Atmospheric Chemistry Surrogate Modeling With Sparse Identification of Nonlinear Dynamics

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 2
• Main Authors: Yang, Xiaokai, Guo, Lin, Zheng, Zhonghua, Riemer, Nicole, Tessum, Christopher W.
• Format: Journal Article
• Language: English
• Published: 01.06.2024
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000132

Modeling atmospheric chemistry is computationally expensive and limits the widespread use of chemical transport models. This computational cost arises from solving high‐dimensional systems of stiff differential equations. Previous work has demonstrated the promise of machine learning (ML) to accelerate air quality model simulations but has suffered from numerical instability during long‐term simulations. This may be because previous ML‐based efforts have relied on explicit Euler time integration—which is known to be unstable for stiff systems—and have used neural networks which are prone to overfitting. We hypothesize that parsimonious models combined with modern numerical integration techniques can overcome this limitation. Using a small‐scale mechanism to explore the potential of these methods, we have created a machine‐learned surrogate by (a) reducing dimensionality using singular value decomposition to create an interpretably‐compressed low‐dimensional latent space and (b) using Sparse Identification of Nonlinear Dynamics (SINDy) to create a differential‐equation‐based representation of the underlying dynamics in the compressed latent space with reduced stiffness. The root mean square error of ML model prediction for ozone concentration over 9 days is 37.8% of the root mean square concentration across all simulations in our testing data set. The surrogate model is 10× faster with 12× fewer integration timesteps compared to reference model and is numerically stable in all tested simulations. Overall, we find that SINDy can be used to create fast, stable, and accurate surrogates of a simple photochemical mechanism. In future work, we will explore the application of this method to more detailed mechanisms and their use in large‐scale simulations. Plain Language Summary Atmospheric chemistry modeling is computationally expensive due to complex dynamic processes among numerous chemical species. Machine learning techniques have the potential to accelerate these models. We mathematically group the chemical species to reduce their number and apply a machine‐learning algorithm based on sparse regression (SINDy) to create a fast, stable, and accurate surrogate model for a simplified photochemical mechanism. The machine‐learned surrogate model is 10× faster than the reference model and is numerically stable in all tested nine‐day simulation cases. Key Points We apply the sparse identification of nonlinear dynamics to create a machine‐learned surrogate model for an atmospheric chemical mechanism We train the model to learn the chemical dynamics in a compressed latent space with 10× speedup overall The surrogate model can make stable predictions for all modeled chemical species without noticeable error accumulation