Something else
Currently we produce an observation cube for each forecast (shape time, station). We need an operator that combines this into a single cube, by looking for unique stations, ensuring there is station data for the full period (all forecasts and lead-times, if not drop that station), and then construct the appropriate forecast_reference_time dimension.
This is already done on an experimental branch, but it needs lifting, tidying, and tests added https://github.com/MetOffice/CSET/pull/2419/changes#diff-05cfcf0a4d084ae6a866040193ee8f83127a832011b334ea5418ddef8801b134.
Something else
Currently we produce an observation cube for each forecast (shape time, station). We need an operator that combines this into a single cube, by looking for unique stations, ensuring there is station data for the full period (all forecasts and lead-times, if not drop that station), and then construct the appropriate forecast_reference_time dimension.
This is already done on an experimental branch, but it needs lifting, tidying, and tests added https://github.com/MetOffice/CSET/pull/2419/changes#diff-05cfcf0a4d084ae6a866040193ee8f83127a832011b334ea5418ddef8801b134.