摘要
叶片含水量是反映作物生理特性的一个重要参数,对生态环境的研究具有重要意义。采用小波分析方法,分析叶片含水量对反射率的影响特征,建立综合利用多波段信息的作物叶片水分含量反演模型。基于PROSPECT模型的辐射传输理论,推导出由叶片反射率光谱的小波系数反演叶片水分含量CW的理论模型。利用六种常用的小波函数,对叶片组分水、干物质和白化基本层的吸收光谱进行小波分解。选取对水分变化最敏感,同时对其他组分不敏感的分解尺度和波段位置,找到能稳定突出水的光谱特征的小波系数。结果表明:bior1.5小波函数在尺度为200nm,波段位置为1 405和1 488nm的小波系数具有上述特征。建立由叶片反射率光谱的bior1.5小波系数反演叶片水分含量CW的反演模型,模型有两个转换系数a和Δ都受叶片结构参数N的影响。利用PROSPECT模型生成模拟光谱数据集,校正建立的叶片水分含量反演模型中的两个转换系数a和Δ,并与LOPEX93实验光谱数据集结合验证反演模型。结果表明:反演模型不仅比传统基于植被指数的统计模型在精度上有提高(反演值与实测值的R2最高达到0.987),而且更加稳定,普适性更高。研究表明,小波分析方法在利用高光谱数据反演作物叶片水分含量方面具有独特的优势。
Leaf water content is a fundamental physiological characteristic parameter of crops ,and plays an important role in the study of the ecological environment .The aim of the work reported in this paper was to focus upon the retrieval of leaf water con-tent from leaf-scale reflectance spectra by developing a physical inversion model based on the radiative transfer theory and wavelet analysis techniques .A continuous wavelet transform was performed on each of leaf component specific absorption coefficients to pick wavelet coefficients that were identified as highly sensitive to leaf water content and insensitive to other components .In the present study ,for identifying the most appropriate wavelet ,the six frequently used wavelet functions available within MATLAB were tested .Two bior1.5 wavelet coefficients observed at the scale of 200 nm are provided with good performance ,their wave-length positions are located at 1 405 and 1 488 nm ,respectively .Two factors （a and Δ） of the predictive theoretical models based on the bior1.5 wavelet coefficients of the leaf-scale reflectance spectra were determined by leaf structure parameter N .We built a database composed of thousands of simulated leaf reflectance spectra with the PROSPECT model .The entire dataset was split into two parts ,with 60% the calibration subset assigned to calibrating two factors （a and Δ） of the predictive theoretical model . The remaining 40% the validation subset combined with the LOPEX93 experimental dataset used for validating the models .The results showed that the accuracy of the models compare to the statistical regression models derived from the traditional vegetation indices has improved with the highest predictive coefficient of determination （R2 ） of 0.987 ,and the model becomes more robust . This study presented that wavelet analysis has the potential to capture much more of the information contained with reflectance spectra than previous analytical approaches which have tended to focus on using a small number of optimal wavebands while dis-carding the majority of the spectrum .
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2015年第1期167-171,共5页
Spectroscopy and Spectral Analysis
基金
国家(973计划)项目(2010CB950702)资助
关键词
高光谱特征
叶片含水量
小波分析
反演模型
Hyper-spectral features
Leaf water content
Wavelet analysis
Inversion model
作者简介
方美红，1986年生，南京大学国际地球系统科学研究所博士研究生e-mail：fmh19860616@126．com
通讯联系人e-mail：juweimin@nju．edu．cn
