Toward Stable, General Machine‐Learned Models of the Atmospheric Chemical System

• Published in: Journal of geophysical research. Atmospheres 2020-12, Vol.125 (23), p.n/a
• Main Authors: Kelp, Makoto M., Jacob, Daniel J., Kutz, J. Nathan, Marshall, Julian D., Tessum, Christopher W.
• Format: Article
• Language: English
• Subjects: Accumulation; Air pollution; atmospheric chemical mechanism; Atmospheric chemistry; Atmospheric chemistry models; Atmospheric conditions; Atmospheric models; chemical mechanism; Chemical speciation; Climate change; Climate models; Computer applications; Error reduction; Geophysics; Initial conditions; Learning algorithms; Machine learning; model emulation; Modelling; Ozone; Particulate matter; Random noise; Simulation; Solvers; surrogate model; Suspended particulate matter; Training
• ISSN: 2169-897X; 2169-8996
• DOI: 10.1029/2020JD032759

Atmospheric chemistry models—components in models that simulate air pollution and climate change—are computationally expensive. Previous studies have shown that machine‐learned atmospheric chemical solvers can be orders of magnitude faster than traditional integration methods but tend to suffer from numerical instability. Here, we present a modeling framework that reduces error accumulation compared to previous work while maintaining computational efficiency. Our approach is novel in that it (1) uses a recurrent training regime that results in extended (>1 week) simulations without exponential error accumulation and (2) can reversibly compress the number of modeled chemical species by >80% without further decreasing accuracy. We observe an ~260× speedup (~1,900× with specialized hardware) compared to the traditional solver. We use random initial conditions in training to promote general applicability across a wide range of atmospheric conditions. For ozone (concentrations ranging from 0–70 ppb), our model predictions over a 24‐hr simulation period match those of the reference solver with median error of 2.7 and