Research

Projects

Our current work builds AI-enabled tools for understanding and predicting the climate system. Earlier projects use machine learning and satellite data to study aerosol–cloud interactions and to emulate climate models.

Current projects

MOSAIC

An agentic scientific pipeline for discovering interpretable climate physics from observations.

  • machine learning
  • climate modeling
  • equation discovery
  • observations

JAX-GCM

A fully differentiable atmospheric model spanning fast SPEEDY physics to full ECHAM physics with online aerosol

  • machine learning
  • climate modeling
  • differentiable
  • aerosols

GAIA Initiative

Harnessing AI to understand, forecast, and steward Earth's complex systems

  • machine learning
  • earth science
  • ocean science

ClimateBench2

A probabilistic benchmark for evaluating physics-based, data-driven and hybrid climate models on an equal footing.

  • machine learning
  • climate modeling
  • benchmarking

Earlier work

Detecting ship-tracks

Using machine learning to detect ship-tracks in satellite imagery and quantify how shipping emissions brighten clouds.

  • aerosols
  • shallow clouds
  • satellite
  • machine learning

Ship-track viewer

An interactive map of every ship-track detected by our machine learning algorithm, with MODIS timestamps for each track.

  • aerosols
  • shallow clouds
  • satellite

Pockets of open cells

A deep-learning census of 8,491 pockets of open cells across the Californian, Peruvian and Namibian stratocumulus decks.

  • aerosols
  • shallow clouds
  • satellite
  • machine learning

Climate model emulation

Developing climate model emulators for better parameter estimation and calibration.

  • emulation
  • climate modeling

ClimateBench v1

Explore future warming under different emissions scenarios with an online emulator built on ClimateBench v1.

  • emulation
  • benchmarking
  • machine learning

Browse by topic