Support "hdate" metadata field instead of "date" for reforecast datasets #104
Description
What happened?
Some datasets, in particular reforecasts and climatology-style products, use the MARS metadata key hdate to indicate the start time for which a forecast is issued. In these datasets, the date key instead represents the model version or the time at which the reforecast was run, which is a different concept.
At the moment, covjsonkit always uses the date key when encoding the time axis, and ignores hdate. When requests are coming via polytope, this leads to incorrect time axes for datasets where hdate is the intended forecast reference time.
This makes it currently impossible to use covjsonkit (via polytope-mars) correctly for climatology and reforecast datasets that rely on hdate.
When hdate is present in the MARS request or response metadata, covjsonkit should be able to use it as the forecast reference time so that hdate, not date, controls the time axis in the resulting CovJSON.
Ideally, this would work transparently for different polytope features, such as timeseries and polygons.
request = {
"class": "ce",
"stream": "efcl",
"type": "sfo",
"model": "lisflood",
"date": "20230101", # model version
"hdate": "20250911", # this should control the time dimension in the covjson output
"time": "1200",
"levtype": "sfc",
"expver": "8888",
"domain": "g",
"param": "240023",
"step": "6",
"origin" : "ecmf",
"feature": {
# Should also work for other requests like polygons etc.
"type": "timeseries",
"points": [[50.73746373267438, 7.107723168102066]],
"axes": ["latitude", "longitude"],
"time_axis": "step",
},
}Related issue in earthkit-data: ecmwf/earthkit-data#714
What are the steps to reproduce the bug?
Version
develop
Platform (OS and architecture)
macos,atos
Relevant log output
Accompanying data
No response
Organisation
No response