Analysis of quality metadata in the GEOSS Clearinghouse
-
Upload
paula-diaz -
Category
Data & Analytics
-
view
63 -
download
0
description
Transcript of Analysis of quality metadata in the GEOSS Clearinghouse
Analysis of the Quality Metadata in
GEOSS Clearinghouse
QUAlity aware VIsualisation for the Global Earth Observation
system of systems
SEVILLANO Eva1, DÍAZ Paula 2, NINYEROLA Miquel1, MASÓ Joan2 , ZABALA Alaitz1, PONS Xavier1
1 UAB Universitat Autònoma de Barcelona.2 CREAF Centre for Ecological Research and Forestry Applications.
Objectives
• To get a first analysis of the data quality in the Clearinghouse
• Analyze the quality contained in the metadata (ISO 19115)– Quality indicators
www.geoviqua.org
– Quality indicators– Lineage– Usage
• Start building components for the GEO Portal– Quality Broker– Quality searcher– Quality visualization
Methodology
97203 XML documents
CSW
GEOSS Clearinghouse
www.geoviqua.org
• Harvest all XML documents, ISO 19115. (October 2011)
Methodology
97203 XML documents
Database
GestBD
Xpathextraction
CSW
GEOSS Clearinghouse
www.geoviqua.org
• Massive extraction of MD quality elements– Quality indicators– Lineage– Usage
Overall Results
• Total metadata records in the Clearinghouse– 97203
• Total number of quality indicators– 52187
www.geoviqua.org
– 52187
• Metadata records with quality indicators– 19107
• Metadata records with lineage– 10899 (9261 process, 3771
source)
• Metadata with usage– 1226
Quality Scope
• 19.66% Metadata records with quality indicators– 2.7 quality indicator per metadata record
www.geoviqua.org
1. Quality indicators
• 19.66% Metadata records with quality
www.geoviqua.org
Quality indicators
• 19.66% Metadata records with quality– 2.7 QI/MD
www.geoviqua.org
Quality indicators
www.geoviqua.org
Quality Indicators in IDEC Metadata
Quality indicators – Comparison Clearinghouse - IDEC
www.geoviqua.org
PositionalAccuracy95.38%
ThematicAccuracy
2.60%
Completeness0.46%
TemporalAccuracy
0.06%Logical
Consistency0.02%
Quality Indicators in IDEC Metadata
GEOSS
IDEC
Quality indicator result
85.8% (22275 QI)
14.18% (3669 QI
mainly conformance to INSPIRE)
www.geoviqua.org
0.02% (5 QI)
19115-2 Extension for "per pixel"
quality
Quality indicator result
www.geoviqua.org
Quality indicators - Quantitative
1000012000140001600018000
Nu
mb
er
of
qu
ali
ty e
lem
en
ts
Quality elements - Quantitative measures
www.geoviqua.org
02000400060008000
10000
Nu
mb
er
of
qu
ali
ty e
lem
en
ts
Complete value
Declare value
Quantitative type
Quality indicators - Qualitative
600800
100012001400
Nu
mb
er
of
qu
ali
ty e
lem
en
ts
Quality elements - Conformance measures
www.geoviqua.org
0200400600
Nu
mb
er
of
qu
ali
ty e
lem
en
ts
Conformance to specification
Declare conformance
Conformance type
Coverage result (ISO19115-2 extension)
• Clearinghouse record ID: 273234, 273232, 273233, 273235, 273236)• Only 5 records use this. Bad news for visualizing data + quality maps
• Title: OMNO2e:OMI Column Amount NO2:ColumnAmountNO2CS30
<gmd:DQ_QuantitativeAttributeAccuracy><gmd:measureDescription>
<gco:CharacterString>The 'version 003' product is the second public rele ase. It is based on improved radiance calibration. For details, please see docum ent:
www.geoviqua.org
radiance calibration. For details, please see docum ent: http://disc.sci.gsfc.nasa.gov/Aura/OMI/OMTO3e_v003. shtml </gco:CharacterString>
</gmd:measureDescription><gmd:result><gmi:QE_CoverageResult >
<gmi:spatialRepresentationType><gmd:MD_SpatialRepresentationTypeCode codeList="./resources/codeList.xml#MD_SpatialRepresentationTy peCode " codeListValue="grid ">grid </gmd:MD_SpatialRepresentationTypeCode></gmi:spatialRepresentationType>
<gmi:resultFile gco:nilReason="missing " /> <gmi:resultFormat>
<gmd:MD_Format><gmd:name><gco:CharacterString>CF-netCDF</gco:CharacterString></gmd:name></gmd:MD_Format>
</gmi:resultFormat></gmi:QE_CoverageResult ></gmd:result>
</gmd:DQ_QuantitativeAttributeAccuracy>
2. Lineage
www.geoviqua.org
2. Lineage
www.geoviqua.org
2. Lineage
www.geoviqua.org
LI_ProcessStep with LI_Source Example
Clearinghouse record ID 131007 (simplified)• Compile survey input data from the best and most cu rrent survey records.
