The project aims to develop a fast method for computing the effects of changes in groundwater systems. This could be from wells, pollution, heavy precipitation etc.. This fast method should be developed by a combination of geostatistics, hydrological simulations and machine learning.
Project start: 01-09-2020
Project ending: 31-08-2023
BU/department/ industry to which the project is associated: Water
Cooperation-partners: Aarhus University, Department of Geoscience
Advisory board: Troels Norvin Vilhelmsen (Niras) & Thomas Mejer Hansen (Aarhus University)
The ultimate goal of the project is to develop an applicable, very fast, and highly accurate computation method for administrative groundwater work concerning well placement, pollution and future climate challenges.
We are currently working on replicating groundwater flow simulations from heavy numerical models using fast neural networks. Results are promising in both accuracy and speed. In addition, we wish to include estimates of uncertainty on the results from the geological model and hydraulic parameters. With a complete set-up, it will be possible to perform accurate and fast groundwater flow simulations with results as probabilities instead of single values.
This will help improve decision support in groundwater management and increase the number of investigations feasible to perform within a limited timeframe.
Project status: Ongoing
5 main keywords about the project: Machine Learning, Groundwater, Geostatistics, Neural Networks, Numerical Models.
Educational background: BSc Theoretical Physics, MSc Geophysics
Hopes for the future works of the project: I hope that some of the work will be implemented in BEST.
The importance of cooperating with NIRAS: NIRAS offers expert guidance on groundwater modelling and helps the project focus on industrial impact and not just academic.
