Application of Singular Spectrum Analysis (SSA)
for the Reconstruction of Annual Phenological
Profiles of NDVI Time Series Data
Jose Edgardo L. ABAN
Department of Science and Technology
Gen. Santos Ave., Bicutan Tagig, Metro Manila, Philippines
Tel: + 632-837-2071 to 80 loc. 2100-2109 Fax:+632-837-3168
Email: jelaban@dost.gov.ph
Ryutaro TATEISHI
Center for Environmental Remote Sensing (CEReS)
Chiba University
1-33 Yayoi, Inage, Chiba-ken, Japan
Tel: +81-43-290-3965 Fax: +81-43-290-3857
JAPAN
Email: aban@ceres.cr.chiba-u.ac.jp
ABSTRACT
The Singular Spectrum Analysis (SSA) is proposed as an alternative novel technique used to
reduce noise in the Normalized Difference Vegetation Index (NDVI) time-series annual
phenology profile from the Advanced Very High Resolution Radiometer (AVHRR) data. The
SSA preserves more valuable elements of the NDVI profile based on data-adaptive basis
functions. This paper describes a detailed investigation into the suitability of the SSA for NDVI
profile reconstruction. The ability of the SSA to decompose the time series into several
principal components, distinguish between the main signal and noise, and subsequently
reconstruct the original time series through the use of reconstructed components (RC) are
demonstrated in this study.
INTRODUCTION
Since 1982, the National Oceanic and Atmospheric Administration’s (NOAA) Advanced Very
High Resolution Radiometers (AVHRR) has provided daily Earth observations. From these
data, NDVI profiles have been built and compiled through the years. Such multi-annual
profiles data have been used to monitor primary production (Prince and Tucker, 1986; Fung et
al., 1987), to study the dynamics of major biomes (Malingreau, 1986), for land cover
classification (Townshend et al., 1991), and for estimation of crop yield (Bartholome’, 1991).
Changes in the NDVI derived from AVHRR data are usually indicative of changes in the
surface conditions, most predominantly changes in vegetation. However, there are other
extrinsic factors that cause changes in the overall NDVI profile, among which are cloud
contamination, atmospheric variability and bi-directional effects (Gutman, 1991; Tucker and
Sellers, 1986; Justice et al., 1991; Soufflet et al., 1991; Simpson and Stitt, 1998); these changes
are usually considered as undesirable noise in vegetation studies. A widely used method to
reduce this noise is the Maximum Value Composite (MVC) technique proposed by Holben
(1986) and Prince and Justice (1991). In spite of corrections, the data still contain unwanted
variation or “noise” (Gutman, 1991; Tucker and Sellers, 1986; Justice et al., 1991; Soufflet et
al., 1991; Simpson and Stitt, 1998).
In this study we explore a relatively new technique called the Singular Spectrum Analysis (SSA).
The SSA is similar to an extensively used technique of PCA, whereby it undertakes a linear
transformation of a set of image bands to create a new band set with images that are
uncorrelated and are ordered in terms of the amount of variance explained in the original data
(Johnston, 1980; Mather, 1987, ). whereby the data is decomposed into a set of uncorrelated
eigenvectors. The SSA was developed independently by Broomhead and King (1986a, 1986b)
and Danilov and Zhigljavsky (1997). SSA is a nonparametric method and assumes no specific
model of the process being analyzed. Unlike previous methods of NDVI profile extraction, it
uses data-adaptive basis functions, the lag-covariance matrix, rather than threshold limits, or the
pre-determined measures of central tendency (i.e. mean, median, or mode), which are basically
non-adaptive techniques. Though numerous studies have employed the SSA technique in a
variety of environmental as well as econometric applications, the method has never been applied
on the derivation of the NDVI profile, for which this present study aims to demonstrate.
