Forecast p levels might not always be 10,50 and 90

[peterdudfield]

Current we assume forecast levels are 10,50 and 90. I think 10 and 90 are optional.

Infact, in the future we could be saving different amounts. I wonder if we should actually save

• expected value (p50)
• plevels in a dictionary
  And then we are flexible to save and use what we want

[devsjc]

I was expecting this, in the python notebook there's an example of saving p75 to the metadata object. I did wonder about putting them all on metadata, for consistency, but - doesn't everything we have at the moment save 90 and 10 values?

[peterdudfield]

Good point, I think most do, but I cant gauarantee it.
We pull the NESO forecast, and that does have p10 or p90, so that would be a good example of one.
Also the Ned-nl forecast doesnt have these.

Perhaps making them optional would be good (or put in metadata), i dont mind

[devsjc]

This is now handled with the new other_statistics_fractions fields in the various RPCs involving p levels.