Abstract
Basaltic lava flows are one of the most common landforms across the Earth and other rocky planets in the solar system. Their presence indicates a clear geological event in the cooling history of a planet, where molten rock breeched from the crust, interacted with the surface, and then froze in place. As a result, basaltic lava flows can characterize the environments of a planetary system, both internal and external, past and present, that few other landforms can provide. One characteristic of basalt, crystallinity, refers to the proportion of crystalline minerals versus glass within the lithology of frozen lava, where glass only forms when molten material rapidly froze by contacting cold atmospheres, water, or ice. Crystallinity can be approximated from the magnitude of in-situ visible-and-near-infrared reflectance averaged from 500 to 1000 nm (R500-1000), but this method has not been adapted to complex flow surfaces nor to remote sensing images of basalt lava flows which would enable regional- and planetary scale interpretations of cooling history.
In this dissertation, I investigate the R500-1000 parameter on the morphologically complex Krafla lava flow, Iceland, to test if reflectance varies with morphology. I find that surface exposure controls reflectance independent of morphology, where preservation of the exterior glass-rich rim or physical erosion into the crystalline interior directly affects R500-1000. Some morphologies show preference for different exposure types, where smooth pahoehoe surfaces may be glass-rich or crystalline if exfoliated, while rough pahoehoe can remain glass-rich if not overturned. Once overturned, rough crystalline undersides of pahoehoe are difficult to distinguish from a’a.
To adapt these observations to remote sensing, I produced a morphological map of several lava types to segment multiple spectral parameters sensitive to reflectance magnitude and diagnostic absorptions. These maps support evidence of shifts in reflectance due to glass-rich and crystalline mixtures per lava type that are proportional to in-situ, which approximates crystallinity, but are limited to derive crystallinity or R500-1000 directly due to physical and instrumentational effects that are difficult to resolve on remotely sensed images of rough, dark, basalt surfaces.
To account for these effects within satellite images, I adapt a machine learning model to apply the relationships between satellite and in-situ reflectance to predict flow-scale R500-1000 as a proxy for crystallinity across Krafla and a second flow Jordan Craters, Oregon, as a test for robustness. These models support earlier results, indicating that rougher lava types are more crystalline on average while exposure type controls reflectance overall. Field observations confirm this, where glass-rich surfaces are predicted with lower R500-1000, while tilted, disrupted, and exfoliated surfaces with crystalline exposures are predicted with higher R500-1000 proportionally, independent of surface roughness. These results are consistent across the surfaces of both flows, providing evidence that crystallinity can be approximated from satellite images of basalt lava flows despite containing effects from instrumentational noise and the geometry of surface roughness.