sPlot open - An environmentally-balanced, open-access, global dataset of vegetation plots

This manuscript is still work in progress

This manuscript (permalink) was automatically generated from fmsabatini/sPlotOpen_Manuscript@0f8d658 on October 30, 2020.

Authors

Francesco Maria Sabatini^1,2,†, Jonathan Lenoir^3,†, Tarek Hattab⁴, Elise Arnst⁵, Milan Chytrý⁶, Jürgen Dengler^7,8,9, Valério De Patta Pillar¹⁰, Patrice De Ruffray¹¹, Stephan M. Hennekens¹², Ute Jandt², Florian Jansen¹³, Borja Jiménez-Alfaro¹⁴, Jens Kattge¹⁵, Aurora Levesley¹⁶, Oliver Purschke¹⁷, Brody Sandel¹⁸, Fahmida Sultana¹⁹, Svetlana Aćić²⁰, Emiliano Agrillo²¹, Miguel Alvarez²², Iva Apostolova²³, Mohammed A.S. Arfin Khan²⁴, Isabelle Aubin²⁵, Marijn Bauters^26,27, Yves Bergeron²⁸, Erwin Bergmeier²⁹, Idoia Biurrun³⁰, Anne D. Bjorkman³¹, Laura Casella³², Luis Cayuela³³, Tomáš Černý³⁴, Victor Chepinoga³⁵, János Csiky³⁶, Renata Ćušterevska³⁷, Els De Bie³⁸, Michele De Sanctis²¹, Panayotis Dimopoulos³⁹, Mohamed Abd El-Rouf Mousa El-Sheikh^40,41, Brian Enquist⁴², Manfred Finckh⁴³, Emmanuel Garbolino⁴⁴, Melisa Giorgis⁴⁵, Valentin Golub⁴⁶, Alvaro G. Gutierrez⁴⁷, Mohamed Z. Hatim⁴⁸, Guillermo Hinojos Mendoza⁴⁹, Norbert Hölzel⁵⁰, Jürgen Homeier⁵¹, Wannes Hubau^52,53, Adrian Indreica⁵⁴, John Janssen¹², Birgit Jedrzejek⁵⁵, Norbert Jürgens⁴³, Zygmunt Kącki⁵⁶, Ali Kavgacı⁵⁷, Elizabeth Kearsley⁵⁸, Michael Kessler⁵⁹, Andrey Korolyuk⁶⁰, Hjalmar Kühl^9,61, Flavia Landucci⁶², Hongyan Liu⁶³, Tatiana Lysenko⁶⁴, Corrado Marcenò³⁰, Jesper Erenskjold Moeslund⁶⁵, Jonas V. Müller⁶⁶, Jérôme Munzinger⁶⁷, Jalil Noroozi⁶⁸, Arkadiusz Nowak⁶⁹, Viktor Onyshchenko⁷⁰, Gerhard E. Overbeck⁷¹, Aníbal Pauchard⁷², Robert K. Peet⁷³, Aaron Pérez-Haase^74,75, Tomáš Peterka⁶², Gwendolyn Peyre⁷⁶, Oliver L. Phillips¹⁶, Vadim Prokhorov⁷⁷, Valerijus Rašomavičius⁷⁸, Rasmus Revermann⁴³, John S. Rodwell⁷⁹, Eszter Ruprecht⁸⁰, Solvita Rūsiņa⁸¹, Cyrus Samimi⁸², Joop H.J. Schaminée¹², Marco Schmidt⁸³, Urban Šilc⁸⁴, Željko Škvorc⁸⁵, Anita Smyth⁸⁶, Zvjezdana Stančić⁸⁷, Zhiyao Tang⁶³, Ioannis Tsiripidis⁸⁸, Milan Valachovič⁸⁹, Kim André Vanselow⁹⁰, Kiril Vassilev²³, Eduardo Vélez-Martin⁹¹, Roberto Venanzoni⁹², Alexander Christian Vibrans⁹³, Risto Virtanen^9,94,95, Henrik von Wehrden⁹⁶, Viktoria Wagner⁹⁷, Donald A. Walker⁹⁸, Desalegn Wana⁹⁹, Karsten Wesche^9,100,101, Timothy Whitfeld¹⁰², Wolfgang Willner¹⁰³, Susan Wiser⁵, Thomas Wohlgemuth¹⁰⁴, Sergey Yamalov¹⁰⁵, Helge Bruelheide^1,2

