Statistical Analysis for Pollutants in Precipitation Observed in a Selected Region

Abstract

<p>The study of pollution and its effects on the environment has long been an interest for many researchers. Short and long term affects as well as future predictions based on past observations are important factors to consider when one undertakes such a study. The purpose of this thesis is to observe the long term changes and trends of pollutants in precipitation selected from the north-eastern region of the United States of America and Canada. The data was collected on a weekly basis between 1995 to 2006 on air pollutants Ammonium. Nitrate. any type of Sulphate. and Hydron (NH4: N03. XS04. and H+. respectively). In total. 19 different stations 'vvere investigated. Two types of statistical models were fit to the data which include the gamma regression and the random effect model. The gamma regression assumed independence of any spatial and temporal factors. This was used to conceptualize the overall trend and yearly fit. The preliminary analysis found strong evidence of spatial dependence. but temporal dependence was so weak that it could be ignored. The random effect model has been adopted to handle dependencies caused by any underlying mechanisms. Pollutant NH4 had no significant factors resulting from the fitting of the random effect model and the Year effect was non-significant. In the result for pollutant N03: the coefficient of Year was significant and decreasing. Pollutant XSO'l was revealed to have a significant and decreasing Year effect. The random effect model did not produce any significant factors for pollutant H+. Overall. the random effects model is a more reasonable approach to fitting this data because it considers spatial dependence where the gamma regression assumes independent responses.</p>