comparison to other traits. A reason for the ferences in trait responses. We suggest that higher predictive accu- fewer community mean data points for wood density is that species racy can be achieved for traits that respond primarily to large-scale mean values are generally considered appropriate because interspe- environmental factors, such as specific leaf area, whereas predictive cific variation in wood density is larger than the intraspecific varia- accuracy at this extent and resolution would be lower for traits such tion; therefore, new wood density data are rarely collected (Siefert as LNC that might respond primarily to small-scale environmental et al., 2015). Additionally, most wood density measures available variation. Nevertheless, the results of our trait models can improve were collected before 1980, meaning that taking global species aver- the mechanistic understanding of global plant trait–environment ages of wood density does not consider any evolutionary responses relationships and contribute to answering the central question in to changes in the local environment. Finally, wood density is rarely functional ecology of the predictability of traits by the environ- reported because of the challenges of measuring it in small shrubs ment. Our assessment opens a new avenue to test impacts of global and low vegetation (Perez-Harguindeguy et al., 2013). change on trait distributions and, eventually, on plant communities. BOONMAN et al. | 1045 Ultimately, the results of our global trait models are contingent on REFERENCES the quality and distributions of the trait data. Modelling efforts will Ackerly, D. D., & Cornwell, W. K. (2007). A trait-based approach to com- be enriched greatly by improving data collection or data availability munity assembly: Partitioning of species trait values into within- and among-community components. Ecology Letters, 10, 135–145. in areas where no or few data are available and for locations char- https://doi.org/10.1111/j.1461-0248.2006.01006.x acterized by unique environmental conditions. Albert, C. H., Thuiller, W., Yoccoz, N. G., Soudant, A., Boucher, F., Saccone, P., & Lavorel, S. (2010). Intraspecific functional variability: AC K N OW L E D G M E N T S Extent, structure and sources of variation. Journal of Ecology, 98, 604–613. https://doi.org/10.1111/j.1365-2745.2010.01651.x C.C.F.B., M.A.J.H. and L.S. were supported by the European Research Baker, T. R., Phillips, O. L., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore, A., Counsil (ERC project CoG SIZE 647224). A.B.-L. was supported by a … Martínez, R. V. (2004). Variation in wood density determines spa- Juan de la Cierva-Incorporación grant (IJCI-2017-31419) from the tial patterns in Amazonian forest biomass. Global Change Biology, 10, Spanish Ministry of Science, Innovation and Universities. J.P. was 545–562. https://doi.org/10.1111/j.1529-8817.2003.00751.x Bjorkman, A. D., Myers-Smith, I. H., Elmendorf, S. C., Normand, S., Rüger, funded by the European Research Council (ERC Synergy grant ERC- N., Beck, P. S. A., … Weiher, E. (2018). Plant functional trait change SyG-2013-610028 IMBALANCE-P), P.H. was funded by Conselho across a warming tundra biome. Nature, 562, 57–62. https://doi. Nacional de Densenvolvimento Científico e Tecnológico (CNPq org/10.1038/s4158​6-018-0563-7 grant no. 309617/2016-2), N.A.S. was supported by the Vidi grant Blonder, B., Lamanna, C., Violle, C., & Enquist, B. J. (2014). The n-dimen- 016.161.318 by The Netherlands Organization for Scientific research sional hypervolume. Global Ecology and Biogeography, 23, 595–609. https://doi.org/10.1111/geb.12146 (NWO), and V.G.O. thanks the Russian Science Foundation (RSF # 19- Bloomfield, K. J., Cernusak, L. A., Eamus, D., Ellsworth, D. S., Prentice, 14-00038) for financial support. We thank Enio E. Sosinksi (Conselho I. C., Wright, I. J., … Atkin, O. K. (2018). A continental-scale assess- Nacional de Densenvolvimento Científico e Tecnológico, São Paulo ment of variability in leaf traits: Within species, across sites and Research Foundation), for his data contribution. The study has between seasons. Functional Ecology, 32, 1492–1506. https://doi. org/10.1111/1365-2435.13097 been supported by the TRY initiative on plant traits (http://www. Borchert, R. (1998). Responses of tropical trees to rainfall seasonality try-db.org). The TRY initiative and database are hosted, developed and its long-term changes. In A. Markham (Ed.), Potential impacts of and maintained by J. Kattge and G. Bönisch (Max Planck Institute climate change on tropical forest ecosystems (pp. 241–253). Dordrecht, for Biogeochemistry, Jena, Germany). TRY is currently supported by The Netherlands: Springer. Butler, E. E., Datta, A., Flores-Moreno, H., Chen, M., Wythers, K. R., DIVERSITAS/Future Earth and the German Centre for Integrative Fazayeli, F., … Reich, P. B. (2017). Mapping local and global variabil- Biodiversity Research (iDiv) Halle–Jena–Leipzig. ity in plant trait distributions. Proceedings of the National Academy of Sciences of the United States of America, 114, E10937–E10946. DATA AVA I L A B I L I T Y S TAT E M E N T https://doi.org/10.1073/pnas.17089​8 4114 Canadell, J., Jackson, R., Ehleringer, J., Mooney, H. A., Sala, O. E., & Community mean trait data used in this paper and the predicted trait Schulze, E.