利用辽宁锦州地区2013年生长季不同土壤水分控制条件下的春玉米冠层高光谱数据, 及对应的植株叶面积指数(leaf area index, LAI)数据, 分析在不同发育期内不同生长状况下的春玉米冠层高光谱特征及其与植株叶面积指数的关系。 采集并计算共313组有效样本, 包括350～2 500 nm波段范围光谱的反射率、 反射率倒数的对数、 反射率一阶导数及LAI, 应用多元逐步线性回归法和偏最小二乘回归法, 对剔除了受大气水分影响较为严重光谱波段的其他波段数据进行降维, 构建叶面积指数的全波段冠层高光谱数据模型, 并进行精度检验与比较。 结果表明, 春玉米LAI与光谱反射率在可见光波段(350～680 nm)、 红外波段(1 430～1 800和1 950～2 450 nm)均呈显著的负相关; 反射率倒数的对数在对应区间为显著的正相关; 反射率一阶导数则在可见光和近红外波段(350～1 350 nm)存在较显著相关波段。 三种全波段冠层高光谱数据在春玉米LAI的线性回归中, 偏最小二乘法在以冠层反射率为自变量的模型构建中, 比多元逐步线性回归拟合度好, 其总均方根误差为0.480 7; 以冠层光谱反射率的倒数的对数及一阶导数为自变量, 应用逐步线性回归法建模, 拟合度较好, 其总均方根误差分别为0.333 5和0.348 8; 三种光谱数据的春玉米LAI两种回归算法中, 以冠层反射率倒数的对数为自变量, 应用逐步线性回归方法建模的拟合度最佳。

Based on the leaf area index (LAI) and canopy hyperspectral data during growing season of spring maize under different soil moisture conditions in Jinzhou, Liaoning province in 2013, the relationship between LAI and the characteristics of canopy hyperspectral in different development periods with different growth status were analyzed. Canopy spectral reflectance, its logarithm of the reciprocal and its first derivative in 350～2 500 nm of 313 valid data sets were collected and calculated, after rejecting the bands which were serious influenced by the atmospheric water content. Multivariate step linear regression (MSLR) and partial least squares regression (PLS) were used as the dimensionality reduction methods to establish the maize LAI models, and the models' precision were compared and tested respectively. The results show that, the LAI of spring maize has significant negative correlation with the spectral reflectance of visible band (350～680 nm), and infrared band (1 430～1 800 and 1 950～2 450 nm), but it has significant positive correlation with the logarithm of the reflectance reciprocal in these bands. The reflectance first derivative and LAI have significant positive correlation bands in visible band and infrared band (350～1 350 nm). Linear regression algorithm of spring maize LAI with the whole band of hyperspectral data, using PLS with the spectral reflectance as the independent variable to establish the LAI model, the fitting degree is better than that of MSLR; the root mean square error (RMSE) is 0.480 7, and using MSLR with the logarithm of the reflectance reciprocal and the reflectance first derivative as the independent variable, have better fitting degree than that of PLS, the RMSE are 0.333 5 and 0.348 8 respectively. Use MSLR with the logarithm of the spectral reflectance reciprocal as the independent variable to establish the maize LAI model, the fitting degree is better in the three canopy hyperspectral data of spring maize of the two regression algorithm.