陕北地区GIMMS、SPOT--VGT和MODIS归一化植被指数的差异分析

为了选择适合监测陕北黄土高原地区植被的最佳遥感序列数据,精确监测陕北黄土高原退耕前后的植被变化,选取GIMMS、SPOT-VGT和MODIS3种常用的遥感数据,运用相关分析和均方根误差分析方法,比较3种遥感数据在陕北黄土高原植被空间分布、归一化差异植被指数(NDVI)季节变化和年际变化3个方面的异同。结果表明:1)在植被空间分布方面,GIMMS/NDVI、SPOT-VGT/NDVI和MODIS/NDVI在大范围上的空间分布格局基本一致,但通过分布图分析可以看出,MODIS遥感数据由于其地物分辨率高及NDVI动态范围大的优点,比SPOT-VGT和GIMMS数据更适合于反映植被类型多样的陕北黄土高原地区植被的空间分布。2)在季节变化方面,3种遥感数据NDVI季节变化之间存在极显著的相关关系。其中,均方根误差分析结果表明,MODIS/NDVI与GIMMS/NDVI之间的差异明显大于MODIS/NDVI与SPOT-VGT/NDVI之间的差异;不同季节3种遥感数据NDVI差异也不同,夏季由于云雨较多,3种遥感数据NDVI之间差异最大。3)在年际变化方面,MODIS和SPOT-VGT数据反映出陕北黄土高原地区NDVI在1999—2007年间呈显著增加趋势,而GIMMS/NDVI却未呈现显著变化,说明GIMMS/NDVI在反映陕北高原地区植被年际变化方面存在显著缺陷。通过相关分析可以看出,GIMMS/NDVI和MODIS/NDVI年际变化之间的相关系数随植被覆盖度的升高而降低,尤其在针阔混交林区,其NDVI相关系数甚至为负值,表明GIMMS传感器对高覆盖度植被变化的响应不太敏感,与其他两者相比更易受水气和云的干扰。因此,GIMMS/NDVI不能作为历史均值NDVI直接应用到MODIS应用模型中,尤其在反映高覆盖度植被年际变化方面。

Comparison and validation of GIMMS,SPOT-VGT and MODIS global NDVI products in the Loess Plateau of northern Shaanxi Province,northwestern China

To select an optimum time series in remote sensing data for monitoring vegetation changes, especially before and after implementation of the Sloping Land Conversion Program, we analyzed spatial distributions, seasonal and annual variations of vegetation in the Loess Plateau of Shaanxi Province, based on the normalized difference vegetation index (NDVI) derived from three kinds of remote sensing data, i.e. GIMMS, SPOT-Vegetation (SPOT-VGT) and MODIS. The NDVI data were compared using linear correlation and root mean-square error (RMSE) analyses. The results showed that: 1) there were no significant differences in large scale spatial distributions among the GIMMS/NDVI, the SPOT-VGT/NDVI and the MODIS/NDVI data sets. However, the MODIS/NDVI data was more efficient in monitoring spatial vegetation distributions, with high spatial resolution and a wide range in the NDVI; 2) In seasonal changes of vegetation, significant correlations were found among the three remote sensing data sets. RMSE between the MODIS/NDVI and GIMMS/NDVI was higher than that between the MODIS/NDVI and SPOT-VGT/NDVI. We also found that differences among the three remote sensing data sets varied between seasons, with the largest difference occurring during summers due to more rainy and cloudy days; 3) In annual changes of vegetation, significant increases in both MODIS/NDVI and SPOT-VGT/NDVI were shown from 1999-2007, but no significant variation in GIMMS/NDVI, suggesting that this data set has some limitations in monitoring the annual vegetation changes. The correlation coefficient of annual variation between the MODIS/NDVI and GIMMS/NDVI decreased with an increase in vegetation coverage, and became even negative for coniferous and broad-leaf mixed forests, indicating that the GIMMS sensor was less sensitive to the variation of vegetation with high coverage due to changes of atmospheric water vapor and clouds. We conclude that GIMMS/NDVI has some limitations in representing annual variation of vegetation with high coverage, and is not suitable as a substitution of historical MODIS/NDVI data.