Marie-Françoise Voidrot, Frédéric Guillaud, Chris Little, Peter Trevelyan, Matt Wardle, John Schattel , Ben Domenico, Aaron Braeckel, Jeff de La Beaujardière, Josef Matula, Michal Weis
In the OGC Met Ocean DWG Teleconference on 9 November 2009, the main ideas were that:
We would recommend to use the classical TIME for 'validity date/time'
. This is not common meteorological practice, but safest for naïve users. DIMENSIONS are
compulsory for the servers but not for the clients. Using TIME for 'validity date/time' allows the use of clients unable to handle DIMENSIONS.
The 'validity date/time' should not be confused with the 'validity period' which limits when a forecast can be used. The 'validity period' might not
include the 'validity date/time'. E.g. a 12 hour forecast may only be use between 5 and 9 hours, even though there may be 3 hours of validity remaining.
Periods and durations could be speficied according to ISO8601. [There are a number of profiles, including W3C
, of ISO8601, limiting some of the over flexible options.]
Several forecasts may be available and valid for a given TIME. The meteorological community call some of these 'runs' of a particular forecast model, and there may be
several models too. This idea is prevalent in oceanographic forecasting too.
RUN and RANGE would be specified via DIMENSIONS that could be DIM_RUN_START_TIME and DIM_FORECAST_OFFSET.
Some considered that the IBL proposals of DIM_RUN and DIM_FORECAST were ambiguous but still have to check that these latter proposals are the best,
considering WMO vocabulary or other proposals. (Chris will check that).
NWS or NOAA participant (Jeff) said they also have a DIM_RUN_ID to manage several runs for the same RUN_START_TIME.
We will have to clarify the use of this. Perhaps DIM_RUN_ID could be used for EPS for example.
We recommend to use the WMO vocabulary as far as possible
as it is already standardised and translated into the 6 working languages of WMO
(English, French, Spanish, Russian, Arabic and Chinese)
We talked about publishing policies, should we wait for the all run to be finished to publish the WMS(es)?
Some use cases where identified :
- how to manage that the best run is not always the last (NWS suggested via the Default, Fred wondered if this has to be handled via CSW BEST_RUN, LAST_RUN or ALL_RUNS requests)
- how to handle the users that don't want to wait until all run available (Ben talked about the need for local model initialisation but this would rather be for WCS)
Defaults should be used for all DIMENSIONS but precedences should be defined
to handle inconsistent definitions or requests (TIME not.equal.to DIM_RUN_START_TIME+DIM_FORECAST_OFFSET)
We also talked about defining one service by model or for all but didn't analyse all pros and cons.
There are practical implications for the size of GetCapabilities
documents and queries.
- 21 Nov 2009