Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Data hosting infrastructure for primary biodiversity data


Published on

Today, an unprecedented volume of primary biodiversity data are being generated worldwide, yet significant amounts of these data have been and will continue to be lost after the conclusion of the projects tasked with collecting them. To get the most value out of these data it is imperative to seek a solution whereby these data are rescued, archived and made available to the biodiversity community. To this end, the biodiversity informatics community requires investment in processes and infrastructure to mitigate data loss and provide solutions for long-term hosting and sharing of biodiversity data.

We review the current state of biodiversity data hosting and investigate the technological and sociological barriers to proper data management. We further explore the rescuing and re-hosting of legacy data, the state of existing toolsets and propose a future direction for the development of new discovery tools. We also explore the role of data standards and licensing in the context of data hosting and preservation. We provide five recommendations for the biodiversity community that will foster better data preservation and access: (1) encourage the community’s use of data standards, (2) promote the public domain licensing of data, (3) establish a community of those involved in data hosting and archival, (4) establish hosting centers for biodiversity data, and (5) develop tools for data discovery.

The community’s adoption of standards and development of tools to enable data discovery is essential to sustainable data preservation. Furthermore, the increased adoption of open content licensing, the establishment of data hosting infrastructure and the creation of a data hosting and archiving community are all necessary steps towards the community ensuring that data archival policies become standardized.

BMC Bioinformatics 2011, 12(Suppl 15):S5 doi:10.1186/1471-2105-12-S15-S5

Published in: Technology
  • DOWNLOAD THIS BOOKS INTO AVAILABLE FORMAT (2019 Update) ......................................................................................................................... ......................................................................................................................... Download Full PDF EBOOK here { } ......................................................................................................................... Download Full EPUB Ebook here { } ......................................................................................................................... Download Full doc Ebook here { } ......................................................................................................................... Download PDF EBOOK here { } ......................................................................................................................... Download EPUB Ebook here { } ......................................................................................................................... Download doc Ebook here { } ......................................................................................................................... ......................................................................................................................... ................................................................................................................................... eBook is an electronic version of a traditional print book THIS can be read by using a personal computer or by using an eBook reader. (An eBook reader can be a software application for use on a computer such as Microsoft's free Reader application, or a book-sized computer THIS is used solely as a reading device such as Nuvomedia's Rocket eBook.) Users can purchase an eBook on diskette or CD, but the most popular method of getting an eBook is to purchase a downloadable file of the eBook (or other reading material) from a Web site (such as Barnes and Noble) to be read from the user's computer or reading device. Generally, an eBook can be downloaded in five minutes or less ......................................................................................................................... .............. Browse by Genre Available eBooks .............................................................................................................................. Art, Biography, Business, Chick Lit, Children's, Christian, Classics, Comics, Contemporary, Cookbooks, Manga, Memoir, Music, Mystery, Non Fiction, Paranormal, Philosophy, Poetry, Psychology, Religion, Romance, Science, Science Fiction, Self Help, Suspense, Spirituality, Sports, Thriller, Travel, Young Adult, Crime, Ebooks, Fantasy, Fiction, Graphic Novels, Historical Fiction, History, Horror, Humor And Comedy, ......................................................................................................................... ......................................................................................................................... .....BEST SELLER FOR EBOOK RECOMMEND............................................................. ......................................................................................................................... Blowout: Corrupted Democracy, Rogue State Russia, and the Richest, Most Destructive Industry on Earth,-- The Ride of a Lifetime: Lessons Learned from 15 Years as CEO of the Walt Disney Company,-- Call Sign Chaos: Learning to Lead,-- StrengthsFinder 2.0,-- Stillness Is the Key,-- She Said: Breaking the Sexual Harassment Story THIS Helped Ignite a Movement,-- Atomic Habits: An Easy & Proven Way to Build Good Habits & Break Bad Ones,-- Everything Is Figureoutable,-- What It Takes: Lessons in the Pursuit of Excellence,-- Rich Dad Poor Dad: What the Rich Teach Their Kids About Money THIS the Poor and Middle Class Do Not!,-- The Total Money Makeover: Classic Edition: A Proven Plan for Financial Fitness,-- Shut Up and Listen!: Hard Business Truths THIS Will Help You Succeed, ......................................................................................................................... .........................................................................................................................
    Are you sure you want to  Yes  No
    Your message goes here
  • Be the first to like this

Data hosting infrastructure for primary biodiversity data

  1. 1. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 RESEARCH Open AccessData hosting infrastructure for primarybiodiversity dataAnthony Goddard1*, Nathan Wilson2, Phil Cryer3, Grant Yamashita4 Abstract Background: Today, an unprecedented volume of primary biodiversity data are being generated worldwide, yet significant amounts of these data have been and will continue to be lost after the conclusion of the projects tasked with collecting them. To get the most value out of these data it is imperative to seek a solution whereby these data are rescued, archived and made available to the biodiversity community. To this end, the biodiversity informatics community requires investment in processes and infrastructure to mitigate data loss and provide solutions for long-term hosting and sharing of biodiversity data. Discussion: We review the current state of biodiversity data hosting and investigate the technological and sociological barriers to proper data management. We further explore the rescuing and re-hosting of legacy data, the state of existing toolsets and propose a future direction for the development of new discovery tools. We also explore the role of data standards and licensing in the context of data hosting and preservation. We provide five recommendations for the biodiversity community that will foster better data preservation and access: (1) encourage the community’s use of data standards, (2) promote the public domain licensing of data, (3) establish a community of those involved in data hosting and archival, (4) establish hosting centers for biodiversity data, and (5) develop tools for data discovery. Conclusion: The community’s adoption of standards and development of tools to enable data discovery is essential to sustainable data preservation. Furthermore, the increased adoption of open content licensing, the establishment of data hosting infrastructure and the creation of a data hosting and archiving community are all necessary steps towards the community ensuring that data archival policies become standardized.Introduction of data [4]. As a result, biodiversity data collected todayToday, an unprecedented volume of primary biodiversity are as endangered as the species they represent. Theredata are being generated worldwide [1], yet significant are also important questions of access to and interpreta-amounts of this data have been and will continue to be tion of the data. Inaccessible data are effectively lostlost after the conclusion of the projects tasked with col- until they are made accessible. Moreover, data that arelecting them [2]. Gray et al. [3] make a distinction misrepresented or easily misinterpreted can result inbetween ephemeral data, which, once collected, can conclusions that are even more inaccurate than thosenever be collected again, and stable data, which can be that would be drawn if the data were simply lost.recollected. The extinction of species, habitat destruc- Although it is in the best interest of all who createtion and related loss of rich sources of biodiversity make and use biodiversity data to encourage best practices toephemeral a significant amount of data that have his- protect against data loss, the community still requirestorically been assumed to be stable. Whether the data additional effective incentives to participate in a sharedare stable or ephemeral, however, poor record keeping data environment and to help overcome existing socialand data management practices nevertheless lead to loss and cultural barriers to data sharing. Separated silos of data from disparate groups presently dominate the cur-* Correspondence: rent global infrastructure for biodiversity data. There are1 Center for Library and Informatics, Woods Hole Marine Biological some examples of projects working to bring the dataLaboratory, Woods Hole, MA 02543, USA out of those silos and to encourage sharing between theFull list of author information is available at the end of the article © 2011 Goddard et al; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
