玉米叶片氮含量的高光谱估算及其品种差异

Hyperspectral diagnosis of leaf N concentration of maize and variety difference

准确、快速、及时地对玉米氮营养状况做出判断是氮肥合理施用的基础。该研究在水培条件下对3个玉米品种(组合)叶片氮含量(LNC)的高光谱敏感波段、估算模型及其品种差异进行了探讨。结果表明,LNC与不同波段叶片光谱反射率的相关性存在品种差异,但3个品种(组合)都在500～649nm和691～730nm表现极显著的负相关关系,并在同一波长获得最高的相关系数,说明可以利用统一的波段来预测不同品种的LNC。依品种建立了LNC与归一化差值光谱指数(NDSI)或比值光谱指数(RSI)的定量关系模型,NDSI(714,554)和RSI(714,554)所建模型的拟合度最好,直线和指数模型拟合度均达到极显著水平,可以用来估算玉米LNC。玉米LNC估算中,以该品种数据所建模型的估算偏差最低,利用综合模型或其他品种模型则加大了估算偏差,高估与低估的最高偏差分别为35.6%和32.7%。在利用高光谱技术进行玉米氮营养状况诊断的研究及应用中,应考虑品种间差异。

The accurate,rapid and timely monitoring and diagnosis of maize nitrogen（N）nutrition is the base of the reasonable application of N fertilizer.By using the method of potted plant experiment to precisely control the supply of N,the sensitive spectral band and diagnosis model of three kinds of maize（hybrid）for leaf N concentration（LNC） monitoring and their variety difference were studied.The results show that：1）There is variety difference in correlation between the LNC and the leaf spectrum reflectance within the range of 350-2 500 nm,but it presents significant negative correlation at F=0.01 between 500-649 nm and 691-730 nm for all the three varieties（hybrid）,and gets the highest correlation coefficient at the same wavelength.All of those show that an unified spectral band can be used to diagnose LNC in different varieties.2）The quantitative equations of different varieties are established between maize LNC and normalized difference spectral index（NDSI）or ratio spectral index（RSI）.The fitting degree is the best based on NDSI （714,554）and RSI（714,554）and is significantly at F=0.01 for the linear and exponential equations,so the derived regression models can be reliably used for estimating LNC in maize.3）The deviation of estimation for maize LNC is the lowest when using the regression model derived from the same variety,but the deviation increases when using a model derived from multi-varieties or other variety.The maximum deviation of up or down estimation are 35.6%and 32.7%, respectively.The variety difference should be taken account,when using hyperspectral technique to diagnose the maize plant N status in researches and applications.