MetPetDB: A database for metamorphic geochemistry

MetPetDB: A database for metamorphic geochemistry

Frank S. Spear, Benjamin Hallett, and Joseph M. PyleDepartment of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, New York 12180, USA([email protected])

Sibel Adalı, Boleslaw K. Szymanski, Anthony Waters, Zak Linder, Shawn O. Pearce,Matthew Fyffe, Dennis Goldfarb, Nickolas Glickenhouse, and Heather Buletti

Department of Computer Science, Rensselaer Polytechnic Institute, Troy, New York 12180, USA

[1] We present a data model for the initial implementation of MetPetDB, a geochemical database specificto metamorphic rock samples. The database is designed around the concept of preservation of spatialrelationships, at all scales, of chemical analyses and their textural setting. Objects in the database (samples)represent physical rock samples; each sample may contain one or more subsamples with associatedgeochemical and image data. Samples, subsamples, geochemical data, and images are described withattributes (some required, some optional); these attributes also serve as search delimiters. All data in thedatabase are classified as published (i.e., archived or published data), public or private. Public andpublished data may be freely searched and downloaded. All private data is owned; permission to view, edit,download and otherwise manipulate private data may be granted only by the data owner; all such editingoperations are recorded by the database to create a data version log. The sharing of data permissions amonga group of collaborators researching a common sample is done by the sample owner through the projectmanager. User interaction with MetPetDB is hosted by a web-based platform based upon the Java servletapplication programming interface, with the PostgreSQL relational database. The database web portalincludes modules that allow the user to interact with the database: registered users may save and downloadpublic and published data, upload private data, create projects, and assign permission levels to projectcollaborators. An Image Viewer module provides for spatial integration of image and geochemical data. Atoolkit consisting of plotting and geochemical calculation software for data analysis and a mobileapplication for viewing the public and published data is being developed. Future issues to address includepopulation of the database, integration with other geochemical databases, development of the analysistoolkit, creation of data models for derivative data, and building a community-wide user base. It is believedthat this and other geochemical databases will enable more productive collaborations, generate moreefficient research efforts, and foster new developments in basic research in the field of solid earthgeochemistry.

Components: 7166 words, 6 figures, 7 tables.

Keywords: metamorphic petrology; database; geochemistry; geoinformatics.

Index Terms: 3660 Mineralogy and Petrology: Metamorphic petrology; 3665 Mineralogy and Petrology: Mineral

occurrences and deposits; 3690 Mineralogy and Petrology: Field relationships (1090, 8486).

Received 31 July 2009; Revised 5 October 2009; Accepted 12 October 2009; Published 4 December 2009.

Spear, F. S., et al. (2009), MetPetDB: A database for metamorphic geochemistry, Geochem. Geophys. Geosyst., 10, Q12005,

doi:10.1029/2009GC002766.

G3G3GeochemistryGeophysics

Geosystems

Published by AGU and the Geochemical Society

AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES

GeochemistryGeophysics

Geosystems

Article

Volume 10, Number 12

4 December 2009

Q12005, doi:10.1029/2009GC002766

ISSN: 1525-2027

ClickHere

for

FullArticle

Copyright 2009 by the American Geophysical Union 1 of 15

http://dx.doi.org/10.1029/2009GC002766

1. Introduction

[2] Current research in the geological sciences gen-erates a huge amount of visual and numeric data, andthe recognition of the need to catalog and share thisdata (e.g., see white papers at http://www.commu-nitytechnology.org/nsf_ci_report, http://www.adec.edu/nsf/nsfcyberinfrastructure.html, and http://www.nsf.gov/news/special_reports/cyber/index.jsp)as a value-added research tool has spurred thedevelopment of a number of organizations (EARTH-CHEM http://www.earthchem.org/earthchemWeb/index.jsp; GEON (Geosciences Earth network) http://www.geongrid.org/; SESAR http://www.geosamples.org/) and databases (GEOROC http://georoc.mpch-mainz.gwdg.de/georoc/Start.asp; NAVDAT http://navdat.kgs.ku.edu/; and PETDB http://www.petdb.org/petdbWeb/index.jsp) specific to handling, sharing,and manipulating data generated in the research andanalysis of earth materials.

[3] A compelling justification for this developmentis the belief that fundamentally new science willemerge from the scientific disciplines that possesshighly evolved cyberinfrastructure. A leading exam-ple in the geosciences is the use of ocean basaltchemistry data mined from PETDB [Lehnert et al.,2000]; indeed,Lehnert and Langmuir [2007] state thatover 150 articles have cited the geochemical databasePetDB as the primary data source in a publicationsince 2002 [e.g., Spiegelman and Kelemen, 2003;Salters and Stracke, 2004; Hirschmann et al., 2003].These papers make use of large data sets that wouldhave been virtually impossible for an individual, orsmall group of collaborators, to collate until the adventof the aforementioned databases.

