摘要
分析比较几种常见宽波段植被指数和高光谱窄波段植被指数与大豆叶面积指数(LAI)的相关性及其预测力,通过建立不同植被指数与LAI之间的统计回归模型,发现各植被指数均与LAI曲线相关,相关关系可分为两种不同的模式——幂函数关系和指数函数关系。对于全部植被指数和窄波段植被指数而言,一阶微分绿度植被指数1DZ-DGVI对大豆LAI的估算效果最佳;而对于宽波段植被指数而言,以宽波段三角植被指数B-TVI的效果最佳。通过对比发现,在估算大豆LAI方面,窄波段植被指数并没有表现出明显的优势性,有些植被指数甚至还没有其对应宽波段植被指数表现的好。不论从回归分析结果的均方根误差RMSE来看,还是从模型检验的均方根误差RMSE和相对误差RE来看,B-TVI的表现与1DZ-DGVI的表现相差不多,因为两者回归分析和模型检验的RMSE分别相差0.0153、0.0083,模型检验的相对误差仅相差0.0043,这表明宽波段光谱植被指数可以用来监测大豆LAI。图2,表3,参22。
The correlation analyses between soybean LAI and selected vegetation indices (VIs), including some hyperspectral vegetation indices based on waveform analysis of the reflectance across the vegetation red edge were carried out, and the performance of the VIs was evaluated on the basis of their capability to accurately estimate LAI in the paper. The results indicated that the curvilinear relations, including a power relation and an exponent relation, existed between selected VIs and LAI by the regressive analysis between them. The performance of first-order derivative green vegetation index (IDZ-DGVI) was optimal for all the VIs and narrow-band VIs, while broad triangular vegetation index (B-TVI) was best for broad-band VIs. The performance of narrow-band VIs had not been obviously predominant for estimating soybean LAI and even some of them performed worse in LAI estimate accuracy than their broad-band counterparts. The performance of IDZ-DGVI was only slightly better than B-TVI with calibration RMSE and validation RMSE, RE as criteria. The RMSE differences of their regressive analyses and model validations were only 0.0153, 0.0083 respectively, and the RE difference was 0.0043, which indicated that soybean LAI estimation by means of the vegetation index constructed by near-infrared band, red band and green band from broad-band spectral reflectance was reasonable.
出处
《农业系统科学与综合研究》
CSCD
2007年第4期503-508,共6页
System Sciemces and Comprehensive Studies In Agriculture
基金
中国科学院知识创新重点项目(KZCX3-SW-356)
关键词
大豆LAI
植被指数
估算模型
soybean LAI
vegetation indices
estimate models