基于TIMESAT的3种时序NDVI拟合方法比较研究——以藏北草地为例

Comparison of Three NDVI Time-series Fitting Methods based on TIMESAT——Taking the Grassland in Northern Tibet as Case

以藏北地区2007-2009年MODIS 16 d合成的NDVI时间序列为例,介绍了基于TI MESAT2.3软件的3种主要拟合算法——非对称高斯函数（AG）拟合、双Logistic曲线（D-L）拟合和Savitzky-Golay（S-G）滤波法的基本原理和实现流程;重点从拟合重建NDVI时间序列对原始ND-VI值上包络线的拟合效果及保持原始高质量NDVI点值真实值的程度两个方面,分析比较3种算法的特点。结果表明：①3种拟合算法均能不同程度提高整个区域的NDVI平均值,AG与D-L拟合法处理后的NDVI时间序列与原始NDVI曲线的整体特征较S-G滤波方法更加吻合;②AG与D-L拟合重建的NDVI时间曲线在生长季峰期高于上包络线,S-G滤波法处理结果低于上包络线,3种方法中AG拟合结果与上包络线最为接近;③在保持原始高质量NDVI值真实性方面,AG与D-L拟合法处理结果相似,除生长季曲线的峰期外,均优于Savitzky-Golay滤波法。该研究结论为基于NDVI时间序列进行陆地系统生态环境各方面研究中数据去噪预处理的方法选择提供参考。

MODIS 16 days composited NDVI time-series of 2007-2009 in northern Tibet are taken as study case to compare the characters of three principal fitting methods,that is the double Logistic function fitting（D-L）,asymmetric Guassian function fitting（AG）,and Savitzky-Golay filtering（S-G） methods.To begin first,the basic principles and implementation process of the three algorithms are introducted based the TIMESAT 2.3 program.Then,the results of NDVI time-series fitting based on the three methods are emphatically compared and analyzed,from the aspects of fitting effect for the upper envelope curve of original NDVI series and the ability of preserving the high-quality NDVI fidelity.The results show that the three fitting methods would raise the mean value of NDVI samples to some degree,and that the AG and D-L fitting algorithms generate more consistently reconstructed NDVI time-series to the original NDVI temporal curve than the S-G filtering method.Secondly,the fitting NDVI values of AG and D-L methods are higher than points of the upper envelope curve,the S-G filtering method is opposite.Among the three algorithms,AG fitting produce the most approximative results.Besides,the AG and D-L methods perform extremely similarity to keep the fidelity of high-quality NDVI samples,and their fitting NDVI series are better than that of S-G filtering except the peak period of growing season.