Study on Circulation Classification Based Surface Ozone Concentration Prediction Model
LIANG Zhuo-ran,GU Ting-ting,YANG Xu-chao,DU Rong-guang,ZHONG Hong-lin,Qi Bing
DOI: https://doi.org/10.3969/j.issn.1000-6923.2017.12.008
2017-01-01
Abstract:In this study, a new surface ozone concentration level simulation model was developed by combining the Lamb-Jenkinson objective circulation classification method and the stepwise linear regression model, and successfully applied in Hangzhou. The influence of local meteorological factors and large-scale circulation factors, which are from large scale reanalysis data between 2011 and 2016, to the ozone concentration level are analysed first. Our results showed a strong seasonal trend of the surface ozone concentration in Hangzhou. Ozone concentration level was much higher during spring and summer, and frequently exceeded the national standard which caused ozone pollution events, especially in May. Local meteorological conditions had a significant impact on the ozone concentration level. Both solar radiation and daily maximum temperature are positively correlated to ozone concentration level, but the relative humidity and precipitation are negatively correlated to it. In addition, we also classified the local cyclonic circulation observations using Lamb-Jenkinson objective circulation classification method, and identified that different cyclonic circulation types had different impact to the ozone concentration. The anti-cyclonic circulation was the dominant type and accounted for 26.5% of all the cyclonic circulation in Hangzhou, while the circulation from northwest had the least frequency of 0.6%. Ozone pollution events happened most frequently (23.8%) under the south airflow controlled circulation condition, but had the lowest frequency under north airflow controlled circulation condition (3.7%). Our results proved that the performance of the ozone concentration level simulation model will improve greatly by considering the seasonal circulation types and the meteorological factors using the stepwise regression method, which could increase the correlation between model prediction and observations up to 0.87, and our new model also improves the simulation of the extreme high ozone concentration pollution events significantly, it successfully predicted 15ozone pollution events out of 24events that happened in 2016, with a high TS score of 52%.