Using machine learning to quantify sources of light-absorbing water-soluble humic-like substances (HULISws) in Northeast China. (15th December 2022)

• Record Type: Journal Article
• Authors: Hong, Yihang; Cao, Fang; Fan, Mei-Yi; Lin, Yu-Chi; Bao, Mengying; Xue, Yongwen; Wu, Jiyan; Yu, Mingyuan; Wu, Xia; Zhang, Yan-Lin
• Is Part Of: Atmospheric environment. Volume 291(2022)
• Publication Date: 2022-12-15
• Subjects: Humic-like substances -- Heating season -- Optical properties -- Source apportionment -- FLEXPART model -- Machine learning algorithm; Air -- Pollution -- Periodicals; Air -- Pollution -- Meteorological aspects -- Periodicals; 551.51
• DOI: 10.1016/j.atmosenv.2022.119371
• Languages: English
• ISSNs: 1352-2310

Abstract

Abstract: Water-soluble humic-like substances (HULISws ) are a group of heterogeneous organic compounds in the atmosphere, greatly impacting climate change and human health. As the most significant biomass burning (BB) and fossil fuel burning polluted area in China, Northeast China Plain could generate more light absorbing HULISws to the atmosphere. In this work, daily fine particle (PM2.5 ) samples were collected in Changchun, located in Northeast China Plain, from October 17th to November 29th, 2016, to investigate the potential sources of optical properties of HULISws . Here, the whole sampling period was divided into three sub-periods using the density of fire spots and the timeline of the central heating: the non-heating (October 17th to October 24th), early heating (October 25th to November 13th), and normal heating (November 14th to November 29th) periods. The mean mass concentrations of HULISws were 4.2 ± 1.2, 8.6 ± 3.6, and 2.4 ± 1.3 μg m −3 during the non-heating, early heating, and normal heating periods, respectively. The positive matrix factorization (PMF) model results suggested that the contribution of primary BB, secondary BB, and fossil fuel burning emissions were 46%, 24%, and 19% during the non-heating period, 22%, 29%, and 17% during the early heating period, and 13%, 16%, and 50% during the normal heating period, respectively. Combining the PMF results with the random forest (RF) algorithm, the contribution of each source to the optical properties was quantified. Here, BB still dominated the optical properties of HULISws in Northeast China. During the study period, the primary BB and secondary BB contributed 25% and 34% in the light absorption coefficient (Abs), 22% and 33% in the mass absorption exponent (MAE), and 17% and 33% in the absorption Ångström exponent (AAE), respectively. Other sources like cooking and fossil combustion contributed 25% and 35% in Abs, 35% and 26% in MAE, and 16% and 33% in AAE, respectively. Highlights: 72% biomass burning emission contributed 40%–50% to light absorption parameters. 4% cooking emission contributed 16%–35% to light absorption parameters. This work provides a fast and easy way to quantify the contribution of sources to the light absorption properties.