  2. 2. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 2 of 14 projects. Examples include the Global Biodiver- for adopting and implementing data management/pre-sity Information Facility (GBIF) data portal [5] and the servation changes.Encyclopedia of Life (EOL) [6]. Each of them works to We acknowledge that biodiversity data range widely,bring together data from their partners [7,8] and makes both in format and in purpose. Although some authorsthose data available through their application program- carefully restrict the use of the term ‘data’ to verifiableming interfaces (APIs). In addition, there are projects, facts or sensory stimuli [23], for the purposes of thisincluding ScratchPads [9] and LifeDesks [10], that allow article we intend a very broad understanding of thetaxonomists to focus on their area of expertise while term covering essentially everything that can be repre-automatically making the data they want to share avail- sented using digital computers. Although not all biodi-able to the larger biodiversity community. versity data are in digital form, we restrict our Although there is a strong and healthy mix of plat- discussion here to digital data that relate in some way toforms and technologies, there remains a gap in stan- organisms. Usually the data are created to serve a speci-dards and processes, especially for the ‘long tail’ of fic purpose, ranging from ecosystem assessment to spe-smaller projects [11]. In analyzing existing infrastruc- cies identification to general education. We alsoture, we see that large and well-funded projects predic- acknowledge that published literature is a very impor-tably have more substantial investments in tant existing repository for biodiversity data, and everyinfrastructure, making use of not only on-site redun- category of biodiversity data we discussed may be pub-dancy, but also remote mirrors. Smaller projects, on the lished in the traditional literature as well as more digi-other hand, often rely on manual or semi-automated tally accessible forms. Consequently, any of the datadata backup procedures with little time or resources for discussed here may have relevant citations that arecomprehensive high availability or disaster recovery important for finding and interpreting the data. How-considerations. ever, given that citations have such a rich set of stan- It is therefore imperative to seek a solution to this dards and supporting technologies, they will not bedata loss and ensure that data are rescued, archived and addressed here as that would distract from the primarymade available to the biodiversity community. Although purpose of this paper.there are broad efforts to encourage the use of bestpractices for data archiving [12-14], citation of data Categories of data and examples[15,16] and curation [17] as well as large archives that We start this section with a high-level discussion of theare focused on particular types of biology-related data, types of data and challenges related specifically to biodi-including sequences [18-20], ecosystems [21] and obser- versity data. This is followed by a brief review of variousvations (GBIF) species descriptions (EOL), none of these types of existing Internet-accessible biodiversity dataefforts are focused on the long-term preservation of bio- sources. Our intent is not to be exhaustive, but ratherdiversity data. To this end the biodiversity informatics to demonstrate the wide variety of biodiversity datacommunity requires investment in trustworthy processes sources and to point out some of the differencesand infrastructure to mitigate data loss [22] and needs between them. The choice of sites discussed is based onto provide solutions for long-term hosting and storage a combination of the authors’ familiarity with the parti-of biodiversity data. We propose the construction of cular tools and their widespread use.Biodiversity Data Hosting Centers (BDHCs), which are Most discussions of the existing data models forcharged with the task of mitigating the risks presented managing biodiversity data are either buried in whathere by the careful creation and management of the documentation exists for specific systems, for example,infrastructure necessary to archive and manage biodiver- the Architectural View of GBIF’s Integrated Publishingsity data. As such, they will provide a future safeguard Toolkit (IPT) [24] or, alternatively, the data model isagainst loss of biodiversity data. discussed in related informally published presentations, In laying out our vision of BDHCs, we begin by cate- for example John Doolan’s ‘Proposal for a Simplifiedgorizing the different kinds of biodiversity data that Structure for EMu’ [25]. One example of a data modelare found in the literature and in various datasets. We in the peer-reviewed literature is Rich Pyle’s Taxonomerthen lay out some features and capabilities that BDHC [26] system. However, this is again focused on a particu-should possess, with a discussion of standards and best lar implementation of a particular data model for a par-practices that are pertinent to an effective BDHC. ticular system.After a discussion of current tools and approaches for Raw observational data are a key type of biodiversityeffective data management, we discuss some of the data and include textual or numeric field notes, photo-technological and cultural barriers to effective data graphs, video clips, sound files, genetic sequences or anymanagement and preservation that have hindered the other sort of recorded data file or dataset based on thecommunity. We end with a series of recommendations observation of organisms. Although the processes
  3. 3. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 3 of 14 in collecting these data can be complex and Catalog of Fishes [33]. Several sites have begun develop-may be error prone, these data are generally accepted as ing biodiversity data for the semantic web [34] includingthe basic factual data from which all other biodiversity the Hymenoptera Anatomy Ontology [35] and Taxon-data are derived. In terms of sheer volume, raw observa- [36]. In addition, many projects includetional data are clearly the dominant type of data. extensive catalogs of names because names are key toAnother example is nomenclatural data. These data are bringing together nearly all biodiversity data [37]. Exam-the relationships between various names and are thus ples include the Catalog of Life [38], WoRMs [39], ITISvery compact and abstract. Nomenclatural data are gen- [40] and ION [41]. Many sources for names, includingerally derived from a set of nomenclatural codes (for those listed above, are indexed through projects such asexample, the International Code of Botanical Nomencla- the Global Names Index [42] and uBio [43].ture [27]). Although the correct application of these Some sites focus on curated museum collections (forcodes in particular circumstances may be a source of example, Natural History Museum, London [44], andendless debate among taxonomists, the vast majority of the Smithsonian Institution [45], among many others).these data are not a matter of deep debate. However, Others focus on descriptions of biological taxa rangingdescriptions of named taxa, particularly above the spe- from original species descriptions (the Biodiversity Heri-cies level, are much more subjective, relying on both the tage Library [46] and MycoBank [47]), up-to-date tech-historical literature and fundamentally on the knowledge nical descriptions (FishBase [48], the Missouri Botanicaland opinions of their authors. Digital resources often do Garden [49], and the Atlas of Living Australia [50]), anda poor job of acknowledging, much less representing, descriptions intended for the general public (Wikipediathis subjectivity. [51] and the BBC Wildlife Finder [52]). For example, consider a system that records observa- As described above, identifications are often conflatedtions of organisms by recording the name, date, location with raw observations or names. However, there are aand observer of the organism. Such a system can con- number of sites set up to interactively manage newlyflate objective, raw observational data with a subjective proposed identifications. Examples include iSpot [53],association of that data with a name based on some iNaturalist [54], [55] and Nationale Data-poorly specified definition of that name. Furthermore, bank Flora en Fauna [56], which accept observations ofmost of the raw observational data are then typically any organisms, and sites with more restricted domains,discarded because the observer fails to record the raw such as Mushroom Observer [57], eBird [58] or Bug-data that they used