基于小波变换的毛竹叶片净光合速率高光谱遥感反演  

Retrieval of leaf net photosynthetic rate of moso bamboo forests using hyperspectral remote sen-sing based on wavelet transform

作　　者：孙少波[1,2] 杜华强[1,2] 李平衡[1,2] 周国模[1,2] 徐小军[1,2] 高国龙[1,2] 李雪建[1,2] 

机构地区：[1]浙江省森林生态系统碳循环与固碳减排重点实验室,浙江临安311300 [2]浙江农林大学环境与资源学院,浙江临安311300

出　　处：《应用生态学报》2016年第1期49-58,共10页Chinese Journal of Applied Ecology

基　　金：浙江省杰出青年科学基金项目(LR14C160001);国家自然科学基金项目(31370637;61190114);国家林业局‘948’项目(2013-4-71);浙江省本科院校中青年学科带头人学术攀登项目(pd2013239)资助~~

摘　　要：在对毛竹林叶片高光谱反射率数据进行小波变换的基础上,寻找和确定最佳的小波植被指数反演毛竹林叶片的净光合速率(P_n).结果表明:理想的高频小波植被指数反演得到的P_n精度高于低频小波植被指数和光谱植被指数,其中,由小波分解第一层高频系数构建的归一化植被指数、比值植被指数和差值植被指数与P_n之间的相关性最好,R^2为0.7,均方根误差(RMSE)较低,为0.33;而低频小波植被指数反演P_n的精度低于光谱植被指数.由各层理想小波植被指数所构建的多元线性模型反演得到毛竹叶片P_n与实测P_n之间具有显著的相关关系,R^2为0.77,RMSE为0.29,且精度明显高于基于光谱植被指数所构建的多元线性模型.与光谱植被指数反演毛竹P_n的敏感波段仅局限于可见光波段相比,小波植被指数探测的敏感波长范围更广,包含了可见光及多个红外波段.高光谱数据在经过小波变换后能够发现更多反映毛竹P_n的细节信息,且整体反演精度比原始光谱有了显著提高,研究结果为基于高光谱遥感反演植被P_n提供了一种新的可选方法.This study focused on retrieval of net photosynthetic rate（Pn） of moso bamboo forest based on analysis of wavelet transform on hyperspectral reflectance data of moso bamboo forest leaf.The result showed that the accuracy of Pn retrieved by the ideal high frequency wavelet vegetation index（VI） was higher than that retrieved by low frequency wavelet VI and spectral VI.Normalized difference vegetation index of wavelet（NDVIw）,simple ratio vegetation index of wavelet（SRw）and difference vegetation index of wavelet（Dw） constructed by the first layer of high frequency coefficient through wavelet decomposition had the highest relationship with Pn,with the R^2 of 0.7 and RMSE of 0.33;low frequency wavelet VI had no advantage compared with spectral VI.Significant correlation existed between Pn estimated by multivariate linear model constructed by the ideal wavelet VI and the measured Pn,with the R^2 of 0.77 and RMSE of 0.29,and the accuracy was significantly higher than that of using the spectral VI.Compared with the fact that sensitive spectral bands of the retrieval through spectral VI were limited in the range of visible light,the wavelength of sensitive bands of wavelet VI ranged more widely from visible to infrared bands.The results illustrated that spectrum of wavelet transform could reflect the Pn of moso bamboo more in detail,and the overall accuracy was significantly improved than that using the original spectral data,which provided a new alternative method for retrieval of Pn of moso bamboo forest using hyper spectral remotely sensed data.

关 键 词：毛竹 净光合速率 反演 小波变换 高光谱遥感 

分 类 号：S795[农业科学—林木遗传育种] S771.8[农业科学—林学]