Forecasting of surface ozone concentrations 24 hours in advance using neural networks
This source preferred by Emili Balaguer-Ballester
Authors: Ballester, E.B., Olivas, E.S., Carrasco-rodriguez, J.L. and Del Valle-tascon, S.
Journal: International conference o Neural Networks and Applications
This data was imported from Scopus:
Authors: Ballester, E.B., Olivas, E.S., Carrasco-Rodriguez, J.L. and Del Valle-Tascon, S.
Journal: Advances in Neural Networks and Applications
This study test the usefulness of neural network models for the prediction of 24-hours ahead O3 concentration. For a rural environment in eastern Spain, the O3, meteorological, and nitrogen oxides (NO and NO2) data for the three ozone seasons (March to October) periods have been used to develop models to predict one-day ahead ambient O3 concentrations. The problem is examined initially for the univariate case, and it is extended to include additional meteorological and nitrogen oxides parameters in the process of estimating the optimum models. Three performance measures, namely, root mean square error, coefficient of determination, and index of agreement yield similar results for both univariate and multivariate models. The results indicate that the developed neural models predict the O 3 time series more effectively compared to the previous procedures based on dynamical system theory. Under conditions too demanding for advanced physico/chemical models, the univariated models may offer useful alternative to derive O3 prediction directly from O3 data time series. The MLP models can be used for EU public advisories.