摘要
健康落叶松与遭受病害落叶松的冠层光谱反射率曲线具有明显差异,利用反映这种差异的光谱特征参数建立回归模型,可为反演落叶松冠层光合色素含量进而诊断落叶松健康状况提供方法和途径。以吉林省延边州敦化、和龙两市林场中健康的和遭受落叶松早落病侵害的人工落叶松林为调查对象,在对野外采集的14个落叶松冠层样本进行光谱测量及光合色素含量测量的基础上,选取8个对落叶松冠层光合色素含量变化反映敏感的光谱参数参与建立其光合色素含量的一元线性回归和多元逐步回归模型。研究结果表明,不同健康程度的落叶松冠层光谱曲线在其可见光及近红外波段有3个比较明显的特征差异处,分别位于光谱曲线的"绿峰"、"红谷"和"红边"位置。利用反映这些差异的8个光谱特征参数建立落叶松冠层光合色素含量的回归模型,除"红边"这一参数回归效果不令人满意外,其余7个参数均得到了较好的回归效果,其中利用峰谷波长差Dgr建立的关于总叶绿素和叶绿素b含量的一元回归模型R2值分别达到0.8428和0.7498,利用NDGI建立的关于叶绿素a和类胡萝卜素含量的一元回归模型R2值分别达到0.8758和0.7897;多元逐步回归模型的回归效果与一元回归模型相比,各判定系数R2值均有所提高,总叶绿素、叶绿素a、b和类胡萝卜素含量的回归模型R2值分别达到0.885、0.910、0.839和0.862。
The difference between the spectral reflectance curves of healthy and diseased larch canopies is obvious. In this work, we chose the larch plantations as the study object. Firstly, we selected 14 larch canopy samples and measured their spectral reflectance and photosynthetic pigment contents. Then we selected 8 spectral parameters to set up a series of linear regression models. The results show that 3 remarkable different regions occur in the visible and near infrared bands. The correlation between the spectral parameters and the photosynthetic pigment contents are good except the parameter REP. The R^2 values of the single-variable linear regression models for estimation of the contents of total chlorophyll,chlorophyll a, chlorophyll b and carotenoid are 0. 8428,0. 8758,0. 7498 and 0. 7897 respectively. The multivariable regression models were also set up and their performance was better than the single variable regression models. The R^2 values of the multivariable regression models for estimation of the contents of total chlorophyll, chlorophyll a,chlorophyll b and carotenoid are 0. 885, 0. 910, 0. 839 and 0. 862 respectively. This work provides an approach for the retrieval of the forest canopy photosynthetic pigment contents by using hyperspectral data.
出处
《遥感技术与应用》
CSCD
2008年第3期264-271,345,I0002,共10页
Remote Sensing Technology and Application
基金
中国科学院遥感应用研究所遥感科学国家重点实验室开放研究基金
关键词
高光谱
落叶松冠层
光合色素含量
健康
回归模型
Hyperspectrum
Larch canopy
Photosynthetic pigment content
Health
Regression models