An improved trend vegetation analysis for non-stationary NDVI time series based on wavelet transform.

The aim of this paper is to improve trend analysis for non-stationary Normalized Difference Vegetation Index (NDVI) time series (TS) over different areas in Tunisia based on the wavelet transform (WT) multi-resolution analysis (MRA-WT), statistical test, and meteorological data. The MRA-WT was applied in order to decompose the TS into different components. However, the most challenge for TS analysis using MRA-WT laid in the selection of two optimum parameters: the level of decomposition and mother wavelet (MW). In this work, both factors were investigated. Firstly, the level of decomposition was calculated for 18 different MWs, and secondly the energy to Shannon entropy ratio criterion was investigated to choose the most suitable MW. The Mann-Kendall test (MK) and Sen's slope were applied to the last approximation component in order to analyze long-term vegetation changes. Finally, the influence of meteorological data for trend was analyzed. The results were first computed for different sites in Tunisia using MODIS NDVI TS from 2001 to 2017. The obtained results proved the importance of MW selection. Level 5 was considered for the TS as the best level of decomposition for long-term vegetation changes. The Daubechies and Symlets MWs (db9 and sym4) showed the highest energy to entropy ratio for three selected vegetation canopies. A combination of the two MW was proposed to derive a trend vegetation analysis at image level. A degradation in the forest area and a few increases in cropland and vegetation area were presented.