Spatiotemporal modeling of vegetation dynamics in a changing environment: combining Earth observation and machine learning

Samenvatting

Vegetation, including forests and grasslands, is essential for Earth's environmental health, offering services like carbon sequestration, soil stabilization, and water regulation. Traditional forest monitoring methods, such as field surveys and National Forest Inventories (NFI), face limitations in scalability and responsiveness to dynamic ecological changes. The integration of Earth observation data and machine learning has revolutionized forest monitoring, providing comprehensive and high-resolution data. Satellites offer extensive spatial coverage, while machine learning enhances data processing and analysis, identifying patterns and changes in forest landscapes. This thesis aims to integrate these technologies with traditional methods to improve vegetation dynamics understanding and forest ecosystem monitoring. Four research questions guide the work, focusing on climate change impacts, forest tree species distribution, trends in forest disturbances, and the effect of coordinate precision in NFI data on species classification accuracy. The thesis emphasizes creating robust, accurate, and timely forest monitoring systems for effective management and conservation.