Add domain_query option

[blychs]

Solves Issue #278 .
As suggested by @rschwant , adds optional keyword "domain_query".
To avoid breaking old code, sets the default value for "domain_query" as True.
EDIT: I forgot to mention: I have tested this with airnow_wrfchem.ipynb, adding False domain_query and giorgi and EPA regions.

Cheers
Pablo

[zmoon]

I think the proposed config format (domain_query=False) is kind of confusing. In a sense it's still a query, just based on automatic lat/lon box. To me this feels more like a "domain query" than the original using site metadata already present in the dataframe.

Also, the Giorgi regions actually are rectangles so that's fine. But I would be skeptical about suggesting people use the EPA region rectangles, which are pretty rough estimates.

I have a little project for assigning to precise EPA regions. But it uses GeoPandas which is a heavy dependency.

[rschwant]

@zmoon, should we call it something else then?

domain_metadata=True (default, use metadata from observation file to determine domain)
domain_metadata=False (use build in features in MONETIO to determine metadata since it is not provided?)

Open to other options too?

[zmoon]

should we call it something else then?

I feel it could be done just with domain_type with a special syntax. For example

• domain_type: auto-region:giorgi -- select on the fly using the rectangles (this PR method)
• domain_type: giorgi_region -- current behavior, using column giorgi_region in the dataframe

[rschwant]

I like this method too. This may be easier to explain in the docs too.

[blychs]

OK, I'll try to add that now.

[blychs]

I think the proposed config format (domain_query=False) is kind of confusing. In a sense it's still a query, just based on automatic lat/lon box. To me this feels more like a "domain query" than the original using site metadata already present in the dataframe.

Also, the Giorgi regions actually are rectangles so that's fine. But I would skeptical about suggesting people use the EPA region rectangles, which are pretty rough estimates.

I have a little project for assigning to precise EPA regions. But it uses GeoPandas which is a heavy dependency.

If we end up adding optional regionmask, that would include geopandas as a requirement.

[blychs]

If anyone (probably @zmoon or @rschwant) could review this, we can accept it while creating a new issue to update it to something more robust in the future, as suggested by Zach.

[zmoon]

@blychs do you want to mention these new option in the docs YAML section?

[blychs]

I don't understand the failed tests... I don't believe I changed any of that?

[blychs]

@zmoon One of the tests is failing because it does not find noaa https://www.ncdc.noaa.gov/crn/ . Is it just out or does that link require updates?

[rschwant]

We are linking to data that normally is stable, but is located at the Asheville office, so it's likely that the recent hurricane took down some of their servers. Let's see if it's back up and running tomorrow and go from there. We can re-run the checks then.

[zmoon]

It seems the NCEI/NCDC websites are still having issues: https://www.ncei.noaa.gov/node/6696

[rschwant]

@zmoon do you approve this now, since you have looked at this closer than me? If so, I will merge without the checks passing as it might take NOAA awhile to fix the servers in Asheville.

[zmoon]

@blychs were you able to test that it seems to work?

[blychs]

Wait, something is wrong. Nothing breaks (I had tested it) but it seems to not be doing anything. I'll fix it right away.

[blychs]

OK, it was in effect being skipped. I had tested it with the airnow_wrfchem example, nothing broke, and I was happy with it.
Now it should be fine. I just tested it with that example, running it with and without auto-region:epa and auto-region:giorgi.
I don't have an aircraft example though.
We were not supporting anything like this for satellite before. This might work for it but I haven't tested it.