Data Models for Radio Astronomy in the VO

0.15 MB·English

by  J. D. Santander-Vela

1
DATAMODELSFORRADIOASTRONOMYINTHEVO
J.D.Santander-Vela1,2
1InstitutodeAstrof´ısicadeAndaluc´ıa-CSIC
2AMIGAcollaboration
9
0
0
2

n
ABSTRACT antennaarchive(soonto be published),and the DSS-63
a
J archive.

6
DataModelsareanessentialpartofautomaticdatapro- Earlyduringthedevelopmentphase,adecisionwasmade
1
cessing,butevenmoresowhentryingtotietogetherdata thatwewouldnotjustprovidea VOcompatibilitylayer

] coming frommanydifferentdata sources, as is the case ontopofthesearchives’infrastructures,butthattheinter-
M fortheInternationalVirtualObservatory. Inthistalkwe nal archive organisationshould reflect existing VO data
will review the differentdata modelsused in the IVOA, modelsinordertoassurethattheVOinterfacewouldbe
I
. whichpartsofthatDataModellingworkarestillincom- abletoprovidethemostmetadata.
h
plete, specially in radio wavelengths, and the work the
p
AMIGA group has done within the IVOA Data Mod-
-
o elling Working Group to overcome those shortcomings 2. DATAMODELS
r both in missing data models and support for Radio As-
t
s tronomy.
a Datamodelsarethedetaileddescriptionofthesetofen-
[
Keywords: VirtualObservatory;DataModelling;Radio titiesneededforinformationstorageinaparticularfield,
1 and specifyboth the data beingstored, and the relation-
Astronomy.
v shipsbetweenthem. Datamodelsarepartofthehidden
0 VO infrastructureastronomerswould normallyneverbe
2 involvedwith,butknowledgeofdatamodelscanenhance
5 1. INTRODUCTION theopportunitiesfortheexploitationoftheVO.
2
1. Within the VO, data models apply not only directly to
0 TheAMIGAproject(AnalysingtheinterstellarMedium the scientific data, but to the metadata describing them.
9 of Isolated GAlaxies) was born in 2003, and intends to As the way to structure information dependson the ap-
0 provideastatisticalcharacterisationofastrictlyselected plicationdomain,VOdatamodelsdescribeastronomical
: sampleofisolatedgalaxiescomposedbymorethan1000 datasetsinawaythatisasinstrumentindependentaspos-
v
i objects, by means of multi-wavelengthdata, and with a sible,toensurethatthesamedescriptioncanbeusedfor
X particular emphasis on radio data at cm, mm, and sub- datawithdifferentprovenance.Usersmustalsobeableto
r mm wavelengths. All these data are being periodically querythosedatamodelstobeabletofinddatasetswhich
a releasedviathewebpageoftheproject1,whichprovides complywithcertainproperties.
aVirtualObservatory(VO)ConeSearchinterface.
The IVOA Data Modelling Working Group (DMWG)
AMIGA+isthenaturalextensiontoAMIGA,withthree started an an effort to provide a complete data model
differentgoals:exploitationoftheAMIGAcatalogue,se- for astronomical observations, the Data Model for Ob-
lectingthebestcandidatesforadetailedstudyofisolated servations(McDowelletal.,2005). Oneofthemostim-
galaxies; scientific extension to the millimetre and sub- portant parts of it was the Characterisation of datasets,
millimetrerange;andparticipationinthedevelopmentof thatis,thecompletespecificationofwherethosedatasets
systemsallowingtheaccessanddisplayoflargeradioas- couldbefoundinthespatial,temporal,andspectralaxes,
tronomicaldatabases,bothsingle-dishandinterferomet- with more axes avalaible (i.e., polarisation) for suitable
ric,withintheVOframework. datasets. The Observation data model was put on hold,
and the Data Model for Astronomical Dataset Charac-
During the AMIGA+ projects two VO-compliant radio terisation(McDowelletal.,2007)wasstarted.
