Search

Air pollution in urban areas is an important environmental problem related to the quality of life in cities, due to its potential significant effects on human health. Particulate matter of 2.5 fraction, PM2.5, are particles with the diameter less than 2.5 μm, being an air pollutant type of special concern for sensitive people, such as children and elderly people, as they can penetrate the lungs and can cause serious health problems. Thus, it is desirable to have an efficient PM2.5 forecasting system which informs the population about the air pollution episodes, with exceedances of the PM2.5 allowed concentration limit, and provide some advices to reduce the effects on sensitive people health. The development of such a system is performed under the ROKIDAIR research project. This paper focuses on the application of a PM2.5 forecasting method based on artificial intelligence, and proposes a short-term PM2.5 forecasting model that uses an adaptive neuro-fuzzy inference system (ANFIS). The proposed method is applied to an hourly PM2.5 dataset from Ploiesti city, an industrial town from Romania, with a history of air pollution episodes. The forecasting models have two types of inputs, only PM2.5 concentrations and PM2.5 concentrations plus temperature, and the results obtained with the two models are compared using statistical indexes. [ABSTRACT FROM AUTHOR]