The Biological and Chemical Oceanography Data Management O ce (BCO-DMO) and the Rolling Deck to Repository (R2R) program are two key data repositories for oceanographic research, supported by the U.S. National Science Foundation (NSF). R2R curates digital data and documentation generated by environmental sensor systems installed on vessels from the U.S. academic research eet, with support from the NSF Oceanographic Technical Services and Arctic Research Logistics Programs. BCO-DMO human-curates and maintains data and metadata including biological, chemical, and physical measurements and results from projects funded by the NSF Biological Oceanography, Chemical Oceanography, and Antarctic Organisms & Ecosystems Programs. These two repositories have a strong connection, and document several thousand U.S. oceanographic research expeditions since the 1970's. Recently, R2R and BCO-DMO have made their metadata collections available as Linked Data, accessible via public SPARQL endpoints. In this paper, we report on these datasets.
Researchers in the geosciences are challenged by the volume and heterogeneity
of data types and formats, and the di culty in discovering, accessing, and
integrating data sets from multiple sources [
Geoscience researchers are therefore seeking methods and tools that allow them to more easily share, discover, access, and reuse data. Currently, a very important role to this end is played by large-scale data repositories, which warehouse data for redistribution and inspection. Each repository usually caters for a specialized subcommunity of researchers, and is highly specialized and focused on particular purposes.
In the meantime, the number of such repositories, which can be accessed on the World Wide Web, abounds. It thus comes as no surprise that they each come with their own modes of access, visualizations, tools, data structures, etc. So, while access to relevant research data is now much easier in principle, diversity and heterogeneity continue to provide signi cant barriers to discovery and access.
At the same time, global issues such as climate change and deforestation, together with a growing understanding of the many interrelationships between di erent subdisciplines, impose the necessity to consider Earth as a single but very complex system. This drives the need to not only discover and access data, but also to integrate information accross elds and disciplines. This importance is witnessed, e.g., by the National Science Foundation's funding of the EarthCube program, which aims at providing \unprecedented data sharing" across the geosciences.1
Linked data, of course, provides a basic means to this end. Unfortunately, while the uptake of linked data in the earth sciences is growing, it also remains relatively slow. But as repository metadata begins to be published as linked data, it gathers momentum due to the additional opportunities provided by publishing in this shared format which decreases the barrier to reuse.
Another advantage of advancing linked data solutions for the geosciences
emerges when considering the sociocultural bene ts. For example, existing data
compilations such as the Global Multi-Resolution Topography synthesis [
In this paper, we present linked datasets providing content from the two key ocean science repositories in the U.S., The Biological and Chemical Oceanography Data Management O ce (BCO-DMO) and the Rolling Deck to Repository (R2R) program. We will rst discuss the speci c relevance of these repositories and their datasets for their research elds (Section 2), then provide more details about the corresponding linked datasets and their availability (Section 3), before concluding (Section 4).
With their global capability and diverse array of sensors, the U.S. academic research eet is an essential mobile observing platform for ocean science. Data collected on every expedition are of high value, especially given the high costs and increasingly limited resources for ocean exploration. The Rolling Deck to Repository (R2R) program2 is funded by NSF to provide stewardship of environmental sensor data routinely collected by the U.S. academic research eet, working in close collaboration with the University-National Oceanographic Laboratory System (UNOLS) and the NOAA National Data Centers.
R2R maintains a catalog of vessels, instrument systems, expeditions, datasets, investigators, organizations, funding awards, cruise reports, and navigation tracks (see Figure 1) { every NSF-funded oceanographic cruise on a vessel in the academic eet creates records in R2R. As such, R2R ensures preservation of and
access to U.S. national oceanographic research data resources, and provides a central gateway through which data from oceanographic expeditions is routinely cataloged and securely transmitted to national long-term archives including the National Geophysical Data Center (NGDC) and National Oceanographic Data Center (NODC). R2R thus provides essential data documentation for each expedition, and tools to improve documentation of the wide array of shipboard data acquisition activities typical of modern expeditions.
R2R also conducts post-cruise quality assessment to document the quality of data as originally delivered from vessels and provides feedback to cruise operators regarding the data quality. The main objective is focused on identifying occurrences of suspicious data, and not to assess the scienti c value of the data. That is, R2R aims to preserve the data and the accompanying metadata to capture as much as possible the orignal intent as they were collected or acquired during expedition. The quality assessment is realized through a series of (mostly) automated tests such as checking whether appropriate metadata exists, searching for possible errors in le formats, as well as collecting summaries of record-level testing of data. All of these are done without making changes to the original raw data les.
