|
A Neural Network-Based Approach for Real-Time Measurement of the Concentration of Gaseous Pollutants in Tehran Using MODIS
|
Mina Saleh   , Reza Shah-Hosseini * , Zahra Bahramian , Sara Khanbani |
| School of Surveying & Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran |
|
|
Abstract: (1561 Views) |
Nowadays, gas pollutants are considered as an important challenge in big cities. Due to the fact that gaseous pollutants have negative effects on human health and destroy the environment, there are several methods for predicting the concentration of gaseous pollutants such as carbon monoxide (CO), nitrogen dioxide (NO2) and sulfur dioxide (SO2). The aim of the present research is to calculate the concentration of gaseous pollutants in real time using MODIS sensor data, including night and day surface temperature, aerosol light depth, vegetation index and the data from the ground stations monitoring the concentration of the pollutants using multi-layer perceptron neural network. The perceptron neural network had the best performance with 8 neurons, 4 of which in the input layer, 3 in the middle layer, and one in the output layer. 80% of the data were considered as the training data and 20% as the test data; and 15% of the training data were considered for Validation of the network. Using the aforementioned training and experimental data, the parameters of the number of periods and the learning rate were subjected to sensitivity analysis and the most suitable parameters were selected. In the next step, the random forest regression method was used to compare the results. The results showed that the multilayer perceptron neural network performed better than the random forest regression. In this research, the qualitative analysis of the pollutant concentration map and the pollutants’ relationship with the land use and the existing roads around each of the air quality control monitoring stations was done. The data of Tehran city were used as a 6-year time series from 1393 to 1399. The accuracy evaluation of the proposed method using the experimental data shows 86% accuracy for measuring carbon monoxide (CO) and nitrogen dioxide (NO2) pollutants and 92% accuracy for sulfur dioxide (SO2) one.
|
|
| Keywords: Air Pollution, Perceptron Neural Network, Gaseous Pollutants, GIS, MODIS |
|
|
Full-Text [PDF 3047 kb]
(370 Downloads)
|
Type of Study: Research |
Subject:
RS
Received: 2023/07/23 | Accepted: 2023/12/24 | ePublished ahead of print: 2024/02/18 | Published: 2024/03/4
|
|
|
|
|
|
|
| Send email to the article author |
|
|
