Air pollution prediction by using an artificial neural network model

Heidar Maleki, Armin Sorooshian, Gholamreza Goudarzi, Zeynab Baboli, Yaser Tahmasebi Birgani, Mojtaba Rahmati

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Abstract: Air pollutants impact public health, socioeconomics, politics, agriculture, and the environment. The objective of this study was to evaluate the ability of an artificial neural network (ANN) algorithm to predict hourly criteria air pollutant concentrations and two air quality indices, air quality index (AQI) and air quality health index (AQHI), for Ahvaz, Iran, over one full year (August 2009–August 2010). Ahvaz is known to be one of the most polluted cities in the world, mainly owing to dust storms. The applied algorithm involved nine factors in the input stage (five meteorological parameters, pollutant concentrations 3 and 6 h in advance, time, and date), 30 neurons in the hidden phase, and finally one output in last level. When comparing performance between using 5% and 10% of data for validation and testing, the more reliable results were from using 5% of data for these two stages. For all six criteria pollutants examined (O3, NO2, PM10, PM2.5, SO2, and CO) across four sites, the correlation coefficient (R) and root-mean square error (RMSE) values when comparing predictions and measurements were 0.87 and 59.9, respectively. When comparing modeled and measured AQI and AQHI, R2 was significant for three sites through AQHI, while AQI was significant only at one site. This study demonstrates that ANN has applicability to cities such as Ahvaz to forecast air quality with the purpose of preventing health effects. We conclude that authorities of urban air quality, practitioners, and decision makers can apply ANN to estimate spatial–temporal profile of pollutants and air quality indices. Further research is recommended to compare the efficiency and potency of ANN with numerical, computational, and statistical models to enable managers to select an appropriate toolkit for better decision making in field of urban air quality. Graphical abstract: [Figure not available: see fulltext.].

Original languageEnglish (US)
JournalClean Technologies and Environmental Policy
DOIs
StatePublished - Jan 1 2019

Fingerprint

Air pollution
Air quality
artificial neural network
air quality
atmospheric pollution
Neural networks
prediction
Health
Air Pollutants
index
pollutant
dust storm
Public health
Carbon Monoxide
Air
Mean square error
Agriculture
Neurons
Dust
public health

Keywords

  • ANN
  • AQHI
  • AQI
  • Criteria air pollutants

ASJC Scopus subject areas

  • Environmental Engineering
  • Environmental Chemistry
  • Management, Monitoring, Policy and Law

Cite this

Air pollution prediction by using an artificial neural network model. / Maleki, Heidar; Sorooshian, Armin; Goudarzi, Gholamreza; Baboli, Zeynab; Tahmasebi Birgani, Yaser; Rahmati, Mojtaba.

In: Clean Technologies and Environmental Policy, 01.01.2019.

Research output: Contribution to journalArticle

Maleki, Heidar ; Sorooshian, Armin ; Goudarzi, Gholamreza ; Baboli, Zeynab ; Tahmasebi Birgani, Yaser ; Rahmati, Mojtaba. / Air pollution prediction by using an artificial neural network model. In: Clean Technologies and Environmental Policy. 2019.
@article{9ec5491e0ce548ce8ba5c2eb560efea2,
title = "Air pollution prediction by using an artificial neural network model",
abstract = "Abstract: Air pollutants impact public health, socioeconomics, politics, agriculture, and the environment. The objective of this study was to evaluate the ability of an artificial neural network (ANN) algorithm to predict hourly criteria air pollutant concentrations and two air quality indices, air quality index (AQI) and air quality health index (AQHI), for Ahvaz, Iran, over one full year (August 2009–August 2010). Ahvaz is known to be one of the most polluted cities in the world, mainly owing to dust storms. The applied algorithm involved nine factors in the input stage (five meteorological parameters, pollutant concentrations 3 and 6 h in advance, time, and date), 30 neurons in the hidden phase, and finally one output in last level. When comparing performance between using 5{\%} and 10{\%} of data for validation and testing, the more reliable results were from using 5{\%} of data for these two stages. For all six criteria pollutants examined (O3, NO2, PM10, PM2.5, SO2, and CO) across four sites, the correlation coefficient (R) and root-mean square error (RMSE) values when comparing predictions and measurements were 0.87 and 59.9, respectively. When comparing modeled and measured AQI and AQHI, R2 was significant for three sites through AQHI, while AQI was significant only at one site. This study demonstrates that ANN has applicability to cities such as Ahvaz to forecast air quality with the purpose of preventing health effects. We conclude that authorities of urban air quality, practitioners, and decision makers can apply ANN to estimate spatial–temporal profile of pollutants and air quality indices. Further research is recommended to compare the efficiency and potency of ANN with numerical, computational, and statistical models to enable managers to select an appropriate toolkit for better decision making in field of urban air quality. Graphical abstract: [Figure not available: see fulltext.].",
keywords = "ANN, AQHI, AQI, Criteria air pollutants",
author = "Heidar Maleki and Armin Sorooshian and Gholamreza Goudarzi and Zeynab Baboli and {Tahmasebi Birgani}, Yaser and Mojtaba Rahmati",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/s10098-019-01709-w",
language = "English (US)",
journal = "Clean Technologies and Environmental Policy",
issn = "1618-954X",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Air pollution prediction by using an artificial neural network model