^✉ — To whom correspondence should be addressed: francesco.sabatini@botanik.uni-halle.de
^† — These authors contributed equally to this work

Abstract

Vegetation provides the foundation of life on Earth. Assessing biodiversity status and trends in plant communities is therefore critical to understand and quantify the effects of global change on ecosystems. Here, we present the largest dataset of vegetation plots (i.e. species co-occurrence or community composition data) ever released in open access. It contains information on 91,031 vegetation plots recording the cover or abundance of each plant species that occurs in a plot of a given surface area at the date of the botanical survey. Plots were derived from 103 local to regional datasets. To improve the representation of Earth’s environmental conditions, plots were resampled from a larger pool of vegetation plots using an environmentally stratified sampling design. Each vegetation plot comes with information on community-weighted means and variances of key plant functional traits. Our open-access dataset can be used to explore global patterns of diversity at the plant community level, as ground truthing data in remote sensing applications or as a baseline for biodiversity monitoring.

Background & Summary

Biodiversity is facing a global crisis (1). As many as 1 million species are estimated to be already facing extinction, mostly as a consequence of anthropogenic impacts, land-use and climate change (1). The rates of biodiversity redistribution and homogenization are also accelerating (2; 3). Biological assemblages are becoming progressively more similar to each other globally, as local biodiversity and endemic species go extinct and are replaced by introduced exotic species or by more widespread and competitive native species (1; 3). This has profound potential impacts on human and ecosystem health (4; 5). For instance, many terrestrial and marine species are shifting their geographical distribution as a response to climate change (2), including animals hosting pathogens transmissible to humans (6; 7; 8).

Vegetation, i.e., the assemblage of plant species, is no exception to this biodiversity crisis (9; 10; 3). This is worrisome, since terrestrial vegetation accounts for 80% (450 Gt C) of the living biomass on Earth (11). Given the central role of vegetation in ecosystem productivity, stability and functioning (10), assessing biodiversity status and trends in plant communities is paramount, for other life compartments and human societies alike.

Monitoring plant biodiversity trends requires adequate data across a range of scales (12). Large independent collections of plant occurrence data do exist at the global or continental extent via the Botanical Information and Ecology Network (BIEN) (13), the Global Inventory of Floras and Traits (GIFT) (14) or the Global Biodiversity Information Facility (GBIF) (https://www.gbif.org/). However, all these occurrence-only databases either neglect how individual plant species co-occur and interact locally to form plant communities, or are collected at spatial resolutions (e.g., one‐degree grid cells) which are too coarse to assess biodiversity trends at the most relevant scale of local plant communities (15).

Yet, there is a long-lasting tradition among botanists to record the cover or abundance of each plant species that occurs in a vegetation plot of a given size (i.e. surface area) at a given time (e.g. 16). Compared to species-level data, vegetation-plot data present many advantages. First, they contain information on which plant species co‐occur together in the same locality at a given moment in time (17). This built-in feature of vegetation plots is a necessary prerequisite for testing hypotheses related to biotic interactions among plant species (i.e. plant-plant interactions). It can also provide crucial information on where and when a species is absent, therefore improving current species distribution models (18). Being spatially explicit, vegetation plots can be resurveyed through time to assess potential changes in plant species composition relative to a baseline (19; 20, 3). As they normally contain also information on the relative cover or abundance of each species, vegetation plots are more adequate to detect subtle biodiversity changes, compared to data based on the occurrence of individual species only (21).

