Estudo de níveis de ozônio troposférico usando métodos de estatística univariada e multivariada para duas localidades da cidade do Rio de Janeiro

Abstract

With increasingly intense urban and industrial development, currently a key challenge is to eliminate or reduce the impact caused by emissions of pollutants into the atmosphere. In the year 2012, the Rio de Janeiro hosted the Rio +20, the United Nations Conference about Sustainable Development, where representatives from around the world participated. At the time, among other issues discussed were the green economy and sustainable development. The tropospheric O3 presents itself as an extremely crucial variable due to its strong environmental impact, and knowing the behavior of the parameters that affect the air quality of a region, is useful for predicting scenarios. The chemistry of atmospheric sciences and meteorology are highly nonlinear and thus the forecasts of air quality parameters are hard to be determined. Air quality depends on emissions, meteorology and topography. The observed data were Nitrogen Dioxide (NO2), Nitrogen Monoxide (NO), Nitrogen Oxides (NOx), Carbon Monoxide (CO), Ozone (O3), Scalar Wind Speed (VEV), Global Solar Radiation (RSG), Temperature (TEM), Relative Humidity (UR) and collected through the mobile station monitoring the Secretaria do Meio Ambiente (SMAC) of Rio de Janeiro City in two places in the metropolitan area at Pontíficia Universidade Católica (PUC-Rio) and the Universidade do Estado do Rio de Janeiro (UERJ) at years 2011 and 2012. This study had three objectives: (1) to analyze the behavior of the variables, using the method of principal components analysis (PCA) of exploratory analysis; (2) propose forecasts of O3 levels from primary pollutants and meteorological factors, comparing the effectiveness of nonlinear methods like as artificial neural networks (ANN) and support vector machine regression (SVM-R), from primary pollutants and meteorological factors and finally, (3) perform data classification method using support vector machine classification (SVM-C). The PCA technique showed that for PUC dataset, variables NO, NOx and VSV have a greater impact on the concentration of O3 and the UERJ data set had the temperature (TEM) and Global Solar Radiation (RSG) as the most important variables. The results from the nonlinear regression techniques ANN and SVM obtained were very closely and acceptable to UERJ dataset presenting coefficient of determination (R2) for validation, 0.9122 and 0.9152 and Root Mean Square Error (RMECV) 7.66 and 7.85, respectively. As for the PUC and PUC + UERJ datasets, both techniques, obtained less satisfactory results. For these datasets, the SVM proved results slightly higher, and PCA, SVM and ANN had demonstrated their robustness presenting themselves as useful tools for understanding, classification and prediction scenarios for air quality