European Commission logo
INSPIRE Community Forum

FoI Encoding

Hi Ilkka,

here the counterhijack of the subtopic FoI encoding! ;)

Comments inline with KS: prefix and blue

By Ilkka RINNE 2 hours ago

Hello Kathi,

Sorry for hijacking your discussion topic, not to worry I'm not asking you to return to the drawing table :-) The issue is real however: It's currently very difficult to refer from inside OMSO Observation instances to the EF facilities which were used for creating them.

KS: first question is why you didn't create a subtype of SF_SpatialSamplingFeature for this purpose? what we did for the AQD e-Reporting. Example further below.

For FMI case I just used a plain sams:SF_SpatialSamplingFeature instance as the FOI value, and coded the weather observation station name and id in a custom element used as th value of the sampledFeature:

<om:featureOfInterest>
  <sams:SF_SpatialSamplingFeature gml:id="sampling-feature-1-1-fmisid">
    <sam:sampledFeature>
      <target:LocationCollection gml:id="sampled-target-1-1">
        <target:member>
          <target:Location gml:id="obsloc-fmisid-101007-pos">
            <gml:identifier         codeSpace="http://xml.fmi.fi/namespace/stationcode/fmisid">101007</gml:identifier>
            <gml:name codeSpace="http://xml.fmi.fi/namespace/locationcode/name">Helsinki Rautatientori</gml:name>
            <gml:name codeSpace="http://xml.fmi.fi/namespace/locationcode/geoid">-16011680</gml:name>
            <gml:name codeSpace="http://xml.fmi.fi/namespace/locationcode/wmo">2934</gml:name>
            <target:representativePoint xlink:href="#point-101007"/>
            <target:region codeSpace="http://xml.fmi.fi/namespace/location/region">Helsinki</target:region>
          </target:Location>
        </target:member>
      </target:LocationCollection>
    </sam:sampledFeature>
    <sams:shape>
      <gml:MultiPoint gml:id="mp-1-1-fmisid">
        <gml:pointMember>
          <gml:Point gml:id="point-101007" srsName="http://www.opengis.net/def/crs/EPSG/0/4258" srsDimension="2">
            <gml:name>Helsinki Rautatientori</gml:name>
            <gml:pos>60.17169 24.94460</gml:pos>
          </gml:Point>
        </gml:pointMember>
      </gml:MultiPoint>
    </sams:shape>
  </sams:SF_SpatialSamplingFeature
</om:featureOfInterest>

(using MultiPoint and LocationCollection due to fact that there data for may be more than one station in the same response)

KS: you mean you list multiple stations in the LocationCollection, then multiple geometries in the MultiPoint. and then crosslink via the target:representativePoint xlink:href entry? why not provide multiple SpatialSamplingFeature instances?

KS: ok, now the promised AQD example. FoI is the air bubble around the inlet (AQD_Sample), sampledFeature is the representative area around the station. While i don't have a link to the station from the sample, it would be no problem to add, just as we've added other relevant information such as distances to buildings/curbs. (in AQD we reference the station from the observation). here the AQD example:

<!-- AQD_Sample  gml:id="SamplingFeature_1" -->
<gml:featureMember>
    <aqd:AQD_Sample gml:id="SamplingFeature_1">
        <sam:sampledFeature xlink:href="http://data.umweltbundesamt.at/inspire/AT.UBA.AQD/SampledFeature_1"/>
        <sams:shape>
            <gml:Point gml:id="SamplingFeaturePoint_1" srsDimension="2" srsName="urn:ogc:def:crs:EPSG::4258">
                <gml:pos>19.6856534817236 48.6886762376192</gml:pos>
            </gml:Point>
        </sams:shape>
        <aqd:inspireId>
            <base:Identifier>
                <base:localId>SamplingFeature_1</base:localId>
                <base:namespace>AT.UBA.AQD</base:namespace>
            </base:Identifier>
        </aqd:inspireId>
        <aqd:inletHeight uom="http://dd.eionet.europa.eu/vocabulary/uom/length/m">5</aqd:inletHeight>
        <aqd:buildingDistance uom="http://dd.eionet.europa.eu/vocabulary/uom/length/m">6</aqd:buildingDistance>
        <aqd:kerbDistance uom="http://dd.eionet.europa.eu/vocabulary/uom/length/m">7</aqd:kerbDistance>
    </aqd:AQD_Sample>
</gml:featureMember>
<!-- AQD_RepresentativeArea  gml:id="SampledFeature_1"-->
<gml:featureMember>
    <aqd:AQD_RepresentativeArea gml:id="SampledFeature_1">
        <sam:sampledFeature xlink:href="http://sweet.jpl.nasa.gov/2.2/realm.owl#Atmosphere"/>
        <sams:shape>
            <gml:Polygon gml:id="SampledFeaturePolygon_1" srsName="urn:ogc:def:crs:EPSG::4258">
                <gml:exterior>
                    <gml:LinearRing>
                        <gml:posList srsDimension="2" count="4">19.6856534817236 48.6886762376192 19.6850950126256 48.6892127722526 19.6850150763855 48.6898202553015 19.6850478566741 48.6906926606387 </gml:posList>
                    </gml:LinearRing>
                </gml:exterior>
            </gml:Polygon>
        </sams:shape>
        <aqd:inspireId>
            <base:Identifier>
                <base:localId>SampledFeature_1</base:localId>
                <base:namespace>AT.UBA.AQD</base:namespace>
            </base:Identifier>
        </aqd:inspireId>
        <aqd:description>A description of the representative ares</aqd:description>
        <aqd:documentation>http://www.umweltbundesamt.at</aqd:documentation>
    </aqd:AQD_RepresentativeArea>
</gml:featureMember>

