论文部分内容阅读
叶片作为植物冠层的基本组成元素,其自身的光学特性直接影响着遥感所能获得的植物冠层反射光谱。从原理上讲,叶片的光学特性不仅取决于其内部生化组分含量的多少,还与其物理结构密切相关。因此对叶片内部物理结构进行估算有助于分离其对叶片光谱的影响,从而提高叶片生化信息反演的精度。在基于叶片内部辐射传输过程的PROSPECT模型中,叶片内部结构用一个假想的叶肉结构参数N来描述。PROSPECT模型模拟光谱发现,N对叶片反射率和透过率均影响显著,且影响范围涵盖400—2500nm的全部波段。本文利用水稻叶片实测光谱和生化数据尝试了3种N的估算方法,包括两种经验方法和一种模型反演方法,并对其进行比较。结果表明,由于两种经验方法都基于N和表观叶面积(SLA)之间的非线性经验公式,因此两者具有内在的数学关系。运用模型反演方法估算的N可在实测水稻光谱和模型模拟光谱间得到最小RMSE,且其在数值上小于两种经验方法的估算值。以N为因变量,叶片光谱反射率为自变量,运用逐步线性回归分析建立了N的光谱估算模型,550nm,816nm,1210nm和1722nm四个波段被选入模型,回归效果较好,为N的估算提供了一种新的经验方法。
As the basic component of plant canopy, its own optical properties directly affect the plant canopy reflectance spectrum obtained by remote sensing. In principle, the optical properties of leaves depend not only on the amount of biochemical content of its internal, but also closely related to its physical structure. Therefore, estimating the internal physical structure of the leaf helps to separate its influence on the leaf spectrum, thereby improving the accuracy of leaf biochemical information inversion. In the PROSPECT model based on the radiative transfer process inside the blade, the internal structure of the blade is described by an imaginary mesophyll structure parameter N. PROSPECT model simulated spectra found that N has a significant impact on leaf reflectivity and transmittance, and the impact of covering all the 400-2500nm band. In this paper, three kinds of estimation methods of N are tried using the measured spectra and biochemical data of rice leaves, including two empirical methods and a model inversion method, and compared. The results show that both of them have an inherent mathematical relationship because both empirical approaches are based on a nonlinear empirical formula between N and apparent leaf area (SLA). The estimated N using the model inversion method yields a minimum RMSE between the measured rice spectra and the model simulated spectra, and is numerically less than estimated by two empirical methods. Using N as the dependent variable and spectral reflectance of leaf as independent variable, spectral estimation model of N was established by stepwise linear regression analysis. The four bands of 550 nm, 816 nm, 1210 nm and 1722 nm were selected into the model and the regression effect was good and N Estimation provides a new empirical method.