pfsspy is a python package for carrying out Potential Field Source Surface modelling, a commonly used magnetic field model of the Sun and other stars.

Note

If you find any bugs or have any suggestions for improvement, please raise an issue here: https://github.com/dstansby/pfsspy/issues

Citing

If you use pfsspy in work that results in publication, please cite the Journal of Open Source Software paper at https://doi.org/10.21105/joss.02732. A ready made bibtex entry is

@article{Stansby2020,
  doi = {10.21105/joss.02732},
  url = {https://doi.org/10.21105/joss.02732},
  year = {2020},
  publisher = {The Open Journal},
  volume = {5},
  number = {54},
  pages = {2732},
  author = {David Stansby and Anthony Yeates and Samuel T. Badman},
  title = {pfsspy: A Python package for potential field source surface modelling},
  journal = {Journal of Open Source Software}
}

Contents

Installing

pfsspy requires

• python >= 3.7

• sunpy >= 3.0

and can be installed from PyPi using

pip install pfsspy

This will install pfsspy and all of its dependencies. In addition to the core dependencies, there are two optional dependencies (numba, streamtracer) that improve code performance. These can be installed with

pip install pfsspy[performance]

Examples

Using pfsspy

Magnetic field along a field line

Magnetic field along a field line

How to get the value of the magnetic field along a field line traced through the PFSS solution.

First, import required modules

import astropy.constants as const
import astropy.units as u
import matplotlib.pyplot as plt
import sunpy.map
from astropy.coordinates import SkyCoord

import pfsspy
from pfsspy import tracing
from pfsspy.sample_data import get_gong_map

Load a GONG magnetic field map

gong_fname = get_gong_map()
gong_map = sunpy.map.Map(gong_fname)

Out:

Files Downloaded:   0%|          | 0/1 [00:00<?, ?file/s]

pfsspy.mrzqs200901t1304c2234_022.fits.gz:   0%|          | 0.00/242k [00:00<?, ?B/s]

pfsspy.mrzqs200901t1304c2234_022.fits.gz:   0%|          | 100/242k [00:00<06:57, 578B/s]


Files Downloaded: 100%|##########| 1/1 [00:00<00:00,  2.24file/s]
Files Downloaded: 100%|##########| 1/1 [00:00<00:00,  2.24file/s]

The PFSS solution is calculated on a regular 3D grid in (phi, s, rho), where rho = ln(r), and r is the standard spherical radial coordinate. We need to define the number of rho grid points, and the source surface radius.

nrho = 35
rss = 2.5

From the boundary condition, number of radial grid points, and source surface, we now construct an Input object that stores this information

pfss_in = pfsspy.Input(gong_map, nrho, rss)
pfss_out = pfsspy.pfss(pfss_in)

Now take a seed point, and trace a magnetic field line through the PFSS solution from this point

tracer = tracing.FortranTracer()
r = 1.2 * const.R_sun
lat = 70 * u.deg
lon = 0 * u.deg

seeds = SkyCoord(lon, lat, r, frame=pfss_out.coordinate_frame)
field_lines = tracer.trace(seeds, pfss_out)

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Set MJD-OBS to 59093.544444 from DATE-OBS'.
  FITSFixedWarning)

From this field line we can extract the coordinates, and then use .b_along_fline to get the components of the magnetic field along the field line.

From the plot we can see that the non-radial component of the mangetic field goes to zero at the source surface, as expected.

field_line = field_lines[0]
B = field_line.b_along_fline
r = field_line.coords.radius
fig, ax = plt.subplots()

ax.plot(r.to(const.R_sun), B[:, 0], label=r'$B_{r}$')
ax.plot(r.to(const.R_sun), B[:, 1], label=r'$B_{\theta}$')
ax.plot(r.to(const.R_sun), B[:, 2], label=r'$B_{\phi}$')
ax.legend()
ax.set_xlabel(r'r / $R_{\odot}$')
ax.set_ylabel(f'B ({B.unit})')

plt.show()

HMI PFSS solutions

HMI PFSS solutions

Calculating a PFSS solution from a HMI synoptic map.

This example shows how to calcualte a PFSS solution from a HMI synoptic map. There are a couple of important things that this example shows:

• HMI maps have non-standard metadata, so this needs to be fixed

• HMI synoptic maps are very big (1440 x 3600), so need to be downsampled in order to calculate the PFSS solution in a reasonable time.

First import the required modules

import os

import astropy.units as u
import matplotlib.pyplot as plt
import sunpy.map
from sunpy.net import Fido
from sunpy.net import attrs as a

import pfsspy
import pfsspy.utils

Set up the search.

Note that for SunPy versions earlier than 2.0, a time attribute is needed to do the search, even if (in this case) it isn’t used, as the synoptic maps are labelled by Carrington rotation number instead of time

time = a.Time('2010/01/01', '2010/01/01')
series = a.jsoc.Series('hmi.synoptic_mr_polfil_720s')
crot = a.jsoc.PrimeKey('CAR_ROT', 2210)

Do the search.

If you use this code, please replace this email address with your own one, registered here: http://jsoc.stanford.edu/ajax/register_email.html

result = Fido.search(time, series, crot,
                     a.jsoc.Notify(os.environ["JSOC_EMAIL"]))
files = Fido.fetch(result)

Out:

Export request pending. [id=JSOC_20211018_1007_X_IN, status=2]
Waiting for 0 seconds...
1 URLs found for download. Full request totalling 6MB

Files Downloaded:   0%|          | 0/1 [00:00<?, ?file/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:   0%|          | 0.00/6.17M [00:00<?, ?B/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:   0%|          | 100/6.17M [00:00<2:02:54, 836B/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:   0%|          | 21.2k/6.17M [00:00<00:58, 105kB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:   1%|1         | 74.7k/6.17M [00:00<00:22, 265kB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:   4%|3         | 220k/6.17M [00:00<00:09, 644kB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:   8%|8         | 515k/6.17M [00:00<00:04, 1.32MB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:  17%|#6        | 1.04M/6.17M [00:00<00:02, 2.37MB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:  28%|##8       | 1.75M/6.17M [00:00<00:01, 3.76MB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:  45%|####4     | 2.75M/6.17M [00:00<00:00, 5.59MB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:  59%|#####8    | 3.61M/6.17M [00:01<00:00, 6.47MB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:  75%|#######5  | 4.63M/6.17M [00:01<00:00, 7.41MB/s]

hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits:  92%|#########2| 5.68M/6.17M [00:01<00:00, 8.09MB/s]


Files Downloaded: 100%|##########| 1/1 [00:01<00:00,  1.40s/file]
Files Downloaded: 100%|##########| 1/1 [00:01<00:00,  1.40s/file]

Read in a file. This will read in the first file downloaded to a sunpy Map object

hmi_map = sunpy.map.Map(files[0])
print('Data shape: ', hmi_map.data.shape)

Out:

Data shape:  (1440, 3600)

Since this map is far to big to calculate a PFSS solution quickly, lets resample it down to a smaller size.

hmi_map = hmi_map.resample([360, 180] * u.pix)
print('New shape: ', hmi_map.data.shape)

Out:

New shape:  (180, 360)

Now calculate the PFSS solution

nrho = 35
rss = 2.5
pfss_in = pfsspy.Input(hmi_map, nrho, rss)
pfss_out = pfsspy.pfss(pfss_in)

Using the Output object we can plot the source surface field, and the polarity inversion line.

ss_br = pfss_out.source_surface_br
# Create the figure and axes
fig = plt.figure()
ax = plt.subplot(projection=ss_br)

# Plot the source surface map
ss_br.plot()
# Plot the polarity inversion line
ax.plot_coord(pfss_out.source_surface_pils[0])
# Plot formatting
plt.colorbar()
ax.set_title('Source surface magnetic field')

plt.show()

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/pfsspy/output.py:95: UserWarning: Could not parse unit string "Mx/cm^2" as a valid FITS unit.
See https://fits.gsfc.nasa.gov/fits_standard.html for the FITS unit standards.
  warnings.warn(f'Could not parse unit string "{unit_str}" as a valid FITS unit.\n'
/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Invalid DATE-OBS format '2018.11.09_12:30:52_TAI''.
  FITSFixedWarning)
/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'unitfix' made the change 'Changed units:
  'degree' -> 'deg'.
  FITSFixedWarning)
/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/sunpy/util/decorators.py:378: SunpyMetadataWarning: Missing metadata for observer: assuming Earth-based observer.

  new_val = prop(instance)
/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/pfsspy/output.py:95: UserWarning: Could not parse unit string "Mx/cm^2" as a valid FITS unit.
See https://fits.gsfc.nasa.gov/fits_standard.html for the FITS unit standards.
  warnings.warn(f'Could not parse unit string "{unit_str}" as a valid FITS unit.\n'

Open/closed field map

Open/closed field map

Creating an open/closed field map on the solar surface.

