def _spectrum_from_component(self, layer, component, wcs, mask=None):
data = SpectralCube(component.data, wcs)
if mask is not None:
data = data.with_mask(mask)
if self._data_operation.currentIndex() == 1:
spec_data = data.mean((1, 2))
elif self._data_operation.currentIndex() == 2:
spec_data = data.median((1, 2))
else:
spec_data = data.sum((1, 2))
spec_data = Spectrum1DRef(spec_data.data,
unit=spec_data.unit,
dispersion=data.spectral_axis.data,
dispersion_unit=data.spectral_axis.unit,
wcs=data.wcs,
name=layer.label)
# Store the relation between the component and the specviz data. If
# the data exists, first remove the component from specviz and then
# re-add it.
if layer in self._specviz_data_cache:
old_spec_data = self._specviz_data_cache[layer]
dispatch.on_remove_data.emit(old_spec_data)
self._specviz_data_cache[layer] = spec_data
dispatch.on_add_to_window.emit(data=spec_data,
style={'color': layer.style.rgba[:3]})
def _spectrum_from_component(self,
layer,
component,
wcs,
mask=None,
cid=None):
data = SpectralCube(component.data, wcs)
if mask is not None:
data = data.with_mask(mask)
# Update the associated data attribute in the plugin
self._spec_ops.spectral_data = data
self._spec_ops.component_id = cid
if self._data_operation.currentIndex() == 1:
spec_data = data.mean((1, 2))
elif self._data_operation.currentIndex() == 2:
spec_data = data.median((1, 2))
else:
spec_data = data.sum((1, 2))
spec_data = Spectrum1DRef(spec_data.data,
unit=spec_data.unit,
dispersion=data.spectral_axis.data,
dispersion_unit=data.spectral_axis.unit,
wcs=data.wcs,
name=layer.label)
spec_layer = Spectrum1DRefLayer.from_parent(spec_data)
# Store the relation between the component and the specviz data. If
# the data exists, first remove the component from specviz and then
# re-add it.
old_spec_layer = self._specviz_data_cache.get(layer)
self._specviz_data_cache[layer] = spec_layer
if old_spec_layer is None:
dispatch.on_add_to_window.emit(layer=spec_layer,
style={
'color': layer.style.rgba[:3],
'line_width': 3
},
vertical_line=True)
else:
dispatch.replace_layer.emit(old_layer=old_spec_layer,
new_layer=spec_layer,
style={
'color': layer.style.rgba[:3],
'line_width': 3
})
def _spectrum_from_component(self, layer, component, wcs, mask=None):
data = SpectralCube(component.data, wcs)
if mask is not None:
data = data.with_mask(mask)
spec_data = data.sum((1, 2))
spec_data = Spectrum1DRef(spec_data.data,
unit=spec_data.unit,
dispersion=data.spectral_axis.data,
dispersion_unit=data.spectral_axis.unit,
wcs=data.wcs)
# Store the relation between the component and the specviz data. If
# the data exists, first remove the component from specviz and then
# re-add it.
if layer in self._specviz_data_cache:
old_spec_data = self._specviz_data_cache[layer]
dispatch.on_remove_data.emit(old_spec_data)
self._specviz_data_cache[layer] = spec_data
dispatch.on_add_to_window.emit(spec_data)
class MultiResObs(object):
"""
Object to hold observations of the same object at different resolutions.
This is intended for matching and analyzing high-res (interferometric)
and low-res (single dish) observations.
"""
def __init__(self, highres, lowres):
super(MultiResObs, self).__init__()
self.highres = SpectralCube.read(highres)
self.lowres = SpectralCube.read(lowres)
self.highres_convolved = None
self.lowres_convolved = None
self.lowbeam = self.lowres.beam
self.highbeam = self.highres.beam
self.combined_beam = self.lowbeam.convolve(self.highbeam)
def apply_mask(self, highres_mask=None, lowres_mask=None):
'''
Apply a pre-made mask to either of the data cubes.
