Publication and dissemination of datasets in taxonomy: ZooKeys working example

A concept for data publication and semantic enhancements proposed by ZooKeys and applied in the mile-stone paper by Miller et al. (2009) is described. For the fi rst time in systematic zoology, a unique combination of data publication and semantic enhancements is applied within the mainstream process of journal publishing, to demonstrate how: (1) All primary biodiversity data underlying a taxonomic monograph are published as a dataset under a separate DOI within a paper; (2) Th e occurrence dataset is separately discoverable and accessible through GBIF data portal (data.gbif.org) simultaneously with the publication; (3) Th e occurrence dataset is published as a KML (Keyhole Markup Language) fi le under a distinct DOI to provide an interactive experience in Google Earth; (4) All new taxa (42) are registered at ZooBank during the publication process (mandatory for ZooKeys); (5) All new taxa (42) are provided to Encyclopedia of Life through XML mark up on the day of publication (mandatory for ZooKeys). It is proposed to clearly distinguish between static and dynamic datasets in the way they are published, preserved and cited.


Introduction
Th e publication of datasets is currently being extensively discussed in scientifi c publishing.Th e need for open access to research data consequently ensuring its preservation, dissemination and reuse is recognized as a strategic goal in several documents at governmental and international levels.For instance, at the European Union (EU) level it is recognized that long-term sustainability will be achieved not only by digital data storage but even more so by increasing probability of data reuse.Such a perspective is refl ected by the statement of the Council of Europe on "the importance of better access to unprocessed data and repository resources for data and material that allows fresh analysis and utilisation beyond what the originator of the data had envisaged" (2832nd COMPETITIVENESS (Internal market, Industry and Research) Council meeting, Brussels, 22 and 23 November 2007).Moreover, in 2008 the 7 th Framework Program of EU opened a special call entitled "Rehabilitation of data from biodiversity-related projects funded under previous framework programmes".
Data publication has become a common practice in other branches of science (i.e.Altman and King 2007).In a recently published white paper of OECD (Green 2009), there are described working examples of a quite simple approach to data publishing.Publishing of data is not primarily a technical problem.Th e main question is how to publish data under the open access model and how to motivate data collectors and creators?Th ere are also other unresolved questions related to metadata, consequent data use and reuse, conventions of citation, author's and/or institution's recognition, challenges of publishing dynamic datasets (databases) and so on.
Th e benefi ts from data publishing for authors and for society seem obvious.Since the data is online and freely available, authors gain broader recognition for their work.But more importantly, published data when well-described and validated, contribute to data discovery, access, and publishing infrastructure such as the Global Biodiversity Information Facility (GBIF) and larger data aggregations through organizations like ZooBank, Morphbank, Encyclopedia of Life (EOL).Increasingly, analyses and metaanalyses of aggregated data will unlock new potentials for academic science, applied science (i.e., conservation), and informed public discourse leading to better public policy.Th e indexing practices described here would establish authorship of datasets, crediting researchers and institutions for their productivity and initiative in piloting non-traditional forms of publication.Sponsors, funding agencies and society in general would benefi t from the full disclosure, aggregation and reuse of the results of publicly funded scientifi c research.
In systematic zoology we have already excellent examples of semantic enhancements to research papers (i.e., Pyle et al. 2008;Fisher and Smith 2008;Talamas et al. 2009), however we still lack a comprehensive "full-life-cycle" model for data management, from a manuscript through the peer-reviewed publication process to data aggregation, dissemination, and stable long-term deposition.Th is particularly concerns primary biodiversity (mostly based on but not limited to specimen and observationby-locality) data.
We describe here a vision for data publication and dissemination in taxonomy.Th e most ambitious illustration of this publishing process (Fig. 1) is the milestone paper of Miller et al. (2009) published in the present issue.Following its declared policy to develop and apply innovative ways of publishing towards a "web-based" taxonomy (Penev et al. 2008), ZooKeys, in close cooperation with the authors and GBIF, provide in this paper for the fi rst time in systematic zoology a unique combination of data publication and semantic enhancements (Shotton et al. 2009), applied within the mainstream process of journal publishing: 1.All primary biodiversity data underlying a taxonomic monograph are published as a dataset under a separate DOI within the paper 2. Th e occurrence dataset is indexed to GBIF simultaneously with the publication 3. Th e occurrence dataset is published as a KML (Keyhole Markup Language) fi le under a distinct DOI to provide an interactive experience in Google Earth.Th e interactive map features collection occurrence data for all specimens in the monograph and links to collections of images for each species posted on Morphbank.Data can be fi ltered to display or hide any family, genus, or species.4. All new taxa ( 42) are registered at ZooBank during the publication process (mandatory for ZooKeys) 5.All new taxa (42) are provided to Encyclopedia of Life through XML mark up on the day of publication (mandatory for ZooKeys)