As of April 28, 2015, R2R hosts data from 24 in-service vessels, 4,356 cruises, and a total of 18,238,775 archived les. The R2R website has an average of over 60,000 page views per month. 2.2
BCO-DMO The Biological and Chemical Oceanography Data Management O ce (BCODMO)3 was created to serve principal investigators funded by the NSF's Biological Oceanography, Chemical Oceanography and Antarctic Organisms & Ecosystems Programs as a facility where marine biogeochemical and ecological data and information developed in the course of scienti c research can easily be disseminated, protected, and stored on short and intermediate time-frames. The Data Management O ce also provides research scientists and others with the tools and systems necessary to work with marine biogeochemical and ecological data from heterogeneous sources with increased e cacy. To accomplish this, two data management o ces were united in 2006 and enhanced to provide a venue for submission of electronic data and metadata and other information for open distribution via the World Wide Web. The BCO-DMO data system can accommodate many di erent types of data including biological, chemical, and physical measurements and results. The system provides access to the data (numbers, images, and/or documents) in a consistent manner, with su cient metadata, so that others can make full use of these data for their own purposes. The existence of su cient metadata enables the discovery and accurate reuse of data by more than just the initial investigators who collect and process the data. The BCODMO data system is not simply a catalog of data resources, but a system that takes full advantage of a MySQL database storing documentation (metadata)
for each data set, and a data management backend that allows data to reside at multiple sites (including the originating investigator's location if they wish).
The o ce manages existing and new data sets from individual scienti c investigators and collaborative groups of investigators, and continues to make these available online. The o ce works with principal investigators and other data contributors on data quality control; maintains an inventory and program thesaurus of strictly de ned eld names; generates metadata Directory Interchange Format records required by federal agencies; ensures submission of data to national data centers; supports and encourages data synthesis by providing new, online, webbased display tools; and facilitates regional, national, and international data and information exchange. The data being served provide the scienti c investigators with an opportunity to explore the complex and multifaceted data sets wherever they reside world-wide and to collaborate with colleagues in addressing pressing environmental questions, problems, and challenges. The BCO-DMO collection of data sets supports synthesis and modeling activities, reuse of oceanographic data for new research endeavors, availability of \real data" for teachers/students at school and college level to use in their classes, and provides decision-support eld data for policy-relevant issues. Figure 2 shows a sample screen shot.
In terms of data quality, BCO-DMO employs an approach that is laregly people-intensive. Here, BCO-DMO provides data managers who work closely with investigators to ensure su cient metadata are collected and preserved to assist discovery, use, and reuse tasks. Collected metadata include information regarding design of experiments, instruments employed, as well as all the steps in processing eld measurements into the nal form of the data. Beyond the collection, data managers also coordinate closely with data contributors to decide how to organize and present the data in the best way possible. By employing this approach, BCO-DMO feels that higher quality data can be obtained and reused e ectively.
As of April 28, 2015, BCO-DMO hosts 7,490 datasets including information about 1,799 researchers, 2,127 deployments, and 512 projects, that span the full range of oceanographic measurements from research cruises, timeseries sites, laboratory and mesocosm experiments, and synthesis and modeling projects. The BCO-DMO site typically has over 6,500 page views each month. 3 3.1
R2R The R2R linked dataset currently consists of over 530,000 triples, which are accessible via SPARQL Endpoint.4 Machine-readable metadata is available at http://data.rvdata.us/.well-known/void. A Snorql interface is also provided5 for exploring the SPARQL Endpoint, and an entry point URL is provided for Semantic Web browsers.6 A navigable HTML view is also available.7 The SPARQL endpoint is fed from the internal R2R database and is therefore upto-date. Bulk download is possible at http://www.rvdata.us/outgoing/lod/ rvdata.us.20150430.ttl.gz. R2R data are currently under Creative Commons CC BY-NC-SA 3.0 US license.
The RDF graph structure underlying the R2R linked dataset uses a set of
interlinked ontology design patterns which are described elsewhere [
BCO-DMO The BCO-DMO linked dataset9 has machine-readable metadata accessible at http://www.bco-dmo.org/.well-known/void. The whole dataset currently con
sists of over 2,170,000 triples. The triples are accessible via a SPARQL Endpoint and a Virtuoso SPARQL Browser10 is provided for exploring the SPARQL Endpoint. The SPARQL endpoint is fed from the internal BCO-DMO database and is therefore up-to-date. Bulk download is also possible via the URIs pointed to by the void:dataDump property within the machine-readable metadata. BCODMO data are currently under Creative Commons CC BY-SA 3.0 license.
