Please use this identifier to cite or link to this item:
|Title:||Ensemble techniques to improve air quality assessment: Focus on O3 and PM|
|Authors:||MONTEIRO A.; RIBEIRO I.; TCHEPEL O.; CARVALHO A.; MARTINS H.; SA E.; FERREIRA J.; GALMARINI Stefano; MIRANDA A.i.; BORREGO C.|
|Citation:||ENVIRONMENTAL MODELING \& ASSESSMENT|
|Type:||Articles in Journals|
|Abstract:||Five air quality models were applied over Portugal for July 2006 with an ensemble purpose. These models were used, with their own meteorology, parameterizations, boundary conditions and chemical mechanisms, but with the same emission data. The validation of the individual models and its ensemble for ozone (O3) and particulate matter (PM) was performed using monitoring data from 22 background stations over Portugal. After removing the bias from each model, different ensemble techniques were applied and compared. Besides the median (MED), several weighted ensemble approaches were tested and intercompared: static (SLR) and dynamic (DLR) multiple linear regressions (using less-square optimization method) and the Bayesian Model Averaging (BMA) methodology. The goal of the comparison is to estimate to what extent the ensemble analysis is an improvement with respect to the single model results. The obtained results revealed that no one of the 4 tested ensembles clearly outperforms the others on the basis of statistical parameters and probabilistic analysis (reliability and resolution properties). Nevertheless, statistical results have shown that the application of the weights slightly improves ensemble performance when compared to those obtained from the median ensemble. The same statistical analysis together with the probabilistic measures demonstrate that the SLR and BMA methods are the best performers among the assessed methodologies.|
|JRC Institute:||Institute for Environment and Sustainability|
Files in This Item:
There are no files associated with this item.
Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.