基于支持向量机的星载GNSS-R土壤湿度反演方法

Spaceborne GNSS-R soil moisture inversion method based on support vector machine

土壤湿度对水循环、陆气能量交换具有重要影响,获取高时空分辨率的土壤湿度对科学研究和生态环境评价具有重要意义.文章提出了一种基于支持向量机的星载GNSS-R土壤湿度反演方法.首先,基于CYGNSS数据提取地表反射率和信号入射角参数,并提取SMAP数据中的植被光学厚度和地表粗糙度作为辅助参数;然后,利用支持向量机初步构建土壤湿度反演模型,并通过更换核函数和灵活使用网格搜索、随机搜索的方法优化获得星载GNSS-R土壤湿度反演模型;最后,以SMAP数据中的土壤湿度为参考值评估CYGNSS土壤湿度反演模型.结果表明文章提出的方法反演土壤湿度的均方根误差和相关系数分别为0.077 cm^(3)/cm^(3)和0.84.该研究表明星载GNSS-R土壤湿度反演结果与SMAP土壤湿度具有良好的一致性,为高时空分辨率的土壤湿度反演提供了一种有效的新思路.

Soil moisture has an important effect on water cycle and land air energy exchange.It is important to obtain soil moisture with high spatial and temporal resolution for scientific research and ecological environment evaluation.In this paper,a spaceborne GNSS-R soil moisture inversion method based on support vector machine is proposed.First,the parameters of surface reflectance and signal incident angle are obtained based on CYGNSS data,and the vegetation optical thickness and surface roughness in SMAP data are extracted as auxiliary parameters.Then,the soil moisture inversion model is initially constructed by using the support vector machine,and the spaceborne GNSS-R soil moisture inversion model is optimized by replacing the kernel function and using grid search and random search methods flexibly.Finally,the CYGNSS soil moisture inversion model was evaluated using the soil moisture in the SMAP data as a reference value.The results show that the root mean square error and correlation coefficient of the method proposed in this paper are 0.077 cm^(3)/cm^(3)and 0.84,respectively.This study shows that the inversion results of spaceborne GNSS-R soil moisture are in good agreement with SMAP soil moisture,which provides an effective new idea for soil moisture inversion with high spatial and temporal resolution.