to make the determination or Guide [59]. The GBIF data portal provides access tobecause the system does not provide a convenient way observational data from a variety of sources, includingto associate raw data, such as photographs, with a parti- many government organizations. Many sites organizecular determination. Similarly, the person making the their data according to a single classification that reflectsidentification typically neglects to be specific about what the current opinions and knowledge of the resourcedefinition of the name they intend. Although this exam- managers. Examples of such sites whose classificationsple raises a number of significant questions about the come reasonably close to covering the entire tree of lifelong-term value of such data, the most significant issues are the National Center for Biotechnology Informationfor our current purpose relate to combining data from (NCBI) taxonomy [60], the Tree of Life Web [61], themultiple digital resources. The data collected and the Catalog of Life [38], the World Registry of Marine Spe-way that different digital resources are used are often cies (, and the Interimvery different. As a result it is very important to be able Register of Marine and Nonmarine Genera [62]. EOLto reliably trace the origin of specific pieces of data. It is [6] gathers classifications from such sources such andalso important to characterize and document the various allows the user to choose a preferred classification fordata sources in a common framework. accessing its data. Finally, some sites, such as TreeBase Currently, a wide variety of biodiversity data are avail- [63] and PhylomeDB [64], focus on archiving computerable through digital resources accessible through the generated phylogenies based on gene sequences.Internet. These range from sites that focus on photo-graphs of organisms from the general public such as Features of biodiversity data hosting centersgroups within Flickr [28] to large diverse systems Most of the key features of a BDHC are common to theintended for a range of audiences such as EOL [6] to general problem of creating publicly accessible, long-very specialized sites focused on the nomenclature of a term data archives. Obviously, the data stored in thespecific group of organisms, such as Systema Dipter- systems are distinctive to the biodiversity community,orum [29] or the Index Nominum Algarum [30]. Other such as images of diagnostic features of specific taxa.sites focused on nomenclature include the International However, the requirements to store images and thePlant Names Index [31], Index Fungorum [32] and the standards used for storing them are widespread. In
  4. 4. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 4 of 14, significant portions of the data models, and ‘standard’, then the total amount of work that has to bethus communication standards and protocols, are dis- done by community goes up as the square of the num-tinctive to the biodiversity community. For example, the ber of players (roughly speaking, an N2 problem, with Nspecific types of relationships between images, observa- being the number of players). If, however, some com-tions, taxon descriptions and scientific names. munity standards are used, the amount of work for the Government, science, education and cultural heritage community is N×S (with S being the number of stan-communities, among others, are faced with many of the dards used in the community). As S approaches N,same general challenges that face the biodiversity infor- there is no point in standardizing, but as S becomesmatics community when it comes to infrastructure and much less than N, the total amount of support requiredprocesses to establish long-term preservation and access by the community to deal with accounting for the var-of content. The NSF DataNet Program [65] was created ious standards decreases. A full account of standardswith this specific goal in mind. The Data Conservancy and their importance to robust technical systems is[66] and the DataOne Project [67], both funded by beyond the scope of this article. However, we acceptDataNet, are seeking to provide long-term preservation, that the use of standards to facilitate data preservationaccess and reuse for their stakeholder communities. The and access is very important for BDHCs. In this context,US National Oceanographic and Atmospheric Adminis- we present a general overview of common standardstration [68] is actively exploring guidelines for archiving used in biodiversity systems in the next section.environmental and geospatial data [69]. The widerscience and technology community has also investigated Tools, services and standardsthe challenges of preservation and access of scientific Here we summarize some of the tools, services, andand technical data [70]. Such investigations and projects standards that are used in biodiversity informatics pro-are too great to cover in the context of this paper. How- jects. Although an exhaustive list is beyond the scope ofever, every effort should be made by the biodiversity this article, we attempt to present tools that are widelyinformatics community to build from the frameworks known and used by most members of the community.and lessons learned by those who are tackling thesesame challenges in areas outside biodiversity. Tools for replicating data The primary goal of BDHCs is to substantially reduce TCP/HTTP protocols allow simple scripting of com-the risk of loss of biodiversity data. To achieve this goal mand-line tools, such as wget or curl, to transfer data.they must take long-term data preservation seriously. When this is not an option and there is access via aFortunately, this need is common to many other areas login shell such as OpenSSH, other standard tools canof data management and many excellent tools exist to be used to replicate and mirror data. Examples of thesemeet these needs. In addition to the obvious data sto- are rsync, sftp and scp.rage requirements, data replication and effective meta- Peer-to-Peer (P2P) file-sharing includes technologiesdata management are the primary technologies required such as BitTorrent [71], which is a protocol for distri-to mitigate against the dangers of data loss. buting large amounts of data. BitTorrent is the frame- Another key feature of BDHCs is to make the data work for the BioTorrents project [72], which allowsglobally available. Although any website is, in a sense, researchers to rapidly share large datasets via a networkglobally available, the deeper requirements are to make of pooled bandwidth systems. This open sharing allowsthat availability reliable and fast. Data replication across one user to collect pieces of the download from multipleglobally distributed data centers is a well-recognized providers, increasing the efficiency of the file transferapproach to making data consistently available across while simultaneously providing the downloaded bits tothe world. other users. The BioTorrents website [73] acts as a cen- Finally, the use and promotion of standards is a key tral listing of datasets available to download and the Bit-feature for BDHCs. Standards are the key component to Torrent protocol allows data to be located on multipleenabling effective, scalable data transfer between inde- servers. This decentralizes the data hosting and distribu-pendently developed systems. The key argument for tion and provides fault tolerance. All files are then integ-standards in BDHCs is that they reduce the need for a rity checked via checksums to ensure they are identicalset of systems that need to communicate with each on all nodes.other to create a custom translator between each pair ofsystems. Everyone can aim at supporting the standards Tools for querying data providersrather than figuring out how to interoperate with every- OpenURL [74] provides a standardized URL format thatone else on a case by case basis. If the number of enables a user to find a resource they are allowed toplayers is close to the number of standards, then there access via the provider they are querying. Originallyis no point in standardizing. If each player has its own used by librarians to help patrons find scholarly articles,
