Discussion #2681

Spatial data set vs. spatial object

Added by Michael Lutz over 4 years ago. Updated over 4 years ago.

Status:Assigned
Priority:Normal
Assignee:James Passmore

Description

I would like to discuss to what we map, in the WCS technical guidelines, the concepts of "spatial data set" and "spatial object" and whether we provide guidelines for both sets of operations included in the network service Regulation:

  • PART A: Download Operations (referred to as "pre-defined data set download service operations" in the current guidelines for WFS) - these are mandatory to be provided
  • PART B: Direct Access Download Operations - these are only to be provided "where applicable", i.e. they are basically optional

For the actual download operations, we have the following:

  • Part A: Get Spatial Data Set: allows the retrieval of a Spatial Data Set (note that this is the IR definition, while the Directive says that download services shall also enable the download of parts of spatial data sets)
  • Part B: Get Spatial Object: allows the retrieval of Spatial Objects based upon a query

"Spatial data set" and "spatial object" are defined as follows in the Directive:

  • ‘spatial data’ means any data with a direct or indirect reference to a specific location or geographical area
  • ‘spatial data set’ means an identifiable collection of spatial data
  • ‘spatial object’ means an abstract representation of a real-world phenomenon related to a specific location or geographical area

I think there is a relatively clear mapping from "spatial data set" to "coverage". As we discussed in the Ispra workshop in Oct 2014, according to the definition, "coverage" could also be mapped to "spatial object" (1 coverage is one spatial object). But I think, when this was written, people rather thought of spatial object in the vector world (e.g. a "building", a "river") and assumed that a spatial data set is basically a collection of spatial objects.

So for me the question is: do we try to provide guidance only the Get Spatial Data Set or also the Get Spatial Object operation? Looking at these operations in detail:

  1. Get Spatial Data Set operation
    • request shall include
      • Language
      • Spatial Data Set Identifier
      • Coordinate Reference System
    • response shall be the requested Spatial Data Set in the requested language and in the requested Coordinate Reference System
  2. Get Spatial Object operation
    • request shall include
      • Language [same as Get Spatial Data Set operation]
      • Spatial Data Set Identifier [same as Get Spatial Data Set operation]
      • Coordinate Reference System [same as Get Spatial Data Set operation]
      • Query, including any combination of the following search criteria (with logical and comparison operators):
        • Unique Resource Identifier of Spatial Data Set
        • all relevant key attributes and the relationship between Spatial Objects as set out in Regulation (EU) No 1089/2010; in particular
          • the Unique Identifier of Spatial Object and
          • the temporal dimension characteristics, including the date of update
        • bounding box, expressed in any of the Coordinate Reference Systems listed in Regulation (EU) No 1089/2010
        • Spatial Data Theme
    • response shall include:
      • the set of Spatial Objects which complies with Regulation (EU) No 1089/2010 and fulfils the search criteria in the query, in the requested language and in the Coordinate Reference System.
      • Spatial Objects metadata: at least the INSPIRE metadata elements of the set of Spatial Objects.

(1) can be easily mapped to the GetCoverage operation in WCS 2.0 core - as shown by James in the current TG draft.

For a mapping of (2), we would need to

  • map "the set of Spatial Objects which (...) fulfils the search criteria in the query, in the requested language and in the Coordinate Reference System" to the result of a GetCoverage operation including some sub-setting.
  • understand how to map the required search criteria, in particular "all relevant key attributes and the relationship between Spatial Objects (incl. the Unique Identifier of Spatial Object and the temporal dimension characteristics, including the date of update)

But it could be another option to say that we only do a mapping for (1), which, given the functionalities already offered by the GetCoverage operation in WCS 2.0 core, already covers much (but maybe not all) of the requirements for (2).

real-world-to-feature.pdf - Real world to Feature via Spatial Objects (347 KB) James Passmore, 15 Feb 2016 11:52 am

History

#1 Updated by Michael Lutz over 4 years ago

Dear MIWP-7b experts,

please have a look at and provide your views on this fundamental question that we need to address before progressing with the work on the WCS guidelines.

Cheers,

Michael

#2 Updated by Peter Baumann over 4 years ago

Let me start from the naive definition: "a coverage is a special kind of feature". Based on this, a coverage would be an object. Which also makes sense when it comes to identification: if you make a coverage the object (=first-class citizen) then you identify by its name, that's strightforward. If you make a pixel the first-class citizen then you identify it by coverage name + direct position coordinate, which is much more involved (because this identification, in a GetCoverage request, is distributed over several parameters, and you can have coordinate wildcards, etc.). Bottom line: coverage = object I find easiest and consistent.

Which means, as Michael points aout above, that we have no operation "Get Spatial Object Set by bounding box". Nobody had requested that for the Core, but such a thing is in EOI-WCS where it is called DescribeEOCOverageSet. Given a space/time bbox and a collection (a StitchedMosaic or a DatasetSeries) it delivers those objects contained in the set that intersect with the search box. A flag allows to switch between an "overlap" and a "fully contained" semantics.

If INSPIRE finds that such an operation is needed then I would gladly support a change request to add this to OGC WCS (likely as an extension) in a way we find meaningful (and as coherent with EO-WCS as possible).

 

#4 Updated by James Passmore over 4 years ago

The attached PDF is an attempt to quantify how I understand the relationship of mapping the real world, through the classification of Spatial Objects, to delivery through a download service as features whether that be through Atom, WFS, or WCS.

Also available in: Atom PDF