Overplotting markers and artists¶
For the example in the following page we start from the example introduced in Initializing axes with world coordinates.
Apart from the handling of the ticks, tick labels, and grid lines, the
WCSAxes class behaves like a normal Matplotlib
Axes instance, and methods such as
scatter(), and so on will work and plot the
data in pixel coordinates by default.
In the following example, the scatter markers and the rectangle will be plotted in pixel coordinates:
# The following line makes it so that the zoom level no longer changes, # otherwise Matplotlib has a tendency to zoom out when adding overlays. ax.set_autoscale_on(False) # Add a rectangle with bottom left corner at pixel position (30, 50) with a # width and height of 60 and 50 pixels respectively. from matplotlib.patches import Rectangle r = Rectangle((30., 50.), 60., 50., edgecolor='yellow', facecolor='none') ax.add_patch(r) # Add three markers at (40, 30), (100, 130), and (130, 60). The facecolor is # a transparent white (0.5 is the alpha value). ax.scatter([40, 100, 130], [30, 130, 60], s=100, edgecolor='white', facecolor=(1, 1, 1, 0.5))
All such Matplotlib commands allow a
transform= argument to be passed,
which will transform the input from world to pixel coordinates before it is
passed to Matplotlib and plotted. For instance:
will take the values passed to
scatter() and will
transform them using the transformation passed to
transform=, in order to
end up with the final pixel coordinates.
WCSAxes class includes a
method that can be used to get the appropriate transformation object to convert
from various world coordinate systems to the final pixel coordinate system
required by Matplotlib. The
get_transform() method can
take a number of different inputs, which are described in this and subsequent
sections. The two simplest inputs to this method are
For example, if your WCS defines an image where the coordinate system consists of an angle in degrees and a wavelength in nanometers, you can do:
ax.scatter(, [3.2], transform=ax.get_transform('world'))
to plot a marker at (34deg, 3.2nm).
ax.get_transform('pixel') is equivalent to not using any
transformation at all (and things then behave as described in the Pixel
For the special case where the WCS represents celestial coordinates, a number
of other inputs can be passed to
'fk4': B1950 FK4 equatorial coordinates
'fk5': J2000 FK5 equatorial coordinates
'icrs': ICRS equatorial coordinates
'galactic': Galactic coordinates
In addition, any valid
astropy.coordinates coordinate frame can be passed.
For example, you can add markers with positions defined in the FK5 system using:
ax.scatter(266.78238, -28.769255, transform=ax.get_transform('fk5'), s=300, edgecolor='white', facecolor='none')
In the case of
plot(), the positions of the center of the markers is transformed, but the markers themselves are drawn in the frame of reference of the image, which means that they will not look distorted.
Transformations can also be passed to Astropy or Matplotlib patches. For example, we can
get_transform() method above to plot a quadrangle
in FK5 equatorial coordinates:
from astropy import units as u from astropy.visualization.wcsaxes import Quadrangle r = Quadrangle((266.0, -28.9)*u.deg, 0.3*u.deg, 0.15*u.deg, edgecolor='green', facecolor='none', transform=ax.get_transform('fk5')) ax.add_patch(r)
However, it is very important to note that while the height will indeed be 0.15 degrees, the width will not strictly represent 0.3 degrees on the sky, but an interval of 0.3 degrees in longitude (which, depending on the latitude, will represent a different angle on the sky).
In other words, if the width and height are set to the same value, the resulting polygon will not be a square.
The same applies to the
Circle patch, which will not actually produce a circle:
from matplotlib.patches import Circle r = Quadrangle((266.4, -28.9)*u.deg, 0.3*u.deg, 0.3*u.deg, edgecolor='cyan', facecolor='none', transform=ax.get_transform('fk5')) ax.add_patch(r) c = Circle((266.4, -29.1), 0.15, edgecolor='yellow', facecolor='none', transform=ax.get_transform('fk5')) ax.add_patch(c)
If what you are interested is simply plotting circles around
sources to highlight them, then we recommend using
scatter(), since for the circular
marker (the default), the circles will be guaranteed to be
circles in the plot, and only the position of the center is
To plot ‘true’ spherical circles, see the Spherical patches section.
Quadrangle is the recommended patch for plotting a quadrangle, as opposed to Matplotlib’s
The edges of a quadrangle lie on two lines of constant longitude and two lines of constant latitude (or the equivalent component names in the coordinate frame of interest, such as right ascension and declination).
The edges of
Quadrangle will render as curved lines if appropriate for the WCS transformation.
Rectangle will always have straight edges.
Here’s a comparison of the two types of patches for plotting a quadrangle in
ICRS coordinates on
from matplotlib.patches import Rectangle # Set the Galactic axes such that the plot includes the ICRS south pole ax = plt.subplot(projection=wcs) ax.set_xlim(0, 10000) ax.set_ylim(-10000, 0) # Overlay the ICRS coordinate grid overlay = ax.get_coords_overlay('icrs') overlay.grid(color='black', ls='dotted') # Add a quadrangle patch (100 degrees by 20 degrees) q = Quadrangle((255, -70)*u.deg, 100*u.deg, 20*u.deg, label='Quadrangle', edgecolor='blue', facecolor='none', transform=ax.get_transform('icrs')) ax.add_patch(q) # Add a rectangle patch (100 degrees by 20 degrees) r = Rectangle((255, -70), 100, 20, label='Rectangle', edgecolor='red', facecolor='none', linestyle='--', transform=ax.get_transform('icrs')) ax.add_patch(r) plt.legend(loc='upper right')
filename = get_pkg_data_filename('galactic_center/gc_bolocam_gps.fits') hdu = fits.open(filename) ax.contour(hdu.data, transform=ax.get_transform(WCS(hdu.header)), levels=[1,2,3,4,5,6], colors='white')
In the case where you are making a plot of a celestial image, and want to plot a circle that represents the area within a certain angle of a longitude/latitude,
Circle patch is not appropriate, since it will result in a distorted shape (because longitude is not the same as the angle on the sky).
For this use case, you can instead use
SphericalCircle, which takes a tuple of
Quantity or a
SkyCoord object as the input,
Quantity as the radius:
from astropy import units as u from astropy.coordinates import SkyCoord from astropy.visualization.wcsaxes import SphericalCircle r = SphericalCircle((266.4 * u.deg, -29.1 * u.deg), 0.15 * u.degree, edgecolor='yellow', facecolor='none', transform=ax.get_transform('fk5')) ax.add_patch(r) #The following lines show the usage of a SkyCoord object as the input. skycoord_object = SkyCoord(266.4 * u.deg, -28.7 * u.deg) s = SphericalCircle(skycoord_object, 0.15 * u.degree, edgecolor='white', facecolor='none', transform=ax.get_transform('fk5')) ax.add_patch(s)