phersh2
commented
1 month ago Kathi, this is a good point on removing a duplication of the spatial feature in TSML.
Since TSML is derived from both Observations and Coverages standards, and both of those have a spatial component, it's best to not have both spatial aspects from those respective standards in TSML.
The link to Figure 25 - Coverage Domain in a Timeseries is referring back to the prior TSML version 1.3, and showcases that the TS_DomainObject, being derived from Coverages (ISO 19123:2005) has the spatial component in the UML (spatialComponent of type GM_object).
But we also have the spatial aspect from Observations.
Moving forward, we propose to remove the spatial aspect from Coverages and focus on the keeping just the spatial aspect from Observations.
This would also maintain consistency with the requirements class in the TSML conceptual OGC doc 15-043r3 which denotes the spatial aspect in TSML derives from Observations, not Coverages:
An OM_Observation whose result is a discrete coverage that varies in time (c.f. OM_DiscreteCoverageObservation). _Here the OMObservation feature type provides the spatio-temporal context for the series.
8.4 Requirements Class: Timeseries Observation
Requirement /req/uml-timeseries-observation/resultDomain
The spatial domain of the timeseries result shall be consistent with the featureOfInterest of the observation.
Requirement /req/uml-timeseries-observation/featureOfInterest
_The featureOfInterest shall be of type or subtype of SFSpatialSamplingFeature, as defined by ISO 19156, which includes MonitoringFeature as defined in clause 8.20 of this standard, or any other Feature Type (for non-spatial timeseries).
Thus, we have a purely temporal coverage in the TSML 2.0 "Coverage Domain in a Timeseries".
So, moving forward, we need to ascertain is if the spatialElement is there because it is used, or there because it was the only way to satisfy the old Coverage version.
We believe it is to satisfy the old Coverage version, but would like to ask if that spatial Element from the old Coverages is used at all?
(we will keep this issue out of a covJSON encoding for now, as we plan to fold Observations into the resulting Coverage if we extended CovJSON)
I'm a bit confused on the TS_DomainObject. To date, my assumption had been that this is purely temporal, the times for which we have values. However, now looking at Figure 25 - Coverage Domain in a Timeseries, I see that there is also a link to a spatialElement, basically a duplicate to the FoI in the paired Observation.
When encoding as TVP I believe this becomes a mute point, but I'd be interested if anybody uses a coverage encoding where they have a 2-dimensional (or 3-D) domain by adding spatial axes (extent 1) for the FoI location?
Alternatively, does anybody utilize this for moving observations, with the domain being a set of time-location pairs?