Prediction of Tropospheric Ozone at Rio De Janeiro City From Primary Polluents and Meteorological Factors Using Artificial Neural Networks (Ann) and Support Vector Machine (Svm) Regression

Abstract

Tropospheric ozone is well known as an extremely important factor due to its strong environmental impact. The air quality depends on emissions, meteorology and topography. NO 2 , NO, NOx, CO, O 3 , scalar wind speed, solar radiation, temperature, and relative humidity (HUM) were monitored. These data sets were collected by the mobile station monitoring located at PUC-Rio and UERJ between 2011 and 2012, from the Secretary of the Environment of Rio de Janeiro. This study aimed at the prediction of O 3 from primary pollutants and meteorological factors. The obtained results of ANN and SVM regression techniques were acceptable to the dataset UERJ presenting coefficient of determination (R 2) for validation, 0.9122 and 0.9152, and the square root of the mean square error of cross validation (RMECV) 7.66 and 7.85, respectively. Regarding the data set of PUC-Rio, the two techniques gave inferior results.