This package is intended to provide scripts to glue the different pieces of the LSST difference imaging code into a single pipeline that can be used for DESC data challenges. This code has been tested with the LSST stack version w.2018.39. It is not compatible with previous releases and has not been tested with later ones. The default configuration is designed to work with Data Challenge 1.2 data.

This package requires additional dataset definitions in the obs_lsstCam mapper. These are currently availble on the dc2/run1.2i_diff of obs_lsstCam.

The baseline pipeline includes three steps:

• Difference image creation at the visit level that produces DIASources and difference images
• Association of the DIASources into DIAObjects
• Running forced photometry at the positions of the DIAObjects on all visits

It is assumed that you have a repo where the data has been processed all the way through the coadd and multiBandDriver stages. Here is the command to create difference images for a set of visits:

imageDifferenceDriver.py [repo] --rerun [rerun]  --id visit=[visit list]  -C dia_pipe/config/ImageDifferenceDriver.py --cores [cores]

This is a driver script that inherits from ctrl_pool so it can be submitted to a batch system or run locally with all the standard commands.

For each visit given, it will find the corresponding overlapping coadd tracts/patches. It is currently configured to use the science images or deepCoadd_calexp data products because that is what is available. This can be changed once there are other coadds produced. The coadd is then psf-matched to the visit and subtracted. The default algorithm is Alard-Lupton, but the Zogy method is also availible. More information on the algorithms and settings can be found ??.

The ouptut of this stage is a DIASource catalog and difference image for every CCD.

The association is done separately for regions on the sky. For each coadd patch we get all the overlapping DIASources (within the inner boundary of the tract/patch so there are no duplicate objects in different catalogs) and associate them into DIAObjects based soley on position. The association command is:

associationDriver.py [repo] --rerun [rerun] --id tract=[tract] filter=u^g^r^i^z^y --cores [cores]

It will fetch DIASources only from the filters explicitly given. This is also a batch task that will parallelize over the number of patches in the tract. Therefore, it doesn't make sense to give more cores than the number of patches.

There are currently two association algorithms:

• One based on the MultiMatch algorithm in the LSST stack. This requires loading all objects into memory before matching and is not meant to handle large number of objects.
• A simple matching algorithm that keeps a running list of DIAObjects and adds to the list for each new DIASource catalog. This should scale much better than the MultiMatch approach and is currently the default.

It should be fairly simple to add additional association algorithms.

By default the list of visits is chosen by using those that were used to construct the coadd image. If you want to use a different set of visits you can can append the option:

--selectId visit=12345^12346^12347

We can do forced photometry at the positions of the DIAObjects for all the visits:

forcedPhotCcdDiaDriver.py [repo] --rerun [rerun] --id visit=[visit list] --cores [cores]

This should produce an output for every visit.

There are three other forced photometry algorithms. The first two will produce an output for each visit.

1. Forced photometry on difference images at the position of the template catalogs. This will get all template objects from the band the exposure was observed in and within the sky bounding box and perform forced photometry. Objects that have the position of the parent object outside of the bounding box will be removed from the catalog. This can be called by:

   forcedPhotCcdTemplateDia [repo] --rerun [rerun] --id visit=[visit list]

   you can also run this via the forcedPhotCcdDiaDriver.py script by setting a configuration file:

   from lsst.dia.pipe.forcedPhotCcdTemplateDia import ForcedPhotCcdTemplateDiaTask
   config.forced.retarget(ForcedPhotCcdTemplateDiaTask)

2. Forced photometry on the template images at the position of the diaSrc catalog. This will take the diaSrc catalog and find all template images that contain these objects and perform forced photometry. A detect_isPrimary flag is added to select unique objects because there will be overlap between different tracts/patches. This can be called by:

   forcedPhotCoaddDiaSrc [repo] --rerun [rerun] --id visit=[visit list]

   this can also be run via the forcedPhotCcdDiaDriver.py script by setting a configuration file:

   from lsst.dia.pipe.forcedPhotCoaddDiaSrc import ForcedPhotCoaddDiaSrcTask
   config.forced.retarget(ForcedPhotCoaddDiaSrcTask)

The other forced photometry algorithm performs forced photometry on the template images at the position of the diaObject catalog. This is run on tracts and patches and has a different structure than the previous two scripts:

forcedPhotCoaddDia.py [repo] --rerun [rerun] --id tract=[tract] patch=[patch] filter=[filter]

It has it's own driver script that can be run by:

forcedPhotCoaddDiaDriver.py [repo] --rerun [rerun] --id tract=[tract] patch=[patch] filter=[filter] --cores [cores]

This package adds a selector to limit the input visits based on seeing and time that are fed into coaddDriver.py. Here is an example config file:

from lsst.dia.pipe.selectImages import TimeBestSeeingWcsSelectImagesTask
config.select.retarget(TimeBestSeeingWcsSelectImagesTask)
config.select.minMJD = 59580.0
config.select.maxMJD = 60310.0
config.select.minPsfFwhm = 2.5
config.select.maxPsfFwhm = 3.5

The FWHM values are specified in pixels.

By default the LSST difference imaging software will only get the template from a single tract. If the overlap area between tracts is more than the size of a CCD then this is not an issue, however if the overlap area is small then some CCDs will not have complete template coverage. 'dia_pipe' includes an alternate template routine that will get all overlapping tracts for a given CCD. In the region of overlap the template (and PSF) will be an average of the overlaping tracts. To enable this algorithm with imageDifferenceDriver:

from lsst.dia.pipe.getMultiTractTemplate import GetCoaddAsMultiTractTemplateTask
config.imageDifference.getTemplate.retarget(GetCoaddAsMultiTractTemplateTask)