Handling of Ensemble Model output in CovJson responses

[juerg1s]

Ensemble Predictions Systems (EPS) are common datasets in the Met. CovJSON has no domain to handle them. EPS are handled as normal timeseries with an extra dimension (members) It is possible to define new domains or dimensions by using extensions ( check https://covjson.org/cookbook/ ) Extensions should be grouped in prefixes. If one registers prefixes at https://covjson.org/prefixes/ it would be interoperable.

An attempt for a prefix based domain/axis is made in the repsonses of

https://interpol-dev.met.no/collections/roms-barents-eps/instances

{
"type": "Coverage",
"domain": {
"type": "Domain",
"domainType": "eps:EnsemblePointSeries",
}}

with an axis:

"ranges": {
"current_direction": {
"type": "NdArray",
"dataType": "float",
"axisNames": [
"t",
"eps:member"
],

This is just work at a dev-state. There is a need for handling EPS in the community - the important thing is to agree on something

[havardf]

This should probably be synced with allowing for custom dimensions for ensemble members in collections, so the naming is consistent.

[havardf]

It might make sense to register a prefix common to this profile or rather to EUMETNET, so if we extend CoverageJSON with multiple properties we can reuse that prefix. Like eumetnet:EnsemblePointSeries and eumetnet:foo