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

Area of Science:

• Environmental Science
• Remote Sensing
• Data Analysis

Background:

• Non-stationary Normalized Difference Vegetation Index (NDVI) time series (TS) pose challenges for accurate trend analysis.
• Wavelet Transform (WT) offers potential for multi-resolution analysis (MRA-WT) of complex time series data.

Purpose of the Study:

• To improve trend analysis for non-stationary NDVI time series in Tunisia.
• To optimize parameter selection (decomposition level and mother wavelet) for MRA-WT.
• To investigate the influence of meteorological data on vegetation trends.

Main Methods:

• Applied Multi-Resolution Analysis Wavelet Transform (MRA-WT) to decompose NDVI time series.
• Investigated decomposition levels and selected optimal mother wavelets (MWs) using the energy to Shannon entropy ratio criterion.
• Utilized Mann-Kendall (MK) test and Sen's slope estimator for trend analysis on the approximation component.
• Analyzed the impact of meteorological data on identified vegetation trends.

Main Results:

• Identified Level 5 as the optimal decomposition level for long-term vegetation change analysis.
• Daubechies (db9) and Symlets (sym4) mother wavelets demonstrated the highest energy to entropy ratio.
• A combined MW approach was proposed for image-level trend analysis.
• Observed forest area degradation and increases in cropland and vegetation cover in Tunisia from 2001-2017.

Conclusions:

• Mother wavelet selection is crucial for accurate NDVI time series trend analysis.
• The optimized MRA-WT method provides reliable insights into vegetation dynamics.
• Results indicate significant land cover changes, highlighting the need for targeted environmental management.