GSA 2020 Connects Online

Paper No. 207-14
Presentation Time: 5:02 PM

PIXELS AND PINGOS: UTILIZING COMPUTER VISION TO DETECT PINGOS ON EARTH, MARS, AND CERES


UDELL, Kathrine1, HUGHSON, Kynan H.G.2, SCHMIDT, Britney E.2, SIZEMORE, Hanna G.3, SCULLY, Jennifer E.C.4, RAYMOND, Carol A.4 and RUSSELL, C.T.5, (1)School of Earth & Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst Drive, Atlanta, GA 30332, (2)Georgia Institute of Technology, Atlanta, GA 30332, (3)Planetary Science Institute, Tucson, AZ 85719-2395, (4)Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, (5)Earth, Planetary and Space Sciences/IGPP, University of California, Los Angeles, 603 Charles Young Drive, 3845, Los Angeles, CA 90095

Pingos are mound-like structures formed through the freezing of groundwater under specific periglacial hydrological conditions [1]. These dynamic water-based systems are of interest on other planetary bodies due to the implications of the existence of subsurface water ice resources. On Earth, these features are found in Canada, Alaska, Russia, and elsewhere where the necessary conditions exist. Similar landforms have been documented on Mars on Utopia Planitia [e.g. 2 & 3], and recent high-resolution images from NASA’s Dawn spacecraft revealed pingo-like morphologies on the dwarf planet Ceres in Occator and Urvara craters [4,5]. These potential hydrological systems on Mars and Ceres may be, or could have been, important habitats for life or nexus for pre-biotic chemistry. Additionally, identifying and characterizing the water resource potential of possible extraterrestrial pingos is relevant for future human space exploration.

This project is a combined effort between manually identifying potential pingo features on Earth, Mars, and Ceres, and developing a computational pipeline that can identify these features to a level of accuracy similar to human identification using computer vision and machine learning techniques. We examined these features on Earth on the Tuktoyaktuk peninsula at a scale of 1:10,000, on Mars within Utopia Planitia at a scale of 1:2,500, and on Ceres within Occator Crater at a scale of 1:10,000. Earth and Ceres imagery are both at a resolution of ~3m/pixel and Mars imagery is ~1m/pixel. Manually mapping these features allowed us to then check our code output for accuracy.

Currently, preliminary maps of pingo features on Mars, Ceres, and Earth have been completed manually. A code has been developed that successfully identifies pingo features on Earth with a high degree of accuracy and will be used to help map potential pingo-like features on Mars and Ceres.

References: [1] Mackay J. R. (1998) Geo. Phys et Quart., 52, 271-323. [2] Dundas C. M. & McEwen A. S. (2010) Icarus, 205, 244-258. [3] Soare R. J. et al. (2019) Icarus, in press. [4] Schmidt B. E. et al. (Accepted) Nat. Geo. [5] Hughson K. H. G. et al. (2020) LPSC LI.

Acknowledgments: Planet Team (2017). Planet Application Program Interface: In Space for Life on Earth., the Dawn Data Archive on the PDS SmallBodies Node, Mars HiRISE on the PDS Mars