基于不同光谱变换的剑湖茭草鲜生物量估测研究

通过实地采集剑湖湿地茭草反射光谱和现场测量鲜生物量，基于24种光谱变换对茭草反射光谱特征进行分析，选取16种光谱变换筛选全波段（350～2 350 nm）中对茭草鲜生物量敏感的特征波段，构建其鲜生物量估测模型。结果表明：不同形式的光谱变换更容易分析光谱特征，对数倒数和倒数的变换增强了可见光波段的特征。对数倒数一阶微分变换增强了近红外波段的特征，倒数二阶微分和对数倒数二阶微分增强了短波红外的特征，4～5尺度的连续小波变换适合分析原始光谱特征。连续小波变换后最大相关系数为0.734；其次为二阶微分变换，最大相关系数为−0.730。基于立方根二阶微分变换构建的多元回归模型对茭草鲜生物量估测效果最佳，R²、RMSE、P和RPD分别为0.88、1 044.90 g/m²、83.95%、2.64。

The Study on Fresh Biomass Estimation of Zizania latifolia Based on Different Spectral Transformations of Spectral Reflectance

The reflectance spectrum were collected and fresh biomass were measured for Zizania latifolia in situ of Jianhu Wetland, and the analysis of reflectance spectral for Z. latifolia based on 24 spectral transformations. Then establishment of the fresh biomass estimation models based on the characteristic bands selection were sensitive to biomass on analysis of the full bands (350−2 350 nm) of 16 spectral transformations. The results showed that spectral characteristics can be more easily analyzed by the different spectral transformation. The visible band characteristics was enhanced by logarithmic reciprocal and reciprocal transformation. The near infrared band characteristics was enhanced by first-order derivative of logarithmic reciprocal transformation. The short-wave infrared band characteristics was enhanced by second-order derivatives of reciprocal and second-order derivative of logarithmic reciprocal. And 4−5 scale continuous wavelet transformation were suitable for the original spectral characteristics analyzing. The highest correlation coefficient of 0.734 after continuous wavelet transformation, and followed second-order derivative transformation with the highest correlation coefficient of −0.730 in all spectral transformations. The best effective model on estimating fresh biomass of Z. latifolia was multivariate regression model based on second-order derivative of cubic root transformation, R2, RMSE, P and RPD were 0.88, 1 044.90 g/m2, 83.95% and 2.64, respectively.