Is CDB able to handle raster of arbitrary data, for example meteo datasets (wind field, cloud cover etc..) or even simulation results ?
To some extent yes. As you know, there is no meteo dataset in the specification today. However, the OGC CDB specification clearly separates the conceptual model (volume 11 of the OGC spec) from the current enumeration of CDB datasets. This allows for new datasets definitions which would remain aligned with the core model of CDB. For instance, the division of tiles into datasets and LOD is in the core model but not the specific raster encoding formats.
As for the "time varying" type of information you refer in your question, there is no guidance in the specification at the moment to handle this.
Now your question "Is CDB able" might be interpreted from a specification stand point (as answered above) or might be viewed as "Are the currently available CDB toolsets able to handle arbitrary raster data". Let me answer the later now.
The CDB API is able to handle up to 3 channel TIFF encoding. Its support has been validated for current CDB dataset encoding. I suspect that some of the meteo data you refer to might go beyond the current support.
For time varying information or simulation results, where not guided in the specification, some specific implementations exists. Those have been typically challenging when dealing with fast changing rasters due to the CDB requirement to update all LODs for a given change to preserve consistency.
I am not aware of any implementation of weather information as CDB extensions similar to your provided list.
The closest the current CDB specification goes toward time varying simulation would be tide information (elevation sub-dataset). In this case, the CDB dataset stores the high and low tide "static" information sufficient for a tide simulation to use as input, mix with simulated time of day to compute the current tide height at a given location/time.
So, to recap, new raster datasets can be added to CDB as extensions and proposed for standardization as long as they are aligned with the CDB core conceptual model. Time varying information is not standardized in the specification. Still, the concept of "CDB data change" is supported in the CDB API distributed by Presagis today.
Ok thanks Hermann for this detailled answer.
Actually, I was not really thinking about time varying data. Let say I have a georeferenced tiff image containing ground temperature at a specific time or whatever else, is there any way I can store it in a CDB terrain and where shall I put it in the tree folder ?
I'm not talking about tools either.
I know I can store these kind of rasters in a CDB terrain: elevation, bathimetry, imagery, all these with LOD but is it possible to store a raster of data of one dimension like temperature as well ?
In your answer, it seems that it might be possible: "This allows for new datasets definitions which would remain aligned with the core model of CDB." Where can I find more details and example about it in the specifications ?
Yes, you can create a new dataset to store temperature.
You could create a new dataset of raster, give it a number and a name (not conflicting with existing CDB enumerations) and store it in the CDB under each geocel folder as such:
The content can be a single channel TIFF, 32bit float for temperature.
However, the "time" information is not stored. If you need to store different files for different "time" it got collected, you need to define this upfront such that client know which file to load. In imagery, this is done with "quarterly" or "monthly" index. Those are coded into the "S" and "T" as different index, leading to different files for each "time slice".
Ok. Thanks Hermann. That's exactly what I needed to know.