5种植被指数识别青藏高原季节牧场比较研究

Identification of seasonal pasture in QinghaiTibetan Plateau based on five vegetation indices

摘要:

季节牧场(冷季牧场和暖季牧场)面积是载畜量计算的重要指标，对于草畜平衡季节动态管理具有重要指导意义。本研究基于 Landsat 8 影像构建 5 种对植被扰动敏感的植被指数(NBR、CSI、 NBR2、NDMI 和 NDVI)，通过 5 种植被指数对青藏高原地区两类最主要的高寒草地(高寒草甸和高寒草原)的季节牧场可利用面积进行识别，并通过区分度(M)和分类精度定量评价各指数识别季节牧场的能力。结果表明：1)受到放牧家畜的啃食和践踏等干扰，在暖季放牧末期，冷季牧场和暖季牧场光谱差异明显；2)5 种植被指数均能很好地区分高寒草地的季节牧场，M 值均高于 2. 38，总体精度均高于 87. 33%；3)基于近红外(NIR)和短波红外(SWIR2)波段构建的 NBR 指数对高寒草甸和高寒草原的季节牧场分类效果最好，总体精度均高于 93. 33%，Kappa 系数均大于 0. 83。

Abstract:

The usable area of seasonal pasture(cold and warm)is an important parameter for the calculation of livestock carrying capacity，which has instructive significance for seasonal dynamic management of forage-livestock balance. In this study，five vegetation indices(NBR，CSI，NBR2，NDMI and NDVI)which are sensitive to vegeta- tion disturbance were constructed based on Landsat 8 images to identify the seasonal usable area of the two main al- pine grasslands(alpine meadow and alpine steppe)in the Qinghai-Tibet Plateau. The abilities of the five vegetation indices for the identification of seasonal pasture were quantitatively evaluated by separability index M and classifica- tion accuracy. The results showed that the spectral differences between cool and warm season pastures were obvious at the end of warm season grazing due to disturbances from gnawing and trampling of grazing livestock. Five vegeta- tion indices distinguished the seasonal pasture of alpine grassland. The M values were all higher than 2. 38，and the overall accuracy was higher than 87. 33%. The NBR index based on NIR and SWIR2 band was the best vegetation index for the classification of season pastures in alpine grasslands，with the overall accuracy was higher than 93. 33%. The Kappa coefficients were all higher than 0. 83.