astronomical archives were developed: the IRAM 30m
We have built a complete observation data model
1http://amiga.iaa.csic.es/ for single-dish radio telescopes, the Radio Astro-
2
nomical DAta Model for Single-dish telescopes fieldcontainingtheRightAscensioncanbeidentifiedby
(RADAMS) (Santander-Velaetal., 2008), based on the UCD atom pos.eq.ra, while a photometric flux
thosetwodocuments. in the V band can be identified by the two UCD atoms
phot.flux; em.opt.V. This provides both a uni-
fiedvocabularytoidentifyanyastrophysicalquantity,and
anautomaticknowledgediscoverytoolforfieldswithar-
3. DATAMODELELEMENTS
bitrary relationships. In fact, UCDs were born out of a
jointCDS/ESOdataminingeffort(Ortizetal.,1999).
WhendefiningaVOdatamodel,wehavetospecify:
However,UCDscanonlyprovidedatakindinformation,
but not relationship information. In a sense, they are a
kindofspecialisedunit,complementary—orthogonal—
Entities Being the data model building blocks, they
to physical units: in the same way that quantities with
group related attributes within a data model. They
the same physical units can be very different in nature
can be mapped to Classes in Object-Oriented Pro-
(i.e.,decaytimeforanisotopeversusoscillationperiod),
gramming(OOP),orElementsinXML.
fields with identical UCDs can also be related to differ-
Fields They are the actual data elements of the model. entreal-wordphenomena. Inordertoallowsuchdeeper
They map to Attributes in OOP, and they can be relationshipstobeexpressed,anddisambiguatemetadata
mapped to Attributes or to Elements without chil- fieldsUTypeswereborn.
dreninXML.
UTypes are created from a hierarchical data model
Relationships The different entities and fields have hi- by enumerating the different parents a particular field
erarchicalorrelationalrelationships:anobservation has in that hierarchy. For instance, a field con-
projects has projectedobservations, and all entities taining the Right Ascension in equatorial coordinates
which share a common project ID are related, for for where an instrument was pointed to corresponds
instance. Forthedatamodeltobeuniquelydefined to the spatial coverage characterisation, in particular
thoserelationshipsmustbemadeexplicit. to the Location property, and thus it would sport a
UType of characterisation.coverage.spa-
Datatypes For computers to be able to correctly inter- tial.location, the UCD would be pos.eq.ra,
pretadatastreamaDatatypeneedstobespecified. anditsunitscouldbeanyangularunit.
Forinstance,objectIDscouldbeIntegers,butthey
are normally textual, so String data must be used. Butevenwiththehelpofunits,UCDsandUTypes,some-
Wecouldconsidertherestrictionswhichcanbede- times it can be difficult to tag a particular piece of data
finedforcomplexdatatypesinXMLaspartofthe withmeaningfulsemantics,speciallyfordatawhichdoes
datatyping. nothaveadirectplaceina VOdatamodel. Forthatwe
canborrowtechniquesfromtheSemanticWeb(aneffort
Units Nophysicalquantitycanbespecifiedwithoutpro-
forprovidingwebdocumentswithsemantics,sothat,for
viding its units. Physical-data related Fields need
instance, a table of camera prices can be tagged so that
Units to be specified, or Units have to be a fixed
software tools can identify in it prices, if possible be-
property of certain Fields, but they either need to
longingtodigitalcameras,eventoparticularbrands),and
existasanimplicitattributeofa particularfield, or
provide one or more standardised astronomical vocabu-
tohavetheirowndedicatedField. laries. TheIVOASemanticsWG3 hasstartedrecreating
controlledvocabulariessuch as UCDs in Semantic Web
Semantics As observation metadata are related to real-
form, and even the IAU thesaurushas been recreated in
world elements and quantities, VO data models
thatway(Derriereetal.,2008). Weareusingtheminthe
should specify semantics —i.e., what is exactly
IRAM30marchiveinordertoprovidesemanticstodata
meant in the real world by a particular field— to
comingfromantennaengineeringterms.
avoid ambiguities. Most of VO semantics are pro-
vided via Unified ContentDescriptors (UCDs) and
UTypes.