'''
if highres_mask is not None:
self.highres = self.highres.with_mask(highres_mask)
if lowres_mask is not None:
self.lowres = self.lowres.with_mask(lowres_mask)
def match_coords(self):
'''
Match the spatial and spectral coordinates of the cubes.
'''
# TODO: use Skycoords to convert extrema into the same frame
# Are either of the cubes in the correct frame?
if self.highres.header['CTYPE1'] != self.lowres.header['CTYPE1']:
raise TypeError("ctypes do not match. Are observations in the "
"same projection? Highres: " +
self.highres.header['CTYPE1'] +
" Lowres: " + self.lowres.header['CTYPE1'])
elif self.highres.header['CTYPE2'] != self.lowres.header['CTYPE2']:
raise TypeError("ctypes do not match. Are observations in the "
"same projection? Highres: " +
self.highres.header['CTYPE2'] +
" Lowres: " + self.lowres.header['CTYPE2'])
# Determine which cube should be slice down so both have the same
# spatial coverage.
limits_high = []
limits_low = []
low_long = self.lowres.longitude_extrema
high_long = self.highres.longitude_extrema
low_lat = self.lowres.latitude_extrema
high_lat = self.highres.latitude_extrema
if low_long[0] < high_long[0]:
limits_low.append(high_long[0])
limits_high.append("min")
else:
limits_high.append(low_long[0])
limits_low.append("min")
if low_long[1] > high_long[1]:
limits_low.append(high_long[1])
limits_high.append("max")
else:
limits_high.append(low_long[0])
limits_low.append("min")
if low_lat[1] > high_lat[1]:
limits_low.append(high_lat[1])
limits_high.append("min")
else:
limits_high.append(low_lat[0])
limits_low.append("min")
if low_lat[0] < high_lat[0]:
limits_low.append(high_lat[0])
limits_high.append("max")
else:
limits_high.append(low_lat[0])
limits_low.append("max")
# Apply common slicing to both
self.highres = \
self.highres.subcube(xlo=limits_high[0], xhi=limits_high[1],
ylo=limits_high[2], yhi=limits_high[3])
self.lowres = \
self.lowres.subcube(xlo=limits_low[0], xhi=limits_low[1],
ylo=limits_low[2], yhi=limits_low[3])
# Now match the spectral extends
low_spec = \
self.highres.spectral_extrema[0] if \
self.highres.spectral_extrema[0] > self.lowres.spectral_extrema[0]\
else self.lowres.spectral_extrema[0]
high_spec = \
self.highres.spectral_extrema[1] if \
self.highres.spectral_extrema[1] < self.lowres.spectral_extrema[1]\
else self.lowres.spectral_extrema[1]
self.highres = self.highres.spectral_slab(low_spec, high_spec)
self.lowres = self.lowres.spectral_slab(low_spec, high_spec)
def convert_to(self, unit=u.K, freq=1420.40575177*u.MHz):
'''
Convert both sets to common brightness units.
'''
convert_high = unit != self.highres.unit
convert_low = unit != self.lowres.unit
high_unit = self.highres.unit
low_unit = self.lowres.unit
if unit == u.K:
if convert_high and 'Jy' in high_unit.name:
self.highres = \
self.highres.to(unit, self.highbeam.jtok_equiv(freq))
if convert_low and 'Jy' in low_unit.name:
self.lowres = \
self.lowres.to(unit, self.lowbeam.jtok_equiv(freq))
else:
raise NotImplementedError("Only supporting Jy/beam -> K right now.")
def convolve_to_common(self, verbose=False, use_dask=True,
block=(256, 256)):
'''
Convolve cubes to a common resolution using the combined beam.