– BLM database of the index to all official (microfilm, CD, other) BLM survey records.– USFS survey records.– Private land surveyor records– GCDB Data Collection Attribute Definitions Version 2.0, Appendix A, 2/14/1991. Survey records
used - source abbreviations.• Compile listings of known locations of PLSS corners .
– USGS topographic quadrangles and other sources.– USC&GS published coordinate data.– NGS published coordinate data.– BLM global positioning Data.– USFS global positioning data.
• Coordinates of control stations are entered into a control data base with associated reliabilities.
• Topologically correct GIS coverages are modified to use FGDC compliant naming
www.geoviqua.org
• Topologically correct GIS coverages are modified to use FGDC compliant naming conventions and then loaded into the LSI database. These layers can then be downloaded as shapefiles through the LSI website.
• GCDB Data was downloaded for Kiowa and Cheyenne Cou nties, Colorado .– C:\f\gis_data\sand\zipped\kiowa\twnshp.shp.xml
• Metadata imported and data was exported from region s format to shapefile format• Dataset copied .
– C:\f\gis_data\sand\data\basedata\plss\ck_gcdb_region_township• Source Contribution: Survey data in the form of off icial (microfilm, CD, other) survey
and BLM, abstracted into a vector digital format.on line• Source Contribution: Survey and control data from t he Cartographic Feature File
(CFF) data set.disc• Source Contribution: Digitized control data from st andard topological quadrangle
sheets.disc
LI_Lineage: LI_Source
• 6.02% metadata records (5851) contain direct list of the data sources.– 1.85% (1798) with temporal extent
class LI_Source_only
LI_Lineage
+ statement :CharacterString [0..1]+ scope :DQ_Scope [0..*]
constraints{"source" role is mandatory if LI_Lineage.statement and "processStep" role are not documented}
Metadata Information::MD_Metadata
+resourceLineage0..*
www.geoviqua.org
• Gives credit (attribution, and eventually some trust on them)
• If quality indicators are not provided for the dataset, the quality indicators from sources can be a clue.
LI_Source
+ description :CharacterString [0..1]+ sourceSpatialResolution :MD_Resolution [0..1]+ sourceReferenceSystem :MD_ReferenceSystem [0..1]+ sourceCitation :CI_Citation [0..1]+ sourceMetadata :CI_Citation [0..*]+ scope :DQ_Scope [0..*]
constraints{"description" is mandatory if "scope" is not documented}{"scope" is mandatory if "description" is not documented}
and "processStep" role are not documented}{"processStep" role is mandatory if LI_Lineage.statement and "source" role are not documented}
+source 0..*
LI_Lineage: LI_ProcessStep
• 8.26% metadata records (8035) contain the direct list of the processes without sources– 292 (0.30%) contain date
class From_LI_ProcessStep_to_LI_Source
LI_Lineage
+ statement :CharacterString [0..1]
+ scope :DQ_Scope [0..*]
constraints
Metadata Information::MD_Metadata
+resourceLineage0..*
www.geoviqua.org
• With the order of these processes.
• If quality indicators are not provided for the dataset, it’s difficult to infer resource quality with only a process list
LI_ProcessStep
+ description :CharacterString
+ rationale :CharacterString [0..1]
+ stepDateTime :TM_Primitive [0..*]
+ processor :CI_ResponsiblePartyInfo [0..*]
+ reference :CI_Citation [0..*]
+ scope :DQ_Scope [0..*]
constraints
{"source" role is mandatory if LI_Lineage.statement
and "processStep" role are not documented}
{"processStep" role is mandatory if
LI_Lineage.statement and "source" role are not
documented}
+processStep 0..*
Complete Provenance:MD_ProcessStep with MD_Source
• 1.26% metadata records (1226 ) with more complete provenance process .
• How and when the data sources where used
class From_LI_ProcessStep_to_LI_Source
LI_Lineage
+ statement :CharacterString [0..1]+ scope :DQ_Scope [0..*]
constraints{"source" role is mandatory if LI_Lineage.statement and "processStep" role are not documented}
Metadata Information::MD_Metadata
+resourceLineage0..*
www.geoviqua.org
sources where used
• If quality indicators are not provided for the dataset, we can deduce which sources have more influence in the quality of the final result
LI_Source
+ description :CharacterString [0..1]+ sourceSpatialResolution :MD_Resolution [0..1]+ sourceReferenceSystem :MD_ReferenceSystem [0..1]+ sourceCitation :CI_Citation [0..1]+ sourceMetadata :CI_Citation [0..*]+ scope :DQ_Scope [0..*]
constraints{"description" is mandatory if "scope" is not documented}{"scope" is mandatory if "description" is not documented}
LI_ProcessStep
+ description :CharacterString+ rationale :CharacterString [0..1]+ stepDateTime :TM_Primitive [0..*]+ processor :CI_ResponsiblePartyInfo [0..*]+ reference :CI_Citation [0..*]+ scope :DQ_Scope [0..*]
and "processStep" role are not documented}{"processStep" role is mandatory if LI_Lineage.statement and "source" role are not documented}
+processStep 0..*
+source
0..*
Complete provenance in ISO19115-2
• LI_ProcessStep includes a LE_Processing that has a runTimeParameters attribute that allows us describing the exact list of parameters used in the execution.
