Application of ensemble forecast and linear regression method in improving PM10 forecast over Beijing areas

Si Huang, Xiao Tang, Wenshuai Xu, Zhe Wang, Huansheng Chen, Jie Li, Qizhong Wu, Zifa Wang

Research output: Contribution to journalArticle

11 Citations (Scopus)

Abstract

In this study, ensemble forecast combined with linear regression method is used to reduce the uncertainty in air quality models. Firstly, the PM10 forecasts by three models (NAQPMS, CAMx and CMAQ) in EMS-Beijing are evaluated over Beijing areas. In order to improve the forecast performance, the linear regression method (REG) is used to combine the forecast results of the three models and is compared with the ensemble mean method. The results show that for single model forecast, great difference exists among different models and no model performs much better for all statistic indexes than the other two models. Overall, CMAQ performs better in tendency prediction, while NAQPMS has smaller root mean square errors than the other two models. Ensemble mean method presents poor performance in improving the PM10 forecasts from the three models. On the other hand, REG brings significant improvement of the PM10 forecast. When an appropriate training length (36 days) is applied, the root mean square errors of PM10 forecast over 28 stations of Beijing is reduced by 32%~43% when using REG and the bias decreased considerably to 5.8 μg·m-3. This result implies that REG can greatly improve forecast performance than single model and ensemble mean forecast. Furthermore, the REG also greatly improve capturing of pollution episode forecast.

Original languageEnglish
Pages (from-to)56-64
Number of pages9
JournalHuanjing Kexue Xuebao/Acta Scientiae Circumstantiae
Volume35
Issue number1
DOIs
Publication statusPublished - Jan 6 2015

Fingerprint

Linear regression
Mean square error
forecast
method
Air quality
Pollution
air quality
Statistics
pollution

All Science Journal Classification (ASJC) codes

  • Environmental Engineering
  • Environmental Chemistry
  • Environmental Science(all)

Cite this

Application of ensemble forecast and linear regression method in improving PM10 forecast over Beijing areas. / Huang, Si; Tang, Xiao; Xu, Wenshuai; Wang, Zhe; Chen, Huansheng; Li, Jie; Wu, Qizhong; Wang, Zifa.

In: Huanjing Kexue Xuebao/Acta Scientiae Circumstantiae, Vol. 35, No. 1, 06.01.2015, p. 56-64.

Research output: Contribution to journalArticle

Huang, Si ; Tang, Xiao ; Xu, Wenshuai ; Wang, Zhe ; Chen, Huansheng ; Li, Jie ; Wu, Qizhong ; Wang, Zifa. / Application of ensemble forecast and linear regression method in improving PM10 forecast over Beijing areas. In: Huanjing Kexue Xuebao/Acta Scientiae Circumstantiae. 2015 ; Vol. 35, No. 1. pp. 56-64.
@article{9ca2105f5ea6490898ef70ab90b466b2,
title = "Application of ensemble forecast and linear regression method in improving PM10 forecast over Beijing areas",
abstract = "In this study, ensemble forecast combined with linear regression method is used to reduce the uncertainty in air quality models. Firstly, the PM10 forecasts by three models (NAQPMS, CAMx and CMAQ) in EMS-Beijing are evaluated over Beijing areas. In order to improve the forecast performance, the linear regression method (REG) is used to combine the forecast results of the three models and is compared with the ensemble mean method. The results show that for single model forecast, great difference exists among different models and no model performs much better for all statistic indexes than the other two models. Overall, CMAQ performs better in tendency prediction, while NAQPMS has smaller root mean square errors than the other two models. Ensemble mean method presents poor performance in improving the PM10 forecasts from the three models. On the other hand, REG brings significant improvement of the PM10 forecast. When an appropriate training length (36 days) is applied, the root mean square errors of PM10 forecast over 28 stations of Beijing is reduced by 32{\%}~43{\%} when using REG and the bias decreased considerably to 5.8 μg·m-3. This result implies that REG can greatly improve forecast performance than single model and ensemble mean forecast. Furthermore, the REG also greatly improve capturing of pollution episode forecast.",
author = "Si Huang and Xiao Tang and Wenshuai Xu and Zhe Wang and Huansheng Chen and Jie Li and Qizhong Wu and Zifa Wang",
year = "2015",
month = "1",
day = "6",
doi = "10.13671/j.hjkxxb.2014.0728",
language = "English",
volume = "35",
pages = "56--64",
journal = "Huanjing Kexue Xuebao / Acta Scientiae Circumstantiae",
issn = "0253-2468",
publisher = "Science Press",
number = "1",

}

TY - JOUR

T1 - Application of ensemble forecast and linear regression method in improving PM10 forecast over Beijing areas

AU - Huang, Si

AU - Tang, Xiao

AU - Xu, Wenshuai

AU - Wang, Zhe

AU - Chen, Huansheng

AU - Li, Jie

AU - Wu, Qizhong

AU - Wang, Zifa

PY - 2015/1/6

Y1 - 2015/1/6

N2 - In this study, ensemble forecast combined with linear regression method is used to reduce the uncertainty in air quality models. Firstly, the PM10 forecasts by three models (NAQPMS, CAMx and CMAQ) in EMS-Beijing are evaluated over Beijing areas. In order to improve the forecast performance, the linear regression method (REG) is used to combine the forecast results of the three models and is compared with the ensemble mean method. The results show that for single model forecast, great difference exists among different models and no model performs much better for all statistic indexes than the other two models. Overall, CMAQ performs better in tendency prediction, while NAQPMS has smaller root mean square errors than the other two models. Ensemble mean method presents poor performance in improving the PM10 forecasts from the three models. On the other hand, REG brings significant improvement of the PM10 forecast. When an appropriate training length (36 days) is applied, the root mean square errors of PM10 forecast over 28 stations of Beijing is reduced by 32%~43% when using REG and the bias decreased considerably to 5.8 μg·m-3. This result implies that REG can greatly improve forecast performance than single model and ensemble mean forecast. Furthermore, the REG also greatly improve capturing of pollution episode forecast.

AB - In this study, ensemble forecast combined with linear regression method is used to reduce the uncertainty in air quality models. Firstly, the PM10 forecasts by three models (NAQPMS, CAMx and CMAQ) in EMS-Beijing are evaluated over Beijing areas. In order to improve the forecast performance, the linear regression method (REG) is used to combine the forecast results of the three models and is compared with the ensemble mean method. The results show that for single model forecast, great difference exists among different models and no model performs much better for all statistic indexes than the other two models. Overall, CMAQ performs better in tendency prediction, while NAQPMS has smaller root mean square errors than the other two models. Ensemble mean method presents poor performance in improving the PM10 forecasts from the three models. On the other hand, REG brings significant improvement of the PM10 forecast. When an appropriate training length (36 days) is applied, the root mean square errors of PM10 forecast over 28 stations of Beijing is reduced by 32%~43% when using REG and the bias decreased considerably to 5.8 μg·m-3. This result implies that REG can greatly improve forecast performance than single model and ensemble mean forecast. Furthermore, the REG also greatly improve capturing of pollution episode forecast.

UR - http://www.scopus.com/inward/record.url?scp=84921634148&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84921634148&partnerID=8YFLogxK

U2 - 10.13671/j.hjkxxb.2014.0728

DO - 10.13671/j.hjkxxb.2014.0728

M3 - Article

AN - SCOPUS:84921634148

VL - 35

SP - 56

EP - 64

JO - Huanjing Kexue Xuebao / Acta Scientiae Circumstantiae

JF - Huanjing Kexue Xuebao / Acta Scientiae Circumstantiae

SN - 0253-2468

IS - 1

ER -