中文摘要：

为了在作物氮素管理中实现叶片氮含量的实时无损估测，以不同类型小麦品种在不同施氮水平下连续3年田间试验为基础，研究了小麦叶片氮含量与冠层高光谱参数的定量关系。结果表明，叶片氮含量随着施氮水平的增加而提高，冠层光谱反射率在不同叶片氮含量水平下存在明显差异。叶片氮含量的敏感波段主要存在于近红外平台和可见光区，其中，红边区域最为显著。红边及面积类参数REPIE、SDr-SDb和FD729与叶片氮含量关系密切，方程拟合决定系数R^2分别为0.829、0.806和0.856，估计标准误差SE分别为0.278、0.295和0.271；模拟宽光谱波段组合类参数方程拟合精度较低，标准误差较大，以AVHRR-GVI为变量模拟方程，R^2和SE分别为0.786和0.315；多波段组合类参数方程拟合效果较好，以mND705为变量建立方程，其R^2和SE分别为0.836和0.275。经不同年际独立数据检验，红边及面积类参数表现最好，以REPIE、SDr-SDb和FD729三个参数为变量，模型预测的RMSE分别为0.418、0.380和0.395，相对误差RE分别为14.4%、15.1%和15.2%；模拟宽光谱波段组合类参数与多波段组合类参数比较，模拟宽光谱波段组合模型预测效果更好，以AVHRR-GVI和mND705为变量建立模型，RMSE分别为0.436和0.408，RE分别为17.3%和16.7%。以上结果表明，红边及面积类参数与叶片氮含量关系密切且表现稳定，利用REPIE、SDr-SDb和FD729三个参数可以对小麦叶片氮含量进行可靠的监测。

英文摘要：

Crop nitrogen status is a key index for evaluating crop growth,increasing yield and improving quality.Non-destructive﹠quick assessment of leaf nitrogen concentration is very necessary for management of nitrogen nutrition in crop production.This study investigated the quantitative relationships of leaf nitrogen concentration to canopy hyperspectral reflectance in wheat with three field experiments consisting of different variety types and nitrogen levels in three growing seasons.The results showed that the nitrogen concentration in wheat leaf increased with increasing of applying nitrogen rates,and changes in canopy spectral reflectance with varied leaf nitrogen concentration were all highly different.The sensitivity bands occurred during visible light and near infrared region mostly,and a close correlation existed between red-edge district and leaf nitrogen concentrations.The relationships between twenty vegetation indices from references,the first derivatives and leaf nitrogen concentrations were analyzed using statistical models.As a result,the area and red edge spectral parameters were significantly correlated with leaf nitrogen concentrations,and the relationships to REPIE,SDr-SDb and FD729 were all highly significant,with the determination of coefficients（R^2）as 0.829,0.806 and 0.856,and the standard errors（SE）as 0.278,0.295 and 0.271,respectively.The relationships between the combination of several single bands and leaf N concentration were better than those of the combination of simulated wide spectra bands in terms of R^2 and SE.Regression models of leaf N concentrations to mND705 and AVHRR-GVI were established with R^2 of 0.836 and 0.786,SE of 0.275 and 0.315,respectively.The validation results from independent datasets indicated that the area and red edge spectral parameters were the best to predict leaf N concentrations,with RMSE of 0.418,0.380 and 0.395,RE of 14.4%,15.1% and 15.2% for using REPIE,SDr-SDb and FD729,respectively.The combination of several single bands were slightly lower as to