Vegetation-plot data are very fragmented, though, as they typically stem from a myriad of research projects. As such, these data often suffer from the usual trade-off in biodiversity data: Collections have either fine-grain spatial resolutions but small spatial extents, or vice versa (22). Furthermore, with their disparate sampling protocols, standards and taxonomic resolutions, aggregating and harmonizing vegetation plot data proves extremely challenging (23). It is not surprising, therefore, that these data have only been rarely used in global‐scale biodiversity research until recently (24; 25).

The sPlot initiative tries to close this data gap. It leverages on several existing local to regional vegetation-plot datasets, to create a harmonized and comprehensive global geo-database of terrestrial plant species assemblages (26). Established in 2013, sPlot currently contains more than 1.9 million vegetation plots, and is fully integrated with the TRY database (27), from which it derives information on plant functional traits. The sPlot database is increasingly being used to study continental- to global-scale vegetation patterns, such as the relative contribution of regional vs. local factors on the global patterns of fern richness (28), the mechanisms underlying the spread and abundance of native vs. invasive tree species (29), and worldwide trait–environment relationships in plant communities (23).

Here, we provide an open-access data set composed of 91,031 plots, which is representative of the environmental space covered by the sPlot database. Plots stem from 103 databases, and span across 115 countries (Figure 1). This resampled dataset (sPlot Open - hereafter) is composed of: (1) plot-level information, including metadata and basic vegetation structure descriptors; (2) the species composition of each vegetation plot, including species cover or abundance information when available; and (3) community-level functional diversity indices derived from the TRY database (27).

Methods

Vegetation plot data sources

We started from the sPlot database v2.1 (created October 2016), which contains 1,121,244 vegetation plots and 23,586,216 species records stemming from 110 different vegetation‐plot datasets of regional, national or continental extent. Some of the 110 datasets stem from regional or continental initiatives (see 26 for more information). For instance: 48 vegetation-plot datasets derive from the European Vegetation Archive (EVA) (17), three major African datasets from the Tropical African Vegetation Archive (TAVA), multiple vegetation datasets in the USA from the VegBank archive (30; 31). Data from other continents (South America, Asia) or countries were contributed as separate datasets. The metadata of each individual vegetation-plot dataset stored in sPlot are managed through the Global Index of Vegetation‐Plot Databases (GIVD; 32), using the GIVD identifier as the unique dataset identifier.

Resampling method

Data in the sPlot database are unevenly distributed across continents and biomes (see 23). Mid-latitude regions in developing countries (mostly Europe, the USA and Australia) are overrepresented, while regions in the tropics and subtropics are underrepresented, which is a typical geographical bias in biodiversity data (e.g., 33; 2). To reduce this imbalance to the extent possible, we performed a stratified resampling approach, using several environmental variables available at the global extent as sampling strata. We considered 30 climatic and soil variables. For climate we complemented the 19 bioclimatic variables from CHELSA (34), as well as two variables reflecting growing-season warmth (growing degree days above 1 °C - GDD1 - and 5 °C - GDD5), which we calculated based on CHELSA bioclimatic variables. In addition we considered an index of aridity (AR) and a model for Potential Evapotranspiration (PET - 35). For soil, we extracted seven variables from the SOILGRIDS database (36), namely: soil organic carbon content in the fine earth fraction, cation exchange capacity, pH, as well as the fractions of coarse fragments, sand, silt and clay.