BCO-DMO uses a manually designed ontology for data organization, which
was reported on in [
The reader may suspect some overlaps exist between R2R and BCO-DMO, given that there are actually only dozens of oceanographic research vessels deployed for eld observation, etc. The map-based interfaces also look similar. Indeed, there is a strong partnership between R2R and BCO-DMO, which makes linking their content between each other particularly attractive and potentially impactful. R2R housed data about the vessels, the route navigated during an expedition, as well as narrative description of activities performed during the expedition. It also hosted data obtained from on-board sensors and devices xed to the vessels, such as those from CTD11 instruments or multibeam sensors. On the other hand, data obtained from devices personally brought by the researchers (and thus are not xed permanently to the vessels) are not kept by R2R, but rather by other repositories, particularly BCO-DMO. In this context, R2R and BCODMO are linked two each other via (meta)data about persons and they agree on oceanoraphic cruise identi ers. This linking is of high quality as both data repository maintainers closely cooperate to identify the overlap. For cruise identi ers in particular, there are only about a few dozens research vessels actively used for the U.S. oceanography research, and R2R essentially acts as the gateway of data from the whole eet of vessels before data being deposited and catalogued in other long-term archives. As such, determining the mapping between the two datasets and checking the redundancy become relatively manageable. Furthermore, both linked datasets provide external links to DBpedia, more precisely they map a liations (organizations), scienti c instruments (devices), and research programs to DBpedia using skos:exactMatch links, these were discovered through string matching.
It is important to note that although BCO-DMO has information about cruises, it does not host the detailed navigation data and other kinds of data pertinent to the vessels of which the vessel operators are reposnsible { these are hosted by R2R. BCO-DMO is more focused on data from speci c researchers who run research projects. This means that BCO-DMO would have more detailed data about observations and measurements made during a research expedition. In addition, BCO-DMO does not limit its operation solely on oceanographic data coming from expeditions aboard research vessels, but also those from deployments via other platforms, such as moorings, satellite, land-based platforms, or submarine-based platforms, although oceanographic data from vessel-based expeditions constitute signi cant chunk of the BCO-DMO repository. 4
As Semantic Web technologies are on the rise in applications, the publication of metadata as linked datasets by major geoscience data repositories is likely going to be a driver of future developments. As data becomes available as linked data, its reusability increases, and this includes the development of linked data based 11 conductivity, temperature, and depth of the ocean data discovery and access. In this paper, we have presented the linked datasets providing metadata for the two major oceanographic data repositories, R2R and BCO-DMO.
Besides the obvious potential these linked datasets have for leveraging Semantic Web technologies for the geosciences, these datasets also lend themselves to Semantic Web research, as they pose interesting and challenging problems while at the same time are \real" datasets, as opposed to the often arti cial or academically produced benchmarks. For example, they provide an excellent playground for investigations into ontology matching due to the various degrees of overlap between sub-domains, widely di erent scales, and due to the fact that the utilization of spatio-temporal aspects will likely be critical. They also provide a realistic setting for co-reference resolution problems, solutions of which would have immediate bene cial bene t to the data repositories. Particularly interesting is the fact that, while the datasets are of signi cant size, they still center around a relatively clearly de ned research community, thus certain variables can more easily be controled. Di erent ways to refer to places, e.g. via coordinates or gazetteer names, and di erent ways to refer to chemicals, e.g. by name or formula, etc. provide additional challenging dimensions for co-reference resolution research.
From a much wider perspective, of course, the development of Semantic Web methods and tools for on-the- y integration of major geoscience data repositories would have immediate major impact on the work of geoscientists in practice. Providing linked data for some repositories { or even for most repositories { can only be a very small rst step in this endeavour, which requires major advances in methods. Some EarthCube projects, among them the GeoLink project which the authors are part of, already pursue this vision.
Acknowledgement The presented work has been partially funded by the National Science Foundation under the award 1440202 \EarthCube Building Blocks: Collaborative Proposal: GeoLink-Leveraging Semantics and Linked Data for Data Sharing and Discovery in the Geosciences."