Prediction of Black Carbon concentration in an urban site by means of different regression methods
- Autores: Carlos Marcos, Sara Segura Redón, Gustavo Camps Valls, Victor Estellés Leal, Roberto Pedrós, María Pilar Utrillas Esteban, José Antonio Martínez Lozano
- Localización: 2nd Iberian Meeting on Aerosol Science and Technology: Proceedings Book RICTA 2014 / coord. por Jordi Grifoll Taverna, Joan Rosell Llompart, 2014, ISBN 978-84-695-9978-5, págs. 91-94
- Idioma: inglés
- Enlaces
-
Resumen
Motor vehicle emissions are one of the major sources of Black Carbon (BC) in urban areas, where it contributes significantly to air pollution. However, quantifying the direct effect of traffic on BC concentration is not straightforward, since meteorological conditions may affect the distribution and transport of this pollutant. In this work we analyse the ability of four different regression methods to predict BC concentrations in the surroundings of a main highway, using traffic and meteorological data as predictors. We observe that, amongst the analysed methods, the best results are obtained with two non-linear models: Kernel Ridge Regression and Gaussian Process Regression. These results suggest that some processes affecting the BC concentration might not be properly described by linear models.
