Forest Health Decline
This project aimed to monitor forest health using remote sensing data, focusing on Sentinel-2 satellite imagery from 2015 to 2024.
By calculating the Normalized Difference Vegetation Index (NDVI), we detected changes in vegetation over time that could indicate stress or tree mortality.
A LiDAR-derived forest mask was applied to limit analysis to vegetated areas.
Different anomaly detection approaches were used – such as year-to-year differences, seasonal comparisons, and z-score standardization.
Visual validation was performed using high-resolution orthophotos from 2017, 2019, 2020, and 2023, confirming the observed patterns.
The method is scalable, reproducible, and offers valuable support for forest managers aiming to detect early signs of decline.
While machine learning was not implemented due to time constraints, it is proposed as a future improvement for automated classification of tree health.
Poster du travail de bachelor de Millena Kidane
