What to do when different models have different numbers of non-null pairs with the obs

[zmoon]

@btang1 encountered this in his #238 situation.

This is not something we usually expect to happen with models, since model fields are usually non-null at all time/space and we usually configure things such that multiple models being compared span the same period. But it's not impossible, for example:

• model's time starts later than the others, so when pairing is done, the model value is set to null for those times before
• search radius setting is too small and prevents pairing to model value, causing model value to be set to null in the pair
• the model data does have null at certain times, perhaps from a post-processing to mask unrealistic values, or something done in a MONETIO reader, in a way that's different from the other models

Ideas:

• check pairs, print warning message if different non-null counts among different obs-model pairs
• drop where obs + all model values are not all non-null
• annotate counts in plots

We are planning to start with the warning message. And then add the option to drop.

[zmoon]

I can start to work on drop obs( option 2).

Can you start with the warning message? Even when dropping we probably want to print a message with the info of what the differences are / what is being dropped.

[btang1]

@zmoon I have a branch to add a warning, and I test it works. code start from: https://github.com/btang1/MELODIES-MONET/blob/PairOBS/melodies_monet/driver.py#L1052 code end at: https://github.com/btang1/MELODIES-MONET/blob/PairOBS/melodies_monet/driver.py#L1062 also changed code at : https://github.com/btang1/MELODIES-MONET/blob/PairOBS/melodies_monet/driver.py#L925