Repository navigation
SkyCoord.separation not symmetric #5185
Description
When calculating separations between SkyCoords in AltAz and ICRS frames, it matters which way round you calculate the separation! For example:
In [1]: from astropy import coordinates as coord, units as u
In [2]: from astropy.time import Time
In [3]: time = Time('2003-04-05 06:07:08')
In [4]: loc = coord.EarthLocation(*(-1463969.3018517173,
-5166673.342234327,
3434985.7120456537)*u.m)
In [5]: altaz_frame = coord.AltAz(location=loc, obstime=time)
In [6]: moon = coord.get_moon(time, location=loc).transform_to(altaz_frame)
In [7]: one_deg_away = coord.SkyCoord(az=moon.az, alt=moon.alt+1*u.deg, frame=moon.frame)
In [8]: icrs_coord = one_deg_away.icrs
In [9]: icrs_coord.separation(moon)
Out[9]: <Angle 0.9999999999666758 deg>
In [10]: moon.separation(icrs_coord)
Out[10]: <Angle 141.88054031044086 deg>
Activity
adrn commentedon Jul 18, 2016
Hm. I suspect this is because information about the (GCRS) moon frame is lost when you get an ICRS frame for the moon, e.g.,
one_deg_away.icrsIf you instead do (with the
SkyCoord):StuartLittlefair commentedon Jul 18, 2016
Indeed, and the case above is rather artificial since the conversion to ICRS is not necessary.
However, I do have a use case where the conversion to ICRS first is quite necessary. It's not that information is lost: the
obstimeis the crucial piece of information and that is stored in the ICRSSkyCoord. It's just that it's being ignored; I suspect that there's an assumption ofobstime=Time('J2000')or similar for all ICRSSkyCoords.But essentially I think the snippet above should work if all was operating as expected.
StuartLittlefair commentedon Jul 28, 2016
I'm closing this issue. There are two things going on here. The first is that I ignored the distance and should have written
The second is that
transforms the
icrs_coordto an altaz (Earth-centered) frame and computes the separation, whilsttransforms the moon to the
ICRSframe and computes the separation.In other words, both calculate the angular separation between two coordinates, but from different points of view. Whilst this is a bit counter-intuitive the code does the correct thing I think.