[4] Although some metamorphic rocks are includedin existing petrologic databases, and at least onedatabase has been developed specifically to addresssome of the needs of metamorphic petrology[Schmatz et al., 1995], there is no current globaldatabase that incorporates the special requirementsof metamorphic geochemistry. Interpretations ofmetamorphic parageneses require not only highprecision chemical analyses of minerals present,but evaluation of the textural context of thoseanalyses. Because of this, a large component ofthe existing metamorphic data is images (photo-micrographs, backscattered electron (BSE), second-ary electron (SE) and cathodoluminescence (CL)images, X-ray maps, etc.) and the location ofanalyses with respect to these images is a criticalcomponent of a metamorphic database. Unfortu-nately, many of the images and analyses collected

during the course of a study never get publishedowing to limitations of print media. Indeed, it isestimated by the authors that less than one percentof all data collected on metamorphic rocks ispublished, which suggests that a potential treasuretrove of information is waiting to be mined, giventhe proper infrastructure.

[5] The preservation of spatial relationships at avariety of scales thus differentiates a geochemicaldatabase for metamorphic rocks from other geo-chemical databases, and provides a framework forits design. In our realization, the basic componentsof a database specific to metamorphic petrologyshould (1) include the incorporation of bulk rockand mineral analyses; (2) provide for the incorpo-ration of images of any type; (3) preserve thespatial relationship among the various images col-lected on a thin section; (4) preserve the relation-ship between analyses and textural setting of theanalyses on the relevant images; (5) provide anintuitive user interface that allows for searching,uploading, and downloading sample information aswell as facilitating collaborations among research-ers; (6) provide a set of tools for recalculation,plotting, and analysis of data; and (7) interfacewith other geological databases. We envision adatabase that is populated not only with published,archive data, but incorporates unpublished datathat complements what is published, and that canalso serve as a research level collaborative tool forresearchers globally.

[6] Ideally, a metamorphic database should incorpo-rate both raw data (analyses, images, etc., collectedon a sample) as well as derivative or interpretativedata for an area (P-T conditions, P-T-t path, coolinghistory, etc.). The present communication describesour progress on the development of the first part ofthe database for raw data. How to treat derivative/interpretative data is far more complex and will bediscussed in a future communication.

2. Data Model

[7] The MetPetDB data model is built around theconcept of a sample. The sample is a physicalentity (a piece of rock) that exists in the database asa virtual entity with associated information, similarto the approach taken by other geochemical data-bases [e.g., Lehnert et al., 2000]. A number ofconsiderations have been taken into account in thedesign of the data model. Typically, a thin sectionis cut from a sample and analytical work (e.g.,electron microprobe analysis) is done on a polished

GeochemistryGeophysicsGeosystems G3G3

spear et al.: metpetdb—a database for metamorphic rocks 10.1029/2009GC002766

2 of 15

thin section that may be different from the originalthin section. Whole rock chemical analyses aredone on yet different parts of the rock, and otherparts of the sample might be used for mineralseparates (e.g., zircons). It is considered imperativethat the chain of evidence of these distinct pieces ofa sample be kept intact, so we have also introducedthe concept of a subsample, which is, as the nameimplies, a piece of the original sample. Chemicalanalyses are always done on subsamples (unlessthe entire sample has been crushed for analysis).Each thin section is a subsample and a piece of therock used for mineral separates is yet anothersubsample. Images may be associated with thesample (e.g., a sample map or outcrop photograph)or an individual subsample (e.g., thin section).Each of these data types has its own set ofattributes, which will be discussed in this section.

3. Sample

[8] Sample attributes are shown in Table 1 withrequired attributes labeled ‘‘R.’’ The minimumrequired information for adding a sample to thedatabase are (1) sample number; (2) location (lat-itude and longitude); (3) rock type (see Table 2);and (4) sample owner (assumed to be the userentering the data), although, of course, additionalinformation is highly desirable. When available,the IGSN (International Geo Sample Number) canprovide the unique sample identifier and in thefuture we expect that all samples will be registered

and have IGSNs. Every sample must have anowner so that data use permissions can be properlyevaluated. At present, samples are uniquely iden-tified in MetPetDB by the sample number + owner(i.e., an individual sample owner must have allunique sample IDs). It is recognized that therequirement of a sample owner will create somespecial issues in the long term (e.g., the death of adata owner), which will be dealt with as individualcases. Latitude and longitude provide location,with the option to provide a location error if asample is not well located. Although moderntopographic maps and GPS provides location with-in a few meters, many older data are not soprecisely located, and it is important that thelocation of a sample is not misrepresented. The

Table 1. Sample Attributes

Sample AttributeRequired (R) orOptional (O)

Number ofPossibleAttributes Comments

Sample number R 1 The unique sample identifier is the sample number + sample ownerSample owner R 1IGSN O 0–1 International Geological Sample NumberSample alias O 0–n Provides for multiple names for the same sampleLongitude (DD.DDDDD) R 1 Five decimal digits provides a location accuracy of around 1 m.Latitude (DD.DDDDD) R 1Location error (m) O 0–1 Estimated error in location of sampleRock type R 1 Rock names given in Table 2Region O 0–n Geographic descriptor. Multiple regions are permissible.Country O 0–1Collector O 0–1Collection date O 0–1Present sample location O 0–1 Where the sample can be foundComment O 0–n Multiple comments possibleReference O 0–n Published data requires a referenceMetamorphic facies O 0–n Multiple facies allowedMinerals present O 0–n Text field. Can be an x (present), a mode, or other characters.Public/Private R 1 A flag defining whether the sample is viewable by public

