Lunar swirls are recognized as broad, bright albedo features in various regions of the Moon. These features are often separated by dark off-swirl lanes or terminate against the dark background, such as lunar maria. Prior mapping of swirls has been done primarily by albedo contrast, which is prone to subjectivity due to differing interpretations from one mapper to the next. Upon closer examination of the on-swirl areas, the albedo in many cases do not appear to be the same everywhere, suggesting that there is a transition from bright to dark.
In a new study, a team of PSI scientists led by Senior Research Associate Frank Chuang applied machine learning techniques to address these issues by identifying the number of different swirl units and then mapping them based on the surface reflectance or light energy recorded by the NASA Lunar Reconnaissance Orbiter (LRO) Narrow Angle Camera (NAC). The reflectance, data that is calibrated using the spacecraft and sun positions relative to the surface, was then selected for two locales with prominent swirls, Reiner Gamma and Mare Ingenii.
Two separate machine learning (computer-coded) algorithms were applied to the data in each region, K-means clustering and Maximum Likelihood Classification. The former is an ‘unsupervised’ technique in which the algorithm defines and maps all of the data into units based on their values and their location. The latter is a ‘supervised’ technique that uses user-defined training areas for each unit type in the classification and mapping process. Results show that the classification maps are a very reasonable match to what is observed for the two study regions. Mapping of lunar swirls typically involve two units, on-swirl and off-swirl, but the machine learning techniques identified and confirmed a third transitionary unit between the two, termed diffuse-swirl. These results show for the first time that lunar swirls can be defined and mapped using quantitative information, thus removing most of the subjectivity involved in prior studies. The data and the statistics generated from the maps also have value in future studies in studying other swirl regions on the Moon.
Chuang is lead author on the paper “Mapping Lunar Swirls with Machine Learning: The Application of Unsupervised and Supervised Image Classification Algorithms in Reiner Gamma and Mare Ingenii” that appears in the Planetary Science Journal. PSI scientists Matthew Richardson, John Weirich, Amanda Sickafoose and Deborah Domingue are co-authors on the paper.