Keywords: atmospheric pollution, PM25, PM10-25, time series, particulate matter, air pollution, air quality, Sao Carlos, Brazil
The temporal trends of PM2.5 and PM10−2.5 fractions of the atmospheric Particulate Matter (PM) concentrations in the city centre of Sao Carlos were studied. The mass concentrations were measured using a dichotomous sampler. The samples were obtained weekly using a 24-hour sampling time, and the period of study covers the years 2001 to 2006. The SARIMA model from the R statistical package was used for modelling. The SARIMA obtained for each series of PM were compared to results obtained with the Holt?Winters method. Both models provided satisfactory results and were sensitive to seasonal and local events. They can, therefore, be successfully used as an analytical tool for similar atmospheric studies elsewhere. The SARIMA results showed better prediction performance, with smaller values of RMSE and RMSE percentage (values of 4.67 and 0.92% for the PM2.5 and 4.55 and 2.73% for PM10−2.5, respectively) than the Holt?Winters models.