  5. 5. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 5 of 14 is now used for any kind of resource on the internet. shared space where all nodes in the network supportThe standard supports linking from indexed databases each other and sustain authoritative copies of the origi-to other services, such as journals, via full-text search of nal content. LOCKSS is OAIS (Open Archival Informa-repositories, online catalogs or other services. OpenURL tion System) [79] compliant and preserves all genresis an open tool and allows common APIs to access data. and formats of web content to preserve both the histori- The Biodiversity Heritage Library (BHL)’s OpenURL cal context and the intellectual content. A describedresolver [75] is an API that was launched in 2009 and ‘LOCKSS box’ collects content from the target sitescontinues to offer a way for data providers and aggrega- using a web crawler similar to the ones that searchtors to access BHL material. Any repository containing engines use to discover content. It then watches thosecitations to biodiversity literature can use this API to sites for changes, allows the content to be cached ondetermine whether a given book, volume, article and/or the box so as to facilitate a web proxy or cache of thepage is available online through BHL. The service sup- content in case the target system is ever down and has aports both OpenURL 0.1 and OpenURL 1.0 query for- web-based administration panel to control what is beingmats, and can return its response in JSON, XML or audited, how often and who has access to the material.HTML formats, providing flexibility for data exchange. DiGIR (Distributed Generic Information Retrieval) [80] is a client/server protocol for retrieving informationTools for metadata and citation exchange from distributed resources. Using HTTP to transportOAI-PMH (Open Archives Initiative Protocol for Meta- Darwin Core XML between the client and server, DiGIRdata Harvesting, usually referred to as simply OAI) [76] is a set of tools to link independent databases into a sin-is an established protocol to provide an application gle, searchable virtual collection. Although it was initi-independent framework to harvest metadata. Using ally targeted to deal only with species data, it was laterXML over HTTP, OAI harvests metadata descriptions expanded to work with any type of information and wasof repository records so that servers can be built using integrated into a number of community collection net-metadata from many unrelated archives. The base works, in particular GBIF. At its core, DiGIR provides aimplementation of OAI must support metadata in the search interface between many dissimilar databasesDublin Core format (a metadata standard for describing using XML as a translator. When a search query isa wide range of networked resources), but support for issued, the DiGIR client application sends the query toadditional representations is available. Once an OAI ser- each institution’s DiGIR provider, which is then trans-vice is initially harvested, future harvests will check only lated into an equivalent request that is compatible withfor new or changed records, making it an efficient pro- the local database. Thus the response can deal with thetocol and one that is easily set up as a repetitive task search even though the details of the underlying data-that runs automatically at regular intervals. base are suppressed, thus allowing a uniform virtual CiteBank [77] is an open access repository for biodi- view of the contents on the network. Network speedversity publications published by the BHL that allows and availability were major concerns of the DiGIR sys-sharing, categorizing and promoting of citations. Cita- tem; nodes would time out before requests could betions are harvested from resources via the OAI-PMH processed, resulting in failed queries. The functionalityprotocol, which seamlessly deals with updates and of DiGIR was superceded by the Tapir protocol.changes from remote providers. Consequently, all data BioCASE (The Biological Collection Access Servicewithin CiteBank is also available to anyone via OAI, for Europe) [81] “is a transnational network of biologicalthus creating a new discovery node. Tools are available collections”. BioCASE provides access to collection andthat allow users to upload individual citations of whole observational databases by providing an XML abstrac-collections of bibliographies. Open groups may be tion layer in front of a database. Although its search,formed around various categories and can assist in mod- retrieval and framework is similar to DiGIR, it is never-erating and updating citations. theless incompatible with DiGIR. BioCASE uses a schema based on the Access to Biological CollectionsTools for data exchange Data (ABCD) schema [82].LOCKSS (Lots of Copies Keep Stuff Safe) [78] is an TAPIR is a current Taxonomic Database Workinginternational community program, based at Stanford Group (TDWG) standard that provides an XML APIUniversity Libraries, that uses open source software and protocol for accessing structured data. TAPIR extendsP2P networking technology to map a large, decentra- features of BioCASE and DiGIR by making a more gen-lized and replicated digital repository. Using off-the- eric method of data interchange. The TAPIR project isshelf hardware and requiring very little technical exper- run by a task group that oversees the development. Thetise, the system preserves an institution’s content while XML request and response for access can be stored in aproviding preservation for other content - a virtual distributed database. TAPIR combines and extends
  6. 6. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 6 of 14 of BioCASE and DiGIR to create a more gen- Linking multiple clusters of Hadoop servers globallyeric means for sharing data. A task group oversees the would present biodiversity researchers with a commu-maintenance of the software as well as the protocol’s nity utility. GBIF developers have already worked exten-standard. sively with Hadoop in a distributed computing Although DiGIR providers are still available, their environment, so much of the background research intonumbers have diminished since its inception and now the platform and its application to biodiversity data hasnumber around 260 [83]. BioCASE, meanwhile, has been done. In a post entitled ‘Hadoop on Amazon EC2approximately 350 active nodes [84]. TAPIR is being to generate Species by Cell Index’ [89], GBIF generatedused as the primary collection method by GBIF, who a ‘species per cell index’ map of occurrence data acrosscontinue to maintain and add functionality to the project. an index of the entire GBIF occurrence record store,Institutions such as the Missouri Botanical Garden host where each cell represented an area of one degree lati-all three services (DiGIR, BioCASE and TAPIR) to allow tude by one degree longitude. This map consisted ofongoing open access to their collections. Many projects over 135 million records. This was processed using 20will endeavor to maintain legacy tools for as long as there Amazon EC2 instances running Linux, Hadoop andis demand. This support is helpful for systems that have Java. This map generation was completed in 472 sec-already invested in legacy tools, but it is important to onds and was used to show that processing tasks couldpromote new tools and standards to ensure their adop- be parsed out over many Hadoop instances to come uption as improvements in tools and standards are made. In with a unified result. Because Hadoop runs in Java it isso doing, there needs to be a clear migration path from very simple to provision nodes and link them togetherlegacy systems to the new one. to form a private computing network.Distributed computing Standards used in the biodiversity communityInfrastructure as a service The biodiversity community uses a large range of data-Cloud infrastructure services provide server virtualiza- related standards ranging from explicit data file formats,tion environments as a service. When a user requires a to extensions to metadata frameworks, to protocols. Lar-number of servers to store data or run processing tasks, ger data objects such as various forms of media,they simply rent computing resources from a provider, sequence data or other types of raw data are addressedwho provides the resources from a pool of available by standardized data file formats, such as JPEG [90],equipment. From here, virtual machines can be brought PNG [91], MPEG [92] and OGG [93]. For more specificonline and offline quickly, with the user paying only for forms of data, standards are usually decided by the orga-the actual services used or resources consumed. The nization governing that data type. The Biodiversitymost well known implementation of this is the Amazon Information Standards [94] efforts are intended toElastic Compute Cloud (Amazon EC2) [85]. address the core needs for biodiversity data. The meta- data standard Darwin Core [95] and the related GBIF-Distributed processing developed Darwin Core Archive [96] file format, in par-Using common frameworks to share processing globally ticular, address most of the categories of data discussedenables researchers to use far more computing power above. The current Darwin Core standard is designed tothan had been previously possible. By applying this addi- be used in XML documents for exchanging data and istional computing power, projects can leverage techni- the most common format for sharing zoological data. Itques such as data mining to open up new avenues of is also worth noting that the Darwin Core Archive fileexploration in biodiversity research. After computational format is explicitly designed to address the most com-targets are selected, individual jobs are distributed to mon problems that researchers encounter in this space,processing nodes until all work is completed. Connect- but not necessarily all of the potential edge cases thating disparate system only requires a secure remote con- could be encountered in representing and exchangingnection, such as OpenSSH [86], over the internet. biodiversity data. Apache Hadoop [87] is a popular open source project The TAPIR [97] standard specifies a protocol forfrom Yahoo that provides a Java software framework to querying biodiversity data stores and, along with DiGIRsupport distributed computing. Running on large clus- and BioCASE, provides a generic way to communicateters of commodity hardware, Hadoop uses its own dis- between client applications and data providers. Biodiver-tributed file system (HDFS) to connect various nodes sity-specific standards for data observations and descrip-and provides resilience against intermittent failures. Its tions are also being developed. For example, the EnvOcomputational method, map/reduce [88], passes small [98] standard is an ontology similar to Darwin Core thatfragments of work to other nodes in the cluster and defines a wide variety of environments and some ofdirects further jobs while aggregating the results. their relationships.