'''
# Create convolution kernels from the combined beam
conv_kernel_high = \
self.combined_beam.as_kernel(wcs_to_platescale(self.highres.wcs))
# if use_dask:
# assert np.alltrue([bl > kern for bl, kern in
# zip(block, conv_kernel_high.shape)])
conv_kernel_low = \
self.combined_beam.as_kernel(wcs_to_platescale(self.lowres.wcs))
# if use_dask:
# assert np.alltrue([bl > kern for bl, kern in
# zip(block, conv_kernel_low.shape)])
high_pad = np.ceil(conv_kernel_high.shape[0] / 2).astype(int)
highres_convolved = np.empty(self.highres.shape)
high_chans = len(self.highres.spectral_axis)
if verbose:
print("Convolving high resolution cube.")
if use_dask:
highres_convolved = auto_dask_map(self.highres, blocks=block,
args=[conv_kernel_high])
# for chan in range(high_chans):
# if verbose:
# print("On Channel: "+str(chan)+" of "+str(high_chans))
# if use_dask:
# da_arr = \
# da.from_array(np.pad(self.highres.filled_data[chan, :, :],
# high_pad, padwithnans),
# chunks=block)
# highres_convolved[chan, high_pad:-high_pad,
# high_pad:-high_pad] = \
# da_arr.map_overlap(
# lambda a:
# convolve_fft(a,
# conv_kernel_high,
# boundary='fill',
# interpolate_nan=True,
# normalize_kernel=True),
# depth=2*high_pad,
# boundary=np.nan).compute()
# else:
# highres_convolved[chan, :, :] = \
# convolve_fft(self.highres.filled_data[chan, :, :],
# conv_kernel_high, boundary='fill',
# interpolate_nan=True, normalize_kernel=True)
update_high_hdr = \
_update_beam_in_hdr(self.highres.header, self.combined_beam)
self.highres_convolved = \
SpectralCube(highres_convolved*self.highres.unit, self.highres.wcs,
header=update_high_hdr)
# Cleanup a bit
del highres_convolved
# Now the low resolution data
lowres_convolved = np.empty(self.lowres.shape)
low_chans = len(self.lowres.spectral_axis)
if verbose:
print("Convolving low resolution cube.")
if use_dask:
lowres_convolved = auto_dask_map(self.highres, blocks=block,
args=[conv_kernel_low])
# for chan in range(low_chans):
# if verbose:
# print("On Channel: "+str(chan)+" of "+str(low_chans))
# lowres_convolved[chan, :, :] = \
# convolve_fft(self.lowres.filled_data[chan, :, :],
# conv_kernel_low, boundary='fill',
# interpolate_nan=True, normalize_kernel=True)
update_low_hdr = \
_update_beam_in_hdr(self.lowres.header, self.combined_beam)
self.lowres_convolved = \
SpectralCube(lowres_convolved*self.lowres.unit, self.lowres.wcs,
header=update_low_hdr)
def flux_recovered(self, plot=True, filename=None, enable_interp=True,
interp_to='lower', diff_tol=1e-4*u.m/u.s):
'''
Check the amount of flux recovered in the high resolution image versus
the low resolution data. If the spectral axes don't match, one is
interpolated onto the other.
Parameters
----------
plot : bool, optional
Enable plotting.
filename : str, optional
Give filename for the plot save file. When specified, the plot
is automatically saved.
interp_to : 'lower' or 'upper', optional
If the spectral axes don't match, interpolated onto the same.
The default 'lower' interpolates to the spectral axis with the
lowest resolution.
'''
# Add up the total intensity in the cubes and compare
# Assumes that there is some masking such that noise
# doesn't dominate
if self.highres_convolved is not None:
high_channel_intensity = \
self.highres_convolved.sum(axis=(1, 2))
else:
high_channel_intensity = self.highres.sum(axis=(1, 2))
Warning("Should run convolve_to_common before. Using unconvolved"
" cube.")
if self.lowres_convolved is not None:
low_channel_intensity = self.lowres_convolved.sum(axis=(1, 2))
else:
low_channel_intensity = self.lowres.sum(axis=(1, 2))
Warning("Should run convolve_to_common before. Using unconvolved"
" cube.")
