So as to not bury the lede, we’re testing a new method for classifying galaxies with Galaxy Zoo Tags! If you want to skip my musings about the evolution of Galaxy Zoo, you can jump directly to the section introducing this new method here.

First, as a quick introduction, my name is Hayley Roberts and I’m a postdoctoral astrophysicist and data scientist for Zooniverse based at the University of Minnesota. I finished my PhD last year at the University of Colorado Boulder, studying a rare phenomenon found in extreme galaxy mergers called OH megamasers. As a postdoc, my work has broadened to studying galaxy evolution through major mergers and starburst galaxies. This has included evaluating how well our galaxy classification schemes work for high redshift, or more distant, galaxies. 

Our Evolving View of Galaxies

When Galaxy Zoo (GZ) launched in 2007, the first campaign utilized data from the Sloan Digital Sky Survey (SDSS), a pioneering survey enabled by innovative instrumentation and data handling techniques. It cannot be overstated how much SDSS data revolutionized our understanding of many aspects of astronomy, but particularly galaxy evolution and morphology, through new insights such as the color-magnitude relation and galaxy environment. However, the galaxies in this SDSS sample have a median redshift of z~0.1 (~1.3 billion light-years away), meaning this first GZ campaign was limited to only our nearest neighboring galaxies. This is reflected in the earliest iteration of the GZ classification workflow, which only asked volunteers to determine if a galaxy was a spiral (edge-on, clockwise, or anti-clockwise), an elliptical, or a merger. Clearly, we had a long way to go before the myriad of potential other options reflected in the current GZ workflow were conceived. 

What drove the expansion of the classification choices in Galaxy Zoo workflow?

Two things: you and the data. Before the launch of the first iteration of GZ, the hope was to get 20,000–30,000 volunteers to participate in the first few months — the actual number ended up surpassing 100,000. This fundamentally altered the outlook on what GZ was expected to be able to do. The first GZ data release included classifications for nearly a million galaxies, an order of magnitude more than previous comparable studies. This data enabled numerous studies on unprecedented scales (such as studying the properties of dust in spiral galaxies, compiling the largest sample of mergers at the time, and exploring the co-evolution of host galaxies and their AGN), and led to new discoveries (including green peas and voorwerpjes). These first couple years demonstrated that your classifications were enabling GZ to achieve extraordinary scientific results. 

Since GZ’s launch, our view of the galaxies in our universe has expanded in number, diversity, and distance. Surveys, such as DESI, have allowed GZ volunteers (you!) to classify millions of galaxy images. The diversity of galaxies classified have inspired aspects of GZ classification tree iterations over the years or entire spinoff projects. However, over the years, the biggest change to the galaxies being classified on GZ has been evident through new technology and telescopes.

Above shows images of a galaxy obtained by four different telescopes or surveys that have starred in different GZ campaigns: SDSS, DECaLS, HST, and JWST. This galaxy, CANDELS J141937.3+525050.3, has a redshift of z=0.73 (~6.5 billion light-years away) and is a spectacular demonstration of just how our view of the universe has evolved. The SDSS and DECaLS images, the first and second panels, show a distant, unremarkable smudge of a galaxy. However, the images taken by HST and JWST, the third and fourth panels, unveil a spectacular merger between three galaxies. The higher sensitivity and resolution achieved over the years has given us unparalleled views of the galaxies in our neighborhood. Additionally, the number of high redshift galaxies in GZ has vastly increased, as well as the maximum redshift of these galaxies, particularly for JWST.

High redshift galaxies, those that are very distant and therefore seen as they were in the early universe, often defy traditional categories like spiral and elliptical galaxies. The universe was much younger and more dynamic when these galaxies formed, leading to a greater variety of shapes and structures. Unlike their more mature counterparts, high redshift galaxies frequently exhibit irregular and clumpy morphologies, complicating morphological measurements and making them harder to classify using traditional frameworks. Understanding high redshift galaxy morphology may require new classification schemes with less rigidity that account for their unique properties and the evolving state of the early universe. 

Introducing Galaxy Zoo Tags: An Experimental New Way to Classify Galaxies

As we push the boundaries of our observations, the classification of high redshift galaxies presents new challenges, as discussed above. To address these challenges, we’re testing an alternative way to classify galaxies called Galaxy Zoo Tags! This workflow allows you to assign multiple “tags” (morphological features) to each galaxy without the restriction of working within the classification tree. This should allow for substantially more flexibility in classifications and feature combinations that might not be fully captured in the traditional classification tree. 

You can start classifying using this new workflow right now on a sample of galaxy images from DECaLS and CEERS data from JWST. All images in this test period have been previously classified using the traditional classification tree. The goals of testing this method are to:

1. Compare the results of the two different classification methods (tagging vs. traditional GZ tree), and 
2. Collect your thoughts and feedback on this new approach.

To help with the second goal, we’ve set up this Google form for you to provide feedback on this new classifying method. It’s also linked in the banner on the GZ Tags page. We would greatly appreciate your time testing this new method out and any thoughts or opinions you may have about it. If you have any questions, you can come chat with us on this Talk board.

We don’t have any plans to remove or replace the current GZ classification system at the moment. The future of this workflow style will be determined both by how helpful/informative the data is and your response. This is just one new approach, but there are likely many others we could pursue. As always, your input, feedback, and ideas are imperative to designing any potentially updated classification schemes, so please do consider testing out this new method and providing feedback via the Google form.  

Thank you all again for your efforts and making this all possible.