Montage Software Fuels Astronomical Advances

With a growing international user base, the Montage image mosaic software has become a valuable tool in the astronomer’s arsenal. “It is especially good at generating products used to make discoveries,” said Bruce Berriman, Project Manager of the NASA Infrared Processing and Analysis Center (IPAC) Infrared Science Archive at the California Institute of Technology (Caltech).

Montage-produced mosaics of astronomical observations can combine millions of images into seamless tapestries. Large objects such as galaxy clusters typically fall across multiple images, and mosaics allow studying these objects as complete units. NASA’s Earth Science Technology Office Computational Technologies (CT) Project funded Montage development to maximize the information obtained from the many terabytes of observations taken by ground-based telescopes and NASA spacecraft.

“Different telescopes point at slightly different parts of the sky. The resolutions are different as well,” Berriman said. “With Montage, you can process these images so that they have the same image parameters, point to the same part of the sky, and sample the sky at the same rate.” Berriman is also project manager of the Montage development team, made up of Caltech and NASA Jet Propulsion Laboratory (JPL) researchers. Caltech astronomy professor Thomas Prince serves as principal investigator.

Since we last reported on Montage (see “Supercomputer Serves Up Giant Mosaic of Milky Way Images,” ESDCDNews, Summer 2003), the software has undergone several major improvements. Notably, Montage is up to 30 times faster than in its initial release. The speed-up is due to enhanced parallelization techniques as well as better algorithms for performing the necessary image reprojections. Montage now also includes utilities for visualizing and managing images, such as slicing a mosaic into more manageable tiles. “We didn’t intend to deliver these utilities,” Berriman said. “Astronomers asked us to do it.”

Much of Montage’s success stems from its flexibility, Berriman said. It offers user choice with independent portable modules that can run on a variety of computing environments, including workstations, clusters, supercomputers, and grids. The Montage team has joined with the University of Southern California’s Information Sciences Institute to implement Montage on the TeraGrid, the most powerful distributed computing environment in the world. This National Science Foundation-funded resource links over 40 teraflops of computing power, nearly 2 petabytes of mass storage, and data analysis and visualization tools over a dedicated 10- to 30-gigabit-per-second national network.

Image above: The Spitzer Wide-area InfraRed Extragalactic Survey (SWIRE) team produced this optical-infrared panchromatic view of the Tadpole Galaxy (UGC10214), the result of a galaxy-galaxy interaction that stretched the outer spiral arm into a long tadpole-like tail. The mosaic combines several thousand infrared Spitzer images with single, large-format optical and near-infrared images (Image credit: Tom Jarrett and the SWIRE team).

On the TeraGrid, Montage is currently available as a Web service for evaluation by astronomers. The service creates mosaics from three multi-terabyte image databases: the Two Micron All Sky Survey (2MASS), the Digitized Palomar Observatory Sky Survey (DPOSS), and the Sloan Digital Sky Survey. Testers’ feedback will help tweak the software for a public service on the IPAC and National Virtual Observatory Web sites, with a debut expected in late summer.

Whether through the TeraGrid or other means, scientists across the United States and in other countries are harnessing Montage’s capabilities. Their applications are as varied as finding new galaxies, probing star-forming regions, and creating an all-sky map.

The most extensive Montage users come from the NASA Spitzer Space Telescope Legacy Teams. As Berriman explained, “The Spitzer Science Center funded data teams and gave them generous observing time on Spitzer. In exchange, they give early public access to data sets. Legacy implies the value of the data.” Spitzer is an infrared telescope with its own internal repository, but the data will eventually come over to the IPAC Infrared Science Archive.

Image above: Caltech’s Michelson Science Center (MSC) created these mosaics of a molecular cloud, where stars can form. The left-hand mosaic uses two images from the optical Digitized Palomar Observatory Sky Survey (DPOSS). The right-hand mosaic incorporates 75 images from the infrared Two Micron All Sky Survey (2MASS). The optical mosaic clearly shows the cloud boundaries, demonstrating the utility of aligning different types of survey data and placing them on the same pixel scale (Image credit: David Ciardi and the MSC team).

One of the legacy teams using Montage is the Spitzer Wide-area InfraRed Extragalactic Survey (SWIRE), which consists of American and British researchers. SWIRE aims to discover new galaxies at redshifts of 2 to 3, or 3.4 billion to 2.2 billion years old given the best-estimated age of the universe. SWIRE has imaged 49 square degrees of the sky, “equivalent to the area covered by about 250 full moons,” said IPAC astronomer Jason Surace. The Spitzer observations come in seven infrared and near-infrared wavelength bands. Comparative optical observations come from the Isaac Newton Telescope in La Palma, Spain. The SWIRE team uses Spitzer’s MOPEX mosaic tool for their own data and Montage for the optical data and comparisons.

“Various Montage tool modules are scripted to mass-process up to thousands of optical observations and then match the projection, scale, and section of the sky in the Spitzer data mosaics,” Surace said. The team then incorporates the mosaics into multi-band visualization tools to show features across wavelengths. “Having all these data in the same reference frame makes comparison on objects across bandpasses trivial, whereas prior to mosaicking, comparison of objects across different scales and projections was too complex and at times not even possible,” Surace said.

Scientists at Caltech’s Michelson Science Center (MSC) have been using Montage to produce mosaics of molecular clouds and star-forming regions. MSC uses 2MASS and DPOSS images in their mosaics. “By creating single images for multiple filters, all registered to the same coordinate system over many degrees, Montage can enable detailed studies of the large-scale distribution of the mass of the clouds and the positions and evolutionary status of young stellar objects over the extent of the molecular clouds,” said MSC’s David Ciardi. “Without a mosaicking tool such as Montage, studies are generally limited to smaller fields of view or to catalog photometry from sky surveys such as 2MASS, which may not accurately reflect the presence of extended emission around young stellar objects.”

Researchers at one of the TeraGrid sites, the San Diego Supercomputer Center (SDSC), have collaborated with the Montage team to build an all-sky map that encompasses the full 2MASS data set—8 terabytes. Berriman describes the effort as a “computational tour de force.” SDSC processed the 2MASS mosaics using as many
as 1,000 or more processors at a time. Even with Montage’s improved performance, combining more than 4 million 2MASS images took approximately 130,000 processor hours, which included time to shake down the TeraGrid hardware. The final product consists of 5,202 mosaics, with a total size of 20 terabytes.

“The Montage team is using this all-sky computing project to learn how to manage image processing at scale while generating scientifically valuable products,” Berriman said. “The initial results are very promising and show that available hardware and software do indeed support large-scale processing. The Montage team is now investigating the scientific value of the all-sky mosaics to ensure that calibration accuracy is preserved and that background rectification is successful across the sky. Ultimately we hope that our investigations have applicability to future astronomical missions and surveys that require image processing at scale.”

Montage team members include Tom Prince, Bruce Berriman, Anastasia Clower Laity, and John Good at Caltech; Joseph Jacob and Daniel Katz at JPL; and Ewa Deelman, Gurmeet Singh, Mei-Hui Su, and Carl Kesselman at the University of Southern California.

http://montage.ipac.caltech.edu
http://swire.ipac.caltech.edu
http://ct.gsfc.nasa.gov

Image above: This mosaic shows dust clouds and nebulosity in the plane of our Galaxy, as derived from images in the 2MASS Second Incremental Data Release. It combines 347 images observed in three near-infrared bands. The full-scale mosaic (this reduction is 1/24 the resolution) contains 12,000 individual pixels on a side (Image credit: Montage team).