Because of a lack of spatially distributed evaluation data flood inundation modelling is commonly performed with simplified flood propagation models that very often consider only one channel roughness parameter value to be calibrated. This concept as such is debatable, as friction values are known to be spatially heterogeneous. Moreover, with (over-) simplified model structures, it is clear that distributed simulations may consistently perform well at a given location whilst underperforming at another. Using only limited field data in model calibration, it is not possible to gather enough information on local model errors to improve the modelling concept. However, new processing procedures of remotely sensed flood imagery allow model calibration to be performed at any desired location. Such spatially distributed data can provide enough information to assess models locally. This local evaluation enables the modeller to define and set spatially distributed patterns of model parameters. Using the 2003 River Alzette (G.D. of Luxembourg) flood event recorded by the ENVISAT satellite, this study demonstrates that by directing the flood-modelling concept towards spatially clustered roughness parameters conditioned on remote sensing, it is possible to identify a model structure that generates acceptable model simulations not only at the reach scale but also locally. Applying the same procedure on earlier events with larger magnitude generates similar parameter clusters that are correlated with those obtained with the 2003 ENVISAT flood event. This illustrates the great potential that remote sensing holds in terms of identifying new ways to test existing model concepts and to contribute to the development of improved flood inundation models. The findings thus support the long-debated hypothesis that remote sensing improves our understanding and thus modelling of complex environmental processes, such as floods.
- Behavioural criteria
- Flood inundation model error
- Monte Carlo-based computation
- Parameter identifiability
- Remote sensing
- Spatially distributed parameter