Research PaperPredicting PM10 concentration in Seoul metropolitan subway stations using artificial neural network (ANN)
Introduction
The indoor air quality of subway systems (known as rapid transit systems, metros, subways, the tube, etc.) can significantly affect the passengers health since these systems are widely used for short-distance transit in metropolitan urban areas in many countries [1], [2], [3], [4]. The particles generated by abrasion during subway operations and the vehicle-emitted pollutants flowing in from the street in particular affect the air quality in underground subway stations. Those particles contain heavy metal elements [5], while the vehicle-emitted pollutants flowing in from the street contain particulate and gaseous pollutants generated by the combustion of fossil fuel [6].
Studies thus far of the air quality in underground subway stations have focused on analysis of the distribution of particulate matter less than 10 μm (PM10) concentrations [7] and gaseous pollutants [8], as well as analysis of pollutants according to the ventilation characteristics of the underground space, the characteristics of train operation and the structure of the station [9], [10], [11]. Marsik et al. [12] and Yu et al. [13] reported that outdoor and indoor PM10 concentrations were positively correlated. There are also many studies on the risks to human health of the component properties of fine dusts generated in underground stations. Seaton et al. [1] and Karlsson et al. [2] reported that magnetite (FeO3) content in these PM10 generated in subways accelerates DNA damage and causes oxidative stress in lungs while Pope et al. [3] reported that prolonged exposure to PM10 can cause cardiovascular disease.
Most metropolitan cities worldwide have atmospheric pollution measurement networks to continuously monitor PM10 concentration in the atmosphere, with Seoul operating 31 atmospheric and 14 roadside networks. However, continuous measurement of PM10 in indoor spaces is done only in the limited number of stations because the measurement systems are expensive and difficult to install and operate for significant periods of time in spaces crowded with people. Therefore, continuous measurement of PM10 concentration on the platform is not a reality, with the systems generally installed and run only for specific periods of time. In addition, it is quite necessary to operate the ventilation system of underground station based on the prediction of air pollutants level because the transfer (dispersion) of air pollutants takes time, especially for PM.
Recent studies have adopted statistical techniques that predict concentrations using the concentration of other pollutants or past measurements. The leading statistical technique is the artificial neural network (ANN), which uses pattern recognition technology used especially in environmental fields [14], [15]. Chellali et al. [16], Grivs et al. [17] and Papanastasiou. [18] reported that ANN-predicted PM10 levels showed the highest performance over other prediction models while Perez [19] reported that ANN showed the superior prediction on PM2.5. Grivas et al. [17] used meteorological information such as wind speed, ambient temperature and relative humidity as inputs to predict fine dust concentration in the atmosphere, while Chaloulakou et al. [20] added wind direction as one of the inputs. Our literature review revealed that most studies predicting PM10 focused on PM10 in outdoor air. While outdoor PM10 are considerably affected by meteorological information, indoor spaces, such as underground subway stations, are affected by a multitude of conditions such as the pollution source or inflow (or outflow) of PM10 through the ventilation systems [7], [9], [21].
Here, this study predicts indoor PM10 concentration on selected subway stations using input variables such as outdoor PM10, the number of subway trains running, and information on ventilation operation. ANN model predicts indoor PM10 using input variables of one hour previous data also. As well, we investigated the relationship between ANN’s performance and the depth of underground subway station. We hope that this predictive tool could provide an effective ventilation strategy of underground station preventing passengers from exposure of PM10.
Section snippets
Target underground subway stations
This study analyzed the PM10 measured hourly on a platform in six selected major underground subway transfer stations (A1, A2, B1, B2, C1 and C2) in Seoul as shown in Fig. 1. The measurement method and the detailed results are presented in our previous study [7]. The platform at station A1 was the nearest to the ground at 9.8 m below the street surface, while the platform at station C2 was the furthest at 23.1 m. The platforms are divided into “island” type, in which transportation in both
Use of outdoor public data and indoor measured data
Official data on the air pollution monitoring network and the roadside monitoring network near (within 2.9 km) each underground subway station (www.airkorea.or.kr) were obtained as the outdoor concentration of our study as indicated in Fig. 1. Table 2 presents the measurement campaign period and the average with range of PM10_out and PM10_in.
PM10_out collected from the air pollution monitoring network near each of the six station showed the average concentration range of 31 ∼ 52 μm/m3 while PM10_in
Conclusion
- 1.
Linear regression analysis between PM10_out and PM10_in showed that the average correlation coefficient (R2) of the 6 analyzed stations was 0.43 (0.18 ∼ 0.63), whereas ANN model including additional input variables of subway frequency (SF) and ventilation rate (VR) showed a significant increase of correlation coefficients, 0.65 (0.39 ∼ 0.81).
- 2.
ANN model with previous time variables, which predicts the current PM10_in showed a high correlation between the predicted and actual measured values for all
Acknowledgement
This work was supported by the Railway Technology Development Program(17RTRP-B082486-04) funded Korea Ministry of Land, Infrastructure and Transport (MOLIT).
References (28)
- et al.
A new look at inhalable metalliferous airborne particles on rail subway platforms
Sci. Total Environ.
(2015) - et al.
A multivariate study for characterizing particulate matter (PM10 PM2.5, and PM1) in Seoul metropolitan subway stations
Korean J. Hazard. Mater.
(2015) - et al.
Indoor air quality control for improving passenger health in subway platforms using an outdoor air quality dependent ventilation system
Build. Environ.
(2015) - et al.
Subway platform air quality: assessing the influences of tunnel ventilation: train piston effect and station design
Atmos. Environ.
(2014) - et al.
Influence of mechanical ventilation system on indoor carbon dioxide and particulate matter concentration
Build. Environ.
(2014) - et al.
Prediction of contamination potential of groundwater arsenic in Cambodia Laos, and Thailand using artificial neural network
Warter Res.
(2011) - et al.
Artificial neural network models for prediction of PM10 hourly concentrations in the Greater Area of Athens, Greece
Atmos. Environ.
(2006) - et al.
Prediction of PM2.5 concentrations several hours in advance using neural networks in Santiago
Chile. Atmos. Envrion.
(2000) - et al.
Anoticeable shift in particulate matter levels after platform screen door installation in a Korean subway station
Atmos. Environ.
(2012) - et al.
Neural network prediction model for fine particulate matter (PM 2. 5) on the US–Mexico border in El Paso (Texas) and Ciudad Juárez (Chihuahua)
Environ. Modell. Software
(2005)
A hybrid ARIMA and artificial neural networks model to forecast particulate matter in urban areas: the case of Temuco, Chile
Atmos. Environ.
Online monitoring and interpretation of periodic diurnal and seasonal variations of indoor air pollutants in a subway station using parallel factor analysis(PARAFAC)
Energy Build.
The London underground dust and hazards to health
Occup. Environ. Med.
Subway particles are more genotoxic than street particles and induce oxidative stress in cultured human lung cells
Chem. Res. Toxicol.
Cited by (137)
Large-scale spatiotemporal deep learning predicting urban residential indoor PM<inf>2.5</inf> concentration
2023, Environment InternationalHybrid approach based on response surface methodology and artificial neural networks coupled with genetic algorithm (RSM-GA-ANN) for the Prediction and optimization for the Photodegradation of dye using nano ZnO anchored glass fiber under solar light irradiation
2023, Journal of the Taiwan Institute of Chemical Engineers