-D. (1996). Maximum rooting depth of vegetation types at distributions maps generated in this study are provided via FigShare, the global scale. Oecologia, 108, 583–595. https://doi.org/10.1007/ as are all R scripts regarding data analyses (http://doi.org/10.6084/ BF003​29030 m9.figsh​are.11559852). Owing to regulations of the major trait data- Chave, J., Coomes, D., Jansen, S., Lewis, S. L., Swenson, N. G., & Zanne, A. E. (2009). Towards a worldwide wood eco- base used, the raw data cannot be shared openly. nomics spectrum. Ecology Letters, 12, 351–366. https://doi. org/10.1111/j.1461-0248.2009.01285.x ORCID Chen, S., Wang, W., Xu, W., Wang, Y., Wan, H., Chen, D., … Bai, Y. (2018). Coline C. F. Boonman https://orcid.org/0000-0003-2417-1579 Plant diversity enhances productivity and soil carbon storage. Ana Benítez-López https://orcid.org/0000-0002-6432-1837 Proceedings of the National Academy of Sciences USA, 115, 4027– 4032. https://doi.org/10.1073/pnas.17002​98114 Aafke M. Schipper https://orcid.org/0000-0002-5667-0893 Díaz, S., Kattge, J., Cornelissen, J. H. C., Wright, I. J., Lavorel, S., Dray, S., Wilfried Thuiller https://orcid.org/0000-0002-5388-5274 … Gorné, L. D. (2016). The global spectrum of plant form and func- Madhur Anand https://orcid.org/0000-0002-2425-4502 tion. Nature, 529, 167–171. https://doi.org/10.1038/natur​e16489 Bruno E. L. Cerabolini https://orcid.org/0000-0002-3793-0733 Elith, J., Kearney, M., & Phillips, S. (2010). The art of modelling range-shift- ing species. Methods in Ecology and Evolution, 1, 330–342. https://doi. Johannes H. C. Cornelissen https://orcid. org/10.1111/j.2041-210X.2010.00036.x org/0000-0002-2346-1585 Firn, J., McGree, J. M., Harvey, E., Flores-Moreno, H., Schütz, M., Andres Gonzalez-Melo https://orcid.org/0000-0003-0069-456X Buckley, Y. M., … Risch, A. C. (2019). Leaf nutrients, not specific leaf Wesley N. Hattingh https://orcid.org/0000-0002-3626-5137 area, are consistent indicators of elevated nutrient inputs. Nature Ecology and Evolution, 3, 400–406. https://doi.org/10.1038/s4155​ Pedro Higuchi https://orcid.org/0000-0002-3855-555X 9-018-0790-1 Daniel C. Laughlin https://orcid.org/0000-0002-9651-5732 Franklin, J., Serra-Diaz, J. M., Syphard, A. D., & Regan, H. M. (2016). Vladimir G. Onipchenko https://orcid.org/0000-0002-1626-1171 Global change and terrestrial plant community dynamics. Proceedings Josep Peñuelas https://orcid.org/0000-0002-7215-0150 of the National Academy of Sciences USA, 113, 3725–3734. https://doi. org/10.1073/pnas.15199​11113 Lourens Poorter https://orcid.org/0000-0003-1391-4875 Grime, J. P. (1998). Benefits of plant diversity to ecosystems: Immediate, Nadejda A. Soudzilovskaia https://orcid.org/0000-0002-9584-2109 filter and founder effects. Journal of Ecology, 86, 902–910. Mark A. J. Huijbregts https://orcid.org/0000-0002-7037-680X Hengl, T., De Jesus, J. M., Heuvelink, G. B. M., Ruiperez Gonzalez, M., Luca Santini https://orcid.org/0000-0002-5418-3688 Kilibarda, M., Blagoti, A., …Kempen, B. (2017). SoilGrids250m: Global 1046 | BOONMAN et al. gridded soil information based on machine learning. PLoS ONE, 12, effects on plant community composition trajectories depend upon e0169748. https://doi.org/10.1371/journ​al.pone.0169748 management legacies. Global Change Biology, 24, 1722–1740. https:// Karger, D. N., Conrad, O., Böhner, J., Kawohl, T., Kreft, H., Soria-Auza, doi.org/10.1111/ijlh.12426 R. W., … Kessler, M. (2017). Climatologies at high resolution for the Poorter, L., Castilho, C. V., Schietti, J., Oliveira, R. S., & Costa, F. R. C. earth's land surface areas. Scientific Data, 4, 170122. https://doi. (2018). Can traits predict individual growth performance? A test in a org/10.1038/sdata.2017.122 hyperdiverse tropical forest. New Phytologist, 219, 109–121. https:// Kattge, J. (2019). TRY plant trait database. Retrieved from www.try-db. doi.org/10.1111/nph.15206 org/ Poorter, L., van der Sande, M. T., Arets, E. J. M. M., Ascarrunz, N., Kattge, J., Díaz, S., Lavorel, S., Prentice, I. C., Leadley, P., Bönisch, Enquist, B. J., Finegan, B., … Peña-Claros, M. (2017). Biodiversity G., … Wirth, C. (2011). TRY – A global database of plant and climate determine the functioning of Neotropical forests. traits. Global Change Biology, 17, 2905–2935. https://doi. Global Ecology and Biogeography, 26, 1423–1434. https://doi. org/10.1111/j.1365-2486.2011.02451.x org/10.1111/geb.12668 Keddy, P. A. (1992). A pragmatic approach to functional ecology. Purschke, O., Schmid, B. C., Sykes, M. T., Poschlod, P., Michalski, S. G., Functional Ecology, 6, 621–626. Durka, W., … Prentice, H. C. (2013). Contrasting changes in taxo- Kimball, S., Gremer, J. R., Angert, A. L., Huxman, T. E., & Venable, D. L. nomic, phylogenetic and functional diversity during a long-term (2012). Fitness and physiology in a variable environment. Oecologia, succession: Insights into assembly processes. Journal of Ecology, 101, 169, 319–329. 857–866. https://doi.org/10.1111/1365-2745.12098 Larjavaara, M., & Muller-Landau, H. C. (2010). Rethinking the value of Reich, P. B. (2014). The world-wide ‘fast–slow’ plant economics spec- high wood density. Functional Ecology, 24, 701–705. https://doi. trum: A traits manifesto. Journal of Ecology, 102, 275–301. https:// org/10.1111/j.1365-2435.2010.01698.x doi.org/10.1111/1365-2745.12211 Lavorel, S., & Garnier, E. (2002). Predicting changes in community com- Reich, P. B., & Oleksyn, J. (2004). Global patterns of plant leaf N and P position and ecosystem functioning from plant traits: Revisiting in relation to temperature and latitude. Proceedings of the National the Holy Grail. Functional Ecology, 16, 545–556. https://doi. Academy of Sciences of the United States of Ameica, 101, 11001–11006. org/10.1046/j.1365-2435.2002.00664.x https://doi.org/10.1073/pnas.04035​88101 Madani, N., Kimball, J. S., Ballantyne, A. P., Affleck, D. L. R., van Bodegom, Reich, P. B., Oleksyn, J., & Tjoelker, M. G. (1996). Needle respiration and