Keywords
pm10 forecast, ann, optimal interpolation
Start Date
1-7-2012 12:00 AM
Abstract
Over recent years, the high levels of air pollution have become a quite important problem due to their direct impact on human health. Due to these health risks, EU directive (2008/50/EC) recommends member states to ensure that timely information about actual and forecasted levels of pollutants are provided to the public. In order to follow these guidelines, prevent critical episodes and inform the public, environmental authorities need fast and reliable forecasting systems. In literature, air pollution forecasting systems can be broadly split in two main category: (1) deterministic systems, solving mass-balance differential equations for a large number of pollutant in atmosphere and (2) non deterministic data driven systems, based on the identification of stochastic models using measurement data collected by monitoring networks. In this work, an optimal interpolation technique has been used to integrate daily PM10 concentrations forecasted by artificial neural networks in 120 monitoring stations with the results of a deterministic chemical transport model. The methodology has been applied to a Northern Italy domain, characterized as one of the most polluted and industrialized area in Europe. Among measurement data from 2003-2007, year 2008 has been selected to test the performance of the integrated modelling system. The evaluation shows very good performances for all the stations, with high agreement in both mean value and 95th percentile computed over the entire year and correlation coefficient usually higher than 0.7 and small values of normalized mean error.
A combined Neural Network and Optimal Interpolation approach for PM10 forecast over Po Valley
Over recent years, the high levels of air pollution have become a quite important problem due to their direct impact on human health. Due to these health risks, EU directive (2008/50/EC) recommends member states to ensure that timely information about actual and forecasted levels of pollutants are provided to the public. In order to follow these guidelines, prevent critical episodes and inform the public, environmental authorities need fast and reliable forecasting systems. In literature, air pollution forecasting systems can be broadly split in two main category: (1) deterministic systems, solving mass-balance differential equations for a large number of pollutant in atmosphere and (2) non deterministic data driven systems, based on the identification of stochastic models using measurement data collected by monitoring networks. In this work, an optimal interpolation technique has been used to integrate daily PM10 concentrations forecasted by artificial neural networks in 120 monitoring stations with the results of a deterministic chemical transport model. The methodology has been applied to a Northern Italy domain, characterized as one of the most polluted and industrialized area in Europe. Among measurement data from 2003-2007, year 2008 has been selected to test the performance of the integrated modelling system. The evaluation shows very good performances for all the stations, with high agreement in both mean value and 95th percentile computed over the entire year and correlation coefficient usually higher than 0.7 and small values of normalized mean error.
