Mimicking atmospheric photochemical modeling with a deep neural network

Xing, Jia; Zheng, Shuxin; Li, Siwei; Huang, Lin; Wang, Xiaochun; Kelly, James T.; Liu, Chang; Jang, Carey; Zhu, Yun; Zhang, Jia; Bian, Jiang; Liu, Tie-Yan; Hao, Jiming

doi:10.1016/j.atmosres.2021.105919

Cited by 11 publications

(9 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The VAE-encoder employs the same model structure as the VAE-decoder, but the NO x emission replaces the current-day NO 2 as the output of the VAE-encoder, and the latter becomes one of its features. The model structure is nearly identical to that used in our previous study, , with the exception that we exclude VOC emissions from the features, which may have uncertainties but for which there are currently very few observations. Further studies are recommended to constrain both NO x and VOC emissions at the same time using additional observations such as HCHO and O 3 concentrations retrieved from satellites.…”

Section: Methodsmentioning

confidence: 99%

“…19 The machine-learning-based method was also used to estimate the surface NO 2 emissions from satellite data. 20 Recent research has shown that a neural network (NN) trained with CTM (noted as NN-CTM) can accurately capture the nonlinearity of photochemical reactions, 21 and it has been successfully used to adjust emissions by backpropagating the gradient of the loss function and measuring the difference between the CTM predictions and observations. 22 When compared to the traditional CTM, the NN-CTM has the advantage of providing an efficient prediction of pollution concentration under various emission and meteorological conditions, allowing us to easily modulate the model inputs to create user-preferred outputs (i.e., to be consistent with the observations).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Rapid Inference of Nitrogen Oxide Emissions Based on a Top-Down Method with a Physically Informed Variational Autoencoder

Xing

Zheng

et al. 2022

Environ. Sci. Technol.

Self Cite

View full text Add to dashboard Cite

Accurate timely estimation of emissions of nitrogen oxides (NO x ) is a prerequisite for designing an effective strategy for reducing O3 and PM2.5 pollution. The satellite-based top-down method can provide near-real-time constraints on emissions; however, its efficiency is largely limited by efforts in dealing with the complex emission–concentration response. Here, we propose a novel machine-learning-based method using a physically informed variational autoencoder (VAE) emission predictor to infer NO x emissions from satellite-retrieved surface NO2 concentrations. The computational burden can be significantly reduced with the help of a neural network trained with a chemical transport model, allowing the VAE emission predictor to provide a timely estimation of posterior emissions based on the satellite-retrieved surface NO2 concentration. The VAE emission predictor successfully corrected the underestimation of NO x emissions in rural areas and the overestimation in urban areas, resulting in smaller normalized mean biases (reduced from −0.8 to −0.4) and larger R 2 values (increased from 0.4 to 0.7). The interpretability of the VAE emission predictor was investigated using sensitivity analysis by modulating each feature, indicating that NO2 concentration and planetary boundary layer (PBL) height are important for estimating NO x emissions, which is consistent with our common knowledge. The advantages of the VAE emission predictor in efficiency, flexibility, and accuracy demonstrate its great potential in estimating the latest emissions and evaluating the control effectiveness from observations.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Rapid Inference of Nitrogen Oxide Emissions Based on a Top-Down Method with a Physically Informed Variational Autoencoder

Xing

Zheng

et al. 2022

Environ. Sci. Technol.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Since the physical transport process is mainly affected by instantaneous advection and diffusion conditions under single timestep and the integration of physical module and chemical module required to be conducted in one timestep, a deep U‐shape convolutional neural network (U‐Net) was selected to construct the deep learning model in this study (Dolz et al., 2018; Xing et al., 2022). U‐Net used encoder, decoder and skip connections to extract image features while reducing information loss during encoding and decoding (Ronneberger et al., 2015).…”

Section: Model Development Of the Deep Learning Surrogatementioning

confidence: 99%

Approximating Three‐Dimensional (3‐D) Transport of Atmospheric Pollutants via Deep Learning

