Digitization Workflows for Paleontology Collections

Palaeontologia Electronica palaeo-electronica.org

Digitization workflows for paleontology collections

Talia S. Karim, Roger Burkhalter, Úna C. Farrell, Ann Molineux, Gil Nelson, Jessica Utrup, and Susan H. Butts

ABSTRACT

The development of digitization workflows is an essential part of any formalized large-scale digitization program. Paleontological collections literature has addressed the need for, and utility of, digitized collections for nearly four decades, but no modern, community-vetted set of digitization workflows to accomplish this goal has been widely adopted. With the advent of the U.S. National Science Foundation’s (NSF) Advancing the Digitization of Biodiversity Collections (ADBC) program in 2011, iDigBio, NSF’s national coordinating center for facilitating digitization, in collaboration with broad community representation from numerous institutions, launched a series of working groups to address workflow development across all major preparation types. Workflow modules have been developed for pre-digitization curation, data entry, imaging objects (catalogs, field notes and other materials not stored with specimens, labels, two- and three-dimensionally preserved specimens), image processing, and proactive digitization. Modules and the tasks they include may be implemented in any order and customized for specific configurations and institutional parameters. The workflows are made publicly available for download and customization at GitHub and via the iDigBio documentation pages. A review of platforms for electronic data publishing through online aggregators, a crucial step in any digitization program, is also provided.

Talia S. Karim. University of Colorado Museum of Natural History, Boulder, Colorado 80503, USA. [email protected] Roger Burkhalter. Sam Noble Museum, University of Oklahoma, 2401 Chautauqua Avenue, Norman, Oklahoma 73072, USA. [email protected] Úna C. Farrell. Department of Geological Sciences, Stanford University, 450 Serra Mall, Stanford, California 94305, USA. [email protected] Ann Molineux. Non-vertebrate Paleontology Lab, University of Texas, 10100 Burnet Road, Austin, Texas 78758, USA. [email protected] Gil Nelson. Department of Biological Sciences, Florida State University, Tallahassee, Florida 32303, USA. [email protected] Jessica Utrup. Yale University, Peabody Museum of Natural History, Division of Invertebrate Paleontology, 170 Whitney Avenue, PO Box 208118, New Haven, Connecticut 06520, USA. [email protected] Susan H. Butts. Yale University, Peabody Museum of Natural History, Division of Invertebrate Paleontology, 170 Whitney Avenue, PO Box 208118, New Haven, Connecticut 06520, USA. [email protected]

Karim, Talia S., Burkhalter, Roger, Farrell, Úna C., Molineux, Ann, Nelson, Gil, and Butts, Susan H. 2016. Digitization workflows for paleontology collections. Palaeontologia Electronica 19.3.4T: 1-14 palaeo-electronica.org/content/2016/1569-digitization-workflows

Keywords: workflow; digitization; imaging; data publishing; paleontology; collections

Submission: 4 May 2015 Acceptance: 2 August 2016

INTRODUCTION BACKGROUND The development and documentation of effec- Paleontologists started discussing “electronic tive workflows is a primary component in the digiti- data processing” nearly four decades ago, fore- zation of neontological and paleontological shadowing the advent of digitization and digital collections (Granzow-de la Cerda and Beach, data storage and display (see e.g., Glenister, 1977; 2010; Vollmar et al., 2010; Beaman and Cellinese, Mello, 1969; Mello and Collier, 1972) and many 2012; Haston et al., 2012; Nelson et al., 2012; paleontology collections have had some form of Weiss, 2013; Wolniewicz, 2009). Nevertheless, digital specimen database for decades. The Pale- there is no current workflow documentation outlin- ontological community has also long recognized ing best or recommended practices for the digitiza- the need for publishing these data and making tion of paleontological collections. Until recently, them discoverable via aggregation sites. The discussions dealing with computerization focused PaleoPortal, initially funded in 2003 by the National mostly on electronic database design (e.g., Morris, Science Foundation, included one of the early 2000) or defining the role digitization should play in attempts at aggregation of digital museum records the paleontology community (MacLeod and Gural- via DiGIR protocol. It served as a single access nick, 2000). In the last decade, large scale digitiza- point for searching multiple collections across the tion projects in the United States (with iDigBio) and United States. In May 2012, Integrated Digitized Europe (with the EUROPEANA Portal and more Biocollections (iDigBio), Yale Peabody Museum, recently, Synthesis 3) have initiated a number of Botanical Research Institute of Texas, and the Bio- conversations and collaborations about how best diversity Institute at the University of Kansas co- to capture specimen data quickly, publish the data sponsored the Developing Robust Object-to- in an easily accessible manner with sufficient data Image-to-Data (DROID) workflow workshop in quality (for review see Smith and Blagoderov, Gainesville, FL (iDigBio, 2012a) for the purpose of 2012), with discipline-specific workflows developed designing consensus workflows across various and published (e.g., herbaria, Nelson et al., 2015) domains within the biodiversity collections commu- as a product of these initiatives. However, digitiza- nity. Concomitant with the planning and execution tion of paleontology collections is often not straight- of this workshop, staff from iDigBio conducted a forward, as specimens vary in size and series of on-site visits to a number of museum and dimensionality, no published guide for standard academic collections to document existing digitiza- specimen orientations exists, specimens often sit tion workflows within natural history collections on top of labels obscuring data, and data is often (Nelson et al., 2012). The preliminary results of this stored in multiple formats and in multiple locations survey were incorporated into the DROID work- (e.g., labels with specimens; locality cards, field shop (Nelson, 2012). The primary outcome of the notes, and ledgers away from specimens). A case DROID workshop was a series of working groups study by Weiss (2013) involving collaboration focused on the production and distribution of between a paleontology collection and library sci- preparation-specific digitization workflows for (1) entists recognized several additional concerns flat sheets and packets, (2) pinned objects in trays regarding digitization of paleontological collections, and drawers, and (3) 3-D objects in spirits, jars, including preservation and access of data, meta- drawers, and boxes (iDigBio, 2012b). Nelson and data and thoroughness of the digital record, and Karim both participated in the original DROID work- the fidelity of rendering fossils in 2-D and 3-D shop and the subsequent development of work- images. We address these issues with the generic flows by the 3-D objects working groups. This workflow modules presented herein and offer sug- working group eventually split into two separate gestions on how to turn some of these challenges working groups to develop more specific workflows into opportunities for more comprehensive digitiza- for fluid preserved specimens and fossil speci- tion of a collection. mens, respectively. The iDigBio Paleontology Digi- tization Working Group also organized a series of workshops/symposia (Table 1) and webinars that

2 PALAEO-ELECTRONICA.ORG

TABLE 1. Summary of iDigBio Paleontology Digitization Working Group workshops and symposia contributing to workflow development.

