The Darwin Core Archive standard is a human and machine readable system for exchanging information on biological diversity. It consists of a central data file (e.g. a taxonomy) and any number of extensions (e.g. specimens, images) that point to the central file (a star schema, see Baker, 2013 & Fig. 1). As the data in Scratchpads are predominately about taxa (images of taxa, descriptions of taxa, specimens of taxa,...) we have chosen to use the Scratchpad's taxonomic classification as our central file. The extensions that we use cover descriptions of taxa using the Species Profile Model, the distribution of taxa, images and other media, associated literature references, specimens, vernacular names, the users of the site and comments made on site content. The content of the files may be described using Darwin Core terms, or terms from other schemas. These data files are all plain text (UTF-8) files.

Figure 1.  

The star schema showing the relation of Darwin Core Archive extension files to the core file.

In addition to the data files each archive has a meta.xml file that defines what the fields are in each data file, how rows and fields are terminated in each file, and what type of data are contained in each file (rowType). Of particular interest to our implementation is the fact that for a particular file in the archive the meta.xml can define multiple meanings by using different rowTypes. For example this allows us to treat descriptions.txt as both TDWG Descriptions (as defined at http://rs.gbif.org/terms/1.0/Description) and EoL Media items (as defined at http://eol.org/schema/media/Document). More information on this implementation is in the description.txt section below.

The Drupal module we have written, dwca_export (Fig. 2), provides new default Drupal views (see Views) that match the structures we have specified in our meta.xml file. The Views data export module provides the functionality to export these views as comma-separated values text files. This module also provides the functionality to create these text files in a batch operation (appending several rows at a time to the output file to reduce memory usage) from the command line. This batch operation reduces the memory required to produce the archive, as early implementations of this module for large exports required 1GB of memory, which is a large amount of resource in a shared hosting environment. In most cases one view is used to produce one text file in the DwC-A. The exception is the description.txt which uses one view for each Species Profile Model (SPM) field. A list of the SPM fields we use is at https://git.scratchpads.eu/v/scratchpads.git/blob/695831817b318a4eef42a31de1e61bd6f67145a7:/sites/all/modules/spm/EOL_SPM.xml. The output of these views is concatenated in a later stage of the archive creation process.

Figure 2.  

Data flow from Scratchpad to Darwin Core Archive. The dwca_export module (orange) defines a number of custom Drupal Views (queries) that collect the data required for archive generation from the Scratchpad (blue) and combines them with the meta.xml which describes the information in the archive. For eMonocot Scratchpads the emonocot_dwca module (green) provides an intermediary function replacing the Scratchpads internal unique identifiers with those used throughout the eMonocot project (see eMonocot modifications section).

To conserve server resources the DwC-As are generated as a background process, triggered by a change being made to the content of the site. It is also possible for the Scratchpad administration team to rebuild the archives on demand using Drush via the command line (see Drush section below).

The functionality provided by this module can be enabled by the maintainer on any Scratchpad site through the Tools page found at Admin > Structure. Once enabled, the archive file will become available at [scratchpads.url]/dwca.zip after it has been created by the system.

This module is for Scratchpad maintainers who wish to share the content of their site with third party aggregators.

The classification.txt file forms the core of our star schema and is consistent between all third party users.

Scratchpads as a system are designed to facilitate classifications conforming to any (or no) code of nomenclature. While we change the Scratchpad interface to match the terminology of a particular code on a per-classification basis, Darwin Core standardises the terminology used for the taxonomic and nomenclatural status of a taxon (http://rs.gbif.org/vocabulary/gbif/taxonomic_status.xml, http://rs.gbif.org/vocabulary/gbif/nomenclatural_status.xml). It is therefore necessary for us to map from the ontologies we use (derived from ITIS, the Integrated Taxonomic Information System) to the DwC standards. This mapping is shown in Table 2.

The field used are shown in Table 1.

Descriptions used for describing the identification and biology of taxa in Scratchpads are defined using the TDWG Species Profile Model. A precise definition of the fields that we support is available from the relevant part of the Scratchpads Git repository and match the fields used by EoL in the Lifedesks project.

In Scratchpads it is possible to create one or more taxon description nodes for every taxon. These nodes allow entry of text into any of the fields described above. In a DwC-A each of these fields is considered a record in itself (Fig. 3). For this reason the universally unique identifier of a Taxon Description (the Scratchpad node identifier) is not enough to uniquely identify each row in the descriptions.txt file. To address this we append the hash (#) character and the field name to the end of the node UUID to create an additional identifier column for projects such as eMonocot which need it.

Figure 3.  

A single Taxon Description node on a Scratchpad corresponds to one or more rows in the description.txt file.

The way that the eMonocot Portal and EoL handle textual descriptions of taxa and narratives on taxon biology is very different and our implementation needs to accommodate both. The eMonocot portal ingests these data using the GBIF Taxon Description extension to DwC-A whereas EoL expects them to be treated as media items associated with a taxon using their own media extension (http://eol.org/schema/media/Document). In order to fulfil our aim of creating a single archive for all consumers we need to provide different definitions of the descriptions file in the meta.xml. For this reason in the meta.xml we define two different rowTypes for the same file (Table 3) – this allows the consumer to choose the definition they prefer without the Scratchpads having to create separate Darwin Core Archives.

