Improving the Regionalization of Groundwater Head Dynamics with static environmental features

This study aims to improve the regionalization of groundwater head dynamics using static environmental features. Recent machine- and deep learning studies have explored the use of these features for spatial and temporal imputation (Haaf et al., 2023) or improvement of global models (e.g., Chidepudi et al. (2024); Heudorfer et al. (2024); Nolte et al. (2024)). While physiographic features, including geology, land cover, anthropogenic factors, and topography, have been identified as important predictors of groundwater dynamics at regional and watershed scales (Haaf et al., 2020; Haaf et al., 2023; Rinderer et al., 2017; Zhao et al., 2023), there is still a lack of understanding on how to leverage static features to achieve significant model improvement for groundwater time series regionalization (e.g., Heudorfer et al., 2024; Nolte et al., 2024).

In this study, we use a data-driven, static feature-based approach to regionalize groundwater head duration curves and reconstruct them based on similar donor sites (Haaf et al., 2023). We evaluate the similarity of static features compared to the geographical proximity of donor sites. The data set consists of more than 150 ten-year, daily groundwater head time series in the upper Danube catchment and more than 60 static features at each site.

Our findings suggest that geographical proximity, related to both physiographic and climatic similarity, is the best default approach for selecting donor sites for regionalization. However, in specific cases where the nearest donor sites were located in different hydrogeological regimes, static features significantly improve regionalization. The study demonstrates the potential for improving the regionalization of groundwater dynamics using spatial features in diverse hydrogeological settings. Further research on larger and more diverse data sets is warranted to allow for robust feature selection strategies.

 

References

Chidepudi, S. K. R., Massei, N., Jardani, A., Dieppois, B., Henriot, A., & Fournier, M. (2024). Training deep learning models with a multi-station approach and static aquifer attributes for groundwater level simulation: what’s the best way to leverage regionalised information? EGUsphere, 2024, 1-28. https://doi.org/10.5194/egusphere-2024-794
Haaf, E., Giese, M., Heudorfer, B., Stahl, K., & Barthel, R. (2020). Physiographic and Climatic Controls on Regional Groundwater Dynamics. Water Resources Research, 56(10). https://doi.org/10.1029/2019wr026545
Haaf, E., Giese, M., Reimann, T., & Barthel, R. (2023). Data‐Driven Estimation of Groundwater Level Time‐Series at Unmonitored Sites Using Comparative Regional Analysis. Water Resources Research, 59(7). https://doi.org/10.1029/2022wr033470
Heudorfer, B., Liesch, T., & Broda, S. (2024). On the challenges of global entity-aware deep learning models for groundwater level prediction. Hydrol. Earth Syst. Sci., 28(3), 525-543. https://doi.org/10.5194/hess-28-525-2024
Nolte, A., Haaf, E., Heudorfer, B., Bender, S., & Hartmann, J. (2024). Disentangling coastal groundwater level dynamics in a global dataset. Hydrol. Earth Syst. Sci., 28(5), 1215-1249. https://doi.org/10.5194/hess-28-1215-2024
Rinderer, M., McGlynn, B. L., & van Meerveld, H. J. (2017). Groundwater similarity across a watershed derived from time-warped and flow-corrected time series. Water Resources Research, 53(5), 3921-3940. https://doi.org/10.1002/2016wr019856
Zhao, F.-H., Huang, J., & Zhu, A. X. (2023). Spatial prediction of groundwater level change based on the Third Law of Geography. International Journal of Geographical Information Science, 37(10), 2129-2149. https://doi.org/10.1080/13658816.2023.2248215