应用无人机遥感数据估算沙柳灌丛地上生物量的方法  

Methods for Estimating the Aboveground Biomass of Salix psammophila Shrublands Using UAV Remote Sensing Data

作　　者：孙小添 王海超 韩青池 孙凯 李颖 王继璇 裴志永[1] Sun Xiaotian;Wang Haichao;Han Qingchi;Sun Kai;Li Ying;Wang Jixuan;Pei Zhiyong(Inner Mongolia Agricultural University,Hohhot 010018,P.R.China;Northeast Forestry University)

机构地区：[1]内蒙古农业大学,呼和浩特010018 [2]东北林业大学

出　　处：《东北林业大学学报》2025年第5期74-81,共8页Journal of Northeast Forestry University

基　　金：内蒙古自治区自然科学基金项目(2023QN03029);国家自然科学基金地区科学基金项目(32301665);高校基本科研业务费项目(BR230122)。

摘　　要：沙柳(Salix psammophila)作为一种重要的固碳和生态恢复植物,准确估算其地上生物量(AGB)对评估沙柳在改善生态环境、防治沙漠化和碳中和等方面具有重要意义。本研究以鄂尔多斯市造林总场沙柳国家林木种质资源库的沙柳灌丛为研究对象,采用一阶SG平滑法(FDSG)与标准正态变量变换法(SNV)分别对原始光谱数据进行预处理,结合样方实测沙柳灌丛地上生物量构建高光谱的植被指数,根据最小二乘回归(PLSR)、支持向量机(SVM)、随机森林回归(RFR)构建沙柳灌丛地上生物量估测模型。结果表明:基于标准正态变量变换(SNV)高光谱的归一化植被指数和比值植被指数组合作为输入参数构建的RFR模型对沙柳灌丛地上生物量反演的效果最好,模型的决定系数(R^(2))为0.6356,均方根误差为2.1264,与PLSR和SVM模型中反演精度最高一组模型的决定系数(R^(2))分别相差0.1817和0.3513,均方根误差分别相差0.3769和0.8538。Salix psammophila,as an important plant for carbon sequestration and ecological restoration,accurately estimating its aboveground biomass(AGB)is of great significance for evaluating the role of S.psammophila in improving the ecological environment,preventing desertification,and achieving carbon neutrality.In this study,the S.psammophila shrubbery in the National Forest Germplasm Resources Bank of S.psammophila at the Afforestation Headquarters in Ordos City was taken as the research object.The first-order Savitzky-Golay smoothing method(FDSG)and the standard normal variate transformation method(SNV)were used to preprocess the original spectral data respectively.Combined with the measured aboveground biomass of S.psammophila shrubbery in the sample plots,hyperspectral vegetation indices were constructed.Based on partial least squares regression(PLSR),support vector machine(SVM),and random forest regression(RFR),estimation models for the aboveground biomass of S.psammophila shrublands were established.The results showed that the RFR model constructed with the combination of the normalized vegetation index and the ratio vegetation index based on the hyperspectral data after standard normal variate transformation(SNV)as input parameters has the best performance in retrieving the aboveground biomass of S.psammophila shrublands.The determination coefficient(R^(2))of the model is 0.6356,and the root mean square error is 2.1264.Compared with the models with the highest retrieval accuracy in the PLSR and SVM models,the differences in the determination coefficient(R^(2))are 0.1817 and 0.3513 respectively,and the differences in the root mean square error are 0.3769 and 0.8538 respectively.

关 键 词：沙柳 植被指数 无人机遥感 地上生物量 

分 类 号：S771.8[农业科学—森林工程]