黄河源区高寒草甸群落高度遥感模拟与评价  

Simulation and evaluation of forage height in alpine meadow grassland in headwaters of the Yellow River

作　　者：殷建鹏 高金龙[1,2] 冯琦胜 葛静[1,2] 孟宝平 杨淑霞[1,2] 梁天刚 孟小平[3] Yin Jian-peng;Gao Jin-long;Feng Qi-sheng;Ge Jing;Meng Bao-ping;Yang Shu-xia;Liang Tian-gang;Meng Xiao-ping(State Key Laboratory of Grassland and Agro-ecosystems,College of Pastoral Agriculture Science and Technology,Lanzhou University,Lanzhou 730020,Gansu,China;Key Laboratory of Grassland Livestock Industry Innovation,Ministry of Agriculture,Lanzhou 730020,Gansu,China;Primary school of Mayingtan,Dingxi 748105,Gansu,China)

机构地区：[1]草地农业生态系统国家重点实验室兰州大学草地农业科技学院,甘肃兰州730020 [2]农业部草牧业创新重点实验室,甘肃兰州730020 [3]陇西县马营滩小学,甘肃定西748105

出　　处：《草业科学》2018年第4期737-748,共12页Pratacultural Science

基　　金：国家自然科学基金(31702175;31372367;31672484);长江学者和创新团队发展计划资助(IRT_17R50);中央高校基本科研业务费(lzujbky-2016-186)

摘　　要：草层高度是反映草地生长状况的重要指标之一,与草地地上生物量的监测及载畜力的评估具有紧密联系。目前,对我国天然草地群落高度的监测尚缺乏精确的遥感方法。本研究以黄河源区高寒草甸为研究区,分别基于MODIS植被指数、冠层高度模型以及草地冠层高光谱反射率,构建了高寒草甸草地群落高度的估算模型,并对模型精度进行了评价。结果表明,1)12种MODIS植被指数中,NDVI对草层高度的变化最为敏感,其4种回归模型的R^2均较低,介于0.203~0.241,NDVI指数模型(y=0.789e^(3.186x),R^2=0.241)的拟合效果最好,但反演误差较大(RMSE=4.2cm,CVRMSE=45.7%);2)冠层高度模型在黄河源地区高寒草甸试验区的反演精度较低(RMSE=5.8cm,CVRMSE=62.1%),实际应用误差较大;3)高寒草甸群落高度与519.4-583.17nm波段之间的冠层光谱反射率存在显著负相关关系(P<0.05),基于光谱位置变量R′510.59的线性模型(y=^(-1)56.375x+20.384)相对最优(R^2=0.489,RMSE=3.5cm),较适合反演高寒草甸群落的草层高度。Canopy height in grassland is an important factor for reflecting growth status,and is closely related to monitoring of aboveground biomass and grazing capacity evaluation as well.Presently,accurate remote sensing methods for monitoring canopy height in natural grassland communities are still lacking.In this study,height estimation models based on MODIS vegetation indices,canopy height model,and grassland hyperspectral reflectance were established,and assessed in the source region of the Yellow River,individually.The results showed that:1)NDVI was most sensitive to grassland canopy height in 12 MODIS vegetation indices.However,the R 2 values of four regression models derived from NDVI were low,which were 0.203~0.241.The NDVI exponential model(y=0.789e 3.186 x,R 2=0.241)had the best fitting effect;however,inversion error was greater(RMSE=4.2 cm,CVRMSE=45.7%);2)The accuracy of the canopy height model in alpine meadow was low(RMSE=5.8 cm,CVRMSE=62.1%);3)There was a significant negative correlation between height of alpine meadow community and canopy spectral reflectance ranging within 519.4~583.17 nm band(P<0.05).The linear model y=-156.375x+20.384(R 2=0.489,RMSE=3.5 cm)based on R′510.59 was suitable for estimating the grassland canopy height in alpine meadows.

关 键 词：黄河源区 群落高度 植被指数 冠层高度模型 高光谱反射率 

分 类 号：Q948.15[生物学—植物学]