摘要
在研究了广州市能见度变化特征及低能见度发生的主要影响因素的基础上,利用广州市环境监测站2007-2009年的空气污染物(PM10、SO2、NO2)监测数据及同期地面气象要素(10min平均风速、最大风速、气温、相对湿度、露点温度、气压、24h降水量)观测资料筛选出主要的预报因子,用径向神经网络建立预报模型,并对2009年9月1日到12月25日的能见度进行预报试验。结果表明径向神经网络预报模型在能见度低于10km时预报准确率明显高于统计回归预报方程。采用分级方法统计得出在未出现低能见度情况下,中低能见度,中高能见度预报准确率分别为80%,69.6%,均高于线性回归预报方程(40%,47.8%)。
Based on the study of visibility variation and the main causes for low visibility in Guangzhou during 2007-2009, the appropriate predictors were selected from the concentrations of air pollutants （PM10, SO2, NO2） and surface meteorological elements （10minute average wind speed, maximum wind speed, air temperature, relative humidity, dew point temperature, barometric pressure, rainfall in 24 hours） . The radial basis function neural network was used to establish a prediction model to conduct the forecasting test （Sept. lth, 2009-Dec.25th, 2009） .The results indicated that neural network prediction model had a higher accuracy than the statistical model when visibility was less than 10km.Without extremely-low-value visibility in the test, the prediction accuracy of low-value and moderate-value visibility was 80% and 69.6% respectively, much higher than that produced by the statistical model （40%, 47.8%） .
出处
《气象研究与应用》
2014年第1期17-20,49,共5页
Journal of Meteorological Research and Application
基金
广州市气象局"广州市空气污染影响因子分析与建模研究"课题资助项目
关键词
能见度预报
神经网络
广州
visibility forecasting
neural network
Guangzhou
