91̽��

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.

Abstract:

Plain Language Summary: Aerosols, small particles suspended in the atmosphere, influence cloud properties by acting as nuclei for cloud droplet formation. These aerosol‐cloud interactions (ACI) introduce uncertainties in climate research, making it essential to improve our understanding of them. This paper presents findings from a model intercomparison project that examines the impact of aerosols on clouds and climate in simulations that directly represent cloud processes under idealized equilibrium climate conditions. We show that cloud responses to aerosols vary substantially across models, though certain consistent responses emerge. Specifically, increased aerosol loading generally suppresses initial rain formation, which in turn alters the thermodynamic conditions of the atmosphere. We also discuss how these thermodynamic changes influence the large‐scale atmospheric circulation.

Abstract:

To study and monitor the Sun and its atmosphere, various space missions have been launched in the past decades. With rapid improvement in technology and different mission requirements, the data products are subject to constant change. However, for such long-term studies as solar variability or multi-instrument investigations, uniform data series are required. In this study, we built on and expanded the instrument-to-instrument translation (ITI) framework, which provides unpaired image translations. We applied the tool to data from the Extreme Ultraviolet Imager (EUI), specifically the Full Sun Imager (FSI) on Solar Orbiter and the Atmospheric Imaging Assembly (AIA) on the Solar Dynamics Observatory (SDO). This approach allowed us to create a homogeneous dataset that combines the two extreme ultraviolet (EUV) imagers in the 174/171 Å and 304 Å channels. We demonstrate that ITI is able to provide image calibration between Solar Orbiter and SDO EUV imagers, independent of the varying orbital position of Solar Orbiter. The comparison of the intercalibrated light curves derived from 174/171 Å and 304 Å filtergrams from EUI and AIA shows that ITI can provide uniform data series that outperform a standard baseline calibration. We evaluate the perceptual similarity in terms of the Fréchet inception distance, which demonstrates that ITI achieves a significant improvement of perceptual similarity between EUI and AIA. The study provides intercalibrated observations from Solar Orbiter/EUI/FSI with SDO/AIA, enabling a homogeneous dataset suitable for solar cycle studies and multi-viewpoint investigations.