利用光谱技术监测植被水分状况是了解植被生理状况及生长趋势的重要手段之一。 选择艾比湖湿地自然保护区作为靶区。 采用聚类分析、 变量投影重要性分析(VIP)以及敏感性分析等方法， 对植被不同含水量进行分级， 并针对不同等级的植被含水量进行估算及验证。 结果表明： (1)基于聚类分析中的欧氏距离的方法将植被叶片相对含水量划分为高等、 中等、 低等三个等级， 其范围分别为7076%～8069%， 5327%～7076%， 3100%～5327%。 在中红外与远红外（1 350～2 500 nm）之间， 反射率越低植被含水量越高； 波长380～1 350 nm范围， 无此现象。 (2)应用VIP方法可知， 所选的8种植被水分指数VIP值均超过了08， 说明植被水分指数预测能力均较强且差别不显著。 其中MSI， GVMI与植被叶片相对含水量的非线性三次拟合函数效果最佳， MSI决定系数R2为06575和GVMI决定系数R2为0674 2。 植被叶片相对含水量在30%～45%范围， MSI指数的NE值最低， 在45%～90%范围时， GVMI指数的NE值最低。 NDWI1240指数的NE值在70%左右起伏较大， 说明NDWI1240 指数在植被含水量为70%左右， 预测能力较差。 (3)通过误差分析可知GVMI指数反演的结果误差最小， 不同的植被指数对不同含水量的植被估算结果相差较为明显， 因此分段估算植被含水量是有必要的。 综上所述， 利用高光谱遥感技术对监测艾比湖保护区植被生长及干旱环境提供基础研究。

Monitoring the water status of vegetation by spectral technique is one of the important means to understand the physiological status and growth trend of vegetation. In this study, the Ebinur Lake Wetland Nature Reserve is chosen as the target area. By using cluster analysis, variable importance projection (VIP) and sensitivity analysis method, the vegetation water content was classified, estimated and validated. The Results showed that in the clustering analysis method based on Euclidean distance of the vegetation moisture content is divided into three grades with higher water content, medium water content and low water content, whose ranges are around 7076%~8069%, 5327%~7076% and 31%~5327%, respectively. From 1 350 to 2 500 nm wavelength range, the spectral reflectance of water content is the lowest ,however there is no law from 380 to 1 350 nm wavelength range. By using VIP method, all vegetation water index VIP value of more than 08, indicated that vegetation water index estimation ability of water content of vegetation leaves is strong and the difference is not obvious. The MSI, or GVMI and vegetation water content cubic equation fitting is the best, the fitting coefficients of R2 were 0657 5 and 0674 2 respectively. The RWC in the range of 30%~45%, the MSI value of the NE index is the lowest. In the range of 45%~90%, the GVMI value of the NE index is the lowest. About 70% of NE value NDWI1240 index has undulation, it shows that the NDWI1240 index of the vegetation water content is at about 70% and the prediction ability is poor. Through the error analysis, the error of GVMI exponent inversion is the smallest, different vegetation indices have obvious difference in vegetation estimation results with different water contents. Therefore, it is necessary to estimate vegetation water content. In summary, using hyper spectral remote sensing technology to monitor vegetation growth and drought environment in Ebinur Lake Reserve Area is feasible.The results provide a theoretical basis for the large area inversion of satellite borne hyper spectral sensors for vegetation water content.