parse_cf() mercator projection - no central_latitude

[ahuang11]

~/anaconda3/lib/python3.7/site-packages/metpy/plots/mapping.py in make_mercator(attrs_dict, globe)
    146         kwargs.pop('false_northing', None)
    147 
--> 148     return ccrs.Mercator(globe=globe, **kwargs)
    149 
    150 

TypeError: __init__() got an unexpected keyword argument 'central_latitude'

I think similar to

    # CartoPy can't handle central latitude for Geostationary (nor should it)
    # Just remove it if it's 0.
    if not kwargs.get('central_latitude'):
        kwargs.pop('central_latitude', None)

[dopplershift]

Can you point me to a file somewhere with this attribute? I'd like to use it validate putting in a patch to CartoPy.

[ahuang11]

I think I'm still getting this issue.

I don't think the logic is correct:

    if not kwargs.get('central_latitude'):
        kwargs.pop('central_latitude', None)

With this minimal example:

kwargs = {'central_latitude': 90}
if not kwargs.get('central_latitude'):
    kwargs.pop('central_latitude', None)
print(kwargs)

You can see central_latitude still exists.

I think if you remove the if statement, it works and thus can be simplified to:

kwargs = {'central_latitude': 90}
kwargs.pop('central_latitude', None)
print(kwargs)

[dopplershift]

If you drop a central_latitude of 90, any already projected coordinates won't work. Can you point me to a file with this metadata so I can do a more exhaustive test?

[ahuang11]

Unfortunately the data I work with isn't publicly accessible. I'm not sure I follow your statement though. Isn't the code already supposed to drop central latitude for those projections with the pop?

…

On Tue, Dec 3, 2019, 6:08 PM Ryan May ***@***.***> wrote: If you drop a central_latitude of 90, any already projected coordinates won't work. Can you point me to a file with this metadata so I can do a more exhaustive test? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1101?email_source=notifications&email_token=ADU7FFTLMXKM5QSNSNVTUUTQW3YI7A5CNFSM4IE7DPWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEF3IDQY#issuecomment-561414595>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADU7FFWLX224Z7WOBLOBDELQW3YI7ANCNFSM4IE7DPWA> .

[dopplershift]

The code that's in right now, only drops central_latitude if it's "falsey", meaning that bool(kwargs['central_latitude']) == False--so None or 0 essentially. Those can be dropped without impacting the projection math. If central_latitude is 90, that results in entirely different projected value than 0, unless I'm missing something about Mercator?

[ahuang11]

I see.

I think the issue is that cartopy's Mercator projection has no central_latitude kwd: e.g. ccrs.Mercator(central_latitude=90) will error out.

[dopplershift]

@ahuang11 I'm not sure that it makes sense to ever specify a central latitude (or origin latitude) for Mercator. The CF Conventions documentation on Mercator don't list it as a metadata parameter. Also, the PROJ docs on Mercator only show lon_0 as an option, not lat_0

Testing with PyProj, it looks like lat_0 is silently ignored:

from pyproj import Proj

proj1 = Proj({'proj':'merc', 'lon_0':'-100', 'lat_0':'90'})
proj2 = Proj({'proj':'merc', 'lon_0':'-100', 'lat_0':'0'})

proj1(-100, 50) == proj2(-100, 50)

prints True.