Static and dynamic datasets
A dataset is understood here as a logical fi le -analog or machine-readable -presenting a collection of facts (observations, descriptions or measurements) formally structured in records; each record is structured in fi elds.Within the domain of biological taxonomy, a dataset can be any discrete collection of data underlying a taxonomic papere.g., a list of all occurrence data published in the paper, data tables from which a graph or map is produced, an appendix with morphological measurements, etc.Each dataset has its own DOI within a paper.
We propose a clear distinction between fi xed "data tables" that represent precisely the set or sets of data upon which the analyses and conclusions of a given scientifi c paper are based and dynamic "databases" that represent larger and more extensive collections of data that may or may not include the precise data table or tables that are the referent(s) for a given scientifi c paper.Both data tables and databases are of strong potential scientifi c interest and application but publication of data tables is inextricably linked to a scientifi c paper and the publisher must assure consistent and secure access, in perpetuity, to referent data tables.In other words, a newer version of a data table cannot be uploaded on the journal's website and in this way the publisher guarantees its consistency and perpetuity in time, similarly to the content of an electronic publication.
We propose the use of digital object identifi ers (DOI), semantically related to the DOI of the respective paper, to be assigned to each discrete referent data table.Th is will insure that there is a fi xed and citable entity to which subsequent reviews and revisions can refer.Th e referent data table can be downloaded in conjunction with a paper and critically analyzed.It can also be freely used for subsequent analyses and publications given proper citation and attribution.
Publication and citation of dynamic databases is supplementary and discretionary in the context of a given paper but should certainly be encouraged for the greater good of science.Sustained maintenance of databases is not an obligation for a publisher but is -by emerging scientifi c norms -an obligation for scientists, scientifi c institutions and agencies, libraries and archives.Ultimately we can envision aggregations -at various chronologic, geographic or taxonomic scales -of hundreds of databases composed of thousands of datasets.
Respecting the citation of databases, we note that conventions of citation must include careful specifi cation of the version, date and time of accessioning.Database design should at all points support easy "stamping" of data usage.

Identifi cation and location
We propose also to distinguish between static and dynamic datasets semantically to facilitate harvesting through scripts and other machine-generated methods.Data tables, as the referents of a scientifi c paper, can be identifi ed with the acronym "dt" (from data table) within its DOI, which is a semantical extension of the paper's DOI (working examples from the paper of Miller et al. (2009)

Stand-alone publication of a dataset
"Datasets", as collections of data, can be published separately in a form of a conventional publication containing textual description of key features of the dataset (e.g."metadata") -i.e., introductory information on a taxon/taxa being subject of this dataset, principles of architecture, size, technical description, history of the dataset/database itself, hosting, relations to other datasets, ownership and copyright issues and so on.Within a journal, such a publication may be classifi ed as "Dataset" analogously to other type of publications, i.e., "Editorial", "Research article", or "Correspondence".Formal protocols for what will constitute complete and acceptable metadata for data (data tables, datasets and databases) will need to be prescribed in subsequent analysis.Signifi cant progress on this problem is being made in other domains (Green 2009).

