Air Quality Forecasts Improved by Combining Data Assimilation and Machine Learning With Satellite AOD. Issue 1 (31st December 2021)

• Record Type: Journal Article
• Authors: Lee, Seunghee; Park, Seohui; Lee, Myong‐In; Kim, Ganghan; Im, Jungho; Song, Chang‐Keun
• Is Part Of: Geophysical research letters. Volume 49:Issue 1(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12-31
• Subjects: air quality forecast -- data assimilation -- machine learning -- particulate matter -- random forest; Geophysics -- Periodicals; Planets -- Periodicals; Lunar geology -- Periodicals; 550
• DOI: 10.1029/2021GL096066
• Languages: English
• ISSNs: 0094-8276

Abstract

Abstract: Satellite aerosol optical depth (AOD) data assimilation (DA) using numerical air quality forecast models has shown a limited improvement due to large uncertainties in the AOD observation operator. This study employed a machine learning (ML) algorithm to estimate the ground‐level particulate matter (PM) from the Geostationary Ocean Color Imager (GOCI) AOD through the random forest with high accuracy. Analysis fields were subsequently produced by applying PM estimations to the Weather Research and Forecasting‐Chemistry/three‐dimensional variational DA system. Initialization of the model with the new analysis remarkably reduced the analysis error and increased the forecast skill. The PM10 prediction showed significant benefits for up to 24 forecast hours, whereas PM2.5 prediction was improved for up to six forecast hours. Considering a broad spatial coverage by satellites, the synergistic use of DA and ML can maximize the effectiveness of satellite DA for air quality forecasts at the ground. Plain Language Summary: Due to their broad spatial coverage, satellite aerosol observations (OBS) are now widely used in real‐time air quality monitoring and forecasts. A conventional method in most operational air quality monitoring and forecast centers uses satellite data to fill the spatiotemporal gap in the ground OBS network by combining them with the simulation data from the numerical air quality model. However, this process contains significant intrinsic errors in transforming aerosol products from satellites into the mass concentration values to be predicted. This study suggests using a machine‐learning (ML) algorithm to estimate the ground‐level particulate matter (PM) concentrations from the satellite optical data. This method not only covers broad geographical areas, including mountainous and oceanic regions that are not being covered by ground OBS stations but also improves the initialization process of the numerical air quality forecast model. The PM10 prediction shows significant benefits for up to 24 forecast hours, whereas PM2.5 prediction is improved for up to six forecast hours. Considering a broad spatial coverage by satellites, the synergistic use of data assimilation (DA) and ML can maximize the effectiveness of satellite DA for air quality forecasts at the ground. Key Points: Conventional aerosol data assimilation (DA) suffers from large uncertainties when using satellite aerosol optical depth (AOD) observations Machine learning was applied to estimate ground particulate matter (PM) concentrations from the satellite AOD for use in the conventional 3D‐VAR DA system The modified DA experiment remarkably improves the PM10 and PM2.5 prediction performance compared to the conventional AOD DA experiment