WORKSHOP/MEETING DETAILS LOCATION & DATE Developing Robust Obect-to-Image-to- Sponsors: iDigBio, Yale Peabody University of Florida, Data (DROID) Workshop Museum, Botanical Research Institute of Gainesville, FL Texas, Biodiversity Institute-University of May 2012 Kansas. Goal: design consensus workflows across various collection types. Paleontology Digitization WorkshopSponsors: iDigBio, Yale Peabody Yale Peabody Museum, Museum, Fossil Insect Collaborative New Haven, CT TCN. September, 2013 Goal: discuss digitization efforts across the paleontological community, gather feedback on DROID workflow development. Paleontology Imaging WorkshopSponsors: iDigBio, Jackson School of University of Texas, Geoscience-University of Texas. Austin, TX Goal: share and demonstrate various April, 2014 imaging techniques used for fossil specimens, including 3D scanning, CT scanning, and photogrammetry, and gather feedback on DROID workflow development. Specify for Paleontology CollectionsSponsors: iDigBio, Biodiversity Biodiversity Institute- Institute-University of Kansas. University of Kansas, Goal: provide training on how to use the Lawrence, KS Specify 6 collections database software May, 2014 for paleontology collections, gather feedback on how to structure “paleo context” data within the Specify 6 data model, and gather feedback for DROID workflow development.

Digitization Symposium- Advancing the Sponsors: iDigBio, GSA-Geoinformatics Geological Society Digitization of Paleontology and Division, Paleontological Society, of America (GSA) Geoscience Collections: Projects, Society for the Preservation of Natural Annual Meeting, Programs, & Practices History Collections, University of Colo- Vancouver, BC rado Museum of Natural History, Yale October, 2014 Peabody Museum. Outcome: 34 oral and 4 poster presen- tations sharing implementations of digitization projects across the geosci- ences.

addressed various aspects of paleontological spec- approach to workflow development. The four pri- imen digitization with the aim of gathering ideas mary modules for which workflows were developed from the broader paleontological community about are listed in Figure 1. their workflows and to draw these efforts together Figure 2 presents, in modified business pro- into a comprehensive document of workflow mod- cess modeling format, one implementation of these ules. The results of these working groups were modules in an illustrative workflow that essentially then synthesized into generalized workflow mod- parallels the sequence listed above. The figure ules for digitizing various aspects of paleontologi- depicts the main workflow in the central lane with cal collections and they are presented herein. one-time and episodic tasks shown in adjacent par- allel lanes. It also emphasizes (with a loop symbol) DIGITIZATION WORKFLOWS which of the modules encompass iterative processes. Note that quality control and conservation Following the lead of DROID, output from pre- are associated with more than one module, as will vious iDigBio-sponsored workflow working groups, likely be the case in any specific implementation. and the digitization task clusters enumerated by These workflows are not introduced as best Nelson et al. (2012), the paleontology digitization practices, a term that we define to mean broadly workflows working group pursued a modular vetted, rigorously tested, and critically measured

3 KARIM ET AL.: DIGITIZATION WORKFLOWS

Workflow Modules Module 0: Pre-Digitization Curation Module 1: Data Entry Module 2: Imaging Module 2A: Imaging catalogs, field notes, and other materials not stored with specimens Module 2B: Imaging labels associated with specimens Module 2C: Imaging three-dimensionally preserved specimens Module 2D: Imaging two-dimensionally compressed specimens Module 2E: Image processing Module 3: Proactive Digitization

FIGURE 1. List of workflow modules developed by the iDigBio DROID and Paleontology working groups.

protocols and procedures in use with substantial grams, and other differences between institutions, consistency across a community of users. Instead, a modular approach facilitates institutional adapta- the workflows presented here constitute a set of tions and encourages the creation of workflows recommended practices drawn from numerous customized to the institution (Haston et al., 2012). published and unpublished sources and known Editable and PDF versions of these workflows are practices of institutional leaders in the digitization also available for download and customization at: of neontological and paleontological specimens; github.com/iDigBioWorkflows/PaleontologyDigiti- they are reflective of common, successful, and zationWorkflows. implementable practices. For simplicity in accounting for all tasks in a The format of each module is standardized particular module, we have used a linear rather into four columns and n rows (Table 2). Column 1 than iterative representation, although iterative includes the Task ID of each task to be accom- sub-processes are occasionally referenced within plished in the workflow. Tasks are numbered a particular task’s explanations and comments sec- sequentially. Column 2 contains the name of the tion (see third column of Table 2). For instance, in task, sometimes expanded into a short statement Module 2D–Imaging Two Dimensional Com- that describes the activity to be accomplished. Col- pressed Fossils, steps T4-T11 are listed as being umn 3 elaborates on the tasks; often expressing iterative in the explanations and comments section variations in strategies for executing the task, for step T4. This set of steps, which includes tasks details of the task, and further explanations of what such as lens setup, specimen cleaning and posi- the task requires and/or is intended to accomplish. tioning, and adjusting lighting, can be completed in Column 4 itemizes the resources needed to a continuous series for each specimen and then accomplish the task, mentions relevant institutional repeated for multiple specimens, or can be protocols and manuals, and frequently includes lit- adjusted as needed for a particular project. erature citations, links to related websites, and Although this approach risks obscuring embedded resources such as software and equipment. iterative sequences within a particular module, it The modules address all aspects of digitiza- reflects our goal of providing a comprehensive tion and provide a framework within which an menu of digitization tasks while leaving specific undigitized collection could incorporate curatorial implementations and order of execution at the dis- work and conservation. They also provide alterna- cretion of users. tive approaches for those currently digitizing but Some processes referenced in the workflows uncertain how to handle a particular type of media, (e.g., focus stacking) are dependent on wide varia- and provide other solutions to digitizing problems tions in specimen size and composition, making a experienced in their present protocol. The detailed universally relevant accounting of workflow steps documentation for these modules is available at difficult. We have highlighted these processes iDigBio’s website (iDigBio, 2016). within the workflows, usually providing generic The decision to use a modular approach in the guidelines for implementation accompanied by rel- development of these workflows stems from their evant references. Recommendations of specific generic nature. Given the breadth in database software is beyond the scope of these workflows, management systems, imaging equipment and but we do reference common software used in configurations, institutional infrastructure, varia- paleontological collections digitization. tions in physical facilities, goals for digitization pro-

4 PALAEO-ELECTRONICA.ORG

FIGURE 2. Example implementation of the workflow modules described herein.

