基于高光谱遥感的不同品种花生冠层叶面积指数的通用估算模型

摘要

为了快速、无损地监测花生冠层LAI，获取其长势信息，于2014年通过大田试验选用5个品种花生作为供试品种，使用ASD FieldSpec HandHeld便携式野外光谱仪采集花生不同生育时期的冠层高光谱数据，同时使用SUNSCAN冠层分析系统实测花生冠层叶面积指数（LAI），并应用光谱微分技术和统计分析技术，分别分析4种光谱形式和6种植被指数与LAI的相关关系，建立估算模型。结果表明：高光谱反射率及其光谱变换形式中最优波段与LAI的相关性均极显著，其中一阶微分光谱ρ'在793nm波段处构建的估测方程对花生冠层 LAI 的估算效果最好（r=-0.5391，P＜0.01，RE=0.2497），其模拟值与实测值的拟合度达极显著水平（R=0.4435，P＜0.01）；一阶微分光谱ρ'在734nm 波段处 LAI 的实测值与模拟值的拟合效果最好（ R=0.5485， P＜0.01）。6种植被指数所选的最优组合波段与 LAI 均通过了0.01水平的显著性检验（r≥0.5731，P＜0.01），其中归一化植被指数NDVI[760，976]对花生冠层LAI的估算效果最好（r=-0.6421， P＜0.01，RE=0.2167），模拟值与实测值的拟合度达极显著水平（R=0.6731，P＜0.01），且优于ρ'对 LAI的估算效果；LAI实测值与模拟值拟合效果最好的为DVI[760，976]（R=0.6893，P＜0.01）。研究结果表明一阶导数光谱和植被指数对花生冠层LAI的估算精度均较高，植被指数的估算精度尤高，研究同时进一步证实了导数光谱和植被指数能较好地消除土壤、大气等环境背景信息的影响。%In order to monitor peanut canopy effectively and nondestructively and to get their growth information, hyperspectral data of peanut canopy was measured with ASD FieldSpec and canopy leaf area index was measured with SUNSCAN during different growing stages in the field experiment, including five cultivars with different ecological types. The correlations between four spectral forms, six vegetation index and LAI were analyzed, and the estimation models were established, by using spectral derivative technique and statistical analysis technique. The results showed that the correlation between the optimal bands of hyperspectral reflectance, its transformation forms and LAI were 1% significant, LAI could be estimated better by the first derivative spectra ρ' at 793 nm(r=-0.5391, P<0.01, RE=0.2497), the correlation coefficient between simulating data and testing data was 1% significant (R=0.4435, P<0.01); simulating LAI and testing LAI could be fitted mostly by the first derivative sPectra ρ ' at 734nm(R=0.5485, P<0.01). The correlations between the optimal bands of six vegetation index and LAI were 1% significant(r≥0.5731, P<0.01), LAI could be estimated mostly by NDVI [760, 976](r=-0.6421,P<0.01, RE=0.2167), the correlation coefficient between simulating data and testing data was 1% significant(R=0.6731, P<0.01); simulating LAI and testing LAI could be fitted mostly by DVI [760, 976](R=0.6893, P<0.01). The results indicated that the estimation accuracy of peanut canopy was higher than that of with ρ' and vegetation index, especially vegetation index. It confirmed further that ρ' and vegetation index played a great role on eliminating the effect of environment background including soil and atmosphere.