Logistic函数方法拟合多时序NDVI数据的改进研究

Improved Logistic Model for Fitting Time-series NDVI Data

单Logistic函数曲线拟合法是NDVI时间序列重建及物候遥感中关键物候期划分的重要方法之一。虽然该方法不需要设定阈值或经验系数、较适应于不同环境区域的物候遥感监测,但是在山区NDVI噪音较大的情形下,其拟合精度仍会受到较大影响。选取秦岭中部山区为研究区,在分析了多年NDVI时间序列数据特征基础上,利用山区NDVI数据序列最大值相对于最小值更为稳定的特性,对传统单Logistic模型求解方法进行改进,采用更为稳定的参数构建模型以提高NDVI时间序列重建的精度。基于秦岭样区MODIS NDVI多年遥感数据,分别在保持植被生长季特征能力和保留高质量原始真值程度两方面对原方法与改进方法的计算结果进行比较。研究表明改进的方法在上述两个方面都具有更好的效果。在植被指数噪音较大的山区,改进的方法对NDVI时间序列重建表现出更好的适用性,可为复杂的山区物候遥感相关研究提供参考。

The single Logistic function curve fitting is one of the most important methods of the NDVI time series reconstruction and key phenophase division in phenology remote sensing. The method is more adap- ted to complex environment as it does not need to set the threshold or empirical coefficient. However, in some mountain areas,the fitting precision of single logistic function is still low if the NDVI values are con- taminated too much. In this paper, using the NDVI time series data of the sampling area in the Qinling Mountains from 2001 to 2013 to analysis the stability of the maximums and the minimums of the NDVI curves, we found that the maximums are more stable in the NDVI time series data than the minimums, and then we modified the form of the single logistic model on the basis of the above analysis. Finally, a more stable method was proposed to improve the accuracy of the NDVI time series reconstruction with large noise in complex mountains. In this paper,in order to measure the ability to keep vegetation growth season features and retain high quality original true values of the modified method and original method, the means of correlation coefficient and root mean squared error of the two methods for fitting NDVI data of the sam- pling area in the Qinling Mountains from 2011 to 2013 were calculated. And then the two indices of the modified method were evaluated by comparing with the original one. The results show that the modified method has a better performance than the original one in both above aspects. The modified method has a better applicability in the NDVI time series reconstruction with large noise,which can provide reference for the research related phenology remote sensing in complex mountains.