不同数据源归一化植被指数在中国北方草原区的应用比较

中国北方草原区生产力在区域碳水循环、农牧业发展中举足轻重。归一化植被指数（Normalized Difference Vegetation Index,NDVI）广泛应用于生产力的计算,然而目前来源众多的NDVI数据反映中国北方草原植被时空动态的一致性仍未可知。本研究利用2000—2015年3个来源NDVI数据集（MODIS NDVI、GIMMS NDVI和SPOT NDVI）并以国际上公认的数据准确性较高的MODIS NDVI为基准对比分析了中国北方草原区NDVI时空动态的一致性,并选取适宜的NDVI产品揭示研究区NDVI长期的时空格局。结果表明：整个中国北方草原区以及部分草原类型（高寒草甸、高寒草原、高寒荒漠、温带荒漠草原）GIMMS NDVI和MODIS NDVI 2套数据集无论是数值范围,还是年际波动和变化趋势具有较高一致性（二者在高寒草甸、高寒草原、高寒荒漠、温带荒漠草原的相关系数分别为0.60、0.47、0.51、0.74）,而SPOT NDVI数值远高于其他2个数据集,尤其是在青藏高原草原区,SPOT NDVI数值每年较另外两套数据集约偏高0.15,表明该区域使用SPOT数据应慎重。部分温带草原类型（典型草原和草甸草原）GIMMS NDVI和SPOT NDVI数据集在年际波动以及变化趋势上具有较高的一致性（相关系数分别为0.85和0.60）,但温带草原区3种数据集NDVI数值范围整体相差不大,小于0.06。基于上述结果,本研究进一步采用时间序列最长且与MODIS NDVI一致性最好的GIMMS NDVI分析了研究区NDVI的时空动态,发现1982—2015年中国北方草原区NDVI整体呈增加趋势,25%的区域达显著水平（p<0.05）,主要集中在温带草原区;高寒草原区NDVI大部分区域变化不显著且有一定比例的区域NDVI呈显著下降趋势。本研究可以为模型数据集选择和预测中国北方草原区植被对未来气候变化的响应提供科学依据。

Assessment of Normalized Difference Vegetation Index from Different Data Sources in Grassland of Northern China

Productivity of grasslands in northern China plays an important role in regional carbon-water cycle and the development of agriculture and husbandry. Normalized Difference Vegetation Index (NDVI) has been widely used as an indicator of net primary productivity. However, it still remains unclear about the consistency among numerous NDVI datasets in characterizing the spatial and temporal dynamics of grasslands in Northern China. In this study, taking MODIS NDVI as a benchmark dataset, three NDVI datasets (MODIS NDVI, GIMMS NDVI, and SPOT NDVI) were used to compare and analyze the spatial and temporal consistency of NDVI in the grassland of northern China from 2000 to 2015. The most suitable NDVI datasets were selected to reveal the spatial and temporal patterns of NDVI in the study area. Our results show that in terms of the inter-annual variability and changing trend, GIMMS NDVI and MODIS NDVI presented high consistency over the entire grassland area, especially in alpine grassland area including alpine meadow, alpine grassland, alpine desert, and part of temperate grassland area (i.e., desert steppe), with correlation coefficients of 0.60, 0.47, 0.51, and 0.74 respectively. While SPOT NDVI values were much higher than those of the other two datasets, especially in alpine grasslands on Qinghai-Tibet Plateau, with a higher NDVI of 0.15 on average, which implied that caution should be taken when using SPOT NDVI to analyze vegetation dynamics or model productivity in alpine grasslands. GIMMS NDVI and SPOT NDVI displayed relatively high consistency in both temporal variability and changing trend in part of typical and meadow steppes, with correlation coefficients of 0.85 and 0.60, respectively, however, all the three NDVI datasets were highly consistent in their variation ranges in this area, with differences of NDVI less than 0.06. Based on GIMMS NDVI datasets, i.e. the one with the longest time series and highest consistency with MODIS NDVI, we further analyzed the spatial and temporal patterns of NDVI in the study area. We found that NDVI increased generally from 1982 to 2015, with 25% of grassland areas (mainly in temperate grassland area) being significant (p<0.05). There was no significant change of NDVI for the entire alpine grassland area though a significant decreasing trend occurred in a small proportion of the region. Our study has implications for model communities to select datasets and provides an advanced understanding of the responses of vegetation to future climate change in the grassland of northern China.