Accurate predictions of groundwater level changes are crucial for the sustainable management of groundwater resources. In view of climatic changes and declining groundwater levels in some regions, it is becoming increasingly important to understand how these resources will evolve over different time periods. As groundwater levels decline and abstraction increases, pressure on the resource grows, thereby intensifying conflicts over its use between the public water supply, agriculture, and industry.
In this context, the KIMoDIs joint project developed an AI-based prototype for a monitoring, data management, and information system, focusing on groundwater level predictions across various spatial scales. In addition to a comprehensive, state-wide analysis of Brandenburg, four local pilot areas with specific issues (e.g. land use conflicts, ecological impacts, or salinisation risks) were examined in detail.
An operational workflow has been developed for seasonal groundwater level predictions in the state of Brandenburg, which calculates forecasts for the next 16 weeks. An integrated early warning function highlights monitoring sites when the predicted value for the respective calendar week is unusually low. Robust medium- and long-term model projections were calculated for around 270 monitoring sites under various climate projections (CMIP5). In the RCP 8.5 scenario, a decline in groundwater levels between 8% and 26% is expected for the period 2071–2100. In contrast, in the RCP 2.6 scenario, a smaller decline ranging between 1% and 11% is expected. These ranges highlight the importance of considering climate uncertainties to ensure robust long-term projections and water management adaptation strategies.
Another focus was the development of abstraction scenarios to model the impact of water withdrawals on groundwater levels. Due to variations in data availability, quality and objectives, this focus was limited to a specific pilot region. This demonstrated how abstraction scenarios can be systematically integrated into decision-making processes. The results demonstrate that future groundwater level development is determined by the interaction of climate and usage, and that both factors must be considered in planning.
The prototype also consolidates a wide range of additional information, creating a robust basis for decision-making. As well as groundwater level trends, the model incorporates abstraction data from individual districts, and comprehensive information on groundwater salinisation in Brandenburg. To this end, extensive datasets comprising over 23,000 analyses were subjected to genetic evaluation, providing a spatial and temporal overview of salinisation trends in the state's groundwater.
The developed prototype clearly demonstrates the potential of AI-based predictions, integrated monitoring data and modern analytical methods for forward-looking groundwater management. Following the project’s conclusion, the industrial partner Mapular will primarily be responsible for commercial exploitation.
The Federal Ministry of Research, Technology and Space (BMFTR) is funding the joint project "KIMoDIs" within the "Sustainable Groundwater Management" funding initiative under the federal research programme on water "Wasser:N", which is part of the BMFTR strategy "Research for Sustainability (FONA)".
Literature
Final report
- BRODA, S., KUNZ, S. & WETZEL, M. (2026): Verbundprojekt KI-basiertes Monitoring-, Datenmanagement- und Informationssystem zur gekoppelten Vorhersage und Frühwarnung vor Grundwasserniedrigständen und -versalzung – KIMoDIs in der Fördermaßnahme Nachhaltige Grundwasserbewirtschaftung (LURCH). – Sachbericht zum Verwendungsnachweis. DOI: 10.34657/42765
Papers
- CHIOGNA, G., MARCOLINI, G., ENGEL, M. et al. (2024): Sensitivity analysis in the wavelet domain: a comparison study. – Stoch Environ Res Risk Assess 38, 1669–1684. DOI: 10.1007/s00477-023-02654-3
- DOLL, F., LIESCH, T., WETZEL, M., KUNZ, S. & BRODA, S. (2026): Validation strategies for deep learning-based groundwater level time series prediction using exogenous meteorological input features. – Geoscientific Model Development, 19(7), 2657–2675. DOI: 10.5194/gmd-19-2657-2026
- ENGEL, M. & KÖRNER, M. (2024): Sentinel-2 Tiling Scheme Grid-Overlay for Efficient I/O-Operations Based on Spherical Voronoi Polygons and Local Optimization. – In IEEE International Geoscience and Remote Sensing Symposium. DOI: 10.1109/IGARSS53475.2024.10640984
- KUNZ, S., SCHULZ, A., WETZEL, M., NÖLSCHER, M., CHIABURU, T., BIESSMANN, F. & BRODA, S. (2025): Towards a global spatial machine learning model for seasonal groundwater level predictions in Germany. – Hydrology and Earth System Sciences, 29(15), 3405–3433. DOI: 10.5194/hess-29-3405-2025
- RECKE, T., WETZEL, M. & HERMSDORF, A. (2025): KIMoDIs – KI-basiertes Monitoring-, Datenmanagement- und Informationssystem zur gekoppelten Vorhersage und Frühwarnung vor Grundwasserniedrigständen und -versalzung. – Brandenburgische Geowissenschaftliche Beiträge Bd. 32/2025.