First, import required modules

import astropy.constants as const
import astropy.units as u
import matplotlib.colors as mcolor
import matplotlib.pyplot as plt
import numpy as np
import sunpy.map
from astropy.coordinates import SkyCoord

import pfsspy
from pfsspy import tracing
from pfsspy.sample_data import get_gong_map

Load a GONG magnetic field map

gong_fname = get_gong_map()
gong_map = sunpy.map.Map(gong_fname)

Set the model parameters

nrho = 60
rss = 2.5

Construct the input, and calculate the output solution

pfss_in = pfsspy.Input(gong_map, nrho, rss)
pfss_out = pfsspy.pfss(pfss_in)

Finally, using the 3D magnetic field solution we can trace some field lines. In this case a grid of 90 x 180 points equally gridded in theta and phi are chosen and traced from the source surface outwards.

First, set up the tracing seeds

r = const.R_sun
# Number of steps in cos(latitude)
nsteps = 90
lon_1d = np.linspace(0, 2 * np.pi, nsteps * 2 + 1)
lat_1d = np.arcsin(np.linspace(-1, 1, nsteps + 1))
lon, lat = np.meshgrid(lon_1d, lat_1d, indexing='ij')
lon, lat = lon*u.rad, lat*u.rad
seeds = SkyCoord(lon.ravel(), lat.ravel(), r, frame=pfss_out.coordinate_frame)

Trace the field lines

print('Tracing field lines...')
tracer = tracing.FortranTracer(max_steps=2000)
field_lines = tracer.trace(seeds, pfss_out)
print('Finished tracing field lines')

Plot the result. The to plot is the input magnetogram, and the bottom plot shows a contour map of the the footpoint polarities, which are +/- 1 for open field regions and 0 for closed field regions.

fig = plt.figure()
m = pfss_in.map
ax = fig.add_subplot(2, 1, 1, projection=m)
m.plot()
ax.set_title('Input GONG magnetogram')

ax = fig.add_subplot(2, 1, 2)
cmap = mcolor.ListedColormap(['tab:red', 'black', 'tab:blue'])
norm = mcolor.BoundaryNorm([-1.5, -0.5, 0.5, 1.5], ncolors=3)
pols = field_lines.polarities.reshape(2 * nsteps + 1, nsteps + 1).T
ax.contourf(np.rad2deg(lon_1d), np.sin(lat_1d), pols, norm=norm, cmap=cmap)
ax.set_ylabel('sin(latitude)')

ax.set_title('Open (blue/red) and closed (black) field')
ax.set_aspect(0.5 * 360 / 2)

plt.show()

Dipole source solution

Dipole source solution

A simple example showing how to use pfsspy to compute the solution to a dipole source field.

First, import required modules

import astropy.constants as const
import astropy.units as u
import matplotlib.patches as mpatch
import matplotlib.pyplot as plt
import numpy as np
import sunpy.map
from astropy.coordinates import SkyCoord
from astropy.time import Time

import pfsspy
import pfsspy.coords as coords

To start with we need to construct an input for the PFSS model. To do this, first set up a regular 2D grid in (phi, s), where s = cos(theta) and (phi, theta) are the standard spherical coordinate system angular coordinates. In this case the resolution is (360 x 180).

nphi = 360
ns = 180

phi = np.linspace(0, 2 * np.pi, nphi)
s = np.linspace(-1, 1, ns)
s, phi = np.meshgrid(s, phi)

Now we can take the grid and calculate the boundary condition magnetic field.

def dipole_Br(r, s):
    return 2 * s / r**3


br = dipole_Br(1, s)

The PFSS solution is calculated on a regular 3D grid in (phi, s, rho), where rho = ln(r), and r is the standard spherical radial coordinate. We need to define the number of rho grid points, and the source surface radius.

nrho = 30
rss = 2.5

From the boundary condition, number of radial grid points, and source surface, we now construct an Input object that stores this information

header = pfsspy.utils.carr_cea_wcs_header(Time('2020-1-1'), br.shape)
input_map = sunpy.map.Map((br.T, header))
pfss_in = pfsspy.Input(input_map, nrho, rss)

Using the Input object, plot the input field

m = pfss_in.map
fig = plt.figure()
ax = plt.subplot(projection=m)
m.plot()
plt.colorbar()
ax.set_title('Input dipole field')

Out:

Text(0.5, 1.0, 'Input dipole field')

Now calculate the PFSS solution.

pfss_out = pfsspy.pfss(pfss_in)

Using the Output object we can plot the source surface field, and the polarity inversion line.

ss_br = pfss_out.source_surface_br

# Create the figure and axes
fig = plt.figure()
ax = plt.subplot(projection=ss_br)

# Plot the source surface map
ss_br.plot()
# Plot the polarity inversion line
ax.plot_coord(pfss_out.source_surface_pils[0])
# Plot formatting
plt.colorbar()
ax.set_title('Source surface magnetic field')

Out:

Text(0.5, 1.0, 'Source surface magnetic field')

Finally, using the 3D magnetic field solution we can trace some field lines. In this case 32 points equally spaced in theta are chosen and traced from the source surface outwards.

fig, ax = plt.subplots()
ax.set_aspect('equal')

# Take 32 start points spaced equally in theta
r = 1.01 * const.R_sun
lon = np.pi / 2 * u.rad
lat = np.linspace(-np.pi / 2, np.pi / 2, 33) * u.rad
seeds = SkyCoord(lon, lat, r, frame=pfss_out.coordinate_frame)

tracer = pfsspy.tracing.FortranTracer()
field_lines = tracer.trace(seeds, pfss_out)

for field_line in field_lines:
    coords = field_line.coords
    coords.representation_type = 'cartesian'
    color = {0: 'black', -1: 'tab:blue', 1: 'tab:red'}.get(field_line.polarity)
    ax.plot(coords.y / const.R_sun,
            coords.z / const.R_sun, color=color)

# Add inner and outer boundary circles
ax.add_patch(mpatch.Circle((0, 0), 1, color='k', fill=False))
ax.add_patch(mpatch.Circle((0, 0), pfss_in.grid.rss, color='k', linestyle='--',
                           fill=False))
ax.set_title('PFSS solution for a dipole source field')
plt.show()

Overplotting field lines on AIA maps

Overplotting field lines on AIA maps

This example shows how to take a PFSS solution, trace some field lines, and overplot the traced field lines on an AIA 193 map.

First, we import the required modules

import os

import astropy.units as u
import matplotlib.pyplot as plt
import numpy as np
import sunpy.io.fits
import sunpy.map
from astropy.coordinates import SkyCoord

import pfsspy
import pfsspy.tracing as tracing
from pfsspy.sample_data import get_gong_map

Load a GONG magnetic field map

gong_fname = get_gong_map()
gong_map = sunpy.map.Map(gong_fname)

Load the corresponding AIA 193 map

if not os.path.exists('aia_map.fits'):
    import urllib.request
    urllib.request.urlretrieve(
        'http://jsoc2.stanford.edu/data/aia/synoptic/2020/09/01/H1300/AIA20200901_1300_0193.fits',
        'aia_map.fits')

aia = sunpy.map.Map('aia_map.fits')
dtime = aia.date

The PFSS solution is calculated on a regular 3D grid in (phi, s, rho), where rho = ln(r), and r is the standard spherical radial coordinate. We need to define the number of grid points in rho, and the source surface radius.

nrho = 25
rss = 2.5

From the boundary condition, number of radial grid points, and source surface, we now construct an Input object that stores this information

pfss_in = pfsspy.Input(gong_map, nrho, rss)

Using the Input object, plot the input photospheric magnetic field

m = pfss_in.map
fig = plt.figure()
ax = plt.subplot(projection=m)
m.plot()
plt.colorbar()
ax.set_title('Input field')

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Set MJD-OBS to 59093.544444 from DATE-OBS'.
  FITSFixedWarning)

