中文摘要：

高光谱信息是探测植物体内氮素含量状况的重要手段,而植物体中的氮素与水体含氮量息息相关.本研究区为以再生水为主要补给水水源的北京门城湖湿地公园,通过获取区内典型的再生水氮净化挺水植物芦苇和香蒲叶片的高光谱数据,并在室内测定对应样点的水体总氮含量指标,探讨基于典型湿地挺水植物高光谱数据对水体总氮进行遥感探测的可行性.采用4种高光谱参数（光谱指数、归一化差值指数、＂三边＂参数及吸收特征参数）分别建立一元线性模型、逐步多元回归模型和偏最小二乘模型,根据决定系数（R2）和均方根误差（RMSE）进行模型精度检验.结果表明：逐步多元回归和偏最小二乘模型的预测精度高于一元线性模型.3种模型对芦苇的拟合效果均优于香蒲.偏最小二乘模型对芦苇的拟合效果最优（R2=0.854,RMSE=0.647）.500~700 nm是反映水氮含量的最佳波段范围,绿峰与红谷反射率的比值与水体总氮含量具有较强的相关性,尤其是吸收特征参数能够较好地预测水体总氮含量.

英文摘要：

Hyperspectral reflectance information is a crucial method to detect total nitrogen content in plant leaves, meanwhile, vegetation nitrogen content has a strong relationship with nitrogen in water. Taking Mencheng Lake Wetland Park supplied with reclaimed water as study area, the vege- tation hyperspectral data （Phragmites australis and Typha angustifolia）, and the content of total ni- trogen in water were detected to investigate the feasibility of estimating total nitrogen content in re- claimed water based on hyperspectral reflectance information from emergent plants. We established simple linear regression model, stepwise multiple linear regression model and partial least square re- gression model based on four hyperspectral indices （spectra/indices, normalized difference indices, trilateral parameters, absorption feature parameters）, respectively. The accuracy of these models was coefficient of determination （R2 ） and root mean square error （RMSE）. The results showed that stepwise multiple linear regression model and partial least square regression model predicted more accurately than simple linear regression model, and the accuracy of prediction models based on P. australis reflectance spectra was higher than those on T. angustifolia. Partial least square regression model was the most useful explorative tool for unravelin~ the relationship between snectral reflec-tance of P. australis and total nitrogen content in water with R2 of 0. 854 and RMSE of 0. 647. 500-700 nm was the best band range for detecting water total nitrogen content. The reflectance ratio of green peak and red valley could be effectively predicted by the absorption feature parame- ters.