Evaluation of consistency among three NDVI products applied to High Mountain Asia in 2000–2015
Evaluation of consistency among three NDVI products applied to High Mountain Asia in 2000–2015
复制标题
DOI：
10.1016/j.rse.2021.112821
复制
发表时间：
2022-02
影响因子：
13.5
通讯作者：
Yongchang Liu;Zhi Li;Yaning Chen;Yupeng Li;Hongwei Li;Qianqian Xia;P. M. Kayumba
中科院分区：
文献类型：
--
作者：
Yongchang Liu;Zhi Li;Yaning Chen;Yupeng Li;Hongwei Li;Qianqian Xia;P. M. Kayumba
目前的研究评估了三个归一化差异植被指数（NDVI）数据集，即GIMMS，MODIS和SPOT之间的一致性，以表征2001-2015年亚洲高山（HMA）的高山植被动态（绿化和布朗宁）。这些数据集的效用进行了探索，以评估植被的变化在不同的时空尺度和海拔高度，并比较其空间趋势和分布格局。除了皮尔逊相关系数进行定量分析的一致性和不一致性的每个数据集，一个NDVI质量控制（QC）层和陆地卫星NDVI也被用来评估的结果。结果表明，GIMMS的NDVI均值最高，SPOT的最低。然而，GIMMS也显示出天山和青藏高原的布朗宁趋势，速率为-0.3 × 10− 3/a，而MODIS和SPOT则显示出绿化趋势（TSMODIS= 0.5 × 10− 3/a，TSSPOT= 0.6 × 10− 3/a，TPMODIS= 0.9 × 10− 3/a，TPSPOT= 1.6 × 10− 3/a）。此外，MODIS-SPOT显示出最高的相关性（RGREEN= 0.73; RBROWN= 0.47），其次是MODIS-GIMMS和GIMMS-SPOT。总体而言，NDVI趋势的一致性似乎是较高的TS。一致性绿变像素主要分布在TP中部向东北部延伸的区域和TP西部向TS东部延伸的区域，占32.14%，一致性布朗宁像素集中在TP西南部和TS中部，占8.32%。不一致像素占59.54%，其中39.21%的不一致绿变像素广泛分布于HMA，20.58%的不一致布朗宁像素在TS中部和TP南部相对明显。本研究为后续植被动态研究中数据的选择和重建提供了基准推断。
The current study evaluates consistency among three Normalized Difference Vegetation Index (NDVI) datasets, namely GIMMS, MODIS and SPOT, to characterize alpine vegetation dynamics (greening and browning) across High Mountain Asia (HMA) in 2001–2015. The utility of these datasets is explored to evaluate the vegetation's variability at different spatial-temporal scales and, elevation, and to compare their spatial trends and distribution patterns. In addition to the Pearson correlation coefficients performed to quantitatively analyze the consistency and inconsistency of each dataset, an NDVI quality control (QC) layer and Landsat NDVI are also used to evaluate the findings. The results indicate that the GIMMS has the highest NDVI mean, while SPOT has the lowest. However, GIMMS also showed a browning trend for both Tianshan (TS) and the Qinghai Tibet Plateau (TP) at a rate of −0.3 × 10−3per year, whereas MODIS and SPOT exhibit a greening trend (TSMODIS= 0.5 × 10−3per year, TSSPOT= 0.6 × 10−3per year, TPMODIS= 0.9 × 10−3per year, TPSPOT= 1.6 × 10−3per year). Furthermore, MODIS-SPOT shows the highest correlation (RGREEN= 0.73; RBROWN= 0.47), followed by MODIS-GIMMS, and GIMMS-SPOT. The overall, NDVI trend consistency appears to be higher in TS. Finally, the consistent greening pixels mainly distributed in central TP stretching to the northeastern part, and in western stretching to eastern TS, account for 32.14%, while 8.32% of consistent browning pixels are concentrated in southwestern TP and central TS. The inconsistent pixels account for 59.54%, with 39.21% of inconsistent greening pixels being widely distributed across HMA, and 20.58% of inconsistent browning pixels being relatively pronounced in central TS and southern TP. This study provides baseline inferences for the selection and reconstruction of data in follow-up studies on vegetation dynamics.