Accelerating Radiative Transfer in 3D Simulations of the Venusian Atmosphere

This work introduces an approach to enhance the computational efficiency of 3D Global Circulation Model (GCM) simulations by integrating a machine-learned surrogate model into the OASIS GCM^[1]. Traditional GCMs, which are based on repeatedly numerically integrating physical equations governing atmospheric processes across a series of time-steps, are time-intensive, leading to compromises in spatial and temporal resolution of simulations. This research improves upon this limitation, enabling higher resolution simulations within practical timeframes.

Speeding up 3D simulations holds significant implications in multiple domains. Firstly, it facilitates the integration of 3D models into exoplanet inference pipelines, allowing for robust characterisation of exoplanets from a previously unseen wealth of data anticipated from post-JWST instruments^[2-3]. Secondly, acceleration of 3D models will enable higher resolution atmospheric simulations of Earth and Solar System planets, enabling more detailed insights into their atmospheric physics and chemistry.

This work builds upon previous efforts in both exoplanet science and Earth climate science to accelerate 3D atmospheric models. Prior work in exoplanet science primarily relies on extrapolating 1D models to approximate certain 3D atmospheric variations^[4-6], which often introduces both known and unknown biases. Earth climate science, benefiting from a wealth of high-resolution observational measurements, has had a different set of methods employed, namely machine-learned surrogate models trained on such observations^[7-8]. This work employs machine-learned surrogate model techniques, benchmarked in Earth climate science^[8], to be used in general planetary climate models.

Our method replaces the radiative transfer module in OASIS with a recurrent neural network-based model trained on simulation inputs and outputs. Radiative transfer is typically one of the slowest and lowest resolution components of a GCM, thus providing the largest scope for overall model speed-up. The surrogate model was trained and tested on the specific test case of the Venusian atmosphere, to benchmark the utility of this approach in the case of non-terrestrial atmospheres. This approach yields promising res­­ults, with the surrogate-integrated GCM demonstrating above 99.0% accuracy and 10 times CPU speed-up compared to the original GCM.

In conclusion, this work presents a method to accelerate 3D GCM simulations, offering a pathway to more efficient and detailed modelling of planetary atmospheres.

 

References:

[1] Mendonca J. & Buchhave L., ‘Modelling the 3D Climate of Venus with OASIS’, 2020, Volume 496, Pages 3512-3530, Monthly Notices of the Royal Astronomical Society

[2] Gardner, J. P., “The James Webb Space Telescope”, Space Science Reviews, vol. 123, no. 4, pp. 485–606, 2006. doi:10.1007/s11214-006-8315-7.

[3] Tinetti, G., “Ariel: Enabling planetary science across light-years”, arXiv e-prints, 2021. doi:10.48550/arXiv.2104.04824.

[4] Q. Changeat and A. Al-Refaie, “TauREx3 PhaseCurve: A 1.5D Model for Phase-curve Description,” Astrophys J, vol. 898, no. 2, p. 155, Aug. 2020, doi: 10.3847/1538-4357/ab9b82.

[5] K. L. Chubb and M. Min, “Exoplanet atmosphere retrievals in 3D using phase curve data with ARCiS: application to WASP-43b,” Jun. 2022, doi: 10.1051/0004-6361/202142800.

[6] P. G. J. Irwin et al., “2.5D retrieval of atmospheric properties from exoplanet phase curves: Application toWASP-43b observations,” Mon Not R Astron Soc, vol. 493, no. 1, pp. 106–125, Mar. 2020, doi: 10.1093/mnras/staa238.

[7] Yao, Y., Zhong, X., Zheng, Y., & Wang, Z. 2023, Journal of Advances in Modeling Earth Systems, 15, e2022MS003445, doi: 10.1029/2022MS003445

[8] Ukkonen, P. 2022, Journal of Advances in Modeling Earth Systems, 14, e2021MS002875, doi: 10.1029/2021MS002875