Presentations
- BRODA, S. (2026): KI-basiertes Monitoring-, Datenmanagement- und Informationssystem zur gekoppelten Vorhersage und Frühwarnung vor Grundwasserniedrigständen und -versalzung. – Abschlussveranstaltung der BMBF-Fördermaßnahme „Nachhaltige Grundwasserbewirtschaftung“ (LURCH), Frankfurt am Main.
Download (PDF, 7MB, File does not meet accessibility standards) - BRODA, S. (2025): KI in der Hydrogeologie – Erfahrungen zwischen Innovation und Realität. – 27. Symposium Strategien zur Sanierung von Boden und Grundwasser, Frankfurt a. M.
- DOLL, F., BRODA, S., HEUDORFER, B., KUNZ, S., THULLNER, M., WETZEL, M. & LIESCH, T. (2024): Die Eignung von Modellen des maschinellen Lernens für die gekoppelte Vorhersage von Grundwasserständen und Grundwasserversalzung. – 27. Tagung der Fachsektion Hydrogeologie e. V. in der DGGV e. V., RWTH Aachen University.
- DOLL, F., LIESCH, T., HABBEL, B., PAXIAN, A., DALANE, C., VOIGT, J., WETZEL, M., KUNZ, S. & BRODA, S. (2026): Verbesserung der Vorhersage von Grundwasserständen durch abgeleitete meteorologische Eingangsvariablen in Deep-Learning-Modellen. – FH-DGGV-Tagung 2026, Leipzig.
- ENGEL, M. & KÖRNER, M. (2024): Sentinel-2 Tiling Scheme Grid-Overlay for Efficient I/O-Operations based on Spherical Voronoi Polygons and Local Optimization. – IEEE International Symposium on Geoscience and Remote Sensing (IGARSS), Athens.
- ENGEL, M., KUNZ, S., WETZEL, M. & KÖRNER, M. (2024): Multiresolution Analysis based Assessment of Agricultural Effects on Groundwater Levels. – ESA Agriculture Under Pressure, Frascati.
- HABBEL, B., LIESCH, T., OHMER, M., WETZEL, M., CLOS, P. & BRODA, S. (2026): Automatisierte Datenvorverarbeitung für die Zeitreihenmodellierung: Das Python-Paket hytise. – FH-DGGV-Tagung 2026, Leipzig.
- KUNZ, S., SCHULZ, A. NÖLSCHER, M., WETZEL, M., CHIABURU, T., BIESSMANN, F. & BRODA, S. (2024): Global machine learning (ML) models for short-term groundwater level prediction. – Geosaxonia 2024, Dresden.
- KUNZ, S., WETZEL, M., ENGEL, M. & BRODA, S. (2025): Deep Learning Models for Seasonal Groundwater Level Prediction. – EGU General Assembly 2025, Vienna.
- KUNZ, S., WETZEL, M., GONZÁLEZ, E., ELBRACHT, J., SCHULZ, A., CHIABURU, T., NÖLSCHER, M., BIESSMANN, F. & BRODA, S. (2026): Integration von Grundwasserentnahmedaten in ML-Modelle: Potenziale für verbesserte Grundwasserprognosen und die Bildung von Nutzungsszenarien. – FH-DGGV-Tagung 2026, Leipzig.
- SCHULZ, A., KUNZ, S., NÖLSCHER, M., WETZEL, M., BRODA, S. & BIESSMANN, F. (2024): Entwicklung eines globalen Modells für kurzfristige Grundwasserstandsvorhersagen unter Anwendung des Temporal Fusion Transformers. – 27. Tagung der Fachsektion Hydrogeologie e. V. in der DGGV e. V., RWTH Aachen University.
- THULLNER, M., DOLL, F., WETZEL, M., KUNZ, S., HÜSKKE, K., BRODA, S., LIESCH, T. & HUBNER, G. (2026): Faktoren für Salinitätsschwankungen der Süßwasserlinse unterhalb der Insel Langeoog. – FH-DGGV-Tagung 2026, Leipzig.
- WETZEL, M. (2024): Künstliche Intelligenz in der Hydrogeologie. – Berlin-Brandenburger Brunnentage, Berlin.
- WETZEL, M., KUNZ, S., DOLL, F., HABBEL, B., LIESCH, T. & BRODA, S. (2026): Datengetriebene Modellierung von Grundwasserganglinien: Herausforderungen unterschiedlicher Reaktionsdynamik. – FH-DGGV-Tagung 2026, Leipzig.