The EoL agents file is a non-standard (not conforming to the star schema) extension to DwC-A that records information about people who have contributed to the other content in the archive (e.g. authors of textual content, the photographer of images). As this rowType is only understood by EoL the presence of this file in the archive does not cause problems for other consumers of the DwC-A. The structure of this file is described in Table 4.

For the eMonocot project there is a need for comments on data in the portal and comment replies in the Scratchpad to be synchronised across both platforms. This synchronisation was implemented using the DwC-A, ensuring that this one export mechanism could serve all of the project's needs. This extension to the archive does not conform to the star schema as it does not link directly to the central classification.txt. Despite this the rowType is, at present, only understood by eMonocot, and is ignored by other projects reading the archive. The data model used is a subset of the Open Annotation Data Model (http://www.w3.org/ns/oa#Annotationt). The structure of this file is described in Table 5.

This file contains information from the Scratchpad site's bibliography. The structure of the file is described in Table 6.

This table contains the museum/herbarium specimens and observation records present within a Scratchpad. At present Scratchpads use DwC 1.2.1 as the internal specimen/observation standard. This is currently being upgraded to DwC 1.4. Table 7.

The vernacular names extension records common names, their language, where in the world they are used and any remarks (Table 8).

The eMonocot project makes extensive use of its own unique identifiers derived from the World Checklist of Monocots (WCM). A separate Scratchpads/Drupal module, emonocot_dwca, handles the replacement of the Scratchpads UUID with the eMonocot/WCM identifier. In addition, the licences for taxon descriptive content from eMonocot sites are derived through a link to a publication in the Scratchpad bibliography. A licence field on that publication is used to assign a licence to all the descriptive content that has been copied from that resource. This function is not present in standard Scratchpads, and the emonocot_dwca module ensures that the correct licence is applied to content when it is exported in the DwC-A.

One additional Drush command is provided by the module dwca_rebuild. This function needs to be called twice to build an archive. The first time it is called it creates a number of processes run via the POSIX nohup command (POSIX, the Portable Operating System Interface, is a set of standards for Unix-like operatinng systems) to prevent problems with timeouts when connected to a server via SSH. These processes run through the individual views used to create the DwC-A files and output their contents to text files. The second time it is run the various description views are aggregated into a single file, and this, along with the other exported text files are then zipped to create the archive.

This feature is used by the Scratchpads Team (or the server administrator where the site is not hosted by the Scratchpad project) and is not available to Scratchpads users.

Example usage

From within a site directory: drush dwca_rebuild

Using an alias: drush @cypriioideae.e-monocot.org dwca_rebuild

This module has dependencies on a number of standard Drupal contributed modules (Views, Views Data Export) and also the Darwin Core module developed for the Scratchpads project. While the module may run correctly on a non-Scratchpad Drupal installation this is not a supported use. The module can be enabled on a Scratchpad site through the Structure > Tools interface, which will automatically enable all dependencies.

A number of tools exist that can perform validation on a DwC-A file, for example the tools developed by EoL and GBIF. While these tools are useful they are insufficient for complete verification of an archive. The EoL validator is specifically designed to interpret content destined for the Encyclopedia of Life project, and as such, only reports errors when, for example, licences incompatible with EoL's terms and conditions are used, although in this case there is no requirement for using an EoL licence in the definition of a DwC-A. The EoL validator also reports errors when an image is listed multiple times in the media extension, a situation that may arise when a photograph on a Scratchpad site contains two identifiable species. In contrast the GBIF validator checks for basic compliance with the schema (testing structure rather than content), and no validator is available for the eMonocot project.

In order to provide a consistent and correct DwC-A from a Scratchpad to any of our content partners the decision was made to create our own validator. This validator checks the structure of the archive, ensuring that all files and columns specified in the meta.xml are present in the zip file. Where columns should contain structured content (e.g. a URL, ISSN, ISBN) the content in each row is checked for conformity using regular expression matching. A complete list of checks currently performed is available on the validator wiki.

Checking the structure of the file allows us to check for errors such as column transposition during any future development to the dwca_export module. The precise checking of data in each row also allows us to check the data in the Scratchpad. This gives us the potential to add a feature that warns users of potential errors in the content present within their Scratchpad. For example, running the validator against a number of sites identified several occasions where ISSNs had incorrectly been entered into the ISBN field. At present the validator is run as a stand-alone tool – future work will provide Scratchpad maintainers with a list of identified errors with links to edit content flagged by the validator.

Once the module is enabled the Darwin Core Archive is available for use by our content partners in ViBRANT and eMonocot as well as by anybody else who has an interest in the data. Users of the data must abide by the licence (generally Creative Commons) that the site maintainer has specified. It should be noted that individual rows of data are likely to have different licences.

By default all of the content in the site that is available to the public via the standard web interface is available in the archive. Content creators can hide content from both the Scratchpad website and the Darwin Core Archive by marking that content as 'Unpublished'. It should be noted that due to the caching of web pages and the archives being created by a cron task that content may take a few hours to disappear from both the website and the archive.

At present only content types specified by the Scratchpad team (Specimens, Taxonomy, References, Images, etc.) can be mapped into the Darwin Core Archive. In the future we plan to provide an interface to allow custom content types created by site maintainers to be mapped into the archive.