MENTOR - Machine learning based, nationwide groundwater level prediction

Country / Region: Germany

Begin of project: October 1, 2020

End of project: September 30, 2026

Status of project: April 15, 2024

Data-driven models, especially artificial neural networks (ANN), have proven their suitability for modelling and predicting groundwater levels many times. In particular, ANNs are characterised by the lower effort and are less dependent on field data availability compared to conventional numerical, physical-based modelling approaches, especially when it comes to (supra-)regional problems. New developments in the field of deep learning promise a significant improvement of already existing prediction approaches.

A reliable forecast of groundwater levels is, for example, the basis for deriving water availability for drinking water supply and agricultural irrigation, the delineation of potential soil subsidence zones due to extremely low groundwater levels in connection with droughts and/or water abstraction, the delineation of areas of potential groundwater floodings to protect transport infrastructure, buildings and agricultural land, as well as the development of suitable avoidance and adaptation strategies.

Input data, NARX model and groundwater level forecast Source: KIT & BGR

The aim is to further improve the method developed in the project phase I, which enables nationwide short-, medium- and long-term forecasts of groundwater levels and spring discharges, particularly with regard to extreme events (droughts/heat waves and extreme/high precipitation). To this end, the existing approaches, which are based on artificial neural networks and are already suitable for weekly, monthly and seasonal forecasts at individual groundwater measuring points, are being further developed, but new promising deep learning approaches are also being pursued. In order to close spatial gaps and thus improve the regionalisation of forecasts, accompanied by pilot magnetic resonance soundings, the number of reference measuring points defined so far is to be extended. Furthermore, the prediction quality is to be improved by adding additional input parameters such as river levels, water withdrawals etc. and by applying deep learning approaches. Data basis are groundwater data of the national monitoring networks as well as weather and climate forecasts and climate projections of the German Meteorological Service (DWD). The forecasts will be made available on BGR's website via a web application, updated according to the forecast periods.

Version 1 of the GRUVO web application was released to the public in mid-April and is available at https://gruvo.bgr.de.

Literature:

Technical Paper

• WUNSCH, A. & LIESCH, T. (2020): Entwicklung und Anwendung von Algorithmen zur Berechnung von Grundwasserständen an Referenzmessstellen auf Basis der Methode Künstlicher Neuronaler Netze. - Abschlussbericht Projektphase I, 183 S., 61 Abb., 11 Tab., 13 Anh.; KIT, Karlsruhe. doi: 10.5445/IR/1000136522

Papers

• WUNSCH, A., LIESCH, T. & BRODA, S. (2022b): Deep learning shows declining groundwater levels in Germany until 2100 due to climate change. - Nat Commun 13, 1221. doi: 10.1038/s41467-022-28770-2
• WUNSCH, A., LIESCH, T. & BRODA, S. (2022a): Feature-based Groundwater Hydrograph Clustering Using Unsupervised Self-Organizing Map-Ensembles. - Water Resour. Manage., 36(1): 39-54. doi: 10.1007/s11269-021-03006-y
• WUNSCH, A., LIESCH, T. & BRODA, S. (2021): Groundwater level forecasting with artificial neural networks: a comparison of long short-term memory (LSTM), convolutional neural networks (CNNs), and non-linear autoregressive networks with exogenous input (NARX). - Hydrol. Earth Syst. Sci. 25: 1671-1687. doi: 10.5194/hess-25-1671-2021
• WUNSCH, A., LIESCH, T. & BRODA, S. (2021): Feature-basiertes Clustering von Umweltzeitreihen mit Self-Organizing-Map-Ensembles. - In: REUSSNER, R.H., KOZIOLEK, A. & HEINRICH, R. (Hrsg.): Informatik 2020. Gesellschaft für Informatik, Bonn. (p. 1035-1041). doi: 10.18420/inf2020_98
• WUNSCH, A., LIESCH, T. & BRODA, S. (2018): Forecasting Groundwater Levels using nonlinear Autoregressive Networks with exogenous Input (NARX). - J. Hydrol. 567: 743-758. doi: 10.1016/j.jhydrol.2018.01.045

Conference contributions

• BRODA, S., WUNSCH, A. & LIESCH, T. (2018): Wochen-, Monats-und Jahreszeitenvorhersage von Grundwasserständen mit künstlichen neuronalen Netzen. 26. Tagung der Fachsektion Hydrogeologie e. V. in der DGGV e. V., Ruhr-Universität Bochum.
• BRODA, S., WUNSCH, A., LIESCH, T., GOLDSCHEIDER, N. & REICHLING, J. (2017): Weekly, monthly and seasonal Forecasting of Groundwater Levels using Artificial Neural Networks. - 44th IAH Congress, Dubrovnik, Croatia.
• NÖLSCHER, M., HEBER, M., CLOS, P., ZAEPKE, M., STOLZ, W. & BRODA, S. (2024): Aktueller Zustand und Vorhersage der Grundwasserstände – eine neue bundesweite Fachanwendung. - 29. Tagung der Fachsektion Hydrogeologie e. V. in der DGGV e. V., RWTH Aachen University.
• WUNSCH, A., LIESCH, T. & BRODA, S. (2021): Using Convolutional Neural Networks to evaluate Long-Term Groundwater Trends in Germany. - 48th IAH Congress, Brussels, Belgium.
• WUNSCH, A., LIESCH, T. & BRODA, S. (2020): Groundwater Level Forecasting with Artificial Neural Networks: A Comparison of LSTM, CNN and NARX. - AGU Fall Meeting, San Francisco, CA, USA.
• WUNSCH, A., LIESCH, T. & BRODA, S. (2019): Uncover Similarities of Groundwater Dynamics with Machine Learning based Hydrograph Clustering. - AGU Fall Meeting, San Francisco, CA, USA.

Partner:

Karlsruhe Institute of Technology (KIT), Institute of Applied Geosciences - Division of Hydrogeology