Collections Digitization Policy/Strategic Plan asset plan template created by the National Museum of Natural History is useful in developing Tasks within several of the workflow modules such a plan: www.idigbio.org/wiki/images/2/20/ suggest the existence of an institutional strategic NMNH_Digital_Asset_Plan_Template.pdf. plan or a collection-level policy or framework for Decision making at the institutional level can guidance in making curatorial and digitization- also be useful in finding solutions to long-term data related decisions. Detailed, well thought out, and storage and data publishing. The University of Col- written digitization policies are important in plan- orado Museum of Natural History has worked with ning a digitization program (Scoble, 2010). They the campus Research Computing Facility to are commonplace in many biodiversity collections acquire large amounts of storage for images and (New York Botanical Garden, 2004) and recom- other digital data being generated by a wide variety mended herein as an essential component for of digitization projects across the museum. Several ensuring consistency of practice across staff and museum staff members have continued to work time. Their development should precede the with Research Computing to develop data man- launch of a digitization program, and their revision agement tools desired by the museum as a whole. should be ongoing to reflect emerging technolo- This relationship between Research Computing gies, new discoveries in techniques and pro- and the museum was made possible by leveraging cesses, and continual refinement of workflows. support from museum leadership and cross disci- Topics to be covered within these documents plinary recognition at the collections-level of the include: digitization rationale, digitization goals, need for long-term data storage. Similarly, Yale data and metadata standards (Häuser et al., 2005), University has developed a collaborative frame- image standards, technology replacement and work for supporting the lifecycle management and data management plans, project planning and use of Yale's digital assets. Collections from multi- management, a digital asset management plan, ple units within the university can now be searched and standards for data publication. The digital

5 KARIM ET AL.: DIGITIZATION WORKFLOWS

TABLE 2. Example of modular workflow format from Module 0: Pre-Digitization Curation and Setup.

Module 0: Pre-Digitization Curation and Setup Task ID Task NameExplanations and Comments Resources T1 Prioritize specimens, Varies by institution. Should follow Institutional policy, collection objects, institutional digitization policies and project guidelines, ledgers, field notes, and guidelines. active research catalogs to digitize. criteria, etc.

T2 Note damage to object to Route to conservation workflow as Institutionally specific be digitized that needs necessary, based on institutional policy or curation guidelines. immediate attention. curatorial practices.

T3 Update specimen This may happen prior to the digitization According to taxonomy (and related of any taxonomic group. Some institutions institutional protocol authority files) as update specimens with expert and procedures necessary. determinations prior to digitization. Others and/or project record determinations from the label in requirements. anticipation of community involvement in helping correct determinations. from a single platform that combines digitized infor- within a single collection or institution bear the mation from the art galleries, libraries, academic same identifier, and the uniqueness of these identi- departments, and the Yale Peabody Museum. This fiers is usually enforced by database conventions. platform can be explored at: dis- For instance, many database programs require cover.odai.yale.edu/ydc/. that entries into the catalog number field be unique, thus the software will not allow you to create multi- Workflow Modules ple specimen records with the same catalog num- Pre-digitization curation and set up. The pre- ber. Using this built-in safety net works well for digitization curation and set up module (Module 0) ensuring that duplicate catalog numbers are not constitutes a group of tasks to be accomplished assigned within a single collection. prior to digitization (Figure 2). This module GUIDs on the other hand, are universally addresses numerous curatorial tasks that might unique and ensure that no two specimens across otherwise be put off until time is available, resulting the universe of collections bear the same identifier. in substantial opportunity for collection improve- As records of specimens are aggregated with ment. Topics covered include: assessing curatorial those from other institutions, the use of GUIDs will and digitization priorities; selecting and configuring ensure that all references to a specimen can be a database management system; updating con- resolved to a single physical object stored in a spe- trolled taxonomies within a database; making and cific natural history collection. The critical point updating taxonomic determinations; defining data here is that the GUIDs should be assigned to the entry standards; determining the types of objects to specimen records by their initial creators. Retroac- digitize (e.g., specimens, labels, field notes, led- tively physically affixing a GUID to every specimen gers, catalogs, etc.); assessing and noting speci- in a collection is not feasible, especially for large men/collection object damage, and repairing if time collections. A more realistic option is for a GUID to allows and damage is minor, or routing to conser- be assigned to each electronic specimen record by vation if necessary; determining image file naming the institution creating the record, and in fact many conventions; assigning/updating local catalog num- collections management software packages bers; selecting and implementing a strategy for already do just this (e.g., Specify [Specify Software creating, assigning, and associating globally Project, University of Kansas, Lawrence; speci- unique identifiers (GUID) to specimens/collection fyx.specifysoftware.org/] and EMu [Axiell Group, objects. Sweden; emu.kesoftware.com/] will automatically We emphasize the importance of assigning assign a GUID to each specimen record). The criti- catalog numbers at this stage before digitization cal point here is that the GUIDs should be has started as this piece of information is required assigned to the specimen records by their initial for most digital records (e.g., specimen records in a creators and not secondarily by subsequent data database) to be created. These locally unique cata- aggregators so as to avoid records possessing log numbers also ensure that no two specimens multiple GUIDs. The GUID can then be tracked

