基于遥感估算方法的干旱区植被覆盖度适应性评价  

Adaptive evaluation of vegetation coverage estimation in arid region based on remote sensing technology

作　　者：谢秋霞[1] 孙林[1] 韦晶[1] 李秀瑞[1] 

机构地区：[1]山东科技大学测绘科学与工程学院,青岛266590

出　　处：《生态学杂志》2016年第4期1117-1124,共8页Chinese Journal of Ecology

基　　金：山东省杰出青年基金(2012JQB01025);国家科技支撑计划项目(2012BAH27B04);国家自然科学基金项目(41501400);所长基金项目(Y5SJ0600CX)资助

摘　　要：与浓密植被覆盖区相比,半干旱、干旱地区因植被覆盖较少、空间分布零散的特点,致使利用遥感手段提取植被覆盖度(vegetation fractional coverage,VFC)的精度较低。针对上述问题,本文首先分别讨论了回归模型法、像元二分法和混合像元分解法3种典型的VFC提取方法在干旱区VFC反演应用中的适应性及限制性因素;然后通过分析干旱区纯土壤端元植被指数值(NDVIs)对植被覆盖度的敏感性,将基于地面光谱测量的NDVIs应用于传统方法进行改进,以新疆大黄山典型干旱区为例,使用Landsat-8 OLI数据,进行植被覆盖度反演实验,最后使用实测VFC数据对反演结果进行精度验证。验证结果表明:基于混合像元分解理论的全约束最小二乘法在干旱区植被覆盖度的反演精度最高,反演值与实测值间的相关性(R^2)达到0.989,其次为改进的像元二分法(R^2=0.848)和回归模型法(R^2=0.827)。Compared with dense vegetation area,arid and semi-arid areas show a lower accuracy in vegetation fractional coverage（ VFC） calculation using remote sensing method,due to its poor and scattered vegetation coverage. To solve the above problems,restrictive and adaptation factors of three typical traditional VFC inversion methods,the regression model method,pixel dichotomy method and the spectral unmi-xing method in arid areas were analyzed in this paper. Then the traditional methods were improved based on ground-measured spectrums through analysis of the sensitivity of NDVISto vegetation cover in arid areas. A VFC inversion experiment was carried out in a typical arid area in Xinjiang using the Landsat-8 OLI data. The measured VFC data were used to verify the experiment results. The results showed that the different methods showed certain difference for VFC inversion accuracy. The fully constrained least squares linear spectral unmixing method based on spectral unmixing showed the highest accuracy in the arid area,with highest consistency with the ground measured VFC（ R^2= 0. 989）,followed by the pixel dichotomy method（ R^2= 0.848） and the regression model method（ R^2= 0.827）.

关 键 词：干旱区 植被覆盖度 回归模型法 像元二分法 混合像元分解法 

分 类 号：Q948[生物学—植物学]