About us

As part of the collaborative Carbon4D project with Goethe University Frankfurt, two new publications have been released. The studies provide new insights into the spatial patterns of soil organic carbon stocks and the influence of landscape characteristics on soil respiration, contributing to a better understanding of carbon dynamics across landscapes. The papers have been published in Geoderma Regional and Geoderma:
Baumberger M; Haas B; Borken W; Nowosad J; Giese L; Klein-Raufhake T; Hamer U; Meyer N; Meyer H. (2026): How the landscape influences soil respiration: Explaining spatio-temporal patterns with interpretable machine learning. Geoderma 472: 117904.
Haas B; Baumberger M; Müller M; Schweers J; Hülsmann L; Lehndorff E; Meyer H; Meyer N. (2026): Regional spatial and vertical patterns of SOC stocks in a low mountain landscape in Germany. Geoderma Regional 46: e01102. 

Our latest publication in ERDKUNDE discusses key methodological challenges in spatial machine learning, including validation, uncertainty quantification, algorithm development, software implementation, and reproducibility. The paper advocates for more rigorous and transparent spatial prediction workflows, with particular attention to realistic model evaluation and the communication of uncertainty. This collaborative contribution emerged in part from discussions at the Advances in Spatial Machine Learning 2025 workshop, bringing together perspectives from researchers across the field. We are grateful to all workshop participants whose insights helped shape the ideas presented in the paper. Read the article: https://www.erdkunde.uni-bonn.de/article/view/3471

Our working group participated in two workshops on remote sensing for environmental monitoring. Jan Steen was part of the Young Modellers in Ecology Workshop 2026 at the Alfred Wegener Institut on Sylt and presented results from our ongoing project "Ready for Take-Off!?" and led a sessions on computational methods for geodata. Marvin Ludwig was speaker at the Conference of the Center for Nature Conservation and Sustainable Energy (KNE) in Berlin on how to integrate drone remote sensing into conservation monitoring in solar parks. We had many insightful discussions with other researchers, politics and administration.

Our research group participated in this year’s general assembly of the European Geosciences Union  with several contributions presenting current work on spatial modelling, machine learning, and deep learning in the context of Earth observation. In addition to engaging discussions around our research, the conference provided excellent opportunities for networking, scientific exchange, and the development of new ideas and collaborations.

Following the successful first edition in Münster last year, that we initiated, the second Advances in Spatial Machine Learning workshop took place this year in Tartu, Estonia and was organized by Evelyn Uuemaa and Marta Jemeljanova. The workshop brought together researchers working at the intersection of spatial data science, machine learning, and Earth observation to discuss recent methodological advances and emerging research challenges. Further information: https://jakubnowosad.com/advsml_2026/

Over the past years, our team has developed a range of methods to support the application of machine learning for spatial data. These methods have been implemented in our R package CAST. From March 11–13, we held our second CAST Package Developer Days. During these three days, the development team came together to revise the package, improve existing functions, and implement new features. More information about the package and its documentation can be found here: https://hannameyer.github.io/CAST/

Since February, Fyeqa Akram has joined our research team as a research associate and doctoral candidate. She will be working on our Twente–Münster collaboration project, “Quantifying and Modelling Peat Breathing with Satellite Radar Data,” which we supervise jointly with Ling Chang from the University of Twente. We are very much looking forward to the collaboration!

We are pleased to announce that R Coding for Ecology (Springer, 2025) has now been published with contributions from our group! The book features our CAST package, which implements key methods for spatial predictive modeling developed by our group. These methods are illustrated in a reproducible way using the example of mapping plant species richness. In addition, Jakub authored several other chapters, including contributions on landscapemetrics, cblindplotR, and cartogram.

Read more here: https://link.springer.com/book/10.1007/978-3-031-99665-8

Maiken Baumberger successfully defended her dissertation on 31.10.2025 and was awarded a doctorate in natural sciences. The title of her thesis is: „Landscape-scale modelling of soil properties in space, depth and time: an interpretable machine learning approach”. Congratulations from your working group!

With 14 students from Landscape Ecology in Münster and Physical Geography in Frankfurt, we spent 17 days on excursion – from Helsinki to Hamningberg in Varanger and back through western Finland. Along the way, we explored various topics in physical geography. We examined the omnipresent traces of the Ice Age, studied Finland’s agriculture and forestry, got to know Aapa and Palsa mires, and visited peatland restoration projects. In Lapland, we focused on cultural and ecological aspects of reindeer grazing as well as the traditions of the Sámi. The impacts of climate change on these ecosystems were a recurring central theme. We spent most of the trip camping, which allowed us to experience and enjoy the landscape with its numerous lakes particularly intensively, alongside the scientific program.
 

On July 3, our working group day took place at the Wersehaus. We discussed and developed the strategies and content of the working group together and then paddled, barbecued and had a great day in a convivial atmosphere.

From March 24th to 25th, we hosted an international workshop at our institute, focusing on advances in spatial machine learning (see https://jakubnowosad.com/advsml_2025/). Seventeen colleagues from various institutions joined us to discuss key topics such as model validation, uncertainty assessment, and interpretability. The event was inspiring, and we look forward to continuing the discussions with follow-up activities.

In April, Jan Steen starts as our new working group member and as a PhD student in our new project "Ready for Take-Off?!". Funded by the DBU and with our projects partners, we will utilize and develop new usecases of drone remote sensing for an improved habitat monitoring.

In our new projekt Sensorlök, we want to integrate environmental sensors into teaching at the Institute of Landscape Ecology in order to train students to implement innovative methods of environmental monitoring in practice. Led by Marvin Ludwig, a sensor network will be set up by students in our teaching and research area “Lüntener Wald” in the summer term '25 that automatically measures, transmits and evaluates groundwater levels, microclimate and soil parameters.

Our new project “Ready for Take-Off!?” starts in April. Together with the Biologischen Station Zwillbrock, the LWL and the Nationalpark Vorpommersche Boddenlandschaft, we are developing and evaluating drone remote sensing methods to support the flora-fauna-habitat monitoring of grassland and heath habitat types. The focus is on the reliable estimation of habitat quality criteria and the applicability of the methods for users in practice.

In a joint publication by the Animal Ecology working group and ourselves, and led by our former Master's student Alina Krämer, we modelled the habitat suitability of the sand lizard at its north-western distribution limit. The results have now been published in the Journal of Biogeography: https://doi.org/10.1111/jbi.15099

From Januray on, Darius Görgen will join our research team as a scientific assistant and PhD student.
He will work in our new project "Deep learning in Space and time", which is part of the Transregio 391 - Spatio-temporal Statistics for the Transition of Energy and Transport (https://trr391.tu-dortmund.de/).
In our subproject, we will explore new methods for using deep learning in the context of spatio-temporal earth observation data.