KS: I don't like the idea of embedding the station information into the sampledFeature as it's semantically wrong. However, i'm wondering if it might make sense to define an INSPIRE Sample type which has an INSPIRE ID and optionally a link back to the station. Then one could derive all further INSPIRE relevant O&M FoI bits from this.

Ideas?

:)

Kathi

 

 

  • Ilkka RINNE

    Hi Kathi, thanks for spawning this into a separate thread.

    Why we didn't create a subtype of SF_SpatialSamplingFeature for modelling an observation station or instrument is a good question. I guess we could change this in the future, and add a property called "site" etc. to refer to EF MonitoringFacility via WFS GetFeature request.

    It seems that in your AQD example you think of a sample (an air specimen captured at place x and time y) as the samplingFeature,is this correct? Are the AQD_Sample and AQD_RepresentativeArea both directly extending SF_SpatialSamplingFeature? Did you consider using SF_Specimen as parent type for the AQD_Sample?

    In FMI case the samplingFeature is seen more as an instrument (sampler) capable of providing estimates of the properties of it's surrounding media.

  • Ilkka RINNE

    The reasoning behind the multiple locations and points needs a bit more explaining.

    At FMI they were not very fond the verbosity of the WaterML timeseries model, which is mandated by the D2.9 chapter 6.6.1.1.2. PointTimeSeriesObservation. They needed a model which would be able to provide observation (or forecast) data for several locations, several time steps and several observable properties in a reasonably concise format. In the WaterML 2.0 timeseries (using time-value pairs) you can only provide the value for only one property at each point in time, which results in duplicating all the timestamps for each or the observable properties.

    So my alternative solution was to use GridSeriesObservation, with an irregular 4-D grid having it's grid points at the position of each of the observation stations (first 3 dimensions) and at each time step (4th dimension). The result of the GridSeriesObservation is grid coverage providing values for all the observed properties at each of the grid points.

    In order to match the geometries of the samplingFeature to the result grid, the samplingFeature shape also needed to contain all these points - as the measurement data is provided at these discrete points only, not in between them. The sampledFeature in the point observation case is also defined as a collection of real world locations (places), not stations, even though the geometries of these places are (singular) points (as specified by the D2.9 for point time series data). The station code provided is really just a location identifier, a precise place name.

    The GridSeriesObservation timeseries solution is reasonable concise (as XML goes), but also quite a bit more complex due to the long lists of grid coordinates and the values which must be mapped to those points be the client. So in the end FMI ended up providing point data in both formats.

    For your last suggestion about the Sample type, is this close to what you had in mind:

    <element name="EstimatingAppliance" type="ef:EstimatingApplianceType" substitutionGroup="sams:SF_SpatialSamplingFeature">
            <annotation>
                <documentation>EstimatingAppliance is a physical appliance or instrument capable of creating estimates
                of properties of the environment, like an observation instrument or station located within an 
                EnvironmentalMonitoringFacility.
                </documentation>
            </annotation>
        </element>
        <complexType name="EstimatingApplianceType">
            <annotation>
                <documentation>EstimatingAppliance</documentation>
            </annotation>
            <complexContent>
                <extension base="sams:SF_SpatialSamplingFeatureType">
                    <sequence>
                        <element name="inspireId" type="base:IdentifierPropertyType" minOccurs="0">
                            <annotation>
                                <documentation>-- Definition --
                                    External object identifier.</documentation>
                            </annotation>
                        </element>
                        <element name="facility" type="ef:EnvironmentalMonitoringFacilityPropertyType" minOccurs="0" />
                    </sequence>
                </extension>
            </complexContent>
        </complexType>

     

  • Katharina SCHLEIDT

    By Katharina SCHLEIDT

    Hi Ilkka,

    to your first response, a few answers:

    to my understanding of O&M, providing the measurement instrument as samplingFeature is semantically wrong:

    From 19156: This feature is the real-world object whose properties are under observation, or is a feature intended to sample the real-world object

    Linking this sample with the station seems clean, but claiming you're measuring the humidity (or whatever) of a sensor doesn't logically work.

