Submitted as: model description paper
25 Jul 2022
Submitted as: model description paper | 25 Jul 2022
Status: this preprint is currently under review for the journal GMD.

Development of an LSTM-Broadcasting deep-learning framework for regional air pollution forecast improvement

Haochen Sun1,2, Jimmy C. H. Fung1,3,4, Yiang Chen3, Zhenning Li3, Dehao Yuan5, Wanying Chen3, and Xingcheng Lu3,4 Haochen Sun et al.
  • 1Department of Mathematics, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong SAR, China
  • 2Department of Computer Science and Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong SAR, China
  • 3Division of Environment and Sustainability, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong SAR, China
  • 4Atmospheric Research Center, HKUST Fok Ying Tung Research Institute, Guangzhou, China
  • 5Department of Computer Science, University of Maryland, College Park, Maryland, USA

Abstract. Deep-learning frameworks can effectively forecast the air pollution data for individual stations by decoding time-series data. However, most of the existing time-series-based deep-learning models use offline spatial interpolation strategies and thus cannot reliably project the station-based forecast to the spatial region of interest. In this study, the station-based long short-term memory (LSTM) technique was extended for spatial air quality forecasting by combining a novel deep-learning layer termed the broadcasting layer, which incorporates a learnable weight decay parameter designed for point-to-area extension. Unlike most existing deep-learning-based methods that isolate the interpolation from the model training process, the proposed end-to-end LSTM-broadcasting framework can consider the temporal characteristics of the time series and spatial relationships among different stations. To validate the proposed deep-learning framework, PM2.5 and O3 forecasts for the next 48 h were obtained using 3D chemical transport model simulation results and ground observation data as the inputs. The root mean square error associated with the proposed framework was 40 % and 20 % lower than those of the Weather Research Forecast–Community Multiscale Air Quality model and an offline combination of the deep-learning and spatial interpolation methods, respectively. The novel LSTM-broadcasting framework can be extended for air pollution forecasting in other regions of interest.

Haochen Sun et al.

Status: open (until 19 Sep 2022)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on gmd-2022-164', Anonymous Referee #1, 08 Aug 2022 reply

Haochen Sun et al.

Haochen Sun et al.


Total article views: 189 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
141 44 4 189 15 4 4
  • HTML: 141
  • PDF: 44
  • XML: 4
  • Total: 189
  • Supplement: 15
  • BibTeX: 4
  • EndNote: 4
Views and downloads (calculated since 25 Jul 2022)
Cumulative views and downloads (calculated since 25 Jul 2022)

Viewed (geographical distribution)

Total article views: 168 (including HTML, PDF, and XML) Thereof 168 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 08 Aug 2022
Short summary
This study developed a novel deep-learning layer, the broadcasting layer, to build an end-to-end LSTM based deep-learning model for regional air pollution forecast. By combining the ground observation, WRF-CMAQ simulation, and the Broadcasting-LSTM deep-learning model, forecast accuracy has been significantly improved when compared to other methods. The broadcasting layer and its variants can also be applied in other research areas to supersede the traditional numerical interpolation methods.