Using the multiple linear regression model and the time-stratified approach to PM10

Abstract

Air Quality Modeling gained great importance in atmospheric pollution because of its negative effects on the environment and human health. In our study, the rate of the PM10 has been effected by a large number of elements such as (O3,CO2,SO3,NO,NOX). It is also effected by other explanatory variables such as wind speed, temperature and others. Nine independent variables have been used in the application of the multiple regression model over three years of metrological datasets. The seasonal influences for weekly, seasonally periods lead up to difficult analyzing and forecasting. Classification the datasets into weekly and seasonally then using Time-stratified (TS) approach and taking only the compatible observations may solve the difficulty and lead to more accurate results. Reducing the number of variables may also lead to more accurate results. The multiple linear regression (MLR) model is the most common for studying like this number of variables.. Three years of Malaysian meteorological datasets are studied in this research. It was noted that the results of prediction in a method MLR one of the best methods to use such data. Appling the Time-stratified approach for PM10 data according to the weekly and seasonally pattern were better and more accurate for MLR comparing to the original data.