祁连山东缘高寒草地植物群落叶绿素高光谱反演模型的建立

Construction of chlorophyll hyperspectral inversemodel of alpine grassland community inEastern Qilian Mountains

摘要:

祁连山高寒草地是我国北方草地的重要组成部分,近年来退化较严重,导致其产量、质量下降。利用ASD地物光谱仪采集了祁连山东缘高寒草地植物群落光谱数据,并利用Person相关系数法、PCA主成分分析法和VIP变量投影重要性法等方法,筛选了与高寒草地植物群落叶绿素相关性较高的原始光谱波段与植被指数,通过多元逐步回归与多元线性回归建立了叶绿素反演模型,为祁连山东缘高寒草地的遥感监测提供技术支撑。结果如下:1)原始光谱384、528、721 nm波段的反射率与高寒草地植物群落的叶绿素相关系数较高,共筛选出522个原始光谱波段作为多元逐步回归变量;2)植被指数RVI、SAVI、NDVI₆₇₀、VARI、PSRI、ARVI、RGI、GI、OSAVI、GNDVI与群落叶绿素显著相关,且优于单波段原始光谱;3)利用筛选出敏感波段的原始光谱反射率与植被指数指标建立反演模型,其中原始光谱多元逐步回归模型精度(R²=0.889)最高,且模型检验结果较好(R²=0.9161,RMSE=0.05),可作为高寒草地植物群落叶绿素的反演模型。

Abstract:

The alpine grassland of Qilian Mountains is an important part of grassland in northern China.The serious soil degradation of that area in recent years leads to a significant decline in grass yield and quality.The chlorophyll content of forage plants can effectively reflect the growth status of vegetation.Traditional methods for chlorophyll determination is not only time-consuming,but also could cause a certain degree of damage to grassland vegetation.Therefore,the use of hyperspectral technology to monitor plant quality and degradation status of grassland in a real-time manner would be very important.The ASD ground spectrometer was used to collect the spectral data from plant community of the alpine grassland in the Eastern Qilian Mountains.Person correlation coefficient method,PCA principal component analysis and VIP variable projection importance method were used to screen the chlorophyll in plant community of the alpine grassland.The original spectral band and vegetation index of the chlorophyll inversion model were established using multiple stepwise regression and multiple linear regression methods.Our results showed that the reflectivity of the original spectral bands at 384 nm,528 nm,and 721 nm and the chlorophyll correlation coefficient of the alpine grassland plant community are high.A total of 522 original spectral bands were screened as multiple stepwise regression variables.NDVI₆₇₀,VARI,PSRI,ARVI,RGI,GI,OSAVI,GNDVI are significantly correlated with chlorophyll,and the correlation between vegetation index and chlorophyll was superior to the single-band original spectrum.An inversion model was developed using the original spectral reflectance and vegetation index of the sensitive band ,particularly the original spectral multiple stepwise regression model showed the highest accuracy (R²=0.889) with excellent model test results (R²=0.916 1,RMSE=0.05).