6 PALAEO-ELECTRONICA.ORG

back to the physical specimen via the institutionally from images of those paper records (Figure 2) unique catalog number, which is also a part of the (Farrell, 2014). Specific data entry protocols will specimen record. largely depend on the database software in use. In Many different types of GUIDs exist (see e.g., the majority of cases data are entered by hand on Page, 2009; Richards et al., 2011; Guralnick et al., a record-by-record basis, or by uploading previ- 2014), but there is no clear consensus on which ously prepared spreadsheets (see, for example, type should be used broadly for natural history col- Module 3: Proactive Digitization below). Incorporat- lections (see Guralnick et al., 2015). One possible ing voice recognition software into the workflow option for a resolvable identifier for fossil speci- may help speed up data entry, although it has been mens would be to use an International Geo Sample used primarily for data capture during imaging Number (IGSN) (www.geosamples.org/igsnabout) (Butts, 2013) and to streamline file renaming after in place of an automatically generated GUID. In the image capture. Currently, optical character recog- case of IGSNs, the sample is registered with a nition (OCR) software is not widely used in paleon- unique nine-digit alphanumeric code that serves as tological collections, and is not included in this the GUID. A key aim of the IGSN system is to pro- module due to the difficulties associated with hand- vide a robust way to link other data, such as pub- written labels and the need to handle specimens to lished analytical results, back to the original remove the label, which is generally located under- physical sample or specimen. The Non-vertebrate neath the specimen. OCR has been successfully Paleontology Lab at the University of Texas at Aus- implemented in other disciplines where labels are tin (UT-NPL) has started using IGSNs and has easily visible, such as herbarium sheets (Barber, adopted them into their digitization workflow as a 2012; Barber et al., 2013; Haston et al., 2012; Laf- way to link specimens back to other types of data. ferty and Landrum, 2009), and may prove effective In practice, the actual assignment of GUIDs is for some paper records in paleontological collec- most likely to occur as part of Module 1: Data tions, such as typed locality cards or catalog led- Entry, and not as part of Module 0. However, we gers. Additionally, once paper records have been recommend that the strategy for their creation and imaged, collections could take advantage of crowd- assignment should happen as part of Module 0. sourcing projects, such as Notes from Nature Guralnick et al. (2015) discuss several ways to (www.notesfromnature.org/), which allow citizen deal with the application of GUIDs to new and leg- scientists to transcribe data. acy data and also advocate that whatever type of In addition to taxonomic and geographical GUID is used, that it should be resolvable. In prac- data, paleontologists have to consider geological tice, however, many paleontology collections are time and stratigraphic information. Relational data- opting for the blanket assignment of GUIDs to bases that accommodate these data elements specimens by their collections management soft- (e.g., Arctos (arctosdb.org/), EMu, Specify) deal ware (e.g., University of Kansas, University of Col- with the relationships between the specimen orado, and University of Texas using Specify; Yale record, the locality, and the geological context in Peabody Museum and Smithsonian National different ways, in part a reflection of the lack of Museum of Natural History using EMu). This consensus within the paleontological community. option, in which the databasing software automati- When choosing a software program it is important cally generates and associates a GUID (in the form to consider not only how closely the database of a universally unique identifier (UUID)) with each structure matches local institutional records, but specimen record in the database, is becoming also which version will allow for the most efficient widespread because it offers an easy and quick databasing and subsequent data querying (see solution to creating an identifier that can then be Karim and Farrell, 2014; Morris, 2005). mapped to the Darwin Core occurrenceID field Once data have been entered, they can be (Darwin Core Task Group, 2015) when data are subsequently augmented and updated (Figure 2). published. Whether auto-generated UUIDs or In particular, many institutions have separate geo- resolvable identifiers such as IGSNs or URIs (e.g., referencing workflows, which may take advantage Page, 2008) become the standard has yet to be of georeferencing tools that are integrated into the determined. database (e.g., GEOLocate in Specify). In other Data entry. The data entry module (Module 1) cases, it can be more efficient to georeference deals with the process of entering specimen, local- localities before uploading spreadsheets of data in ity, and associated data into a database, whether order to take advantage of collaborative georefer- from catalog/ledger sheets, specimen labels, or encing projects. Previously digitized source materi-

7 KARIM ET AL.: DIGITIZATION WORKFLOWS als (e.g., field notebooks; see Module 2A) may also stored separate from specimens and are often be attached once specimen records have been imaged independent of related specimens. Recom- created (iDigBio, 2014a). mended equipment includes scanners or digital Maintaining clean data is a constant concern, SLRs, depending upon whether the pages can be and quality control steps are recommended at mul- laid flat or perhaps cut and fed through a sheet tiple stages in this module. In some collections, feeder, or must be kept tightly bound. Whichever student databasing staff prepare spreadsheets, method is chosen, a file naming procedure, file for- which can be checked by a qualified supervisor mat, and relevant metadata fields to capture should prior to upload. In other cases, or additionally, a be determined before starting. A data quality check combination of standard queries designed to catch of both image and associated metadata is an mistakes (e.g., unexpectedly blank fields, contra- important part of the workflow. dictory information) and spot-checking are Module 2B addresses the digitization of labels employed to catch and correct any errors within the stored with specimens/collection objects; they may database. be imaged with specimens or separately. The Imaging objects. The imaging module (Module 2) workflow assumes that only labels are being is designed for large-scale imaging projects that imaged although it recommends that label imaging will allow researchers and the public to gain a bet- should be coincident with specimen imaging to ter understanding of material held in a particular minimize specimen handling. Maintaining appropri- collection and the condition of that material. This ate connection between label and specimen is a module consequently does not cover detailed priority; ideally both label and specimen should research-quality imaging (e.g., SEM). It does cover carry the specimen number so that they can be suggested equipment and methods for effective easily reassociated. It is useful if the file naming imaging of several types of objects as well as protocol matches that used for the specimens. aspects of file naming, file formats, image process- Labels, especially historic labels, may be fragile ing, and file storage and archiving. Transporting and may need to be repaired before imaging. They and storing materials for imaging is included in the may also require cleaning to reveal more legible workflow. The module is subdivided into several handwriting. Technicians should be trained in rele- sub-modules based on the type of object being vant conservation techniques. imaged; these are listed in Figure 1. Each sub- Separate workflows are presented for imaging module requires similar steps to prepare and three-dimensionally (Module 2C) and two-dimen- arrange the material for imaging, quality control, sionally (Module 2D) preserved specimens, with and return of specimens to their original location in the main differences being that the former includes the repository. Variations in workflow and method- steps for blackening and whitening, more compli- ology arise from the type of material imaged. cated specimen mounting, and accounts for imag- File naming protocols vary widely among col- ing the specimen in multiple views. The use of lections, but should at least include the catalog focus stacking (also referred to as focal plane number and ensure that there are no spaces, peri- merging, z-stacking, or focus blending) is included ods, or special characters in file names. A stan- in both workflows and will depend on several vari- dardized file and folder naming system was ables, including the size and three-dimensional implemented at CUMNH that includes that catalog relief of the specimen, equipment, and final desired number and the lens and magnification at which image quality. While focus stacking is primarily uti- the image was acquired. This file name helps with lized for three-dimensionally preserved specimens, semi-automated scale bar insertion, bulk image it can also be useful for capturing fine details on upload into Specify, and downstream data quality two-dimensionally preserved compression fossils control issues. File names at the SNOMNH also (e.g., fossil insects and plants). We realize that include view for 3-D specimens (e.g., dorsal, ven- these two workflow scenarios do not fully cover all tral, lateral). UT-NPL also use a standardized file types of fossil specimens, but they do cover many naming protocol, which includes the catalog num- types, and they are intended to serve as a guide ber and view. Yale Peabody Museum employs the for developing workflows for additional preparation institutional plus division acronyms as a prefix, the types. catalog number, and an abbreviated object orienta- Before specimen imaging is undertaken it is tion as the suffix. important to consider the types of specimens being Module 2A covers the digitization of cards, imaged and what type of images are required for ledgers, and field notebooks, which are typically the digitization project, as this will inform many