P. M., Reich, P. B., … Running, S. W. (2018). Future global productivity nitrogen concentration in Scots pine populations from a broad latitu- will be affected by plant trait response to climate. Scientific Reports, dinal range: A common garden test with field-grown trees. Functional 8, 2870. https://doi.org/10.1038/s4159​8-018-21172​-9 Ecology, 10, 768–776. Maire, V., Wright, I. J., Prentice, I. C., Batjes, N. H., Bhaskar, R., Ross, D. S., & Ketterings, Q. (1995). Recommended methods for de- van Bodegom, P. M., … Santiago, L. S. (2015). Global effects of termining soil cation exchange capacity. Recommended Soil Testing soil and climate on leaf photosynthetic traits and rates. Global Procedures for the Northeastern United States, 493, 62–69. Ecology and Biogeography, 24, 706–717. https://doi.org/10.1111/ Siefert, A., Violle, C., Chalmandrier, L., Albert, C. H., Taudiere, A., geb.12296 Fajardo, A., … Wardle, D. A. (2015). A global meta-analysis of the Marmion, M., Parviainen, M., Luoto, M., Heikkinen, R. K., & Thuiller, W. relative extent of intraspecific trait variation in plant communi- (2009). Evaluation of consensus methods in predictive species distri- ties. Ecology Letters, 18, 1406–1419. https://doi.org/10.1111/ bution modelling. Diversity and Distributions, 15, 59–69. https://doi. ele.12508 org/10.1111/j.1472-4642.2008.00491.x Šímová, I., Violle, C., Svenning, J.-C., Kattge, J., Engemann, K., Sandel, B., Moles, A. T. (2018). Being John Harper: Using evolutionary ideas to … Enquist, B. J. (2018). Spatial patterns and climate relationships of improve understanding of global patterns in plant traits. Journal of major plant traits in the New World differ between woody and her- Ecology, 106, 1–18. https://doi.org/10.1111/1365-2745.12887 baceous species. Journal of Biogeography, 45, 895–916. https://doi. Moles, A. T., Warton, D. I., Warman, L., Swenson, N. G., Laffan, org/10.1111/jbi.13171 S. W., Zanne, A. E., … Leishman, M. R. (2009). Global patterns Stoklosa, J., Daly, C., Foster, S. D., Ashcroft, M. B., & Warton, D. I. (2015). in plant height. Journal of Ecology, 97, 923–932. https://doi. A climate of uncertainty: Accounting for error in climate variables org/10.1111/j.1365-2745.2009.01526.x for species distribution models. Methods in Ecology and Evolution, 6, Moreno-Martínez, Á., Camps-Valls, G., Kattge, J., Robinson, N., 412–423. https://doi.org/10.1111/2041-210X.12217 Reichstein, M., van Bodegom, P., … Running, S. W. (2018). A meth- Swenson, N. G., & Enquist, B. J. (2007). Ecological and evolution- odology to derive global maps of leaf traits using remote sensing and ary determinants of a key plant functional trait: Wood density climate data. Remote Sensing of Environment, 218, 69–88. https://doi. and its community-wide variation across latitude and elevation. org/10.1016/J.RSE.2018.09.006 American Journal of Botany, 94, 451–459. https://doi.org/10.3732/ Ordoñez, J. C., Van Bodegom, P. M., Witte, J. M., Wright, I. J., Reich, ajb.94.3.451 P. B., & Aerts, R. (2009). A global study of relationships be- Swenson, N. G., & Weiser, M. D. (2010). Plant geography upon the basis tween leaf traits, climate and soil measures of nutrient fertil- of functional traits: An example from eastern North American trees. ity. Global Ecology and Biogeography, 18, 137–149. https://doi. Ecology, 91, 2234–2241. https://doi.org/10.1890/09-1743.1 org/10.1111/j.1466-8238.2008.00441.x The Plant List. (2013). Version 1.1. Retrieved from www.thepl​antli​st.org Pearson, R. G., & Dawson, T. P. (2003). Predicting the impact of climate Thomas, S. C., & Bazzaz, F. A. (1999). Asymptotic height as a predictor of change on the distribution of species: Are bioclimate envelope mod- photosynthetic characteristics in Malaysian rain forest trees. Ecology, els useful? Global Ecology and Biogeography, 12, 361–371. https://doi. 80, 1607–1622. org/10.1109/CLOUD.2010.16 Thuiller, W., Brotons, L., Araújo, M. B., & Lavorel, S. (2004). Effects of Pérez-Harguindeguy, N., Díaz, S., Garnier, E., Lavorel, S., Poorter, H., restricting environmental range of data to project current and future Jaureguiberry, P., … Cornelissen, J. H. C. (2013). New handbook for species distributions. Ecography, 27(2), 165–172. standardised measurement of plant functional traits worldwide. Thuiller, W., Guéguen, M., Renaud, J., Karger, D. N., & Zimmerman, N. Australian Journal of Botany, 61, 167–234. E. (2019). Uncertainty in ensembles of global biodiveristy scenarios. Perring, M. P., Bernhardt-Römermann, M., Baeten, L., Midolo, G., Blondeel, Nature Communications, 10(1), 1–9. https://doi.org/10.1038/s4146​ H., Depauw, L., … Verheyen, K. (2017). Global environmental change 7-019-09519​-w BOONMAN et al. | 1047 Thuiller, W., Lafourcade, B., Engler, R., & Araújo, M. B. (2009). APPENDIX BIOMOD – A platform for ensemble forecasting of spe- cies distributions. Ecography, 32, 369–373. https://doi. org/10.1111/j.1600-0587.2008.05742.x DATA SOURCES Tyree, M. T., & Sperry, J. S. (1989). Vulnerability of xylem to cavitation Ackerly, D. (2004). Functional strategies of chaparral shrubs in re- and embolism. Annual Review of Plant Biology, 40(1), 19–36. https:// lation to seasonal water deficit and disturbance. Ecological doi.org/10.1146/annur​ev.pp.40.060189.000315 Monographs, 74, 25–44. Van Bodegom, P. M., Douma, J. C., & Verheijen, L. M. (2014). A fully traits-based approach to modeling global vegetation distribution. Adler, P. B., Milchunas, D. G., Lauenroth, W. K., Sala, O. E., & Burke, I. Proceedings of the National Academy of Sciences USA, 111, 13733– C. (2004). Functional traits of graminoids in semi-arid steppes: A 13738. https://doi.org/10.1073/pnas.13045​51110 test