1 GAUSSCLUMPS Decompose a 3-dimensional spectral line data set into Gaussian shaped clumps
The GAUSSCLUMPS algorithm was written to iteratively decompose a
3-dimensional data cube of two spatial coordinates and one spectral co-
ordinate into a series of Gaussian shaped clumps. GAUSSCLUMPS starts by
fitting a 3-dimensional clump locally to the maximum of the input lbv
cube. It then subtracts this clump from the cube, creating a residual
map, and then continues with the maximum of this residual map.
The procedure is repeated until a stop criterion is met, for instance
when the maximum of the residual maps drops below the 3 sigma level.
When fitting a Gaussian to the maximum, a modified chi-squared function
is minimized. Three "stiffness" parameters control the fitting, ensuring
that a local clump is fitted and subtracted.
GAUSSCLUMPS was developed by Stutzki & Guesten (1990, ApJ, 356, 513) who
applied it to their M17SW data sets. It was lateron described and ana-
lyzed in more detail by Kramer et al. (1998, A&A, 329, 249) who also ap-
plied it to data cubes of various Galactic clouds. The papers of e.g.
Heithausen et al. (1998) and Simon et al. (2000) give more examples of
the application of gaussclumps to large scale CO maps.
The general recommendation is to make sure that the data is fully sam-
pled. For the "stiffness" parameters, use the initial guess of 1,1,1
which usually works out well. It helps a lot if the emission drops to
near zero at the map edges. Otherwise GAUSSCLUMPS may find clumps which
are partly outside the mapped region. Make sure that GAUSSCLUMPS does
not decompose noise into clumps, for instance by setting a 3sigma
threshold. GAUSSCLUMPS does always create clumps which are larger than
the resolutions. However, the intrinsic, deconvolved FWHMs may become
very small. It is therefore recommended to select the clumps which are
intrinsically larger than a certain fraction of the resolution (e.g.
50%), from the output files, for further analysis and discard the rest
which anyway only contains only a tiny fraction of the total mass.
Gaussclumps has also been applied to dust continuum maps, i.e. maps of
only two dimensions by e.g. Mookerjea, Kramer et al. (2004, A&A, 426,
119). The input data were prepared by adding two adjacent, empty "ve-
locity" planes to the original 2d data cube. This simple trick "fools"
the algorithm to run without problems.
The following files are created by gaussclumps: gaussclumps3d.out show-
ing the details of the decomposition into clumps step by step and gauss-
clumps3d.list which lists the parameters of all clumps found, line by
line.
Blanked pixels are ignored. The blanking value and tolerance are defined
in the image header.
For more details, please contact Juergen Stutzki (stutzki@ph1.uni-
koeln.de) or Carsten Kramer (kramer@iram.es).
The meaning of some parameters is evident. Here, we offer some explana-
tion for the not so obvious ones. The distributed gaussclumps.init con-
tains reasonable starting values.
2 INP_IMAGE$
TASK\CHARACTER "input file" INP_IMAGE$
The input cube should be ordered in lbv
2 RMS$
TASK\CHARACTER "rms [K]" RMS$
rms of the input cube
2 FWHM_BEAM$
TASK\REAL"beam FWHM [m] or [s]" FWHM_BEAM$
Angular resolution of the input cube in arcmin or arcsec, e.g. 13S
2 FWHM_START$
TASK\INTEGER "initial guess [beam FWHM]" FWHM_START$
Initial guess of the FWHM of the Gaussian clump to be fitted in units of
the input cube resolution, e.g. 1.5.
2 STIFFNESS$
TASK\INTEGER "stiffness S0[3]" STIFFNESS$
Stiffness parameters control the fitting of the Gaussian clump to the
maximum of the map.
S0[1] biases the fit to stay below the observed intensity but stay near
the observed intensities. S0[2] biases the fitted amplitude to stay
close to the observed maximum. S0[3] biases the fitted position to stay
near the position of the observed maximum. S0=1,1,1 are good starting
values and are often sufficient.
2 THRESHOLD$
TASK\INTEGER "threshold [K]" THRESHOLD$
Stop iterating when the maximum of the residual map drops below the
threshold.
2 MAXCLUMPS$
TASK\INTEGER "max number of clumps" MAXCLUMPS$
Stop iterating when the maximm of number of clumps is exceeded.
2 CONTRAST$
TASK\INTEGER "contrast [fraction of maximum]" CONTRAST$
To be written by C.Kramer or J.Stutzki.
2 MININTEGRAL$
TASK\INTEGER "minimum integral [fration of total]" MININTEGRAL$
Stop iterating when the integral of the residual map drops below a frac-
tion of the integral of the original input map.
2 APERTURE_FWHM$
TASK\INTEGER "aperture FWHM [initial guess]" APERTURE_FWHM$
The width of the Gaussian function weighting the observed points in the
vicinity of the maximum biases the fit to find individual clumps rather
than an averaged, smeared-out, broad clump. A small width introduces a
bias to fit small-scale structure. The width is given in units of the
resolution of the input map.
2 APERTURE_LMTS$
TASK\INTEGER "aperture cutoff [initial guess]" APERTURE_LMTS$
To speed-up computation, no data outside the aperture cutoff is consid-
ered for the fit.
1 ENDOFHELP