基于高光谱的云南胶园土壤全氮含量反演  

Inversion of Soil TotalNitrogen Content inYunnan Rubber Plantation Based onHyperspectral

作　　者：刘忠妹 黎小清[1] 陈桂良[1] 丁华平[1] 许木果 杨春霞[1] LIU Zhongmei;LI Xiaoqing;CHEN Guiliang;DING Huaping;XU Muguo;YANG Chunxia(Yunnan Institute of Tropical Crops,Jinghong,Yunnan 666100,China)

机构地区：[1]云南省热带作物科学研究所,云南景洪666100

出　　处：《热带农业科技》2023年第1期35-41,共7页Tropical Agricultural Science & Technology

基　　金：云南省应用基础研究计划(2019FD006);省所热带作物科技创新专项资金(RF2022-6)。

摘　　要：以在云南西双版纳橡胶园采集的225个土壤样本为研究对象,进行风干土壤样品光谱数据采集和全氮含量测定,对土壤反射率(R)进行重采样(5 nm)后进行倒数的对数(log(1/R))、多元散射校正(MSC)、标准正态交换(SNV)的光谱变换处理和SG平滑或导数变换模式优选获取最佳光谱数据,采用竞争性自适应重加权算法(CARS)、连续投影算法(SPA)和CARS-SPA算法对土壤全氮光谱特征波段进行选择,并分别应用偏最小二乘法(PLS)和支持向量机(SVM)构建胶园土壤全氮估算模型。结果表明,研究区内土壤全氮含量与光谱反射率呈负相关,基于光谱变换和预处理,log(1/R)结合SG导数变换是最佳的土壤全氮反演光谱变换模式,其中,导数阶数为1,SG滤波窗口为43,多项式次数为9。综合比较各变量筛选与建模组合发现,用CARS方法从全波段中筛选出的特征波长建立的SVM模型(R^(2)=0.816,RMSE=0.201,RPD=2.266)效果最好,用于估测研究区土壤全氮含量具有良好的可行性,可为开展云南胶园土壤全氮高效测试和实施橡胶树精准施肥提供技术支持。Taking 225 soil samples collected from Xishuangbanna rubber plantation in Yunnan as the research object,the spectral data collection and total nitrogen content determination of air-dried soil samples were carried out.The soil reflectance(R)was resampled(5 nm),and the optimal spectral data were obtained by the reciprocal logarithm(log(1/R)),multiple scattering correction(MSC),standard normal variate(SNV)and SG smoothing or derivative transformation mode.The competitive adaptive reweighted sampling(CARS),successive projections algorithm(SPA)and CARS-SPA were used to select the spectral characteristic bands of soil total nitrogen,and partial least squares(PLS)and support vector machine(SVM)were used to construct the estimation model of soil total nitrogen in rubber plantation.The results showed that the soil total nitrogen content in the study area was negatively correlated with the spectral reflectance.Based on spectral transformation and pretreatment,log(1/R)combined with SG derivative transformation was the best spectral transformation model for soil total nitrogen inversion,where the derivative order was 1,the SG filter window was 43,the polynomial degree was 9.By comprehensively comparing the variable selection and modeling combination,it was found that the SVM model(R^(2)=0.816,RMSE=0.201,RPD=2.266)established by the characteristic wavelengths selected by the CARS method from the full band had the best effect,which was used to estimate the soil total nitrogen content in the study area.It has good feasibility and can provide technical support for the efficient test of soil total nitrogen in Yunnan rubber plantations and the implementation of precise fertilization of Yunnan rubber trees.

关 键 词：橡胶园 土壤全氮 高光谱 偏最小二乘 支持向量机 

分 类 号：S794.1[农业科学—林木遗传育种]