海面风速和有效波高(significant wave height,SWH)是海洋环境中的关键参数,两者之间关系密切.全球导航卫星系统反射测量(global navigation satellite system reflectometry,GNSS-R)可有效反演海面风速和SWH,然而已有研究局限于单一参数的反演.为此,提出一种基于机器学习算法的海面风速和SWH联合反演方法.首先通过质量控制获取有效的气旋全球导航卫星系统(cyclone global navigation satellite system,CYGNSS)观测数据,进而分别采用随机森林、极端梯度提升、轻量梯度提升机、决策树和自适应增强算法建立联合反演模型,并对比分析其反演性能.经实验表明:极端梯度提升更适用于海面风速和SWH的联合反演,均方根误差分别为 0.91 m/s和0.20 m,皮尔逊相关系数分别达到 0.90 和 0.96.相对于传统的单一参数反演,本文方法能够实现对海面风速和SWH高效又准确的反演.
Joint inversion of sea surface wind speed and significant wave height based on machine learning
Sea surface wind speed and significant wave height(SWH)are key parameters in the marine environment,with a closely interrelated relationship.The global navigation satellite system reflectometry(GNSS-R)can effectively invert sea surface wind speed and SWH,yet existing studies have been limited to the inversion of single parameter.Thus,this paper presents a method for the joint inversion of sea surface wind speed and SWH based on machine learning algorithms.Initially,valid observational data from the cyclone global navigation satellite system(CYGNSS)were acquired through quality control.Subsequently,joint inversion models were constructed using random forest,extreme gradient boosting,light gradient boosting machine,decision tree,and adaptive boosting algorithms,and their inversion performances were comparatively analyzed.It is experimentally shown that the extreme gradient boosting is more suitable for the joint inversion of sea surface wind speed and SWH,and the root mean square errors are 0.91 m/s and 0.20 m,and the correlation coefficients reach 0.90 and 0.96,respectively.Compared with the traditional single parameter inversion,the method in this paper can realize efficient and accurate inversion of sea surface wind speed and SWH.
国家科技期刊平台
NSTL
万方数据