  7. 7. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 7 of 14 Community acceptance of standards is an ongoing has already scanned approximately 30 million pages ofprocess and different groups will support various ones core biodiversity literature. The project estimates thatat different times. The process for determining a biodi- there are approximately 500 million total pages of coreversity specific standard is generally organized by the biodiversity literature. Of that, 100 million are out ofBiodiversity Information Standards community (orga- copyright and are waiting to be scanned by the project.nized by TDWG) centered on a working group or mail- Interoperability presents a further technological bar-ing list and often has enough exposure to ensure that rier to data access and interpretation. In many cases pri-those with a stake in the standard being discussed will mary biodiversity data cannot be easily exchangedhave an opportunity to comment on the proposed stan- between a source system and a researcher’s system. Thedard. This process ensures that standards are tailored to standards discussed here help to address this problem,those who eventually use them, which further facilitates but must be correctly implemented in both the sourcecommunity acceptance. Nomenclatural codes in taxon- and destination systems.omy, for example, are standardized, and members of Finally, there is no existing general solution for user-this discipline recognize the importance of these driven feedback mechanisms in the biodiversity space.standards. Misinterpretation of biodiversity data is largely the result of incomplete raw observations and barriers to the pro-Barriers to data management and preservation pagation of data. The process of realizing that two rawTechnological barriers observations refer to the same taxon (or have someWe distinguish between technological and social/cul- more complex relationship) takes effort, as does the pro-tural barriers to effective data sharing, management and pagation of that new data. To be successful, such a sys-preservation. Technological barriers are those that are tem would need to be open to user-driven feedbackprimarily due to devices, methods and the processes and mechanisms that allow local changes for immediate use,workflows that use those devices and methods. Social/ which propagate back to the original source of the data.cultural barriers are those that arise from both explicit Much of this data propagation could be automated by aand tacit mores of the larger community as well as pro- wider adoption of standards and better models for thecedures, customs and financial considerations of the dynamic interchange and caching of data.individuals and institutions that participate in the datamanagement/preservation. Social and cultural barriers The primary technological cause of loss of biodiversity The social and cultural barriers that cause data loss aredata is the poor archiving of raw observations. In many well known and not specific to biodiversity data [99]. Incases raw observations are not explicitly made or they addition to outright, unintentional data loss, there areare actively deleted after more refined taxon description well known social barriers to data sharing. Members ofinformation has been created from them. The lack of the scientific community can be afraid to share dataarchival storage space for raw observations is a key since this might allow other scientists to publishcause of poor archiving, but simply having a policy of research results without explicitly collaborating with orkeeping everything is not scalable. At minimum there crediting the original creator(s) of the data. Oneshould be strict controls governing what data are approach for addressing this concern is forming dataarchived. There also must to be a method for identifying embargoes, in which the original source data are ren-data that either are not biodiversity-related or have little dered unavailable for a specified period of time relativevalue for future biodiversity research. Lack of access to to its original collection or subsequent publication. Inprimary biodiversity data can be due to a lack of techni- addition, there are no clear standards for placing a valuecal knowledge by the creators of the data regarding how on biodiversity data. This results in radically differentto digitally publish them, and often this is caused by a beliefs about the value of such data, and in extremelack of funding to support publication and long-term cases can lead scientists to assert rights over data withpreservation. In cases in which data do get published, expensive or severely limited licensing terms.there remain issues surrounding maintenance of those The financial cost of good archiving and the moredata sources. Hardware failures and format changes can general cost of hardware needed, software developmentresult in data becoming obsolete and inaccessible if con- and the cost of data creation are all other forms ofsideration is not given to hardware and file format social barriers to the preservation of biodiversity dataredundancy. [100]. The cost of creating new raw observations con- Another primarily technological barrier to the access tinues to drop, but some types of raw observational dataand interpretation of biodiversity data is that much of are still extremely expensive or otherwise difficult tothe key literature is not available digitally. Projects such acquire, especially if significant travel or field work areas the BHL are working to address this problem. BHL required to create the data. Similarly, access to museum
  8. 8. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 8 of 14 or living organisms can be crucial to accu- between the recommendations, as having completed onerately interpreting existing data. There are efficient recommendation will make it easier to pursue others.mechanisms to facilitate such access within the aca-demic sphere, but for scientists outside of the academic 1. Short-term: encourage community use of dataworld this can be a significant barrier to creating more standardsmeaningful data. The biodiversity community must work in the short Overcoming these barriers will require effective term with associations such as the TDWG standardsincentives. Ultimately the advantages of having shared, body to strongly promote biodiversity standards in datacentralized access to the data should serve as its own transfer, sharing and citation. Key players, such as dataincentive. This has already happened for genetic providers, must ensure that they work to adopt thesesequence data as reflected by the widespread adoption standards as a short-term priority and encourage otherof GenBank [18]. In comparison, biodiversity data have projects to follow suit. The adoption of standards isa much greater historical record than sequence data paramount to the goal of achieving sustainable data pre-and more importantly a set of legacy conventions that servation and access and paves the way for interoper-were largely created outside of the context of digital ability between datasets, discovery tools and archivingdata management. As a result, investment in accurately systems [104]. The processes for standards establish-modeling these conventions and providing easy-to-use ment and community acceptance need to be as flexibleinterfaces for processing data that conform to these and inclusive as possible, to ensure the greatest adoptionconventions is likely to have greater return in the long of standards. Therefore, an open, transparent and easilyrun. Providing more direct incentives to the data crea- accessible method for participating in the standards pro-tors could be valuable as way to get the process cess is important to ensure that standards meet thestarted. community’s needs. An interesting example is the use of unique identifiers. Popular solutions in the biodiversityLegal and mandatory solutions community include GUIDs [105], URLs [106] and LSIDsMajor US grant funding organizations including the [107]. This debate has not resolved on a single agreed-National Institutes of Health [101] and the National on standard and it is not clear that it ever will. There-Science Foundation [102] are now requiring an up-front fore, it is important that BDHCs support all such sys-plan for data management, including data access and tems, most likely through an internally managedarchiving as part of every grant proposal. The National additional layer of indirection and replication of theScience Foundation is taking steps to ensure that data sources when possible.from publicly funded research are made public [103].Such policies, although designed to ensure that data are 2. Short-term: promote public domain licensing ofmore accessible, have implications for data archiving. contentMandatory data archiving policies are likely to be very To minimize the risk of loss, biodiversity data, includingeffective in raising awareness of issues surrounding data the images, videos, classifications, datasets and data-loss and archiving. However, such policies are neither bases, must all be replicable without asking permissionstrictly necessary to ensure widespread adoption of data of any rights holders associated with that data. Sucharchiving best practices, nor are they a sufficient solu- data must be placed under as liberal a license as possibletion on their own. Adoption of these policies will assist or, alternatively, placed into the public domain. Thisin ensuring that a project’s data are available and position should be adopted in the short term by existingarchivable. data providers. To avoid long-term issues, these licenses should allow additional data replication without seekingRecommendations and discussion permission from the original rights holders. A variety ofHere we recommended five ways in which the commu- appropriate licenses are available from Creative Com-nity can begin to implement the changes required for mons [108]. ‘Non-commercial’ licenses should besustainable data preservation and access: (1) Encourage avoided if possible because this makes it more difficultthe community’s use of data standards, (2) promote the for the data to be shared and used by sites that maypublic domain licensing of data, (3) establish a commu- require commercial means to support their projects. Innity of those involved in data hosting and archival, (4) these cases, an open, ‘share-alike’ license that preventsestablish hosting centers for biodiversity data, and (5) the end user from encumbering the data with exclusivedevelop tools for data discovery. licenses should be used. We prioritize recommendations here from short-term Open source code should be stored in both an online(immediate to 1 year), medium-term (1 to 5 years), to repository, such as github [109] or Google code [110],long-term (5 or more years). There is inter-dependency and mirrored locally at a project’s institution for