8 PALAEO-ELECTRONICA.ORG tasks in the final workflow. For instance, should the processed in a macro that generates (1) composite specimens be imaged from multiple angles? images of specimens at different orientations and Should relief and other details be enhanced (such (2) two spreadsheets for automated import to EMu as whitening with ammonium chloride sublimate, (Butts, 2013). The first spreadsheet is used for see Teichert, 1948; Kier et al., 1965; Jago, 1973; importing images to the EMu Multimedia module, Marsh and Marsh, 1975; Feldmann, 1989; Zachos, and the second spreadsheet associates each mul- personal commun., 2008; Hegna, 2010) prior to timedia record to its object record. imaging? Will all specimens in a lot be imaged or We reiterate that while many photographic just a single exemplar? Will individual specimens techniques are available to researchers, these on a slab be imaged separately or will one wider modules are designed for capturing a large number shot of the whole slab suffice? Will focus stacking of high-quality images in a short amount of time be utilized for more three-dimensionally preserved that can be of use to researchers, teachers, and specimens? How will the lighting be arranged? the general public. They are not necessarily Strongly directed light can bring out surface relief in intended to be publication quality, although in some specimens, but can also introduce artifacts (Haug cases they are. We expect that researchers will and Haug, 2014). There are obvious advantages to assess the type and condition of material held having clear written guidelines on such questions using these images; higher quality images can before the imaging begins for each project. then be requested as "virtual loans" or the physical The initial steps of the specimen imaging specimen can be requested so that more special- workflows guide the user through some of these ized imaging (e.g., stereo-pair images, reflectance basic questions, and continue with how specimens transformation imaging (RTI), infrared imaging) will be transported to the imaging area, camera can be undertaken by the researcher. and software setup, specimen selection and clean- Module 2E outlines the tasks associated with ing, and image capture, and rehousing. Tasks such image processing, such as processing image as specimen transport might seem trivial, but can stacks, applying file names, inserting scale bars, become important if the collection is housed on a archiving images, and creating derivatives. Possi- different floor, or even building, from the imaging ble variations in implementation of these tasks fol- station; a well thought out plan is necessary from low. If stacking software is being utilized, images the start of the digitization project. Imaging labels are typically processed (i.e., stacked) immediately can also be done in conjunction with imaging spec- after the images are captured so that any speci- imens, as is noted in these workflow modules, and mens that are out of focus can be reimaged before might be a more efficient way of capturing two dif- they are put away. In some instances, stacks are ferent types of collections data at once. The batch processed overnight, with the caveat that KUMIP and UT-NPL rely heavily on label data for some may need to be reimaged. Archiving "mas- database record entry and quality control, and ter" files and creating smaller derivatives to be label images have become an important part of quickly and easily distributed to users are critical quality control in their data entry module. steps in this module, and we recommend that insti- We also recognize the opportunity for using tutions establish policies governing these pro- voice recognition software to capture data (e.g., cesses to ensure efficiency and consistent results catalog numbers and specimen views) concur- (Crick, 2005; iDigBio, 2012c). The use of digital rently with specimen imaging as part of this mod- scale bars and other text inserted into an image ule. Relevant information can be read into a digital during this module varies widely. Many institutions spreadsheet for later reuse in file naming or poten- skip this step by using a physical scale bar inserted tially primary data capture. An added bonus to at the time of image capture, while other institu- using voice recognition software for data capture tions insert digital scale bars into images as part of while imaging is that the ability to enter data derivative production. The UT-NPL uses a semiau- “hands-free” facilitates manipulation of the speci- tomatic process to insert a digital scale bar based men during imaging. The Yale Peabody Museum on the scale embedded in the image file. The (YPM) has pioneered the use of voice recognition CUMNH utilizes a script to retrieve the catalog software in paleontology collections (Butts, 2013). number, lens, and magnification information, all Data types captured include: image number, cata- contained in the file name, to insert a caption into log number, and suffix abbreviation indicating the the lower left corner of the image that include the orientation of the specimen (ventral, dorsal, hinge, catalog number and digital scale bar in black text etc.). The data collected on the spreadsheet are over a small white rectangle. This task is partially

