Self-Organization Maps Classification And Integrated Forecast Of PM_2.5Typical Pollution Weather In Eastern China

In recent years,with rapid development of China's eastern economy,air pollution seriously increased in eastern China.Eastern region is main economic region of China,so it is a current important issue that air pollution was well predicted in eastern china.This paper used the Empirical Orthogonal Function to study PM_2.5 pollution situation and the Self-Organizing Feature Map to study weather situation in eastern China.Based on this,this paper used multivariate linear regression integration model to predict daily average concentration of PM_2.5.The results shows:?1?The Empirical Orthogonal Function got three main spatial distribution modes of PM_2.5 daily average concentration in eastern china,and that are consistent distribution?north opposite with south type?North and south ends opposite with middle type.?2?The Self-Organizing Feature Map got nine weather situations,and they are Mongolian High?Inland High?High-pressure Front?Northeast High?Uniform Pressure Field?L High Pressure?The Coastal Trough?Northeast into the sea high pressure and east china into the sea high pressure.?3?Mongolian High?Northeast High?Uniform Pressure Field?Northeast into the sea high pressure and east china into the sea high pressure easily caused PM_2.5 pollution in Beijing?Tianjin?Hebei and Henan.Uniform Pressure Field and L High Pressure easily caused PM_2.5 pollution in Yangtze River Delta region.Inland High and High-pressure Front difficultly caused PM_2.5 pollution in eastern china.?4?Mongolian High?L High Pressure and east china into the sea high pressure had the longer life span in winter half year.During 48 hours,weather situation was easier to change that includes Northeast High->Northeast into the sea high pressure,Uniform Pressure Field->east china into the sea high pressure,Northeast into the sea high pressure->The Coastal Trough.?5?Multivariate linear regression integration model was used to forecast PM_2.5 daily average concentration in Beijing?Nanjing?Xiamen?Shanghai and Zhengzhou,and the result of Multivariate linear regression integration model is better than the three MOS models.Multivariate linear regression integration model is better than the three MOS models in forecasting PM_2.5 daily average concentration in Beijing?Nanjing?Xiamen.Shanghai and ZhengzhouThe reasonable accuracy of PM_2.5 prediction displayed in this study show that the multivariate linear regression integration model approaches perform well during the study periods and have great potential to be regional air pollution prediction in operation mode.