We stratified our sampling effort based on the following procedure. First we ran a global principal component analysis (PCA) of the 30 above-mentioned environmental variables. We considered the full environmental space of all terrestrial habitats on Earth at a spatial resolution of 2.5 arcmin, totaling 8,384,404 terrestrial grid cells, irrespective of whether a grid cell hosted vegetation plots from the sPlot database v2.1 or not. We then subdivided the environmental space represented by the first two principal components (PC1–PC2), accounting for 47% and 23% of the total variation on PC1 and PC2, respectively, into a 100 × 100 grid. This PC1-PC2 bidimensional space was subsequently used to balance our sampling effort across all PC1-PC2 grid cells for which vegetation plots are available. Before projecting vegetation plots from the sPlot database v2.1 onto this PC1-PC2 environmental space, we removed vegetation plots: from wetlands; from anthropogenic vegetation types; without geographical coordinates; and with a location uncertainty higher than 3 km for those having geographical coordinates. This led to a total of 799,400 out of the initial set of 1,121,244 vegetation plots. When projecting the 799,400 vegetation plots in the PC1-PC2 grid, we calculated how many vegetation plots occurred in each PC1-PC2 grid cell. For those grid cells with more than 50 vegetation plots (n = 858), we randomly selected up to 50 vegetation plots using the heterogeneity-constrained random resampling algorithm from [37]. This approach optimizes the selection of a random subset of vegetation plots that encompasses the highest variability in species composition while avoiding peculiar and rare communities, which may represent outliers. We based the quantification of variability in plant species composition among the 50 randomly selected vegetation plots by computing the mean and the variance of the Jaccard’s dissimilarity index (38) between all possible pairs of vegetation plots for a given random selection of 50 vegetation plots (n = 1225). More precisely, for a given PC1-PC2 grid cell containing more than 50 vegetation plots, we generated 1,000 random selections of 50 vegetation plots and ranked the 1,000 random selections according to the mean (ascending order) and variance (descending order) value. Ranks from both sortings were summed for each random selection, and the random selection with the lowest summed rank was considered as the most representative of the focal grid cell. In case a grid cell contained fewer than 50 plots, we retained all of them. In this way, we reduced the imbalance towards over-sampled climate types, while ensuring the resampled dataset to be representative of the entire environmental gradient covered by the sPlot database. We repeated the resampling procedure three times to get three different possibilities of a random selection of 50 vegetation plots per PC1-PC2 grid cell with, initially, more than 50 vegetation plots. Vegetation plots selected during the first iteration were our first choice, while we considered the vegetation plots additionally selected in the second and third iteration as reserves when asking for the permission to release the data as open access to each dataset’s contributor(s).

Permission to release the data as open access

The resampling procedure resulted in a preliminary potential selection of 98,383 vegetation plots (first choice) and 51,634 vegetation plots flagged as reserves (second or third choice for the subset of PC1-PC2 grid cells with more than 50 vegetation plots available). Being the sPlot database a consortium of independent datasets, whose copyright belongs to the data contributor, we used this preliminary potential selection to ask each dataset’s custodian (i.e., either the owner of a dataset or its authorized representative in case of a collective dataset) for permission to release the data of each selected vegetation plot as open access. For 8,070 vegetation plots, permission could not be granted, for instance because the data are unpublished, confidential or sensitive. For these vegetation plots, we used the reserve pool to randomly select replacements, for which such permission could be granted. We imposed the constraint that each vegetation plot in the reserve should belong to the same environmental stratum, i.e., the same PC1-PC2 grid cell, of the confidential vegetation plot. Note that 2,380 PC1-PC2 grid cells (11.7% of total) had one more confidential vegetation plots (median = 1, mean = 3.4, max = 171) that could not be replaced from the reserve pool.

Trait information

For each vegetation plot for which open access has been granted, we computed the community weighted means for eighteen plant functional traits derived from the TRY database v3.0 (27). These traits were selected among those traits that describe the leaf, wood and seed economics spectra (39; 40), and are known to either affect different key ecosystem processes or respond to macroclimatic drivers or both (26). The eighteen plant functional traits were: (1) leaf area [mm²]; (2) stem specific density [g cm^-3]; (3) specific leaf area [m²kg^-1]; (4) leaf carbon concentration [mg g^-1]; (5) leaf nitrogen concentration [mg g^-1]; (6) leaf phosphorus concentration [mg g^-1]; (7) plant height [m]; (8) seed mass [mg]; (9) seed length [mm]; (10) leaf dry matter content [g g^-1]; (11) leaf nitrogen per area [g m^-2]; (12) leaf N:P ratio [g g^-1]; (13) leaf 𝛿¹⁵N [per million]; (14) seed number per reproductive unit; (15) leaf fresh mass [g]; (16) stem conduit density [mm^-2]; (17) dispersal unit length [mm]; and (18) conduit element length [μm].

