Forecasting O3 levels in industrial area surroundings up to 24 h in advance, combining classification trees and MLP models

Özet

A two-step methodology was developed to forecast tropospheric ozone (O3) concentration levels, k hoursahead (k ¼ 1, 8, 12, 24), combining meteorological, air quality and industrial emissions data, across threeair quality monitoring stations in Sines Portuguese region. Firstly, the best O3 concentration predictorshave been identified through Classification and Regression Trees techniques; then Multilayer Perceptronmodels were adopted to forecast O3 levels for each monitoring site.The obtained generalization model performances are very good to classify in advance the expected classof O3 concentration level. Performance results vary from 70% of success to forecast O3 class above 70 mg/m324 h in advance up to 99% to predict the next hour in advance. These successful results are favorable to beimplemented in a real-time tool for health and environmental advisories, allowing the forecast of airpollutants concentrations up to 24 h ahead, improving the local air quality management systems.