大量研究利用PathfinderAVHRR NDVI资料分析植被状况与气温、降水等气候要素之间的关系。许多分析指出Pathfinder资料包含误差 ,并分析这些资料误差对大尺度NDVI 气温耦合关系检测结果的影响。利用奇异值分解方法 (SVD) ,通过比较不同NDVI资料误差情况下北半球春季NDVI对气温变化响应的时空特征的差异 ,对资料误差造成的分析结果的可靠性进行判断。考虑了 4种误差形式 ,分别是不同强度的连续误差、不连续误差、强火山喷发造成的误差及趋势误差。分析结果表明 ,利用SVD分析大尺度的NDVI 气温耦合特征时 ,允许的NDVI资料误差的最大上限阈值大致在 0 5σ左右。PathfinderAVHRR NDVI原始资料包含的误差很可能低于此阈值 ,得到的分析结果有较高的可信度。此外 ,在不知道NDVI原始资料误差的情况下进行植被对气候变化响应的检测时 ,可以借鉴此方法对结果的可靠性进行检查和验证。

The pathfinder AVHRR NDVI data are widely used for analyzing the response of vegetation condition to climate change. However, the Pathfinder NDVI data sets contain a lot of random errors due to various factors such as cloud, aerosols, instrumental interruption and changes and so on. To what extent these errors can influence the results of NDVI/climate connections? In the present work,the authors analyzed the impacts of NDVI errors on NDVI/temperature coupling in spring, by using singular value composition analysis technique. Four kinds of errors are tested, including the continuous, discontinuous and volcanic errors, and the error trends as well. Results show that when the errors are less than half the original NDVI standard deviation the SVD spatial modes and temporal changes are stable and independent of the error. However, if the error gets larger, the results would be distorted apparently. Comparison experiments conducted in the present study suggest that the upper limit of potential errors for reasonable results is about 0 5 standard deviations of the original NDVI series. Actual errors are likely to be within this limit, thus the results from the original data are acceptable. In addition, the methods applied in the study can be applied to the similar research to test the significance and robustness of the results in case of unknowing how large the error is.