    As for why i didn't specialize SF_Specimen for AQD_Sample is because the way a specimen is defined in the standard, it's a representative bit the FoI that's removed and stored. Thus, all the attributes available pertain to things like current location, samplingMethod, size - none of which make sense for the air bubble around a sensor or intake. In this case, it corresponds far better with SF_SamplingFeature, which is where i left it

    :)

  • Katharina SCHLEIDT

    By Katharina SCHLEIDT

    Now to the 2nd post:

    for a coverage result this could make sense. would really like to have some examples for O&M returning coverages - anybody else have any?

     

    To the proposed INSPIRE Sample type, similar to what you're proposing, but still different, reasoning:

    • definately not estimatingAppliance (or sensor, or other measurementDevice aphorism). should just be called Sample (like our INSPIRE Process is called Process). Again, the FoI is not the measurement device, please reread the standard if this isn't clear
    • At least the facility reference needs to be properly modelled to allow for xlink (don't want to duplicate the entire facility data here)

    Advantage of defining an INSPIRE Sample is that the base information (inspireId and Facility) would then always be provided in the same place. Anybody have other ideas on what else this basic type would need?

     

  • Ilkka RINNE

    Hi Kathi,

    I think you are correct, FOI is not intended to be a measurement device or sampler. But still ISO 19156 talks about "stations" as one category of sampling feature types:

    4.17
    sampling feature
    feature which is involved in making observations concerning a domain feature


    EXAMPLE Station, transect, section or specimen

    (4 Terms and definitions, page 5)

    9.1.1.3 Observed property is a proxy

    The procedure for obtaining values of the property of interest may be indirect, relying on direct observation of a more convenient parameter which is a proxy for the property of interest. Application of an algorithm or processing chain obtains an estimate of the ultimate property of interest.

    [snip]

    9.1.2 Role of sampling features
    Sampling features are artefacts of an observational strategy, and have no significant function outside of their role in the observation process. The physical characteristics of the features themselves are of little interest, except perhaps to the manager of a sampling campaign.

    EXAMPLE A “station” is essentially an identifiable locality where a sensor system or procedure may be deployed and an observation made. In the context of the observation model, it connotes the “world in the vicinity of the station”, so the observed properties relate to the physical medium at the station, and not to any physical artefact such as a mooring, buoy, benchmark, monument, well, etc.

    (9.1 The context for sampling, page 20) 

    "Station" is also mentioned as one of the "common names for sampling features" in chapter 9.1.3.

    So based on that I would not say that the semantic wrongness of thinking of an observation station as a sampling feature can be justified by referring to ISO 19156.

    As you noted a sample is also one typical type of sampling feature, as explained in chapter "9.1.1.2 Proximate feature-of-interest embodies a sample design":

    For various reasons, the domain feature may not be fully accessible. In such circumstances, the procedure for estimating the value of a property of the domain feature involves sampling in representative locations.

  • Ilkka RINNE

    Some live examples of serving O&M coverages from FMI Open Data WFS:

    (hmm, I might have to apply for a new api-key after exposing this one :-))

  • Ilkka RINNE

    Despite my comments on the wording of ISO 19156 I think we are nicely approaching a synthesis here: The "station" in the 19156 sampling feature sense is an identifiable locality where a sensor system may be deployed. This is actually very close to the Sample in the sense you Kathi used it: an "air bubble around a sensor or intake". So it's some kind of vicinity (area) inside which the measurement instrument(s) are sampling the air or other media. This feature has no other function than capture the non-singular piece of space containing the medium used for the sampling at the moment of taking the sample.

    I agree that the semantics of the MonitoringFacility is a bit different: it's a place for people and/or machines to work and produce information about the properties of the surrounding environment. So if I may propose to change the name of my previously proposed feature type simply to SamplingArea. It may or may not have an InspireId and may or may not have a direct link to a Facility (via the "relatedFacility" property, of type ef:EnvironmentalMonitoringFacilityPropertyType?).

    I would like this solution because It could also be used for atmospheric balloon soundings, which have very long trajectory in the atmosphere as the shape of their sampling feature, but are still connected to the observation station where the balloon was sent.

     

  • Katharina SCHLEIDT

    By Katharina SCHLEIDT

    As to your balloons, not too different from the stuff we've got in the O&M Guidelines in section 5.2.2 Feature of Interest Curve

    :)

Environmental Monitoring & Observations

Environmental Monitoring & Observations

Environmental Monitoring Facilities, Observations and Measurements