Computational Ecology

In the research area of Computational Ecology, we combine ecological questions with data-driven and mathematical modeling approaches to better understand dynamic processes in microbial communities. The focus is on analyzing marine time series from Arctic ecosystems, employing methods from network theory, machine learning, and environmental modeling. The goal is to quantify ecological stability, identify key organisms, and improve predictions of changes in Arctic marine ecosystems.

Although our research primarily focuses on the Arctic, we also conduct comparative analyses with other marine time series, for example, from Atlantic, subpolar, or temperate regions. The aim is to better understand general patterns of microbial dynamics. The methods applied here can be transferred to other ecological systems, such as soil or plant microbiomes, as investigated within CEPLAS.

Our work is strongly interdisciplinary and based on long-term collaborations with biological, oceanographic, and data science partner institutions. Bachelor's and Master's students are actively involved in many of these projects, working with us on data analysis, method development, and visualizations.

Time Series Analysis

The analysis of ecological time series is a central method for detecting seasonal, interannual, or climate-driven changes in microbial communities. We use high-resolution long-term data from remote access sampler (RAS), stationed in the Arctic and Atlantic Oceans, which continuously measure biological and physical parameters. These data include not only microbial abundances (based on amplicon sequencing) but also environmental factors such as temperature, salinity, oxygen, and ice cover.

Using Fourier transformation, clustering techniques, recurrent models, and similarity analyses, we reconstruct activity patterns, stability-relevant periods, and transitions between system states. The goal is to elucidate regulatory mechanisms and ecological tipping points.

Methods: Networks, AI, Modeling

Our methodological focuses include:

• Network Analyses: We reconstruct microbial-ecological interaction networks (e.g., through Co-Occurrence or Cross-Convergent Mapping) and identify their structural properties. Using Energy Landscape Analysis, we model the stability of such systems under various environmental conditions.
• Artificial Intelligence: We use Deep Learning models to identify and quantify, for example, zooplankton in high-resolution image data (e.g., from the LOKI system). Our self-developed tool, DeepLOKI, allows for automated, fast, and objective evaluations of large image volumes.
• Mathematical Modeling: We use Energy Landscape Analysis to investigate ecological stability and dynamics. This method represents possible system states as an "energy landscape," where valleys indicate stable communities and hills represent transition states. By embedding time series data into these landscapes, we can identify stable configurations, responses to disturbances, and transition pathways between states. This method provides a simple visualization and assessment of the stability of microbial networks under changing conditions.

These methods form a transferable set of tools for a wide range of ecological questions and are explicitly applicable to terrestrial systems such as plant or soil microbiomes.

Marine Datasets & Cooperations

Our research is based on unique datasets from the Arctic Ocean, collected over years within the framework of the long-term observatories LTER HAUSGARTEN and FRAM. Particularly noteworthy are autonomous samplers that continuously collect samples even during the polar night – a globally unique dataset.

Cooperation partners include:

• The Alfred Wegener Institute (Bremerhaven)
• British Antarctic Survey (BAS)
• The CEPLAS Graduate School for Plant Sciences (in the context of transferable methods)
• Various bioinformatics groups in Germany

The resulting datasets are publicly accessible and serve as a basis for comparative analyses with marine and terrestrial microbiomes.

PolarBot

PolarBot is an innovative science communication project that uses an AI-powered chatbot to convey research on polar regions and climate change in an understandable and interactive way. The chatbot is designed in the form of a user-friendly web platform for people of all ages and knowledge levels. The goal is to playfully spark interest in polar research topics and provide scientifically sound answers – for interested laypeople as well as for schoolchildren, families, and students.

In addition to its use on the website, PolarBot is also designed for use in exhibitions, educational formats, or events. Visitors can ask specific questions about displayed images or videos and receive direct context-related explanations – complementing classic exhibition concepts with an individual, dialogue-based component. The technical implementation is carried out in cooperation with the Conversational AI platform Cognigy, based in Düsseldorf.

We also regularly work with Bachelor's and Master's students in the development and content design of the chatbot – particularly in the areas of Data Literacy, Communication Design, and Environmental Education.

PolarBot can be found at the following link: http://polarbot.hhu.de/

Research Questions

Our research addresses the following central questions:

• How do climate change, Atlantification, and ice melt affect the structure and function of marine microorganism communities?
• Which species are considered key species or indicators of ecological stability?
• How do interaction networks change throughout the year and under varying environmental conditions?
• Which seasonal patterns are stable, and which are subject to climate-driven shifts?
• How can methods of time series analysis, network analysis, and AI be transferred to other systems – such as soil or plant microbiomes?

The goal is to develop robust ecological indicators for environmental change through model-based and data-driven research – both in marine and terrestrial areas.

Contact: Dr. Ovidiu Popa, Dr. Ellen Oldenburg