A globally-applicable framework for compound flood risk modeling

Low-lying coastal deltas are prone to floods as these areas are often densely populated and face flooding from fluvial (discharge), coastal (surge and waves) and pluvial (rainfall) drivers. If these drivers co-occur, they can cause or exacerbate flooding, and are referred to as compound flood events. Most compound flood studies have either investigated the statistical dependence between drivers or used hydrodynamic models to assess the physical interactions between drivers, but few have combined both aspects to examine extreme flood levels for e.g. risk assessments. Furthermore, hydrodynamic compound flood models are often setup at a local scale, require many person hours to set up and are based on local data, making these hard to scale up. Hence, the need for globally-applicable compound flood risk modelling remains. 

We developed a globally-applicable framework for compound flood risk modelling. It consists of a local hydrodynamic SFINCS model which is automatically set up based on global datasets after several processing steps and loosely coupled to global models using HydroMT (https://deltares.github.io/hydromt_sfincs/latest/). We applied to the Sofala province of Mozambique where we validated it for two historical tropical cyclone events and used it for a compound flood risk analysis. For the validation, we compared flood extents from the global and local flood models with observed flood extents from remote sensing. Our analysis shows that the local model, while based on the same data, has a higher accuracy compared to the global model. This is due to a more complete representation of flood processes and an increased spatial resolution. We also analyzed the compound flood dynamics and show that the areas where water levels are amplified by interactions between flood drivers vary significantly between events. Finally, we also calculated the compound flood risk from fluvial, pluvial and coastal drivers based on a large stochastic event set of plausible (compound) flood conditions derived from ~40 years of reanalysis data. We find that coastal flood drivers cause the largest risk in the region despite a more widespread fluvial and pluvial flood hazard as most exposure is affected by elevated sea levels. Flood risk increases when accounting for the observed dependence between flood drivers compared to independence and this difference is mainly attributed to events with large return periods. Since the model setup and coupling is automated, reproducible, and globally-applicable, the presented framework offers a way forward towards large scale compound flood risk modelling.