Table 2. MetPetDB Recognized Rock Names

Rocks

Metapelite types slate, phyllite, schist, gneiss, migmatiteCarbonate types marble, calc-silicateMetabasite types greenschist, amphibolite, blueschist,

eclogiteTextural types granofels, hornfels, skarnMineral types quartzite, jadeitite, glaucophanite,

serpentinite, garnetite, pyroxenite,cordierite-anthophyllite

Deformation types mylonite, cataclasiteProtolith types metapelite, metaarkose, metagreywacke,

metabasite, metacarbonate, metagranite,metaigneous, metavolcanic


spear et al.: metpetdb—a database for metamorphic rocks 10.1029/2009GC002766spear et al.: metpetdb—a database for metamorphic rocks 10.1029/2009GC002766

3 of 15

rock types shown in Table 2 are restrictive relativeto the total range of metamorphic rock names thathave been used in the literature. However, it wasdecided that a compact, simple list is preferable toan exhaustive one because it helps avoid the use ofambiguous names. It is fully recognized that manyresearchers will wish to use their own favorite (insome cases locally derived) rock names, and this isencouraged. MetPetDB is designed to accommo-date as many of these as desired in comment fields,which are fully searchable. Next, if the data arefrom a published article (i.e., archived data) aunique reference to the publication is required.We have followed the approach of NAVDAT inadopting the 10-digit GeoRef Accession Number asthe unique publication identifier. Finally, publisheddata are given the owner name ‘‘PUBLICATION’’and are not editable (to preserve the publishedrecord) whereas unpublished data are always edit-able by the owner.

[9] Optional data include the country of origin(recognizing that the country may no longer exist),and the ‘‘region’’ from which the sample wascollected. Multiple regions are allowed because itis recognized that petrologists will wish to locatesamples using multiple geographic names. Forexample, a sample from the first author’s collectioncomes from the ‘‘Eastern Alps,’’ the ‘‘TauernWindow,’’ the ‘‘Hohe Tauern’’ the ‘‘Großvenedigerregion,’’ the ‘‘Froznitztal,’’ and the ‘‘Tauern Eclo-gite Zone.’’ Multiple regions might make searchingmore difficult initially, but it is planned to developabsolute geographic outlines for specific regionsbased on the database population that will facilitatesearching.

[10] A list of minerals and modes is optional databecause not all published papers provide a list ofminerals. Furthermore, it is specifically not requiredto list a mineral assemblage, because the term ‘‘as-semblage’’ carries the implication of a stable equilib-rium assemblage at some metamorphic condition,and that is most definitely an interpretation. Addi-tionally, the metamorphic facies (or grade) attribute(Table 3) can be multivalued, due to the recognitionthat, for the same suite of rocks, one author may usethe term ‘‘amphibolite facies’’ whereas another au-thor may select ‘‘staurolite zone’’ as the descriptor ofmetamorphic grade. This will not complicatesearches because it will be simple to choose bothsearch criteria. Items such as collector and collectiondate are self-explanatory. The attribute ‘‘present sam-ple location’’ is included so that individuals whowishto use a sample can find it.

[11] Note that it is not necessary for chemicalanalyses to be available for a sample to be enteredinto the database. In this way, MetPetDB differsfrom other geochemical databases (e.g., PetDB,and GeoRoc) in that it is, fundamentally, a databaseof metamorphic samples, similar to NAVDAT forvolcanic rocks. This does not limit in any way theusefulness of MetPetDB and it provides the oppor-tunity of recording the types of rocks found in ametamorphic terrane even when only a singlesample has been taken from the field.

4. Subsample

[12] The data model recognizes that the mainportion of analytical effort is expended on aliquotsderived from the sample itself. Such aliquots maytake the form of thin sections, polished thin sec-tions, rock chips, or mineral separates. Such sam-ple aliquots are conceptually accommodated by thedata model in the form of a subsample. Requiredsubsample attributes include sample name, sub-sample name, subsample type (thin section, pol-ished thin section, rock chip, or mineral separate),and subsample owner (Table 4). Subsample owneris listed because it is recognized that a differentperson may ‘‘own’’ a thin section or mineralseparate from the original sample owner.

5. Images

[13] It is recognized in the design of MetPetDBthat most textural information in a metamorphicrock is visual. In addition, the location of analyseswithin minerals (e.g., core versus rim) is critical to

Table 3. MetPetDB Recognized Metamorphic GradeNames

Metamorphic Grades

Metamorphic Facies zeolite, prehnite-pumpellyite,greenschist, amphibolite,epidote amphibolite, granulite,blueschist, eclogite, hornfels

Mineral Zonesin Metapelites

chlorite zone, biotite zone,garnet zone, staurolite zone,staurolite-kyanite zone,kyanite zone, sillimanite zone,andalusite zone,sillimanite-K feldspar zone,garnet-cordierite zone,migmatite zone

Extreme metamorphism ultra high pressure,ultra high temperature



4 of 15

their interpretation and a chemical analysis of azoned mineral without this knowledge is signifi-cantly less valuable than one that is well located.

