Keywords

SVM, PAH, PM2.5, Urban maps, health

Location

Session E2: Environmental Modeling of Human Health Effects from Global to Local Scale

Start Date

18-6-2014 2:00 PM

End Date

18-6-2014 3:20 PM

Abstract

Epidemiological studies about health effects of air quality are often based on data inferred by monitoring stations, and the issue of constructing pollutants exposure maps is crucial for improving such studies. The study about Polycyclic Aromatic Hydrocarbons (PAHs) exposure in urban areas is the aim of the EXPAH LIFE+ Project, so an integrated approach, based on measurements and modeling techniques, has been applied to simulate PAHs concentration in the urban area of Rome in one year period (June 2011 - May 2012). Support Vector Machines (SVMs) have been applied to forecast PAHs concentrations starting from actual measurements. After a feature selection process, the SVM has been trained and tested with blind samples, showing very significant results. Then, the same SVM has been used for building daily PAHs exposure maps. In our work, since not all the actual measurements were available, new indices have been defined for assessing the maps. All the outputs produced by the SVM have been also compared with those obtained by two applications of chemical transport models (FARM be and FARM fc). The overall results suggest the applicability of SVM methods in estimating daily and annual PAHs exposure in urban areas.

Share

COinS
 
Jun 18th, 2:00 PM Jun 18th, 3:20 PM

Estimation of PAHs concentration fields in an urban area by means of Support Vector Machines

Session E2: Environmental Modeling of Human Health Effects from Global to Local Scale

Epidemiological studies about health effects of air quality are often based on data inferred by monitoring stations, and the issue of constructing pollutants exposure maps is crucial for improving such studies. The study about Polycyclic Aromatic Hydrocarbons (PAHs) exposure in urban areas is the aim of the EXPAH LIFE+ Project, so an integrated approach, based on measurements and modeling techniques, has been applied to simulate PAHs concentration in the urban area of Rome in one year period (June 2011 - May 2012). Support Vector Machines (SVMs) have been applied to forecast PAHs concentrations starting from actual measurements. After a feature selection process, the SVM has been trained and tested with blind samples, showing very significant results. Then, the same SVM has been used for building daily PAHs exposure maps. In our work, since not all the actual measurements were available, new indices have been defined for assessing the maps. All the outputs produced by the SVM have been also compared with those obtained by two applications of chemical transport models (FARM be and FARM fc). The overall results suggest the applicability of SVM methods in estimating daily and annual PAHs exposure in urban areas.