Signal Processing Methods for Flood Early Warning Systems

We present in a data-driven approach for detection of anomalies in earthen dam (dike) behaviour that can indicate the onset of flood defence structure failure. This approach is implemented in the UrbanFlood early warning system's Artificial Intelligence component that processes dike measurements in on-line manner. Suggested approach successfully detected anomalies registered during the non-destructive piping experiment at the Zeeland dike (the Netherlands).