[14] Image attributes are shown in Table 5. Imagesmay be associated with either a sample (e.g., aphotograph of the outcrop, the sample, or a mapshowing sample locations from a published paper)or a subsample (e.g., a micrograph of a thinsection). A number of different image types aredefined, as listed in Table 6, including, but notlimited to, photographs, photomicrographs, SEMimages (BSE, SE, and CL), and X-ray maps.Additional image types will be added as needed.A scale attribute (image width in mm) is providedif this information is available. Image files alwayscontain headers with information relating to thenumber of pixels in the X and Y dimensions, andthis information can be extracted from the file. Theimage resolution (size of an individual pixel) canthus be calculated if the full width dimension isprovided.

[15] A key aspect of preserving the usefulness ofmetamorphic data is preservation of the spatialrelationship among minerals. For example, is aparticular muscovite crystal located in the shearplane or the Q-F domain of a crenulation cleavage?Is a particular monazite located in the core or near

the rim of the garnet host? In other words, thespatial relationship of minerals with respect to eachother (i.e., the texture) is critical for the interpreta-tion of the chemistry of these phases. To preservethese relationships, MetPetDB incorporates a sub-sample grid system whereby all images and anal-yses can be located with respect to each other forthe individual subsample. It is perhaps simplest tothink of this as a GIS (Geographic InformationSystem) for thin sections where the data includeimages and analyses. The system includes animage viewer that permits the user to registerimages with respect to one another and to locateand view analyses with respect to these images. Anexample of the image viewer is given below in thesection on user interface.

[16] Some images include more than a singlesample as, for example, a photograph of an outcropor a scanned image of a map on which samplelocations are marked. MetPetDB accommodatesthis by permitting the same image to be linked toany number of samples.

6. Chemical Analyses

[17] Chemical analyses form the core of any geo-chemical database although, as noted above, it is

Table 4. Subsample Attributes

AttributeRequired (R) orOptional (O)


Sample number R 1 IGSN, if availableSubsample name R 1 e.g. SS01, SS02 etc.Subsample type R 1 Types include thin section, polished thin section, rock chip, mineral separateSubsample owner R 1 Subsamples may be owned by individuals different than the sample ownerPublic/private R 1 A flag defining whether the subsample is viewable by public

Table 5. Image Attributes

Image AttributeRequired (R) orOptional (O)


Sample number R 1 Images require a sample. The same image (e.g., a map) may beuploaded for numerous samples.

Subsample name O 0–1 e.g. SS02 or blank if it is a whole sample imageImage file name R 1Image type R 1 See Table 6Image scale O 0–1 Image width in mmOwner R 1Collector O 0–1Comments O 0–nDescription 0 0–1



5 of 15

not a requirement that chemical analyses exist for asample to be entered into MetPetDB. Requiredattributes for chemical analyses include (Table 7)(in addition to information about the sample andsubsample): (1) mineral name (either identity ofmineral or ‘‘bulk rock’’ for a whole-rock analysis);(2) analytical method; (3) element or species name;(4) units of concentration (e.g., weight percentoxide, ppm, etc.); (5) concentration value; (6)analysis owner. Optional attributes provide infor-mation about the analytical facility, analysis date

and analyst, publication reference, and supportinganalysis information (e.g., spot number, elementprecision, analysis total).

[18] Two different types of data attributes areincluded to provide spatial information about ananalysis. The ‘‘reference image name’’ refers to animage on which the analysis is located. In somecases this may be a scanned image (e.g., photo-graph or BSE image) with analysis spots located byhand. If the location of the analysis spot on this

Table 6. Image Types and Attributes

Image Type Comments

Field notes Scan of notes taken in the fieldDrawing Hand sketch or published figureMap Geologic or other map with sample locationsCross sectionFigure Miscellaneous figure, may be compositeField photo Field photograph of sample locationPhotograph General sample imageryThin section scan Special image of entire thin sectionPhotomicrograph-transmitted plane polarizedPhotomicrograph-transmitted crossed polarsPhotomicrograph-reflected plane polarizedPhotomicrograph-reflected crossed polarsSecondary electron image SEBackscattered electron image BSECathodoluminescence image CLX-ray element distribution map element name (R) Element name is required for X-ray maps

Table 7. Analysis Attributes

AttributeRequired (R) orOptional (O)


Sample number R 1Subsample number R 1 Analyses require a subsampleMineral name R 1 ‘‘Bulk’’ for a whole rock analysisAnalytical method O 0–1 e.g. EMP, LA-ICPMS, SIMSLocation of analytical facility O 0–1Reference O 0–n Published data require a referenceAnalysis date O 0–1Analyst O 0–1Spot number R 1 Unique spot number identifierReference image name O 0–1 Image name or file name on which analysis spot is locatedImage X coordinate O 0–1 Location of spot on reference imageImage Y coordinate O 0–1 Location of spot on reference imageX stage coordinate of analysis O 0–1 Stage location of spot analysisY stage coordinate of analysis O 0–1 Stage location of spot analysisElement or species name R 1–n e.g. Si or SiO2

Value RUnits R Wt % or ppmPrecision O Precision unit is required if precision is providedPrecision type R

Analysis total O 0–1Comment O 0–n Multiple comments are permissible



6 of 15

reference image is known, then the coordinates canbe specified. Finally, it is common on manyelectron microprobes that the X-Y stage coordi-nates are saved with the analysis data. The refer-ence frame for these coordinates change every timea sample is reloaded into the microprobe, but for asingle session they can provide useful spatialinformation. For example, the X-Y stage coordi-nates for points taken on a line traverse provideinformation about the distance between analyses.