5. ROLEOFDATAMODELSINTHEVO
4. SEMANTICS, UCDS,UTYPESANDIVOAVO- We canidentifyintheVOfourdifferentphases,andwe
CABULARIES canseethatinallofthemdatamodelsplayacentralrole:
UCDs are a controlled vocabulary2, under the supervi- Discovery Datasets available in the VO have to be
sionoftheIVOASemanticsWG,whichprovidesalistof discoverable for them to appear automatically in
atomswhichcanbeusedtoidentifyfieldsascorrespond- VO tools. The VO Registry holds data for exist-
ing to specific astronomical quantities. For instance, a ing datasets so that they can be easily discovered.
2http://www.ivoa.net/Documents/latest/ 3http://www.ivoa.net/cgi-bin/twiki/bin/view/
UCDlist.html IVOA/IvoaSemantics
3
Data Model for Observation
Observation Curation
ObsData
Policy
Characterization Provenance Target Packaging
Coverage Resolution Precision Sensitivity
Accuracy
Data Model for
Dataset Characterisation
Figure1. RADAMSstructure: isacombinationoftheObservationandCharacterisationdatamodels,whichhavebeen
completelyspecifiedforsingledish radiotelescopes. Datamodelswhich arecompletelynewintheirdefinitionforany
kindofinstrumentaremarkedingray.
ThedatamodelforSpace-TimeCoordinates(STC), datamodel,thatapartfrombeinganintegralpartof
datasetCharacterisation(CharDM),Resourcemeta- the Observationdata model(ObsDM), it should be
data (ResDM), and the UCD and IVOA thesaurus builtinastand-aloneformsothatitcanbeapplied
(IVOAT)arerelevantinthisphase. tonewlygenerated,non-observationaldata.
Evaluation Datasetshavetobeevaluatedinordertoas-
sess their applicability to the kind of analysis we
6. RADAMSSTRUCTUREANDPROPERTIES
might wish to perform; for instance, in orderto do
imagemosaicingweneedacertaincoordinateover-
lap,andinordertodoimagestackingweneedanal-
AfterhavingpresentedtheimportanceofVOdatamodels
mostcompleteoverlap,andcomparableresolutions.
for the different activities involved in VO data queries,
The main data model involved in this phase is the
analyses and transformations, we will present the main
CharDM. RADAMSfeatures.
DataAccess There is an implicit data model in the
Figure1showstheRADAMSmainstructure,whichcan
IVOAdataaccessprotocols,theDataAccessLayer,
be seenasacombinationoftheObservationdatamodel
which is centredon targets(coordinateswith toler-
andtheCharDMdatamodel,butfullyspecifyingclasses
ances/searchradii),andusesseveralpropertiesfrom
that were only laid out by the ObsDM. Some particular
theCharDM,suchastheCoverageinseveralaxes.
adaptations of the CharDM to radio astronomy, specifi-
callyfortheSensitivityclass, havebeenperformed,and
Transformation Whencreatinganewdataset,ortrans-
theTargetclassisabletodealwithradiocatalogues.
forming an existing one, a new CharDM instance
needs to be created. If the transformed data set is
However, RADAMS main contributions come from the
a spectrum, the Spectral data model (SpecDM) is
ObsDMclasseswhichhavebeencompletelyspecified:
needed bothfor obtainingthe completedescription
of the originaldata and describing the transformed
product. Thereisnoexistingdatamodelyetforim- Packaging We have developed a VOPack packaging
ages or for more complex data within the VO. In standard, which embeds characterisation informa-
addition, in order to trace the origin of the trans- tionforanypackagedobservationset,andcanspec-
formed image we would need to use a Provenance ifytherecursiveinclusionofotherpackages.
