Re: Clarification required: BP5 "describing properties that change over time" [SEC=UNCLASSIFIED]

In yesterday's BP sub-group call (minutes [1]) we talked about this issue.
The question from my previous email [2] remained unanswered ... but not any
more,

@eparsons wanted the "simple" approach, where we just version the
information (at dataset, spatial thing or property level) to be retained
... but agreed that for situations like tracking the GPS location of a
fishing vessel, it made sense to use the "timeseries coverage" approach for
the time-varying property. In the fishing vessel example, this means that
time is the independent variable (the domain) and latitude & longitude are
the dependent variable (the range).

@byron agreed that this was a pretty common pattern.

There are, of course, many ways to serialise this information ... including
the "time-value-pair" (TVP) model used in OGC TimeseriesML (as opposed to
the domain - range - range-set construct that is normally used for
coverages).

@brucebannerman's point is still valid ... sometimes you've got data that
varies in 2 or 3 spatial dimensions that _also_ varies with time - in which
case we're talking about a spatio-temporal coverage; creating a "cube" of
data. It seems complicated, but it's really just an extension of the simple
case.

It's unlikely that this will make it into the BP doc ahead of TPAC (sorry)
... but it's coming.

HTH, Jeremy

[1]: https://www.w3.org/2016/09/07-sdwbp-minutes
[2]: https://lists.w3.org/Archives/Public/public-sdw-wg/2016Aug/0150.html

On Fri, 26 Aug 2016 at 10:29 Linda van den Brink <l.vandenbrink@geonovum.nl>
wrote:

> It happens!
>
>
>
> Meanwhile, we haven’t had real discussion about this yet. To me it sounds
> sensible to add something about (simple) time series. If anyone can provide
> some content for this.
>
>
>
> I created an issue (https://github.com/w3c/sdw/issues/356) and referenced
> it in the BP.
>
>
>
> *Van:* Bruce Bannerman [mailto:B.Bannerman@bom.gov.au]
> *Verzonden:* woensdag 24 augustus 2016 12:12
> *Aan:* Ed Parsons; Jeremy Tandy; SDW WG Public List
> *Onderwerp:* Re: Clarification required: BP5 "describing properties that
> change over time" [SEC=UNCLASSIFIED]
>
>
>
> Sorry. Strike that comment.
>
> It has been a bad day.
>
> B
> ------------------------------
>
> *From:* Bruce Bannerman <B.Bannerman@bom.gov.au>
> *Sent:* Wednesday, 24 August 2016 7:45:34 PM
> *To:* Ed Parsons; Jeremy Tandy; SDW WG Public List
> *Subject:* Re: Clarification required: BP5 "describing properties that
> change over time" [SEC=UNCLASSIFIED]
>
>
>
> OK Ed.
>
> Not much use to me then.
>
> Bruce
> ------------------------------
>
> *From:* Ed Parsons <eparsons@google.com>
> *Sent:* Wednesday, 24 August 2016 7:00:57 PM
> *To:* Jeremy Tandy; Bruce Bannerman; SDW WG Public List
> *Subject:* Re: Clarification required: BP5 "describing properties that
> change over time" [SEC=UNCLASSIFIED]
>
>
>
> I agree simple is good!  remember the intended audience for the BP include
> the majority of web users who have simple spatial data sets to publish..
>
> Ed
>
>
>
> On Wed, 24 Aug 2016, 09:12 Jeremy Tandy, <jeremy.tandy@gmail.com> wrote:
>
> Thanks Bruce. Good points ... I was thinking in terms of the simplest
> situation and there are, of course, these other cases you've reminded me
> of.
>
> The question remains for wider discussion: should we add timeseries
> coverage as a "possible implementation" for BP 5?
>
> Thanks.
>
> On Wed, 24 Aug 2016 at 01:45, Bruce Bannerman <B.Bannerman@bom.gov.au>
> wrote:
>
> Hi Jeremy,
>
>
>
> wrt: '(time-series are one-dimensional coverages)'
>
>
>
> I suspect that I'm missing the intent here, IMO this is but one type
> of time-series.
>
>
>
>
>
> I believe that we can have n-dimensional time-series of traditional
> coverages as well, e.g.:
>
>    - ​a time-series of climate grids representing the distribution of
>    maximum temperature across a continent;
>    -
>    - ​​a time-series of n-dimensional climate models across a continent,
>    or the globe;
>    - a time-series of n-dimensional satellite observations (i.e.
>    multi-spectral images along a swathe over multiple passes over a given
>    region); and
>    - a time-series of derived data across a region, e.g. a time-series of
>    NDVI analyses derived from multi-spectral satellite imagery.
>
>
>
> And of course in many instances we need to relate the observed property of
> the areal coverage time-series back to the type of one-dimensional
> time-series that you have referred to.
>
>
>
> Bruce
>
>
>
>
> ------------------------------
>
> *From:* Jeremy Tandy <jeremy.tandy@gmail.com>
> *Sent:* Wednesday, 24 August 2016 12:50 AM
> *To:* SDW WG Public List
> *Subject:* Clarification required: BP5 "describing properties that change
> over time"
>
>
>
> All-
>
>
>
> BP5 [1] addresses the situation where the properties of a spatial thing
> are updated due to some new measurement or assertion. For example, the
> boundary geometry of flooded area will change as the flood levels rise and
> fall, or the details of a land parcel may be updated due to a new survey.
>
>
>
> In this case, the resource is a spatial thing whose properties are
> captured as a series of discrete snapshots through time. Those snapshots
> may be created on a fixed schedule (e.g. as a result of an annual survey)
> or ad-hoc (e.g. some one notices that things have changed).
>
>
>
> There is a second case where properties change over time: when the
> property is a time dependent variable that is being measured or estimated
> according to some sampling regime. In this case, the resource seems to be a
> "time-series" of property values. For example, the water level of a canal,
> the wind speed measured at routine intervals by an anemometer, or the GPS
> position of a fishing vessel.
>
>
>
> (time-series are one-dimensional coverages)
>
>
>
> So in the example case of the canal and its water-level, the canal (a
> spatial object) has a property "water-level" that refers to a water-level
> time-series coverage. As coverages are themselves features (so says OGC et
> al) then we can also say that our water-level time-series coverage is also
> a spatial thing (albeit a conceptual spatial thing that you can't trip over
> in the real world).
>
>
>
> Is this second case a valid "possible approach to implementation" that
> should be added to BP 5? If so, please can someone in the working group
> suggest some criteria as when to use the "property snapshot" pattern (case
> 1) or the "time-series coverage" pattern (case 2).
>
>
>
> (for me it's subjective, but seems to be related to the intent to
> establish a coherent sampling regime for a property value as opposed to an
> ad-hoc update)
>
>
>
> Jeremy
>
>
>
> BTW, I've deliberately skipped over the "value assignment" concern (e.g.
> how values are assigned to properties) as described in ISO 19109 §7.4.10
> ValueAssignment - for example, how OM_Observation can be used to relate the
> "observed value" to a "feature of interest" with all the metadata about the
> observation or measurement process. I don't think this is relevant for this
> BP.
>
>
>
>
>
> [1]: http://w3c.github.io/sdw/bp/#desc-changing-properties
>
> --
>
> *Ed Parsons *FRGS
> Geospatial Technologist, Google
>
> Google Voice +44 (0)20 7881 4501
> www.edparsons.com @edparsons
>