Re: UCR ISSUE-75: Coverage requirement for quality metadata in scope?

Issue-75 was closed at the Lisbon F2F. I have just rephrased the Quality
metadata requirement and renamed it to Quality per sample
<http://w3c.github.io/sdw/UseCases/SDWUseCasesAndRequirements.html#QualityPerSample>
.

Regards,
Frans

On 14 September 2016 at 15:33, Jon Blower <j.d.blower@reading.ac.uk> wrote:

> Hi Francs,
>
>
>
> Yes, good thoughts. I’m happy with the phrasing, “It should be possible to
> describe properties of data quality (e.g. uncertainty) per data sample”.
>
>
>
> Ø  By the way, are there other quality properties next to uncertainty
> that should be possible to record per sample in a coverage?
>
>
>
> Yes – quality flags are another common use case.
>
>
>
> Ø  If it is only uncertainty that matters, couldn't that be adequately
> addressed by using significant digits?
>
>
>
> No – that’s precision, which is a different thing. Uncertainty can be
> expressed in a load of different ways. The ideal situation (for some) is to
> have a probability density function for each sample value – but most people
> collapse this to a “x plus or minus y” kind of simplified representation.
> Summary statistics are a middle ground.
>
>
>
> (It’s worth noting that if you don’t convey a view of uncertainty then you
> are actually transmitting no information at all… We get around this by
> either assuming that values are exact or by applying some implicit
> knowledge about uncertainty. But that’s probably a diversion into
> beard-stroking.)
>
>
>
> Cheers,
>
> Jon
>
>
>
> *From: *Frans Knibbe <frans.knibbe@geodan.nl>
> *Date: *Wednesday, 14 September 2016 13:42
> *To: *Jon Blower <sgs02jdb@reading.ac.uk>
> *Cc: *SDW WG Public List <public-sdw-wg@w3.org>
> *Subject: *Re: UCR ISSUE-75: Coverage requirement for quality metadata in
> scope?
>
>
>
> Hi Jon,
>
>
>
> Thank you for your thoughts. It seems to me that your first point, quality
> information for the entire dataset, does not really justify the explicit
> requirement. It is something that should already be covered by general best
> practices for data on the web, right?
>
>
>
> But data quality data per sample value seems to be something that needs
> specific guidance. By the way, is it OK to use 'sample' instead of 'pixel'?
> Not all coverage data are raster data...
>
>
>
> So how about rephrasing the requirement to "It should be possible to
> describe properties of data quality (e.g. uncertainty) per data sample"?
>
>
>
> By the way, are there other quality properties next to uncertainty that
> should be possible to record per sample in a coverage? If it is only
> uncertainty that matters, couldn't that be adequately addressed by using
> significant digits?
>
>
>
> Regards,
>
> Frans
>
>
>
>
>
>
>
> On 13 September 2016 at 20:31, Jon Blower <j.d.blower@reading.ac.uk>
> wrote:
>
> Hi Frans,
>
>
>
> I agree that data quality is a general issue. However, there may be some
> specific things to say about coverages. For example:
>
>
>
> 1.       “General” quality information (e.g. sensor accuracy) could be
> recorded as metadata with the coverage.
>
> 2.       However, if we want **per-pixel** quality information in the
> coverage, this implies that we need a set of range values for the quality
> information (e.g. one set of range values for temperature, and one for the
> uncertainty on the temperature, howsoever measured). It also means that
> there ought to be a way to link these two sets of range values together, to
> advertise to clients that they should consider those two things together.
>
>
>
> There are no widely-used standards for per-pixel quality information as
> far as I know, but we have proposed such a mechanism in CoverageJSON [1].
> In a previous project (GeoViQua) we developed a way to do this in WMS too
> [2]. Both are incomplete and could be further developed.
>
>
>
> (UncertML provides a vocabulary for probability density functions and
> statistics, which can be reused here.)
>
>
>
> Best wishes,
> Jon
>
>
>
> [1] https://covjson.org/spec/#parametergroup-objects
>
> [2] http://dx.doi.org/10.3390/ijgi4041965
>
>
>
> *From: *Frans Knibbe <frans.knibbe@geodan.nl>
> *Date: *Friday, 9 September 2016 14:57
> *To: *SDW WG Public List <public-sdw-wg@w3.org>
> *Subject: *UCR ISSUE-75: Coverage requirement for quality metadata in
> scope?
> *Resent-From: *<public-sdw-wg@w3.org>
> *Resent-Date: *Friday, 9 September 2016 14:57
>
>
>
> Hello everyone,
>
>
>
> I just realised ISSUE-75 <https://www.w3.org/2015/spatial/track/issues/75>
> did not have its own e-mail thread yet. Of course that won't do. So this is
> the official thread for ISSUE-75.
>
>
>
> The issue is about the quality metadata requirement
> <http://w3c.github.io/sdw/UseCases/SDWUseCasesAndRequirements.html#QualityMetadata>.
> It is a requirement for the Coverage deliverable. It currently reads "It
> should be possible to describe properties of the data quality, e.g.
> uncertainty."
>
>
>
> The need to describe data quality is a general data issue, so it does not
> seem in scope for the Coverage deliverable and perhaps it does not need to
> be mentioned in the UCR document.
>
>
>
> Should we decide to keep the requirement, perhaps it could just as well be
> related to the other deliverables.
>
>
>
> Is there something about coverage data that justify making this
> requirement explicit for the coverage deliverable?
>
>
>
>
>
> Greetings,
>
> Frans
>
>
>
>
>
>
>