of grazing histories. Journal of Applied Ecology, 41, 653–663. Vitousek, P. M., & Sanford, R. L. Jr (1986). Nutrient cycling in moist trop- Baddeley, J. A., Woodin, S. J., & Alexander, I. J. (1994). Effects of ical forest. Annual Review of Ecology and Systematics, 17, 137–167. increased nitrogen and phosphorus availability on the photosyn- Webb, C. T., Hoeting, J. A., Ames, G. M., Pyne, M. I., & LeRoy Poff, N. (2010). A structured and dynamic framework to advance traits- thesis and nutrient relations of three Arctic dwarf Shrubs from based theory and prediction in ecology. Ecology Letters, 13, 267–283. Svalbard. Functional Ecology, 8, 676–685. https://doi.org/10.1111/j.1461-0248.2010.01444.x Baraloto, C., Paine, C. E. T., Poorter, L., Beauchene, J., Bonal, D., Went, F. W. (1953). The effect of temperature on plant growth. Annual Domenach, A. M., … Chave, J. (2010). Decoupled leaf and stem Review of Plant Physiology, 4, 347–362. https://doi.org/10.1146/ annur​ev.pp.04.060153.002023 economics in rain forest trees. Ecology Letters, 13, 1338–1347. Wright, I. J., Reich, P. B., Cornelissen, J. H. C., Falster, D. S., Barrett, R. T. S., Hollister, R. D., Oberbauer, S. F., & Tweedie, C. E. Groom, P. K., Hikosaka, K., … Westoby, M. (2005). Modulation (2015). Arctic plant responses to changing abiotic factors in of leaf economic traits and trait relationships by climate. northern Alaska. American Journal of Botany, 102, 2020–2031. Global Ecology and Biogeography, 14, 411–421. https://doi. org/10.1111/j.1466-822x.2005.00172.x Baruch, Z., & Goldstein, G. (1999). Leaf construction cost, nutrient Wright, I. J., Reich, P. B., Westoby, M., Ackerly, D. D., Baruch, Z., Bongers, concentration, and net CO2 assimilation of native and invasive F., … Villar, R. (2004). The worldwide leaf economics spectrum. species in Hawaii. Oecologia, 121, 183–192. Nature, 428, 821–827. Bassow, S. L., & Bazzaz, F. A. (1997). Intra- and inter-specific vari- Yang, Y., Zhu, Q., Peng, C., Wang, H., Xue, W., Lin, G., … Li, S. (2016). A novel approach for modelling vegetation distributions and ana- ation in canopy photosynthesis in a mixed deciduous forest. lysing vegetation sensitivity through trait-climate relationships in Oecologia, 109, 507–515. China. Scientific Reports, 6, 24110. https://doi.org/10.1038/srep2​ Berner, L. T., Alexander, H. D., Loranty, M. M., Ganzlin, P., Mack, M. 4110 C., Davydov, S. P., & Goetz, S. J. (2015). Biomass allometry for Zomer, R. J., Trabucco, A., Bossio, D. A., & Verchot, L. V. (2008). Climate change mitigation: A spatial analysis of global land suitability for alder, dwarf birch, and willow in boreal forest and tundra ecosys- clean development mechanism afforestation and reforestation. tems of far northeastern Siberia and north-central Alaska. Forest Agriculture, Ecosystems and Environment, 126, 67–80. https://doi. Ecology and Management, 337, 110–118. org/10.1016/j.agee.2008.01.014 Bjorkman, A. D., Elmendorf, S. C., Beamish, A. L., Vellend, M., & Henry, G. H. R. (2015). Contrasting effects of warming and in- B I O S K E TC H creased snowfall on Arctic tundra plant phenology over the past two decades. Global Change Biology, 21, 4651–4661. Coline Boonman is a PhD candidate at the Radboud University Blonder, B., Buzzard, V., Simova, I., Sloat, L., Boyle, B., Lipson, R., in The Netherlands, and her research focuses on local-scale vari- … Enquist, B. J. (2012). The leaf-area shrinkage effect can bias ation and macroecological patterns of plant functional traits. Her paleoclimate and ecology research. American Journal of Botany, main interests are the adaptability of plants to different environ- 99, 1756–1763. ments and the effect of natural and anthropogenic disturbances Blonder, B., Vasseur, F., Violle, C., Shipley, B., Enquist, B. J., & Vile, on vegetation patterns. D. (2015). Testing models for the leaf economics spectrum with leaf and whole-plant traits in Arabidopsis thaliana. AoB PLANTS, 7, plv049. S U P P O R T I N G I N FO R M AT I O N Blonder, B., Violle, C., Bentley, L. P., & Enquist, B. J. (2011). Venation Additional supporting information may be found online in the networks and the origin of the leaf economics spectrum. Ecology Supporting Information section. Letters, 14, 91–100. Blonder, B., Violle, C., & Enquist, B. J. (2013). Assessing the causes and scales of the leaf economics spectrum using venation net- How to cite this article: Boonman CCF, Benítez-López A, works in Populus tremuloides. Journal of Ecology, 101, 981–989. Schipper AM, et al. Assessing the reliability of predicted plant Bloomfield, K. J., Cernusak, L. A., Eamus, D., Ellsworth, D. S., Prentice, trait distributions at the global scale. Global Ecol Biogeogr. I. C., Wright, I. J., … Atkin, O. K. (2018). A continental-scale as- 2020;29:1034–1051. https://doi.org/10.1111/geb.13086 sessment of variability in leaf traits: Within species, across sites and between seasons. Functional Ecology, 32, 1492–1506. 