基于Logistic回归的肇庆市区雾天气的预报模型
|
摘要
利用2015—2018年高要区国家基本气象常规观测资料,选取11个与雾天气事件相关的气象要素,通过主成分分析降维得到新的公共因子后构建Logistics回归方程,并根据ROC曲线、约登指数选出作为判别雾天气事件的最优临界值。结果表明:气压、温度露点差、相对湿度、水汽压、绝对湿度、露点、10 min平均风速、能见度、总云量、低云量、日照与雾天气现象具有很好的相关性;从这11个气象要素中提取3个具有代表性的主成分因子分别为水汽因子、辐射因子和风速因子。建立的Logistic回归预报模型对有雾和无雾的判别效果较为理想,比判别大雾和轻雾效果好。
Eleven foggy weather-related meteorological elements were selected based on conventional observations from national basic sites in the Gaoyao District, Zhaoqing from 2015 to 2018, a logistic regression equation was constructed from new common factors obtained by principle component analysis and dimension reduction, and optimal critical values were determined to identify foggy weather events following the ROC curve and Youden's index. The result is shown as follows. The meteorological elements of air pressure, temperature-dew point difference, relative humidity, vapor pressure, absolute humidity, dew point, 10 min average wind speed, visibility, total cloud amount, low-cloud amount, and sunshine duration were well correlated with the foggy weather phenomenon. Three representative principle components were extracted from the 11 elements, namely, the factors of water vapor, radiation and wind speed. The set-up logistic regression forecasting model was better in discriminating foggy from fogless weather than heavy fog from light fog.
Eleven foggy weather-related meteorological elements were selected based on conventional observations from national basic sites in the Gaoyao District, Zhaoqing from 2015 to 2018, a logistic regression equation was constructed from new common factors obtained by principle component analysis and dimension reduction, and optimal critical values were determined to identify foggy weather events following the ROC curve and Youden's index. The result is shown as follows. The meteorological elements of air pressure, temperature-dew point difference, relative humidity, vapor pressure, absolute humidity, dew point, 10 min average wind speed, visibility, total cloud amount, low-cloud amount, and sunshine duration were well correlated with the foggy weather phenomenon. Three representative principle components were extracted from the 11 elements, namely, the factors of water vapor, radiation and wind speed. The set-up logistic regression forecasting model was better in discriminating foggy from fogless weather than heavy fog from light fog.
引文
[1]唐洁,梁慕慧,李厚伟,等.肇庆地区大雾气候特征分析[J].广东气象,2012,34(3):32-35.
[2]吕晶晶,牛生杰,张羽,等.湛江东海岛一次春季海雾的宏微观结构及边界层演变特征[J].气象学报,2014,72(2):350-365.
[3]张弘豪,黄先伦,殷宏南,等.阳江大雾气候特征和地面要素分析[J].广东气象,2016,38(6):42-46.
[4]杨凤群,林苗青,翁武坤.近35年南澳岛大雾气候特征及其影响因素分析[J].广东气象,2011,33(5):33-35.
[5]孙喜艳,张晓东,黄卫东,等.亚锦赛期间汕尾一次海雾过程的分析[J].广东气象,2015,37(5):31-34.
[6]陈国贞,陈锐,唐洁,等.1965—2012年紫金县雾的气候特征分析[J].广东气象,2015,37(6):35-38.
[7]叶树春,梁域,胡淑娟,等.云浮地区大雾天气的成因分析及预报[J].广东气象,2012,34(3):29-31.
[8]罗鹍,陈玲,尹淑娴,等.东莞市典型天气的辐射特征及影响因子分析[J].广东气象,2015,37(3):43-47.
[9]中国气象局.地面观测规范[S].北京:气象出版社,2003:20-23.
[10]陈卫中,潘晓平,倪宗瓒.Logistic回归模型在ROC分析中的应用[J].中国卫生统计,2007(1):22-24.
[11]解坤,张俊芳.基于KMO-Bartlett典型风速选取的PCA-WNN短期风速预测[J].发电设备,2017,31(2):86-91.
[12]孙振宇.多元回归分析与Logistic回归分析的应用研究[D].南京:南京信息工程大学,2008.
[13]毛程燕,龚理卿,陈江锋,等.浙江中西部大雾天气的气候特征及潜势预报方法[J].广东气象,2018,40(2):10-13.
[14]庞古乾,伍志方,郭春迓,等.广东省前汛期分区强对流潜势预报方法研究[J].热带气象学报,2016,32(2):265-272.
[15]Zhu Y H.Study on probability estimation of haze in Beijing based logistic regression model[M].America:America Proceedings of Workshop,2017.
[16]冯齐友.成都城市雾霾成因及其治理对策研究[D].成都:成都理工大学,2017.
[2]吕晶晶,牛生杰,张羽,等.湛江东海岛一次春季海雾的宏微观结构及边界层演变特征[J].气象学报,2014,72(2):350-365.
[3]张弘豪,黄先伦,殷宏南,等.阳江大雾气候特征和地面要素分析[J].广东气象,2016,38(6):42-46.
[4]杨凤群,林苗青,翁武坤.近35年南澳岛大雾气候特征及其影响因素分析[J].广东气象,2011,33(5):33-35.
[5]孙喜艳,张晓东,黄卫东,等.亚锦赛期间汕尾一次海雾过程的分析[J].广东气象,2015,37(5):31-34.
[6]陈国贞,陈锐,唐洁,等.1965—2012年紫金县雾的气候特征分析[J].广东气象,2015,37(6):35-38.
[7]叶树春,梁域,胡淑娟,等.云浮地区大雾天气的成因分析及预报[J].广东气象,2012,34(3):29-31.
[8]罗鹍,陈玲,尹淑娴,等.东莞市典型天气的辐射特征及影响因子分析[J].广东气象,2015,37(3):43-47.
[9]中国气象局.地面观测规范[S].北京:气象出版社,2003:20-23.
[10]陈卫中,潘晓平,倪宗瓒.Logistic回归模型在ROC分析中的应用[J].中国卫生统计,2007(1):22-24.
[11]解坤,张俊芳.基于KMO-Bartlett典型风速选取的PCA-WNN短期风速预测[J].发电设备,2017,31(2):86-91.
[12]孙振宇.多元回归分析与Logistic回归分析的应用研究[D].南京:南京信息工程大学,2008.
[13]毛程燕,龚理卿,陈江锋,等.浙江中西部大雾天气的气候特征及潜势预报方法[J].广东气象,2018,40(2):10-13.
[14]庞古乾,伍志方,郭春迓,等.广东省前汛期分区强对流潜势预报方法研究[J].热带气象学报,2016,32(2):265-272.
[15]Zhu Y H.Study on probability estimation of haze in Beijing based logistic regression model[M].America:America Proceedings of Workshop,2017.
[16]冯齐友.成都城市雾霾成因及其治理对策研究[D].成都:成都理工大学,2017.