Application of a linear regression model in the study of PM10 distribution at urban districts of Hanoi city

Abstract

This study applied the linear regression model to Landsat 8 OLI satellite images and in-situ measurement data to create a distribution map of PM10 concentration, thereby supporting the
monitoring and assessing the air quality in urban districts of Hanoi. The PM10 concentration is estimated based on the correlation between atmospheric reflectance, which is determined from Landsat 8 OLI images using visible bands and the PM10 concentration measured at ground level. Through linear regression analysis, we have found the most appropriate regression function corresponding to R and R2 values (0.971 and 0.943 respectively) and low root mean square error (RMSE = 7.75 μg/m3). The results showed that some areas have very high PM10 concentration, reaching over 300 μg/m3, scattered distribution of industrial parks, industrial clusters, traffic routes where there is high traffic density or there are new urban areas and projects under construction. Meanwhile, some suburban areas have much lower concentration of PM10, only up to 50 μg m3 distributed northwest and northeast of the city center.