|Paper title||Refining and Validating a Surface Water Flood Forecasting System Using Sentinel 1 SAR data|
|Form of presentation||Poster|
Flooding is recognised as an environmental hazard that affects more people than any other environmental hazard. It is also anticipated to affect a higher proportion of the global population and incur rising costs in the future due to rapid urbanization, increasing settlements in floodplains, climate change, and variability.
To meet these challenges Previsico have developed their FloodMap Live software to provide high resolution, real-time flood forecasts based on a predictive flood modelling system. Flood forecasts, such as those provided by Previsico, enable actions to be taken to reduce loss of life and property in the event of a flood and help to identify genuine insurance claims post-flood.
Yet flood models require integration and validation with external data sources, such as satellite imagery, for re-calibration of model predictions and to demonstrate prediction effectiveness. Independent information from satellite data enables refinements to be made to flood models, in turn supporting more accurate forecasts of flooding evolution.
Following a successful collaboration with the University of Leicester, Previsico is developing a flood extent product derived from Sentinel 1 radar imagery that will provide near-real-time information on flood location and extent in both urban and rural areas. Synthetic Aperture Radar (SAR) was chosen for the satellite product as data collection is not impeded by cloud cover or a lack of illumination and can acquire data over a site during day or night time under almost all weather conditions. Furthermore the Sentinel 1 SAR-C instrument provides dual polarisation capability, very short revisit times and rapid product delivery. This satellite product will allow Previsico to refine it’s model in order to offer more accurate and validated flood models to their customers ensuring they can respond to a flood event in a targeted and efficient manner.
Here we will present our progress so far in developing and utilising Sentinel 1 SAR data to refine and validate a commercial flood model. Results from the initial version of this Sentinel 1 flood product were encouraging, as shown in Figure 1 for an area over Doncaster and Rotherham in the UK which were affected by flooding in November 2019. The method also performed well in non-flood events suggesting it is fairly robust even when inundation has not occurred. Further comparisons against external data sources such as Copernicus EMS showed promise and allowed us to identify improvements to the code, either to be implemented in the prototype product or in future versions of this product. Comparisons to flood forecasts from Previsico’s flood modelling system were also performed and the results from this will be presented.