Master's thesis; AI-based Species Distribution Modeling using Remote Sensing

Master's thesis; AI-based Species Distribution Modeling of Predators in Scandinavia using Remote Sensing

Background
Understanding the distribution of large predators, such as bears, lynx, and wolves, is crucial for effective wildlife management and conservation in Scandinavia. However, monitoring these species across vast and often remote landscapes is challenging and resource-intensive. Species distribution models (SDMs) are powerful tools that combine species occurrence data with environmental variables to predict the geographic distribution of species. When integrated with high-resolution remote sensing data, these models can provide a scalable and cost-effective approach to understanding habitat suitability and predicting predator distribution. This project will leverage predator observation data from ROVBASE, a comprehensive database for predator inventories in Norway and Sweden, in combination with remote sensing imagery and AI-based methods to develop robust species distribution models. Further data sources may be explored and fused in the final solution, such as data from iNaturalist.

Description
The aim of this project is to develop and evaluate species distribution models for large predators in Scandinavia using a combination of remote sensing data and machine learning. In collaboration with relevant authorities in Norway and Sweden, the project will systematically investigate how environmental factors derived from satellite imagery (such as land cover, vegetation indices, topography, and human activity) influence the distribution of key predator species. 

Key Responsibilities

1. Preprocess and integrate predator observation data from ROVBASE with various remote sensing datasets. 

2. Implement and train deep learning models for species distribution modeling. 

3. Analyze the robustness of the models and identify the key environmental variables driving predator distribution. 

4. Explore workflows for scaling the approach to larger geographical areas and for long-term monitoring programs.

Expected outcome
A proof-of-concept workflow for automated, large-scale species distribution modeling of predators. The results will contribute to more accurate habitat assessments for key Scandinavian predators and provide valuable decision support for wildlife management, conservation efforts, and conflict mitigation. We will investigate what data sources are most informative for training the models and what modeling strategies are the most successful.

Who are you?
The project is suitable for a motivated student with a strong background in Python and machine learning. An interest in applied AI, remote sensing, and environmental applications is suitable, and experience with machine learning frameworks (e.g. PyTorch, TensorFlow) is meriting. Background knowledge in GIS, image analysis, or ecology is an advantage.

What We Offer

• Supervision by an interdisciplinary team of researchers with expertise in machine learning, remote sensing, and climate-related applications.
• Access to relevant datasets and computational resources.
• An opportunity to contribute to cutting-edge research with potential for high-impact publications.
• A collaborative and stimulating research environment in the RIDR research group.
• After a successful project completion, RISE will pay each student SEK 30 000.

Terms

• Location: Gothenburg
• Time: January to June, 2026
• Suitable for a team of two students.

Welcome with your application!
Please team up and apply together with someone. Interested candidates should submit the following in TeamTailor by November 14:

• A curriculum vitae (CV).
• Academic transcripts.
  

Be brief in your application. List relevant meriting projects and experience (only).
Contact: Olof Mogren, olof.mogren@ri.se. 

We encourage students that are interested in research to apply, and will provide the support needed to write and submit a paper to a relevant conference at the end of the project if the project is successful and the student wants to do so. This can be very beneficial if you want to apply for a PhD position in the future. We encourage students to pair up with their applications and work together on the project. We look forward to receiving your applications and exploring the fascinating world of AI for nature together.

Recommended reading
Lange, C., Cole, E., Van Horn, G., and Mac Aodha, O. (2023). Active learningbased species range estimation. 37th Conference on Neural Information Processing Systems (NeurIPS 2023).
Teng, M., Elmustafa, A., Akera, B., Bengio, Y., Abdelwahed, H. R., Larochelle, H., and Rolnick, D. (2023). SatBird: Bird species distribution modeling with remote sensing and citizen science data. 37th Conference on Neural Information Processing Systems (NeurIPS 2023).
Abdelwahed, H. R., Teng, M., and Rolnick, D. (2024). Predicting species occurrence patterns from partial observations. Tackling Climate Change with Machine Learning workshop at ICLR 2024.