Right now I interpolate the data for plotting and skill using a 30 points limit to avoid huge extrapolations. That is fine for most SECOORA applications due to the high resolution of the data and most of the models.
To make this more generic we can find a max and min valid dates for all the series (excluding low temporal resolution models) and interpolate only within that range.
On second thought. Excluding the low resolution is not OK. I'd rather have a useless comparison, but show that the models exists, than omitting these models. They are very few and only global HYCOM comes to mind.
Right now I interpolate the data for plotting and skill using a 30 points limit to avoid huge extrapolations. That is fine for most SECOORA applications due to the high resolution of the data and most of the models.
To make this more generic we can find a max and min valid dates for all the series (excluding low temporal resolution models) and interpolate only within that range.