A Big Data processing strategy for hybrid interpretation of flood embankment multisensor data

Abstract

The assessment of flood embankments is a key component of a country’s comprehensive flood protection. Proper and early information on the possible instability of a flood embankment can make it possible to take preventative action. The assessment method proposed by the ISMOP project is based on a strategy of processing huge data sets (Big Data). The detection of flood embankment anomalies can take two analysis paths. The first involves the computation of numerical models and comparing them with real data measured on a flood embankment. This is the path of model-driven analysis. The second solution is data-driven, meaning time series are analysed in order to detect deviations from average values. Flood embankments are assessed based on the results of model-driven and data-driven analyses and information from preprocessing. An alarm is triggered if a critical value is exceeded in one or both paths of analysis. Tests on synthetic data demonstrate the high efficiency of the chosen methods for assessing the state of flood embankments.