中文摘要：

干旱区荒漠植物的叶绿素定量反演是动态监测、快速有效评估植物生物量及长势的有效方法.利用便携式可见-近红外光谱仪FieldSpecPro3测定绿洲、盐碱地及沙漠3种生境内的芦苇高光谱值,对高光谱数据一阶微分以及红边参数与叶绿素含量进行了相关分析,选取最佳红边参数建立经验估算模型与神经网络模型,并评估检验.模型显示,三种生境下均为二项式回归模型的决定系数最佳,检验精度的决定系数（R2）分别为0.8466、0.8672和0.7935,均方根误差RMSE（root-mean-squareerror）分别为2.3601、1.4112和2.8002;BP神经网络模型的检验精度的决定系数（R2）分别为0.9147、0.9331和0.8813,RSME分别为1.4010、0.9964和0.5559.结果表明,利用BP（backpropagation）神经网络估算的模型精确度显著提高,可作为芦苇叶绿素高光谱反演的有效模型而使用,为荒漠植物叶片叶绿素的光谱特征反演提供了借鉴,为监测荒漠植物生长、产量估算及动态监测等提供可行的手段.

英文摘要：

The quantitative inversion of chlorophyll of desert plants in arid areas is an effective approach of dynamic monitoring and fast assessment of biomass and plant growth. Hyper spectral reflectances of Phragmites in Oasis habitat, saline area and desert habitat were measured by FieldSpecPro3 portable spectrometer. Correlation analysis was carried on hyperspectral data of first order and red edge parameters and chlorophyll content, meanwhile, red edge parameters were chose to establish optimal experience estimation model and the BPNN model. Models showed that the determination coefficients （R2） of binomial regression model reached the best, which were 0.8466, 0.8672 and 0.7935; RMSE was 2.3601, 1.4112 and 2.8002. For BPNN model, R2 of accuracy test separately was 0.9147, 0.9331 and 0.8813; RMSE was 1. 4010, 0.9964 and 0.5559. Thus, the accuracy of BPNN was the best model and it could be widely used to be a kind of good chlorophyll hyperspectral inversion model, in providing reference for desert plant chlorophyll spectral characteristics of the inversion and providing a convenient and feasible way for monitoring plant growth and physiological characteristics.