Text(0.5, 1.0, 'Input field')

We can also plot the AIA map to give an idea of the global picture. There is a nice active region in the top right of the AIA plot, that can also be seen in the top left of the photospheric field plot above.

ax = plt.subplot(1, 1, 1, projection=aia)
aia.plot(ax)

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Set MJD-OBS to 59093.541723 from DATE-OBS'.
  FITSFixedWarning)

<matplotlib.image.AxesImage object at 0x7fdc37046550>

Now we construct a 5 x 5 grid of footpoitns to trace some magnetic field lines from. These coordinates are defined in the helioprojective frame of the AIA image

hp_lon = np.linspace(-250, 250, 5) * u.arcsec
hp_lat = np.linspace(250, 500, 5) * u.arcsec
# Make a 2D grid from these 1D points
lon, lat = np.meshgrid(hp_lon, hp_lat)
seeds = SkyCoord(lon.ravel(), lat.ravel(),
                 frame=aia.coordinate_frame)
fig = plt.figure()
ax = plt.subplot(projection=aia)
aia.plot(axes=ax)
ax.plot_coord(seeds, color='white', marker='o', linewidth=0)

Out:

[<matplotlib.lines.Line2D object at 0x7fdc3700aa50>]

Plot the magnetogram and the seed footpoints The footpoints are centered around the active region metnioned above.

m = pfss_in.map
fig = plt.figure()
ax = plt.subplot(projection=m)
m.plot()
plt.colorbar()

ax.plot_coord(seeds, color='black', marker='o', linewidth=0, markersize=2)

# Set the axes limits. These limits have to be in pixel values
# ax.set_xlim(0, 180)
# ax.set_ylim(45, 135)
ax.set_title('Field line footpoints')
ax.set_ylim(bottom=0)

Out:

(0.0, 179.5)

Compute the PFSS solution from the GONG magnetic field input

pfss_out = pfsspy.pfss(pfss_in)

Trace field lines from the footpoints defined above.

tracer = tracing.FortranTracer()
flines = tracer.trace(seeds, pfss_out)

Plot the input GONG magnetic field map, along with the traced mangetic field lines.

m = pfss_in.map
fig = plt.figure()
ax = plt.subplot(projection=m)
m.plot()
plt.colorbar()

for fline in flines:
    ax.plot_coord(fline.coords, color='black', linewidth=1)

# Set the axes limits. These limits have to be in pixel values
# ax.set_xlim(0, 180)
# ax.set_ylim(45, 135)
ax.set_title('Photospheric field and traced field lines')

Out:

Text(0.5, 1.0, 'Photospheric field and traced field lines')

Plot the AIA map, along with the traced magnetic field lines. Inside the loop the field lines are converted to the AIA observer coordinate frame, and then plotted on top of the map.

fig = plt.figure()
ax = plt.subplot(1, 1, 1, projection=aia)
aia.plot(ax)
for fline in flines:
    ax.plot_coord(fline.coords, alpha=0.8, linewidth=1, color='white')

# ax.set_xlim(500, 900)
# ax.set_ylim(400, 800)
plt.show()

# sphinx_gallery_thumbnail_number = 5

GONG PFSS extrapolation

GONG PFSS extrapolation

Calculating PFSS solution for a GONG synoptic magnetic field map.

First, import required modules

import astropy.constants as const
import astropy.units as u
import matplotlib.pyplot as plt
import numpy as np
import sunpy.map
from astropy.coordinates import SkyCoord

import pfsspy
from pfsspy import coords, tracing
from pfsspy.sample_data import get_gong_map

Load a GONG magnetic field map

gong_fname = get_gong_map()
gong_map = sunpy.map.Map(gong_fname)

The PFSS solution is calculated on a regular 3D grid in (phi, s, rho), where rho = ln(r), and r is the standard spherical radial coordinate. We need to define the number of rho grid points, and the source surface radius.

nrho = 35
rss = 2.5

From the boundary condition, number of radial grid points, and source surface, we now construct an Input object that stores this information

pfss_in = pfsspy.Input(gong_map, nrho, rss)


def set_axes_lims(ax):
    ax.set_xlim(0, 360)
    ax.set_ylim(0, 180)

Using the Input object, plot the input field

m = pfss_in.map
fig = plt.figure()
ax = plt.subplot(projection=m)
m.plot()
plt.colorbar()
ax.set_title('Input field')
set_axes_lims(ax)

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Set MJD-OBS to 59093.544444 from DATE-OBS'.
  FITSFixedWarning)

Now calculate the PFSS solution

pfss_out = pfsspy.pfss(pfss_in)

Using the Output object we can plot the source surface field, and the polarity inversion line.

ss_br = pfss_out.source_surface_br
# Create the figure and axes
fig = plt.figure()
ax = plt.subplot(projection=ss_br)

# Plot the source surface map
ss_br.plot()
# Plot the polarity inversion line
ax.plot_coord(pfss_out.source_surface_pils[0])
# Plot formatting
plt.colorbar()
ax.set_title('Source surface magnetic field')
set_axes_lims(ax)

It is also easy to plot the magnetic field at an arbitrary height within the PFSS solution.

# Get the radial magnetic field at a given height
ridx = 15
br = pfss_out.bc[0][:, :, ridx]
# Create a sunpy Map object using output WCS
br = sunpy.map.Map(br.T, pfss_out.source_surface_br.wcs)
# Get the radial coordinate
r = np.exp(pfss_out.grid.rc[ridx])

# Create the figure and axes
fig = plt.figure()
ax = plt.subplot(projection=br)

# Plot the source surface map
br.plot(cmap='RdBu')
# Plot formatting
plt.colorbar()
ax.set_title('$B_{r}$ ' + f'at r={r:.2f}' + '$r_{\\odot}$')
set_axes_lims(ax)

Finally, using the 3D magnetic field solution we can trace some field lines. In this case 64 points equally gridded in theta and phi are chosen and traced from the source surface outwards.

fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')

tracer = tracing.FortranTracer()
r = 1.2 * const.R_sun
lat = np.linspace(-np.pi / 2, np.pi / 2, 8, endpoint=False)
lon = np.linspace(0, 2 * np.pi, 8, endpoint=False)
lat, lon = np.meshgrid(lat, lon, indexing='ij')
lat, lon = lat.ravel() * u.rad, lon.ravel() * u.rad

seeds = SkyCoord(lon, lat, r, frame=pfss_out.coordinate_frame)

field_lines = tracer.trace(seeds, pfss_out)

for field_line in field_lines:
    color = {0: 'black', -1: 'tab:blue', 1: 'tab:red'}.get(field_line.polarity)
    coords = field_line.coords
    coords.representation_type = 'cartesian'
    ax.plot(coords.x / const.R_sun,
            coords.y / const.R_sun,
            coords.z / const.R_sun,
            color=color, linewidth=1)

ax.set_title('PFSS solution')
plt.show()

# sphinx_gallery_thumbnail_number = 4

Finding data

Examples showing how to find, download, and load magnetograms.

HMI data

HMI data

How to search for HMI data.

This example shows how to search for, download, and load HMI data, using the sunpy.net.Fido interface. HMI data is available via. the Joint Stanford Operations Center (JSOC).

The polar filled radial magnetic field synoptic maps are obtained using the ‘hmi.synoptic_mr_polfil_720s’ series keyword. Note that they are large (1440 x 720), so you will probably want to downsample them to a smaller resolution to use them to calculate PFSS solutions.

For more information on the maps, see the synoptic maps page on the JSOC site.

First import the required modules

import os

import sunpy.map
from sunpy.net import Fido
from sunpy.net import attrs as a

import pfsspy.utils

Set up the search.