[19] It is noted here the analytical method attributecould be expanded significantly to encompass aseries of method-specific subattributes and meta-data; this topic has been discussed at length (e.g.,‘‘Data Reporting in Geochemistry,’’ an EarthChemworkshop held at Lamont-Doherty Earth Observa-tory in April, 2007). For example, the analyticalmethod attribute = Electron Microprobe (EMP)generates a whole suite of subsidiary attributesrelating to machine settings (e.g., diffraction crys-tal, calibration standard, detector gas), analyticalconditions (accelerating voltage, analysis current,analysis time), and postanalysis processing (back-ground fitting, ZAF correction routine). Instru-ment-specific analytical attributes are, at present,beyond the scope of this data model, but may beincorporated in future versions.

7. Data Ownership and Sharing

[20] MetPetDB has been designed to fill threemajor functions: It is intended to be a repositoryof published data on metamorphic rocks (sampleinformation, chemical analyses, etc.), it is intendedto store supplemental, supportive data that comple-ments published data but is too voluminous to bepublished by traditional means, and it is intendedto be a platform to facilitate the collaboration ofresearch projects among geoscientists around theglobe. This section discusses aspects of data per-missions and sharing that are implemented inMetPetDB.

7.1. Data Types and Data Ownership

[21] All data housed in MetPetDB is tagged withan owner. Data that are published is tagged with‘‘PUBLICATION’’ as the owner, and are viewableby any user. Published data are immutable; itscontent in the database is a mirror image of theoriginal data in the publication, and, except underspecial circumstances by a system administrator,cannot be added to or edited in any way, with theexception of addition of comments.

[22] All other data have a designated owner, whichis typically the individual who collects and uploadsthe data. An owner may label data as ‘‘private,’’ inwhich case it is viewable only by the owner andmembers of specified projects through the projectmanager (see below), or ‘‘public,’’ in which case itwill be viewable by anyone, but only modifiable bythe owner. Users can choose to make specificsamples and subsamples public selectively. How-ever, to make a subsample public, its parent samplemust be made public.

[23] Ownership of data will be prescribed at thelevels of sample and subsample (see Tables 1 and4). This level of ownership demarcation was con-sidered necessary because it is easily envisionedthat individual A might be the owner (collector) ofa sample, whereas individual B is the owner of athin section from that sample. Individual B mighthave done a considerable amount of work on thisthin section, including publishing a subset of thedata (images and analyses) collected.

7.2. Scientific Collaboration:‘‘Project Manager’’

[24] A major goal of MetPetDB is to facilitateresearch collaborations between individuals andresearch groups anywhere on the planet; suchcollaborations are likely to involve new and un-published data that collaborators wish to share witheach other, but not yet the general public.

[25] This sharing of data from one or more samplesgrouped under a collective research effort betweenindividuals or groups is handled through the proj-ect manager. Collections of samples that pertain toa specific research effort are called ‘‘projects.’’Projects have a project chief plus participants, withthe chief having responsibility for overseeing theorganization of the effort. For example, the projectchief has responsibility for adding (or removing)participants from the project. Projects can containboth public and private data with read/write per-missions being granted to project participants bythe data owners and handled through the projectmanager. Data and individuals can belong to mul-tiple projects, but each project can have only oneproject chief.

8. Database Implementationand Hardware

[26] MetPetDB’s implementation is based upon theJava servlet application programming interface



7 of 15

(API); this choice gives us immediate access to theextensive library of open source Java software. Inparticular, we are interested in leveragingmany of theWeb applications and geographic information sys-tems frameworks that are typically available for theJava platform. PostgreSQL was chosen as the rela-tional database due to its strong adherence to the SQLstandards, and its relative ease of administration. ThePostGIS extensions to PostgreSQL provided a com-pelling platform for our GIS needs such as efficientlysearching for points in bounding boxes.

[27] Object metadata for samples, subsamples,images and chemical analyses are stored directlyin a PostgreSQL database using a data model thatsatisfies the well-known good design rules (a.k.a.normalization) for database programming. Mostmetadata constraints such as required values andvalid value ranges are expressed and enforced bythe database, with the MetPetDB application auto-matically mining the constraints out of the databaseand implementing the same logic within the Javaapplication. This arrangement allows the databaseto provide complete enforcement of constraints,but also allows the application interface to offerimmediate targeted feedback to the user whenincorrect data are supplied.

