Unterstützung 2025
Carla Felder | Institute of Geological Sciences | University of Bern
Pioneering Cosmogenic 39Ar Surface Exposure Dating: The AD 1717 rock avalanche in the Ferret Valley (Mont Blanc Massif)
In the Ferret Valley (Mont Blanc Massif, Italy), a massive historical rock avalanche in AD 1717 deposited chaotic boulders and irregular ridges along the valley floor. These deposits, spanning a 2 km-long area, have been extensively studied using radiocarbon and cosmogenic surface exposure dating techniques. This historical rock avalanche provides a unique natural laboratory to explore the potential of surface-exposure dating with cosmogenic 39Ar (half-life 269 years), a novel application of cosmogenic nuclides to key problems in the Earth sciences.
My MSc thesis focuses on the establishment of the cosmogenic 39Ar surface exposure dating. As a first step, I will calibrate the 39Ar production rate by comparing the 39Ar concentrations from both whole-rock and feldspar samples with the known age of the rock avalanche and existing 10Be data.
During my fieldwork in autumn 2025, I carried out detailed geomorphological mapping of the avalanche deposits and collected samples for 39Ar exposure dating. Sample preparation will be carried out in Bern, and the cosmogenic 39Ar analysis will take place at the University of Heidelberg (Germany).
The MSc project is supervised by Prof. Dr. Naki Akçar (University of Bern) and co-supervised Prof. Dr. Werner Aeschbach (University of Heidelberg), Dr. Philip Deline (Université Savoie Mont-Blanc), and Prof. Dr. Susan Ivy-Ochs (ETH Zurich).
I’m grateful to CH-QUAT for supporting my research by covering travel and accommodation costs during my fieldwork in the Ferret Valley and for making the laboratory visit to Heidelberg possible.
Contact : carla.felder@students.unibe.ch
Robin von Allmen | Geoecology | Department of Environmental Sciences | University of Basel
Battling obstacles towards fully automated pollen analysis using Machine Learning and archaeological reference collections.
Following the first successful applications of modern machine learning models to both modern and fossil pollen assemblages, automated palynology has recently gained renewed attention. Despite this progress, several critical challenges remain unresolved. In particular, the collection of sufficient training images for fossil pollen identification and the transfer of established machine learning workflow applications to other domains are still underdeveloped. As part of an effort to preserve the extensive pollen reference collection at Uppsala University, comprising approximately 1,200 specimen slides, I digitized the most palynologically relevant taxa. I further complemented this dataset by collecting fresh reference samples of corresponding species across the Swiss mountain range and by assembling a preliminary set of pollen images from a late Holocene ice-core assemblage. Using this combined image material, I trained machine learning models on fossil, reference, and modern datasets and conducted cross-validation to assess how well each model performed both within and across domains.
I presented the results of this pilot project at SEB25 in Antwerpen, Belgium. The findings demonstrated that while each model performed well on images belonging to its own training domain, even on previously unseen samples, accuracy dropped substantially when models were applied to images from different domains. Notably, the model trained on the reference collection achieved the most consistent accuracy across all datasets, whereas the model trained exclusively on modern samples performed very well on modern images but showed the weakest cross-domain generalizability. On another note, a short transfer learning phase with a minimal dataset was sufficient to elevate cross-domain accuracy to similar levels as within-domain accuracy.
I am deeply grateful to CH-QUAT for the opportunity to travel to Antwerpen and share these results at SEB25. Presenting this work allowed me to highlight the ongoing challenge of AI adaptability in palynology and helped to initiate the discussion with colleagues about the need to digitize additional reference collections to develop more robust, generalizable AI models in order to facilitate the collection of fossil and modern image data. The network I built at SEB25 will be invaluable as I progress in my Early Career, fostering interdisciplinary collaboration and potentially supporting future funding efforts and projects.
