Improving flood forecasting based on optimized flood mapping from radar images

Funded by: Fonds Wetenschappelijk Onderzoek (FWO)

Researcher: ir. Lisa Landuyt

Promotors: Prof. Dr. ir. Frieke Van Coillie​, Prof. Dr. ir. Niko Verhoest

Duration: 2016 – 2020

Project objectives

This collaboration between the FORSIT lab and the LHWM lab, Ghent University, aims at improving flood inundation models by incorporating flood observations from remotely sensed radar imagery. However, these observations are prone to some uncertainty due to the inherent noise in satellite imagery and due the flood mapping procedure, which often involves some subjective choices. 

Therefore, a novel object-based approach will be developed aiming at an improved flood mapping procedure. The uncertainty in the mapped flood extent will be assessed and the approach will be compared to other existing techniques, with respect to both accuracy and uncertainty level. Then, the flood inundation model will be updated using the flood observations through a data assimilation framework, accounting for the uncertainties in both the observations and the model states. Furthermore, the impact of the data assimilation framework on the flood predictions will be investigated. Finally, the methodology of using radar data for flood monitoring and of using observations for updating flood models can be disseminated to stakeholders.