Peer-review
Th e publication -either as a scientifi c paper containing referent data tables -or a stand-alone publication of a dataset -will be subject to standard peer-review processes in accordance with the editorial requirements of specifi c journals.

Citation
Citation of a data will vary in correspondence with the form of publication and preservation.

Dissemination and usage
ZooKeys was the fi rst journal to off er mandatory ZooBank registration and to supply species descriptions of all new taxa to Encyclopedia of Life on the day of publication.Th anks to that new taxa described on the pages of the journal become quickly known to the world.What happens, however, with the species-by-occurrence records (i.e., what is normally called "primary biodiversity data")?
Description of a new species is not easy and it requires investnments of time, energy, knowledge, and resources.Similarly, the collection of specimen records requires serious eff ort.At ZooKeys we are convinced that each species-by-occurrence record collected on Earth has its own discrete value and deserves proper registration, publication, preservation and dissemination with proper attribution for authors, data suppliers and publishers.
Still the job is not complete merely with publication of data.Th e larger challenge is to accumulate a dataset in a repository, or repositories, where it will be preserved, integrated and reused.Th e role of such discovery and mobilisation of primary biodiversity is major reason for GBIF's existence.Th e recently launched Integrated Publish-ing Toolkit (ipt.gbif.org)by GBIF allows data to be published in the form of a dataset using the DarwinCore protocol.Besides the standard DarwinCore fi elds describing taxonomic position and rank of each taxon, locality, collectors, ID of specimens in a collection, etc., the data table published as a downloadable Excel fi le as Appendix B (doi: 10.3897/zookeys.11.160-app.B.dt) in the paper of Miller et al. (2009) was completed with the following additional fi elds: § LSID (ZooBank) of each taxon § DOI of the dataset § DOI of the publication § Citation Th e dataset was published through GBIF simultaneously with the publication of the paper, allowing its reuse within the scope of an unlimited number of cross-cutting, larger datasets compiled at larger geographic, taxonomic or chronologic scales, e.g. of all spider species recorded in China, all plant and animal species recorded in the same localities or region and so on.Th e LSID link of each species off er unlimited possibilities for use of each specimen record in any branch of science and nature conservation.
However, how to use the data independently, in the form "as-it-is-published"? Under the terms of the Creative Commons Attribution License anyone can download the dataset and use it, provided that the original author and source are credited.Alternatively, the Creative Commons "CC 0" license may be used -it places work in the public domain -this license relies on conventional norms of citation to insure proper attribution.Th e same paper however, off ers one more innovative way to display and use this particular dataset.In Appendix C, the data are published in the form of KML, a fi le format which allows interactive mapping in Google Earth, and incorporates display on maps of all localities and species, descriptions of each specimen records, hierarchical fi ltering and mapping of higher taxa (genus, family), and links to species descriptive morphology in Morphbank.It remains only to include the links to EOL species pages, which will be easy to do, when EOL starts to provide LSIDs to their species pages prior to the day of publication.Another great feature of our time to dream for!We believe that our proposed approach will motivate authors to publish data and to receive recognition in the form of citations, future co-authorship, and credentialing for career development.We look forward to the time when data discovery and publishing initiatives like GBIF will automatically index such datasets and incorporate them along with the correspondent descriptive metadata details.

Fig. 1 .
Fig. 1.Th e ZooKeys model for data publication and semantic enhancements workfl ow in taxonomy in the present issue): To distinguish between static and dynamic datasets, DOI of data, if published as a database, would have the form: doi: 10.3897/zookeys.11.160-app.B.db ("db" marks up a dynamic dataset to diff er it from "dt" which means a static dataset or data table).DOI of another data table, i.e. data underlying a graph, within the same publication would have the form: doi: 10.3897/zookeys.11.160-fi g.2.dt