The Moon’s South Pole−Aitken basin, or SPA for short, is one of the oldest and largest impact structures in the Solar System. The more scientists learn about it, the more they can piece together lunar history and the history of the Solar System.
Covering the central portion of this basin are three different surface material types – maria, cooled pools of lava that appear dark gray; cryptomaria, maria that are lighter in color and partially hidden; and expansive light-colored swaths of lunar plains. Distinguishing between the three by eye can be difficult and time consuming for scientists.
A group of scientists led by Planetary Science Institute Senior Research Associate Frank Chuang set out to see if they could use machine learning techniques to identify and map these material types using albedo – the amount of light or radiation reflecting from a surface – and topography data from the Lunar Reconnaissance Orbiter.
The team applied two machine learning algorithms to the data. The first, called K-means clustering, is what’s known as an ‘unsupervised’ technique in which the algorithm defines and maps the data into units based on the raw data values and their location. The second, called Maximum Likelihood Classification, is a ‘supervised’ technique that relies on user-defined training areas for each unit type in the identification and mapping process. The team used this same technique in a 2022 paper to map lunar swirls, a different albedo feature seen on the Moon.
Their results show that machine learning are as, and possibly more effective, than humans in detecting cryptomaria.
“After applying our algorithms, not only do we find that the mapped cryptomaria agree fairly well with previous studies of cryptomaria, but these are in fact sites where maria was present earlier and are now incompletely covered by a non-volcanic deposit or spotty in places such that the underlying maria remains exposed at the surface,” Chuang said. “If cryptomaria are areas where lunar volcanic materials were present in the past – for example, maria that flowed across and/or infilled parts of the Moon – then our findings suggest that the total amount of maria in SPA are likely underestimated. Thus, the amount of internal heat or energy to produce that volcanic material was likely greater in the past compared to what’s currently known and that is ultimately what our team was trying to understand.”
The findings by Chuang and his group are now published in The Planetary Science Journal as part of a special focus issue on the South Pole-Aitken basin. Other co-authors include PSI Postdoctoral Research Scientist Matthew Richardson and Senior Scientist Deborah Domingue, as well as Jennifer Whitten of the National Air and Space Museum and Daniel Moriarty of NASA Goddard Space Flight Center. The PSI portion of this work was supported by NASA’s Lunar Data Analysis Program (80NSSC20K1425).
PSI’s Alan Fischer assisted with this story.