Suêtes: An end-to-end differentiable non-hydrostatic limited-area dynamical core
Abstract
We present Suêtes, a fully differentiable, non-hydrostatic, limited-area atmospheric dynamical core implemented in JAX. Suêtes includes both a semi-implicit semi-Lagrangian scheme and a Split-Explicit Runge-Kutta scheme with acoustic substepping, enabling applications from coarse regional simulations to convection-permitting dynamics. Its terrain-following geometry, spatial discretization, time integration, initial conditions, lateral boundary treatment, and included physical parameterizations are all compatible with reverse-mode automatic differentiation. A single reverse-mode pass can therefore compute sensitivities of forecast diagnostics to selected initial and boundary fields, physical parameters, topography, and computational geometry. We provide a series of numerical experiments---including a tracer-source inverse problem, a sensitivity analysis of a three-dimensional squall-line, adversarial perturbation construction an ERA5-driven downslope hurricane force wind event, topography optimization, and terrain-following coordinate optimization---demonstrating the versatility of Suêtes as a framework for gradient-based diagnosis and model development at regional and convection-permitting scale.
Related
- Tropical Cyclone-Following Dataset for Limited-Area Machine Learning Applications
- Meso-NH-ISO v1.0: a water stable isotopologue scheme in the non-hydrostatic mesoscale atmospheric model Meso-NH – application to a 2D West African squall line
- Phase-resolved coastal wave modelling using a non-hydrostatic multilayer formulation
- DiffSWE2d: a differentiable Shallow Water Equations solver for end-to-end flood and tsunami modelling
- A library for differentiable signal processing and machine learning on the sphere
BibTeX
@misc{whittaker2026suêtes,
title = {{Suêtes: An end-to-end differentiable non-hydrostatic limited-area dynamical core}},
author = {Tim Whittaker and Seth Taylor and Elsa Cardoso-Bihlo and Alejandro Di Luca and Alex Bihlo},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.36381},
eprint = {2609.36381},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.36381},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Suêtes: An end-to-end differentiable non-hydrostatic limited-area dynamical core
AU - Tim Whittaker
AU - Seth Taylor
AU - Elsa Cardoso-Bihlo
AU - Alejandro Di Luca
AU - Alex Bihlo
PY - 2026
DA - 2026/09/28
PB - arXiv
DO - 10.48550/arxiv.2609.36381
UR - https://arxiv.org/abs/2609.36381
AB - We present Suêtes, a fully differentiable, non-hydrostatic, limited-area atmospheric dynamical core implemented in JAX. Suêtes includes both a semi-implicit semi-Lagrangian scheme and a Split-Explicit Runge-Kutta scheme with acoustic substepping, enabling applications from coarse regional simulations to convection-permitting dynamics. Its terrain-following geometry, spatial discretization, time integration, initial conditions, lateral boundary treatment, and included physical parameterizations are all compatible with reverse-mode automatic differentiation. A single reverse-mode pass can therefore compute sensitivities of forecast diagnostics to selected initial and boundary fields, physical parameters, topography, and computational geometry. We provide a series of numerical experiments---including a tracer-source inverse problem, a sensitivity analysis of a three-dimensional squall-line, adversarial perturbation construction an ERA5-driven downslope hurricane force wind event, topography optimization, and terrain-following coordinate optimization---demonstrating the versatility of Suêtes as a framework for gradient-based diagnosis and model development at regional and convection-permitting scale.
ER -