[28] We expect a fairly large collection of imagesin the system. Accordingly, image files are notstored within PostgeSQL due to the difficultiesassociated with backing up and restoring multi-gigabyte table spaces (data files used by the data-base). Instead, the image files are stored in theUNIX file system in separate directories. Thelocation of the file is determined by taking the filecontent and hashing it to filename using the SHA-1algorithm. As SHA-1 provides a fairly uniformdistribution over filenames, disk arrays can bedivided into 256 partitions, using the first byte ofthe SHA-1 hash to automatically determine thepartition that will hold the image file.

[29] Extensive use of AJAX (Asynchronous Java-Script and XML) through the Google Web Toolkit(GWT) for the front-end interface allows the enduser’s web browser to efficiently request only theraw data required to update each page instead ofreloading the current page after a request. Aftersimple gzip compression has been applied, the sizeof the transmitted data packet averages well under1 kilobyte, facilitating easy server retrieval of therelevant object metadata and transfer to the client atwire speed. As the bulk of most per page process-ing costs is generally the HTML production (andnot the database query processing), pushing this

load onto the client browsers reduces the serverside costs, making the entire MetPetDB applicationmore responsive to all users.

[30] To ensure high-availability, fault tolerance, andadequate storage for images, we have purchased adatabase server and a web server (2 Quad CPUswith1.8TB of storage) for the production site of theMetPetDB. We are currently in process of config-uring these servers. The database server supportsRAID-6 for fault tolerance and fast access to datawhere the failure of one or even two disks does notcause any data loss. All machines are currentlyhoused in the Computer Science Department Labo-ratory at Rensselaer. An existing 300Gb server iscurrently being used as the development platform.

9. User Experience

[31] MetPetDB is accessed through a web portal.The database home page (Figure 1) has options for(1) login or registering; (2) viewing samples; (3)managing projects; (4) searching the database; and(5) uploading new data. A user wishing to explorea suite of samples sees the requested informationdisplayed in a number of different pages. The listof selected samples (these may be part of a project,a suite found from searching, or a user’s entiresample collection) are displayed on the sample listpage, a page that shows, in addition to the samplenumber, a list of other sample attributes (Figure 2).For each page a default set of attributes is dis-played, and these are fully configurable by eachindividual user. Selecting a sample brings up thesample page (Figure 3), which displays generalcharacteristics of a sample, as well as sampleimages and a list of the subsamples. Selecting asubsample brings up the subsample page withconfigurable attributes about available images andanalyses displayed (Figure 4).

[32] Chemical analyses may be of a mineral or thebulk rock. A list of the minerals for which analysesare available is provided and the user can choose toview the entire set of analyses for a particularmineral, or to download the analyses to a spread-sheet format. Note that one attribute for chemicalanalyses is the reference image, which is the imageon which the spot chemical analysis is located, andanalysis locations can be viewed on these images aswell. Images for a subsample are shown as thumb-nails on the subsample page (Figure 4) and individ-ual images can be examined in pop-up windows.

[33] Alternatively, images can be displayed on theimage viewer (Figure 5). The image viewer is a



8 of 15

module designed to display any or all of the imagesin a subsample in their correct spatial positionswith respect to each other. The image viewerincorporates the concept of a subsample map (orimage grid) onto which images can be placed. Eachimage can be scaled and rotated in order to locate

them in their proper positions. Transparency can beadjusted to aid in registration and the entire imagemap can be viewed at a wide range of scales. Thepurpose of developing the image viewer is topreserve the textural setting of all data, which isan essential part of the interpretative process.

Figure 2. Screenshot of MetPetDB web page showing sample listing. Attributes shown in this and other pages areconfigurable by the user. Major options include sorting the listing, adding or editing sample information, and addingsamples to projects.

Figure 1. Screenshot of MetPetDB home page. Major user operations include managing sample collections,managing projects, searching the database, and uploading new data.



9 of 15

Figure 3. Screenshot of MetPetDB sample detail page. Sample location, sample images, rock type, mineralspresent, and numbers of subsamples are shown.



10 of 15

Figure 4. Screenshot of MetPetDB subsample detail page providing a summary of the information available on thespecific subsample including image thumbnails and analyses.



11 of 15

[34] Figure 5 shows an example of an image mapfrom a subsample (thin section) from sample V6B,a garnet-biotite-sillimanite migmatite from the Val-halla Complex, British Columbia [Spear and Par-rish, 1996]. At wide zoom, one BSE image andthree photomicrographs are located and scaled withrespect to the thin section image (Figure 5a).Figure 5b shows an enlargement of a part of thethin section image with a Ca X-Ray map overlay.Analysis spots are shown in their proper locationon the sample in Figure 5c. In fully annotatedsamples, the analysis spots are linked directly tothe analyses in the database.

9.1. Searching the Database

[35] MetPetDB is searchable by nearly all of thedata attributes including words found in commentfields. Searchable items include (but are not limitedto) rock type, minerals present, metamorphic grade,latitude and longitude, region name, collector,publication, availability of images (e.g., Ca X-raymap), availability of analyses (e.g., samples withscapolite analyses), and composition range of anal-yses (e.g., 2.5 < MgO < 7.5). Figure 6 shows anexample of the MetPetDB search page.