Note that for SunPy versions earlier than 2.0, a time attribute is needed to do the search, even if (in this case) it isn’t used, as the synoptic maps are labelled by Carrington rotation number instead of time

time = a.Time('2010/01/01', '2010/01/01')
series = a.jsoc.Series('hmi.synoptic_mr_polfil_720s')

Do the search. This will return all the maps in the ‘hmi_mrsynop_small_720s series.’

result = Fido.search(time, series)
print(result)

Out:

Results from 1 Provider:

152 Results from the JSOCClient:
TELESCOP  INSTRUME WAVELNTH CAR_ROT
-------- --------- -------- -------
 SDO/HMI HMI_SIDE1   6173.0    2097
 SDO/HMI HMI_SIDE1   6173.0    2098
 SDO/HMI HMI_SIDE1   6173.0    2099
 SDO/HMI HMI_SIDE1   6173.0    2100
 SDO/HMI HMI_SIDE1   6173.0    2101
 SDO/HMI HMI_SIDE1   6173.0    2102
 SDO/HMI HMI_SIDE1   6173.0    2103
 SDO/HMI HMI_SIDE1   6173.0    2104
 SDO/HMI HMI_SIDE1   6173.0    2105
 SDO/HMI HMI_SIDE1   6173.0    2106
     ...       ...      ...     ...
 SDO/HMI HMI_SIDE1   6173.0    2238
 SDO/HMI HMI_SIDE1   6173.0    2239
 SDO/HMI HMI_SIDE1   6173.0    2240
 SDO/HMI HMI_SIDE1   6173.0    2241
 SDO/HMI HMI_SIDE1   6173.0    2242
 SDO/HMI HMI_SIDE1   6173.0    2243
 SDO/HMI HMI_SIDE1   6173.0    2244
 SDO/HMI HMI_SIDE1   6173.0    2245
 SDO/HMI HMI_SIDE1   6173.0    2246
 SDO/HMI HMI_SIDE1   6173.0    2247
 SDO/HMI HMI_SIDE1   6173.0    2248
Length = 152 rows

If we just want to download a specific map, we can specify a Carrington rotation number. In addition, downloading files from JSOC requires a notification email. If you use this code, please replace this email address with your own one, registered here: http://jsoc.stanford.edu/ajax/register_email.html

crot = a.jsoc.PrimeKey('CAR_ROT', 2210)
result = Fido.search(time, series, crot,
                     a.jsoc.Notify(os.environ['JSOC_EMAIL']))
print(result)

Out:

Results from 1 Provider:

1 Results from the JSOCClient:
TELESCOP  INSTRUME WAVELNTH CAR_ROT
-------- --------- -------- -------
 SDO/HMI HMI_SIDE1   6173.0    2210

Download the files. This downloads files to the default sunpy download directory.

files = Fido.fetch(result)
print(files)

Out:

Export request pending. [id=JSOC_20211018_1007_X_IN, status=2]
Waiting for 0 seconds...
1 URLs found for download. Full request totalling 6MB

Files Downloaded:   0%|          | 0/1 [00:00<?, ?file/s]
Files Downloaded: 100%|##########| 1/1 [00:00<00:00,  2.70file/s]
Files Downloaded: 100%|##########| 1/1 [00:00<00:00,  2.70file/s]
['/home/docs/sunpy/data/hmi.synoptic_mr_polfil_720s.2210.Mr_polfil.fits']

Read in a file. This will read in the first file downloaded to a sunpy Map object. Note that HMI maps have several bits of metadata that do not comply to the FITS standard, so we need to fix them first.

hmi_map = sunpy.map.Map(files[0])
pfsspy.utils.fix_hmi_meta(hmi_map)
hmi_map.peek()

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Invalid DATE-OBS format '2018.11.09_12:30:52_TAI''.
  FITSFixedWarning)
/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'unitfix' made the change 'Changed units:
  'degree' -> 'deg'.
  FITSFixedWarning)

Parsing ADAPT Ensemble .fits files

Parsing ADAPT Ensemble .fits files

Parse an ADAPT FITS file into a sunpy.map.MapSequence.

Necessary imports

import matplotlib.pyplot as plt
import sunpy.io
import sunpy.map
from matplotlib import gridspec

from pfsspy.sample_data import get_adapt_map

Load an example ADAPT fits file, utility stored in adapt_helpers.py

adapt_fname = get_adapt_map()

ADAPT synoptic magnetograms contain 12 realizations of synoptic magnetograms output as a result of varying model assumptions. See [here](https://www.swpc.noaa.gov/sites/default/files/images/u33/SWW_2012_Talk_04_27_2012_Arge.pdf))

Because the fits data is 3D, it cannot be passed directly to sunpy.map.Map, because this will take the first slice only and the other realizations are lost. We want to end up with a sunpy.map.MapSequence containing all these realiations as individual maps. These maps can then be individually accessed and PFSS solutions generated from them.

We first read in the fits file using sunpy.io :

adapt_fits = sunpy.io.fits.read(adapt_fname)

adapt_fits is a list of HDPair objects. The first of these contains the 12 realizations data and a header with sufficient information to build the MapSequence. We unpack this HDPair into a list of (data,header) tuples where data are the different adapt realizations.

data_header_pairs = [(map_slice, adapt_fits[0].header)
                     for map_slice in adapt_fits[0].data]

Next, pass this list of tuples as the argument to sunpy.map.Map to create the map sequence :

adapt_maps = sunpy.map.Map(data_header_pairs, sequence=True)

adapt_map_sequence is now a list of our individual adapt realizations. Note the .peek()` and ``.plot() methods of MapSequence returns instances of sunpy.visualization.MapSequenceAnimator and matplotlib.animation.FuncAnimation1. Here, we generate a static plot accessing the individual maps in turn :

fig = plt.figure(figsize=(7, 8))
gs = gridspec.GridSpec(4, 3, figure=fig)
for i, a_map in enumerate(adapt_maps):
    ax = fig.add_subplot(gs[i], projection=a_map)
    a_map.plot(axes=ax, cmap='bwr', vmin=-2, vmax=2,
               title=f"Realization {1+i:02d}")

plt.tight_layout(pad=5, h_pad=2)
plt.show()

pfsspy utilities

Re-projecting from CAR to CEA

Re-projecting from CAR to CEA

The pfsspy solver takes a cylindrical-equal-area (CEA) projected magnetic field map as input, which is equally spaced in sin(latitude). Some synoptic field maps are equally spaced in latitude however, which is a plate carée (CAR) projection.

This example shows how to use the pfsspy.utils.car_to_cea function to reproject a CAR projection to a CEA projection that pfsspy can take as input.

import matplotlib.pyplot as plt

from pfsspy import sample_data, utils

Load a sample ADAPT map, which has a CAR projection

adapt_maps = utils.load_adapt(sample_data.get_adapt_map())
adapt_map_car = adapt_maps[0]

Out:

Files Downloaded:   0%|          | 0/1 [00:00<?, ?file/s]

pfsspy.adapt40311_03k012_202001010000_i00005600n1.fts.gz:   0%|          | 0.00/3.11M [00:00<?, ?B/s]

pfsspy.adapt40311_03k012_202001010000_i00005600n1.fts.gz:   0%|          | 100/3.11M [00:00<1:28:53, 582B/s]

pfsspy.adapt40311_03k012_202001010000_i00005600n1.fts.gz:  11%|#1        | 349k/3.11M [00:00<00:01, 1.58MB/s]

pfsspy.adapt40311_03k012_202001010000_i00005600n1.fts.gz:  31%|###1      | 965k/3.11M [00:00<00:00, 3.38MB/s]

pfsspy.adapt40311_03k012_202001010000_i00005600n1.fts.gz:  68%|######8   | 2.13M/3.11M [00:00<00:00, 6.34MB/s]


Files Downloaded: 100%|##########| 1/1 [00:00<00:00,  1.34file/s]
Files Downloaded: 100%|##########| 1/1 [00:00<00:00,  1.34file/s]

Re-project into a CEA projection

adapt_map_cea = utils.car_to_cea(adapt_map_car)

Out:

/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/astropy/wcs/wcs.py:710: FITSFixedWarning: 'datfix' made the change 'Set MJD-OBS to 58849.000000 from DATE-OBS'.
  FITSFixedWarning)
/home/docs/checkouts/readthedocs.org/user_builds/pfsspy/envs/1.0.0/lib/python3.7/site-packages/sunpy/util/decorators.py:378: SunpyMetadataWarning: Missing metadata for observer: assuming Earth-based observer.

  new_val = prop(instance)

Plot the original map and the reprojected map

plt.figure()
adapt_map_car.plot()
plt.figure()
adapt_map_cea.plot()

plt.show()

• 
• 

pfsspy information

Examples showing how the internals of pfsspy work.

pfsspy magnetic field grid

pfsspy magnetic field grid

A plot of the grid corners, from which the magnetic field values are taken when tracing magnetic field lines.

