Spatial Analysis

New parameter EXCLUDE_SIGMA in ML source detection

Exclusion of close-by sources in source detection

Concerning source detection: DETECT/MAXLIK, COMPUTE/UPPER, DETECT/SOURCES

Identification and removal of neighbouring sources

Close-by other sources within the cut radius can significantly affect the results of source detection. Therefore the maximum-likelihood source detection tries to identify these sources and excludes them from the fit.

The identification and removal of potential other sources in the extraction radius (= cut radius) goes like this:

• select a ring with inner radius = 0.4 * cut radius, outer radius = 1.0 * cut radius.
• divide this ring into 8 sectors
• integrate counts in each sector
• determine average
• if (sectorcounts(i) - average) > threshold * sigma(average) then exclude sector(i)
• exclude also adjacent sector(i-1) and sector(i+1)
• delete all photons from the excluded sectors
• determine remaining area, expressed as fraction of the cut radius circle (this fraction is stored in the solst.tbl in column :FA).

The threshold for exclusion of a sector was hardcoded to 5 (sigma) in EXSAS versions older than JAN96. In newer versions this parameter is available in the parameterfiles for DETECT/MAXLIK (sppar.par) and COMPUTE/UPPER (ulpar.par) as parameter EXCLUDE_SIGMA. This parameter is included in the extended parameter set. To access it, use CREATE/PARFILE with the [expert] option.

What can go wrong?

It has been found in rare occasions, that the initial source positions as supplied by the merged list (mplst) or the user-list are inaccurate. In this cases the actual source position is offset with respect to the cut radius (extraction) circle. This in turn can trigger the exclusion mechanism to erroneously identify and exclude some sectors containing excess photons of the source itself as sectors with close-by other sources. This happens preferrably to strong sources.

Those sources will be truncated and the following ML-fit will yield source counts which are too low. In extreme cases, the source to be fitted will be excluded almost totally from the fit, resulting in ML fluxes too low by orders of magnitude.

It should be noted, that this effect has been observed only for a negligible fraction of a large sample of tested sources. Usually the input positions from the LDETECT/MDETECT fits are precise enough to avoid significant decentering of the cut radii. Care should be taken if the user supplies an input list with source positions. It should be carefully checked that those positions are consistent with the actual source positions within the FWHM.

If you suspect that one of your sources may be affected by this self-excluding, check column :FA in your output SOLST. If it is not 1 then some sectors were excluded. If FA<1,it is recommended that you repeat the ML detection with parameter EXCLUDE_SIGMA set to a higher value and check the results against those obtained before.

S. Döbereiner
send e-mail to xray-info @ mpe . mpg . de

© X-Ray Group at MPE (group)
last update:1998/12/07 12:39:37, editor of this page:Stefan Döbereiner