4
Policy ThePolicyclassthatwehavedevelopedisableto AKNOWLEDGMENTS
accommodate different user roles determined from
the user identification and the explicit or implicit
policiesforeachparticulardataset. TheauthorwishestoacknowledgeEmilioGarc´ıaforhis
support in the early development of the RADAMS, his
Provenance Initiallyspecifiedonlyforradioastronomy discussionswithStephaneLeonforclarificationsanden-
data,weareworkingtomakeitmoregeneral,andto richmentoftheRADAMS,andthedirectionofJose´Fran-
beabletouseitoutsideofobservationalscopes,so ciscoGo´mezandLourdesVerdes-Montenegroduringmy
that software tools which provide dataset transfor- Master Thesis, were most of this work was developed.
mationscandocumenttheoriginoftheirprocessed FurtherdiscussionswithJ.E.RuizdelMazoandV.Espi-
datasets. garesduringtheimplementationphasearealsoacknowl-
edged, as well as the discussions inside the IVOA Data
ModelingWorkingGroup.Partofthisworkhasbeenper-
The RADAMS has been implemented in two archives: formed with the fundingof Spanish EducationMinistry
the Telescope Access for Public Archive System projects AYA 2005-07516-C02and AYA2005-24102-E,
(TAPAS) for the IRAM 30m, to be announced in the andtheEuropeanFP7EuroVO-AIDAproject.
February 2009 IRAM Newsletter, and the scientific
archiveofNASA’sDeepsSpace Centerin Madrid70 m
antenna,DSS-63,andhasshownthatVOdatamodelscan
REFERENCES
beusedasa blueprintforarchiveswhicharebeingbuilt
fromscratchforVOcompatibility.
Derriere,S.,Gray,A.J.G.,Gray,N.,etal.2008,Vo-
cabulariesintheVirtualObservatory,IVOAProposed
Recommendation Version 1.13, International Virtual
7. FUTUREWORK
ObservatoryAlliance
McDowell, J., Bonnarel, F., Chilingarian, I., et al.
Inthefuture,weplantohavetheRADAMSProvenance 2007, Data Model for Astronomical DataSet Char-
data model, part of the author’s thesis (to be published acterisation, IVOA Working Draft v1.1, International
in 2009), contributed to the IVOA DMWG, and have it VirtualObservatoryAlliance
integratedwiththeObsDM.
McDowell, J., Bonnarel, F., Giaretta, D., et al. 2005,
WearestartingalsoacollaborationwiththeALMASci- Data Model for Observation, IVOA Data Modelling
entificArchiveteaminordertocreateinthefuturebotha WG InternalDraftv0.23,InternationalVirtualObser-
datamodelforALMAdatacubeswhichcanbeintegrated vatoryAlliance
in a futureIVOA data modelformultidimensionaldata,
and providing VO compatibility to the ALMA Science Ortiz, P.F., Ochsenbein,F., Wicenec,A.,&Albrecht,
Archive. M. 1999,inAstronomicalSocietyofthePacificCon-
ference Series, Vol. 172, AstronomicalData Analysis
SoftwareandSystemsVIII,ed.D.M.Mehringer,R.L.
Plante,&D.A.Roberts,379–+
8. CONCLUSIONS
Santander-Vela, J. D., Garcia, E., Gomez, J. F., et al.
2008,ArXive-prints
Data models are one of the three bases interoperability
relies on. As the data models for astrophysical obser-
vations’ metadatamust be interoperableacrossdifferent
softwarepackagesandinstrumentdomains,noteveryde-
tail can be included in the data models, or can be de-
scribedusingIVOA’sUCDsandUTypes. Semanticweb
technologies, such as thesauri expressed in W3C stan-
dards,canbeusedtoprovideadditionalsemantics.
However,theIVOAhigh-leveldatamodellingeffortsare
complete enough as to have guided the developmentof
theRADAMSdatamodel. TheRADAMShasbeensuc-
cessfully usedtocreate twooperationalradioastronom-
ical archivesfor two very differentantennas and instru-
mentsystems,suchastheIRAM30mandtheDSS-63.
In the future, the ObsDM must be complete enough to
supportthe complexdatasets to be producedby ALMA
and radio telescopes further ahead in time, such as the
LOFARandotherSKApathfinders.