Predictive multivariate modelling for anticipatory drainage management in coastal lowlands

Climate change related sea level rise and increased winter precipitation are contributing to an increase in flood hazards in low-lying coastal regions of Germany. The magnitude of flood events in these areas is largely dependent on the capacity of the drainage infrastructure such as canals, sluices or pumps. As the drainage capacity varies depending on the technical and environmental conditions, drainage operations are especially under pressure when compound events like an inland flood and a storm surge occur simultaneously.

To gain insight into the factors that impact drainage system capacity, we analyse sea level, hydrometeorological and operational datasets from coastal lowland catchments using multivariate statistical and machine learning-based approaches, e.g. rank correlation and random forests. The analysis indicates complex multi-level correlations of rainfall, wind direction and speed, and tidal water levels with inland flooding and helps to identify combinations of influencing factors that reduce drainage capacity and increase flood hazards. This information is useful to anticipate flood events and assist water management bodies in adjusting drainage operations in advance to mitigate resulting risks.