MSc thesis topic: Can we make reliable wall-to-wall Dutch forest structure maps with Machine Learning methods?
The Dutch Forest Inventory (NFI) delivers reliable information about the state and development of the forest in the Netherlands and fulfils the requirements in terms of reporting about e.g. forest extent and biomass volume. However, collecting data is a time consuming task.
Remote sensing, and LiDAR in particular, is regularly mentioned as potential alternative for the fieldwork, but has so far not been able to convince. However, combining multiple datasources (e.g. AHN-LiDAR derived forest characteristics, multitemporal satellite data) with machine learning techniques might be a good step forward. The current NFI observations can be used for training and validation of the models that will be developed to create a nation-wide forest structure map, containing forest elements as recorded through forest inventories at the moment.
The ultimate goal is to create a remote sensing based forest structure maps of the Netherlands. This is probably a more ambitious goal than can be achieved in one thesis, but let’s make a start!
- Investigate the possibilities to create wall-to-wall forest structure maps based on publicly available LiDAR and satellite data using Machine Learning
- Create forest structure maps of the Netherlands, with known (in)accuracies
- Potentially, create forest change maps
Sensing & measuring; Integrated Land Monitoring