Notice how the spacing becomes larger at the poles, and closer to the source surface. This is because the grid is equally spaced in \(\cos \theta\) and \(\log r\).

import matplotlib.pyplot as plt
import numpy as np

from pfsspy.grid import Grid

Define the grid spacings

ns = 15
nphi = 360
nr = 10
rss = 2.5

Create the grid

grid = Grid(ns, nphi, nr, rss)

Get the grid edges, and transform to r and theta coordinates

r_edges = np.exp(grid.rg)
theta_edges = np.arccos(grid.sg)

The corners of the grid are where lines of constant (r, theta) intersect, so meshgrid these together to get all the grid corners.

r_grid_points, theta_grid_points = np.meshgrid(r_edges, theta_edges)

Plot the resulting grid corners

fig = plt.figure()
ax = fig.add_subplot(projection='polar')

ax.scatter(theta_grid_points, r_grid_points)
ax.scatter(theta_grid_points + np.pi, r_grid_points, color='C0')

ax.set_ylim(0, 1.1 * rss)
ax.set_theta_zero_location('N')
ax.set_yticks([1, 1.5, 2, 2.5], minor=False)
ax.set_title('$n_{r}$ = ' f'{nr}, ' r'$n_{\theta}$ = ' f'{ns}')

plt.show()

Tracer performance

Tracer performance

A quick script to compare the performance of the python and fortran tracers.

import timeit

import astropy.coordinates
import astropy.units as u
import matplotlib.pyplot as plt
import numpy as np
import sunpy.map

import pfsspy

Create a dipole map

ntheta = 180
nphi = 360
nr = 50
rss = 2.5

phi = np.linspace(0, 2 * np.pi, nphi)
theta = np.linspace(-np.pi / 2, np.pi / 2, ntheta)
theta, phi = np.meshgrid(theta, phi)


def dipole_Br(r, theta):
    return 2 * np.sin(theta) / r**3


br = dipole_Br(1, theta)
br = sunpy.map.Map(br.T, pfsspy.utils.carr_cea_wcs_header('2010-01-01', br.shape))
pfss_input = pfsspy.Input(br, nr, rss)
pfss_output = pfsspy.pfss(pfss_input)
print('Computed PFSS solution')

Trace some field lines

seed0 = np.atleast_2d(np.array([1, 1, 0]))
tracers = [pfsspy.tracing.PythonTracer(),
           pfsspy.tracing.FortranTracer()]
nseeds = 2**np.arange(14)
times = [[], []]

for nseed in nseeds:
    print(nseed)
    seeds = np.repeat(seed0, nseed, axis=0)
    r, lat, lon = pfsspy.coords.cart2sph(seeds[:, 0], seeds[:, 1], seeds[:, 2])
    r = r * astropy.constants.R_sun
    lat = (lat - np.pi / 2) * u.rad
    lon = lon * u.rad
    seeds = astropy.coordinates.SkyCoord(lon, lat, r, frame=pfss_output.coordinate_frame)

    for i, tracer in enumerate(tracers):
        if nseed > 64 and i == 0:
            continue

        t = timeit.timeit(lambda: tracer.trace(seeds, pfss_output), number=1)
        times[i].append(t)

Plot the results

fig, ax = plt.subplots()
ax.scatter(nseeds[1:len(times[0])], times[0][1:], label='python')
ax.scatter(nseeds[1:], times[1][1:], label='fortran')

pydt = (times[0][4] - times[0][3]) / (nseeds[4] - nseeds[3])
ax.plot([1, 1e5], [pydt, 1e5 * pydt])

fort0 = times[1][1]
fordt = (times[1][-1] - times[1][-2]) / (nseeds[-1] - nseeds[-2])
ax.plot(np.logspace(0, 5, 100), fort0 + fordt * np.logspace(0, 5, 100))

ax.set_xscale('log')
ax.set_yscale('log')

ax.set_xlabel('Number of seeds')
ax.set_ylabel('Seconds')

ax.axvline(180 * 360, color='k', linestyle='--', label='180x360 seed points')

ax.legend()
plt.show()

This shows the results of the above script, run on a 2014 MacBook pro with a 2.6 GHz Dual-Core Intel Core i5:

Testing pfsspy

Tests of pfsspy.

Open flux

Open flux

Comparing total unsigned flux to analytic solutions. This is done as a function of the number of radial grid cells used in the PFSS calculation.

First, import required modules

import functools

import astropy.units as u
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import scipy.integrate
from helpers import brss_pfsspy
from tqdm import tqdm

from pfsspy import analytic


def open_flux_analytic(l, m, zss):
    Br = analytic.Br(l, m, zss)
    Br = functools.partial(Br, zss)

    def absBr(theta, phi):
        return np.abs(Br(theta * u.rad, phi * u.rad)) * np.sin(theta)

    res = scipy.integrate.nquad(absBr, ranges=[[0, np.pi], [0, 2 * np.pi]])
    return res[0]


def open_flux_numeric(l, m, zss, nrho):
    nphi = 360
    ns = 180
    br = brss_pfsspy(nphi, ns, nrho, zss, l, m)
    return np.sum(np.abs(br)) * (4 * np.pi) / nphi / ns

Set the source surface height, and the (l, m) values to investigate

zss = 2
lms = [10]

fig, ax = plt.subplots()
# nrhos = np.arange(10, 51, 2)
nrhos = [10, 50]

df = pd.DataFrame(index=nrhos, columns=lms)

for lm in lms:
    l = lm // 10
    m = lm % 10
    print(f"l={l}, m={m}")
    flux_analytic = open_flux_analytic(l, m, zss)
    flux_numeric = []
    for nrho in tqdm(nrhos):
        flux_numeric.append(open_flux_numeric(l, m, zss, nrho))
    flux_numeric = np.array(flux_numeric)
    flux_ratio = flux_numeric / flux_analytic
    df[lm] = flux_ratio

Save a copy of the data

df.to_csv('open_flux_results.csv')
df = pd.read_csv('open_flux_results.csv', index_col=0)

for lm in lms:
    l = lm // 10
    m = lm % 10
    color = {1: 'tab:blue', 2: 'tab:orange', 3: 'tab:green'}[l]
    marker = {0: 'o', 1: 10, 2: 11}[m]
    ax.plot(nrhos, df[str(lm)],
            marker=marker, color=color,
            label=f'l={l}, m={m}')

ax.legend()
ax.set_ylim(1)
ax.set_xlim(8)
ax.yaxis.grid(linestyle='--')
ax.set_xlabel('$n_{r}$')
ax.set_ylabel(r'$\Phi_{pfsspy} / \Phi_{analytic}$')
plt.show()

Spherical harmonic comparisons

Spherical harmonic comparisons

Comparing analytical spherical harmonic solutions to PFSS output.

First, import required modules

import astropy.units as u
import matplotlib.pyplot as plt
import numpy as np
import sunpy.map
from matplotlib.gridspec import GridSpec

import pfsspy
from pfsspy import analytic

Setup some useful functions for testing

def theta_phi(nphi, ns):
    # Return a theta, phi grid with a given numer of points
    phi = np.linspace(0, 2 * np.pi, nphi)
    s = np.linspace(-1, 1, ns)
    s, phi = np.meshgrid(s, phi)
    theta = np.arccos(s)
    return theta * u.rad, phi * u.rad


def brss_pfsspy(nphi, ns, nrho, rss, l, m):
    # Return the pfsspy solution for given input parameters
    theta, phi = theta_phi(nphi, ns)

    br_in = analytic.Br(l, m, rss)(1, theta, phi)

    header = pfsspy.utils.carr_cea_wcs_header('2020-1-1', br_in.shape)
    input_map = sunpy.map.Map((br_in.T, header))

    pfss_input = pfsspy.Input(input_map, nrho, rss)
    pfss_output = pfsspy.pfss(pfss_input)
    return pfss_output.bc[0][:, :, -1].T


def brss_analytic(nphi, ns, rss, l, m):
    # Return the analytic solution for given input parameters
    theta, phi = theta_phi(nphi, ns)
    return analytic.Br(l, m, rss)(rss, theta, phi).T

Compare the the pfsspy solution to the analytic solutions. Cuts are taken on the source surface at a constant phi value to do a 1D comparison.

ls = [1, 2, 3]
fig = plt.figure(tight_layout=True)
gs = GridSpec(len(ls), len(ls) + 1)


for i, l in enumerate(ls):
    ax0 = fig.add_subplot(gs[i, 0])
    axs = [fig.add_subplot(gs[i, j], sharey=ax0) for j in range(1, l + 1)]
    axs = [ax0] + axs
    for j, m in enumerate(list(range(l + 1))):
        ax = axs[j]
        nphi = 359
        ns = 179
        rss = 2.5
        nrho = 20

        br_pfsspy = brss_pfsspy(nphi, ns, nrho, rss, l, m)
        br_actual = brss_analytic(nphi, ns, rss, l, m)

        ax.plot(br_pfsspy[:, 180], label='pfsspy')
        ax.plot(br_actual[:, 180], label='analytic')
        ax.set_title(f'l={l}, m={m}')
        if i == 0 and j == 0:
            ax.legend()

plt.show()

Open flux

Open flux

Comparing total unsigned flux to analytic solutions. This is done on a fixed number of radial grid points, and plotted as a function of spherical harmonic.

