摘要
连续一致的NDVI时间序列数据是陆地表面特征长期监测的基础和前提.AVHRR NDVI和MODIS NDVI作为时间记录最长和时空分辨率较高数据的典型代表,是未来植被动态监测极为重要的数据源.深入理解两种数据之间的关系,是延续陆地植被长期监测的关键.利用2000—2006年重叠时段的GIMMS NDVI和MODIS NDVI数据,在青藏高原整体、亚区域、植被类型和像元等多尺度对比分析了两种数据的数值差异和动态变化的一致性,并使用495幅20 km×20 km的Landsat影像计算的NDVI,独立地评估了两种数据集的性能.结果表明:GIMMS NDVI和MODIS NDVI捕获青藏高原月尺度物候变化的能力基本相同(显著性水平大多达到0.001);不同植被类型之间两种数据的相似性差异显著,高覆盖的林地一致性较差,均质化较强的草地、农田的一致性较强;像元尺度,两种数据集在82%的研究区域显著一致;在反映植被空间分布方面,MODIS NDVI的数值更接近Landsat NDVI,而GIMMS NDVI在植被动态变化上与Landsat NDVI更相像,不同植被类型之间差异显著,林地MODIS NDVI与Landsat一致性更好,而草地、农田则是GIMMS NDVI更好.融合两种数据,建立一致的NDVI时间序列数据是可行的.在耦合数据时,需要考虑不同植被类型、不同物候期、不同空间尺度对结果的影响.对于针叶林、阔叶林等植被类型,以及物候过渡期的春秋季进行两种数据集成时需要慎重处理.
Consistent NDVI time series are basic and prerequisite in long-term monitoring of land surface properties. Advanced very high resolution radiometer （AVHRR） measurements provide the longest records of continuous global satellite measurements sensitive to live green vegetation, and moderate resolution imaging spectroradiometer （MODIS） is more recent typical with high spatial and temporal resolution. Understanding the relationship between the AVHRRderived NDVI and MODIS NDVI is critical to continued longterm monitoring of ecological resources. NDVI time series acquired by the global inventory modeling and mapping studies （GIMMS） and Terra MODIS were compared over the same time periods from 2000 to 2006 at four scales of Qinghai-Tibet Plateau （whole region, sub-region, biome and pixel） to assess the level of agreement in terms of absolute values and dynamic change by independently assessing the performance of GIMMS and MODIS NDVI and using 495 Landsat samples of 20 km ×20 km covering major land cover type. High correlations existed between the two datasets at the four scales, indicating their mostly equal capability of capturing seasonal and monthly phenological variations （mostly at 0.001 significance level）. Similarities of the two datasets differed significantly among different vegetation types. The relative low correlation coefficients and large difference of NDVI value between the two datasets were found among dense vegetation types including broadleaf forest and needleleaf forest, yet the correlations were strong and the deviations were small in more homogeneous vegetation types, such as meadow, steppe and crop. 82% of study area was characterized by strong consistency between GIMMS and MODIS NDVI at pixel scale. In the Landsat NDVI vs. GIMMS and MODIS NDVI comparison of absolute values, the MODIS NDVI performed slightly better than GIMMS NDVI, whereas in the comparison of temporal change values, the GIMMS data set performed best. Similar with comparison results of GIMMS a
出处
《应用生态学报》
CAS
CSCD
北大核心
2014年第2期533-544,共12页
Chinese Journal of Applied Ecology
基金
国家自然科学基金项目(41001055)
中央级公益性科研院所基本科研业务专项(2012-YSKY-13)资助
