Palynology — the study of organic-walled microfossils, including pollen, spores, and algal remains — plays a vital role in biostratigraphy and in reconstructing past climates and environments. However, conventional palynological analysis relies on specialists manually identifying and counting microfossils under a light microscope, making it a complex and time-intensive process.
This project investigates how artificial intelligence can transform this traditional workflow. You will work with palynological samples (Triassic) that have already been analysed by experienced palynologists, creating an opportunity to compare established expert interpretations with an AI-supported digital approach.
The work will involve digitizing microscopic slides and using “Scampi”, a cloud-based software platform developed by Equinor and the University of Tromsø (Martinsen et al. 2024). “Scampi” is designed to automate and accelerate the identification of palynological species, enabling the automatic recognition and counting of organic-walled microfossils.
A key part of the project will be assessing the quality of the digital data produced by the software. This may include comparing taxonomic identifications, specimen counts, relative abundances, and preservation categories with the existing conventional counts. Particular attention will be given to identifying taxa or specimen types that are difficult for the AI system to recognize, such as poorly preserved, folded, fragmented, or overlapping microfossils.
The student will also gain practical experience in the preparation and handling of palynological slides, digital microscopy, image quality control, and data management. Depending on the project scope and the student’s interests, there may also be opportunities to contribute to the improvement of training datasets by validating or correcting AI-generated identifications.
The resulting datasets can be used to investigate how closely AI-supported assemblages reproduce established biostratigraphic and palaeoenvironmental interpretations. The project will therefore address not only whether automated methods are faster, but also whether they are sufficiently robust and transparent for routine use in research and applied geological workflows.
The central aim is to evaluate the advantages, limitations, and reliability of AI-supported digital palynology compared with conventional analogue methods. It is an excellent opportunity for a student with an interest in microfossils, microscope-based work, slide preparation, image analysis, and machine-learning tools.
Ref:
- Martinsen et al. 2024. *The 3-billion fossil question: How to automate classification of microfossils.* *Artificial Intelligence in Geosciences* 5, 100080. https://doi.org/10.1016/j.aiig.2024.100080