Figure 5. Mock-up of Image Viewer. (a) Full view of a polished thin section. Background image is a thin sectionscan. Also shown are several photomicrographs and one BSE image. Boxes show locations of Figures 5b and 5c.(b) Enlargement of part of the Image Viewer showing overlay of a Ca X-ray map. The transparency of each image canbe adjusted. (c) Enlargement of a part of the BSE image showing location of analysis spots. Analysis spots for whicha reference image and reference image coordinates are specified can be plotted automatically on the grid.



12 of 15

[36] Mineral-delimited searches are hierarchicalusing a mineral classification tree. This approachensures that selecting ‘‘chain silicate’’ will searchon all pyroxenes and amphiboles but that selecting‘‘glaucophane’’ will find only those samples inwhich glaucophane has been listed. Users are ableto refine searches so that only the desired informa-tion is returned. Once a search is completed, resultsare viewable by any of the viewing pages describedabove, or plotted on a map view such as GoogleMaps or Google Earth.

9.2. Managing Projects

[37] The Web portal allows sample owners tocreate and manage projects. Managing a projectincludes selecting samples to be included in theproject and collaborators to be a part of the project.

10. Evolution of and Future Directionsfor MPDB

[38] The future of MetPetDB (or any other geolog-ical database) depends on its usefulness to the

geologic community. We believe that a well-designed and comprehensive database for meta-morphic geochemistry will be compelling to thegeologic community, but a number of significantissues need to be addressed before this becomes areality. In particular, the database must be populat-ed with sufficient high quality data and searchesmust return useful information. Second, there mustbe significant value added to compel individuals toput forth the effort that will be required to helppopulate the database. Third, there needs to beseamless integration with other databases so thatindividuals can find the information they seekwithout concern for where those data reside. Fi-nally, there is a need to demonstrate that there existscientific problems that can only be solved by largecollections of data and advanced search algorithms.

10.1. Populating the Database

[39] MetPetDB will only be as useful as the quan-tity and quality of data it contains. As mentionedearlier, it is estimated that less than one percent ofmetamorphic data are published, and the question

Figure 6. Screenshot of MetPetDB search page. This image shows the results of a search for rock types = gneiss,metapelite, and schist that contain mineral = garnet.



13 of 15

must be addressed how the other >99% of data willbe incorporated into the database.

[40] In general, data are added to MetPetDB in twoways either as individual samples or by bulkuploading. Bulk uploading of sample data andassociated images has been implemented throughthe use of spreadsheets and is the preferred methodof uploading large quantities of data. Althoughthere is some initial organizational effort to preparedata for uploading, much data is already stored inspreadsheets so this is not as daunting a task asmight first be thought. It is our hope that over timethe data currently stored on researchers hard drivesand file cabinets will find their way into thedatabase as researchers and students find it desir-able to do so. For example, an excellent place for astudent to start a project in a new area is to becomeaware of all previous work that has been done inthe region. If information on a region resides in thestudent’s advisors file cabinets, this becomes anexcellent opportunity to further populate the data-base. Once the backlog of existing data isuploaded, populating the database with newlycollected data should become routine. Eventually,it is anticipated that direct uploading from analyt-ical instruments (e.g., electron microprobes) maybe possible so that users will need to expend only aminor amount of effort.

[41] Published data will be digitized and incorpo-rated into the database by the MetpetDB workinggroup over time. This effort has already begun and,as of this writing, we have mined approximately120 articles, dating between 1962 and 2005. Jour-nals currently represented include American Jour-nal of Science, The American Mineralogist,Contributions to Mineralogy and Petrology, Jour-nal of Petrology, and Journal of MetamorphicGeology. Although the initial effort to add thewealth of published information will be large, itis hoped that at some point publication in anestablished journal will include addition of therelevant data to the database, thus ensuring thedatabase is up to date with the most currentpublished works.

10.2. MetPetDB User Toolkit

[42] A compelling reason to upload one’s own datainto MetPetDB, in addition to facilitating collabo-rations with colleagues, will be the ability to makeuse of the extensive toolkit that is in the process ofbeing developed. Code is being written to docommon tasks such as composition plotting andmineral formula recalculations, including ones re-

quiring stoichiometric constraints to estimate ferriciron content (e.g., pyroxenes and amphiboles). Thetoolkit will contain a complete set of publishedthermobarometry routines [e.g., Spear et al., 1991]that can be used on any database analyses and it isplanned to provide links to facilitate approachessuch as the ‘‘average P-T calculation’’ approach[e.g., Powell and Holland, 1994; Berman, 1991].Additional analysis tools will include software forage calculation (from U, Th, Pb) measurements,pseudosection construction, and contour plotting ofP-T diagrams.

[43] Future plans for MetPetDB include the incorpo-ration of derivative or interpretative data. These termsrefer to the fact that some very useful informationabout metamorphic rocks is only obtained aftersignificant recalculation, analysis, and interpretationof the basic composition and image data. Informationsuch as P-T conditions, maximum temperature orpressure, the P-T path, the age of chemical domainsof monazite, or the timing of crenulation develop-ment all fit the category of requiring significantinterpretation of raw data. Such information is in avery real sense the desired result of many metamor-phic studies, and would be of great use to researchersin a wide range of fields outside of metamorphicpetrology. We do not yet have a satisfactory datamodel for these types of information, but it is plannedto incorporate such derivative data into MetPetDB inthe future.

