Spot stars exploding over Australia with Snapshot Supernova

Snapshot Supernova asks the world’s citizen scientists and armchair astronomers to help discover stars exploding over Australia. It’s the latest project form the Zooniverse crowdsourcing service. The Zooniverse’s co-founder and Oxford University astrophysicist Chris Lintott explained to me how the project evolved from earlier supernova-hunting projects to make its first supernova sighting within hours of the project’s launch.

Burn bright and leave a stellar remnant

In this view from the Hubble Space Telescope, the Crab Nebula is all that remains of a massive supernova that shone in the skies almost a millennia ago. Recorded by astronomers in China and Japan in 1054, it remained visible for two years.

One of two fates await an aging star as its fusion reactions fizzle out. Most stars, including the Sun, become white dwarfs and slowly radiate their residual heat way like the dying embers of a campfire. The most massive stars, however, go out with a bang. Without the pressure from a giant star’s internal fusion reaction, its core collapses under its own mass as gravity crushes it form a neutron star. The sudden release of energy blasts the star’s outer layers away in an explosion so powerful that it shines with a light billions of times brighter than the original star. Credit: Nasa, Esa, J. Hester and A. Loll (Arizona State University)

Both the spectrum of that light and the way it fades over time tell astronomers a lot about the nature of the original star and the physical processes inside it. The trick is making the observations as close to the supernova’s beginning as possible. The giant mountaintop telescopes wouldn’t be suited for supernova searches even if they weren’t already heavily oversubscribed by the world’s astronomers.

Dedicated supernova search programs - both professional and amateur - use relatively small telescopes to scan the entire sky over and over again every night. Automated software searches the data for signs of a supernova and flags it for scientists to review before notifying the astronomy community. But sometimes software isn’t that smart.

Amateurs: as good as experts, better than machines

Galaxy Zoo: Supernova was the Zooniverse’s first supernova-spotting project. It used data from the Palomar Transient Factory’s 1.2-meter telescope. The PTF’s software generated 500 supernova candidates every night which a handful of scientists had to review. Only a few of those candidates - and sometimes none of them - would turn out to be supernovae. Hardly a good use of their time. Fortunately, Galaxy Zoo: Supernova’s 13,000 volunteers did the work just as well. They reviewed before-and-after images and flagged those they thought showed a supernova. Thanks to the wisdom of the crowd, the project’s scientists could be confident they had a real supernova when an image got flagged a dozen times. Galaxy Zoo: Supernova’s citizen scientists ended up discovering more than 3,000 supernovae over the project’s two year run.

Contributors to Snapshot Supernova see three images. The two on the left are pictures of the same part of the sky taken at different times. The third is a subtraction image that only shows the things that are different. Stars and galaxies should disappear in the subtraction image, leaving the supernova sitting in the middle of the crosshairs. This image shows the first supernova discovered by Snapshot Supernova’s citizen scientists. Credit: ANU/Skymapper Southern Sky Survey/Zooniverse

Six years later Snapshot Supernova picks up where Galaxy Zoo: Supernova left off. It uses data from Australia’s 1.35-meter Skymapper Telescope. Volunteers compare before-and-after images in a process similar to the original project. Chris Lintott explained:

We tried to keep things close to the original because we know that it worked!

“Worked” is putting it mildly. The project launched last Wednesday. Within two days more than 36,000 people around the world made over 1.5 million classifications - and discovered their first supernova. As the citizen scientists settle in, the Snapshot Supernova team is helping them improve their work. Lintott explained:

We did add extra help text because people needed the support. We added a system for people to discuss what they’ve seen as that’s proved very useful in other projects. Our science team have been tagging things [in the forum] with #rsgood and #rsbad to provide extra samples.

The citizen scientists have time to practice, though. Even with the huge effort of the past two days they are only 8% of the way through. They have a lot more supernovae to spot.

Human and machine - better together

Spotting supernovae isn’t the project’s only goal. The original Galaxy Zoo: Supernova created a massive data set that PTF scientists used to make machine learning search algorithms that can spot supernovae better than any human. Machine learning is a limited form of artificial intelligence. Designed to learn from its own mistakes, the algorithm crunches through a set of known data and rewrites itself as it goes along. As Lintott explained:

We can already see that [the Skymapper algorithm] misses some things that our volunteers catch. We know from Galaxy Zoo: Supernova that one of the barriers to good machine learning is the need for a spectacularly large training set, and that’s what we’ll provide.

That doesn’t mean Snapshot Supernova’s citizen scientists are putting themselves out of a job. The only efficient way to create those spectacularly large training sets is through the combined efforts of tens of thousands of normal everyday people. New survey projects like Pan-Starrs will run into the same challenges with their early software. They will need help from the world’s amateur supernova spotters. Lintott concluded our exchange with this enticing statement:

I’m convinced the real future is in systems that combine humans and machines, deciding on the fly what’s missing. You’ll be hearing lots more about this shortly.

Want to learn more?

If you liked this article and want to see more, please consider supporting Small Steps to Space.