中文摘要：

牧草生物量是天然和人工草场产出的一个重要指标,能直接反映牧草的长势,它的实时监测是制定合理的草地管理和放牧制度的前提。然而,传统的样方法费时、费力很难满足草地生物量监测的实时性要求。近年来,随着高光谱遥感的发展,为草地的快速、无损的监测提供了可能。通过对内蒙古天然和人工草地生物量及其冠层光谱数据分析,探讨了不同类型光谱指数在估算牧草生物量的鲁棒性。研究表明,由于冠层结构和生物量的影响,不同类型草地冠层的光谱反射存在巨大的差异,从而使现有光谱指数对天然草地和不同品种人工草地牧草生物量的预测能力存在显著的差异（R2=0.00~0.65）。在荒漠草原低生物量的条件下,受土壤背景影响不同光谱指数预测能力较小,而在青贮玉米高生物量条件下,基于红光的光谱指数容易发生饱和失去敏感性。把不同草地类型的数据结合后进行简单比率和归一化窄波段光谱指数的波段优化,与出版的光谱指数相比优化后的归一化光谱指数（normalized difference spectral index,NDSI）显著提高了牧草生物量的预测能力,预测生物量模型的决定系数最高（R2=0.72）。敏感性分析进一步证明了基于波段优化算法的优化光谱指数NDSI,（ration spectral index,RSI）有最低的噪声,从而能更好的预测牧草生物量。波段优化算法可以有效提高遥感估算草地牧草生物量的精度,光谱指数及其优化波段的选择会直接影响模型的预测能力。

英文摘要：

As an important indicator of forage production, aboveground biomass will directly illustrate the growth of forage grass. Therefore, Real-time monitoring biomass of forage grass play a crucial role in performing suitable grazing and manage ment in artificial and natural grassland. However, traditional sampling and measuring are time-consuming and labor-intensive. Recently, development of hyperspectral remote sensing provides the feasibility in timely and nondestructive deriving biomass of forage grass. In the present study, the main objectives were to explore the robustness of published and optimized spectral indices in estimating biomass of forage grass in natural and artificial pasture. The natural pasture with four grazing density （control, light grazing, moderate grazing and high grazing） was designed in desert steppe, and different forage cultivars with different N rate were conducted in artificial forage fields in Inner Mongolia. The canopy reflectance and biomass in each plot were measured during critical stages. The result showed that, due to the influence in canopy structure and biomass, the canopy reflectance have a great difference in different type of forage grass. The best performing spectral index varied in different species of forage grass with different treatments （R^2= 0. 00-0. 69）. The predictive ability of spectral indices decreased under low biomass of desert steppe, while red band based spectral indices lost sensitivity under moderate-high biomass of forage maize. When band combinations of simple ratio and normalized difference spectral indices were optimized in combined datasets of natural and artificial grassland, optimized spectral indices significant increased predictive ability and the model between biomass and optimized spectral indi ces had the highest R^2 （R^2 =0.72） compared to published spectral indices. Sensitive analysis further confirmed that the optimized index had the lowest noise equivalent and were the best performing index in estimating biomass. In conclusion, opt