Postdoctoral Researcher Position in Ecological -Guided Machine Learning at Aarhus Univ

Department of Ecoscience - Freshwater Ecology - Faculty:
Technical Sciences

Deadline 2 Mar 23:59 CET

Expected start 1 Sep

Fixed term full-time position 1 Sep 2026 - 28 Feb 2029

A 2.5-year position as Postdoctoral Researcher within the field of Freshwater Ecology is available at the Department of Ecoscience, Aarhus University. The expected start date is 1st of September 2026 or as soon as possible thereafter. The position seeks to strengthen and complement the Department’s ongoing activities in freshwater ecology, particularly aquatic ecosystem modelling and water quality modelling, with focus on Knowledge-Guided Machine Learning.

The position is a rewarding opportunity to be integrated in an excellent freshwater group. The department’s research and advisory activities are project‑based with a solid tradition in cross‑disciplinary research and international collaboration.

The postdoctoral researcher will work with Dr. Robert Ladwig on developing hybrid models that integrate limnological knowledge into machine learning models following the paradigm of Knowledge-Guided Machine Learning (KGML). The position is part of an on‑going project on “
Integrating AI into Aquatic Ecosystem Models to Decode Ecological Complexity
” funded by Villum Fonden. Within that project, the focus is on exploring novel ways to infer information from environmental data to update our scientific conceptual models. The data foundation will come from the long‑term lake monitoring initiative LTER‑DK, which includes several Danish lakes equipped with real‑time and high‑frequency monitoring buoys, incl. Lake Ravn near Aarhus. The monitoring at Lake Ravn consists of three monitoring buoys, an automatic profiling station, and a weather station.

Further, weekly water quality sampling and turbulence profiles provide additional data for training and testing. A specific aim is to further develop the methodology of modular compositional learning (MCL). Here, an aquatic ecosystem model is decomposed into modular sub‑components that can be either process‑based models and/or deep learning models. MCL has the flexibility to replace any uncertain process description with a deep learning model, which makes it ideal for ecological simulations where precise mathematical descriptions of key processes are lacking but data for training are available.

The MCL methodology will be applied on critical ecological processes, i.e., vertical turbulent diffusion, phytoplankton production and consumption, greenhouse gas emissions, etc., to develop hybrid models. Performance will be compared to several 1D aquatic ecosystem models to evaluate if the respective hybrid MCL models are improving their performance. The overall project aim is on refining current aquatic ecosystem models by building models based on KGML to have improved projections and more robust estimates of uncertainty.

The project provides training and growth opportunities in AI
. Candidates who have strong interest in machine learning, but minimal experience, are encouraged to apply. Along the same lines, candidates with a strong foundation in machine learning, minimal experience in freshwater ecology, but a strong interest to develop models for environmental challenges, are also encouraged to apply.

The postdoctoral researcher will work within an interdisciplinary team of computational and applied scientists. The work will be done in close collaboration with international and interdisciplinary colleagues. There will be close collaborations with colleagues at the University of Wisconsin‑Madison and Virginia Tech (USA). In the Computational Limnology team at the Freshwater Ecology section at Aarhus University, we are striving to have a diverse, fair and inclusive team and work environment.

Team members are supporting each other and help when we see someone physically or mentally struggling. We work respectfully with people from different backgrounds, experiences and nationalities. To collaborate more efficiently and ensure reproducibility, we implement the principles of open…