亚热带人工针叶林冠层叶氮浓度的星载高光谱遥感能力评估(英文)

An Evaluation of Spaceborne Imaging Spectrometry for Estimation of Forest Canopy Nitrogen Concentration in a Subtropical Conifer Plantation of Southern China

冠层叶片氮浓度(CNC)是影响森林生态系统生产力的重要参数之一。本研究探讨了星载成像光谱遥感在估测亚热带红壤丘陵区人工针叶林CNC的表现。分析包括了星载成像光谱数据(Hyperion影像)覆盖的两条样带上的57个野外样方,并将其分为三个子集(A-C)。利用一元回归和偏最小二乘回归方法分析了CNC与成像光谱信息之间的关系。在A-C子集中,CNC与近红外反射率(NIR)之间的相关性一致都呈现显著的正相关关系(R2=0.29,0.33和0.36,P<0.05或P<0.01)。另外,我们利用归一化的氮指数(NDNI)估计森林CNC的变异。在3个子集中,NDNI与CNC都呈现极显著的正相关关系,但相关性不高(R2=0.38,0.20和0.17,P<0.01)。然后利用偏最小二乘方法分析了CNC与整个成像光谱数据(反射率、对数变换和一阶导数变换)之间的相关性,对于各个子集相关性不同且相对微弱。在分析已有数据和对比前人文献基础上,文章分析了影响成像光谱遥感森林CNC的可能原因,并指出研究区人工针叶林单一的冠层结构可能减弱了该地区森林CNC与成像光谱信息之间的相关关系。

Canopy foliar Nitrogen Concentration （CNC） is one of the most important parameters influencing vegetation productivity in forest ecosystems. In this study, we explored the potential of imaging spectrometry （hyperspectral） remote sensing of CNC in conifer plantations in China’s subtropical red soil hilly region. Our analysis included data from 57 field plots scattered across two transects covered by Hyperion images. Single regression and partial least squares regression （PLSR） were used to explore the relationships between CNC and hyperspectral data. The correlations between CNC and nearinfrared relfectance （NIR） were consistent in three data subsets （subsets A-C）. For all subsets, CNC was signiifcantly positively correlated with NIR in the two transects （R2=0.29, 0.33 and 0.36, P＆lt;0.05 or P＆lt;0.01, respectively）. It suggested that the NIR-CNC relationship exist despite a weak one, and the relationship may be weakened by the single canopy structure. Besides, we also applied a shortwave infrared （SWIR） index - Normalized Difference Nitrogen Index （NDNI） to estimate CNC variation. NDNI presented a signiifcant positive correlation with CNC in different subsets, but like NIR, it was also with low coefifcient of determination （R2=0.38, 0.20 and 0.17, P＆lt;0.01, respectively）. Also, the correlations between CNC and the entire spectrum reflectance （or its derivative and logarithmic transformation） by PLSR owned different signiifcance in various subsets. We did not ifnd the very robust relationship like previous literatures, so the data we used were checked again. The paired T-test was applied to estimate the inlfuence of inter-annual variability of FNC on the relationships between CNC and Hyperion data. The inter-annual mismatch between period of ifeldwork and Hyperion acquisition had no inlfuence on the correlations of CNC-Hyperion data. Meanwhile, we pointed out that the lack of the canopy structure variation in conifer plantation area may lead to these weak relationships.