Page 638 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

of particulate matter less than 2.5 microns using machine learning algorithms and identified and selected effective meteorological
parameters. Wei et al.[7] used support vector machine and artificial neural network methods to model the level of air pollution
and proposed the support vector machine method as the superior method for modelling air pollution. Kurani et al.[8] used the
network bayesian method for modelling ozone and PM2.5 pollutant.
In previous research, to classify particulate matter pollution, in some cases only meteorological parameters have been
considered and there was not enough attention to the most effective parameters and also in other cases only one method has been
used for classification. Accordingly, there is no research that fully classifies the pollution class, compares them, and prepares the
classification map by considering the most effective parameters using decision tree algorithms. In this research, an attempt has
been made to pay the above issues. In this regard, in this study, CART and C4.5 decision tree algorithms were used to classify
the concentration of suspended particles less than 2.5 microns and prepare a pollution classification map of the methods.

2. Study area and Dataset

The city of Tehran is located in the southern foothills of the Alborz mountain range, the longitude between 51° 05′-51°
53′East and latitude 35 ° 34 ′-35 ° 59′North with an area of about 700 square kilometers. The altitude of Tehran is between
1050-2000 meters above the sea level. The period of data collection and research is related to the autumn season of 2017 and
2018.

This research uses several types of data:

A) Six-hour data of meteorological parameters, including relative humidity, rainfall, temperature, air pressure, wind
speed, wind direction of synoptic stations in Tehran province received from the Meteorological Organization of the
country.
B) Data related to PM2.5 pollutant concentration that is six hours from the station active measurements of air pollution
in Tehran received from the Air Quality Control Office of Tehran.
C) Related data to the topographic status of the whole city of Tehran received from the National Cartographic Center of
Iran (NCC).
D) Intensity of temperature inversion obtained for 12 hours from observational data of high atmosphere (extracted from
radio sound) by the website of the University of Wyoming[9].
E) All these parameters were determined next to a specific time unit (month of the year, day of the week, hour of the
day) and time dependence was considered for each parameter by creating a time series.

3. Proposed method

The general trend of the proposed method is presented in Fig. 1. In this study, as shown in Fig. 1, by using the Kriging[10]
interpolation method and considering the pixel size of 100 m on the ground, all parameters for the city of Tehran were created
as several layers of information. Accordingly, all points of air pollution in Tehran in addition to all meteorological parameters
were determined on time and concentration of pollutants. Another factor that affects air pollution is the temperature inversion
[11]. Of course, temperature inversion in infected days is not the same amount and intensity, but in some days are different in
terms of intensity and performance [12]. For this purpose, another parameter called inversion intensity is considered along with
other parameters where the unit of it is the Celsius degree per hundred meters. To find this parameter, Equation (1) is used [13,
14] as follows [14]:

= (1)

2