First, import required modules

import functools
import json

import astropy.units as u
import matplotlib.colors as mcolor
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker
import numpy as np
import scipy.integrate
from helpers import brss_pfsspy

from pfsspy import analytic


def open_flux_analytic(l, m, zss):
    Br = analytic.Br(l, m, zss)
    Br = functools.partial(Br, zss)

    def absBr(theta, phi):
        return np.abs(Br(theta * u.rad, phi * u.rad)) * np.sin(theta)
    res = scipy.integrate.nquad(absBr, ranges=[[0, np.pi], [0, 2 * np.pi]])
    return res[0]


def open_flux_numeric(l, m, zss, nrho):
    nphi = 360
    ns = 180
    br = brss_pfsspy(nphi, ns, nrho, zss, l, m)
    return np.sum(np.abs(br)) * (4 * np.pi) / nphi / ns

Set the source surface height, and the (l, m) values to investigate

zss = 2
nrho = 40

results = {}

for l in range(1, 6):
    results[l] = {}
    for m in range(0, l + 1):
        print(f"l={l}, m={m}")
        flux_analytic = open_flux_analytic(l, m, zss)
        flux_numeric = open_flux_numeric(l, m, zss, nrho)
        results[l][m] = flux_numeric / flux_analytic

# open file for writing, "w"
with open("open_flux_harmonics.json", "w") as f:
    # write json object to file
    f.write(json.dumps(results))


with open("open_flux_harmonics.json", "r") as f:
    results = json.load(f, parse_int=int)
print(results)

Plot results

fig, ax = plt.subplots()
norm = mcolor.Normalize(vmin=1, vmax=1.06)
for lstr in results:
    l = int(lstr)
    data = np.atleast_2d(list(results[lstr].values())).T
    im = ax.imshow(data, extent=[l-0.5, l+0.5, -0.5, l+0.5],
                   norm=norm)

fig.colorbar(im, label=r'$\Phi_{pfsspy} / \Phi_{analytic}$')
ax.set_xlim(0.5, l+0.5)
ax.set_ylim(-0.5, l+0.5)
ax.xaxis.set_major_locator(mticker.MultipleLocator(1))
ax.yaxis.set_major_locator(mticker.MultipleLocator(1))


def fmt(x, pos):
    return str(int(x))


ax.xaxis.set_major_formatter(fmt)
ax.yaxis.set_major_formatter(fmt)
ax.set_xlabel('l')
ax.set_ylabel('m')
fig.savefig('flux_harmonics.pdf', bbox_inches='tight')
plt.show()

Analytic dipole field lines

Analytic dipole field lines

First, import required modules

import astropy.constants as const
import astropy.units as u
import matplotlib.pyplot as plt
import numpy as np
from astropy.coordinates import SkyCoord
from astropy.visualization import quantity_support
from helpers import pffspy_output, phi_fline_coords, theta_fline_coords

from pfsspy import tracing

quantity_support()

Out:

<astropy.visualization.units.quantity_support.<locals>.MplQuantityConverter object at 0x7fdc89b76e90>

Compare the the pfsspy solution to the analytic solutions. Cuts are taken on the source surface at a constant phi value to do a 1D comparison.

l = 1
m = 1
nphi = 360
ns = 180
nr = 40
rss = 2

Calculate PFSS solution

pfsspy_out = pffspy_output(nphi, ns, nr, rss, l, m)

rss = rss * const.R_sun

Trace some field lines

n = 90
# Create 1D theta, phi arrays
phi = np.linspace(0, 360, n * 2)
phi = phi[:-1] + np.diff(phi) / 2
theta = np.arcsin(np.linspace(-1, 1, n, endpoint=False) + 1/n)
# Mesh into 2D arrays
theta, phi = np.meshgrid(theta, phi, indexing='ij')
theta, phi = theta * u.rad, phi * u.deg
seeds = SkyCoord(radius=rss, lat=theta.ravel(), lon=phi.ravel(),
                 frame=pfsspy_out.coordinate_frame)

step_size = 1
dthetas = []
print(f'Tracing {step_size}...')
# Trace
tracer = tracing.FortranTracer(step_size=step_size)
flines = tracer.trace(seeds, pfsspy_out)
# Set a mask of open field lines
mask = flines.connectivities.astype(bool).reshape(theta.shape)

# Get solar surface latitude
phi_solar = np.ones_like(phi) * np.nan
phi_solar[mask] = flines.open_field_lines.solar_feet.lon
theta_solar = np.ones_like(theta) * np.nan
theta_solar[mask] = flines.open_field_lines.solar_feet.lat
r_out = np.ones_like(theta.value) * const.R_sun * np.nan
r_out[mask] = flines.open_field_lines.solar_feet.radius

Out:

Tracing 1...

Calculate analytical solution

theta_analytic = theta_fline_coords(r_out, rss, l, m, theta)
dtheta = theta_solar - theta_analytic
phi_analytic = phi_fline_coords(r_out, rss, l, m, theta, phi)
dphi = phi_solar - phi_analytic

fig, axs = plt.subplots(nrows=2, sharex=True, sharey=True)


def plot_map(field, ax, label, title):
    kwargs = dict(cmap='RdBu', vmin=-0.5, vmax=0.5, shading='nearest', edgecolors='face')
    im = ax.pcolormesh(phi.to_value(u.deg), np.sin(theta).value,
                       field, **kwargs)
    ax.contour(phi.to_value(u.deg), np.sin(theta).value,
               field,
               levels=[-0.4, -0.3, -0.2, -0.1, 0.1, 0.2, 0.3, 0.4],
               colors='black', alpha=0.5, linewidths=1)
    ax.set_aspect(360 / 4)
    fig.colorbar(im, aspect=10, ax=ax,
                 label=label)
    ax.set_title(title, size=10)


plot_map(dtheta.to_value(u.deg), axs[0],
         r'$\theta_{pfsspy} - \theta_{analytic}$ (deg)',
         'Error in latitude')
plot_map(dphi.to_value(u.deg), axs[1],
         r'$\phi_{pfsspy} - \phi_{analytic}$ (deg)',
         'Error in longitude')

ax = axs[1]
ax.set_xlim(0, 360)
ax.set_ylim(-1, 1)
ax.set_xlabel('Longitude (deg)')
ax.set_ylabel('sin(Latitude)')

fig.suptitle(f'l={l}, m={m}')
fig.tight_layout()
fig.savefig('error_map.pdf', bbox_inches='tight')

plt.show()

API reference

pfsspy Package

Functions

pfss(input)

Compute PFSS model.

Classes

Input(br, nr, rss)	Input to PFSS modelling.
Output(alr, als, alp, grid[, input_map])	Output of PFSS modelling.

Class Inheritance Diagram

pfsspy.grid Module

Classes

Grid(ns, nphi, nr, rss)

Grid on which the pfsspy solution is calculated.

Class Inheritance Diagram

pfsspy.fieldline Module

Classes

ClosedFieldLines(field_lines)	A set of closed field lines.
FieldLine(x, y, z, output)	A single magnetic field line.
FieldLines(field_lines)	A collection of FieldLine.
OpenFieldLines(field_lines)	A set of open field lines.

Class Inheritance Diagram

pfsspy.tracing Module

Classes

FortranTracer([max_steps, step_size])	Tracer using Fortran code.
PythonTracer([atol, rtol])	Tracer using native python code.
Tracer()	Abstract base class for a streamline tracer.