1048 | BOONMAN et al. Bond-Lamberty, B., Wang, C., & Gower, S. T. (2002). Aboveground Craine, J. M., & Towne, E. G. (2010). High leaf tissue density grass- and belowground biomass and sapwood area allometric equa- land species consistently more abundant across topographic and tions for six boreal tree species of northern Manitoba. Canadian disturbance contrasts in a North American tallgrass prairie. Plant Journal of Forest Research, 32, 1441–1450. and Soil, 337, 193–203. Bond-Lamberty, B., Wang, C., Gower, S. T., & Norman, J. (2002). Leaf Craine, J. M., Towne, E. G., Ocheltree, T. W., & Nippert, J. B. (2012). area dynamics of a boreal black spruce fire chronosequence. Tree Community traitscape of foliar nitrogen isotopes reveals N avail- Physiology, 22, 993–1001. ability patterns in a tallgrass prairie. Plant and Soil, 356, 395–403. Bongers, F., & Pompa, J. (1990). Leaf characteristics of the tropical Craven, D., Braden, D., Ashton, M. S., Berlyn, G. P., Wishnie, M., & rain forest flora of Los Tuxtlas, Mexico. Botanical Gazette, 151, Dent, D. (2007). Between and within-site comparisons of struc- 354–365. tural and physiological characteristics and foliar nutrient content Brown, K. A., Johnson, S. E., Parks, K. E., Holmes, S. M., Ivoandry, T., of 14 tree species at a wet, fertile site and a dry, infertile site in Abram, N. K., … Wright, P. C. (2013). Use of provisioning ecosys- Panama. Forest Ecology and Management, 238, 335–346. tem services drives loss of functional traits across land use inten- de Araujo, A. C., Ometto, J. P. H. B., Dolman, A. J., Kruijt, B., sification gradients in tropical forests in Madagascar. Biological Waterloo, M. J., & Ehleringer, J. R. (2012). LBA-ECO CD-02 C Conservation, 161, 118–127. and N isotopes in leaves and atmospheric CO2, Amazonas, Brazil, Butterfield, B. J., & Briggs, J. M. (2011). Regeneration niche differ- Data set., Oak Ridge National Laboratory Distributed Active A. entiates functional strategies of desert woody plant species. Retrieved from http://doi.org/10.3334/ORNLD​A AC/1097 Oecologia, 165, 477–487. Devictor, V., Mouillot, D., Meynard, C., Jiguet, F., Thuiller, W., & Campetella, G., Botta-Dukát, Z., Wellstein, C., Canullo, R., Gatto, Mouquet, N. (2010). Spatial mismatch and congruence between S., Chelli, S., … Bartha, S. (2011). Patterns of plant trait–environ- taxonomic, phylogenetic and functional diversity: The need for ment relationships along a forest succession chronosequence. integrative conservation strategies in a changing world. Ecology Agriculture, Ecosystems and Environment, 145, 38–48. Letters, 13, 1030–1040. Carbognani, M., Petraglia, A., & Tomaselli, M. (2014). Warming ef- Diemer, M. (1998). Leaf lifespans of high-elevation, aseasonal fects and plant trait control on the early-decomposition in alpine Andean shrub species in relation to leaf traits and leaf habit. snowbeds. Plant and Soil, 376, 277–290. Global Ecology and Biogeography Letters, 7, 457–465. Cerabolini, B. E. L., Brusa, G., Ceriani, R. M., de Andreis, R., Luzzaro, Diemer, M. (1998). Life span and dynamics of leaves of herbaceous A., & Pierce, S. (2010). Can CSR classification be generally ap- perennials in high-elevation environments: “News from the el- plied outside Britain? Plant Ecology, 210, 253–261. ephant's leg”. Functional Ecology, 12, 413–425. Chapin, F. S. I., & Shaver, G. R. (1985). Individualistic growth response Diemer, M., Körner, C., & Prock, S. (1992). Leaf life spans in wild per- of tundra plant species to environmental manipulations in the ennial herbaceous plants: A survey and attempts at a functional field. Ecology, 66, 564–576. interpretation. Oecologia, 89, 10–16. Chapin, F. S. I., Shaver, G. R., Giblin, A. E., Nadelhoffer, K. J., & Ellsworth, D. S., & Reich, P. B. (1993). Canopy structure and vertical Laundre, J. A. (1995). Responses of Arctic tundra to experimental patterns of photosynthesis and related leaf traits in a deciduous and observed changes in climate. Ecology, 76, 694–711. forest. Oecologia, 96, 169–178. Chen, Y., Han, W., Tang, L., Tang, Z., & Fang, J. (2013). Leaf nitrogen and Eskelinen, A., Kaarlejärvi, E., & Olofsson, J. (2017). Herbivory and phosphorus concentrations of woody plants differ in responses to nutrient limitation protect warming tundra from lowland spe- climate, soil and plant growth form. Ecography, 36, 178–184. cies' invasion and diversity loss. Global Change Biology, 23, Cornelissen, J. H. C., Cerabolini, B., Castro-Díez, P., Villar-Salvador, 245–255. P., Montserrat-Martí, G., Puyravaud, J. P., … Aerts, R. (2003). Figueira, A. M. E. S., Miller, S. D., de Sousa, C. A. D., Menton, M. C., Functional traits of woody plants: Correspondence of species Maia, A. R., da Rocha, H. R., & Goulden, M. L. (2008). Effects rankings between field adults and laboratory-grown seedlings? of selective logging on tropical forest tree growth. Journal of Journal of Vegetation Science, 14, 311–322. Geophysical Research, 113, G00B05. Cornelissen, J. H. C., Diez, P. C., & Hunt, R. (1996). Seedling growth, Frenette-Dussault, C., Shipley, B., Léger, J. F., Meziane, D., & Hingrat, allocation and leaf attributes in a wide range of woody plant spe- Y. (2012). Functional structure of an arid steppe plant commu- cies and types. Journal of Ecology, 84, 755–765. nity reveals similarities with Grime's C-S-R theory. Journal of Cornelissen, J. H. C., Quested, H. M., Gwynn-Jones, D., Van Vegetation Science, 23, 208–222. Logtestijn, R. S. P., De Beus, M. A. H., Kondratchuk, A., … Aerts, Freschet, G. T., Cornelissen, J. H. C., van Logtestijn, R. S. P., & Aerts, R. (2004). Leaf digestibility and litter decomposability are related R. (2010). Evidence of the “plant economics spectrum” in a sub- in a wide range of subarctic plant species and types. Functional arctic flora. Journal of Ecology, 98, 362–373. Ecology, 18, 779–786. Fyllas, N. M., Patino, S., Baker, T. R., Bielefeld Nardoto, G., Martinelli, Cornwell, W. K., & Ackerly, D. D. (2009). Community assembly and L. A., Quesada, C. A., … Lloyd, J. (2009). Basin-wide variations in shifts in plant trait distributions across an environmental gradi- foliar properties of Amazonian forest: Phylogeny, soils and cli- ent in coastal California. Ecological Monographs, 79, 109–126. mate. Biogeosciences, 6, 2677–2708. BOONMAN et al. | 1049 Garnier, E., Laurent, G., Bellmann, A., Debain, S., Berthelier, P., diversity and community composition. Ecology and Evolution, 6, Ducout, B., … Navas, M.-L. (2001). Consistency of species rank- 8126–8137. ing based on functional leaf traits. New Phytologist, 152, 69–83. Jurik, T. W. (1986). Temporal and spatial patterns of specific leaf Garnier, E., Lavorel, S., Ansquer, P., Castro, H., Cruz, P., Dolezal, J., … weight in successional northern hardwood tree species. American Zarovali, M. P. (2007). Assessing the effects of land-use change Journal of Botany, 73, 1083–1092. on plant traits, communities and ecosystem functioning in grass- Jurik, T. W., Weber, J. A., & Gates, D. M. (1988). Effects of tempera- lands: A standardized methodology and lessons from an applica- ture and light on photosynthesis of dominant species of a north- tion to 11 European sites. Annals of Botany 99, 967–985. ern hardwood forest. Botanical Gazette, 149, 203–208. Gleason, S. M., Butler, D. W., Ziemińska, K., Waryszak, P., & Westoby, Kaarlejärvi, E., Baxter, R., Hofgaard, A., Hytteborn, H., Khitun, O., M. (2012). Stem xylem conductivity is key to plant water balance Molau, U., … Olofsson, J. (2012). Effects of sarming on shrub across Australian angiosperm species. Functional Ecology, 26, abundance and chemistry drive ecosystem-level changes in a 343–352. forest–tundra ecotone. Ecosystems, 15, 1219–1233. Good, M. K., Morgan, J. W., Venn, S., & Green, P. (2019). Timing of Kaarlejärvi, E., Eskelinen, A., & Olofsson, J. (2017). Herbivores res- snowmelt affects species composition via plant strategy filter- cue diversity in warming tundra by modulating trait-dependent ing. Basic and Applied Ecology, 35, 54–62. species losses and gains. Nature Communications, 8, 419. Grau, O., Ninot, J. M., Pérez-Haase, A., & Callaghan, T. V. (2014). Plant Kattge, J., Knorr, W., Raddatz, T., & Wirth, C. (2009). Quantifying co-existence patterns and high-arctic vegetation composition in photosynthetic capacity and its relationship to leaf nitrogen con- three common plant communities in north-east Greenland. Polar tent for global-scale terrestrial biosphere models. Global Change Research, 33, 19235. Biology, 15, 976–991. Gutiérrez, A. G., Armesto, J. J., & Aravena, J. C. (2004). Disturbance Kazakou, E., Vile, D., Shipley, B., Gallets, C., & Garnier, E. (2006). Co- and regeneration dynamics of an old-growth North Patagonian variations in litter decomposition, leaf traits and plant growth in rain forest in Chiloé Island, Chile. Journal of Ecology, 92, 598–608. species from a Mediterranean old-field succession. Functional Gutiérrez, A. G., & Huth, A. (2012). Successional stages of primary Ecology, 20, 21–30. temperate rainforests of Chiloé Island, Chile. Perspectives in Plant Kichenin, E., Wardle, D. A., Peltzer, D. A., Morse, C. W., & Freschet, Ecology, Evolution and Systematics, 14, 243–256. G. T. (2013). Contrasting effects of plant inter- and intraspecific Guy, A. L., Mischkolz, J. M., & Lamb, E. G. (2013). Limited effects variation on community-level trait measures along an environ- of simulated acidic deposition on seedling survivorship and root mental gradient. Functional Ecology, 27, 1254–1261. morphology of endemic plant taxa of the Athabasca Sand Dunes Klady, R. A., Henry, G. H. R., & Lemay, V. (2011). Changes in high in well-watered greenhouse trials. Botany, 91, 176–181. arctic tundra plant reproduction in response to long-term experi- Han, Q., Kawasaki, T., Nakano, T., & Chiba, Y. (2004). Spatial and sea- mental warming. Global Change Biology, 17, 1611–1624. sonal variability of temperature responses of biochemical pho- Kraft, N. J. B., Valencia, R., & Ackerly, D. D. (2008). Functional traits tosynthesis parameters and leaf nitrogen content within a Pinus and niche-based tree community assembly in an Amazonian for- densiflora crown. Tree Physiology, 24, 737–744. est. Science, 322, 580–582. Han, W., Chen, Y., Zhao, F. J., Tang, L., Jiang, R., & Zhang, F. (2012). Kudo, G. (1996). Intraspecific variation of leaf traits in several decid- Floral, climatic and soil pH controls on leaf ash content in China's uous species in relation to length of growing season. Écoscience, terrestrial plants. Global Ecology and Biogeography, 21, 376–382. 3, 483–489. Harper, K. A., Lavallee, A. A., & Dodonov, P. (2018). Patterns of shrub Kudo, G., Molau, U., & Wada, N. (2001). Leaf-trait variation of tun- abundance and relationships with other plant types within the dra plants along a climatic gradient: An integration of responses forest–tundra ecotone in northern Canada. Arctic Science, 4, in evergreen and deciduous species. Arctic, Antarctic, and Alpine 691–709. Research, 33, 181–190. Hikosaka, K., & Hirose, T. (2000). Photosynthetic nitrogen-use ef- Kurokawa, H., & Nakashizuka, T. (2008). Leaf herbivory and de- ficiency in evergreen broad-leaved woody species coexisting in a composability in a Malaysian tropical rain forest. Ecology, 89, warm-temperate forest. Tree Physiology, 20, 1249–1254. 2645–2656. Hikosaka, K., Nagamatsu, D., Ishii, H. S., & Hirose, T. (2002). Laughlin, D. C., Fulé, P. Z., Huffman, D. W., Crouse, J., & Laliberté, Photosynthesis–nitrogen relationships in species at different E. (2011). Climatic constraints on trait-based forest assembly. altitudes on Mount Kinabalu, Malaysia. Ecological Research, 17, Journal of Ecology, 99, 1489–1499. 305–313. Laughlin, D. C., Leppert, J. J., Moore, M. M., & Sieg, C. H. (2010). Hudson, J. M. G., Henry, G. H. R., & Cornwell, W. K. (2011). Taller and A multi-trait test of the leaf-height-seed plant strategy scheme larger: Shifts in Arctic tundra leaf traits after 16 years of experi- with 133 species from a pine forest flora. Functional Ecology, 24, mental warming. Global Change Biology, 17, 1013–1021. 