Because missing values were particularly widespread in the species-trait matrix, we employed a gap-filling procedure based on hierarchical Bayesian modeling (R package ‘BHPMF’, 41; 42). Gap-filling was performed at the level of individual observations. We then log‐transformed all gap‐filled trait values and averaged each trait by taxon (i.e., at species, or genus level). Additional information on the gap-filling procedure are available in [26].

Community‐weighted means (CWM) and the variances (CWV) were calculated for every plant functional trait j and every vegetation plot k as follows (43):

\[ CWM_{j,k} = \sum_{i}^{n_k} p_{i,k} t_{i,j}\](1)
\[ CWV_{j,k} = \sum_{i}^{n_k} p_{i,k} (t_{i,j} - CWM_{j,k})^2\](2)

where n_k is the number of species with trait information in vegetation plot k, p_i,k is the relative abundance of species i in vegetation plot k calculated as the species’ fraction in cover or abundance of total cover or abundance, and t_i,j is the mean value of species i for trait j.

Data Records

The final dataset that is provided here as open access contains 91,031 vegetation plots from 115 countries and all continents except Antarctica (Figure 1) and stems from 103 constitutive datasets (Table 1). It only contains the species composition of vascular plants, while information on the composition of bryophytes and lichens was discarded since it was only available for a minority of plots (n = 4,963 and n = 3,045, respectively). Information on the size (surface area) of the vegetation survey is available for 61,898 vegetation plots, and ranges between 0.01 m² and 4 ha (mean = 270 m²; median = 78.5 m²). The average number of vascular plant species per vegetation plot ranges between 1 (i.e. monospecific stands) and 270 species (mean = 17.6; median = 13).

By reducing the overrepresentation of vegetation plots in specific environmental conditions, the resampling procedure described above strongly reduced the bias in the distribution of vegetation plots within the environmental niche space. Yet, due to the lack or scarcity of data from some geographical regions, like the tropics, the spatial distribution of vegetation plots remains unbalanced across geographical regions (Figure 1). This is evident when comparing the number of plots across continents or biomes. Europe is by far the best represented continent, with 53,884 vegetation plots. In contrast, Africa and South America have only 4,507 and 5,515 vegetation plots, respectively. The representation of biomes is equally unbalanced. The biomes ‘Temperate midlatitudes’ and ‘Subtropics with winter rain’ have 37,507 and 16,510 vegetation plots, respectively, while none of the other biomes have more than 10,000 vegetation plots (Figure 2).

Finally, the dataset contains a relatively balanced number of forest (n = 25,832) vs. non forest (n = 38,203) vegetation plots, with a minor proportion of plots remaining unassigned (n = 10,050). The assignment of plots to forests and non-forests is based on multiple lines of evidence, including the plot-level information on the cover of the tree layer, as well as traits of species composing a plot, such as growth form and height. In short, a plot record was considered a forest if the cover of the tree layer, or alternatively, the sum of the relative cover of all tree taxa, was greater than 0.25. It was instead considered a non-forest record if the sum of relative cover of low‐stature, non‐tree and non‐shrub taxa was greater than 0.90. For an extensive explanation on this classification scheme, we refer the reader to [26]. Even if the proportion of forest vs. non-forest vegetation plots is relatively well-balanced, the geographical distribution of vegetation plots belonging to different vegetation types is likely not balanced in the geographical space, as it depends on the idiosyncrasies of the constitutive datasets composing the sPlot database. For instance, the data from New Zealand only include plots collected in non-forest ecosystems, while data from Chile only refer to forests. We invite potential users to carefully read the description of each individual dataset in GIVD, or to contact the custodians of each dataset, before using sPlot Open.

