This tutorial provides best-practice principles when standardising traits with terms of the Ecological Trait-data Standard (ETS, https://terminologies.gfbio.org/terms/ets/pages/).

We focus this document on the first of the three use-cases for the ETS vocabulary: The publication of project-specific primary data. To comply with the FAIR principles of scientific data management and aim at findability, accessibility, interoperability and reusability (Wilkinson et al. 2016), trait data should apply the following principles when preparing data for publication:

• Label your dataset! data should be published with appropriate metadata and provide a descriptive title and keywords to be findable by search engines. Keywords should include the general label “trait-data”, and more specific labels such as “functional traits” or “morphometric traits”, as well as the taxa covered by the dataset.

• State terms of use! To enable re-use, the metadata must contain the name of the author or owner of the data, the suggested citation for the data, as well as the license model or terms of use of the data. Recommended are copyright wavers and licenses that permit re-use without seeking authorization from the rights holder, such as Creative Commons (CC) licenses, which also can warrant attribution and citation (CC by) (Kissling et al. 2018). Ideally, these statements are provided by the data-hoster in a machine-readable way, by using standardised fields of metadata or hyperlink references.

• Use independent file formats! For reasons of long-term accessibility, data should not be uploaded in proprietary spreadsheet formats (like ‘.xlsx’) but rather in comma- or tab-separated text files (‘.csv’ or ‘.tsv’) that are compatible with all computing platforms and internationalisation settings by applying a unified character encoding (e.g. UTF-8 or ASCII). For internationalisation, use points as decimal delimiters and set date-formats to ISO 8601, i.e. “YYYY-MM-DD”.

• Use standard terms! Draw column or field names from the ETS vocabulary wherever possible. If co-variates or any ancillary data are not covered by the vocabulary, apply terms of the Darwin Core Standard (DwC).

• Use URI references! Wherever possible, the concepts used to describe the dataset should be referenced to globally accessible Uniform Resource Identifiers (URIs). These references allow data users and software tools to relate the dataset to well-defined concepts of taxa or traits, or measurement units, etc. The ETS-terms ‘traitID’ and ‘taxonID’ are defined to label fields with URIs as values.

• Provide definitions! Metadata must reference the data standards, taxonomic concepts and trait ontologies applied in the dataset as well as any user-defined terms and trait concepts. To provide own trait-concepts, use the terms provided within the ETS section “Terms for traitlists”.

• Publish raw data! When publishing trait-data, priority is on containing measurement values while minimizing the risk of purporting errors through data conversion or aggregation. Thus, data should be published as they have been recorded, i.e. in original units and raw values, along with standardised values that match a unified and compatible scheme of a data standard. This is true for the core terms, i.e. the trait value (the quality) and observation entity, but also for any co-variates reported along with it.

• Keep original values! For the core terms, data providers should complement user-defined entries (such as project specific taxon abbreviations) with standardised values by applying the ‘*Std’-terms. For co-variates, the extensions of the vocabulary provide column names for standardised values only, e.g. elevation above sea-level. These should adhere to accepted standards wherever possible. If any conversion was applied to obtain these data, users should keep the original values in user-defined columns, e.g. by appending ‘*Orig’ to the column name. Also, additional co-variates, e.g. on environmental growth conditions of the specimen, should be provided in user-defined columns. Any user-defined columns must be adequately defined in the accompanying meta-information.

• Report Uncertainty! If publishing aggregate data (e.g. averages for multiple measurements), report the \(n\) and dispersion of the aggregate value. This allows to weigh data from multiple sources when producing compilations. Any sources of bias or noise should be reported along with the data (e.g. observer ID, conditions of measurement etc.). Use terms of the ETS extension “Measurement or Fact”.

• Use Identifiers! Well-defined identifiers (‘IDs’) are key elements to structure the datasets and relate them to complementing datasets, if necessary (Fig. 2 c & d). For instance, for occurrence level data where multiple trait measurements are reported for each individual specimen, the same user-defined entry for ‘occurrenceID’ would link several measurements across the rows of the dataset. Similarly, multivariate measurements, for instance gas chromatography data or x-y-z data of morphometric coordinates could be linked via a ‘measurementID’. In literature data, summarised traits are usually given at the taxon level instead of the individual organism (e.g. reported as means or factorials) and a ‘taxonID’ is the key identifier. In larger compilations, a ‘datasetID’ allow to trace data origin to the primary source and keep track of authorship and original reference. Beyond being just of structural use for the dataset, identifiers are capable of linking own data to consensus taxonomy and trait terminology via URIs, which point to external terminology services.

## Please enter password in TK window (Alt+Tab)