# If the spectral axes are not the same, try interpolating to match
if enable_interp:
better_vres_high = \
np.abs(self.lowres.header['CDELT3']) > \
np.abs(self.highres.header['CDELT3'])
spec_diff = np.abs(np.abs(self.lowres.header['CDELT3']) -
np.abs(self.highres.header['CDELT3']))
spec_unit = self.highres.spectral_axis.unit
if spec_diff * spec_unit < diff_tol:
warnings.warn("Channels are essentially the same. "
"Skipping interpolation. Lower diff_tol"
" to re-enable interpolation.")
else:
# Invert which to interpolate to
if interp_to == "higher":
better_vres_high = not better_vres_high
if better_vres_high:
f = interp1d(np.round(self.highres.spectral_axis.value, 3),
high_channel_intensity.value)
high_channel_intensity = \
f(np.round(self.lowres.spectral_axis.value, 3)) *\
self.highres.unit
else:
f = interp1d(np.round(self.lowres.spectral_axis.value, 3),
low_channel_intensity.value)
low_channel_intensity = \
f(np.round(self.highres.spectral_axis.value, 3)) *\
self.lowres.unit
self.fraction_flux_recovered = \
high_channel_intensity.sum()/low_channel_intensity.sum()
if plot:
p.plot(self.highres.spectral_axis.value,
high_channel_intensity.value)
p.plot(self.lowres.spectral_axis.value,
low_channel_intensity.value)
p.xlabel("Spectral Axis ("+self.highres.spectral_axis.unit.to_string()+")")
p.ylabel("Intensity ("+self.highres.unit.to_string()+")")
if filename is None:
p.show()
else:
p.savefig(filename)
def run_all(self, highres_mask=None, lowres_mask=None, plot=True,
unit=u.K, freq=1420.40575177*u.MHz, filename=True,
use_dask=False, verbose=True):
'''
Run the complete comparisons.
'''
self.apply_mask(highres_mask=highres_mask, lowres_mask=lowres_mask)
self.match_coords()
self.convert_to(unit=unit, freq=freq)
self.convolve_to_common(verbose=verbose, use_dask=use_dask)
self.flux_recovered(plot=plot, filename=filename)
denscube = SpectralCube.read(paths.dpath("H2CO_ParameterFits_{0}dens.fits".format(meastype)))
denscube = denscube.with_mask(okmask)
colcube = SpectralCube.read(paths.dpath("H2CO_ParameterFits_{0}col.fits".format(meastype)))
colcube = colcube.with_mask(okmask)
goodmask_std = stdcube < 0.5
flathead = denscube.wcs.dropaxis(2).to_header()
denscube_lin = SpectralCube(10**denscube.filled_data[:].value,
wcs=denscube.wcs,
mask=okmask)
colcube_lin = SpectralCube(10**colcube.filled_data[:].value,
wcs=denscube.wcs, mask=okmask)
totalcol = colcube_lin.sum(axis=0)
totalcolgood = colcube_lin.with_mask(goodmask).sum(axis=0)
totalcolgoodstd = colcube_lin.with_mask(goodmask_std).sum(axis=0)
denscol = SpectralCube(denscube_lin.filled_data[:] * colcube_lin.filled_data[:], wcs=denscube.wcs,
mask=okmask)
wtdmeandens = np.log10(denscol.sum(axis=0) / totalcol)
mindens_std = denscube.with_mask(goodmask_std).min(axis=0)
mindens_chi2 = denscube.with_mask(goodmask).min(axis=0)
hdu1 = fits.PrimaryHDU(data=wtdmeandens.value,
header=flathead)
hdu1.writeto(paths.dpath("H2CO_ParameterFits_weighted_mean_{0}_density.fits".format(meastype)), clobber=True)
masked_wtdmeans = np.log10(denscol.with_mask(goodmask).sum(axis=0) / totalcolgood)
