Time series of SAR flood maps to calibrate a LISFLOOD-FP hydraulic model

Synthetic Aperture Radar (SAR) satellites are capable of all-weather day and night observations that can discriminate between land and smooth open water surfaces over large scales. Because of this there has been much interest in the use of SAR satellite data to improve our understanding of water processes, in particular for fluvial flood inundation mechanisms.

In past studies it has been proven that integrating SAR derived data with hydraulic models can improve simulations of flooding. The objective of this work is to see how we can improve LISFLOOD-FP Sub-Grid 2D hydraulic model simulations by only using SAR-derived maps of flood extent. We consider the effectiveness of using real SAR data from the European Space Agency’s archive of ENVISAT images to calibrate hydraulic model parameters.

As a test case we applied the method to the River Severn between Worcester and Tewkesbury. We firstly applied the automatic flood mapping algorithm of Giustarini et al. (2013) to ENVISAT ASAR (wide swath mode) flood images to generate a series of SAR-derived flood extent maps. We then created an ensemble of parallel flood extent maps with the hydraulic model (each model representing a unique parameter set). Where there is a favourable comparison between the modelled flood map and the SAR-derived flood extent map we may find the version of the hydraulic model that optimally reproduces the extent of flood water detected on the ground. The model parameters are effectively calibrated using SAR-derived data.

Applying the method to a sequence of SAR acquisitions may provide insight into the advantages, disadvantages and limitations of using a time series of acquired satellite observations.

To complement the investigation we also explore parameter ‘identifiabilty’ within our sequence of available SAR-derived flood extent maps by adopting the DYNIA method proposed by Wagener et al. (2003). We show when in a sequence of images we might most easily detect the looked-for model parameters from the SAR acquisition sequence.