Automated identification of subglacial geomorphological features
To develop an automated method for identifying Rogen moraine features from high-resolution LiDAR data.
Since investigations in the late 1960s, specific moraine features known as ‘Rogen moraines’ (Lundqvist, 1969) have been identified across many regions covered by ice during the last glacial maximum (LGM). These features typically form in areas thought to have been intemperate subglacial conditions, or in areas of subglacially stored water.
The advent of new, high-resolution datasets such as 1m LiDAR DEMs across Norway, and machine-learning becoming more widely available, offer new opportunities to carry out geomorphological mapping in a consistent manner.
The candidate will work to produce an automated method to identify Rogen moraine features within Norway, which can be tested against manually produced feature maps. Previous methods for snow dune identification can be used as a basis to begin this work. The work can be extended towards morphometric characterization of the identified features.
- What are the optimal methods to automatically recognize moraine features in high-resolution LiDAR data? Is the methodology able to identify features with minimal false positive/negative identifications?
- How does the methodology relate to available manual mapping or field observations of features from prior literature and new fieldwork?
- Is the methodology region specific, or can it be applied to other areas and still produce effective results.
- Python / MATLAB or similar (important)
- Interest/experience in machine learning (important)
- Understanding of Remote Sensing / Geospatial data / GIS
- There is potential for participation in fieldwork during this thesis