Zhang

Cheng

et al. 2022

Earth and Space Science

View full text Add to dashboard Cite

Chemical transport model (CTM) is a fundamental research tool for atmospheric environment and is widely used for air quality forecasting, source apportionment, and strategy design of pollution alleviation (Chuang et al., 2018;Shen et al., 2020;Zhang et al., 2014). Physical transport process is the driving force for primary pollutants, thus then affects the concentration of secondary pollutants (Byun & Schere, 2006;Tilt, 2019). It is usually numerically solved based on the Euler's mass continuity equation in CTM and occupies 57%~60% of the entire CTM's computation time (Colella & Woodward, 1984;Ying & Li, 2011;Zhang et al., 2013). Moreover, the exponential growth of computational cost is observed with finer spatial resolution due to the smaller time step and larger iteration steps required (Boffi et al., 2007;Kasim et al., 2020). The solution of physical transport process has become the critical bottleneck of the CTM's computation efficiency.Machine learning, especially deep learning, has become a new research paradigm and acted as partial or whole replacement of complex geoscience models due to its excellent ability for non-linear fitting (Reichstein et al., 2019;Rolnick et al., 2019;Yuan et al., 2020). Specifically, for the surrogate of atmospheric transport process, Lauret et al. ( 2016) combined cellular automata and the artificial neural network (CA-ANN) to calculate the turbulence coefficient in horizontal two-dimensional (2-D) space. The R squared is over 0.7 for most testcases with the computation efficiency increases by ∼1.5 times. Wang and Qian (2018) extended the study to three-dimensional (3-D) space with the same CA-ANN. R squared for a single time step vary between 0.2 and 0.95, and is reduced to almost zero after 100 multiple-time-step. The computation efficiency increased is still ∼1.5 times. Vlasenko et al. ( 2021) emulated the entire CTM and the speed gain is 720 times of acceleration. Only 2-D horizontal distribution of daily average concentration, however, is concerned and the R squared values are only in the range of 0.38-0.67. The comprehensive performance with the aspects of efficiency promotion, long-term consistency, and spatial dimensions was still unsatisfied.

show abstract

“…Meteorology and nonlinear atmospheric chemistry complicate the attribution of air quality responses to changes in precursor emissions . For example, despite unprecedented reductions in anthropogenic emissions during the COVID-19 lockdown, increases in O 3 concentrations were widely observed in urban areas worldwide. − On the other hand, PM 2.5 also did not always decline, as it can be enhanced through secondary formation under unfavorable occasions (e.g., higher humidity, enhanced O 3 ). , Chemical transport models (CTMs) are used widely to simulate air pollutant concentrations under different emission and meteorological conditions. ,, However, CTMs rely highly on the accuracy and resolution of uncertain emission inventories and are computationally expensive for instant air quality management . Machine learning (ML)-based approaches have more flexibility and efficiency in leveraging real-world data and are good at capturing complicated nonlinear relationships .…”

Section: Introductionmentioning

confidence: 99%

Urban–Rural Disparities in Air Quality Responses to Traffic Changes in a Megacity of China Revealed Using Machine Learning

Wen

Zhou

Zhang

et al. 2022

Environ. Sci. Technol. Lett.

View full text Add to dashboard Cite

Assessing the disparities of urban–rural air quality response to changes in emissions is essential for the development of effective air pollution mitigation strategies in megacities. However, meteorology and nonlinear atmospheric chemistry complicate the determination of emission–air quality responses. Here, we established a machine learning (ML)-based air quality simulator based on hourly air quality, meteorology, traffic activity, and other relevant indicators for Chengdu, a megacity in Southwest China. The ML-based simulator exhibits high fidelity in reproducing hourly pollutant concentrations (with cross validation R2 > 0.6 for NO2, O3, and PM2.5). The results indicated similar trends of meteorological impacts but various effects from traffic activities on air quality between urban and rural areas. Truck restriction policies have significantly reduced the impacts of truck traffic on air quality in the urban area. Repartitioning between NO2 and O3 is observed in both urban and rural areas, indicating a VOC-limited regime in winter across Chengdu. Total gaseous oxidant (i.e., OX = NO2 + O3) and PM2.5 concentrations are more sensitive to changes in nontruck (which emit more VOC) traffic than truck (which emit more NOX) traffic. We suggest that effective mitigation policies of OX should be developed according to local features to improve and alleviate winter haze simultaneously.

show abstract

Mimicking atmospheric photochemical modeling with a deep neural network

Cited by 11 publications

References 39 publications

Rapid Inference of Nitrogen Oxide Emissions Based on a Top-Down Method with a Physically Informed Variational Autoencoder

Rapid Inference of Nitrogen Oxide Emissions Based on a Top-Down Method with a Physically Informed Variational Autoencoder

Approximating Three‐Dimensional (3‐D) Transport of Atmospheric Pollutants via Deep Learning

Urban–Rural Disparities in Air Quality Responses to Traffic Changes in a Megacity of China Revealed Using Machine Learning

Contact Info

Product

Resources

About