Accelerating models for multiphase chemical kinetics through machine learning with polynomial chaos expansion and neural networks

Abstract

Abstract. The heterogeneous chemistry of atmospheric aerosols involves multiphase chemical kinetics that can be described by kinetic multi-layer models (KM) explicitly resolving mass transport and chemical reaction. However, KM are computationally too expensive to be used as sub-modules in large-scale atmospheric models, and the computational costs also limit their utility in inverse modelling approaches commonly used to infer aerosol kinetic parameters from laboratory studies. In this study, we show how machine learning methods can generate inexpensive surrogate models based on the kinetic multi-layer model of aerosol surface and bulk chemistry (KM-SUB). We apply and compare two common and openly available methods for the generation of surrogate models, polynomial chaos expansion (PCE) with UQLab and neural networks (NN) through the Python package Keras. We show that the PCE method is well-suited to determine global sensitivity indices of the KM and demonstrate how inverse modelling applications can be enabled or accelerated with NN-suggested sampling. These qualities make them suitable supporting tools for laboratory work in the interpretation of data and design of future experiments. Overall, the KM surrogate models investigated in this study are fast, accurate, and robust, which suggests their applicability as sub-modules in large-scale atmospheric models.

Unveiling the Role of Carbonate Radical Anions in Dust‐Driven SO2 Oxidation

“…In the atmosphere, carbonate radical initiated SO 2 oxidation scheme over a real system metric will be more complicated under diverse variable environment conditions (e.g., RH, light intensity, and accumulating growth of CO 2 concentration) and dust constituents, which is a challenging task. Fortunately, machine‐learning (ML) provides a promising approach to accelerate the prediction of atmospheric chemical characteristics and aerosol constituents (Berkemeier et al., 2023; J. Gao et al., 2022; Qin et al., 2022; Song et al., 2021) through efficient data training and analysis. Following this reasoning, we applied this promising method to reduce the large human resource cost coming from the traditional control variate measurement strategy and to reveal the fundamental driver of significance in triggering the fast sulfate production from multitudinous environmental variables and constituents by the established “train” model.…”

“…in affecting the SO 2 uptake capability in the presence of CO 2 . Fortunately, the booming machine‐learning (ML) approach has been reported to have great potential to predict atmospheric chemicals characteristic and secondary aerosol constituents (Berkemeier et al., 2023; J. Gao et al., 2022; Qin et al., 2022; Song et al., 2021). Additionally, this economic method can dramatically save the cost of labor source and reduce the uncertainties coming from the conventional control variate measurement, and elucidate the nonlinear relationship between the variable feature and considered outcome, making it possible to reveal the fundamental drivers (environmental conditions or constituents) of significance in determining the SO 2 uptake capability from massive quantities of the complicated lab‐based data set of multitudinous characteristics (T. Wang et al., 2023).…”

Acoustic levitation with polarising optical microscopy (AL-POM): water uptake in a nanostructured atmospheric aerosol proxy

“…4 are possible with kinetic multi-layer models, which have become the state-of-the-art in aerosol science. 34,[36][37][38]54 AL-POM only probes birefringent phases, therefore it is specic and could be used for chemical kinetics in a similar way to the use of SAXS as a tool to measure reaction kinetics of specic self- organised structures. 21 Application of a kinetic multi-layer model to those kinetic data could then provide insights into the real-world impact of a particular phenomenon (e.g.…”

A kinetic compass for the design of experiments to determine kinetic parameters

“…A SM can be used to substitute the template model in applications that benefit from low computational cost in exchange for slightly increased model uncertainty. Satisfactory model accuracy can be ensured by a sufficient size of the training data set, and therefore depends on the initial investment of computational resources [20]. SM have helped solving the issue of computational cost in many fields of research, e.g.…”

“…SM can also aid inverse modelling approaches. Berkemeier et al 2023 [20] showed that SM-supported fit ensemble acquisition greatly outperforms regular sampling with the kinetic multi-layer model of aerosol surface and bulk chemistry (KM-SUB) [5] in terms of acquired fits for a given computational effort. However, it remains unclear how SM uncertainty affects the reliability of inverse modelling techniques.…”