Database Organization

sPlot Open is organized into three main matrices.

The ‘header’ matrix contains plot level information for the 91,031 vegetation plots provided in this open access dataset, including metadata (e.g., plot ID, ownership, sampling date, geographical location, positional accuracy), sampling design information (e.g., the total surface area used during the vegetation survey), and a plot-level description of vegetation structure (e.g., vegetation type, percentage cover of each vegetation layer), and vegetation type. Plots in Europe are also classified according to the EUNIS habitat classification (column ‘ESY’), based on the habitat classification expert system described in [45]. For each vegetation plot we further provide information on the dataset it stems from, based on the IDs used in the Global Index of Vegetation-Plot Databases. A brief description of all the 43 variables contained in the header matrix is provided in Table 2.

The ‘DT’ matrix contains data on the species composition of each plot. It is structured in a long format and contains 1,608,610 records, from 39,997 vascular plant taxa, mostly resolved at the species level. For each record we report both the taxon name as originally contributed by the data custodian (column ‘Original_scpecies’), and the taxon name after taxonomic standardization (column ‘Species’). For each entry, we report the species cover//abundance values. These follow different standards across the datasets constituting the sPlot database. We therefore provide both the cover//abundance value as reported in the original data (column ‘Original_abundance’), together with the abundance scale that was originally used (column ‘Abundance_scale’). This can take seven values: ‘CoverPerc’ = percentage cover, ‘pa’ = presence-absence, ‘x_BA’ = basal area (m²/ha, only for woody species), ‘x_IC’ = individual count, i.e., number of individuals in plot, ‘x_SC’ = stem count, i.e., number of stems in plot, ‘x_IV’ = importance value index, ‘x_PF’ = presence frequency. The great majority of entries, however, use the percentage cover scale (n= 1,397,109). Finally, for each entry we calculated a ‘Relative_cover’, i.e., the cover//abundance of a given taxon divided by the total cover//abundance of all taxa in that vegetation plot.

The ‘CWM_CWV’ matrix contains the community-weighted means and variances calculated for each of the 18 functional traits mentioned above. It also contains three additional columns. The column ‘Species_richness’ returns the number of species recorded in each plot. The columns ‘Trait_coverage_cover’ and ‘Trait_coverage_pa’ return respectively the proportion of total cover and species in a plot for which functional trait information was available.

Functional trait information was available for 20,932 species. The average proportion of species in each plot for which we have functional trait information is 0.88 (median = 1). For 47,177 plots the coverage is complete, while only in one plot we have no functional trait information for any of the occurring species. When considering relative cover, the average trait coverage is 0.89. As many as 68,234 and 74,388 plots have functional trait information for more than 80% of the species or 80% of relative cover, respectively.

sPlot Open contains two additional objects. The ‘metadata’ matrix contains plot-level metadata, which provide information on the origin of each individual vegetation plot. This object contains 15 columns, with information on Plot ID, dataset of origin (column ‘GIVD_ID’ - 32), author or surveyor names (columns ‘Releve_author’ and ‘Releve_coauthor’), bibliographic references both at the dataset (column ‘DB_BIBTEXKEY’) and plot level (‘Plot_Biblioreference’ and ‘BIBTEXKEY’). Similarly, the column ‘Project_name’ provide information on the project in which a vegetation plot was collected. When available, we also provide information on the numbering of the plots in the publication where they originally appeared (columns ‘Nr_table_in_publ’, ‘Nr_releve_in_table’), or in the dataset where they were initially stored (‘Original_nr_in_database’). In case of nested plots (n=1,786), we also provide the original plot and subplot IDs (columns: ‘Original_plotID’, ‘Original_subplotID’). The last two columns report plot-level ‘Remarks’, and the unique identifier produced by Turboveg when the vegetation plot was first stored (‘GUID’).

Finally, the object ‘references’, contains all the bibliographic references formatted according to a BibTex standard. Each reference is tagged with a key corresponding to the fields ‘DB_BIBTEXKEY’ and ‘BIBTEXKEY’ in the metadata. We further provide an R function (‘sPlotOpen_citation’) to create reference lists, based on a selection of plots and\or datasets.