• There is a citation of the algorithm used (LI_Algorithm).
class From_LE_ProcessStep_to_LE_Source
LI_Source
+ description :CharacterString [0..1]+ sourceSpatialResolution :MD_Resolution [0..1]+ sourceReferenceSystem :MD_ReferenceSystem [0..1]+ sourceCitation :CI_Citation [0..1]+ sourceMetadata :CI_Citation [0..*]+ scope :DQ_Scope [0..*]
constraints{"description" is mandatory if "scope" is not documented}{"scope" is mandatory if "description" is not documented}
LI_ProcessStep
+ description :CharacterString+ rationale :CharacterString [0..1]+ stepDateTime :TM_Primitive [0..*]+ processor :CI_ResponsiblePartyInfo [0..*]+ reference :CI_Citation [0..*]+ scope :DQ_Scope [0..*]
LI_Lineage
+ statement :CharacterString [0..1]+ scope :DQ_Scope [0..*]
constraints{"source" role is mandatory if LI_Lineage.statement and "processStep" role are not documented}{"processStep" role is mandatory if LI_Lineage.statement and "source" role are not documented}
Metadata Information::MD_Metadata
+processStep 0..*
+resourceLineage
0..*
www.geoviqua.org
used (LI_Algorithm).• All these extensions were done for
the benefit of the EO gridded data, but there are not in the Clearinghouse.
• We can completely evaluate the quality of the resulting product if we know the uncertainties that sources have in their metadata (sourceMetadata citation in LI_Source).
From ISO 19115-2:2009 shown for informative purposes only
Data quality information - Imagery::
LE_ProcessStep
Data quality information - Imagery::LE_ProcessStepReport
+ name :CharacterString+ description :CharacterString [0..1]+ fi leType :CharacterString [0..1]
If "LE_NominalResolution.scanningResolution" is usedthen "LE_Source.scaleDenominator" is required
Data quality information - Imagery::LE_Source
+ processedLevel :MD_Identifier [0..1]+ resolution :LE_NominalResolution [0..1]
Data quality information - Imagery::LE_Processing
+ identifier :MD_Identifier+ softwareReference :CI_Citation [0..1]+ procedureDescription :CharacterString [0..1]+ documentation :CI_Citation [0..*]+ runTimeParameters :CharacterString [0..1]
Data quality information - Imagery::LE_Algorithm
+ citation :CI_Citation+ description :CharacterString
«Union»Data quality information - Imagery::
LE_NominalResolution
+ scanningResolution :Distance+ groundResolution :Distance
"description" is mandatory if "sourceExtent" is not documented
"sourceExtent" is mandatory if "description" is not documented
+report 0..*
+output
0..*
+processingInformation0..1
+algorithm 0..*
3. Usage - User feedback
www.geoviqua.org
• There is one small entry for user feedback in the current ISO-19115:
• MD_Usage– Brief description of ways in which the
resource is currently or has been used
• There are 1.2% (1133) entries– SpecificUsage and – UserContactInfo, only
• All made by the same institution !:
MD_Usage - User feedback
www.geoviqua.org
– Landesvermessung und GeobasisinformationBrandenburg (LGB)
– Tel +49-331-8844-123, Fax. +49-331-8844-16123 – Heinrich-Mann-Allee 103, Potsdam, Brandenburg 14473,
Deutschland– [email protected] – http://www.geobasis-bb.de
Conclusions
• There are many different kinds of quality indicators– There is a lack of a complete description of values provided (no units, missing
measure name, missing evaluation method)
• Quality coverage results (by pixel) are almost inexistent and the the link is not there
• Lineage information is rich in many records, some with more that 100 entries in source or ProcessSteps
www.geoviqua.org
entries in source or ProcessSteps• We have usage examples -> Feedback• Current data is enough to demonstrate search and visualization with
some limitations. Good for GeoViQua.
• Next steps:
– Assess the Quality of Quality Metadata?
– Extend this analysis to other capacity catalogues integrated in the EuroGEOSSBroker
Thank you!Danke!Grazie!Merci!
Ευχαριστίες!
Diolch!Bedankt!
Köszönöm!Ačiū!
Благодарам! Спасибі!Ευχαριστίες!Vďaka!
Tak!Díky!
Tänan!Kiitos!
Благодарам!Dzięki!
Mulţumiri!Хвала!Tack!
Teşekkürler!
Спасибі!Спасибо!Obrigado!
Takk!Gràcies!Gracias!
QUAlity aware VIsualisation for the Global Earth Observation
system of systems