Seminar talk by: Benjamin Robson, University of Bergen:
Automated Analysis of New Zealand Historical Aerial Photographs: Geodetic Glacier Change Detection at Rolleston and Tasman Glaciers
Abstract: Despite New Zealand having over 2500 glaciers, only two are measured routinely in the field making it hard to get an overview of how the cryosphere is responding to climate change. New Zealand has an extensive archive of historical aerial photographs that allow changes to be tracked back to the 1940s, however processing these datasets is a time consuming and partially manual process. In this talk I will show preliminary results from using AI image matchers to automate processing of entire historical surveys and look at glacier changes. I will go through the workflow we have developed and show some early results for two glaciers in New Zealand - Tasman Glacier, the largest in the country, and Rollerston glacier, one of the two glaciers that is actively monitored in the field.
Benjamin Robson is an Associate Professor in Geomatics at the Department of Earth Science, University of Bergen. His research is focused on looking at glacial and periglacial changes using remote sensing, in particular combining aerial, satellite, and UAV based datasets.
Seminar talk by: Sunil Tamang, University of Canterbury, New Zealand:
Accuracy and Ambiguity: Assessing the Influence of Methodological Choices on Automated Rock Glacier Mapping
Abstract: Rock glaciers constitute a critical component of the cryosphere and serve as prominent indicators of permafrost presence in mountainous regions. Accurate and consistent delineation of their spatial extent is essential for advancing research in geomorphology, hydrology, biodiversity, geohazard assessment, and permafrost dynamics. However, mapping rock glaciers remains challenging for two primary reasons. First, they frequently co-occur with or transition into geomorphologically similar landforms that are difficult to distinguish based solely on spectral characteristics in aerial or satellite imagery. Second, their boundaries are inherently ambiguous and dynamic, influenced by evolving topographic and climatic conditions. A conventional approach based on manual digitisation through visual interpretation of geomorphological features are both time-intensive and subjective, thereby limiting scalability and consistency across large spatial extents. Recent advances in geospatial technologies demonstrate the potential of deep learning techniques to automate rock glacier detection and delineation. Nevertheless, existing studies have largely overlooked the quantification and analysis of uncertainty associated with these methods. In this context, the present study investigates how different methodological choices in deep learning workflows influence the accuracy of rock glacier detection and extent delineation. Specifically, it examines key factors such as input data selection, training data definition, sampling strategy, and model architecture, using the Coquimbo Region of Chile as a case study. By systematically evaluating these factors, the study provides insights into the sources of uncertainty in automated rock glacier mapping and contributes toward developing more reliable, transparent, and scalable mapping framework.
Sunil Tamang is a PhD candidate at the School of Earth and Environment, University of Canterbury, New Zealand, and is currently undertaking a research exchange at the University of Bergen.
-------
The GeoHyd seminars are normally held in Auditorium 1 in the Geology Building, Blindern (map) but it is also possible to follow some of the seminars on Zoom.
Videolink to the lectures will be sent to the e-mail list geohyd-alle@geo and geohyd-info@geo. If you want to subscribe to the geohyd-info@geo mailing list send an email to Anita Sørlie.
About the seminar series
The talks in GeoHyd Lunch Seminar Series are offered by the Section for Geography and Hydrology (GeoHyd), Dept. of Geosciences, University of Oslo. The seminars are announced as lunch seminars so bring your lunch if you want to.
The seminars are open for everyone interested, and especially students are welcome.
The Lunch Seminar Team
– Regine, Andi and Anita