An Online‐Learned Neural Network Chemical Solver for Stable Long‐Term Global Simulations of Atmospheric Chemistry

Kelp, Makoto; Jacob, Daniel; Lin, Haipeng; Sulprizio, Melissa P.

doi:10.1029/2021ms002926

Cited by 22 publications

(17 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Maintaining numerical stability with the use of emulators for atmospheric chemistry and other atmospheric parameterizations is a known issue and initial steps have been taken to address it (Brenowitz & Bretherton, 2018;Kelp et al, 2020Kelp et al, , 2022. These recent studies found some performance improvements by using a "recurrent training" scheme, where a model was rolled out in time for n time steps during training, and a loss was calculated on Note.…”

Section: Discussionmentioning

confidence: 99%

“…It could be implemented as is into a 3D model with an operator splitting method: transport, advection, etc,̇ that are solved with traditional ODE solvers and the chemistry is solved with the GRU. This is, for instance, the approach adopted by Keller and Evans (2019) for solving ozone chemistry with random forests in GEOS-Chem and by Kelp et al (2022) with MLPs.…”

Section: Incorporation Of ML Models Into 3d Modelsmentioning

confidence: 99%

“…The Average Time Each Model Required to Advance 300 s of Simulation Time the sum of n time steps, instead of a single time step. However, the models used in these studies were not RNNs, as the network architectures were that of an MLP (Brenowitz & Bretherton, 2018) and an encoder/decoder framework (Kelp et al, 2020(Kelp et al, , 2022, which only utilized feed-forward connections. Rather, training these models relied on the multi-time step loss function as a means to update the model weights using a sequence instead of a single length input.…”

Section: Tablementioning

confidence: 99%

See 2 more Smart Citations

Neural Network Emulation of the Formation of Organic Aerosols Based on the Explicit GECKO‐A Chemistry Model

Schreck

Becker

Gagne

et al. 2022

J Adv Model Earth Syst

View full text Add to dashboard Cite

Secondary organic aerosols (SOA) are formed from oxidation of hundreds of volatile organic compounds (VOCs) emitted from anthropogenic and natural sources. Accurate predictions of this chemistry are key for air quality and climate studies due to the large contribution of organic aerosols to submicron aerosol mass. Currently, only explicit models, such as the Generator for Explicit Chemistry and Kinetics of Organics in the Atmosphere (GECKO‐A), can fully represent the chemical processing of thousands of organic species. However, their extreme computational cost prohibits their use in current chemistry‐climate models, which rely on simplified empirical parameterizations to predict SOA concentrations. This study demonstrates that machine learning can accurately emulate SOA formation from an explicit chemistry model with an approximate error of 2%–8%, up to five days for several precursors and for potentially up to one month for recurrent neural network models, and with 100 to 100,000 times speedup over GECKO‐A, making it computationally useable in a chemistry‐climate model. We generated the training data using thousands of GECKO‐A box simulations sampled from a broad range of initial environmental conditions, and focused on three representative SOA precursors: the oxidation by OH of two anthropogenic (toluene, dodecane), and the oxidation by O3 of one biogenic VOC (α‐pinene). We compare several neural models and quantify their underlying uncertainty and robustness. These are promising results, suggesting that neural network models could be applied to predict SOA in chemistry‐climate models, limited however to the range of environmental conditions that were considered in the training datasets.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Incorporation Of ML Models Into 3d Modelsmentioning

confidence: 99%

Section: Tablementioning

confidence: 99%

See 1 more Smart Citation

Neural Network Emulation of the Formation of Organic Aerosols Based on the Explicit GECKO‐A Chemistry Model

Schreck

Becker

Gagne

et al. 2022

J Adv Model Earth Syst

View full text Add to dashboard Cite

show abstract

“…Machine learning (ML) approaches are an emerging technique for decreasing the computational burden of earth system models with more e cient ML parameterizations, but have documented challenges such as unstable error growth and physical inconsistency which can happen when predicted recurrently (Kelp et al, 2018) or when interacting with other processes in the context of larger models (Rasp et al, 2018;Brenowitz & Bretherton, 2019). One approach towards ML models that can stably interact with other model processes is online training: parameter optimization of ML surrogates while running the entire model (Rasp, 2020;Kelp et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

