Abstract
The current study proposes a novel approach for the multi-model ensemble to be applied in air pollution forecasting. The methodology is based on decomposition of air pollution time series on different components (short-term, daily fluctuations, synoptic scale, etc.) and calibration of the ensemble for each of these components independently taking into account the performance of individual predictors. Therefore, the same model may have a different contribution for the ensemble at high and low frequency fluctuations. The Kolmogorov-Zurbenko (KZ) low-pass filter is used for the time series decomposition. The Fourier analysis is implemented to determine the contribution of different frequencies to the data variance allowing better understanding of the model performance and to define the ensemble weights. The methodology was tested using a group of four different air quality models that were applied over mainland Portugal for the 2006 year, and for main pollutants like O3 and PM10. The approach implemented in this work was compared with one of the most used ensemble technique showing clear advantages.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Eskridge R, Ku JY, Rao ST, Porter PS, Zurbenko IG (1997) Separating different time scales of motion in time series of meteorological variables. Bull Am Meteorol Soc 78(7):1473–1483
Monteiro A, Ribeiro I, Tchepel O, Carvalho A, Sá, Ferreira J, Galmarini S, Miranda AI Borrego C (2011) Bias correction and ensemble techniques to improve air quality assessment: focus on O3 and PM over Portugal. In Proceedings of the 14th international conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, Kos, Greece, 2–6 October 2011, 22–26
Rao ST, Zurbenko IG, Neagu R, Porter PS, Ku JY, Henry RF (1997) Space and time scales in ambient ozone data. Bull Am Meteorol Soc 78(10):2153–2166
Sebald L, Treffeisen R, Reimer E, Hies T (2000) Spectral analysis of air pollutants. Part 2: ozone time series. Atmos Environ 34:3503–3509
Tchepel O, Costa AM, Martins H, Ferreira J, Monteiro A, Miranda AI, Borrego C (2010) Determination of background concentrations for air quality models using spectral analysis of monitoring data. Atmos Environ 44(1):106–114
Vautard R, Schaap PM, Bergström R et al (2009) Skill and uncertainty of a regional air quality model ensemble. Atmos Environ 43(31):4822–4832
Acknowledgments
The authors acknowledge the Portuguese Environmental Protection Agency for the observational dataset support. The work was partly developed under the research projects POCI/AMB/66707/2006 and PTDC/CTE-ATM/103253/2008. Thanks are extended to the Portuguese Science Foundation (FCT) for the grants: SFRH/BPD/66874/2009, SFRH/ BD/60370/2009, SFRH BPD/40620/2007 and SFRH/BD/60474/2009.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Additional information
Questions and Answers
Questioner Name: Nicholas Savage
Q: For use as a forecasting tool you need to use past data to get forecast. Have you tried doing this and how much memory do you think these weights have?
A: Application of this technique to the operational forecast will be implemented as a next step. Taking into account that the methodology presented in this work is based on harmonic functions we expect a good “memory” of the weights for all the components (e.g. daily, synoptic, baseline) with possible exception for the highest frequencies.
Questioner Name: Efisio Solazzo
Q: Have you done any sensitivity test on the threshold value of driving forces? How does the ensemble change?
A: We tested different window size to separate the baseline component and the results confirm a robustness of the applied methodology.
Questioner Name: Rostislav Kouznetsov
Q: Separate weighting of harmonics might result in negative predicted concentrations. How such situations should be treated?
A: In theory it is possible to obtain negative values because of the negative signal in the filter residuals. However, the baseline of time series is always positive and it is very important for the final ensemble. In practice, the negative value means that the ensemble is not properly calibrated and the model weights should be analyzed more carefully.
Questioner Name: Jeremy Silver
Q: Were there gaps in the time-series of observations? If so, how did deal with this?
A: The completeness of the time series was one of the criteria to select the monitoring stations for the current analysis. Moreover, one of the advantages of KZ filter is its good performance in the presence of gaps in the time series.
Rights and permissions
Copyright information
© 2014 Springer Science+Business Media Dordrecht
About this paper
Cite this paper
Tchepel, O. et al. (2014). Improvement of Ensemble Technique Using Spectral Analysis and Decomposition of Air Pollution Data. In: Steyn, D., Builtjes, P., Timmermans, R. (eds) Air Pollution Modeling and its Application XXII. NATO Science for Peace and Security Series C: Environmental Security. Springer, Dordrecht. https://doi.org/10.1007/978-94-007-5577-2_84
Download citation
DOI: https://doi.org/10.1007/978-94-007-5577-2_84
Published:
Publisher Name: Springer, Dordrecht
Print ISBN: 978-94-007-5576-5
Online ISBN: 978-94-007-5577-2
eBook Packages: Earth and Environmental ScienceEarth and Environmental Science (R0)