Assessing vegetation change using satellite imagery analysis

Abstract

This study analysed vegetation changes in the Miombo woodlands of Central Zambia using Sentinel-2 satellite imagery and vegetation indices NDVI and EVI across two time periods: 2017-2018 and 2022-2023. The objective was to assess spatial vegetation dynamics, classify vegetation cover, and evaluate classification accuracy through both threshold-based and supervised machine learning approaches. Ground truth samples were manually collected and utilized for validation. Results revealed significant shifts in vegetation classes over time, with Random Forest classification providing improved accuracy over threshold-based methods. The findings highlight the utility of remote sensing techniques combined with Google Earth Engine for monitoring vegetation trends in Miombo ecosystems, offering valuable insights for sustainable land use planning and ecological assessment in Central Zambia.