Contact: robin.vonallmen@unibas.ch
Dario Rothen | Institute of Geological Sciences | University of Bern
Msc Thesis on the glaciation history of Val Medel
Since the Last Glacial Medel (LGM), the glaciers in the side valleys of the Val Medel have retreated significantly. This study aims to investigate these local glacial systems in two side valleys of the Val Medel (Val Buora and Val Cristallina) in more detail and subsequently model their retreat. To this end, mapping work was carried out this summer in the Val Buora and Val Cristallina areas, and samples of glacial blocks were taken from a section of the Val Buora for dating purposes. This dating is based on the cosmogenic nuclide Be-10. It may also be possible to assess the response of the local glacial systems to large-scale climate fluctuations and thus potentially to better predict their changes in the context of anthropogenic climate change. Furthermore, this study can improve our understanding of the glacial history of the Val Medel/Surselvas region.
I am grateful for the generous support from CH-QUAT, which financed the majority of my fieldwork. Returning to the field several times was crucial to take the necessary samples for dating and to create a detailed field map.
Thank you, CH-QUAT!
Contact: dario.rothen@students.unibe.ch
Maria Grozeva | Geology | ETH Zurich
CH-QUAT support for research visit to University of Aarhus, Denmark
Reconstructing the burial history of glacially deposited sediments in overdeepened valleys remains a central challenge in Quaternary geology. A range of absolute dating techniques—such as luminescence dating, cosmogenic nuclide dating, and helium dating (He/U-Th in pore water)—along with relative methods like the analysis of sedimentary successions, can be used to address this problem.
For my Master’s thesis, I am evaluating the applicability of the CosmoChron age–depth model to cosmogenic ²⁶Al/¹⁰Be data obtained from deep drill-core sediments from the Strassberg Trough in Bülach, northern Switzerland. CosmoChron is a versatile modeling framework that integrates cosmogenic-nuclide concentrations with independent age constraints to reconstruct sediment accumulation and burial histories.
In October 2025, I visited the University of Aarhus to collaborate with some of the developers of CosmoChron. During this visit, I received hands-on training in the model’s implementation and explored how varying initial conditions—such as different pre-burial histories and existing age tie-points (e.g., luminescence minimum ages and helium dating)—influence modeled outcomes and associated uncertainties.
Support from CH-QUAT made this collaboration possible, enabling me to gain direct experience with the model and to strengthen the analytical foundation of my thesis through international scientific exchange.
Contact: mgrozeva@student.ethz.ch
Kannika Wangritthikraikul | Eawag and Department of Earth and Planetary Sciences | ETH Zurich
I am grateful for the support from CH-QUAT, which enabled me to attend and present a poster at the 7th PAGES Open Science Meeting (OSM) in Shanghai, China. This opportunity also allowed me to learn about several interesting ongoing projects from around the world, such as the Human Traces project and the SISAL (Speleothem Isotopes Synthesis and AnaLysis) project. Meeting various experts at the conference provided me with many useful ideas and suggestions for my work through discussions during my poster presentation. This experience was a valuable moment not only in terms of academic growth but also for building connections with researchers in the same field globally, which could benefit potential collaborations in the future. My project focuses on multi-proxy paleoenvironmental reconstruction from the Noboru Umeda peat swamp in eastern Thailand. Situated in mainland Southeast Asia, Thailand lies at the intersection of several sub-monsoonal systems, where monsoon-derived rainfall strongly influences hydrological dynamics and sediment deposition. Peat swamps in this region are highly sensitive to climate variability and serve as important carbon sinks, making them valuable archives for understanding past environmental and climatic changes. This study employs a multiproxy approach—including lipid biomarker analysis, total organic carbon (TOC) content, bulk δ¹³C, carbon-to-nitrogen (C/N) ratios, X-ray fluorescence (XRF) core scanning, and radiocarbon dating—to examine sediment sequences from the Noboru Umeda peat swamp in Trat Province. Radiocarbon dating indicates peat accumulation began around 7700 cal yr BP. A significant increase in TOC and bromine within the top 70 cm of sediment (dated to ~2020 cal yr BP) reflects favorable conditions for peat preservation, such as low depositional energy and high organic matter input. Concurrent decreases in titanium and potassium suggest reduced minerogenic input and potentially weaker summer monsoon intensity. Lipid biomarker data reveal vegetation shifts, highlighting the swamp’s response to Holocene climatic variability.
Contact: Kannika.Wangritthikraikul@eawag.ch