Just a few pixels would let astronomers map surface features like oceans and deserts on an exoplanet
Direct images of exoplanets are rare and lack detail. Future observatories might change that, but for now, exoplanet images don't tell researchers very much. They merely show the presence of the planets as blobs of light.
But a new study shows that only a few pixels can help us understand an exoplanet's surface features.
Astronomers can directly image exoplanets, but only in certain circumstances. Usually, the light from a star washes out the much fainter light from any exoplanets orbiting it. Exoplanets that are very large, very far from their star, or very young are the exceptions. Astronomers can image young planets in infrared because their thermal output is high, while the light from massive exoplanets or exoplanets far from their stars don't have their light washed out as much.
Indistinct images of the exoplanet AB Aur b were enough for one team of researchers to expand our understanding of planet formation. And since most exoplanets are found by examining transit light curves, any actual images of exoplanets are exciting. If the authors of a new study are correct, then even a few pixels of an exoplanet's surface can propel our understanding forward, just like transit light curves have.
The new study is titled "Global Mapping of Surface Composition on an Exo-Earth Using Sparse Modeling," available online at the pre-press site arxiv.org. The lead author is Atsuki Kuwata from the Astronomy Department at the University of Tokyo.
The study focuses on the future when direct imaging of exoplanets becomes viable. At first, these direct images may only provide a few pixels of an exoplanet's surface. The question is, how can we learn as much as possible from a few scant pixels? More than you might think at first, according to this study.
In their paper, the team explains that "the time series of light reflected from exoplanets by future direct imaging can provide spatial information with respect to the planetary surface." They used "sparse modeling" to extract information from direct exoplanet images. Sparse modeling is a machine learning tool that can discover predictive patterns in data, even when the data is sparse or weak.
The researchers employed their sparse modeling on what they call a "toy Earth." They identified surface features useful in exoplanet study. "Applying our technique to a toy model of cloudless Earth, we show that our method can infer sparse and continuous surface distributions and also unmixed spectra without prior knowledge of the planet's surface," they write.
They also applied their technique to actual Earth data from DSCOVR/EPIC. DSCOVR is an NOAA Earth observation satellite, and EPIC is a polychromatic camera on the DSCOVR satellite. EPIC is a powerful tool that provides detailed measurements of ozone, aerosols, cloud reflectivity, cloud height, vegetation properties, and UV radiation estimates at the Earth's surface. The researchers "dumbed down" all this detailed data on Earth's surface as if it was a distant exoplanet they were looking at.
By applying their sparse modeling technique to the DSCOVR/EPIC data, they found patterns that they identified as oceans and cloud cover. They also found two components that they identified as land. "Additionally, we found two components that resembled the distribution of land. One of the components captures the Sahara Desert, and the other roughly corresponds to vegetation, although their spectra are still contaminated by clouds."
Scientists have been working on extracting as much information from sparse data in exoplanet images. One of the methods is called Tikhonov regularization. The image below compares the team's sparse modeling with Tikhonov regularization. "We concluded that sparse modeling provides better inferences of the surface distribution and unmixed spectra than the method based on Tikhonov regularization," the authors write.
This study is a refinement of some previous work, and the results are intriguing. One of the obstacles in this kind of work is that planets rotate. For any results to be valid, scientists have to account for the exoplanet's spin with extreme accuracy. But clouds don't sit still while we take their portraits from tens or hundreds of light-years away. The study had to make accommodations for that. "Additionally, we assumed the surface distribution of the end member as static, but we should also consider the dynamical motion of surfaces, especially for clouds," the team writes in their conclusion.
This work is taking on new significance because upcoming telescopes will start imaging exoplanets directly. This is the realm of our powerful new ground-based telescopes like the upcoming European Extremely Large Telescope (E-ELT) and the Giant Magellan Telescope (GMT.) These telescopes are remarkably powerful and will produce images sharper than space telescopes. The sharpness is necessary to detect direct light from exoplanets and image them.
Currently, direct images of exoplanets don't contain much detail. They're still fascinating and scientifically valuable in some ways, but they don't reveal surface detail.
Artists are another resource in exoplanet images. Skilled illustrators like the ESA's Martin Kornmesser trigger our curiosity and excitement with their data-based portrayals of distant worlds. Without Kornmesser and others spreading exoplanet excitement to the broader public, we'd be in a different place.
In 2015, GMT Project Director Patrick McCarthy told Forbes Magazine that "We should [also] be able to see Jupiter- and Saturn-like planets forming around stars in the Milky Way's Orion and Taurus star-forming complexes with relative ease."
But those images won't be crystal clear, and they won't reveal all of a planet's surface detail. Scientists will still have to tease out as much detail as they can from these images using machine learning, modeling, simulations, and other tools.