黄土高原土壤铁元素含量遥感反演方法

利用光谱技术可快速而高效地获取土壤组分及其空间分布信息,为准确估算黄土高原土壤Fe元素含量与分布特征,通过对榆林东部地区的典型黄土进行野外采集、室内理化分析、光谱测定及其预处理,分析土壤铁含量与反射光谱的相关性,筛选敏感波段,利用偏最小二乘法建模,确定最佳估算模型。光谱反射率及其不同变换后筛选出的敏感波段,主要分布在500、870、1 700、2 200 nm附近;原始反射率(Ref)建模结果相对较稳定且预测效果最佳,预测集相关系数R^2达0.73,均方根误差RMSEP最小;经导数变换(FDR和SDR)及连续统去除CR变换的模型次之,预测集R^2分别达到0.61和0.64;采用波段内插法将土壤Fe含量最优估算模型Ref应用Sentinel-2多光谱遥感影像上,得到土壤Fe含量遥感反演图,结果表明研究区土壤Fe含量的分布特征与地层密切相关;该研究结果可为土壤Fe元素含量遥感分析提供支撑,实现黄土高原土壤铁的光谱快速制图。

Remote Sensing Inversion Method of Soil Iron Content in the Loess Plateau

The information of soil composition and its spatial distribution could be obtained quickly and efficiently by using spectral technology.In order to accurately estimate the content and distribution characteristics of soil Fe elements in the loess plateau,the typical loess in the eastern part of Yulin was collected in the field.Laboratory physical and chemical analysis,spectral determination and pretreatment,analysis of the correlation between soil iron content and reflection spectrum,screening sensitive bands,using partial least squares modeling to determine the best estimation model.The spectral reflectivity and the selected sensitive bands are mainly distributed at 500 nm,870 nm,1 700 nm and 2 200 nm.The original reflectivity(Ref)modeling results are relatively stable and the prediction effect is the best.The prediction set correlation coefficient R^2 is up to 0.73 and the Root Mean Square Error(RMSEP)is the smallest.After derivative transform(FDR and SDR)and continuum removal of CR transform,the prediction set R^2 is 0.61 and 0.64,respectively.The optimal estimation model of soil Fe content(Ref)was applied to the Sentinel-2 multi-spectral remote sensing image to obtain the remote sensing inversion map of soil Fe content by band interpolation.It was found that the distribution characteristics of soil Fe content in the studied area were closely related to the strata.The results of this study can provide support for remote sensing analysis of soil Fe element content and realize rapid spectral mapping of soil iron in the Loess Plateau.