Advecting Superspecies: Efficiently Modeling Transport of Organic Aerosol with a Mass-Conserving Dimensionality Reduction Method

Sturm¹,

Manders²,

Janssen³

et al. 2023

Preprint

View full text Add to dashboard Cite

The chemical transport model LOTOS-EUROS uses a volatility basis set (VBS) approach to represent the formation of secondary organic aerosol (SOA) in the atmosphere. Inclusion of the VBS approximately doubles the dimensionality of LOTOS-EUROS and slows computation of the advection operator by a factor of two. This complexity limits SOA representation in operational forecasts. We develop a mass-conserving machine learning (ML) method based on matrix factorization to find latent patterns in the VBS tracers that correspond to a lower-dimension set of superspecies. Tracers are reversibly compressed to superspecies before transport, and the superspecies are subsequently decompressed to tracers for process-based SOA modeling. This physically interpretable ML method conserves the total concentration and phase of the tracers throughout the process. The superspecies approach is implemented in LOTOS-EUROS and found to accelerate the advection operator by a factor of 1.5 to 1.8. Concentrations remain numerically stable over model simulation times of two weeks, including simulations at higher spatial resolutions than the ML models were trained on. Results from this case study show that this method can be used to enable detailed, process-based secondary organic aerosol representation in air quality operational forecasts in a computationally efficient manner. Beyond this case study, the physically consistent ML approach developed in this work enforces conservation laws that are essential to other Earth system modeling applications, and generalizes to other processes where computational benefit can be gained from a two-way mapping between detailed process variables and their representation in a reduced-dimensional space.

show abstract

“…They implemented it in GEOS-Chem (Kelp et al, 2022), achieving stable one-year simulations for ozone prediction with less than 10 % bias compared to the reference and reducing computational times by a factor of five. The motivation of these previous studies stems from reducing the costs of calculating chemistry, that is usually taking from 50 % to 90 % of the computational costs of running global chemistry models such as GEOS-Chem (Keller & Evans, 2019).…”

Section: Introductionmentioning

confidence: 99%

Towards emulating an explicit organic chemistry mechanism with random forest models

Mouchel‐Vallon

Hodzic

2022

Preprint

View full text Add to dashboard Cite

Predicting secondary organic aerosol (SOA) formation relies either on extremely detailed, numerically expensive models accounting for the condensation of individual species or on extremely simplified, numerically affordable models parameterizing SOA formation for large-scale simulations. In this work, we explore the possibility of creating a random forest to reproduce the behavior of a detailed atmospheric organic chemistry model at a fraction of the numerical cost. A comprehensive dataset was created based on thousands of individual detailed simulations, randomly initialized to account for the variety of atmospheric chemical environments. Recurrent random forests were trained to predict organic matter formation from dodecane and toluene precursors, and the partitioning between gas and particle phases. Validation tests show that the random forests perform well without any divergence over 10 days of simulations. The distribution of errors shows that the sampling of initial conditions for the training simulations needs to focus on chemical regimes where SOA production is the most sensitive. Sensitivity tests show that specializing multiple random forests for a specific chemical regime is not more efficient than training a single general random forest for the entire dataset. The most important predictors are those providing information about the chemical regime, oxidants levels and existing organic mass. The choice of predictors is crucial as using too many unimportant predictors reduces the performances of the random forests.

show abstract

An Online‐Learned Neural Network Chemical Solver for Stable Long‐Term Global Simulations of Atmospheric Chemistry

Cited by 22 publications

References 58 publications

Neural Network Emulation of the Formation of Organic Aerosols Based on the Explicit GECKO‐A Chemistry Model

Neural Network Emulation of the Formation of Organic Aerosols Based on the Explicit GECKO‐A Chemistry Model

Advecting Superspecies: Efficiently Modeling Transport of Organic Aerosol with a Mass-Conserving Dimensionality Reduction Method

Towards emulating an explicit organic chemistry mechanism with random forest models

Contact Info

Product

Resources

About