Modelagem e previsão da qualidade do ar na cidade de Uberlândia - MG

Abstract

The expressive population growth of the Uberlândia city, in the last four decades, in a disordered way and without planning, has caused serious problems in population life quality. One of the most concerning points it is the atmospheric air quality. The main aim of this work was to develop empiric models to predict the PM10 concentration in downtown of Uberlândia-MG. Several linear models, such as AARX, ARMAX, output error and Box-Jenkins, as well as models based on neural networks were evaluated. The models used meteorological variables and vehicles flow as input data. As output predicted by the models, the PM10 concentration is obtained with 3-day horizon. The results showed that good estimates can be obtained by the models, unless for the output-error model. The best estimates of the PM10 concentration were obtained by the Box-Jenkins linear model. The proposed models can be used, among other purposes, for the local public government, as a tool to activate emergency actions during periods of atmospheric stagnation, when PM10 levels in atmosphere can represent risk to public health. The aim of this work was also to verify the relationship between the PM(10) concentration and respiratory diseases in Uberlândia. However, with methodologies used and available data, it was not possible to confirm this relationship.