  9. 9. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 9 of 14 Making open source code publicly available opening them up to a larger community. Allowing com-is an important and often overlooked step for many pro- munity input into biodiversity datasets via comments,jects and must be encouraged by funders, not only for direct access to the data, adding to the data or derivingthe sake of community involvement in a project, but to new datasets from the original data are all powerfulallow others to maintain and further develop the soft- ways of enhancing the original dataset and building aware should a project’s funding conclude. The use of community that cares about and cares for the source software to support the hosting and devel- Wikipedia, which allows users to add or edit definitionsopment of biodiversity informatics tools and projects of any article, is an example of a successful model on ashould also be encouraged by funders. Using common- large scale. In the case of biodiversity, there has beenplace, open source software lowers the barrier to entry success with amateur photographers contributing theirfor collaborators, further increasing the chances of a images of animals and plants and amateur scientistsproject being maintained after funding has ended. helping to identify various specimens. Crowd-sourcing is particularly notable as a scalable approach for enhancing3. Short-term: develop data hosting and archiving data quality. Community input can also work effectivelycommunity to improve the initial quality of the data and associatedEncouraging the use of standards and data preservation metadata [114].practices to the wider biodiversity community requires In 2008, the National Library of Australia, as part ofan investment in training resources. In the short term, the Australian Newspapers Digitization Programthe community, including projects, institutions and (ANDP) and in collaboration with Australian state andorganizations such as GBIF and TDWG, should estab- territory libraries, enabled public access to selected out-lish online resources to familiarize researchers with the of-copyright Australian newspapers. This service allowsprocess of storing, transforming and sharing their data users to search and browse over 8.4 million articlesusing standards, as well as with the concepts of data from over 830,000 newspapers dating back to 1803.preservation and data hosting. Such resources may Because the papers were scanned by Optical Characterinclude screen casts, tutorials and white papers. Even Recognition (OCR) software, they could be searched.better, intensive, hands-on training programs have pro- Although OCR works well with documents that haveven to be effective in educating domain experts and consistent typefaces and formats, historical texts andscientists in technical processes, as can be seen in newspapers fall outside of this category, leading to lesscourses such as the Marine Biology Laboratory’s Medi- than perfect OCR. To address this, the National Librarycal Informatics course, which teaches informatics tools opened up the collection of newspapers to crowd-sour-to medical professionals [111]. Funders and organiza- cing, allowing the user community the ability to correcttions such as GBIF should encourage the establishment and suggest improvements to the OCR text. The volun-of such courses for biologists and others working in bio- teers not only enjoyed the interaction, but also felt thatdiversity. Partners in existing large-scale informatics they were making a difference by improving the under-projects, such as those of the Encyclopedia of Life or standing of past events in their country. By March 2010the Biodiversity Heritage Library, as well as institutions over 12 million lines of text had been improved bysuch as GBIF regional nodes, would be well placed with thousands of users. The ongoing results of this workthe expertise and resources to host such courses. An can be seen on the National Library of Australia’s Trovealternative hands-on method that has been effective is site [115]. Biodiversity informatics projects should con-to embed scientists within informatics groups [112]. sider how crowd-sourcing could be used in their pro- For hosts of biodiversity data, technical proficiency is jects and see this as a new way of engaging users andimportant for any information technology project and the general archiving and sharing are concepts that most sys- Development community to enable biodiversitytems engineers are capable of facilitating. Understanding sharing. A community of developers that have experi-the need for knowledge transfer, a communal group ence with biodiversity standards should be identifiedthat is comfortable with email and mailing lists is criti- and promoted to assist in enabling biodiversitycal for a project’s success. Communication between the resources to join the network of BDHC. In generalgroup can be fostered via the use of open/public wiki, a these developers should be available as consultants forshared online knowledge base that can be continuously funded projects and even included in the original grantedited as the tools, approaches and best practices applications. This community should also consider achange. level of overhead funded work to help enable data Community input and crowd-sourcing. Crowd-sour- resources that lack a replication strategy to join the net-cing [113] is defined as the process of taking tasks that work. A working example of this can be found in taskswould normally be handled by an individual and groups such as the GBIF/TDWG joint multimedia
  10. 10. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 10 of 14 task group, which is working to evaluate and the low latency and high bandwidth available when datashare standards for multimedia resources relevant to are served from a location local to the user. For exam-biodiversity [116]. ple, a project in Europe and a project in Australia who Industry engagement. Industry has shown, in the both serve each other’s data will serve both projects’case of services such as Amazon’s Public Dataset project data from the closest respective location, reducing band-[117] and Google’s Public Data Explorer [118], that width and latency complications when accessing thethere is a great benefit in providing users with access to data. The creation of specialized data archival and host-large public datasets. Even when access to a public data- ing centers, centered on specific data types, would pro-set is simple, working with that dataset is often not as vide for specialized archival of specific formats. Forstraightforward. Amazon provides a way for the public example, with many projects hosting DwC archive filesto interact with these datasets without the onerous within a central hosting center, the center could providegroundwork usually required. Amazon benefits in this services to maintain the data and keep it in a currentrelationship by selling its processing services to those and readable format. By providing this service to a largedoing the research. Engaging industry to host data and quantity of data providers, the costs to each providerprovide access methods provides tangible benefits to the would be significantly lower than if the data providerwider biodiversity informatics community, in terms of were to maintain these formats alone. Such centersits ability to engage users in developing services on top would host both standardized, specific formats and pro-of the data, with dramatically lower barriers to entry vide for a general data store area for raw data that arethan traditional methods, and from data archival or yet to be transformed into a common format.redundancy perspectives. In establishing data hosting infrastructure, the com- munity should pay attention to existing efforts in data4. Medium-term: establish hosting centers for biodiversity management standards and hosting infrastructure, suchdata as those being undertaken by the DataONE project, andKey players, such as institutions with access to appropri- find common frameworks to build biodiversity data spe-ate infrastructure, should act in the medium term to cific solutions.establish biodiversity data hosting centers (BDHCs). Geographically distributed processing. The sheerBDHCs would mostly probably evolve out of existing size of some datasets often makes it difficult to moveprojects in the biodiversity data community. These cen- data offsite for processing and makes it difficult to con-ters will need to focus on collaborative development of duct research if these datasets are physically distantsoftware architectures to deal with biodiversity-specific from the processing infrastructure. By implementingdatasets, studying the lifecycle of both present and common frameworks to access and process these data,legacy datasets. Consideration should be made for these the need to run complex algorithms and tools acrossdata centers to be situated in geographically dispersed slow or unstable internet connections is reduced. It islocations to provide an avenue for load