Articles | Volume 16, issue 23
https://doi.org/10.5194/acp-16-14843-2016
https://doi.org/10.5194/acp-16-14843-2016
Research article
 | 
30 Nov 2016
Research article |  | 30 Nov 2016

Seasonal prediction of winter haze days in the north central North China Plain

Zhicong Yin and Huijun Wang

Abstract. Recently, the winter (December–February) haze pollution over the north central North China Plain (NCP) has become severe. By treating the year-to-year increment as the predictand, two new statistical schemes were established using the multiple linear regression (MLR) and the generalized additive model (GAM). By analyzing the associated increment of atmospheric circulation, seven leading predictors were selected to predict the upcoming winter haze days over the NCP (WHDNCP). After cross validation, the root mean square error and explained variance of the MLR (GAM) prediction model was 3.39 (3.38) and 53 % (54 %), respectively. For the final predicted WHDNCP, both of these models could capture the interannual and interdecadal trends and the extremums successfully. Independent prediction tests for 2014 and 2015 also confirmed the good predictive skill of the new schemes. The predicted bias of the MLR (GAM) prediction model in 2014 and 2015 was 0.09 (−0.07) and −3.33 (−1.01), respectively. Compared to the MLR model, the GAM model had a higher predictive skill in reproducing the rapid and continuous increase of WHDNCP after 2010.

Download
Short summary
Recently, the winter haze pollution over the north central North China Plain has become severe. By treating the year-to-year increment as the predictand, two new statistical schemes were established using the multiple linear regression and the generalized additive model approaches. After cross validation, both of these models could capture the interannual and interdecadal trends and the extremums successfully. Independent tests for 2014 and 2015 also confirmed the good predictive skill.
Altmetrics
Final-revised paper
Preprint