Class Inheritance Diagram

pfsspy.utils Module

Functions

car_to_cea(m[, method])	Reproject a plate-carée map in to a cylindrical-equal-area map.
carr_cea_wcs_header(dtime, shape)	Create a Carrington WCS header for a Cylindrical Equal Area (CEA) projection.
fix_hmi_meta(hmi_map)	Fix non-compliant FITS metadata in HMI maps.
is_car_map(m[, error])	Returns True if m is in a plate carée projeciton.
is_cea_map(m[, error])	Returns True if m is in a cylindrical equal area projeciton.
is_full_sun_synoptic_map(m[, error])	Returns True if m is a synoptic map spanning the solar surface.
load_adapt(adapt_path)	Parse adapt .fts file as a sunpy.map.MapSequence
roll_map(m[, lh_edge_lon, method])	Roll an input synoptic map so that it's left edge corresponds to a specific Carrington longitude.

pfsspy.analytic Module

Analytic inputs and solutions to the PFSS equations.

This sub-module contains functions to generate solutions to the PFSS equations in the case where the input field is a single spherical harmonic, specified with the spherical harmonic numbers l, m.

All angular quantities must be passed as astropy quantities. All radial quantities are passed normalised to the source surface radius, and therefore can be passed as normal scalar values.

Angular definitions

• theta is the polar angle, in the range \(0, \pi\) (ie. the co-latitude).

• phi is the azimuthal angle, in the range \(0, 2\pi\).

Using this module requires sympy to be installed.

Functions

Bphi(l, m, zss)	Analytic phi component of magnetic field on the source surface.
Br(l, m, zss)	Analytic radial component of magnetic field on the source surface.
Btheta(l, m, zss)	Analytic theta component of magnetic field on the source surface.

Improving performance

numba

pfsspy automatically detects an installation of numba, which compiles some of the numerical code to speed up pfss calculations. To enable this simply install numba and use pfsspy as normal.

Streamline tracing

pfsspy has two streamline tracers: a pure python pfsspy.tracing.PythonTracer and a FORTRAN pfsspy.tracing.FortranTracer. The FORTRAN version is significantly faster, using the streamtracer package.

Changelog

1.0.0

New requirements

pfsspy now depends on python >= 3.7, and now does not depend on Matplotlib.

New features

• The max_steps argument to pfsspy.tracers.FortranTracer now defaults to 'auto' and automatically sets the maximum number of steps to four times the number of steps that are needed to span radially from the solar to source surface. max_steps can still be manually specified as a number if more or less steps are desired.

• FieldLines now has a __len__ method, meaning one can now do n_field_lines = len(my_field_lines).

• Added pfsspy.utils.roll_map() to roll a map in the longitude direction. This is particularly helpful to modify GONG maps so they have a common longitude axis.

• Added the pfsspy.analytic sub-module that provides functions to sample analytic solutions to the PFSS equations.

Bug fixes

• pfsspy.utils.carr_cea_wcs_header() now works with versions of sunpy >= 2.0.

• GONG synoptic maps now automatically have their observer information corrected (by assuming an Earth observer) when loaded by sunpy.map.Map.

• The plot settings of input maps are no longer modified in place.

Breaking changes

• The interpretation of the step_size to pfsspy.tracers.FortranTracer has been corrected so that it is the step size relative to the radial cell size. A step size of 0.01 specified in pfsspy<1.0 is approximately equivalent to a step size of 1 in pfsspy 1.0, so you will need to adjust any custom step sizes accordingly.

• Any points on field lines that are out of bounds (ie. below the solar surface or above the source surface) are now removed by the FortranTracer.

• pfsspy.pfss() no longer warns if the mean of the input data is non-zero, and silently ignores the monopole component.

Removals

• Saving and load PFSS solutions is no longer possible. This was poorly tested, and possibly broken. If you have interest in saving and loading being added as a new feature to pfsspy, please open a new issue at https://github.com/dstansby/pfsspy/issues.

0.6.6

Two bugs have been fixed in pfsspy.utils.carr_cea_wcs_header:

• The reference pixel was previously one pixel too large in both longitude and latitude.

• The longitude coordinate was previously erroneously translated by one degree.

Both of these are now fixed.

0.6.5

This release improves documentation and handling of HMI maps. In particular:

• The HMI map downloading example has been updated to use the polar filled data product, which does not have any data missing at the poles.

• pfsspy.utils.fix_hmi_meta() has been added to fix metadata issues in HMI maps. This modifies the metadata of a HMI map to make it FITS compliant, allowing it to be used with pfsspy.

0.6.4

This release adds citation information to the documentation.

0.6.3

This release contains the source for the accepted JOSS paper describing pfsspy.

0.6.2

This release includes several small fixes in response to a review of pfsspy for the Journal of Open Source Software. Thanks to Matthieu Ancellin and Simon Birrer for their helpful feedback!

• A permanent code of conduct file has been added to the repository.

• Information on how to contribute to pfsspy has been added to the docs.

• The example showing the performance of different magnetic field tracers has been fixed.

• The docs are now clearer about optional dependencies that can increase performance.

• The GONG example data has been updated due to updated data on the remote GONG server.

0.6.1

Bug fixes

• Fixed some messages in errors raised by functions in pfsspy.utils.

0.6.0

New features

• The pfsspy.utils module has been added, and contains various tools for loading and working with synoptic maps.

• pfsspy.Output has a new bunit property, which returns the Unit of the input map.

• Added pfsspy.Output.get_bvec(), to sample the magnetic field solution at arbitrary coordinates.

• Added the pfsspy.fieldline.FieldLine.b_along_fline property, to sample the magnetic field along a traced field line.

• Added a guide to the numerical methods used by pfsspy.

Breaking changes

• The .al property of pfsspy.Output is now private, as it is not intended for user access. If you really want to access it, use ._al (but this is now private API and there is no guarantee it will stay or return the same thing in the future).

• A ValueError is now raised if any of the input data to pfsspy.Input is non-finite or NaN. Previously the PFSS computation would run fine, but the output would consist entirely of NaNs.

Behaviour changes

• The monopole term is now ignored in the PFSS calculation. Previously a non-zero (but small) monopole term would cause floating point precision issues, leading to a very noisy result. Now the monopole term is explicitly removed from the calculation. If your input has a non-zero mean value, pfsspy will issue a warning about this.

• The data downloaded by the examples is now automatically downloaded and cached with sunpy.data.manager. This means the files used for running the examples will be downloaded and stored in your sunpy data directory if they are required.

• The observer coordinate information in GONG maps is now automatically set to the location of Earth at the time in the map header.

Bug fixes

• The date-obs FITS keyword in GONG maps is now correctly populated.

0.5.3

• Improved descriptions in the AIA overplotting example.

• Fixed the ‘date-obs’ keyword in GONG metadata. Previously this just stored the date and not the time; now both the date and time are properly stored.

• Drastically sped up the calculation of source surface and solar surface magnetic field footpoints.

0.5.2

• Fixed a bug in the GONG synoptic map source where a map failed to load once it had already been loaded once.

0.5.1

• Fixed some calculations in pfsspy.carr_cea_wcs_header, and clarified in the docstring that the input shape must be in [nlon, nlat] order.

• Added validation to pfsspy.Input to check that the inputted map covers the whole solar surface.

• Removed ghost cells from pfsspy.Output.bc. This changes the shape of the returned arrays by one along some axes.

• Corrected the shape of pfsspy.Output.bg in the docstring.

• Added an example showing how to load ADAPT ensemble maps into a CompositeMap

• Sped up field line expansion factor calculations.

0.5.0

Changes to outputted maps

This release largely sees a transition to leveraging Sunpy Map objects. As such, the following changes have been made:

pfsspy.Input now must take a sunpy.map.GenericMap as an input boundary condition (as opposed to a numpy array). To convert a numpy array to a GenericMap, the helper function pfsspy.carr_cea_wcs_header can be used:

map_date = datetime(...)
br = np.array(...)
header = pfsspy.carr_cea_wcs_header(map_date, br.shape)

m = sunpy.map.Map((br, header))
pfss_input = pfsspy.Input(m, ...)

pfsspy.Output.source_surface_br now returns a GenericMap instead of an array. To get the data array use source_surface_br.data.

The new pfsspy.Output.source_surface_pils returns the coordinates of the polarity inversion lines on the source surface.

In favour of directly using the plotting functionality built into SunPy, the following plotting functionality has been removed:

• pfsspy.Input.plot_input. Instead Input has a new map property, which returns a SunPy map, which can easily be plotted using sunpy.map.GenericMap.plot.

