浙江中南部海雾预报决策树模型研究  

A decision-tree forecasting model of sea fog for the central and southern coastal area of Zhejiang Province

作　　者：俞涵婷 廖晨昕 王可欣 陈华忠 YU Han-ting;LIAO Chen-xin;WANG Ke-xin;CHEN Hua-zhong(Jiaojiang Meteorological Bureau,Taizhou 318000 China;Yuhang Meteorological Bureau,Hangzhou 311100 China;Yuhuan Meteorological Bureau,Yuhuan 317600 China)

机构地区：[1]椒江区气象局,浙江台州318000 [2]余杭区气象局,浙江杭州311100 [3]玉环市气象局,浙江玉环317600

出　　处：《海洋预报》2020年第6期96-101,共6页Marine Forecasts

基　　金：台州市椒江区科技计划项目(192032);台州市气象科技计划项目(TZ2018QN02)。

摘　　要：利用椒江大陈沿海航线上重要站点一江山岛、大陈站和头门岛2015—2018年2—6月的海雾历史观测资料和NCEP/NCAR FNL再分析资料,从海雾的成因中找出大气与海雾的关系。分析的影响因子包括:地面与高空温差(T1000 hPa—T2 m)、(T925 hPa—T2 m)、(T850 hPa—T2 m)、(T975 hPa—T2 m)和1000 hPa相对湿度,低层上升速度分析等。结论如下:(1)暖湿气流本身强弱对大雾无影响,温差才是形成大雾的重要因素,近地层的逆温有利于大雾形成,越低层逆温越强越有利于大雾形成;(2)大雾形成时所需相对湿度基本集中在90以上,950 hPa上较弱的上升速度利于大雾的形成,散度条件对海雾的影响差别不大;(3)通过训练集数据参与模型的建立,模型整体的学习准确率为0.85。将此测试集数据运用于2019年2—6月的大雾数据检验中,成功率为0.8。决策树模型建立的海雾判别流程可在业务中用于浙中南有无海雾的判别。Based on the conventional observational data in Yijiangshan Island,Dachen Island and Toumen Island between February and June from 2015 to 2018 and the NCEP 1°×1°reanalysis data,the relationship between marine atmospheric conditions and sea fog is revealed in this paper.The influence factors analyzed include the temperature difference between 1000 hPa,975 hPa,925 hPa,T850 hPa and the surface layer,relative humidity at 1000 hPa and the ascending speed of the low level air.The results show that the strength of warm and moist advection itself has no effect on the fog and the temperature difference is an important factor in the formation of sea fog.The inversion of the near-surface temperature is conducive to sea fog.The stronger the inversion temperature is,the more favorable it is for the formation of sea fog.The relative humidity required for the formation of sea fog is basically above 90.The weaker ascending air speed at 950 hPa is more conducive to the formation of sea fog,while the divergence conditions have no significant influence to sea fog.A decision-tree model is established by dataset training with an overall learning accuracy rate of 0.85.The model is validated using the sea fog observation from February to June 2019,which reveals a success rate of 0.8.Therefore,the decision-tree model can be operationally applied to identify the sea fog events in the central and southern coastal area of Zhejiang province.

关 键 词：浙江 海雾 分类与回归树 决策树模型 

分 类 号：P732.4[天文地球—海洋科学]