Integrated PM2.5 Estimation Using Satellite Images, Pollution Station Data, Meteorological Data and Artificial Neural Network (Case Study: Tehran City)

Abstract

                                            ABSTRACT

Nowadays, the use of remote sensing to monitor airborne particulates by spatial continuity is made possible by the MODIS Sensor Aerosol Optical Depth (AOD) products. In this study, a simplified aerosol estimation model (SARA) was used to estimate the aerosol optical depth. For this purpose, the MODIS sensor data as inputs to predict PM2.5 near the surface, aerosol optical depth data and meteorological data (wind speed and direction, air temperature, planetary boundary height and relative humidity) and artificial neural network are used. The results of comparing the aerosol optical depth obtained with real PM2.5 data obtained from the contamination stations showed that the highest correlation was related to summer with Pearson coefficient (0.67) and the least Pearson correlation coefficient (0.55) to spring. The results also showed that the use of meteorological data and artificial neural network for predicting land surface PM2.5 was successful. Comparison of PM2.5 output of artificial neural network model with observed PM2.5 values showed seasonal variation with R2 coefficient of 0.51, 0.74, 0.61, 0.62 and RMSE value of 15.2, 7.5, 12.1, 6.59 shows the spring, summer, autumn, and winter seasons, respectively.

Keywords: PM2.5, Artificial Neural Network, MODIS, Tehran, SARA Algorithm