Abstract1. Introduction2. Contents of the Database3. The Relational Database Model4. Structure of the Geochemical Database4.1. Sample-Related Information4.2. Problem of Unique Sample Identification4.3. Analysis-Related Information4.4. Reference-Related Information4.5. Table NORM5. Operational Aspects of the Database6. ConclusionsAcknowledgmentsReferencesFiguresFigure 1Figure 2Figure 3TablesTable 1Table 2Table 3A global geochemical database structure for rocksK. Lehnert, Y. Su, and C. H. LangmuirLamont-Doherty Earth Observatory of Columbia University, Palisades, New York 10964([email protected]; [email protected]; [email protected])B. Sarbas and U. NohlMax-Planck-Institut fuÈr Chemie, Abteilung Geochemie, Postfach 3060, Mainz, D-55020 Germany([email protected]; [email protected])[1] Abstract: This technical brief describes a geochemical and petrological database structure based onthe relational model that has broad applicability to chemical analyses of geological materials. Notablefeatures of the database structure are its comprehensiveness and flexibility. The structure consists of 34interrelated tables, which can accommodate any type of analytical values for all different materials ofrock samples (volcanic glasses, minerals, inclusions, etc.) and for samples from any tectonic setting. Abroad spectrum of supplementary information (metadata) is included that describes the quality of theanalytical data and sample characteristics, such as petrography, geographical location, and samplingprocess, and that can be used to evaluate, filter, and sort the chemical data. All data in the database arelinked to their original reference. The database structure can be implemented in any relational databasemanagement system (RDBMS). It is currently applied in two different rock database projects(RidgePetDB and GEOROC).Keywords: Database; petrology; geochemistry; data management.Index terms: Geochemistry; instruments and techniques.Received October 27, 1999; Revised December 8, 1999; Accepted December 9, 1999; Published May 24, 2000.Lehnert, K., Y. Su, C. H. Langmuir, B. Sarbas, and U. Nohl, 2000. A global geochemical database structure for rocks,Geochem. Geophys. Geosyst., vol. 1, Paper number 1999GC000026 [6408 words, 3 figures, 34 subfigures, 3 tables].1. Introduction[2] Development and testing of models for theEarth's dynamic systems require the scientificcommunity to take full advantage of all dataavailable. In the past this has been attempted byindividuals making spreadsheets on their owncomputer for the particular problem at hand.However, the amount of geochemical dataproduced and published in the Earth sciencesis growing exponentially. Twenty years ago adozen elements on a dozen samples was typicalfor a scientific paper in geochemistry. Todaythere are likely to be 50 elements on severalhundred samples, and the data quantity is solarge that it cannot even be published easily byconventional paper publication. Subsequent in-vestigators often do not even have access to thecomplete data sets, which consisted in part ofunpublished data. Individual compilations ofpublished data take large amounts of time tocreate and are inevitably incomplete, leading toselective and often nonrepresentative compar-isons and conclusions. The rise of interdisci-Technical BriefVolume 1May 24, 2000Paper number 1999GC000026ISSN: 1525-2027Copyright 2000 by the American Geophysical UnionG3G3GeochemistryGeophysicsGeosystemsPublished by AGU and the Geochemical SocietyAN ELECTRONIC JOURNAL OF THE EARTH SCIENCESGeochemistryGeophysicsGeosystemsMay 24, 2000.plinary science also means that scientists indiverse fields need to be able to access datathat can be evaluated for completeness andreliability. In addition, it is now recognized thatthe geochemical data themselves rapidly be-come outdated unless they are supported by thenecessary metadata: exact sample names andlocations, analytical methods and errors, archi-val data, etc. The need for a database thatcontains all the geochemical and petrologicaldata and supporting metadata in a form inwhich it can be accessed readily by the scien-tific community is evident.[3] We have designed a relational databasestructure for chemical and petrological data ofall types of rock samples that includes allessential metadata. Two major databases arecurrently using this structure, the Ridge-PetDB, a petrological database for the oceanfloor compiled at the Lamont-Doherty EarthObservatory [Lehnert et al., 1999] and GEOR-OC, a geochemical database for ocean islandsand other oceanic as well as continental rockscompiled at the Max-Planck-Institut fuÈr Che-mie in Mainz, Germany [Sarbas et al., 1999].2. Contents of the Database[4] A comprehensive geochemical database hasto accommodate far more data than the actualchemical values. Supplementary informationdescribing the data quality as well as theanalyzed samples needs to be incorporatedbecause it is essential for proper evaluation ofthe analytical data and for efficient recoveryand sorting of data. Two types of data in thedatabase, therefore, can be distinguished:1. Primary data comprise all analytical valuesfor rock samples including major oxide andtrace element compositions, radiogenic andstable isotope ratios, noble gas contents,and uranium series for different materialslike whole rocks, volcanic glasses, mineralphases, and melt or mineral inclusions.2. Secondary data include (1) analyticalmetadata that describe the analytical meth-od (technique, laboratory, procedures, er-rors, precision, standard values, correctionprocedures) and the material of the samplethat was analyzed (glass, whole rock,mineral, etc.), (2) sample metadata thatdescribe the rock sample and its prove-nance (petrography, age, sampling techni-que, tectonic setting and geographicalposition of the sampling site, samplingcruise, navigation method, repository, etc.)and (3) reference metadata that givebibliographical information for the refer-ence which reports the analytical values.[5] All of the information listed above repre-sents multidimensional, related data that cannotbe handled reasonably or efficiently in a two-dimensional flat file format database consistingof a number of separate spreadsheets. Searchesfor specific data would be complicated andinefficient because data need to be retrievedfrom each individual file separately. The opti-mal method for organization and delivery ofsuch complex but related data is a