Catalog production

Production of the LAT source catalog

J. Ballet

Since the splinter meeting at Stanford during the consortium meeting in September, we have concentrated on building a pipeline prototype. Because this is the most time consuming step, we have started with the likelihood application. Ideas:

build a script which would be both usable by a task scheduler AND could be run by any user as an ordinary application (for a few sources and ROIs).
blocks for the scheduler should be as big as possible while preserving parallelism (rely as little as possible on the scheduler)

Jim Chiang has written a Python script which chains the calls necessary to the likelihood application. It is called source_analysis.py and is distributed as part of the sane package. It looks like this. The names of the input files are stored in the object passed to the 2^nd step. Only two blocks for the scheduler: preparation (build ROIs, exposure cube) and execution (build source model, build exposure map and call likelihood for a particular ROI). The latter may be run in parallel over many processors. The diffuse emission modeling is treated by a number of image templates (currently two for extragalactic and Galactic).

We have done a number of tests to investigate how likelihood behaves when presented with many weak sources. This is what will happen if we set the threshold of the source detection step low enough to let likelihood decide whether a source is significant or not. It has trouble converging when presented with all the sources at once. A more robust way is to introduce the diffuse emission first, then the bright sources, then the weak ones. A source outside the ROI (but close enough to contribute) should be fixed to the parameters deduced from analyzing another ROI (containing that source).

We have implemented the initial pipeline step, building a number of images and associated exposure maps covering the entire sky in several energy bands. This requires upgrading for use with the new task names and interfaces.

The source detection algorithms have evolved a little. The optimal filter method is now able to provide a significance image of the entire sky in each energy band. This is useful to combine energy bands and improve detection power.

We have installed OPUS 5.4b. Applications are launched as soon as a trigger file exists (and depending on processor availability). We have tested it on two computers sharing data via AFS. Here is the wrapping for the likelihood calls.

The implementation of the catalogAccess (U9) library has moved forward. Remaining to be done are the direct access to the VizieR server, and saving the result to a binary FITS file. Dirk Petry has defined the list of catalogs which will be supported by the catalogAccess library for DC2.

In parallel, Jürgen Knödlseder at CESR has started implementing the sourceIdentify (A2) utility for inclusion in the ScienceTools. The idea is first write a simple general building block which finds spatial coincidences between two catalogs (will exist for DC2), then introduce astrophysical criteria (hardness ratios, variability, ...) at higher level. The first version of that building block exists (and should be checked in shortly).

Seth Digel is leading the discussion on the catalog format (which obviously impacts in its contents). We welcome all input to be sure that we don't leave aside important aspects unwillingly.

For DC2, we aim at a working pipeline for building the point source catalog over one time interval. Nothing fancy on detection of variable sources. We plan to produce the source catalog during the course of the DC2 exercise and present the results at the close-out meeting.

Next step: Introduce one all-sky source detection method (MR_filter) into the pipeline.

Open issue: The diffuse model will be represented in the future by a data cube (RA, Dec, Energy). We will then need an utility to build the diffuse emission map in a given energy band, convolving with the PSF in each energy channel, multiplying by the effective area, and summing the relevant channels. This is needed for the all-sky source search.

Information: We organize a 2^nd meeting on the twin issues of building the catalog and modeling the diffuse emission. The date will be confirmed shortly.