493–501. Iturrate-Garcia, M., O'Brien, M. J., Khitun, O., Abiven, S., Niklaus, Little, C. J., Cutting, H., Alatalo, J., & Cooper, E. J. (2017). Short-term P. A., & Schaepman-Strub, G. (2016). Interactive effects be- herbivory has long-term consequences in warmed and ambient tween plant functional types and soil factors on tundra species high Arctic tundra. Environmental Research Letters, 12, 025001. 1050 | BOONMAN et al. McAllister, C. A., Knapp, A. K., & Maragni, L. A. (1998). Is leaf-level Ordoñez, J. C., van Bodegom, P. M., Witte, J.-P. M., Bartholomeus, R. photosynthesis related to plant success in a highly productive P., van Hal, J. R., & Aerts, R. (2010). Plant strategies in relation to grassland? Oecologia, 117, 40–46. resource supply in mesic to wet environments: Does theory mir- Medlyn, B., Badeck, F.-W., De Pury, D., Barton, C., Broadmeadow, ror nature? The American Naturalist, 175, 225–239. M., Ceulemans, R., … Jarvis, P. (1999). Effects of elevated [CO2] Overbeck, G. E., Müller, S. C., Pillar, V. D., & Pfadenhauer, J. (2005). on photosynthesis in European forest species: A meta-analysis Fine-scale post-fire dynamics in southern Brazilian subtropical of model parameters. Plant Cell and Environment, 22, 1475–1495. grassland. Journal of Vegetation Science, 16, 655–664. Meir, P., Levy, P. E., Grace, J., & Jarvis, P. G. (2007). Photosynthetic Parada, T. J., Jara, C. V., & Lusk, C. H. (2003). Distribución de alturas parameters from two contrasting woody vegetation types in máximas de especies en rodales antiguos de selva Valdiviana, West Africa. Plant Ecology, 192, 277–287. Parque Nacional Puyehue. Bosque, 24, 63–67. Messier, J., McGill, B. J., & Lechowicz, M. J. (2010). How do traits Peco, B., De Pablos, I., Traba, J., & Levassor, C. (2005). The effect vary across ecological scales? A case for trait-based ecology. of grazing abandonment on species composition and functional Ecology Letters, 13, 838–848. traits: The case of dehesa grasslands. Basic and Applied Ecology, Midgley, J. J., van Wyk, G. R., & Everard, D. A. (1995). Leaf attrib- 6, 175–183. utes of South African forest species. African Journal of Ecology, Penuelas, J., Sardans, J., Llusià, J., Owen, S. M., Carnicer, J., 33, 160–168. Giambelluca, T. W., … Niinemets, Ü. (2010). Faster returns on Milla, R., & Reich, P. B. (2011). Multi-trait interactions, not phylogeny, ‘leaf economics’ and different biogeochemical niche in invasive fine-tune leaf size reduction with increasing altitude. Annals of compared with native plant species. Global Change Biology, 16, Botany, 107, 455–465. 2171–2185. Missio, F., Higuchi, P., Silva, A., Longhi, S., Salami, B., Dalla Rosa, A., Pickering, C. M., & Venn, S. (2013). Increasing the resilience of … Bento, M. (2016). Trade-offs and spatial variation of functional Australian alpine flora to climate change and assosiated threats: traits of tree species in a subtropical forest in southern Brazil. A plant functional traits approach. Australia: National Climate IForest, 9, 855–859. Change Adaptation Research Facility. Mitchell, K. A., Bolstad, P. V, & Vose, J. M. (1999). Interspecific Poorter, L. (2009). Leaf traits show different relationships with shade and environmentally induced variation in foliar dark respira- tolerance in moist versus dry tropical forests. New Phytologist, tion among eighteen southeastern deciduous tree species. Tree 181, 890–900. Physiology, 19, 861–870. Poorter, L., & Bongers, F. (2011). Leaf traits are good predictors of Mooney, H. A., Field, C., Gulmon, S. L., & Bazzaz, F. A. (1981). plant performance across 53 rain forest species. Ecology, 87, Photosynthetic capacity in relation to leaf position in desert ver- 1733–1743. sus old-field annuals. Oecologia, 50, 109–112. Powers, J. S., & Tiffin, P. (2010). Plant functional type classifications Morales, D., Gonzalez-Rodriguez, A. M., & Cermak, J. (1996). Laurel in tropical dry forests in Costa Rica: Leaf habit versus taxonomic forests in Tenerife, Canary Islands: The vertical profiles of leaf approaches. Functional Ecology, 24, 927–936. characteristics. Phyton, 36, 251–263. Prentice, I. C., Meng, T., Wang, H., Harrison, S. P., Ni, J., & Wang, G. Mörsdorf, M. A., Baggesen, N. S., Yoccoz, N. G., Michelsen, A., (2011). Evidence of a universal scaling relationship for leaf CO2. Elberling, B., Ambus, P. L., & Cooper, E. J. (2019). Deepened win- New Phytologist, 190, 169–180. ter snow significantly influences the availability and forms of ni- Price, C. A., & Enquist, B. J. (2007). Scaling mass and morphol- trogen taken up by plants in High Arctic tundra. Soil Biology and ogy in leaves: An extension of the WBE model. Ecology, 88, Biochemistry, 135, 222–234. 1132–1141. Müller, S. C., Overbeck, G. E., Pfadenhauer, J., & Pillar, V. D. (2007). Pyankov, V. I., Kondratchuk, A. V., & Shipley, B. (1999). Leaf structure Plant functional types of woody species related to fire distur- and specific leaf mass: the alpine desert plants of the Eastern bance in forest–grassland ecotones. Plant Ecology, 189, 1–14. Pamirs, Tadjikistan. New Phytologist, 143, 131–142. Nauta, A. L., Heijmans, M. M. P. D., Blok, D., Limpens, J., Elberling, Quested, H. M., Cornelissen, J. H. C., Press, M. C., Callaghan, T. V, B., Gallagher, A., … Berendse, F. (2015). Permafrost collapse after Aerts, R., Trosien, F., … Jonasson, S. E. (2003). Decomposition of shrub removal shifts tundra ecosystem to a methane source. sub-Arctic plants with differing nitrogen economies: A functional Nature Climate Change, 5, 67–70. role for hemiparasites. Ecology, 84, 3209–3221. Niinemets, Ü. (1999). Components of leaf dry