| Owing to the impacts on air quality, climate change and human health,atmospheric environment change, especially frequent air compound pollution in East China resulted from human activities and meteorological conditions, has attracted worldwide attention in scientific and public communities over recent years.Therefore, this thesis attempts to establish a Short-term Statistical forecasting model for regional Air Quality in Jiangsu Province in East China, based on the statistical analysis on the environmental and meteorological data combining with the WRF modeling. For this purpose, the research work in this thesis is focused on three issues:1) understanding the relationship between meteorological variations and air quality change, 2) exploring the mechanism on effect of meteorological change on air quality,and 3) establishing statistical forecasting model system for application in operational air quality prediction in Jiangsu Province, China. The major work of this thesis is summarized:（1） Variations in major air pollutants in Nanjing and their meteorological driversIn order to study the seasonal variations major air combined pollutants PM2.5,PM10 and O3 and their influencing factors of meteorology in Nanjing, the environmental monitoring data from January 2013 to February 2015 and the fine meteorological fields in the boundary layer produced by the high resolution WRF modeling were analyzed statistically. From the major air pollutant characteristics,correlations between meteorological factors and pollutant concentrations, as well as the stepwise regression fitting, it is concluded that the seasonal values of PM2.5 and PM10 were the highest in winter and lowest in summer, and daily mean of PM2.5 and PM10 concentrations reached up to 160.6μg·m-3 and 98.0μg·m-3 in winter; and their diurnal changes were distinct from autumn to winter and weak in summer with the similar patterns in four seasons. The daytime peaks of diurnal PM2.5 and PM10 variations in winter delayed 1-2 hours compared to other three seasons. Fine particles dominated atmospheric particles in all seasons and annual mean ratio of PM2.5/PM10 was 0.59. The annual frequencies of PM2.5, PM10 and O3 dominating air pollution wererespectively 51.5%,26.6% and 13.5%. PM2.5 controlled the AQI changes in four seasons,especially during heavy pollution periods, PM2.5 is the primary pollutant with 96% contribution in Nanjing. O3 shifted seasonally between the peak in late spring and early summer and the bottom in late autumn and early winter with a unimodal pattern in diurnal change of O3. O3 was positively related to air temperature and boundary layer height, while PM10, PM2.5 showed the significantly negative correlations with wind speed. The fitting goodness of stepwise regressions for the daily concentrations of PM2.5,PM10 and O38hmax ranged reasonably over 40%-65% in four seasons.（2） Interaction mechanism of major air compound pollution PM2.5 and O3 in NanjingBy using data analysis of air pollutant measurements from 2013 to 2015 in an urban area of East China, Nanjing, we found that the correlation coefficients between major atmospheric combined pollutants of PM2.5 and O3 in hot （from June to August）and cold （from December to February） season were respectively 0.40 and -0.16,passing the confidence level of 99%. Based on the environmental and meteorological data, the cause of the inversion relations between PM2 5 and O3 in the cold and hot seasons was investigated from the aspects of atmospheric oxidation and solar radiation effect of PM2.5. Our study showed that the augmentation of atmospheric oxidation strengthened production of the secondary aerosols with the highest contribution of 26.76% to ambient PM2.5 level. The high O3 concentration in the warm season promoted the formation of the secondary aerosol resulting in a positive correlation between PM2.4 and O3. In cold season with weak atmospheric oxidation,the high PM2.5 concentrations from the primary emission suppressed the surface solar radiation. With the enhanced PM2.5 level, the diurnal pattern of O3 varied with the lower and delayed the peaks, and the daily O3 increment decreased obviously. When the ambient PM2.5 level exceeded 115μg·m-3, the surface O3 concentration dropped to 12.7μg·m-3 at noon with a significant inhibitory effect, leading to a negative correlation between PM2.5 and O3. This observational study revealed the interaction of PM2.5 and O3 in air combined pollution for understanding on seasonal change of atmospheric environment.（3） Establishment of air quality statistical forecasting model in Yangtze River DeltaThe statistical forecasting prediction system of air quality is developed by using the linear stepwise regression equation and artificial neural network model based on principal component analysis. The important predictors are seasonally selected by the correlation coefficients from a set of predictors including air pollutant concentrations and meteorological elements over the prior and current days. The results indicated that the Back-Propagation neural network model is much better than liner regression prediction equation in independent sample test. The fitting goodness of stepwise regressions for the daily concentrations of PM2.5, PM10 and O3 ranged 40%-65% and the accuracy was 43% -55% in four seasons. The accuracy of the BP neural network model for the future 72 hours could reach up to 61%- 77%. The air quality statistical forecasting system is developed, which could potentially be used in the operational air quality forecast. When the length of the training set was increased to 4000 hours,the average accuracy increased to 73.31%Compared with the full time prediction model, the accuracy of the heavy pollution model is improved by 39.98%..It is helpful to improve the stability of the forecast by increasing the prediction set length.Therefore, it is suggested that the long-term historical data should be used to train the neural network and Based on this, the air quality statistical forecasting model in Yangtze River Delta is established. |
PDF Full Text Request