欢迎光临,现在是     设为首页   加入收藏   联系我们
文章摘要
陈有龙,宁雨珂,唐荣年,谢小峰.基于时空独立的随机森林模型对海南热带气温数值预报的订正研究[J].海南大学学报编辑部:自然科学版,2020,38(4):.
基于时空独立的随机森林模型对海南热带气温数值预报的订正研究
Research on tropical temperature correction for numerical forecast in Hainan based on spatiotemporal independence random forest model
投稿时间:2020-07-01  修订日期:2020-07-06
DOI:
中文关键词: 气温 ;数值预报 ;随机森林;海南热带气温
英文关键词: temperature; numerical forecast; random forest; Tropical temperature in Hainan
基金项目:海南省南海气象防灾减灾重点实验室开放基金
作者单位E-mail
陈有龙 海南省气象台 738423138@qq.com 
宁雨珂 海南大学  
唐荣年 海南大学  
谢小峰 海南大学 xfxie@hainanu.edu.cn 
摘要点击次数: 69
全文下载次数: 0
中文摘要:
      采用机器学习方法对气温集合预报进行订正是气象领域研究的热点问题。面向海南省所特有的海岛以及热带特点,并结合海南岛独特的地理地貌,本文设计了基于时空独立的随机森林模型,利用站点的实测数据以及欧洲中期天气预报中心(ECMWF)的模式数据,实现对每个站点未来7天预报时效为3小时的气温精准订正。采用<2°C准确率、<1°C准确率及均方根误差等指标,对ECMWF模式预报温度和本文模型订正气温效果进行评估,结果表明本文所提的订正模型结果显著优于ECMWF模式结果,更接近真实温度值,对ECMWF进行了较好的订正。
英文摘要:
      Revising the temperature forecast with machine learning is critical for meteorological research. Considered the island and tropical characteristics and the unique geographical features of Hainan Island, a random forest model based on spatiotemporal independence was designed in this paper. In this study, we use the forecast data from ECMWF and observed data to achieve the accurate correction for the each station of 7 days forecast, which has 3 hours precision. The proposed method was evaluated by the index of <2°C accuracy, <1°C accuracy and RMSE. The result shows that the proposed method is obviously better than model of ECMWF. It’s more close to the observed temperature value, and the ECMWF result is well revised.
关闭