基于质量权重的Savitzky-Golay时间序列滤波方法

归一化植被指数(NDVI)时间序列数据因含有大量噪声,给其应用带来诸多不便,甚至产生错误结果。自适应Savitzky\|Golay滤波器能够有效地抑制突降噪声,但在对高值噪声的抑制和突降非噪声数据的保护方面存在不足。将MODIS VI产品中的质量因子作为权重,提出基于质量权重的Savitzky\|Golay滤波方法,经验证该方法能够保持高质量NDVI数据的稳定性和相关性,并能够有效抑制噪声的影响。

VI-Quality-Based Savitzky-Golay Method for Filtering Time Series Data

NDVI time-series data contain disturbances that limit their use and even yield false results.Although the adaptive Savitzky-Golay method could effectively filter some sudden fall - noisy data which is assumed traditionally to be contaminated by clouds or poor atmosphere conditions,it cannot preserve some sudden fall data with good quality,and cannot suppress the sudden rise noisy data.Although maximum NDVI values greatly reduce clouds and aerosols,the highest NDVI value does not necessarily correspond to small sensor viewing angles or to the least-contaminated measurement.This paper presents a VI-quality-weighted Savitzky-Golay method which is based on the Savitzky-Golay filter and weighted by VI qualities derived from MODIS VI product.The results illustrate that the quality-weighted methods could filter more noises,especially sudden rise noisy data,effectively preserve high-quality data and meanwhile do not sensibly elevate the values of the whole time-series.It can appropriately fit high quality data among serious fluctuations and better reconstructs wave crests compared with the traditional Distance-weighted Savitzky-Golay method.Statistically,the proposed method here has the following characteristics:(1) it has lower mean variation (or less shift) effect on original NDVI data;(2) it stabilizes high quality NDVI data;and (3) the resulting high quality data are better correlated with original good data,meanwhile the original noise are greatly decorrelated.