It’s a big universe — billions of galaxies each containing billions of stars, and all of it more than 13.8 billion years old.
“Every astronomical event that is not impossible must occur,” said astronomer Tom Matheson.
That knowledge makes the problem Matheson and three colleagues have given themselves fairly complex: Devise a computer system that will winnow the “rarest of the rare” happenings from more than 1 million “alerts” that will be sounded each night from upcoming surveys of the universe, such as the Large Synoptic Survey Telescope (LSST).
“It is the biggest problem in time-domain astronomy,” said Mario Juric, the LSST project scientist for data management.
It is not the only challenge. Over its lifetime, the LSST project will spend as much on computers, software development and data storage and access as it does on the nearly $500 million construction of the telescope, said Juric.
The collaboration formed to find the “rarest of the rare” events is called Antares, for “Arizona-NOAO Temporal Analysis and Response to Events System.”
It pairs Matheson and Abi Saha, both astronomers at the National Optical Astronomy Observatory (NOAO) in Tucson, with Rick Snodgrass and John Kececioglu of the University of Arizona’s computer sciences department.
Antares is not officially linked to the LSST, but the National Science Foundation, primary funder of the LSST, has given it a three-year $611,000 grant to develop a prototype software system.
In its award, the NSF said “the planned Large Synoptic Survey Telescope project could produce a million or more alerts every single night for a decade. Within this veritable flood will be a small number of rare and unusual sources with short lifetimes that must be recognized in real time, or else the opportunity for thorough study will be lost.”
The LSST, currently approaching the final design review that should pave the way for construction beginning next year, will be a huge leap from the current surveys of the cosmos conducted by astronomers. It will produce 1,000 times more data than anything undertaken so far.
This is “big data,” ultimately measured in petabytes — a million-billion pieces of information — requiring the services of the Blue Waters petascale computer at the University of Illinois.
The LSST has a novel 8.4-meter (27.5-foot) mirror crafted at the University of Arizona’s Steward Observatory Mirror Lab that combines the primary and tertiary mirror in one slab of glass. It will become the largest mirror ever employed in a sky survey.
The mirror will be harnessed to the world’s largest digital camera — 3.2 billion pixels of light-capturing technology, capable of spilling out its product every two seconds.
It will take those pictures from a peak in the Atacama Desert in Chile, generally recognized as the best place in the world to do astronomy.
It will produce 15,000 gigabytes (15 trillion bytes) of raw data every night and has committed to making its alerts available within seconds.
The Antares team is in the early stages of figuring out how to sort the alerts issued each night and find the good stuff within seconds.
It is a high-tech problem that the researchers, so far, are approaching in fairly low-tech fashion. They meet each Monday to draw grids on a chalkboard, sometimes jostling one another for a chance to add a column or row to their evolving matrix.
Complex problems need as simple a solution as possible, said Snodgrass, whose presentation on time-domain computer science at a UA meeting about the LSST brought this collaboration together.
Matheson and Saha were at that meeting. They had been meeting about the alert problem and were looking for computer expertise. Toward the end of the day, Snodgrass spoke about his time-domain computing work.
“Tom and I looked at each other and said, this could be it,” said Saha.
Saha said he and Matheson already had “a pre-baked notion of how to do this, and if we had gone down that road it would have been OK for small amounts of data but not for big data and not for the speeds we want.”
They later presented the problem at a meeting with the computer science department, where they enlisted Kececioglu, whose expertise is writing complex algorithms for “nearest neighbors” — finding things that are alike without having to compare them to the universe of known things.
“If you had to compare it to everything in the catalog and find out what it’s most similar to, the straightforward approach, it would take too long,” said Kececioglu.
“It’s really fortuitous that these two guys happened to be here at the university,” said Matheson.
Snodgrass was urged to the meeting by Elliott Cheu, a physicist who also serves as associate dean of the UA College of Science.
Cheu, whose physics group is currently involved in the ATLAS detector at the Large Hadron Collider, sees LSST as “the next big thing” in science and wants the UA to play a big role in it.
Thousands of scientists from throughout the world are expected to join in, and each will have their own interests in what is discovered.
Instant recognition of rare events will change the way astronomy is done, said Matheson. Currently, astronomers compete for time on the world’s giant telescopes by identifying objects of interest, often from time-domain surveys.
In the LSST era, astronomers will be able to reserve time to observe whatever pops up in the sky that night, Matheson said, knowing it will provide a reliable feed of the particular rare events the astronomers study.
“We envision you’ll just ask for time, point your telescope in the general direction of where LSST is observing that night and wait for alerts,” said Juric. “It’s not the way we’re used to doing things.”
Astronomy has always been a needle-in-a-haystack search, said Juric. “The haystack is about to become immense” and machines will have to be concocted to sort through it all.
Each time the LSST snaps a picture of the sky above its peak in the Atacama Desert of Chile, it will produce at least 1,000 “alerts” on phenomena that might be of interest to astronomers. It will take 1,000 such pictures each night.
It will issue an alert on anything that changes. It might be a moving object, perhaps an asteroid that could impact Earth. It could be a point of light in what was dark sky the last time anyone looked. It could be a dimming of a star that might indicate a planet crossing.
Or it could be something nobody has ever seen before.
The computer program for LSST will have a database of everything that is already known about our cosmos — all the other surveys done of the night sky in the Southern Hemisphere. It will be looking for things that change.
Sorting the information is not a brand-new problem, and astronomers are already solving it at smaller scales.
Up on Mount Lemmon, for instance, the telescopes and computers of the Catalina Sky Survey sort nightly through the sky’s moving objects, discarding those things already known in search of undiscovered near-Earth objects that could impact Earth.
Its trove of discarded, non-moving objects is then mined by astronomers at CalTech, who put together the Catalina Real-Time Transient Survey to identify things that change in brightness — rotating pulsars, exploding supernovae and the like.
The challenge for LSST and other large surveys is to do this with a much larger data set, and to do it in seconds rather than hours or days. That requires machines replacing humans.
“There are plenty of projects going on right now that could use something like this. Most rely on a lot of eyeballs to look at it. Once you scale up, that’s not feasible anymore,” said Matheson.
Saha, who is the operations simulation scientist for LSST, said he and Matheson picked this particularly thorny problem because it could lead to the biggest scientific discoveries and because it lends itself to providing an overall framework for the data flow.
Nothing is rejected as the field of objects is narrowed and can be accessed by other scientists with particular interests.
Saha wants the exotic new thing. “We’d like to see something we’ve never seen before and we don’t know what that is. The idea is that there are things lurking there that would have escaped detection before, things that fleetingly occur in a vast universe.”
He compares the goal to a piece of wisdom he picked up from John Steinbeck’s first novel, “A Cup of Gold,” in which the future pirate Sir Henry Morgan consults the village wise man about his plans for a life adventuring at sea.
“The wise man says ‘You’re off to chase the moon, but we never catch the moon. But along the way you might catch some fireflies.’ ”