Machine Learning Analysis Of Martian Valley Networks: Paleoclimatic Implications
Presenter: Lin Ji1
Co-Author(s): Tao Liu, Victor R. Baker
Advisor(s): Victor R. Baker
1Department of Hydrology and Atmospheric Sciences, University of Arizona
Although drainage basin geomorphometry has long been recognized as a potential paleoclimatic indicator for early Mars, previous analyses have been limited to a relatively small number of basin characteristics, with many studies restricted to local scales. We here employ a big data machine learning approach at global scales for both Mars and Earth, considering basin networks, basin shapes, drainage textures, relief, longitudinal profiles, and hypsometric metrics. We find that the arid regions on Earth correspond to small, round, steep, young watersheds, and straight longitudinal profiles, that are primarily concentrated in the mid-latitude, desert zones of the planet. On Mars, for the planet-wide swath of highland valley network dissection along the northern boundary of the planetary dichotomy, we observe smaller, rounder, steeper, younger watersheds, straighter longitudinal profiles, and dominance of first-order channels. We further conclude that (1) the global Martian drainage system is less developed compared to Earth's; (2) over 90% Martian valley network predominantly occurred under arid conditions. The comparison to the Earth data also strongly supports the presence at the time of valley network formation, of an ancient hydrological cycle on Mars capable of producing high-intensity, short-duration rainstorms. Such conditions, near the Mars highlands/lowlands dichotomy, would have been readily facilitated by the contemporaneous presence of a large water body (an “Oceanus Borealis”) on the northern plains of Mars. These findings contribute to a better understanding of the early Mars climate and the associated geomorphological processes on Mars, particularly in regard to the ancient valley network fluvial environments.