Evaluation of Vegetation Conditions for Green Legacy Using Geospatial Technology: A Case of Entoto Natural Park, Addis Ababa, Ethiopia

Monitoring vegetation condition is essential for ecological sustainability, restoration planning, and climate change adaptation, particularly in urban-adjacent conservation areas such as Entoto Natural Park in Addis Ababa, Ethiopia. However, vegetation condition assessments in the park have been limited and lack quantitative evidence based on geospatial approaches. This study evaluates natural vegetation conditions using multispectral remote sensing, spectral indices, and a Random Forest machine learning model. Landsat imagery from 1995, 2005, 2015, and 2025 was processed to generate NDVI, GNDVI, and NDWI indices, which were used to classify vegetation health and analyze temporal trends. The Random Forest classifier was trained using field-based reference samples and validated using out-of-bag accuracy metrics. Results indicate a general improvement in vegetation condition between 1995 and 2025, with higher chlorophyll content and water availability in recently rehabilitated areas, while eucalyptus-dominated zones exhibited comparatively lower moisture and greenness values. The prediction model also forecasted future vegetation conditions, suggesting continued improvement under ongoing restoration programs. This study demonstrates the effectiveness of spectral indices combined with machine learning for vegetation condition monitoring and provides a geospatial foundation to support sustainable management and restoration efforts under Ethiopia’s Green Legacy Initiative.