10.3. Integration With Other Databases

[44] Researchers searching for information are notconcerned with where the retrieved information isstored, but they are concerned that they find all ofthe information relevant to their study. MetPetDBwill be integrated with other databases of the Earth-Chem consortium (http://www.EarthChem.org),which includes PetDB, NAVDAT, and GeoRoc.

10.4. New Directions

[45] We envision that MetPetDB will be useful notonly for researchers, but also for museum curators,university, and secondary school teachers. A sig-nificant potential for exploring the geologic worldwill exist through databases such as MetPetDB andit is our intention to include materials on the Webportal that will assist guest users who may not beversed in geochemistry to find useful and interest-ing information.

[46] The potential also exists for MetPetDB to alterbasic research methodologies within the metamor-



14 of 15

phic petrology community. Hypothesis testingwill bepossible, in many cases, without the time and finan-cial expense required to plan and execute field work,collect, process, and analyze samples. If the quality ofthe data in the database is maintained at a high level,the researcher essentially enters the project at theinterpretation stage, which could, in many cases,dramatically lower the time and cost of researchand allow significant new directions to be explored.We are also in the process of developing a mobileapplication for iPhones, that will allow users to querythe database anywhere, find samples in the vicinity oftheir current location, and take field notes.

[47] Additionally, future data mining algorithmsmay allow researchers to explore aspects of globalgeochemistry that are currently impossible. Forexample, consider a hypothesis that a particulartype of Ca zoning in garnet signifies a particulartype of tectonic evolution. There might be severaldozen Ca zoning maps of garnet in the entiremetamorphic literature today, but there might bemany thousands in a well-populated database.Examining these by hand would be prohibitive,but future searching might incorporate image rec-ognition algorithms that would permit a query suchas ‘‘find all of the samples with Ca zoning in garnetthat demonstrates a core to rim increase of 20% ormore.’’ Sophisticated queries that require largedatabases are already being demonstrated byusers of the igneous geochemistry databases [e.g.,Spiegelman and Kelemen, 2003; Salters andStracke, 2004; Hirschmann et al., 2003] and weanticipate similar applications of MetPetDB.

Acknowledgments

[48] Members of the Metamorphic Petrology Database work-

ing group are thanked for their input and feedback over the

course of this project. We would also like to thank Jennifer

Rodger, Barbara Firebaugh, Samuel Broadaway, Julia Cos-

grove, Ann Cosgrove, Aric Mine and Chris Smith for archived

data extraction and programming error checking routines. John

Schumacher is thanked for his design of the MPDB logo. This

work is funded by NSF grant EAR-0622345.

References

Berman, R. G. (1991), Thermobarometry using multi-equilibrium calculations; a new technique, with petrologicalapplications, Can. Mineral., 29, 833–855.

Hirschmann, M. M., T. Kogiso, M. B. Baker, and E. M. Stolper(2003), Alkalic magmas generated by partial melting ofgarnet pyroxenite, Geology, 31, 481–484, doi:10.1130/0091-7613(2003)031<0481:AMGBPM>2.0.CO;2.

Lehnert, K. A., and C. H. Langmuir (2007), The PetDB datacollection: Impact on science, Geol. Soc. Am. Abstr. Pro-grams, 39, 153.

Lehnert, K., Y. Su, C. H. Langmuir, B. Sarbas, and U. Nohl(2000), A global geochemical database structure for rocks,Geochem. Geophys. Geosyst., 1(5), 1012, doi:10.1029/1999GC000026.

Powell, R., and T. Holland (1994), Optimal geothermometryand geobarometry, Am. Mineral., 79, 120–133.

Salters, V. J., and A. Stracke (2004), Use of the PetDB databasefor calculating the composition of the MORB-source, EosTrans. AGU, 85(47), Fall Meet. Suppl., Abstract SF32A-03.

Schmatz, D. R., M. Engi, and J. E. Lieberman (1995), ParaDIS:A relational database for the consistent documentation andanalysis ofmetamorphicmineral assemblages,Comput.Geosci.,21, 1031–1041, doi:10.1016/0098-3004(95)00039-B.

Spear, F. S., and R. Parrish (1996), Petrology and petrologic cool-ing rates of the Valhalla Complex, British Columbia, Canada,J. Petrol., 37, 733–765, doi:10.1093/petrology/37.4.733.

Spear, F. S., S. M. Peacock, M. J. Kohn, F. P. Florence, andT. Menard (1991), Computer programs for petrologic P-T-tpath calculations, Am. Mineral., 76, 2009–2012.

Spiegelman, M., and P. B. Kelemen (2003), Extreme chemicalvariability as a consequence of channelized melt transport,Geochem. Geophys. Geosyst., 4(7), 1055, doi:10.1029/2002GC000336.



15 of 15

MetPetDB: A database for metamorphic geochemistry

Documents