• pfsspy.Output.plot_source_surface. A map of \(B_{r}\) on the source surface can now be obtained using pfsspy.Output.source_surface_br, which again returns a SunPy map.

• pfsspy.Output.plot_pil. The coordinates of the polarity inversion lines on the source surface can now be obtained using pfsspy.Output.source_surface_pils, which can then be plotted using ax.plot_coord(pil[0]) etc. See the examples section for an example.

Specifying tracing seeds

In order to make specifying seeds easier, they must now be a SkyCoord object. The coordinates are internally transformed to the Carrington frame of the PFSS solution, and then traced.

This should make specifying coordinates easier, as lon/lat/r coordinates can be created using:

seeds = astropy.coordinates.SkyCoord(lon, lat, r, frame=output.coordinate_frame)

To convert from the old x, y, z array used for seeds, do:

r, lat, lon = pfsspy.coords.cart2sph
r = r * astropy.constants.R_sun
lat = (lat - np.pi / 2) * u.rad
lon = lon * u.rad

seeds = astropy.coordinates.SkyCoord(lon, lat, r, frame=output.coordinate_frame)

Note that the latitude must be in the range \([-\pi/2, \pi/2]\).

GONG and ADAPT map sources

pfsspy now comes with built in sunpy map sources for GONG and ADAPT synoptic maps, which automatically fix some non-compliant FITS header values. To use these, just import pfsspy and load the .FITS files as normal with sunpy.

Tracing seeds

pfsspy.tracing.Tracer no longer has a transform_seeds helper method, which has been replaced by coords_to_xyz and pfsspy.tracing.Tracer.xyz_to_coords. These new methods convert between SkyCoord objects, and Cartesian xyz coordinates of the internal magnetic field grid.

0.4.3

• Improved the error thrown when trying to use :class`pfsspy.tracing.FotranTracer` without the streamtracer module installed.

• Fixed some layout issues in the documentation.

0.4.2

• Fix a bug where :class`pfsspy.tracing.FotranTracer` would overwrite the magnetic field values in an Output each time it was used.

0.4.1

• Reduced the default step size for the FortranTracer from 0.1 to 0.01 to give more resolved field lines by default.

0.4.0

New fortran field line tracer

pfsspy.tracing contains a new tracer, FortranTracer. This requires and uses the streamtracer package which does streamline tracing rapidly in python-wrapped fortran code. For large numbers of field lines this results in an ~50x speedup compared to the PythonTracer.

Changing existing code to use the new tracer is as easy as swapping out tracer = pfsspy.tracer.PythonTracer() for tracer = pfsspy.tracer.FortranTracer(). If you notice any issues with the new tracer, please report them at https://github.com/dstansby/pfsspy/issues.

Changes to field line objects

• pfsspy.FieldLines and pfsspy.FieldLine have moved to pfsspy.fieldline.FieldLines and pfsspy.fieldline.FieldLine.

• FieldLines no longer has source_surface_feet and solar_feet properties. Instead these have moved to the new pfsspy.fieldline.OpenFieldLines class. All the open field lines can be accessed from a FieldLines instance using the new open_field_lines property.

Changes to Output

• pfsspy.Output.bg is now returned as a 4D array instead of three 3D arrays. The final index now indexes the vector components; see the docstring for more information.

0.3.2

• Fixed a bug in pfsspy.FieldLine.is_open, where some open field lines were incorrectly calculated to be closed.

0.3.1

• Fixed a bug that incorrectly set closed line field polarities to -1 or 1 (instead of the correct value of zero).

• FieldLine.footpoints has been removed in favour of the new pfsspy.FieldLine.solar_footpoint and pfsspy.FieldLine.source_surface_footpoint. These each return a single footpoint. For a closed field line, see the API docs for further details on this.

• pfsspy.FieldLines has been added, as a convenience class to store a collection of field lines. This means convenience attributes such as pfsspy.FieldLines.source_surface_feet can be used, and their values are cached greatly speeding up repeated use.

0.3.0

• The API for doing magnetic field tracing has changed. The new pfsspy.tracing module contains Tracer classes that are used to perform the tracing. Code needs to be changed from:

  fline = output.trace(x0)
  

  to:

  tracer = pfsspy.tracing.PythonTracer()
  tracer.trace(x0, output)
  flines = tracer.xs
  

  Additionally x0 can be a 2D array that contains multiple seed points to trace, taking advantage of the parallelism of some solvers.

• The pfsspy.FieldLine class no longer inherits from SkyCoord, but the SkyCoord coordinates are now stored in pfsspy.FieldLine.coords attribute.

• pfsspy.FieldLine.expansion_factor now returns np.nan instead of None if the field line is closed.

• pfsspy.FieldLine now has a ~pfsspy.FieldLine.footpoints attribute that returns the footpoint(s) of the field line.

0.2.0

• pfsspy.Input and pfsspy.Output now take the optional keyword argument dtime, which stores the datetime on which the magnetic field measurements were made. This is then propagated to the obstime attribute of computed field lines, allowing them to be transformed in to coordinate systems other than Carrington frames.

• pfsspy.FieldLine no longer overrrides the SkyCoord __init__; this should not matter to users, as FieldLine objects are constructed internally by calling pfsspy.Output.trace

0.1.5

• Output.plot_source_surface now accepts keyword arguments that are given to Matplotlib to control the plotting of the source surface.

0.1.4

• Added more explanatory comments to the examples

• Corrected the dipole solution calculation

• Added pfsspy.coords.sph2cart to transform from spherical to cartesian coordinates.

0.1.3

• pfsspy.Output.plot_pil now accepts keyword arguments that are given to Matplotlib to control the style of the contour.

• pfsspy.FieldLine.expansion_factor is now cached, and is only calculated once if accessed multiple times.

Contributing to pfsspy

pfsspy is a community package, and contributions are welcome from anyone. This includes reporting bugs, requesting new features, along with writing code and improving documentation.

Reporting Issues

If you use pfsspy and stumble upon a problem, the best way to report it is by opening an issue on the GitHub issue tracker. This way we can help you work around the problem and hopefully fix the problem!

When reporting an issue, please try to provide a short description of the issue with a small code sample, this way we can attempt to reproduce the error. Also provide any error output generated when you encountered the issue, we can use this information to debug the issue.

Requesting new features

If there is functionality that is not currently available in pfsspy you can make a feature request on the issue page. Please add as much information as possible regarding the new feature you would like to see.

Improving documentaion

If something doesn’t make sense in the online documentation, please open an issue. If you’re comfortable fixing it, please open a pull request to the pfsspy repository.

Contributing code

All code contributions to pfsspy are welcome! If you would like to submit a contribution, please open a pull request at the pfsspy repository. It will then be reviewed, and if it is a useful addition merged into the main code branch.

If you are new to open source development, the instructions in these links might be useful for getting started:

• Astropy Dev Workflow

• Astropy Dev environment

• Astropy Pull Request Example

Numerical methods

For more information on the numerical methods used in the PFSS calculation see this document.

Synoptic map FITS conventions

FITS is the most common filetype used for the storing of solar images. On this page the FITS metadata conventions for synoptic maps are collected. All of this information can be found in, and is taken from, “Coordinate systems for solar image data (Thompson, 2005)”.

Keyword	Output
CRPIXn	Reference pixel to subtract along axis n. Counts from 1 to N. Integer values refer to the centre of the pixel.
CRVALn	Coordinate value of the reference pixel along axis n.
CDELTn	Pixel spacing along axis n.
CTYPEn	Coordinate axis label for axis n.
PVi_m	Additional parameters needed for some coordinate systems.

Note that CROTAn is ignored in this short guide.

Cylindrical equal area projection

In this projection, the latitude pixels are equally spaced in sin(latitude). The reference pixel has to be on the equator, to facilitate alignment with the solar rotation axis.

• CDELT2 is set to 180/pi times the pixel spacing in sin(latitude).

• CTYPE1 is either ‘HGLN-CEA’ or ‘CRLN-CEA’.

• CTYPE2 is either ‘HGLT-CEA’ or ‘CRLT-CEA’.

• PVi_1 is set to 1.

• LONPOLE is 0.

The abbreviations are “Heliographic Longitude - Cylindrical Equal Area” etc. If the system is heliographic the observer must also be defined in the metadata.

Indices and tables

• Index

• Module Index

• Search Page