9 KARIM ET AL.: DIGITIZATION WORKFLOWS automated by using actions and batch processing captured in the field and imported using the Work- features in Adobe Photoshop. Reviewing existing bench, making subsequent specimen databasing community standards (see e.g., image guidelines progress more rapid, especially in cases where the from AntWeb [California Academy of Sciences, specimens themselves cannot be cataloged in the 2016]) can be helpful when developing this module field. The Smithsonian NMNH is also utilizing simi- at the institutional level, and we encourage other lar methods with EMu FIMS (Field Information institutions to share this information so that broader Management System) in conjunction with a work- community standards can be developed. flow that efficiently transfers data from the field into Proactive digitization. Given the trend toward the EMu database (Hollis, personal commun., greater field-based digitization through which at 2015). The University of Kansas Entomology Col- least some steps in the digitization process occur lections (Short, 2014), VSC herbarium at Valdosta in the field (e.g., recording geographic coordinates State University (Carter, personal commun., 2013), of collection locality) and the sometimes extensive Southwestern Adventist University (Woods and recording of digital images and data of in situ spec- Chadwick, 2014), and the Gray Fossil Site (Wood- imens, the workflows presented here include a ward and Compton, 2014) are also good examples module (Module 3) for what we term proactive digi- of collections that are proactively capturing data in tization. We define this as digitization accom- the field. plished by collectors or field assistants at the time Proactive digitization can also utilize elec- of specimen collection, and prior to deposition in a tronic interfaces designed for Trimble™ or other museum or academic collection, effectively moving geolocation devices; custom interfaces for elec- digitization activities forward in the collecting and tronic tablets, smartphone applications, and similar accessioning processes. In some cases, this might devices; or direct entry into local copies of or wire- include recording text data into an electronic lessly connected collection databases. In some spreadsheet or other input device that can be eas- instances this may entail assigning and recording ily imported into a collection database. This is a rel- catalog numbers and other identifiers in the field atively new and rapidly developing method, but and creating or appropriating existing locality one that fits especially well within paleontology dig- descriptions at the time of collection, depending on itization workflows. Such strategies are currently in the type of collection object being acquired. We use in some domains as follows. anticipate that technological improvements will Ichthyology collection managers at the enhance field collection protocols, significantly Museum of Comparative Zoology at Harvard, for improve digitization efficiency by reducing the example, attempt to collaborate with individual col- quantity of data to be digitized post-collection, and lectors prior to field excursions and generate a that such protocols are likely to become common basic, pre-formatted data entry worksheet that practice in the near future. A recent study in Paleo- matches the museum's database bulk loader and anthropology demonstrates such developments in is also compatible with and complementary to the digital data collection in the field (Reed, et al., collector's workflow (Williston, personal commun., 2015). 2013). A similar project is underway at UT-NPL Data Publishing where spreadsheet formats currently used for pre- database data entry by UT-NPL students is being Although we do not specifically address this adapted for dispersal to all graduate students and topic as a stand-alone module in the digitization faculty who are generating data for future addition workflows described herein, the publication of data to the repository. To encourage use, the template is the logical follow up to all digitization workflows. will be available on the web page The sharing of digitized collections data has long (www.jsg.utexas.edu/npl/research/), the wiki been a bottleneck in digitization workflows (wikis.utexas.edu/display/specify6/Proactive+digiti- (MacLeod and Guralnick, 2000). Most museums zation-field+data), and physically distributed on a today have a digital specimen catalog; however, in flash drive. Users will be able to add additional many cases these data are only searchable inter- fields of relevance, which will further improve the nally by museum staff. Museums rely heavily on IT data fields that are incorporated into the Specify staff (if they exist) to develop online data publishing database. It is a pre-Specify Workbench tool allow- tools and make data available for harvesting. Shar- ing for refinement of data prior to upload. The ing datasets online with the research community Workbench tool in Specify is a useful portable tool. has recently been made much easier due to the For example, data for new locality records can be development of publishing tools (e.g., the Global

10 PALAEO-ELECTRONICA.ORG

Biodiversity Informatics Facility’s Integrated Pub- marine and terrestrial invertebrate fossils. Docu- lishing Toolkit [IPT]), the growth of online data menting and sharing a combined set of the digitiza- aggregators (e.g., iDigBio, VertNet, GBIF), and the tion workflows, protocols, and policies produced by development of data sharing standards (e.g., Dar- members of the DROID working group, many of winCore). IPT is a public web server that is rela- whom are also part of these earliest TCNs, tively easy for a systems administrator to install ensures the availability of tested, vetted, and rec- and can be hosted locally at an institution. Data- ommended practices as new digitization programs sets exported from a local digital specimen catalog, come online. The workflows documented herein typically in a .csv format, can be easily uploaded will encourage coalescence of community-wide using the IPT web interface and detected and standardized processes and serve to alleviate data ingested by a variety of data aggregators, including gaps that result from institutional variation in digiti- those listed above. Aggregators make these data zation practices (Nelson, 2015). We hope that the available for searching, which relieves the local community will benefit from the well-designed and institution of the need for a data portal and the bur- easily implementable set of flexible, modular work- den of publishing. flows that captures what has been gleaned from Some institutions express concern about pub- the several successful paleo digitization programs. lishing data about sensitive fossil localities, including precise georeferences and detailed locality ACKNOWLEDGMENTS descriptions (Weiss, 2013; Barton, et al., 2006), We thank iDigBio for sponsoring the DROID based on concerns for site preservation, parame- workshop, GSA digitization session, and use of ters stipulated by confidential agreements with Adobe Connect. We also thank our home institu- landowners or donors, and legal restrictions. As tions for their continued support of our digitization digitization has become more prevalent, this has efforts. iDigBio is funded by a grant from the become an important issue across the natural his- National Science Foundation's Advancing Digitiza- tory community (Chapman and Grafton, 2008) that tion of Biodiversity Collections Program (Coopera- continues to lack clear consensus. With some tive Agreement EF-1115210). Any opinions, exceptions, Norris and Butts (2014) advocate in findings, and conclusions or recommendations favor of making paleontological locality descrip- expressed in this material are those of the tions publicly available. We do not address efficacy author(s) and do not necessarily reflect the views of data redaction or steps to achieve it in Module 1 of the National Science Foundation. of the workflows, but recognize that institutional policy and federal law often governs the quantity REFERENCES and types of data an institution might choose to expose. For institutions who choose to redact, Barber, A.C. 2012. Advancing Access to Biodiversity physical redaction might occur during data entry Data Using the SALIX Method and Digital Field (e.g., embargoing certain pieces of locality data Guides. Unpublished Master’s Thesis, Arizona State from public data exports), imaging (e.g., covering University, Tempe, Arizona, USA. locality information on a label), or electronically at Barber, A.C., Lafferty, D., and Landrum, L.R. 2013. The SALIX Method: a semi-automated workflow for her- the time data are exposed on the web (e.g., only barium specimen digitization. Taxon, 62:581-590. exporting data to county level or imposing a rule on Barton, M.A., Frakes, B., and Meyer, H.W. 2006. Using a the export query such that only precise data for relational geodatabase to manage paleontological non-sensitive localities are exported). resources at Florissant Fossil Beds National Monu- ment. New Mexico Museum of Natural History and CONCLUSION Science Bulletin, 34:7. Beaman, R. and Cellinese, N. 2012. Mass digitization of The advent of the U.S. National Science scientific collections: New opportunities to transform Foundation’s Advancing Digitization of Biodiversity the use of biological specimens and underwrite biodi- Collections (ADBC) initiative in 2011 has led to a versity science. ZooKeys, 209:7-17. rapid and significant increase in digitization activity Butts, S. 2013. YPM-IP workflow with voice recognition in natural history museums and university-based imaging. Presentation from the iDigBio paleontology biodiversity collections. This is especially true in digitization workshop, Yale Peabody Museum, New the paleobiological sciences. Three Thematic Col- Haven, CT, September 2013. Document accessed lections Networks (TCNs) encompassing 28 pale- 22 January 2015. ontological collections are currently funded through www.idigbio.org/content/ypm-ip-workflow-voice-rec- ADBC, all focusing largely on the digitization of ognition-imaging

