So far, the data in the global forecasting mode are stored in 4D-arrays with dimensions ["ml", "datetime", "lat", "lon"].
However, if the lead time gets larger than the temproal gap between different forecast initialisations time, data gets overwriiten.
To avoid this, the data will be stored in 5D-arrays in the future, where the datetime-dimension is separated into init_time and lead_time with the latter being provided in hours after initialisation, i.e ["init_time", "ml", "lead_time", "lat", "lon"].
So far, the data in the global forecasting mode are stored in 4D-arrays with dimensions
["ml", "datetime", "lat", "lon"]. However, if the lead time gets larger than the temproal gap between different forecast initialisations time, data gets overwriiten. To avoid this, the data will be stored in 5D-arrays in the future, where thedatetime-dimension is separated intoinit_timeandlead_timewith the latter being provided in hours after initialisation, i.e["init_time", "ml", "lead_time", "lat", "lon"].