任萍,陈明轩. 2020. 基于机器学习的复杂地形下短期数值天气预报误差分析与订正?[J]. 气象学报, (0):-, doi:10.11676/qxxb2020.060
基于机器学习的复杂地形下短期数值天气预报误差分析与订正?
Error analysis and correction of short-term numerical weather prediction under complex terrain based on machine learning
投稿时间:2020-01-18  修订日期:2020-05-19
DOI:10.11676/qxxb2020.060
中文关键词:  集成,数值预报,机器学习,XGBoost,线性回归,等权重
英文关键词:Integration, Numerical  prediction, Machine  learning, XGBoost, Linear  regression, Equal weight
基金项目:国家自然科学基金
作者单位E-mail
任萍 中国海洋大学 Pren@ium.cn 
陈明轩 北京城市气象研究院 mxchen@ium.cn 
摘要点击次数: 94
全文下载次数: 92
中文摘要:
      本文初步研发了一套基于机器学习方法XGBoost且考虑地形特征影响的数值预报多模式集成技术,并与传统的等权重平均和线性回归方法的集成效果进行了对比分析。利用北京地区快速更新循环数值预报系统每天8次循环预报给出的近地面2-m温度、2-m相对湿度、10-m风速、10-m风向数据产品,分别基于机器学习方法XGBoost、等权重平均方法、线性回归方法构建了三种体现地形因子影响的多模式预报时间滞后集成模型。试验对比分析了暖季、冷季每日不同时刻的模式预报集成订正效果。研究结果表明:分季节试验中,基于XGBoost的模型对2-m温度、10-m全风速的集成预报结果相对与原始最优预报结果误差,明显优于其他两种传统方法。XGBoost对2-m温度集成的误差可降低11.02—18.09%,对10-m全风速集成误差可降低31.23—33.22%,对10-m风向集成误差可降低4.1—8.23%。而2-m相对湿度的集成预报误差与传统方法接近。整体来看,研发的基于XGBoost的多模式集成预报模型可以充分“挖掘”不同模式或不同时刻快速更新循环预报的优点,有效减少模式的系统性误差,并提供准确性更高的多模式集成确定性预报产品。
英文摘要:
      In this paper, a set of multi-mode integration technology of numerical prediction based on machine learning method XGBoost and considering the influence of topographical features was preliminarily developed. And its integration effect was compared with that of traditional equal weight average and linear regression method. Based on the data products of rapid update cycle numerical prediction system in Beijing, which can provide the cycle prediction including 2-m temperature, 2-m relative humidity, 10-m wind speed and 10-m wind direction near the ground 8 times a day, then three integrated models of multi-model forecast time lag integrated models were construct based on the machine learning method XGBoost, equal weight average method and linear regression method, respectively. The experiment compared and analyzed the effect of the integrated correction of model predictions at different times in the warm and cold seasons every day. The research results indicated that in the seasonal test, the integrated prediction results of the 2-m temperature and 10-m full wind speed based on the XGBoost model are significantly better than the original optimal prediction results, which is significantly better than the other two traditional methods. The error of XGBoost for 2-m temperature integration can be reduced by 11.02-18.09%, the error for 10-m full wind speed integration can be reduced by 31.23-33.22%, and the error for 10-m wind direction integration can be reduced by 4.1-8.23%. The integrated forecast error of 2-m relative humidity is close to the traditional method. As a whole, the developed multi-mode integrated prediction model based on XGBoost can fully "excavate" the advantages of different modes or the rapid updating cycle prediction at different times, therefore effectively reducing the systematic error of the mode, and providing a multi-mode integrated deterministic prediction product with higher accuracy.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