Future-H2O at the EGU 2021
The EGU 2021 was organized in an online format. It enabled great exchange with scientists at different career levels from all over the world and the Future-H2O project benefited from insights into recent advances with data-driven methods in hydrology and for predicting spatial patterns of geochemical variables. Stefan Baltruschat and Annika Nolte presented their methodological ideas for these two research areas as part of the Future-H2O project.
A Random Forest approach for estimating groundwater CO2 concentrations was presented by Stefan Baltruschat. Annika Nolte presented an approach for predicting groundwater levels (GWL) simultaneously at multiple wells from very different coastal regions around the world. A Long-Short-Term-Memory (LSTM) network model is built with GWL data, precipitation and temperature time series and static environmental attributes from global map products (e.g. soil type) and validated and tested on independent parts of the GWL time series (Figure 1).
Figure 1 Deep Learning modeling approach for GWL forecasts. © Nolte, A.
- Estimation of groundwater CO2 concentrations on a catchment scale using Random Forest Presentation held by Stefan Baltruschat at the EGU
- Scale-dependent impacts of natural and anthropogenic drivers on groundwater level dynamics – analysis of shallow coastal aquifers using deep learning Presentation held by Annika Nolte at the EGU