Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry

Due to the complex nature of ambient aerosols arising from the presence of myriads of organic compounds, the chemical reactivity of a particular compound with oxidant/s are studied through chamber experiments under controlled laboratory conditions. Several confounders (RH, T, light intensity, in cha...

Full description

Bibliographic Details
Main Authors: Prashant Rajput, Tarun Gupta
Format: Article
Language:English
Published: Frontiers Media S.A. 2020-12-01
Series:Frontiers in Environmental Science
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fenvs.2020.566136/full
id doaj-ea3be9fa28ef455889acfbfabd9f4026
record_format Article
spelling doaj-ea3be9fa28ef455889acfbfabd9f40262020-12-21T05:49:15ZengFrontiers Media S.A.Frontiers in Environmental Science2296-665X2020-12-01810.3389/fenvs.2020.566136566136Instrumental Variable Analysis in Atmospheric and Aerosol ChemistryPrashant Rajput0Tarun Gupta1Centre for Environmental Health (CEH), Public Health Foundation of Indian, Gurugram, IndiaDepartment of Civil Engineering, Indian Institute of Technology Kanpur, Kanpur, IndiaDue to the complex nature of ambient aerosols arising from the presence of myriads of organic compounds, the chemical reactivity of a particular compound with oxidant/s are studied through chamber experiments under controlled laboratory conditions. Several confounders (RH, T, light intensity, in chamber retention time) are controlled in chamber experiments to study their effect on the chemical transformation of a reactant (exposure variable) and the outcome [kinetic rate constant determination, new product/s formation e.g., secondary organic aerosol (SOA), product/s yield, etc.]. However, under ambient atmospheric conditions, it is not possible to control for these confounders which poses a challenge in assessing the outcome/s accurately. The approach of data interpretation must include randomization for an accurate prediction of the real-world scenario. One of the ways to achieve randomization is possible by the instrumental variable analysis (IVA). In this study, the IVA analysis revealed that the average ratio of fSOC/O3 (ppb−1) was 0.0032 (95% CI: 0.0009, 0.0055) and 0.0033 (95% CI: 0.0001, 0.0065) during daytime of Diwali and Post-Diwali period. However, during rest of the study period the relationship between O3 and fSOC was found to be insignificant. Based on IVA in conjunction with the concentration-weighted trajectory (CWT), cluster analysis, and fire count imageries, causal effect of O3 on SOA formation has been inferred for the daytime when emissions from long-range transport of biomass burning influenced the receptor site. To the best of our knowledge, the IVA has been applied for the first time in this study in the field of atmospheric and aerosol chemistry.https://www.frontiersin.org/articles/10.3389/fenvs.2020.566136/fullcausal inferencemachine learningair pollutionatmospheric chemistryaerosols
collection DOAJ
language English
format Article
sources DOAJ
author Prashant Rajput
Tarun Gupta
spellingShingle Prashant Rajput
Tarun Gupta
Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry
Frontiers in Environmental Science
causal inference
machine learning
air pollution
atmospheric chemistry
aerosols
author_facet Prashant Rajput
Tarun Gupta
author_sort Prashant Rajput
title Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry
title_short Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry
title_full Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry
title_fullStr Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry
title_full_unstemmed Instrumental Variable Analysis in Atmospheric and Aerosol Chemistry
title_sort instrumental variable analysis in atmospheric and aerosol chemistry
publisher Frontiers Media S.A.
series Frontiers in Environmental Science
issn 2296-665X
publishDate 2020-12-01
description Due to the complex nature of ambient aerosols arising from the presence of myriads of organic compounds, the chemical reactivity of a particular compound with oxidant/s are studied through chamber experiments under controlled laboratory conditions. Several confounders (RH, T, light intensity, in chamber retention time) are controlled in chamber experiments to study their effect on the chemical transformation of a reactant (exposure variable) and the outcome [kinetic rate constant determination, new product/s formation e.g., secondary organic aerosol (SOA), product/s yield, etc.]. However, under ambient atmospheric conditions, it is not possible to control for these confounders which poses a challenge in assessing the outcome/s accurately. The approach of data interpretation must include randomization for an accurate prediction of the real-world scenario. One of the ways to achieve randomization is possible by the instrumental variable analysis (IVA). In this study, the IVA analysis revealed that the average ratio of fSOC/O3 (ppb−1) was 0.0032 (95% CI: 0.0009, 0.0055) and 0.0033 (95% CI: 0.0001, 0.0065) during daytime of Diwali and Post-Diwali period. However, during rest of the study period the relationship between O3 and fSOC was found to be insignificant. Based on IVA in conjunction with the concentration-weighted trajectory (CWT), cluster analysis, and fire count imageries, causal effect of O3 on SOA formation has been inferred for the daytime when emissions from long-range transport of biomass burning influenced the receptor site. To the best of our knowledge, the IVA has been applied for the first time in this study in the field of atmospheric and aerosol chemistry.
topic causal inference
machine learning
air pollution
atmospheric chemistry
aerosols
url https://www.frontiersin.org/articles/10.3389/fenvs.2020.566136/full
work_keys_str_mv AT prashantrajput instrumentalvariableanalysisinatmosphericandaerosolchemistry
AT tarungupta instrumentalvariableanalysisinatmosphericandaerosolchemistry
_version_ 1724375536365469696