sharing of data assumed that more researchers would access data thanprocessing and content delivery tasks as well as physical would have the resources to implement their own sto-redundancy. By setting up mechanisms for sharing rage infrastructure for the data. Therefore it is impor-redundant copies of data in hosting centers around the tant for projects to consider the benefits of placingworld, projects can take advantage of increased proces- common processing frameworks on top of their datasing power, whereby the project distributes their com- and providing API access to those frameworks in orderputing tasks to other participating nodes, allowing to provide researchers who are geographically separatedinformatics teams to more effectively use their infra- from the data access to higher speed processing.structure at times when it would otherwise be Data snapshots. Data hosting centers should trackunderused. archival data access and updates in order to understand When funding or institutional relationships do not data access patterns. Given the rapid rate of change inallow mutual hosting and sharing of data, and as a stop datasets, consideration needs to be given to the conceptgap before specialized data hosting centers become of ‘snapshotting’ data. In a snapshot, a copy of the stateavailable, third party options should be sought for the of a dataset is captured at any particular time. The mainbackup of core data and related code. By engaging the dataset may then be added to, but the snapshot willservices of online, offsite backup services, data backup always remain static, thereby providing a revision historycan be automated simply and at a low cost when com- of the data. An example use case for snapshotting apared with the capital expense of backup hardware dataset is an EOL species page. Species pages are dyna-onsite. mically generated datasets comprising user submitted When mutual hosting and serving of data is possible, data and aggregated data sources. As EOL data aregeographically separated projects can take advantage of archived and shared, species pages are updated and
  11. 11. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 11 of 14 Being able to retrieve the latest copy of a data- online data backup. These services would be designed toset may not be as important as retrieving an earlier ver- deal with biodiversity-related data specifically. By tailor-sion of the same dataset, especially if relevant data have ing solutions to the various file formats and standardsbeen removed or modified. Archival mechanisms need used in the biodiversity informatics community, issuesto take snapshotting into consideration to ensure that such as format obsolescence can be better managed. Forhistorical dataset changes can be referenced, located and example, when backing up a mysql file or DarwinCorerecovered. This formalized approach to data stewardship archive file, metadata associated with that file can bemakes the data more stable and, in turn, more valuable. stored in the backup. Such metadata could include the Data and application archiving. Although some data version of DarwinCore or mysql or the date of file crea-exist in standardized formats (for example xls, xml and tion. These data can be later used when migrating for-DwC), others are available only through application-spe- mats in the event of a legacy format becoming obsolete.cific code, commonly a website or web service with Such migrations would be difficult for a large number ofnon-standardized output. Often, these services will store small projects to consider, so by centralizing this pro-datasets in a way that is optimized for the application cess a greater number of projects will have access tocode and is unreadable outside the context of this code. such migrations and therefore the associated data.Whether data archiving on its own is a sustainableoption or whether application archiving is also required 5. Long-term: develop data discovery toolsneeds to be considered by project leaders. Applications The abundance of data made available for public con-that serve data should have a mechanism to export sumption provides a rich source for data-intensivethese data into a standard format. Legacy applications research applications, such as automated markup, textthat do not have the capability to export data into a extraction, machine learning, visualizations and digitalstandard format need to be archived alongside the data. ‘mash-ups’ of the data [119]. As a long-term goal for theThis can be in a long-term offline archive, where the GBIF secretariat and nodes, as well as biodiversity infor-code is simply stored with the data, or in an online matics projects, this process should be encouraged as aarchive where the application remains accessible along- natural incentive. Three such uses of these applicationsside the data. In both cases, BDHC need to store appli- are in geo-location, taxon finding and text-mining.cation metadata, such as software versions and Services for extracting information. By extractingenvironment variables, both of which are critical to place names or geographic coordinates from a dataset,ensuring that the application can be run successfully geo-location services can provide maps of datasets andwhen required. Re-hosting applications of any complex- easily combine these with maps of additional datasets,ity can be a difficult task as new software deprecates fea- providing mapping outputs such as Google Earth’s [120]tures in older versions. One solution that BDHC should kmz output. Taxon finding tools can scan large quanti-consider is the use of virtual machines, which can be ties of text for species names, providing the dataset withcustomized and snapshotted with the exact environment metadata describing which taxa are present in the datarequired to serve the application. However, such solu- and the location within the data in which each taxontions should be seen as a stop-gap. It should be empha- can be found. Using natural language processing, auto-sized that a focus on raw data portability and mated markup tools can analyze datasets and return aaccessibility is the preferred solution where possible. structured, marked-up version of the data. uBio RSS Replicate indexes. Online data indexes, such as those [121] and RSS novum are two tools that offer automatedfrom GBIF and from the global names index, provide a recognition of species names within scientific and mediamethod for linking online data to objects, in this RSS feeds. By identifying data containing species names,instance to names. Centrally storing the index enables new datasets and sources of data can be discovered.users to navigate through the indexes to the huge data- Further development of these tools and integration ofsets in centralized locations. These data are of particular this technology into web crawling software will pave thevalue to the biodiversity data discovery process as well way for automated discovery of online biodiversity dataas being generally valuable for many types of biodiver- that might benefit from being included in the recom-sity research. It is recommended that these index data mended network of biodiversity data stores. The devel-be replicated widely, and mirroring these indexes should opment of web services and APIs on top of datasets andbe a high priority when establishing BDHC. the investment in informatics projects that harvest and Tailored backup services. Given the growing quantity remix data should all be encouraged, both as a means toof data and informatics projects, there is scope for enable discovery within existing datasets and as anindustry to provide tailored data management and incentive for data owners to publish their data publicly.backup solutions to the biodiversity informatics commu- GBIF, funding agencies and biodiversity informatics pro-nity. These tools can range from hosting services to jects should extend existing tools while developing and
  12. 12. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 12 of 14 new tools that can be used to enable data data. The community needs to recognize and proac-discovery. The community should develop a central reg- tively encourage this evolution by developing a sharedistry of such tools and, where practical, provide hosting understanding of what is needed to ensure the preser-for these tools at data hosting centers. vation of biodiversity data and then acting on the Services to enable data push. As new data are pub- resulting requirements. We see the recommendationslished, there is a lag between the publication time and laid out in this paper as a step towards that sharedthe harvesting of this data by other services, such as understanding.EOL or Google. Furthermore, even after articles arepublished, data are in most cases available only in AcknowledgmentsHTML format, with the underlying databases rendered This article has been published as part of BMC Bioinformatics Volume 12inaccessible. Development of services to allow new data Supplement 15, 2011: Data publishing framework for primary biodiversityto be ‘pushed’ to online repositories should be investi- data. The full contents of the supplement are available online at http://www. Publication of the supplementgated, as this would make the new data available for was supported by the Global Biodiversity Information Facility.harvesting by data aggregation services and data hostingcenters. Once the data are discovered they can then be Author details 1 Center for Library and Informatics, Woods Hole Marine Biologicalfurther distributed via standardized APIs, removing this Laboratory, Woods Hole, MA 02543, USA. 2Encyclopedia of Life, Center forburden from the original data provider. Library and Informatics, Woods Hole Marine Biological Laboratory, Woods Once data aggregators discover new data, they should Hole, MA 02543, USA. 3Center for Biodiversity Informatics (CBI), Missouri Botanical Garden, St Louis, MO 63119, USA. 4Center for Biology and Society,cache the data and include them in their archival and Arizona State University, Tempe, AZ 85287, USA.backup procedures. They should also investigate howbest to recover the content provided by a content provi- Competing interests The authors declare that they have no competing interests.der should the need arise. The EOL project follows thisrecommended practice and includes this caching func- Published: 15 December 2011tionality. This allows EOL to serve aggregated datawhere the source data are unavailable, while at the same References 1. Scholes RJ, Mace GM, Turner W, Geller GN, Jürgens N, Larigauderie A,time providing the added benefit of becoming a redun- Muchoney D, Walther BA, Mooney HA: Towards a global biodiversitydant back-up of many projects’ data. This is an invalu- observing system. Science 2008, resource should a project suffer catastrophic data 2. Güntsch A, Berendsohn WG: Biodiversity research leaks data. Create the biodiversity data archives! In Systematics 2008 Programme and Abstracts,loss or go offline. Göttingen 2008. Göttingen: Göttingen University Press;Gradstein SR, Klatt S, Normann F, Weigelt P, Willmann R, Wilson R 2008:.Conclusion 3. Gray J, Szalay AS, Thakar AR, Stoughton C, Vandenberg J: Online Scientific Data Curation, Publication, and Archiving. 