- WETZEL, M., SCHMIDT, L.-K., HERMSDORF, A., KUNZ, S., LIESCH, T., HEUDORFER, B., DOLL, F. & BRODA, S. (2024): Advancing water resource management: Insights and implications from global machine learning models in groundwater prediction. – European Geosciences Union General Assembly (EGU24), Vienna.
Posters
- BRODA, S., KUNZ, S., WETZEL, M., SCHMIDT, L.-K. & HERMSDORF, A., (2025): Predicting Decadal Groundwater Levels in Brandenburg: Deep Learning Approaches for Sustainable Management. – EGU General Assembly 2025, Vienna.
- BRODA, S. & WETZEL, M. (2024): KIMoDIs – KI-basiertes Monitoring-, Datenmanagement- und Informationssystem für das Grundwasser in Brandenburg. – Auftaktveranstaltung der BMBF-Fördermaßnahme „Nachhaltige Grundwasserbewirtschaftung“ (LURCH), Frankfurt am Main.
Download (PDF, 2MB, File does not meet accessibility standards) - DOLL, F., OHMER, M., LIESCH, T., SCHMIDT, L.-K., HERMSDORF, A., WETZEL, M., KUNZ, S. & BRODA, S. (2024): Informationsbasierte Optimierung hydrogeologischer Messnetze – Anwendung von SSPOR auf das hydrogeologische Landesmessnetz Brandenburg. – Statusseminar der BMBF-Fördermaßnahme „Nachhaltige Grundwasserbewirtschaftung“ (LURCH), Frankfurt am Main.
- GONZÁLEZ, E., BRODA, S. & ELBRACHT, J. (2023): KIMoDIs – KI-basiertes Monitoring-, Datenmanagement- und Informationssystem zur gekoppelten Vorhersage und Frühwarnung vor Grundwasserniedrigständen und -versalzung. – Tagung Grundwasserströmungs-Modellierung, Hanover.
- GONZÁLEZ, E., ELBRACHT, J., KUNZ, S., RIECHERS, S. & WETZEL, M. (2026): Ergebnisse aus dem BMBF-Projekt „KIMoDIs“ – Projektgebiet West-Hümmling. – FH-DGGV-Tagung 2026, Leipzig.
- GONZÁLEZ, E., ELBRACHT, J., WETZEL, M. & BRODA, S. (2024): KIMoDIs – Projektgebiet West-Hümmling. – 27. Tagung der Fachsektion Hydrogeologie e. V. in der DGGV e. V., RWTH Aachen University.
- WETZEL, M., KUNZ, S., SCHMIDT-RUDOLPH, L.-K., HERMSDORF, A. & BRODA, S. (2026): KI-gestützte Grundwasserstandsprojektionen: Wie sich Brandenburgs Grundwasserstände bis 2100 entwickeln. – Abschlussseminar der BMBF-Fördermaßnahme „Nachhaltige Grundwasserbewirtschaftung“ (LURCH), Frankfurt am Main.
Download (PDF, 4MB, File does not meet accessibility standards) - WETZEL, M., SCHMIDT, L.-K., HERMSDORF, A., KUNZ, S., LIESCH, T., HEUDORFER, B., DOLL, F. & BRODA, S. (2024): KI-basiertes Grundwassermanagement in Brandenburg: Notwendigkeit und bisherige Entwicklungen. – 27. Tagung der Fachsektion Hydrogeologie e. V. in der DGGV e. V., RWTH Aachen University.
- WETZEL, M., SCHMIDT, L.-K., HERMSDORF, A., KUNZ, S., LIESCH, T., HEUDORFER, M., DOLL, F. & BRODA, S. (2024): KI-basiertes Grundwassermanagement in Brandenburg – Notwendigkeit und bisherige Entwicklungen. – Statusseminar der BMBF-Fördermaßnahme „Nachhaltige Grundwasserbewirtschaftung“ (LURCH), Frankfurt am Main.
Download (PDF, 3MB, File does not meet accessibility standards)
Partner
- Karlsruhe Institute of Technology (KIT)
- Brandenburg State Office for the Environment (LfU)
- Mapular
- German Federal Institute of Hydrology (BfG)
- Technical University of Munich (TUM)
- Germany's National Meteorological Service (DWD)
- State Office for Mining, Energy and Geology of Lower Saxony (LBEG)
- Brandenburg State Office for Mining, Geology and Raw Materials (LBGR)
Funding information
02WGW1662A