A climatology of low clouds over the tropical Atlantic
data processing and analysis from Barbados
In progress
Alarming trend in Earth's net energy imbalance
using dynamical regimes in space to isolate processes at play
The CERES satellite record (2000-2026) of monthly global radiative fluxes shows a worrying trend in Earth's top-of-atmosphere energy imbalance (EEI). To explain this and to predict how it might evolve in the coming decades, we partition the globe into dynamics-based regimes to facilitate a process-based investigation of contributors to the global mean trend.
We allocate importance to all regimes, and we note that the trends in the storm tracks and regions of large scale subsidence are primarily cloud-driven, with the SST pattern dictating stratocumulus feedbacks in the subsidence regions and with poleward storminess shifts controlling the trends in the midlatitudes (plus an aerosol contribution). In regions with climatological convection, we highlight the longwave clear-sky trend which is to first order spatially correlated with water vapor trends. The Antarctic sea ice melt, in regions with higher incoming shortwave radiation and lower endemic cloudiness, is also a driver of the global trend.
Machine learning (ML) on molecular dynamics (MD) simulations
Can graph neural networks (GNNs) replace MD?
Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) simulations of bidisperse sheared granular systems are computationally intensive, and using ML to predict displacements would save time and resources. This project explored the use of GNNs to predict horizontal displacements of particles.
Model development in PyTorch involved data processing, model design, and hyperparameter tuning, and the work was done under Professor Amy Graves in collaboration with Professor Dane Morgan, Ajay Annamareddy, and Siddhant Ranka.
Julia software development for climate modeling
Open-source implementation of ice-phase microphysics scheme
CliMA is an organization led by Professor Tapio Schneider at Caltech and MIT which has developed an open-source, ML-informed climate model in Julia. I worked in the microphysics group under Anna Jaruga to kickstart the development of a 2-moment, bulk, single-category hydrometeor scheme: the predicted particle properties (P3) scheme.
This involved not only implementing the code in Julia but also understanding the physics of the scheme, writing tests and documentation, and testing it in a 1-dimensional column model. Given the importance of clouds in determining Earth's energy balance, and given the major theoretical and observational uncertainties in cloud microphysics, this sort of work is crucial for improving our understanding of the climate system.
NSF REU at Washington State University Lab for Atmospheric Research (LAR)
Air Quality, Atmospheric Chemistry, and Climate Change: Measurements and Modeling in the Pacific Northwest
Although the programming of this REU was based in atmospheric science, I worked with Fabio Scarpare on a project in the College of Agricultural, Human, and Natural Resources Sciences (CAHNRS) to study the effects of agricultural practices on soil hydrology in the context of Brazilian sugarcane cultivation.
The research involved calibration and testing of an in-house agricultural model, CropSyst. In RStudio, I worked with aboveground biomass, soil water content, and leaf area index observations in dialogue with multiyear model output. In the context of aridification and climate change in Brazil's agricultural wetlands, these efforts to understand the effect of surface crop residue on soil hydrology are important for sugarcane agriculture in the region.
