文献类型：期刊文献

中文题名：陕西黄河湿地土壤碳氮含量高光谱估算反演

英文题名：Hyperspectral Estimation and Inversion of Soil Carbon and Nitrogen Contents in the Yellow River Wetland in Shaanxi Province

作者：徐干君[1] 聂磊超[2] 马浩[1] 唐希颖[2] 翟夏杰[2] 赵欣胜[2] 李伟[2]

第一作者：徐干君

机构：[1]国家林业和草原局西北调查规划设计院,陕西西安710048;[2]中国林业科学研究院湿地研究所,北京100091

年份：2022

卷号：18

期号：4

起止页码：10-14

中文期刊名：湿地科学与管理

外文期刊名：Wetland Science & Management

收录：CSTPCD;;国家哲学社会科学学术期刊数据库

基金：黄河流域湿地碳汇核算方法与固碳功能评估研究。

语种：中文

中文关键词：湿地;有机碳;全氮;高光谱;SVM

外文关键词：Wetland;SOC;TN;Hyperspectral;SVM

分类号：S153.621

摘要：为实现湿地土壤有机碳(SOC)和全氮(TN)快速、准确的测定,以陕西黄河湿地自然保护区湿地土壤为研究对象,在室内采集并研究有机碳和全氮与高光谱350~2500 nm波段的定量反演关系,并结合随机森林(RF)模型、反向传播神经网络(BPNN)、偏最小二乘(PLSR)和支持向量机(SVM)建立光谱与有机碳和全氮间的定量反演模型。结果表明:基于原始光谱建立的土壤有机碳反演模型精度更高,经一阶微分变换后的光谱反射率建立的全氮含量反演模型精度更高;采用SVM模型建立的土壤有机碳和土壤全氮模型预测效果最好,基于一阶光谱和原始光谱反射率建立预测模型的决定系数R2分别为0.72、0.64和0.84、0.81,预测均方根误差(rmse)分别为2.91、3.61和0.01、0.01。采用高光谱技术对湿地土壤有机碳和全氮含量进行预测是可行的。
In order to realize the rapid and accurate determination of wetland soil organic matter and total nitrogen,the quantitative inversion relationship between wetland soil(Wetland soil of the Yellow River wetland nature reserve)organic matter or total nitrogen and hyperspectral 350~2500 nm band was studied under indoor conditions.After the spectrum is smoothed and filtered by savitzky Golay,the quantitative inversion model between soil hyperspectral and organic matter and total nitrogen is established by using back propagation neural network(BPNN),partial least squares regression(PLSR),random forest(RF)and(support vector machines,SVM)models.The results show that the soil organic carbon inversion model based on the original spectrum has high precision.The total nitrogen content inversion model established by the first order of differential(FDR)transformation has high precision.the prediction effect of soil organic carbon and soil nitrogen model established by SVM model is the best.The determination coefficients R2 of the prediction model are 0.72,0.64,0.84 and 0.81 respectively,and the root mean square error(RMSE)of prediction are 2.91 and 3.61 and 0.01 and 0.01 respectively.It is feasible to use hyperspectral technology to predict the contents of soil organic matter and total nitrogen in wetland.