Calibrating subgrid parametrizations of single-column ocean models via simulation-based inference
Abstract
Subgrid parametrizations of vertical mixing in ocean models depend on free coefficients that cannot be measured directly and must be calibrated against high-fidelity references such as large-eddy simulations (LES). Existing approaches return point estimates and leave the associated uncertainty unquantified, a limitation when the inverse problem is ill-posed or when distinct parameter configurations fit the data comparably well. Simulation-based inference (SBI) addresses exactly this: given a prior and access to the simulator, it approximates the full posterior over parameters without requiring a tractable likelihood, at a cost set by the number of simulator evaluations. We apply it to \texttt{tunax}, a JAX-based single-column ocean model, to calibrate the coefficients of its k--ε closure. A blockwise PCA summary statistic compresses the simulator output along the depth axis while preserving its forcing--horizon--variable structure, making inference tractable at modest budgets. We compare neural posterior estimation and its sequential variants against a recent training-free approach built on a tabular foundation model. The latter recovers informative posteriors from a few hundred simulator calls, outperforming the trained estimators at every budget considered.
Related
- Subgrid parametrizations mediate the large‐scale effects of soil macroporosity on the water cycle in the ORCHIDEE land surface model
- Building and calibrating a 3Di subgrid flood model for Hanoi, without a surveyed terrain model, drainage data, or rain gauges
- A Likelihood‐Free Framework for Seismic Forecasting Models: Application to Incomplete ETAS Model via Simulation‐Based Inference
- Improving low-cost NO₂ sensor reliability under real-world air quality conditions by calibrating distributions instead of data points
- Evaluating the prediction of subgrid eddy kinetic energy from a machine learning model
BibTeX
@misc{cabezas2026calibrating,
title = {{Calibrating subgrid parametrizations of single-column ocean models via simulation-based inference}},
author = {Luben M. C. Cabezas and Sacha Wendling and Aurèle Gallard and Gabriel Mouttapa and Julien Le Sommer and Pedro L. C. Rodrigues},
howpublished = {HAL},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.13242},
eprint = {2609.13242},
archivePrefix = {arXiv},
url = {https://hal.science/hal-05755180v1},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Calibrating subgrid parametrizations of single-column ocean models via simulation-based inference
AU - Luben M. C. Cabezas
AU - Sacha Wendling
AU - Aurèle Gallard
AU - Gabriel Mouttapa
AU - Julien Le Sommer
AU - Pedro L. C. Rodrigues
PY - 2026
DA - 2026/09/03
PB - HAL
DO - 10.48550/arxiv.2609.13242
UR - https://hal.science/hal-05755180v1
AB - Subgrid parametrizations of vertical mixing in ocean models depend on free coefficients that cannot be measured directly and must be calibrated against high-fidelity references such as large-eddy simulations (LES). Existing approaches return point estimates and leave the associated uncertainty unquantified, a limitation when the inverse problem is ill-posed or when distinct parameter configurations fit the data comparably well. Simulation-based inference (SBI) addresses exactly this: given a prior and access to the simulator, it approximates the full posterior over parameters without requiring a tractable likelihood, at a cost set by the number of simulator evaluations. We apply it to \texttt{tunax}, a JAX-based single-column ocean model, to calibrate the coefficients of its k--ε closure. A blockwise PCA summary statistic compresses the simulator output along the depth axis while preserving its forcing--horizon--variable structure, making inference tractable at modest budgets. We compare neural posterior estimation and its sequential variants against a recent training-free approach built on a tabular foundation model. The latter recovers informative posteriors from a few hundred simulator calls, outperforming the trained estimators at every budget considered.
ER -