Identification of Mars Mesospheric Clouds in Mars Climate Sounder Data Using a Machine-learning Algorithm

Water-ice and CO2-ice clouds have been observed in the mesosphere of Mars, but limited spatial coverage has precluded a comprehensive climatology integral to understanding their formation mechanisms. Several previous analyses have identified populations of Mars Mesospheric Clouds (MMCs), but to date there has not been an exhaustive study across all latitudes, longitudes, and seasons. Mesospheric clouds are observed in limb measurements at visible and infrared wavelengths by the Mars Climate Sounder (MCS) onboard Mars Reconnaissance Orbiter. Due to the observation geometry, the apparent altitude of the cloud rises and falls as the spacecraft moves along its orbit. This leads to high radiance values above ~40 km and arch-shaped features in times series of measured radiance profiles. With over 13 years (6 Mars years) of in-track limb observations, this data set has yet to be exploited to build a climatology of mesospheric clouds.

We constructed an automated cloud detection algorithm using an unsupervised machine-learning approach to catalog and characterize these clouds. We utilize edge detection techniques and a k-means clustering algorithm to identify clouds above 40 km. Here, we present preliminary results of MMCs identified in eight MCS channels covering visible to far infrared wavelengths during one Mars year. We will present latitudinal and longitudinal distributions in each channel. The promising results of this technique suggest it will be valuable in determining a complete climatology of clouds at 3pm and 3am. Furthermore, utilizing all MCS channels will improve our ability to differentiate by composition leading to a better understanding of the physical processes enabling cloud formation in Mars' middle atmosphere.