Sensitivity Analysis for Climate Science with Generative Flow Models

TL;DR

This paper introduces an efficient method for computing sensitivities in climate models using generative flow models and the adjoint state method, significantly reducing computational costs.

Contribution

It applies the adjoint state method to generative flow models for climate sensitivity analysis, enabling fast and reliable gradient computations.

Findings

Reliable gradients can be computed with generative models.

Computational cost reduced from weeks to hours.

Validated sensitivities against model outputs.

Abstract

Sensitivity analysis is a cornerstone of climate science, essential for understanding phenomena ranging from storm intensity to long-term climate feedbacks. However, computing these sensitivities using traditional physical models is often prohibitively expensive in terms of both computation and development time. While modern AI-based generative models are orders of magnitude faster to evaluate, computing sensitivities with them remains a significant bottleneck. This work addresses this challenge by applying the adjoint state method for calculating gradients in generative flow models. We apply this method to the cBottle generative model, trained on ERA5 and ICON data, to perform sensitivity analysis of any atmospheric variable with respect to sea surface temperatures. We quantitatively validate the computed sensitivities against the model's own outputs. Our results provide initial evidence that this approach can produce reliable gradients, reducing the computational cost of sensitivity analysis from weeks on a supercomputer with a physical model to hours on a GPU, thereby simplifying a critical workflow in climate science. The code can be found at https://github.com/Kwartzl8/cbottle_adjoint_sensitivity.