2002 [ the biodiversity data community grows and standards com/apps/pubs/default.aspx?id=64568].development and access to affordable infrastructure is 4. Whitlock MC, McPeek MA, Rausher MD, Rieseberg L, Moore AJ: Datamade available, some of the above-mentioned challenges archiving. Am Nat 2010, 175:145-146. 5. Global Biodiversity Information Facility Data Portal. [].will be met in due course. However, the importance of 6. Encyclopedia of Life. [].the data and the real risk of data loss suggest that these 7. GBIF: Data Providers. [].challenges and recommendations must be faced sooner 8. Encyclopedia of Life: Content Partners. []. 9. ScratchPads. [].rather than later by key players in the community. A 10. LifeDesks. [].generic data hosting infrastructure that is designed for a 11. Heidorn PB: Shedding light on the dark data in the long tail of science.broader scientific scope than biodiversity data may also Library Trends 2008, 57:280-299. 12. Digital Preservation Coalition. [].have a role in access and preservation of biodiversity 13. Library of Congress Digital Preservation. [ Indeed, the more widespread the data becomes, gov].the safer it will be. However, the community should 14. Image Permanence Institute. []. 15. The Dataverse Network Project. [].ensure that the infrastructure and mechanisms it uses 16. DataCite. [].are specific enough to biodiversity datasets that they 17. Digital Curation Centre. [].meet both the short-term and long-term needs of the 18. GenBank. []. 19. DNA Data Bank of Japan. [].data and community. As funding agencies requiring data 20. European Molecular Biology Laboratory Nucleotide Sequence plans becomes more widespread, the biodi- [].versity informatics community will become more 21. The US Long Term Ecological Research Network. [ ].attuned to the concepts discussed here. 22. Klump J: Criteria for the trustworthiness of data-centres. D-Lib Magazine As the community evolves to take a holistic and col- 2011, doi:10.1045/january2011-klump.lective position on data preservation and access, it will 23. Zins C: Conceptual approaches for defining data, information, and knowledge. Journal of the American Society for Information Science andfind new avenues for collaboration, data discovery and Technology 2007, re-use, all of which will improve the value of their
  13. 13. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 13 of 14 IPT Architecture PDF. [ 72. Langille MGI, Eisen JA: BioTorrents: a file sharing service for scientific downloads/detail?name=ipt-architecture_1.1.pdf&can=2&q=IPT data. PLoS ONE 2010, 5:e10071[ +Architecture]. 10.1371/journal.pone.0010071].25. Doolan J: Proposal for a Simplified Structure for Emu.[http://www. 73. BioTorrents. []. 74. OpenURL. []. 20American/john%20doolan_SimplifiedStructure.pps]. 75. BHL OpenURL Resolver Help. [ Pyle R: Taxonomer: a relational data model for managing information openurlhelp.aspx]. relevant to taxonomic research. PhyloInformatics 2004, 1:1-54. 76. Open Archives Initiative. [].27. International Code of Botanical Nomenclature (Vienna Code) adopted by 77. CiteBank. []. the seventeenth International Botanical Congress Vienna, Austria, July 78. LOCKSS. []. 2005. Ruggell, Liechtenstein: Gantner Verlag;McNeill J, Barrie F, Demoulin V, 79. OAIS. []. Hawksworth D, Wiersema, J 2006:. 80. Distributed Generic Information Retrieval (DiGIR). [http://digir.sourceforge.28. Flickr. []. net].29. Systema Dipterorum. []. 81. Biological Collection Access Services. [].30. Index Nominum Algarum Bibliographia Phycologica Universalis. [http:// 82. ABCD Schema 2.06 - ratified TDWG Standard. []. tdwg/CODATA/Schema/].31. International Plant Names Index. []. 83. The Big Dig. [].32. Index Fungorum. []. 84. List of BioCASE Data Sources. [ Catalog of Fishes. []. providers_list.cfm].34. Berners-Lee T, Hendler J, Lassila O: The Semantic Web. Sci Am 2001, 29. 85. Amazon Elastic Compute Cloud (Amazon EC2). [ Hymenoptera Anatomy Ontology. []. ec2/].36. TaxonConcept. []. 86. OpenSSH. [].37. Patterson DJ, Cooper J, Kirk PM, Remsen DP: Names are key to the big 87. Apache Hadoop. []. new biology. Trends Ecol Evol 2010, 25:686-691. 88. Dean J, Ghemawat S: MapReduce: simplified data processing on large38. Catalogue of Life. []. clusters. OSDI’04: Sixth Symposium on Operating System Design and39. World Register of Marine Species. []. Implementation 2004 [].40. Integrated Taxonomic Information System. []. 89. Hadoop on Amazon EC2 to generate Species by Cell Index. [http://41. Index to Organism Names. []. Global Names Index. []. html].43. uBio. []. 90. Joint Photographic Experts Group. [].44. Natural History Museum Collections. [ 91. W3C Portable Network Graphics. []. collections]. 92. Moving Picture Experts Group. [].45. Smithsonian Museum of Natural History: Search Museum Collection 93. []. Records. []. 94. Biodiversity Information Standards (TDWG). [].46. Biodiversity Heritage Library. []. 95. Darwin Core. [].47. Myco Bank. []. 96. Darwin Core Archive Data Standards. [ Fishbase. []. standards-and-tools/publishing-data/data-standards/darwin-core-archives/].49. Missouri Botanical Garden. []. 97. TAPIR (TDWG Access Protocol for Information Retrieval). [http://www.50. Atlas of Living Australia. [].].51. Wikipedia. []. 98. Environmental Ontology. [].52. BBC Nature Wildlife. []. 99. Kabooza Global backup survey: About backup habits, risk factors,53. iSpot. []. worries and data loss of home PCs. 2009 [ iNaturalist. []. globalsurvey.html].55. Artportalen. []. 100. Hodge G, Frangakis E: Digital Preservation and Permanent Access to56. Nationale Databank Flora en Fauna. []. Scientific Information: The State of the Practice. 2004 [http://www.dtic.57. Mushroom Observer. []. mil/cgi-bin/GetTRDoc?Location=U2&doc=GetTRDoc.pdf&AD=ADA432687].58. eBird. []. 101. National Institutes of Health. [].59. BugGuide. []. 102. National Science Foundation. [].60. National Center for Biotechnology Information Taxonomy. [http://ncbi. 103. Scientists Seeking NSF Funding Will Soon Be Required to Submit Data]. Management Plans. [ Tree of Life Web Project. []. cntn_id=116928&org=NSF&from=news].62. Interim Register of Marine and Nonmarine Genera. [ 104. Harvey CC, Huc C: Future trends in data archiving and analysis tools. Adv au/irmng/]. Space Res 2003, 32:347-353, doi: 10.1016/S0273-1177(03)90273-0.63. TreeBASE. []. 105. Leach P, Mealling M, Salz R: A universally unique identifier (UUID) URN64. PhylomeDB. []. namespace. 2005 [].65. Sustainable Digital Data Preservation and Access Network Partners 106. Uniform Resource Locators (URL). Berners-Lee T, Masinter L, McCahill M (DataNet). []. 1994 [].66. Data Conservancy. []. 107. OMG Life Sciences Identifiers Specification. [ DataONE. []. html].68. National Oceanographic and Atmospheric Administration (NOAA). 108. Creative Commons. []. []. 109. GitHub. [].69. Preliminary Principles and Guidelines for Archiving Environmental and 110. Google Code. []. Geospatial Data at NOAA: Interim Report by Committee on Archiving 111. Bennett-McNew C, Ragon B: Inspiring vanguards: the Woods Hole and Accessing Environmental and Geospatial Data at NOAA. USA: experience. Med Ref Serv Q 2008, 27:105-110. National Research Council; 2006. 112. Yamashita G, Miller H, Goddard A, Norton C: A model for bioinformatics70. Hunter J, Choudhury S: Semi-automated preservation and archival of training: the marine biological laboratory. Brief Bioinform 2010, 11:610-615, scientific data using semantic grid services. In Proceedings of the Fifth IEEE doi:10.1093/bib/bbq029. International Symposium on Cluster Computing and the Grid - Volume 01 113. Howe J: Crowdsourcing: Why the Power of the Crowd Is Driving the (CCGRID ‘05). Volume 1. Washington, DC: IEEE Computer Society; Future of Business. Random House Business 2008. 2005:160-167. 114. Hsueh P-Y, Melville P, Sindhwani V: Data quality from crowdsourcing: a71. Bittorrent. []. study of annotation selection criteria. Proceedings of the NAACL HLT
  14. 14. Goddard et al. BMC Bioinformatics 2011, 12(Suppl 15):S5 Page 14 of 14 Workshop on Active Learning for Natural Language Processing. Computational Linguistics 2009, 27-35[ id=1564131.1564137].115. National Library of Australia’s Trove. [].116. GBIF/TDWG Multimedia Resources Task Group. [http://www.keytonature. eu/wiki/MRTG].117. Public Data Sets on Amazon Web Services. [ publicdatasets/].118. Google Public Data Explorer. [].119. Smith V: Data Publication: towards a database of everything. BMC Research Notes 2009, 2:113.120. Google Earth. [].121. uBIo RSS. []. doi:10.1186/1471-2105-12-S15-S5 Cite this article as: Goddard et al.: Data hosting infrastructure for primary biodiversity data. BMC Bioinformatics 2011 12(Suppl 15):S5. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at