两代AVHRR GIMMS NDVI数据集的对比分析——以新疆地区为例

最新发布的1981—2012年的AVHRR GIMMS NDVI3g数据为了解区域植被的近期变化状况提供了数据基础。深入理解该版本与老版本GIMMS NDVIg(1981—2006年)之间的关系,对于使用新数据时充分利用已有老版本的研究结果具有重要意义。以我国西北干旱区的典型区域——新疆为例,研究了两个数据集在反映生长季、春季、夏季和秋季植被现状,植被变化趋势及其对气候变化响应方面的异同。研究结果表明:两个数据集在描述植被活动空间分布、变化趋势及其与气候的相关性方面大体相似,但在数值、动态变化率及其对气候变化响应强度等方面存在的差异也不容忽略。NDVI3g数据生长季和各季节NDVI数值多大于NDVIg,尤其是在夏季和在植被覆盖较好的区域。区域尺度,NDVI3g所反映的植被变化趋势更为平稳,尤其是在夏季和较长的时段,这可能与像元尺度NDVI3g显著增加范围小于NDVIg,而显著减少范围多于NDVIg有关。两个数据集对气温、降水量、潜在蒸散发和湿润指数的响应具有大体一致的空间格局,但对气候因子变化的敏感性存在差异,哪一个数据集更为灵敏依赖于不同的气候因子和时段。一般规律是NDVI3g与热量因子显著正相关的区域小于NDVIg,而与水分因子显著正相关的区域则大于NDVIg。利用长期的生态数据集,尽快理清两个数据集在表征植被变化之间的异同并建立两者的转换关系,对于合理开展植被变化、碳平衡、生态系统服务功能评估等广泛利用NDVI数据的相关研究十分重要。

Comparison of GIMMS NDVI3g and GIMMS NDVIg for monitoring vegetation activity and its responses to climate changes in Xinjiang during 1982-2006

The released third-generation NDVI datasets in 2014, GIMMS NDVI3g, provide a data basis for quantifying recent regional vegetation dynamics over a sufficiently long term. The comparison between the new and old versions (GIMMS NDVIg, from 1981 to 2006) is necessary to link previous studies with future applications of GIMMS NDVI3g in monitoring vegetation activity trends and their responses to climate change. In this study, GIMMS NDVI3g was initially compared with GIMMS NDVIg in an evaluation of spatio-temporal patterns of seasonal vegetation changes in Xinjiang Province, China, at regional and pixel scales during overlapping periods from 1982 to 2006. The influences of climate change (including temperature, precipitation, potential evapotranspiration, and humidity index) on vegetation growth were then analyzed based on GIMMS NDVI3g and GIMMS NDVIg. To better understand the relationships between GIMMS NDVI3g and GIMMS NDVIg, NDVI trends and correlations between NDVI and climatic factors were calculated over multiple nested time series from 18 to 25 starting in 1982. The results indicated that most areas showed an approximate consistency in overall changing trends and correlations with climate variables for both datasets, but differences in many aspects should not be ignored. In most pixels, numerical values of GIMMS NDVI3g were larger than those of GIMMS NDVIg in the growing season, spring, summer, and autumn, particularly in summer, and also in those areas with dense vegetation. At a regional scale, the NDVI trends of GIMMS NDVI3g were smoother than those of GIMMS NDVIg in the growing season and all seasons, particularly in summer and longer periods. At the pixel scale, areas with a significant increase in GIMMS NDVI3g were less than those in GIMMS NDVIg, whereas this was not true in those areas with a significant decrease. The spatial patterns of correlations between GIMMS NDVI3g and four climate variables were approximately similar to those between GIMMS NDVIg and the climate variables, but there were some differences in the sensitivity of both datasets to climate change. Which dataset is more sensitive depends on climate variables and periods. In general, areas with significantly positive correlations between GIMMS NDVI3g and thermal factors were fewer than those of GIMMS NDVIg, whereas positive correlations between NDVI and moisture factors were greater in GIMMS NDVI3g than in GIMMS NDVIg. Integrated other ecological datasets, it is urgent to identify the similarities and differences between the two datasets and to establish a connection between them for reasonably monitoring vegetation dynamics using NDVI datasets.