11 KARIM ET AL.: DIGITIZATION WORKFLOWS

California Academy of Sciences. 2016. AntWeb Partici- Haug, J.T. and Haug, C. 2014. Eoprosopon klugi pation. AntWeb version 6.9.2. Document accessed (Brachyura) - the oldest unequivocal and most "prim- 22 July 2016. itive" crab reconsidered. Palaeodiversity, 7:149-158. www.antweb.org/documentation.do Häuser, C.L., Steiner, A., Holstein, J., and Scoble, M.J. Chapman, A.D. and Grafton, O. 2008. Guide to Best (eds.). 2005. Digital imaging of biological type speci- Practices for Generalising Primary Species-Occur- mens: a manual of best practice. Results from a rence Data. Global Biodiversity Information Facility, study of the European Network for Biodiversity Infor- Copenhagen. mation. European Network for Biodiversity Informa- Crick, M. 2005. Image File Management, p. 41-56. In tion, Stuttgart. Häuser, C.L., Steiner, A., Holstein, J., and Scoble, Hegna, T.A. 2010. Photography of soft-bodied crusta- M.J. (eds.), Digital imaging of biological type speci- ceans via drying, whitening, and splicing. Journal of mens: a manual of best practice. Results from a Crustacean Biology, 30:351-356. study of the European Network for Biodiversity Infor- iDigBio. 2012a. Digitization workflow workshop report. mation. European Network for Biodiversity Informa- Report on the iDigBio Developing Robust Object-to- tion, Stuttgart. Image-to-Data (DROID) workflow workshop, Florida Darwin Core Task Group. 2015. Darwin Core terms: a Museum of Natural History, Gainesville, FL, May quick reference guide. Document accessed 19 Feb- 2012. Document accessed 05 February 2015. ruary 2015. www.idigbio.org/content/digitization-workflow-work- rs.tdwg.org/dwc/terms/ shop-report Farrell, Ú.F. 2014. Label imaging workflow from the KU iDigBio. 2012b. Workflow modules and task lists. Proto- Invertebrate Paleontology collection – a quality con- col developed from the iDigBio Developing Robust trol measure, p. 52. In (editor unknown), Society for Object-to-Image-to-Data (DROID) workflow work- the Preservation of Natural History Collections 29th shop, Florida Museum of Natural History, Gainesville, Annual Meeting, Cardiff, Programme and Abstracts. FL, May 2012. Document accessed 05 February Feldmann, R.M. 1989. Whitening fossils for photographic 2015. purposes, p.342-346. In Feldmann, R.M., et www.idigbio.org/content/workflow-modules-and-task- al.[AUTHOR: editors’ names should be listed] lists (eds.), Paleotechniques. The Paleontological Soci- iDigBio 2012c. Policy on Acceptable Formats for iDigBio- ety, Special Publication 4. hosted Images and Recommendations for the Acqui- Glenister, B.F. 1977. Innovative Programs & Electronic sition, Processing, Storage, and Distribution of Digital Data Processing in Collections Management, p.25- Images. Document accessed 07 February 2015. 28. In Glenister, B.F. (ed.), Fossil Invertebrates - Col- www.idigbio.org/sites/default/files/internal-docs/idig- lections in North American Repositories 1976: A bio-standards/Image_File_Format_Recommenda- Report of the Paleontological Society Ad Hoc Com- tions_and_Standards_20120921.pdf mittee on North American Resources in Invertebrate iDigBio. 2013. Paleo Digitization Workshop. Wiki devel- Paleontology (CONARIP). Paleontological Society, oped for the iDigBio paleontology digitization work- Columbus, OH. shop, Yale Peabody Museum, New Haven, CT, Granzow-de la Cerda, I. and Beach, J.H. 2010. Semi- September 2013. Document accessed 05 February automated workflows for acquiring specimen data 2015. from label images in herbarium collections. Taxon, www.idigbio.org/wiki/index.php/Paleo_Digitiza- 59:1830-1842. tion_Workshop Guralnick, R.P., Cellinese, N., Deck, J., Pyle, R.L., iDigBio. 2014a. Digitizing from Source Materials Work- Kunze, J., Penev, L., Wals, R., Hagedorn, G., Agosti, shop. Wiki developed for the iDigBio digitizing from D., Wieczorek, J., Catapano, T., and Page, R.D. source materials workshop, Yale Peabody Museum, 2015. Community Next Steps for Making Globally New Haven, CT, March 2014. Document accessed Unique Identifiers Work for Biocollections Data. Zoo- 12 February 2015. Keys, 494:133. www.idigbio.org/wiki/index.php/Digitizing-From- doi:10.3897/zookeys.494.9352 Source-Materials Guralnick, R., Conlin, T., Deck, J., Stucky, B.J., and Celli- iDigBio. 2014b. Paleo Imaging Workshop. Wiki devel- nese, N. 2014. The Trouble with Triplets in Biodiver- oped for the iDigBio specimen imaging for paleontol- sity Informatics: A Data-Driven Case against Current ogy workshop, University of Texas, Austin, TX, April- Identifier Practices. PLoS ONE, 9(12): e114069. May, 2014. Document accessed 04 February 2015. doi:10.1371/journal.pone.0114069 www.idigbio.org/wiki/index.php/Paleo_Imag- Haston, E., Cubey, R., Pullan, M., Atkins, H., and Harris, ing_Workshop D.J. 2012. Developing integrated workflows for the iDigBio, 2016. Workflow Modules and Task Lists. iDigBio digitisation of herbarium specimens using a modular website. Document accessed 22 July 2016. and scalable approach. ZooKeys, 209:92-102. www.idigbio.org/content/workflow-modules-and-task- lists

