Presentation #507.02 in the session Icy and Polar Mars.
Mars’ seasonal frost cycle is a critical area of planetary science investigation because frost is a dominant driver for Mars’ climate and surface evolution through the last ~2 billion years (i.e., the Amazonian epoch). Additionally, identification of low-latitude frosted microclimates is important for characterization of potential habitable environments and potential resources for future human explorers. Finally, studies of martian frost may provide an unmatched look into sublimation-driven processes that are also active in the outer solar system and that have no terrestrial analog.
Previous studies of Mars’ seasonal frost cycle traded off between coverage and spatial resolution within manual analysis: scientists took either a global perspective using coarse (>kilometers/pixel) observations, or a local “site” perspective focused on higher-resolution observations (sometimes including multiple datasets). In our study, we utilize data science techniques to identify frost at global scale using both high-resolution and coarse datasets. We also are able to robustly combine different datasets so as to characterize frost type and reduce/quantify uncertainty. For example, using a small set of labeled imagery (see figure), we have trained a machine learning classification model to detect frost in meter-scale visible imagery (HiRISE). We then compared ~kilometer-resolution thermal data (MCS and THEMIS) that was sufficiently close (in time and space) to visible detection so as to mitigate the uncertainty of some visible detections, map the frost distribution at the highest available spatial and temporal resolution, and differentiate between recent CO2 snowfall, CO2 surface frost, and water frost. In the future, we will also integrate meters-scale, global-coverage imagery (CTX) and meters-scale spectral information (CRISM) into our analysis.
Harmonizing these complementary observations will enable a holistic look at the martian seasonal frost cycle over a wide range of relevant spatial and temporal scales. The final data product capturing frost type, timing, and locations, will provide the planetary science community with a new observation-based understanding of where and when different martian frosts form. With it, the community will be empowered to better constrain and validate a wide range of models focused on Mars global/regional atmospheric circulation, volatile transport/budget, and geomorphological activity.