SExtractor Testing (Fall 2006)

T. Evans 2/14/2007

Method

I tested SExtractor on two patch maps in the 353 frequency band, with centers at (0,90) and (30,0) (ecliptic coordinates), 10 degrees and 400 pixels on a side (pixle size = 1.5 arcmin). No optimization of the SExtractor processing was done; I simply set some parameters by guesswork. In the "configuration" file, I changed the following parameters:

PHOT_APERTURES = 3, 5, 7 -- the aperture diameters in pixels, which = 4.5, 7.5, 10.5 arcmin
PIXEL_SCALE = 0 -- use FITS WCS information in the FITS header
SEEING_FWHM = 180. -- stellar FWHM in arcsec (= 3 arcmin or 2 pixels)

I also chose the following output parameters or columns, though not all were actually useful:

NUMBER
FLUX_ISO
FLUXERR_ISO
FLUX_APER(1)
FLUXERR_APER(1)
FLUX_APER(2)
FLUXERR_APER(2)
FLUX_APER(3)
FLUXERR_APER(3)
FLUX_AUTO
FLUXERR_AUTO
KRON_RADIUS
BACKGROUND

The output, therefore, contains fluxes and flux errors for 3 apertures, an isophotal flux and error, and the "modified Kron" flux, flux error, and radius, as well as the background value.

SExtractor Results

SExtractor found 101 sources in the ecliptic pole map and 110 sources in the ecliptic plane map. Since there are only a few input point sources in each patch map (see the next section), it is apparent that most of these sources are part of the large scale background structure in the maps.

Simulation Input Point Sources

There are 3 simulation input point sources found in the area covered by the pole patch map, two of which may be the same source, and there are also 3 input sources in the area covered by the plane patch map, two of which also may be the same source.

Simulation input point sources in pole patch map area

pixel number	ecliptic lon.	ecliptic lat.	galactic lon.	galactic lat.	flux
13506584	293.96819	84.83357	91.05500	27.60800	223.46873
12335344	103.95515	86.28737	100.57288	30.63947	467.83316
12335344	103.97776	86.28734	100.57257	30.64091	502.07602

The input fluxes of the two sources in the same HEALPIX pixel differ by only 7%, so they may be the same source.

Simulation input point sources in plane patch map area

pixel number	ecliptic lon.	ecliptic lat.	galactic lon.	galactic lat.	flux
43567256	34.78564	-2.18395	153.02826	-46.99007	328.85400
43669620	28.80864	2.77807	141.60574	-47.34600	987.18939
43669620	28.80895	2.77744	141.60667	-47.34631	805.90100

In this case, the input fluxes of the two sources in the same HEALPIX pixel differ by 19%. They may still be the same source, but it seems less likely.

Source Comparison

I overlaid the SExtractor sources in red and the simulation intput point sources in green on top of the pole map:

There are two SExtractor sources that line up with the input point sources (in green):

ID	SExt. ecl. lon	SExt. ecl. lat	ap 1 flux	ap 2 flux	ap 1 flux err	ap 2 flux err	Input ecl. lon	Input ecl. lat	Input flux
79	294.17662	+84.84123	157.8	346.2	23.9	39.8	293.96819	84.83357	223.46873
42	104.13945	+86.31014	560.6	1220.4	23.9	39.8	103.95515	86.28737	467.83316
42	104.13945	+86.31014	560.6	1220.4	23.9	39.8	103.97776	86.28734	502.07602

The flux errors happen to be the same for both sources, and all of the values found for this map are in the same ranges. The input flux for the first source is about halfway between the aperture 1 and aperture 2 fluxes found by SExtractor, so perhaps an aperture of 4 pixels would have been a better choice.

Unsurprisingly, SExtractor was unable to separate the second and third input point sources, if they actually are separate sources. And if they are separate sources, the input fluxes should be additive, and the input combined flux should be again about halfway between the aperture 1 and aperture 2 fluxes. If the input sources are actually the same source, the SExtractor aperture fluxes are both too large.

I also did the same exercise for the plane map, with the SExtractor sources in red and the simulation intput point sources in green:

In this patch map, there is only one SExtractor source that lines up with the input point sources (in green):

ID	SExt. ecl. lon	SExt. ecl. lat	ap 1 flux	ap 2 flux	ap 1 flux err	ap 2 flux err	Input ecl. lon	Input ecl. lat	Input flux
--	--	--	--	--	--	--	34.78564	-2.18395	328.85400
25	28.83336	+2.76220	933.3	1921.4	39.0	65.6	28.80864	2.77807	987.18939
25	28.83336	+2.76220	933.3	1921.4	39.0	65.6	28.80895	2.77744	805.90100

I'm not sure why the first input source was not extracted by SExtractor (the green box at the far left of the image, approx. 2/3 down), but it is located in the background structure so perhaps that caused problems.

Again, SExtractor was unable to resolve the second and third input point sources, if they actually are separate sources. And if they are separate sources, the input combined flux should be again about halfway between the aperture 1 and aperture 2 fluxes. If the input sources are actually the same source, the SExtractor aperture 2 flux is too large.

Conclusions

Although we have small number statistics here, it seems that SExtractor is able to find most of the simulation input point sources in the patch maps. The one missing source is probably a result of confusion with the background structure. SExtractor won't always be able to resolve input sources that are close together, especially close sources found in the same HEALPIX pixel, but this is a common problem and very difficult to solve. The best aperture size is probably close to 4 pixels in diameter, or 6 arcmin. Neither the isophotal nor Kron fluxes seemed to correspond very well to the input source fluxes.