Student information
MSc thesis topic: Prediction of forest parameters by linking SAR satellite observations and national forest inventory data
Accurate and timely information on the state of European forests is of great importance for the provision of wood, carbon sequestration and biodiversity conservation. Currently, this information is largely obtained via national forest inventories (NFI’s). These forest plot networks are (re-)measured every 5-10 years. Although these provide very accurate and thematically rich measurements, the large time intervals, course spatial resolution, and methodological inconsistency between countries limits their applicability when faced with increasingly demanding and complex European policy requirements and rapid changes in the forest landscape.
The Sentinel-1 platform, a C-band SAR satellite constellation, provides wall-to-wall coverage at 10m spatial detail, with revisit times of up to 3 days over Europe since 2015. In addition, annual L-band SAR mosaics from ALOS PALSAR-1/2 are available openly. Unlike optical systems, radar is able to penetrate clouds, and is directly sensitive to forest structure and its moisture content. Radar is therefore a potentially invaluable source of information on the state of European forests.
This thesis will combine thematically rich information from NFI’s and temporally dense and high-resolution Sentinel-1 data (and potentially other radar sources such as ALOS PALSAR-1/2) to predict and map important forest parameters such as growing stock and stand density. Spatial modelling and machine learning methods will be used to ‘fill the gaps’ between the coarse spatial resolution of an NFI grid. The modelling accuracy will be assessed using spatial cross-validation.
Software: [Google Earth Engine], R/python, ArcGIS/QGIS
Objectives and Research questions
- Predict wall-to-wall coverage of forest parameters using spatial modelling and machine learning methods by combining radar-based satellite data and NFI information.
- Asses the modelling accuracy using spatial cross-validation.
Requirements
- Geo-scripting course (required)
- Advanced Earth Observation (required)
- Spatial modelling and statistics (optional)
Literature and information
- Silveira et al. 2023. Nationwide native forest structure maps for Argentina based on forest inventory data, SAR Sentinel-1 and vegetation metrics from Sentinel-2 imagery, Remote Sensing of Environment, Volume 285, 113391.
- SAR Forest Monitoring handbook
Theme(s): Sensing & measuring; Modelling & visualisation