Submitted as: model description paper 29 Mar 2021

Submitted as: model description paper | 29 Mar 2021

Review status: this preprint is currently under review for the journal GMD.

Exploring Deep Learning for Air Pollutant Emission Estimation

Lin Huang1,, Song Liu2,3,, Zeyuan Yang4, Jia Xing2,3, Jia Zhang1, Jiang Bian1, Siwei Li5,6, Shovan Kumar Sahu2,3, Shuxiao Wang2,3, and Tie-Yan Liu1 Lin Huang et al.
  • 1Microsoft Research, Beijing, China
  • 2State Key Joint Laboratory of Environmental Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing, China
  • 3State Environmental Protection Key Laboratory of Sources and Control of Air Pollution Complex, Beijing, China
  • 4School of Economics and Management, Tsinghua University, Beijing, China
  • 5School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, China
  • 6State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
  • These authors contributed equally to this work.

Abstract. The inaccuracy of anthropogenic emission inventory on a high-resolution scale due to insufficient basic data is one of the major reasons for the deviation between air quality model and observation results. A bottom-up approach, as a typical emission inventory estimation approach, requires a lot of human labor and material resources, and a top-down approach focuses on individual pollutants that can be measured directly and relies heavily on traditional numerical modelling. Lately, deep neural network has achieved rapid development due to its high efficiency and non-linear expression ability. In this study, we proposed a novel method to model the dual relationship between emission inventory and pollution concentration for emission inventory estimation. Specifically, we utilized a neural network based comprehensive chemical transport model (NN-CTM) to learn the complex correlation between emission and air pollution. We further updated the emission inventory based on backpropagating the gradient of the loss function measuring the deviation between NN-CTM and observations from surface monitors. We first mimicked the CTM model with neural networks (NN) and achieved a relatively good representation of CTM with similarity reaching 95 %. To reduce the gap between CTM and observations, the NN model would suggest an updated emission of NOx, NH3, SO2, VOC and primary PM2.5 which changes by −1.34 %, −2.65 %, −11.66 %, −19.19 % and 3.51 %, respectively, on average of China. Such ratios of NOx and PM2.5 are even higher (~10 %) particularly in Northwest China where suffers from large uncertainties in original emissions. The updated emission inventory can improve model performance and make it closer to observations. The mean absolute error for NO2, SO2, O3 and PM2.5 concentrations are reduced significantly by about 10 %~20 %, indicating the high feasibility of NN-CTM in terms of significantly improving both the accuracy of emission inventory as well as the performance of air quality model.

Lin Huang et al.

Status: open (until 24 May 2021)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Lin Huang et al.

Lin Huang et al.


Total article views: 324 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
240 80 4 324 4 6
  • HTML: 240
  • PDF: 80
  • XML: 4
  • Total: 324
  • BibTeX: 4
  • EndNote: 6
Views and downloads (calculated since 29 Mar 2021)
Cumulative views and downloads (calculated since 29 Mar 2021)

Viewed (geographical distribution)

Total article views: 213 (including HTML, PDF, and XML) Thereof 213 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 19 Apr 2021
Short summary
Accurate estimation of emissions is the prerequisite for effectively controlling air pollution, while current methods either lack sufficient data or representation of nonlinearity. Here we proposed a novel deep learning method to model the dual relationship between emission and pollutant concentration. The emission can be updated through backpropagating the gradient of the loss function measuring the deviation between simulations and observations, resulting in a better model performance.