With the exception of the ‘reference’ file (format .bib), all objects are provided in tab-delimited .txt files. All objects, including the ‘sPlotOpen_citation’ function are also compiled inside an .RData object.

Technical Validation

The sPlot database has a nested structure, and is composed of several individual datasets, each validated and maintained by its respective dataset custodian. In same cases, individual datasets are also collections, whose vegetation plots were provided by their respective owners (the person who performed the actual vegetation survey) or by someone who digitized the original data from the scientific or grey literature. We obviously have no direct control on the individual vegetation plots that we provide here in an open access dataset. Yet, each of these vegetation plots stem from trained professional botanists, or published scientific work, and are accompanied by detailed information on the sampling protocols used, thus ensuring data quality and reliability.

Before having been integrated into the sPlot database, each dataset was further checked for consistency and, if having a different format, was converted to a Turboveg 2 database (46). During this conversion, we checked that all datasets contained the required metadata information, and cross-checked that each plot was located within the geographic scopes of its respective dataset. Finally, we harmonized all the taxonomic names from all datasets, based on the sPlot’s taxonomic backbone (Purschke 2017). This backbone matched all the taxonomic names (without nomenclatural authors) from all datasets in sPlot 2.1 and TRY v3.0 (27) to their resolved version based on the Taxonomic Name Resolution Service web application (TNRS version 4.0; 47; iPlant Collaborative, 2015). This allowed to (1) harmonize all datasets to a common nomenclature, and (2) link the sPlot database to the TRY database (27). All taxa originally denoted at taxonomic ranks lower than species, were aggregated at species level. Additional detail on the taxonomic resolution is reported in [26], while a description of the workflow, including R‐code, is available in [48]

Usage Notes

The sPlot Open database can be downloaded from https://www.idiv.de (link to PlantHub). Users are invited to cite the original sources when using sPlot Open. For some datasets (e.g., AF-00-009, AF-CD-001) the identification of taxa at species level is still in progress. As a rule, we recommend sPlot Open users to get in touch with the custodian(s) of the data they are planning to use (custodian names are reported in https://www.idiv.de/sPlot). The use of data contained in BioTIME should cite original data citations in addition to the present paper. The data included in the present paper represent the subset of sPlot for which we were able to secure permission for making these data open. The additional data in sPlot are available under sPlot’s Governance and Data Property Rules (www.idiv.de/sPlot).

Code Availability

The R code used to produce sPlot Open from the sPlot 2.1 database is contained in the sPlotOpen_code GitHub repository: (https://github.com/fmsabatini/sPlotOpen_Code/). This manuscript was produced using the Manubot workflow (49). The code for reproducing this manuscript is stored in the sPlotOpen_manuscript GitHub repository: (https://github.com/fmsabatini/sPlotOpen_Manuscript).

Acknowledgements

We are grateful to thousands of vegetation scientists who sampled vegetation plots in the field or digitized them into regional, national or international databases. We also appreciate the support of the German Research Foundation for funding sPlot as one of the iDiv (DFG FZT 118, 202548816) research platforms, and the organization of three workshops through the sDiv calls. We acknowledge this support with naming the database “sPlot”, where the “s” refers to the sDiv synthesis workshops.

The study has been supported by the TRY initiative on plant traits (http://www.try-db.org). The TRY initiative and database is hosted, developed and maintained by J. Kattge and G. Bönisch (Max Planck Institute for Biogeochemistry, Jena, Germany). TRY is currently supported by DIVERSITAS/Future Earth and the German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig. Jens Kattge acknowledges support by the Max Planck Institute for Biogeochemistry (Jena, Germany), Future Earth, the German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig and the EU H2020 project BACI, Grant No 640176.

