Fire-affected forest structure and fire severity estimation using satellite-based and airborne LiDAR data and a Neural Network model

This thesis explores how Deep Learning model and satellite LiDAR, GEDI can improve understanding and management of wildfire-affected forests. Focusing on Australia's 2019-2020 "Black Summer" fires, it uses GEDI data to measure forest structure before and after fires. A Convolutional Neural Network (CNN) significantly improved data accuracy, especially in burned areas. Machine learning models also identified key forest features, like foliage density, to predict fire severity. The research highlights the potential of combining remote sensing data with Deep Learning model to enhance fire damage assessment, guide forest restoration, and support more effective wildfire management strategies.