Research Projects
Research Projects
Embedded Files
Development of an automatically-differentiable climate model with Python and JAX
Development of an automatically-differentiable climate model with Python and JAX
The JAX Circulation Model (JCM) is an intermediate-complexity, fully differentiable atmospheric model. It is written in Python and JAX, making it accessible to a broad range of developers, scientists, and educators. This project is free, open-source, and intended for a wide range of applications, building on broader community efforts to make differentiable models available to all. See our GitHub page to find out more! We are collaborating with others to create a Python and JAX model for the entire Earth System.
Code: https://github.com/climate-analytics-lab/jax-gcm
Paper: https://gmd.copernicus.org/articles/19/6451/2026/
Can velocity observations improve predictions of the Tropical Pacific Ocean?
Can velocity observations improve predictions of the Tropical Pacific Ocean?
In this work we assimilate velocity observations from the NOAA Tropical Atmosphere Ocean (TAO) mooring array into a regional simulation of the tropical Pacific Ocean. The goal of this study is to identify whether velocity constraints can improve predictions and simulations of the equatorial Pacific Ocean. We use these results to learn about the predictability of various processes and to identify targeted observations that will further improve Pacific forecasts. The MITgcm assimilation configuration files can be found here.
What processes drive vertical momentum flux in the Pacific Cold Tongue region?Â
What processes drive vertical momentum flux in the Pacific Cold Tongue region?Â
The equatorial Pacific Ocean supports basin-scale waves and strong air-sea coupling. The large scales and strong fluxes mean that this part of the ocean has a significant impact on global weather and climate. In this work we use a data assimilating model, known as an ocean state estimate, to diagnose and understand the processes that move momentum around the equatorial ocean. The scales of these phenomena range from thousands of kilometers to meters, making them challenging to observe and model. Data assimilation works to combine observations with models to create a "best guess" of how our ocean is evolving. Using these tools we investigate the role of internal gravity waves in closing the Cold Tongue momentum budget in our recently published paper in the Journal of Geophysical Research: Oceans.
How can we use side-scan sonar to localize underwater vehicles?
How can we use side-scan sonar to localize underwater vehicles?
Autonomous underwater vehicles (AUVs) are necessary to collect ocean data, monitor coastal environments, and perform labor intensive tasks at sea (e.g., aquaculture monitoring, ocean cleanup, etc). GPS, radio, and camera signals do not penetrate through water, making it challenging to navigate these vehicles. In this project we explored the use of side-scan sonar to autonomously localize an underwater vehicle, given a map of landmarks on the sea floor. They key to our method is the use of an entirely probabilistic framework that is robust to false detections or missed landmarks. See the details of the algorithm and the results of our field tests in Davenport, et al. (2025) which is published in the IEEE Journal of Oceanic Engineering.
Page updated
Google Sites
Report abuse