mass per area – thick- Read, Q. D., Henning, J. A., Classen, A. T., & Sanders, N. J. (2018). ness and density – alter leaf photosynthetic capacity in reverse Aboveground resilience to species loss but belowground resist- directions in woody plants. New Phytologist, 144, 35–47. ance to nitrogen addition in a montane plant community. Journal Ogaya, R., & Peñuelas, J. (2003). Comparative field study of Quercus of Plant Ecology, 11, 351–363. ilex and Phillyrea latifolia: Photosynthetic response to experimen- Reich, P. B., Ellsworth, D. S., Walters, M. B., Vose, J. M., Greshyam, tal drought conditions. Environmental and Experimental Botany, C., Volin, J. C., … Bowman, W. D. (1999). Generelity of leaf trait 50, 137–148. relationships: A test across six biomes. Ecology, 80, 1955–1969. BOONMAN et al. | 1051 Reich, P. B., Oleksyn, J., & Wright, I. J. (2009). Leaf phosphorus in- van der Sande, M. T., Peña-Claros, M., Ascarrunz, N., Arets, E. J. M. fluences the photosynthesis–nitrogen relation: A cross-biome M., Licona, J. C., Toledo, M., & Poorter, L. (2017). Abiotic and analysis of 314 species. Oecologia, 160, 207–212. biotic drivers of biomass change in a Neotropical forest. Journal Rumpf, S. B., Semenchuk, P. R., Dullinger, S., & Cooper, E. J. (2014). of Ecology, 105, 1223–1234. Idiosyncratic responses of high arctic plants to changing snow Villar, R., & Merino, J. (2001). Comparison of leaf construction costs regimes. PLoS ONE, 9, e86281. in woody species with differing leaf life-spans in contrasting eco- Sandel, B., Corbin, J. D., & Krupa, M. (2011). Using plant functional systems. New Phytologist, 151, 213–226. traits to guide restoration: A case study in California coastal Von Holle, B., & Simberloff, D. (2004). Testing Fox's assembly rule: grassland. Ecosphere, 2, 1–16. does plant invasion depend on recipient community structure? Schlesinger, W. H., DeLucia, E. H., & Billings, W. D. (1989). Nutrient- Oikos, 105, 551–563. use efficiency of woody plants on contrasting soils in the west- Vowles, T., Gunnarsson, B., Molau, U., Hickler, T., Klemedtsson, L., ern Great Basin, Nevada. Ecology, 70, 105–113. & Björk, R. G. (2017). Expansion of deciduous tall shrubs but not Simpson, A. H., Richardson, S. J., & Laughlin, D. C. (2016). Soil–cli- evergreen dwarf shrubs inhibited by reindeer in Scandes moun- mate interactions explain variation in foliar, stem, root and re- tain range. Journal of Ecology, 105, 1547–1561. productive traits across temperate forests. Global Ecology and Whitehead, D., Walcroft, A. S., Scott, N. A., Townsend, J. A., Trotter, Biogeography, 25, 964–978. C. M., & Rogers, G. N. D. (2004). Characteristics of photosynthe- Sobrado, M. A. (1991). Cost–benefit relationships in deciduous sis and stomatal conductance in the shrubland species manuka and evergreen leaves of tropical dry forest species. Functional (Leptospermum scoparium) and kanuka (Kunzea ericoides) for the Ecology, 5, 608–616. estimation of annual canopy carbon uptake. Tree Physiology, 24, Soudzilovskaia, N. A., Elumeeva, T. G., Onipchenko, V. G., Shidakov, I. 795–804. I., Salpagarova, F. S., Khubiev, A. B., … Cornelissen, J. H. C. (2013). Williams, M., Shimabokuro, Y. E., & Rastetter, E. B. (2012). LBA-ECO Functional traits predict relationship between plant abundance CD-09 Soil and vegetation characteristics, Tapajos National Forest, dynamic and long-term climate warming. Proceedings of the Brazil, Data set. Retrieved from http://daac.ornl.gov. 10.3334/ National Academy of Sciences USA, 110, 18180–18184. ORNLDAAC/1104 Spasojevic, M. J., & Suding, K. N. (2012). Inferring community assembly Williams-Linera, G. (2000). Leaf demography and leaf traits of mechanisms from functional diversity patterns: The importance of temperate-deciduous and tropical evergreen-broadleaved multiple assembly processes. Journal of Ecology, 100, 652–661. trees in a Mexican montane cloud forest. Plant Ecology, 149, Speed, J. D. M., Austrheim, G., Hester, A. J., & Mysterud, A. (2010). 233–244. Experimental evidence for herbivore limitation of the treeline. Wright, I. J., Reich, P. B., & Westoby, M. (2001). Strategy shifts in Ecology, 91, 3414–3420. leaf physiology, structure and nutrient content between species Street, L. E., Subke, J.-A., Baxter, R., Dinsmore, K. J., Knoblauch, of high- and low-rainfall and high- and low-nutrient habitats. C., & Wookey, P. A. (2018). Ecosystem carbon dynamics differ Functional Ecology, 15, 423–434. between tundra shrub types in the western Canadian Arctic. Wright, I. J., Reich, P. B., Westoby, M., Ackerly, D. D., Baruch, Z., Environmental Research Letters, 13, 084014. Bongers, F., … Gulias, J. (2004). The worldwide leaf economics van der Sande, M. T., Arets, E. J. M. M., Peña-Claros, M., De Avila, spectrum. Nature, 428, 821–827. A. L., Roopsind, A., Mazzei, L., … Poorter, L. (2016). Old-growth Wright, I. J., Westoby, M., & Reich, P. B. (2002). Convergence to- Neotropical forests are shifting in species and trait composition. wards higher leaf mass per area in dry and nutrient-poor habitats Ecological Monographs, 86, 228–243. has different consequences for leaf life span. Journal of Ecology, van der Sande, M. T., Arets, E. J. M. M., Peña-Claros, M., Hoosbeek, 90, 534–543. M. R., Cáceres-Siani, Y., van der Hout, P., & Poorter, L. (2018). Soil Wright, S. J., Kitajima, K., Kraft, N. J. B., Reich, P. B., Wright, I. J., fertility and species traits, but not diversity, drive productivity Bunker, D. E., … Zanne, A. E. (2010). Functional traits and the and biomass stocks in a Guyanese tropical rainforest. Functional growth–mortality trade-off in tropical trees. Ecology, 91, Ecology, 32, 461–474. 3664–3674.