Presenter/Author Information

C. Carnevale
G. Finzi
E. Pisoni
V. Singh
M. Volta

Keywords

ozone forecasting, neural network, co-kriging, spatial interpolation

Start Date

1-7-2008 12:00 AM

Abstract

An integrated forecasting system, consisting of Neural Network (NNs) models and co-kriging techniques, has been developed to forecast maximum eight hours ozone (max8h) concentration, two days in advance, over an urban domain including Milan area of northern Italy. Total numbers of available measurement stations falling within the domain are 23. NNs perform the forecasting at each measurement location and the co-kriging algorithm interpolates the forecasting data all over the domain. NNs have been identified for the period of 2000-2006. Leave-One-Out Cross Validation (LOOCV) has been performed to validate the results of NNs. To perform spatial interpolation of the forecasted maximum daily eight hour (max8h) ozone, co-kriging has been used. For validation of the proposed forecasting system, 5 out of 23 stations have been selected. Year 2004 has been chosen as a test case year to perform the overall forecast. The validation results show good agreement in terms of statistical indexes. The proposed forecasting methodology represents a fast and reliable way to provide decision makers and general public with ozone forecasting data over an urban area.

Share

COinS
 
Jul 1st, 12:00 AM

Neural Networks and Co-Kriging techniques to Forecast Ozone Concentrations in Urban Areas

An integrated forecasting system, consisting of Neural Network (NNs) models and co-kriging techniques, has been developed to forecast maximum eight hours ozone (max8h) concentration, two days in advance, over an urban domain including Milan area of northern Italy. Total numbers of available measurement stations falling within the domain are 23. NNs perform the forecasting at each measurement location and the co-kriging algorithm interpolates the forecasting data all over the domain. NNs have been identified for the period of 2000-2006. Leave-One-Out Cross Validation (LOOCV) has been performed to validate the results of NNs. To perform spatial interpolation of the forecasted maximum daily eight hour (max8h) ozone, co-kriging has been used. For validation of the proposed forecasting system, 5 out of 23 stations have been selected. Year 2004 has been chosen as a test case year to perform the overall forecast. The validation results show good agreement in terms of statistical indexes. The proposed forecasting methodology represents a fast and reliable way to provide decision makers and general public with ozone forecasting data over an urban area.