legoESM: a modular, differentiable, multiscale, AI-ready Earth system model built with AI agents
Abstract
Earth system models (ESMs) have grown tremendously in realism, yet key uncertainties persist in the climate response to greenhouse-gas forcing, particularly due to cloud radiative feedbacks. In addition, their software architecture was not designed for accelerator hardware or modern artificial intelligence (AI). Here we present legoESM, a composable, differentiable, multiscale ESM written in JAX. It builds on decades of community-developed parameterizations and numerical methods, recast in a unified framework by AI coding agents under a human-specified scientific contract and verified through benchmarking. Dynamical cores, physics schemes, grids, complexity levels and components are swappable like building blocks, and can use conventional physics or machine-learned emulators. A single code base spans metre-scale large-eddy simulation to global simulations and weather to climate. End-to-end differentiability enables gradient-based calibration, variational data assimilation and online training. legoESM modular architecture enables systematic evaluation of diverse model variants to explore structural uncertainty and test hypotheses. legoESM produces realistic simulations across scales, reduces land-surface temperature bias through gradient-based calibration, and scales efficiently on GPUs to kilometer-scale simulations. It offers an open, community infrastructure for hypothesis testing, research and teaching in Earth sciences and a template for multiscale physical systems.
Related
- Can LLM Agents Improve Imperfect Earth System Model Parameterizations?
- DiffSWE2d: a differentiable Shallow Water Equations solver for end-to-end flood and tsunami modelling
- Suêtes: An end-to-end differentiable non-hydrostatic limited-area dynamical core
- A library for differentiable signal processing and machine learning on the sphere
- DiffSpecMatch: Differentiable Waveform Inverse Design with Configurable Optimization Modules for Response-Spectrum Matching of Ground Motions
BibTeX
@misc{gentine2026legoesm,
title = {{legoESM: a modular, differentiable, multiscale, AI-ready Earth system model built with AI agents}},
author = {Pierre Gentine and Dhruv Balwada and Aytaç Paçal and Linnia Hawkins and Alistair Adcroft and Hang Fan and Jianing Fang and Aya Lahlou and Kara D. Lamb and David M. Lawrence and Julien Le Sommer and Joseph Mouallem and Juan Nathaniel and Duncan Watson-Parris and Veronika Eyring},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.11883},
eprint = {2610.11883},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.11883},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - legoESM: a modular, differentiable, multiscale, AI-ready Earth system model built with AI agents
AU - Pierre Gentine
AU - Dhruv Balwada
AU - Aytaç Paçal
AU - Linnia Hawkins
AU - Alistair Adcroft
AU - Hang Fan
AU - Jianing Fang
AU - Aya Lahlou
AU - Kara D. Lamb
AU - David M. Lawrence
AU - Julien Le Sommer
AU - Joseph Mouallem
AU - Juan Nathaniel
AU - Duncan Watson-Parris
AU - Veronika Eyring
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.11883
UR - https://arxiv.org/abs/2610.11883
AB - Earth system models (ESMs) have grown tremendously in realism, yet key uncertainties persist in the climate response to greenhouse-gas forcing, particularly due to cloud radiative feedbacks. In addition, their software architecture was not designed for accelerator hardware or modern artificial intelligence (AI). Here we present legoESM, a composable, differentiable, multiscale ESM written in JAX. It builds on decades of community-developed parameterizations and numerical methods, recast in a unified framework by AI coding agents under a human-specified scientific contract and verified through benchmarking. Dynamical cores, physics schemes, grids, complexity levels and components are swappable like building blocks, and can use conventional physics or machine-learned emulators. A single code base spans metre-scale large-eddy simulation to global simulations and weather to climate. End-to-end differentiability enables gradient-based calibration, variational data assimilation and online training. legoESM modular architecture enables systematic evaluation of diverse model variants to explore structural uncertainty and test hypotheses. legoESM produces realistic simulations across scales, reduces land-surface temperature bias through gradient-based calibration, and scales efficiently on GPUs to kilometer-scale simulations. It offers an open, community infrastructure for hypothesis testing, research and teaching in Earth sciences and a template for multiscale physical systems.
ER -