AI Builder and Data Storyteller
When people think about floods, they often picture overflowing rivers or streets filling with water after intense rainfall. Groundwater flooding is different. It rises from below and can affect basements, cellars, and buried infrastructure with little visual warning.
This study focused on a real event in the Alz River valley in southeastern Bavaria during June 2013. Rather than produce one deterministic flood map, the objective was to represent uncertainty explicitly, because groundwater systems are driven by uncertain inputs and incomplete observations.
The modeling workflow combined physical groundwater simulation with Bayesian inference. Sensitivity analysis identified which parameters mattered most, and the DREAM algorithm was used to infer plausible parameter ranges from available observations.
The result was a probability-based risk assessment. One set of maps quantified how likely groundwater levels were to approach the land surface. Another translated that information into a more direct property-level risk signal by estimating the probability of reaching cellar depths at specific locations.
The most useful outcome is transparency: uncertainty becomes part of the result rather than hidden behind a single number. This supports better decisions because stakeholders can see both likely impacts and confidence levels.
The full paper is available here: 10.1016/j.jhydrol.2022.127797.
← Back to Blog