Home

Hello! I'm a fifth and final year astronomy PhD candidate at the University of Michigan. I started here in 2022 after getting a BS in astronomy & physics from the University of Washington. I work with Professor Emily Rauscher studying the atmospheres of hot Jupiters, the most extreme class of planet. My research focuses on the development and use of 3D General Circulation Models (GCMs, a.k.a. Global Climate Models), which solve the equations of fluid dynamics and radiative transfer (coupled to parameterizations of other physical processes) to find the equilibrium state of the atmosphere (see below for some examples!). Most of my science is done using the RM-GCM, developed by Prof. Rauscher's group at U of M. I'm interested in population-level trends across the hot Jupiter range, and in particular how clouds impact these atmospheres and how they look to space-based observatories.

I can be reached via email at thomak@umich.edu and am always happy to chat about my science!

Cloud Radiative Feedback, GCM grids

I run and interpret GCMs with clouds actively included as a source of opacity to explore their impact on climate and observables

Limb Asymmetries

I combine GCMs and CARMA to interpret limb asymmetry signals in hot Jupiter transmission spectra to constrain planetary properties and microphysical models

Model Development

I do a lot of model development to improve the RM-GCM. Most notably, I upgraded the RM-GCM's radiative transfer module to include a correlated-k opacity treatment and parallelized the code

Cloud Radiative Feedback

When clouds are present in an atmosphere, they contribute absorption and scattering in addition to the background gas. This can have a range of impacts on the structure of the planet, in addition to their direct effects on observables. The exact effect a cloud has depends on where it is in the atmosphere, as well as how much scattering and absorption it does (based on its composition, abundance, and the sizes of individual particles).

When reflective clouds are present on the dayside, particularly in the upper atmosphere, this has a cooling effect on the planet as a whole. Reflection from clouds dominates the Bond albedo (the fraction of starlight reflected by the planet) for intermediately-hot Jupiters. Reflective clouds can still result in heating in the cloud layer, however, as the photons that are not reflected are quickly absorbed. This combination of deep-atmosphere cooling and upper atmosphere heating can result in thermal inversions at the cloud deck.

When clouds are absent on the dayside, but present on the nightside (as will be the case for many hot Jupiters), starlight can be absorbed fairly deep into the atmosphere, where redistribution of heat is faster than radiative cooling. This means that the heat is efficiently advected to the nightside, where the clouds trap heat like a blanket. This is analogous to the greenhouse effect, but with opacity varying with location instead of wavelength.

These effects from clouds change both the temperature structure and atmospheric dynamics (e.g. winds), leading to feedbacks that can only be captured by 3-D modeling.

GCM Grids

I spend a lot of my time running and analyzing grids of hot Jupiter GCMs to explore how different parameters and processes (esp. clouds) influence population-level trends. My first paper in grad school used one of these grids to explore how magnetic drag (resistance felt by ions being advected across magnetic field lines) interacts with clouds for planets across a range of temperatures. On the left, I show some of the data products from this work: temperature maps on isobars, with wind patterns overlaid. We found some feedback in the intermediate temperature range, where magnetic drag changes cloud coverage on both the nightside and the terminators (see Kennedy et al. 2025 for more details!).

I've recently submitted a paper on a second grid of GCMs, drilling into how the assumptions we make about clouds (size, composition, vertical extent, and formation efficiency) change their impact on atmospheres and observables. In this work, we used the strength of cloud feedback in our model to set limits on dayside cloud properties (driven by unphysical thermal inversions and too-high Bond albedos in cool hot Jupiters).

Limb Asymmetries with the RM-GCM and CARMA

I'm very interested in the novel power of JWST to disentangle the morning and evening limb contributions to the transmission spectrum, teasing out encoded 3-D information in a previously 1-D measurement. This is especially interesting for clouds, as we expect mornings and evenings to have different temperatures, and therefore different cloud properties. I have contributed a number of GCMs toward observation-focused limb asymmetry studies, and I am currently finishing up a forward-modelling study to interpret the measured JWST limb asymmetries on Kepler-12b using a combination of GCMs and a state-of-the-art cloud microphysics model (exoCARMA). We can explain the cloud properties of the spectra well with GCMs+CARMA models with Mg2SiO4 clouds, but only if the atmosphere is metal-poor (supporting retrieval results) and only when including the influence of the GCM-predicted equatorial jet! We also need to suppress the formation of corundum (Al2O3) clouds, an interesting tension with a number of potential physical implications.

RM-GCM Radiative Transfer Upgrade

Radiative transfer (RT) is hard to do properly in 3D, and is often the most computationally limiting step of a GCM. To deal with this, the field has developed a range of approximations for the the opacity function.

On the fast end, we have double-gray RT, which uses two opacities that are constant in wavelength: one for the starlight, and one for the planet-emitted ("thermal") light. This approach can help us understand the broad strokes of dynamics and energy transport, but loses the influence of lines and bands on heat transport. "Picket-fence" radiative transfer improves upon this by applying lines of equal strength and spacing across the spectrum for thermal channels, combined with any number of starlight channels (usually three). This can be tuned to match the results of more complex models fairly well, and is only about 2.5x slower than double-gray. When I joined the Rauscher group, the RM-GCM could take either of these approaches.

The gold-standard approximation of the opacity function for GCMs is the "correlated-k" method (Lacis & Oinas 1991). In correlated-k, you cut the spectrum up into bins and sort the opacities by strength within each bin before passing light through them. This sorting works because opacities in different conditions are strongly correlated in wavelength (i.e. lines don't move around), so the re-ordering always sends each wavelength to roughly the same place, regardless of temperature and pressure. Sorting the opacity function offers an important advantage-- you need significantly fewer samples of the smooth, sorted opacity function to capture its full dynamic range. This method can retain ~1%-level accuracy and is significantly faster than line-by line calculations, but is much slower than the picket-fence approach.

I added a correlated-k module into the RM-GCM, and parallelized the RT to offset the additional cost of this method, which now runs in the same wall-clock time as the picket-fence code on 32 cores. This upgrade will be released in version 6.0 of the RM-GCM shortly, and adds an extra tool to our kit for approaching more detailed studies of individual planets.

Page updated

Report abuse