Application of Multiple Linear Regression and Geographically Weighted Regression Model for Prediction of PM2.5

The present study deals with the assessment of the spatial distribution of PM2.5 over a decade in Jharkhand state of eastern India, which is a prominent site for mining and industries. Since there are very few monitoring stations on the ground to monitor the air quality of the entire state, satellite data have been utilised. The study period is from the year 2005 to 2016. The selection of the study period is based on the availability of satellite as well as ground station data. Multiple linear regression and geographically weighted regression (GWR) model was employed to predict the concentration of PM2.5 spatially, and the results were compared with the help of Akaike information criterion to identify the better representative model. Results showed that the GWR model performed better in predicting the spatial distribution of PM2.5. PM2.5 concentration of this state exceeds the permissible limit set by the world health organisation. The north-eastern districts of the state (29.36% of the total area) had exceeded even the Indian national ambient air quality standard. The identification of the possible reasons for high concentration was made through visual examination of satellite imageries over the study period. Also, possible health effects were discussed.