Remote sensing
My research on remote sensing has mainly focused on analysis of thermal emission data of asteroids, which are the debris left over from planet formation and fossils of the processes of planetary formation. Asteroids may be natural impact hazards for Earth too. I have a keen interest in surface evolution of these bodies because their surfaces is what we can remotely study with spacecraft and telescopes and link the results to early solar system processes. To do so, I develop advanced computational models to simulate surface evolution processes of airless bodies and analyze spacecraft and ground-based observations, often using machine learning techniques.
Mapping the surface of asteroid (16) Psyche using ALMA data
Cambioni, de Kleer, Shepard 2022 (JGR: planets)
In this paper, my colleagues and I present geographical maps of the metal content and thermal inertia of (16) Psyche, a 200km-sized asteroid long thought to be the core a lost protoplanet. To derive thermal inertia and dielectric constant (proxy for metal content) area-by-area on Psyche, I developed a new thermophysical model to fit these surface properties to temperature data collected with the Atacama Large Millimeter Array (ALMA) at 30 km spatial resolution presented in de Kleer, Cambioni and Shepard, 2021 (which did a disk-integrated fit to the data, finding that Psyche’s surface has a metal content of no less than 20% ) and using the shape model by Shepard, de Kleer, Cambioni et al. 2021. These new maps of Psyche reveal that its surface is highly heterogeneous in metal content and processed by impacts. The most exciting features are large depressions that appear to have particularly high metal content (excavated core materials? Ferrovolcanic deposits?) and a large depression (likely a crater) whose lowlands have lower thermal inertia than the surrounding highlands, indicating variations in either regolith grain size, porosity, or composition.
We look forward to the NASA Psyche mission, which will soon explore asteroid Psyche to test the hypothesis that it is the core of a lost protoplanets, to verify or disprove our interpretations of this unique ALMA dataset!
Top: An artistic rendition of Psyche's surface, featuring both rocky areas and metal-rich areas. Left: top: millimeter-wavelength emissions of Psyche from deKleer et al. 2021. Source: Caltech; bottom: mollweide projection of thermal inertia of Psyche from the study of Cambioni et al. 2022, with altitude contours from the shape model by Shepard et al. 2021. In the news: link, link
Fine-regolith abundance on asteroids controlled by rock porosity
Cambioni, Delbo, Poggiali et al. 2021, Nature
In this paper, my colleagues and I provided evidence that the lack of fine-grained materials (cm-sized fine regolith) on the carbonaceous asteroid (101955) Bennu is due to the high-porosity of its rocks, which get compacted but not fragmented by impacts and experience slow thermal cracking. I led this work as a student collaborator on the NASA OSIRIS-REx sample return mission, training machine learning algorithms to mimic the outcome of expensive, high-resolution thermal simulations of Bennu's surface using the method by Cambioni et al. 2019 (described below). From the results, we inferred that carbonaceous asteroids like Bennu and Ryugu — which are the most populous type of asteroids — should lack fine regolith ponds, while S-type asteroids like Itokawa— which are the second-most populous group, should have terrains rich in fine-regolith, because the rock porosity on the former tend to be than the latter. My current research includes keep testing this prediction, as this has implications for planning sample return missions, planetary defense, as well as meteorite formation.
Asteroids' rocks with higher porosity are compacted by meteoroid impacts rather than excavated. Thermal stresses in a more porous rock are weaker in magnitude than in a denser rock, which means that the former could be less prone to producing fine regolith than is the latter.
In the news: UA Press release, Forbes, DPS 53 Press conference
Distinguishing fine-regolith from rocks using machine learning and thermophysical models
Cambioni et al. 2019, Icarus, (see also Chapter 10 in book "Machine Learning for Planetary Science", edited by Elsevier).
In this paper, my colleagues and I designed a new machine learning (surrogate) thermophysical model to derive the thermal inertia of rocks and their fragments and their relative surface abundances on asteroids from emission in the thermal infrared. Distiguishing rocks from fine regolith eluded previous studies using more conventional techniques that assumed a single ("effective") value of thermal inertia of the surface.
We used this approach to derive the surface abundance of sub-cm regolith on asteroid Itokawa by interpreting ground-based infrared observations. We found that 20% of the surface is covered in cm-sized regolith, consistent with the findings by the JAXA Hayabusa mission that visited the asteroid in 2005, and that the rocks on Itokawa have porosity of 20% ± 4%, suggesting that they are fractured by regolith-forming processes.
Our machine-learning-powered thermophysical approach derives the surface roughness θ, rock abundance RA and the thermal inertias of the regolith and the rocks of asteroid surfaces from observations of the asteroid infrared spectra. ω is the vector of parameters learned by the neural network during training. The thermophysical model used to train the neural networks is that described in Delbo et al. (2015), in book Asteroid IV.
Looking for smooth terrains on asteroid (101955) Bennu
Cambioni et al. 2019, EPSC (see also Walsh et al. 2022, SSR)
In this conference paper, my colleagues and I used machine learning to aid the selection of the sampling site on the surface of asteroid (101955) Bennu, the target of the NASA OSIRIS-REx sample return mission. As a student collaborator on the mission, I customized an existing algorithm for rock classification that was developed by Wagstaff et al. 2013 for Mars to identify smooth terrains which, on the surface of asteroids, are typically associated with the presence of fine regolith that could be sampled by the spacecraft. The search detected three new areas that skipped previous identification by human operators, highlighting that team members experienced fatigue while searching manually for these areas on the surface.
A: an image of the asteroid surface used for training. B: label image, sketching the smooth and rough areas manually identified in the region of image A. C: an image of the target asteroid where the trained machine automatically identified pixels corresponding to smooth areas.