利用光谱特征参数估算病害胁迫下杉木叶绿素含量

摘要

This study was to explore and build the hyper-spectral estimation models of the chlorophyll content of Chinese fir canopy under anthracnose stress, and promote application of the hyper-spectral remote sensing technology to forest pests and disease monitoring. Through aquiring spectral data and the corresponding chlorophyll contents of Chinese fir of different severity levels, the correlation among the pigment contents, the first derivative data and the canopy hyper-spectral data were analyzed, and the estimation models were built by stepwise multiple regression (SMLR) , principal component regression (PCR) and partial least squares regression (PLSR). The high correlation existed between chlorophyll content and original optical spectrum in visible regions (614 — 698 nm) and near infrared regions O724 nm), and tended to be stable in near infrared region. There was also high correlation between chlorophyll content and the first derivative data in the ranges of 424 — 486 nm, 514 — 532 nm, 552—682 nm, 698—755 nm and 762—772 nm. The influence of multicollinearity among parameters was eliminated by the three building model methods, and the three coefficients of determination all passed the 0.01 significance level. But SMLR was the best one, the relative error and RMS error of which was 10. 71% and 0.194, respectively. This study showed significant difference in spectral reflectance under of disease stress of Chinese fir anthracnose of different degrees, and the hype-spectral information could be used to quantificationally estimate chlorophyll contents of Chinese fir canopy under the disease stress with high precision.%为了探索建立炭疽病胁迫下杉木叶绿素含量的高光谱估算模型,促进遥感技术在森林病虫害监测中的应用,通过获取不同发病程度的杉木冠层光谱及相应的叶绿素含量,将冠层光谱数据、一阶微分数据与相应的叶绿素含量分别进行了相关分析.采用逐步回归、主成分回归及偏最小二乘回归方法构建叶绿素含量的估算模型.叶绿素含量与原始光谱在可见光(614～698 nm)和近红外区(724 nm之后)达到极显著相关,且在近红外区基本趋于稳定；与一阶微分光谱在424～486 nm、514～532 nm、552～682 nm、698～755 nm和762～772 nm波段全部达到极显著相关；3种建模方法均消除了参数间多重共线性的影响,模型的决定系数全部达到极显著水平,其中逐步回归模型精度最高,相对误差和均方根误差分别为10.71％和0.194.研究表明受到不同程度炭疽病胁迫的杉木冠层光谱反射差异较大,可利用高光谱信息定量估算病害胁迫下的杉木叶绿素含量,且估算精度较高.