Tundra Trait Team: A database of plant traits spanning the tundra biome

Motivation : The Tundra Trait Team (TTT) database includes field‐based measure‐ ments of key traits related to plant form and function at multiple sites across the tundra biome. This dataset can be used to address theoretical questions about plant strategy and trade‐offs, trait–environment relationships and environmental filtering, and trait variation across spatial scales, to validate satellite data, and to inform Earth system model parameters. Main types of variable contained : The database contains 91,970 measurements of 18 plant traits. The most frequently measured traits (> 1,000 observations each)


| INTRODUC TI ON
Plant traits reflect species' ecological strategies and life histories, and underlie differences in the way plants acquire and use resources. Traits related to plant size and the leaf economics spectrum, for example, represent fundamental trade-offs between the capture and conservation of resources (Díaz et al., 2016;Wright et al., 2004). Because plant traits reflect the direct interaction between a plant and its habitat, variation in plant traits is often closely linked to environmental (including climatic) variation (Moles et al., 2006(Moles et al., , 2009Sandel et al., 2010). As such, plant traits can be used to predict species' responses to environmental and climate change (Fridley, Lynn, Grime, & Askew, 2016;Soudzilovskaia et al., 2013).
Global trait databases (Kattge et al., 2011) have dramatically increased the accessibility of plant trait data over the past decade, but these databases are heavily geographically biased towards temperate regions (e.g. 98% of observations in the TRY trait database were measured south of 60°N). In contrast, the tundra is the most rapidly warming biome on the planet (IPCC, 2013), but until now has been underrepresented in global trait databases, which limits our ability to predict the functional consequences of climate change. This poor geographical coverage of tundra species is especially pronounced in the most remote (e.g. high Arctic, upper alpine) regions. Because intraspecific trait variation is thought to be particularly important in ecosystems such as the tundra where diversity is low and species' ranges are large (Siefert et al., 2015), multi-site trait observations on many individuals are needed to capture the full extent of tundra plant trait variation.
Here, we present the Tundra Trait Team (TTT) database, which contains more than 90,000 unique observations of 18 plant traits on 978 tundra species (Figures 1 and 2, Table 1). The TTT database is unique in its depth and spread. Trait data were collected at 207 unique tundra locations ranging from 47°S (the sub-Antarctic Marion Island) to 79.1°N (Sverdrup Pass, Ellesmere Island, Canada), and include multiple observations on individuals at the same location as well as of the same species at different locations. In addition, 99.8% of the observations in the database are georeferenced, thus allowing trait observations to be linked with environmental data such as gridded climate datasets (e.g. WorldClim, www.worldclim. org, CHELSA, chelsa-climate.org, CRU, crudata.uea.ac.uk, etc.). The TTT database fills a major geographical gap; it contains nearly twice as many high-latitude (≥55°N) observations as the TRY trait database for many key traits ( Figure 3). Trait values in TTT are skewed towards individuals of smaller stature (height and leaf area) relative to values in TRY, likely reflecting improved sampling of the tundra's coldest extremes ( Figure 4).
The TTT database can be used to address wide-ranging theoretical and practical ecological questions. Multiple trait observations on individuals and species at numerous sites across the tundra biome enables the quantification of inter-and intraspecific trait include plant height, leaf area, specific leaf area, leaf fresh and dry mass, leaf dry matter content, leaf nitrogen, carbon and phosphorus content, leaf C:N and N:P, seed mass, and stem specific density. Software format: csv file and GitHub repository with data cleaning scripts in R; contribution to TRY plant trait database (www.try-db.org) to be included in the next version release.  (Wullschleger et al., 2014). We expect that making this dataset publicly available will contribute to future research in these and other unforeseen ways.

| Data acquisition and compilation
Data were submitted directly by the tundra researchers that collected them (see author list and Acknowledgments). These data represent a mix of previously collected data as well as new data collected as part of a multi-site field campaign. In some cases, the submitted trait data have contributed to publications (see Supporting Information Appendix S1 for reference list) but all values in the database are from primary sources (i.e. not extracted from publications).
None of the data contained in the TTT database currently occur in other trait databases (e.g. TRY). All trait data in this version (v. 1.0) of the database are collected on plants growing in situ under natural conditions (i.e. data from experimental treatments were removed). Future updates to the database will also include trait data from experimental treatments (warming, grazing, nutrient addition, snow manipulation, etc.). This will be indicated accordingly in the 'Treatment' column.

| Data curation and quality control
All observations were checked to ensure logical latitude and longitude information and converted to standardized units of measurement. We also removed obviously erroneous or impossible values (e.g. leaf dry matter content values greater than 1 g/g). When possible, suspected errors were checked with the initial data provid- For those species with at least 10 observations of the same trait type, we additionally report an 'error risk' for each observation (see TRY database protocols for more information on the term 'error TA B L E 1 All traits contained in the Tundra Trait Team (TTT) database, including the number of total observations of each trait, the number of unique locations (rounded to the nearest tenth of a decimal degree) at which each trait was measured, and the total number of species for which each trait was measured. The mean, SD, median, and 95% quantiles for each trait are also provided. Leaf d13C and leaf d15N correspond to the leaf carbon isotope signature and the leaf nitrogen isotope signature, respectively

| Data availability and access
The TTT database will be maintained at the GitHub repository

| Data use guidelines
Data are governed by a Creative Commons Attribution 4.0 International copyright (CC BY 4.0). Data are fully public but should be appropriately referenced by citing this data paper. Although not mandatory, we additionally suggest that data users contact and collaborate with data contributors (names provided in the 'DataContributor' column, contact information available through the TTT website: https://tundratraitteam.github.io/) whose datasets F I G U R E 3 Histogram of all observations above 55°N contained in the Tundra Trait Team (TTT; coloured bars) and TRY (grey bars; trydb.org) databases. Bars are stacked, such that the height of the bar corresponds to the total number of observations (TRY + TTT) for that latitude. The first panel ('All Obs') contains all observations for height, specific leaf area (SLA), leaf N, leaf C, leaf P, leaf dry matter content (LDMC), seed mass, leaf area and stem specific density, while subsequent panels show observations for key individual traits. The TTT database more than doubles the number of high-latitude observations available for most traits; this is especially true in Arctic (i.e. above have contributed a substantial proportion (e.g. 5% or greater) of trait observations used in a particular paper or analysis.

| DE SCRIP TI ON OF DATA
The TTT database contains 91,970 observations on 18 plant traits measured in 207 locations across the tundra biome (Figures 1 and 2 Table 2). We have also retained information about the identity of each individual plant ('IndividualID') to facilitate analyses of within-individual trait-trait correlations.
In addition to the trait values themselves, nearly all observations (99.8%) contain information about latitude and longitude of the location where the measurement was taken (Figures 2 and 3).
Elevation was also provided for most observations (70%). The high degree of georeferencing in the dataset enables the extraction of climate and other environmental data corresponding with each trait measurement. In addition, many data contributors provided information about the habitat type ('SubsiteName') in which each individual occurred. The full structure of the database is described in Table 2.

ACK N OWLED G M ENTS
This paper is an outcome of the sTundra working group supported Allen, Nathan Young, Jenny Lowe, and many others to trait data TA B L E 2 Dataset structure. The cleaned Tundra Trait Team (TTT) dataset is provided as a csv file and consists of a single data

Comments
Additional comments provided by the data contributor or collator, usually related to how the measurements were conducted collection, and thank the governments, parks, field stations, and local and indigenous people for the opportunity to conduct research on their land.

R E FE R E N C E S S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section at the end of the article.