Prediction of PM10 Concentrations For Bahía Blanca, Argentina

CHIARVETTO PERALTA L.L.; DIAZ M. F.; BRIGNOLE N. B.

Bahía Blanca

Congreso; XVIII Congreso Argentino de Ciencias de la Computación, CACIC 2012: XIII Workshop Agentes y Sistemas Inteligentes (WASI); 2012

Universidad Nacinal del Sur, Dpto de Ciencias en Ingeniería de la Computación

PM10 non-traditional modelling for e-government development is described in detail. Ambient PM10 concentrations were predicted using meteorological variables as inputs, whose relevance for a generated Artificial Neural Network was analyzed by a feature selection method. The work is specially focused on the surroundings of Bahía Blanca city, its petrochemical pole and Ing. White grain port. Its accuracy was tested with time windows ranging from 2004 to 2006. A trustworthy simulation of the physical phenomena was built. As a result, this predictive model will contribute to the local observatory in order to trigger early-alert warnings.