[M-IS14] Aqua planetology

[MIS14-P06] Rocognition and classification of Martian chaos terrains using imagery marchine learning: Implications for groundwater distribution and hydrological cycles

Hiroki Shozaki², *Yasuhito Sekine¹, Nicholas Guttenberg¹ (1.Earth-Life Science Insitute, Tokyo Institute of Technology, 2.Dept. Earth Planetary Sci., Tokyo Institute of Technology)

Keywords:Mars, hydrological cycles, machine learning

Water activities have played central roles in the environmental evolution on Mars, through creating landforms and secondary minerals, driving geochemical cycles, and possibly supporting life. Knowledge on the distribution and activities of groundwater is critical in the hydrology, the habitability, and in-situ resource utilization for future manned missions. Chaos are fractured depressions that consist of broken blocks and locate on sources of outflow channels. Some of chaos are interpreted to have been formed through outbursts of groundwater triggered by melting of ground ice; thus, they can be clues to understand the distribution and activities of groundwater. On the other hand, other chaos may have been formed through an inflation-deflation of a magma chamber without water activities. A large number of chaos and chaos-like features have been found on Mars. However, the formation mechanisms of these chaos remained uninvestigated.

Here, we perform recognition and classification of chaos based on the formation mechanisms using imagery machine learning. We developed neural network models (i.e., classifiers) to classify chaos formed by water activities (called as water-related chaos), chaos formed by volcano-tectonic activities (called as volcano-tectonic chaos), and non-chaos surface features (e.g., valleys, plains, and craters) using remote-sensing data of chaos previously investigated. Our developed classifiers can recognize chaos and non-chaos surface features with ~97% of test accuracy and can classify water-related chaos and volcano-tectonic chaos with ~95% of test accuracy.
By applying our classifiers to a large number of unclassified chaos, we found two types of chaos terrains on Mars. One is hybrid chaos terrains, in which water-related chaos and volcano-tectonic chaos coexist in one terrain, and the other is water-dominant chaos terrains, in which water-related chaos are predominant. Hybrid chaos terrains are mainly found in the circum-Chryse outflow channels region. Based on our geomorphic analyses, hybrid chaos terrains would have formed through a combination of upwelling of magma and consequent melting of ground ice. The groundwater from the hybrid chaos terrains formed outflow channels and supplied water to water-dominant chaos terrains in the downstream. Water-dominant chaos terrains are mainly located near the dichotomy boundary. The locations of water-dominant chaos terrains coincide with the proposed regions of upwelling of groundwater on ancient Mars. The locations of water-dominant chaos terrains can be potential landing sites for future missions, where frozen ancient groundwater may be available near surface.