AU - Maleki, Heidar

AU - Sorooshian, Armin

AU - Goudarzi, Gholamreza

AU - Baboli, Zeynab

AU - Tahmasebi Birgani, Yaser

AU - Rahmati, Mojtaba

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Abstract: Air pollutants impact public health, socioeconomics, politics, agriculture, and the environment. The objective of this study was to evaluate the ability of an artificial neural network (ANN) algorithm to predict hourly criteria air pollutant concentrations and two air quality indices, air quality index (AQI) and air quality health index (AQHI), for Ahvaz, Iran, over one full year (August 2009–August 2010). Ahvaz is known to be one of the most polluted cities in the world, mainly owing to dust storms. The applied algorithm involved nine factors in the input stage (five meteorological parameters, pollutant concentrations 3 and 6 h in advance, time, and date), 30 neurons in the hidden phase, and finally one output in last level. When comparing performance between using 5% and 10% of data for validation and testing, the more reliable results were from using 5% of data for these two stages. For all six criteria pollutants examined (O3, NO2, PM10, PM2.5, SO2, and CO) across four sites, the correlation coefficient (R) and root-mean square error (RMSE) values when comparing predictions and measurements were 0.87 and 59.9, respectively. When comparing modeled and measured AQI and AQHI, R2 was significant for three sites through AQHI, while AQI was significant only at one site. This study demonstrates that ANN has applicability to cities such as Ahvaz to forecast air quality with the purpose of preventing health effects. We conclude that authorities of urban air quality, practitioners, and decision makers can apply ANN to estimate spatial–temporal profile of pollutants and air quality indices. Further research is recommended to compare the efficiency and potency of ANN with numerical, computational, and statistical models to enable managers to select an appropriate toolkit for better decision making in field of urban air quality. Graphical abstract: [Figure not available: see fulltext.].

AB - Abstract: Air pollutants impact public health, socioeconomics, politics, agriculture, and the environment. The objective of this study was to evaluate the ability of an artificial neural network (ANN) algorithm to predict hourly criteria air pollutant concentrations and two air quality indices, air quality index (AQI) and air quality health index (AQHI), for Ahvaz, Iran, over one full year (August 2009–August 2010). Ahvaz is known to be one of the most polluted cities in the world, mainly owing to dust storms. The applied algorithm involved nine factors in the input stage (five meteorological parameters, pollutant concentrations 3 and 6 h in advance, time, and date), 30 neurons in the hidden phase, and finally one output in last level. When comparing performance between using 5% and 10% of data for validation and testing, the more reliable results were from using 5% of data for these two stages. For all six criteria pollutants examined (O3, NO2, PM10, PM2.5, SO2, and CO) across four sites, the correlation coefficient (R) and root-mean square error (RMSE) values when comparing predictions and measurements were 0.87 and 59.9, respectively. When comparing modeled and measured AQI and AQHI, R2 was significant for three sites through AQHI, while AQI was significant only at one site. This study demonstrates that ANN has applicability to cities such as Ahvaz to forecast air quality with the purpose of preventing health effects. We conclude that authorities of urban air quality, practitioners, and decision makers can apply ANN to estimate spatial–temporal profile of pollutants and air quality indices. Further research is recommended to compare the efficiency and potency of ANN with numerical, computational, and statistical models to enable managers to select an appropriate toolkit for better decision making in field of urban air quality. Graphical abstract: [Figure not available: see fulltext.].

KW - ANN

KW - AQHI

KW - AQI

KW - Criteria air pollutants

UR - http://www.scopus.com/inward/record.url?scp=85066604159&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85066604159&partnerID=8YFLogxK

U2 - 10.1007/s10098-019-01709-w

DO - 10.1007/s10098-019-01709-w

M3 - Article

AN - SCOPUS:85066604159

JO - Clean Technologies and Environmental Policy

JF - Clean Technologies and Environmental Policy

SN - 1618-954X

ER -