Identification of Ext. Sources
Associated with Meteor Streaks

T. Jarrett, IPAC
(990314)
** updated (990315)

Meteor streaks typically induce several false "extended" sources along its trajectory (and even more point sources!); as such, they are a major hindrence toward producing a reliable and clean catalog.

Roc and I have developed an automated streak finder. It uses, as a seed list, point sources with only 1/6 detections (or apparitions; called the n(m) parameter). Since streaks are rapid transients, they should induce 1/6 point sources for the most part (there can of course be other combos, such as 1/5 or 1/4, or possibly, 2/6, 2/5, etc; we have yet to investigate these scenerios). The seed list is fed into my linear "feature" finder. An additional (but even more crucial) seed list uses extended sources with highly inclined orientations (axis ratio < 0.6). Meteor streaks induce false "edge-on" galaxies (of which the real variety are somewhat rare) which align along the streak.

The algor and input/output are described in the SIS str02. A slightly different method toward finding streaks is described here, SIS str01. The results below use the former method.

Using the old 2MASS working database, Roc generated the seed point source list. This seed list covers a larger portion of data (May 1997 to May 1998) than the current "release" database. The linear transient finder was run on this list and a list of extended sources with small axis ratios. It generated the following table of candidate meteor streaks, most of which have been verified as real meteor streaks (see Linda Fullmer's report on a slightly larger sample).

The "severity" of the streak is correlated with the "nhit" parameter (essentially measuring how many 1/6 point sources lie upon the streak) and the "ngal" parameter. The coordinates and position angle of the streak are given by the "ramean,decmean" and "p.a." parmeters, respectively.

The "length" parameter may not give the true length of the streak since the seed list itself is not necessarily complete. The conservative approach is to assume that the streak extends across the coadd. Finally, the "gflag" parameters refers to the presence of a large (known) galaxy nearby. Large galaxies can "look" like streaks since they can be long and linear, and they can also induce 1/6 pt sources (thanks to the copious amount of low surface brightness emission -- susceptible to noise bumps).


Now, finding point or extended sources associated with meteor streaks requires the input candidate meteor streaks and the point or extended source catalogs. For the source catalogs, the required information is: cntr (database ID#), ra, dec, date, scan and stellar source density. Optional information includes the source mags and classification scores (extended sources only). The basic algor is to search around every source in the catalog for linear streaks. If the sources lies within or very close to a streak, it is flagged as such: The proximity parameter is an input parameter (set within the namelist file).

Here I have used as the input catalog the extended source catalog of the first incremental data release (covering dates between june 1997 and Jan 1998). See Census of Extended Sources in the First 2MASS Working Release DB for juicy details. 85 total extended sources in the catalog are identified as pieces of meteor streaks. Visual inspection of each source verifies that all are indeed pieces of streaks.

Verified (and nasty) meteor streaks are found in scans 107 & 109 (07/13/1997), 051 (09/23/1997), and 081 (11/02/1997), 103 (06/22/1997), 051 (07/11/1997), 042 & 091 (10/20/1997), 050 & 056 (11/19/1997), 050 (11/28/1997), 126 (12/17/1997), 082 (01/22/1998), 050 (01/22/1998).

Final caveat: the streak finder appears to be working like a champ. However, what is not obvious is how complete a list of streaks it can generate. What it finds appears to be bright linear streaks. There are surely fainter streaks that are not showing up. The question then is how many false sources are induced by these nefarious meteor streaks?

Related Links: