Development and performance evaluation of statistical models correlating air pollutants and meteorological variables at Pantnagar, India

Abstract

Ambient air quality in respect of SO2, NO2 and total suspended particulate matter (TSPM) was monitored at Pantnagar, India from May, 2008 to April, 2009 and statistically analyzed with meteorological variables such as relative humidity (RH), wind speed (WS), precipitation (P) and mean air temperature (T). TSPM was found to be the major air pollutant causing significant deterioration of air quality with annual mean concentrations of 280μg/m3. Further, weekly mean air pollutant concentrations were statistically analyzed through stepwise multiple linear regression analysis in respect of independent meteorological variables to develop suitable statistical models. Both NO2 and TSPM concentrations were found to have been influenced by meteorological variables with coefficient of determination (R2) of 82.21 and 92.84%, respectively. However, atmospheric SO2 revealed only 22.87% of dependencies on meteorological variables. Partial correlation coefficients revealed that wind speed has the maximum influence (77.80 and 31.50%) on proposed equations for NO2 and SO2, closely followed by weekly mean temperature (73.60 and 24.30%). However, in case of TSPM, individual contribution of ambient temperature (94.40%) was found maximum, followed by relative humidity (86.50%). Model performances were evaluated through both quantitative data analysis techniques and statistical methods. Nearly 98 and 95% of potential error has been explained by the model developed for TSPM and NO2, while in case of SO2, it is found as only 61%. Therefore, performances of models (for TSPM and NO2) to predict ambient weekly mean concentrations based on forecasted weather parameters were found to be excellent, however, performance of model developed for SO2 was found only satisfactory.