Titre stage / Internship title
DeepPlankton: Time series data analysis using Deep Learning to study competition and coexistence in freshwater plankton
Contexte / Context:
Understanding how species coexist and its impacts for maintenance of biodiversity is a central focus in community ecology. Coexistence ability depends on the amount of niche overlap – how much limiting resources two species share, and the competitive ability of a species – how efficiently they exploit these resources. Importantly, these traits are not fixed: They can evolve due to biotic and abiotic environmental interactions.
This Master internship is part of the COMPEX project, a multi-year, eco-evolutionary study of coexistence in natural plankton communities. The experiment consists of outdoor mesocosms (300L tanks) installed and managed over five years in Konstanz, Germany. The focal organisms considered are the freshwater crustacean Daphnia magna and Daphnia pulex, which were included either alone or in competition as mesocosm treatments. Both feed on a diverse algae community originating from Lake Konstanz, and we expect differences in algae community composition based on Daphnia living in coexistence or isolation. Regular sampling of mesocosm daphnia and algae provides a big dataset of time series samples (images were done from samples for further analysis).
Objectifs scientifiques / Scientific objectives:
The specific focus of the master project can be matched to the students main interests. Two directions are possible:
The first direction (P1) is the analysis of Daphnia timeseries data to detect differences between coexistence treatments. We will use machine learning models for automated Daphnia species detection, and assess model uncertainty to realistically inform count data. The resulting count data can be fitted to simple mechanistical population growth model to compare growth patterns between coexistence treatments.
The second direction (P2) focuses on potentially different niches of resources use in the different daphnia species, influencing algae community composition. We will use machine learning species detection for algae communities and analyze how algal diversity and composition may vary across species. Because species detection with diverse algae communities is a more complex methodological tasks, the final focus of this project will be defined with the student according to feasibility and specific interests.
A research stay at the University of Konstanz is encouraged during the project to gain hands-on experience with mesocosm samples and image datasets.
Modalités de réalisation / Implementation terms :
• Species identification of Daphnia and/or algae samples
• Annotation of image datasets
• Machine learning based image classification, Programming in Python or R
• Estimation of classification uncertainty
• Develop computational simulations of model equations (P1)
• Fitting of mechanistic population growth models to Daphnia count data (P1)
• Analysis of algae community compositions (P2)
Conditions particulières / Particular conditions :
Co-supervised by Jelena Pantel and PhD student Sebastian Borgmann (University of Duisburg-Essen)
Place of work : Université Marie et Louis Pasteur (Besançon, France) – Laboratoire Chrono-environnement, University of Konstanz (Germany)
Timing : There is some flexibility around the exact start date
Compétences souhaitées / Desired skills:
• Interest in aquatic, evolutionary and community ecology
• Interest in computational or quantitative methods
• Interest or prior experience in working with programming languages (R, Python)
• Time-management and project organization skills
Commentaires récents