Grantee Research Project Results

Final Report: Machine-learned atmospheric chemical mechanisms

EPA Grant Number: R840012
Title: Machine-learned atmospheric chemical mechanisms
Investigators: Tessum, Christopher , Riemer, Nicole
Institution: University of Illinois Urbana-Champaign
EPA Project Officer: Chung, Serena
Project Period: August 1, 2020 through July 31, 2023 (Extended to July 31, 2024)
Project Amount: $399,469
RFA: Chemical Mechanisms to Address New Challenges in Air Quality Modeling (2019) RFA Text | Recipients Lists
Research Category: Air , Air Quality and Air Toxics , Early Career Awards

Objective:

The objective of the work proposed here is to produce a machine-learned condensed chemical mechanism that provides accurate results in air quality modeling simulations under a wide variety of atmospheric conditions. The work builds on preliminary studies where we have created a neural-network-based chemical mechanism that operates with fewer chemical species and orders-of- magnitude less computational cost than the reference mechanism it was trained on, while reproducing diurnal cycles with low error in 90% of randomly initialized simulations. The project focused on reducing error in the remaining 10% of cases.

Summary/Accomplishments (Outputs/Outcomes):

We have created a machine-learned surrogate of a simplified chemical mechanism by (1) reducing dimensionality using singular value decomposition to create an interpretably-compressed low- dimensional latent space, and (2) using Sparse Identification of Nonlinear Dynamics (SINDy) to create an differential-equation-based representation of the underlying chemical dynamics in the compressed latent space with reduced stiffness.

We have coupled the near-explicit gas-phase Master Chemical Mechanism (MCM) with the state-of-art Particle-resolved Monte Carlo Model for Simulation Aerosol Interactions and Chemistry (PartMC-MOSAIC). The combined multiphase chemical mechanism is a detailed reference for emulation, and we have created datasets for training and testing. We have created a machine-learned surrogate of the gaseous mechanism (MCM) by (1) reducing dimensionality using a linear autoencoder to create an interpretably-compressed low-dimensional latent space, and (2) developing a novel algorithm we call Sparse Identification of Mass Action Dynamics (SIMADy) to create an reaction network-based representation of the underlying chemical dynamics in the compressed latent space ensuring numerical stability.

In other work that is partially supported by this project we have implemented a prototype of a SIMADy surrogate in the GEOS-Chem chemical transport model, we have created a probabilistic version of the SINDy-based atmospheric chemistry surrogate model, we have created a machine-learned surrogate advection operator, and we have begun work on a new framework for geoscientific modeling which is more amenable for integration with machine learning than are current FORTRAN-based methods.

Journal Articles on this Report : 2 Displayed | Download in RIS Format

Publications Views
Other project views:	All 15 publications	2 publications in selected types	All 2 journal articles

Publications
Type	Citation	Project	Document Sources
Journal Article	Kelp MM, Jacob DJ, Kutz JN, Marshall JD, Tessum CW. Toward stable, general machine-learned models of the atmospheric chemical system. Journal of Geophysical Research-Atmospheres 2020;125:e2020JD032759.	R840012 (2021) R840012 (2022) R840012 (2023) R840012 (Final) R835873 (2020) R835873 (Final) R840014 (2023) R840014 (Final)	Full-text: Wiley - HTML Exit Other: AGU - Full Text PDF Exit
Journal Article	Yang X, Guo L, Zheng Z, Riemer N, Tessum CW. Atmospheric chemistry surrogate modeling with sparse identification of nonlinear dynamics. Journal of Geophysical Research:Machine Learning and Computation. 2024;1(2):e2024JH000132.	R840012 (Final)	Full-text: AGU Full-text HTML Exit

Supplemental Keywords:

atmospheric chemistry, chemical mechanism, SINDy

Relevant Websites:

Machine-learned atmospheric chemical mechanisms Exit

Earth Science Machine Learning Exit

Progress and Final Reports:

Original Abstract

The perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.