Journal
SCIENCE OF THE TOTAL ENVIRONMENT
Volume 463, Issue -, Pages 875-883Publisher
ELSEVIER
DOI: 10.1016/j.scitotenv.2013.06.093
Keywords
Artificial neural networks; Air pollution; PM10 forecasting; Regional background; Urban background; Sources
Categories
Funding
- Autonomous Government of Catalonia
- Spanish Ministry of Environment and the Spanish Ministry of Science and Innovation [CGL2010-19464-E/CLI]
- Puglia Region, Italy
- Department of Chemistry of the University of Bad (Italy)
- Institute of Environmental Assessment and Water Research, in the Spanish Research Council
Ask authors/readers for more resources
An artificial neural network (ANN) was developed and tested to forecast PM10 daily concentration in two contrasted environments in NE Spain, a regional background site (Montseny), and an urban background site (Barcelona-CSIC), which was highly influenced by vehicular emissions. In order to predict 24-h average PM10 concentrations, the artificial neural network previously developed by Caselli et al. (2009) was improved by using hourly PM concentrations and deterministic factors such as a Saharan dust alert. In particular, the model input data for prediction were the hourly PMio concentrations 1-day in advance, local meteorological data and information about air masses origin. The forecasted performance indexes for both sites were calculated and they showed better results for the regional background site in Montseny (R-2 = 0.86, SI = 0.75) than for urban site in Barcelona (R-2 = 0.73, SI = 0.58), influenced by local and sometimes unexpected sources. Moreover, a sensitivity analysis conducted to understand the importance of the different variables included among the input data, showed that local meteorology and air masses origin are key factors in the model forecasts. This result explains the reason for the improvement of ANN's forecasting performance at the Montseny site with respect to the Barcelona site. Moreover, the artificial neural network developed in this work could prove useful to predict PM10 concentrations, especially, at regional background sites such as those on the Mediterranean Basin which are primarily affected by long-range transports. Hence, the artificial neural network presented here could be a powerful tool for obtaining real time information on air quality status and could aid stakeholders in their development of cost-effective control strategies. (C) 2013 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available