These minutes have been copied unedited from the Huddle site
The presentation summarizing the output from the conceptual modelling working group @ the 2nd Workshop on the use of OGC/GIS standards in Meteorology can be found here ... it was presented at the Met Ocean DWG Session of the OGC TC, Mountain View (Google) last week.
In summary ...
- we looked at the standards landscape & found significant consensus around the Observations & Measurements model (now ISO19156)
- we identified that WMO, INSPIRE & NextGen/SESAR were committed to exploit the output of this group
- we developed a timeline for meetings over the next 12-months
- we decided to use regular tele-conferences as our check-point meetings
- we got some volunteers to help! (thanks!)
Aaron/Andrew mentioned having 'core' set of feature types that can be extended by specific domains.
In the following presentation, see slide entitled "Composable Weather Model" for WXXM view of such a stack.
Apologies for not updating things for a week; I ran out of time before finishing work for the holiday season & I've only just found a few spare minutes in between getting things prepped for Christmas ...
For the record. I thought we'd capture the agenda ... as we don't seem to be able to refer back to previous meetings
Agenda: 1) Quick recap from working group in Toulouse (Jeremy Tandy to lead) ... why are we here and what have we committed our time to? 2) Brief discussion on INSPIRE conceptual modelling process (Andrew Woolf to lead?) ... we need to be INSPIRE compliant to meet the goals of the group; so let's make sure we take on board any critical constraints 3) Discussion of the use cases that we will use to drive the development of the conceptual model 4) From the use-cases we can derive the scope of the modelling activity
Agenda item 2 (Andrew Woolf)
INSPIRE should not provide too many constraints; the benefit is the provision of a robust framework of guidance & tooling totaling around 100 person-years of effort so far
- ISO TC/211 framework for interoperability, which is formalized in the conceptual model for each thematic domain; the conceptual model is the fundamental point of reference for interoperability
- a methodology (document D2.6) ... which says (i) start with use-cases that describe the need for interoperability, (ii) then look for existing standards, (iii) then identify candidate conceptual models
- a generic conceptual model ... this has a 'layer-cake' approach to describing the world (i) ISO TC/211 standards (ii) utility types (including a 'spatial dataset' type) (iii) useful stereotypes for creating your own 'features'
The standard practice for developing conceptual models is:
- use of Enterprise Architect as the modelling tool ... this includes integrated functionality for geospatial modeling
- use of a subversion as a version-controlled repository for group collaboration
After a brief discussion, the attendees agreed to use Enterprise Architect and subversion as the tools for this activity
Andrew Woolf offered to setup a subversion repository for the conceptual modeling activity and populate with ISO TC/211 models (which now includes Observations & Measurements) and the INSPIRE generic conceptual model
Andrew Woolf also offered to provide an overview of Observations & Measurements in a follow-on telecon ... but for those who can't wait, the most useful resource I have found for O&M is the Solid Earth & Environment GRID (SeeGRID
) twiki maintained by CSIRO ... thanks go to Simon Cox & team for establishing this fantastic information resource!
Chris Little pointed out that OGC Europe are currently distributing a book about Observations & Measurements. If you want to get hold of a copy, please contact Chris directly ...
Chris Little also wanted to give an update regarding WMS issues
He described the ongoing debate at OGC between the 'old way' (2-d map ... cartography inspired) and 'new way' (4-d data cube, that may be 'sliced' to provide a 2-d map)
The issues, which are especially pertinent to meteorology, are yet to be fully resolved ... but Chris indicated that OGC were committed to achieving consistency & commonality to queries across ALL WxS
standards. Most queries need to relate to the structure of the conceptual model NOT the service type through which the query is made. This may take a year-or-2 to resolve; OWS would be the forum in OGC to do this.
Chris may choose to add to the content of this posting if necessary to clarify!
Aaron Braeckel noted: you need a flexible mechanism to get metadata, filtering what is required (either from a catalogue or a capabilities document)
Andrew Woolf responded that he is seeing people use OASIS ebRIM to achieve this ... there is an OGC CSW profile for ebRIM
All agreed that we must consider the queries applied to our data model (and the data model of the query responses) as much as the underpinning conceptual model.