Isabelle Aubin was funded through Natural Sciences and Engineering Research Council of Canada and Ontario Ministry of Natural Resources and Forestry. Yves Bergeron was funded through Natural Sciences and Engineering Research Council of Canada. Idoia Biurrun was funded by the Basque Government (IT936-16). Anne Bjorkman thank the Herschel Island-Qikiqtaruk Territorial Park management, Catherine Kennedy, Dorothy Cooley, Jill F. Johnstone, Cameron Eckert and Richard Gordon for establishing the ecological monitoring programme. Funding was provided by Herschel Island-Qikiqtaruk Territorial Park. Luis Cayuela was supported by project BIOCON08_044 funded by Fundación BBVA. Milan Chytrý, Flavia Landucci, Corrado Marcenò and Tomáš Peterka were supported by the Czech Science Foundation (project no. 19-28491X). Brian Enquist thanks the following individuals and institutions for contributing data to sPlot via the SALVIAS database: Mauricio Bonifacino, Saara DeWalt, Timothy Killeen, Susan Letcher, Nigel Pitman, Cam Webb, The Missouri Botanical Garden, RAINFOR and the Amazon Forest Inventory Network. Alvaro G. Gutiérrez acknowledges FONDECYT 11150835, Project FORECOFUN-SSA PIEF-GA-2010–274798), CONICYT-PAI (82130046). Mohamed Z. Hatim thanks Kamal Shaltout for supervision, and Joop Schaminée for support and funding from the Prince Bernard Culture Fund Prize for Nature Conservation. Jürgen Homeier received funding from BMBF (Federal Ministry of Education and Science of Germany) and the German Research Foundation (DFG Ho3296-2, DFG Ho3296-4). Tatiana Lysenko was funded by Russian Foundation for Basic Research (grant No. 16-04-00747a). Jérôme Munzinger was supported by the French National Research Agency (ANR) with grants INC (ANR-07-BDIV-0008), BIONEOCAL (ANR-07-BDIV-0006) & ULTRABIO (ANR-07-BDIV-0010), by National Geographic Society (Grant 7579-04), and with fundings and authorizations of North and South Provinces of New Caledonia. Gerhard E. Overbeck acknowledges support from Brazil’s National Council of Scientific and Technological Development (CNPq, grant 310022/2015-0). Josep Peñuelas would like to acknowledge the financial support from the European Research Council Synergy grant ERC-SyG-2013-610028 IMBALANCE-P Oliver Phillips was funded by an ERC Advanced Grant (291585, “T-FORCES”) and a Royal Society-Wolfson Research Merit Award. Valério D. Pillar has been supported by the Brazil’s National Council of Scientific and Technological Development (CNPq, grant 307689/2014-0). Solvita Rūsiņa was supported by the University of Latvia grant AAP2016/B041//Zd2016/AZ03 within the “Climate change and sustainable use of natural resources”. Franziska Schrodt was supported by a University of Minnesota Institute on the Environment Discovery Grant, a German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig grant (50170649_#7) and a University of Nottingham Anne McLaren Fellowship. Kim André Vanselow would like to thank W. Bernhard Dickoré for the help in the identification of plant species and acknowledge the financial support from the Volkswagen Foundation (AZ I/81 976) and the German Research Foundation (DFG VA 749/1-1, DFG VA 749/4-1). Evan Weiher was funded by NSF DEB-0415383, UWEC-ORSP, and UWEC-BCDT.

This paper is dedicated to the memory of Dr. Ching-Feng (Woody) Li.

Author contributions

FMS wrote the first draft of the manuscript, with considerable input from JL and HB. JL and TH wrote the resampling algorithm. FMS set up the GiHub projects, curated the database, and produced the graphs. He also coordinated the sPlot consortium. SMH wrote the Turboveg v3 software, which holds the sPlot database. JK provided the trait data from TRY and FS performed the trait data gap filling. HB secured the funding for sPlot as a strategic project of iDiv. All other authors contributed data. All authors contributed to revising the manuscript.

Competing interests

The authors declare no competing interests.

References

Supplementary Material