12 PALAEO-ELECTRONICA.ORG

Jago, J.B. 1973. A hazard in the magnesium oxide Nelson, G., Paul, D., Riccardi, G., and Mast, A.R. 2012. method of whitening fossils. Journal of Paleontology, Five task clusters that enable efficient and effective 47:591-592. digitization of biological collections. ZooKeys, Karim, T.S. and Farrell, U.C. 2014. Specify Paleo Collec- 209:19-45. tions Workshop - training, new technology, and data Nelson, G., Sweeney, P., Wallace, L.E., Rabeler, R.K., models. Report on the iDigBio Specify for paleo col- Allard, D., Brown, H., Carter, J.R., Denslow, M.W., lections workshop, University of Kansas, Lawrence, Ellwood, E.R., and Germain-Aubrey, C.C. 2015. Digi- KS, May 2014. Document accessed 03 February tization workflows for flat sheets and packets of 2015. plants, algae, and fungi. Applications in Plant Sci- www.idigbio.org/content/specify-paleo-collections- ences, 3(9):1-9. workshop-training-new-technology-and-data-models New York Botanical Garden. 2004. New York Botanical Kier, P.M., Grant, R.E., and Yochelson, E.L. 1965. Whit- Garden Virtual Herbarium Best Practices Guide. ening fossils, p. 453-456. In Kummel, B. and Raup, Document accessed 13 January 2015. D.M. (eds.), Handbook of Paleontological Tech- sciweb.nybg.org/science2/VirtualHerbarium.asp niques. W.H. Freeman and Company, San Francisco. Norris, C. and Butts, S. 2014. Let your data run free? Lafferty, D. and Landrum, L. 2009. SALIX, a semiauto- The challenge of data redaction in paleontological matic label information extraction system using OCR. collections. Collection Forum, 28(1-2):113-118. Presentation given at the joint Annual Meeting of the doi:10.14351/0831-4985-28.1.113 Mycological Society of America, American Bryologi- Page, R.D.M. 2008. Biodiversity informatics: the chal- cal and Lichenological Society, American Fern Soci- lenge of linking data and the role of shared identifi- ety, American Society of Plant Taxonomists, ers. Briefings in Bioinformatics, 9(5):345-354. Botanical Society of America, 25-29 July 2009, doi:10.1093/bib/bbn022 Snowbird, Utah. Page, R.D.M. 2009. bioGUID: resolving, discovering, MacLeod, N. and Guralnick, R. 2000. Paleoinformatics, and minting identifiers for biodiversity informatics. p. 31-36. In Lane, R.H., Steininger, F.F., Kaesler, BMC Bioinformatics, 10(suppl. 14):S5. R.L., Ziegler, W., and Lipps, J. (eds.), Fossils and the Reed, D., Barr, W.A., Mcpherson, R.B., Geraads, D. Future: Paleontology in the 21st Century. W. Kramer, Wynn, J.G., and Alemseged, Z. 2015. Digital Data Frankfurt am Main. Collection in Paleoanthropology. Evolutionary Marsh, R.C. and Marsh, L.F. 1975. New techniques for Anthropology, 24:238-249. coating paleontological specimens prior to photogra- doi: 10.1002/evan.21466 phy. Journal of Paleontology, 49:565-566. onlinelibrary.wiley.com/doi/10.1002/evan.21466/epdf Mello, J.F. 1969. Paleontologic data storage and Richards, K., White, R., Nicolson, N., and Pyle, R. 2011. retrieval. Proceedings of the North American Paleon- A Beginner’s Guide to Persistent Identifiers, version tologic Convention, part B: 57-71. 1.0. Released on 9 February 2011. Copenhagen: Mello, J.F. and Collier, F.J. 1972. New procedures in Global Biodiversity Information Facility, 33 pp, acces- recording specimen-related data on fossils. Journal sible online at of Paleontology, 46:776-777. links.gbif.org/persistent_identifiers_guide_en_v1.pdf Morris, P.J. 2000. A data model for invertebrate paleon- Scoble, M. 2010. Natural history collections digitization: tological collections information, p. 105-108. In Rationale and value. Biodiversity Informatics, 7:77- Allmon, W.D. and White, R.D. (eds.), Guidelines for 80, 35KB. journals.ku.edu/index.php/jbi/article/view- the Management and Curation of Invertebrate Fossil Article/3994 Collections. The Paleontological Society, Special Short, A. 2014. Data before specimens: workflows for Publication 10. The Paleontological Society, Boulder. digital management and sharing of field notes and Morris, P.J. 2005. Relational database design and imple- images, p.150. In (editor unknown), Proceedings, mentation for biodiversity informatics. PhyloInformat- 62nd Annual Meeting of the Entomological Society of ics, 7:1-66. America. Entomological Society of America. Nelson, G. 2012. Workflow elements and concepts: com- Smith, V.S. and Blagoderov, V. 2012. Bringing collections mon practices. Presentation at the iDigBio digitizing out of the dark. Zookeys, 209:1-6. biological collections workshop. Document accessed doi:10.3897/zookeys.209.3699 05 February 2015. Teichert, C. 1948. A simple device for coating fossils with www.idigbio.org/content/workflow-elements-and-con- ammonium chloride. Journal of Paleontology, cepts-common-practices 22:102-104. Nelson, G. 2015. Gaps in Biodiversity Data: Vollmar, A., Macklin, J.A., and Ford, L. 2010. Natural his- Challenges for Digitization. Presentation at the tory specimen digitization: challenges and concerns. Global Change Workshop, Missouri Botanical Gar- Biodiversity Informatics, 7:93-112, 188KB. den. Document accessed 05 February 2016. journals.ku.edu/index.php/jbi/article/view/3992/3806 www.idigbio.org/wiki/images/3/3-D/Nelson-Digitiza- Weiss, A. 2013. Wrestling with mosasaurs: results of the tionGaps.pdf Sternberg Museum of Natural History - Forsyth Library fossil digitization pilot project, p. 103-124. In

13 KARIM ET AL.: DIGITIZATION WORKFLOWS

Cool, C. and Ng, K.B. (eds.), Recent Developments museum site and database with open access, p. 701. in the Design, Construction, and Evaluation of Digital In (editor unknown), 2014 GSA Annual Meeting in Libraries: Case Studies. Information Science Refer- Vancouver, British Columbia, Abstracts with Pro- ence, Hershey, PA. grams 46(6). Geological Society of America, Boulder. doi:10.4018/978-1-466-2991-2.ch007 Woodward, B. and Compton, B. 2014. Using GIS to map Wolniewicz, P. 2009. Easily-accessible digital palaeonto- and catalog paleontological and geological speci- logical databases - a new perspective for the storage mens of the Mid-Pliocene Gray Fossil Site, Washing- of palaeontological information. Geologos, 15(3- ton County, Tennessee, USA, p. 823. In (editor 4):181-188. doi:10.2478/v10118-009-0002-1 unknown), 2014 GSA Annual Meeting in Vancouver, Woods, J.A. and Chadwick, A.V. 2014. Development and British Columbia, Abstracts with Programs, 46(6). implementation of an efficient, accessible online Geological Society of America, Boulder.