Andrew Woolf recommended reviewing the latest WFS specification (ISO endorsed) which enables one to provide stored queries packaged with a service implementation, where the query model of the stored queries is independent of the feature being queried (the underlying data model).
Andrew Woolf demonstrated a custom WFS @ Valencia providing stored queries for CSML. Aaron Braeckel & team are currently developing an implementation.
We must be able to control the queries as strongly as the underlying model ... in ISO TC/211 parlance, we need to consider the operations to the model
Agenda item 3 use-cases to support interoperability ... group discussion
Objective is to exchange data easily between communities including: operational meteorology, aviation, INSPIRE, climate & atmospheric science
How can we 'characterize' the data?
- forecast vs. analysis ... the traditional mechanism used by operation meteorologists ... discard: we often seemlessly blur observations into forecasts within a single dataset
- observations vs. imagery ... discard: we routinely simulate observations from numerical models & create 'forecast' imagery
- fundamental topology is critical to classification
Andrew Woolf noted that CSML & Sampling features (from O&M) have adopted this approach. Andrew will publish a list of data communities that have also adopted this approach. Both data managers & practicing scientists routinely use the language of topology to characterize the information they work with.
- continuous vs. discrete ... coverages
Currently this is an active topic of discussion within OGC WCS WG. In the majority of usage, real world continuous phenomena are normally 'discretized' at sampling points (or calculation points within a numerical model). This implies that the data is 'interpolatable' ... you can estimate values in between the discrete points. Evolution of the WCS model (& the underlying ISO TC/211 coverage model) is likely to evolve to reflect this. We will see the boundaries between discrete and continuous coverages blur.
Chris Little noted that we will want to advertise the interpolation mechanisms provided by the service (these may be highly specific to meteorology domain) rather than be restricted to generic interpolation mechanisms.
Within aviation domain, it was noted that WMO CAeM
is supporting co-operative development of new aviation codes using XML, GML & WXXM.
Jeremy Tandy & Pierre Kerherve noted that this activity was covered under that mandate of WMO IPET-MDI ... Bart Nicolai (BelgoControl
) is the liason from CAeM
Aaron Braeckel described the WXXM 1.1 model developed with FAA (?) ... this is a work in progress so consider it DRAFT until notified otherwise. It is, however, almost complete. Aaron posted a copy of the UML & schemas for the model here. Aaron advised that the team are still working on documentation and some of the details of the schema generation.
Essentially, WXXM has a 'soft-typed' weather core (loosely based on CSML) plus a strongly typed aviation element. This is similar to the idea proposed by Andrew Woolf at the Toulouse meeting. He added that we should leave the OPMET and aviation extensions for those respective communities and focus on the 'sweet spot' for a common core that can be re-used to underpin the extensions for those specialized activities. Essentially, you end up with layers of 'specificity' for models. Oliver Newell has a great doc on this ... Dominic Lowe posted a link earlier in this discussion thread.
Chris Little proposed that we restrict our scope to assume that our frame of reference is fixed relative to the earth's surface. He actually used complicated mathematical terms I didn't really understand ... but we got there in the end!
Jeremy Tandy questioned how we would deal with ensembles (multiple models covering the same domain each with 'perturbed' boundary conditions allowing statistical analysis of potential outcomes). Should we use additional ranges within the coverage (like is done for hyper-spectral imagery).
Chris Little noted that although meteorologists often talk about 5th, 6th & nth dimensional datasets (like additional parameters and models) these are undeniably discrete ... there is nothing in between model run 8 and 9 or parameters temperature and wind-speed ... there is also no definitive ordering of the 'axes' ... it doesn't matter which order the ensembles are presented in, you just need to know which is which.
Some of the topologies suggested were:
- trajectory ... aviation flight routes, pollution models
- discrete surface coverage ... areas of designated clear-air turbulence (or other significant weather attribute ... often used for operational decision making, almost always restricted to the '2-d map' metephor)
- discrete curve coverages ... jet-streams with attribute of height
- 2-d, 3-d & 4-d grids
- staggered grids ... i.e. where u & v components of wind speed are on different grids
will be on Tuesday 12th Jan @ 15.00 UTC
Marie-Francoise Voidrot will setup the telcon with OGC.
Many thanks for your participation & have a great christmas!!!
- 09 Feb 2010