NOTE: The galeon mailing list is no longer active. The list archives are made available for historical reasons.
Arliss, Ben, First, my thanks to Ben for kicking off this discussion, which I think highlights some key aspects of interest to the met/ocean domains. My perspectives are inline below.
Arliss, As usual, your questions get right to the heart of the matter. I'll start with my own quick response, with the caveat that other GALEON participants will have perhaps different views (especially on the priorities) and more detailed input as we progress. My responses: 1) Which type(s) of ISO 19123 coverage is this? Gridded Data Structured DiscreteGridPointCoverage Swath DiscreteSurfaceCoverage Point Observation Data Unconnected DiscretePointCoverage Station observation /Timeseries DiscretePointCoverageTrajectoryGeneral DiscretePointCoverage or DiscreteCurveCoverage Profile DiscretePointCoverage Radial DiscreteSurfaceCoverage or DiscreteCurveCoverage
I agree pretty closely with Ben here. See attached picture for the relationships we use in CSML (http://ndg.nerc.ac.uk/csml/). We also make considerable use of the CV_ReferenceableGrid subclass for various domain geometries of the CV_DiscreteGridPointCoverage:
2) What domain dimensions are needed? One of the main points of the aviation-related local forecast use case is to emphasize the need for full 3D plus model-run-time and valid-forecast-time. There is also a need for "time relative to the present" e.g. the latest forecast or observations during the last two hours. Moreover, in the research community, there will be a demand for access to the "raw" data which, in many cases, includes an elevation dimension which is non-spatial, e.g., pressure in the atmosphere or density in the ocean. Perhaps this last topic belongs in the response to the next question relating to axes.
The above diagram (from the CSML User's Manual) illustrates a number of spatio-temporal domain dimensions we require. As Ben mentions, we also use non-length-based coordinates, mainly in the vertical. We believe the new ISO 19111-2 allows such coordinate systems to be integrated with more traditional CRS; the real development needed here is just to catalogue the things (aka EPSG) so the community can reference them. The British Atmospheric Data Centre plans to do this as part of its CF work.
3) What range data structures are needed (called fields and axes in WCS 1.1)? I may have covered some of the needs for special "axes" in item 2) above. But, in general, this is going to require more research on my part. If I understand it correctly, the actual data structures for the fields will be different for different data categories and, in some cases, will be different for different fields within a dataset. This is the area where most of the work is needed in order to come up with "application profiles."
One issue we're grappling with right now is whether you should regard (for instance) the individual time-series measure at a fixed station as part of the same feature/coverage, or whether they should be modelled with separate identities. E.g. we may have diurnal observations of temperature min/max, but wind measurements every 15 minutes. They don't share the same (temporal) domain, but are really both from the same AWS.
4) What are the relative priorities among these types of data? Below are Ben's priorities expressed in terms of the order in which Unidata will be working on them (But please NOTE WELL that others will have other priorities, e.g., the coastal ocean modelling community will be working on irregular grids first whereas hydrology folks will probably be most interested in station obs and radar) 1. Structured Grids (the main focus of GALEON 1 and the CF conventions for netCDF) 2. Station Observations (Currently working on netCDF conventions for Observational datasets) 3. Vertical Profiles (e.g. wind profiler data, balloon or moored buoy CTD soundings) 4. Point Data (e.g. lightning strike data) 5. Radar Radial (e.g., NEXRAD level 2 volume scans) Perhaps some of this would be useful for discussion during the GALEON session at Wednesday's WCS.SWG telecon. -- Ben
galeon archives: