In response to the needs of several astrophysics projects hosted at NERSC, I have begun to create an all-sky digital image based upon the point-and-stare observations taken via the Palomar-QUEST Consortium and the SN Factory + Near Earth Asteroid Team.
This data spans 9 years and almost 20,000 square degrees, with typically 10-100 pointing on a particular part of the sky. In total there are 11 million images. Here is a fits file that contains the number of pointings on a map of the sky. Above you can see the generic layout. The graph goes from about -30 to +60 in declination and from 0-24 hours in right ascension. The grey-white coloring represents about 30 pointings, blue on the order of 70, green about 100 and red approximately 200. At the latter depths we typically achieve 24th magnitude in R-band at S/N = 3. The entire dataset is 90 TB and will create both a temporal and static catalog of astrophysical objects. I was in the unique position of utilizing my expertise in astrophysics imaging and linear algebra (to determine the relative scaling of all the images which overlap with each other) coupled with my parallel processing knowledge to work this through on the NERSC machines. This work has attracted the attention of many groups and will form a very useful dataset for the entire astrophysical community.
Example: Coma Cluster
This project started out as an effort to make a much better reference images from which to make our subtractions in the SN Factory. We would like to use all the images we have acquired over the past several years and co-add them into one deep image.
We have covered ~20,000 sq deg on the sky about 50 separate times. Given the size of our imagers we have approximately 9 million, 6 Mb images. All of this resides now on NERSC Global Filesystem.
We process the images in 2 X 2 square degree chunks. Half of this is a serial task and half is a parallel task. Each part takes roughly the same amount of time ~ 100 wallclock days. Serial: Determine image zeropoint (relative depth with respect to all other images in the co-add) and astrometry (physical scale of the image and where it is on the sky). Parallel: Actual co-addition of 500-1000 images into one 250 Mb image of 2 X 2 square degrees.
To handle the serial part in parallel I wrote mpibatch, a script to perform a round-robin execution of a list of serial commands (or even scripts of several commands) through mpi. It uses a perl script, mknamelist.pl, to convert text list into fortran namelist. The code then reads in the fortran namelist as character strings on the master processor, which spits it out to the slaves to execute, who then ask for more when they are done. A tarball of this code can be found here. Then the threaded code swarp is run to co-add each of the sub-images into one large image.
To the right is a pretty example of the Coma Cluster, which was made by combining over 500 images from 2001-2007. The full fits file can be found here.
Example: High Proper Motion Stars
One of the interesting objects that we often mistake for supernovae are high proper motion stars. Here we have one (labeled by the SN Factory as SNF20061011-008) which moved quite a bit since our reference images in 2001 and 2007. This image sequence to the right is a sum of the images obtained in each year (2001, 2002, 2003, 2005, 2006 & 2007) of this target located at: RA = 46.50662, Dec = -3.51532 (J2000). Over this time it has moved about 8 arcsec.
We have many examples of such stars that make it through our pipeline. We are now considering alternative methods which can exploit this dataset to cull more of these objects out in a separate pipeline. Interested parties can contact me if they would like to work on this program.
New Science: Pre-Imaging of GRBs & SNe
Often it is the case where the GRB afterglow or supernova far overwhelms any sign of its host galaxy. In this case it is necessary to look at pre-imaging of the source. People usually go to the Digital Sky Survey however, their limiting magnitude is typically not that deep (R~20 at best). With our typical coverage we can usually go 2-3 magnitudes deeper on any part of the northern sky. Here are some examples of this work.
GCN on GRB070809.
GCN on GRB070714b.
FITS file for SNF20070825-001. ATEL on SNF20070825-001.
GCN on GRB071025.
GCN on GRB071112C.
FITS file for SNATEL1615. ATEL on Coma SN.