From flood forecast to direct damage prediction: Supporting early action with an Impact-based Forecasting system

The global economic loss caused by weather-related extreme events amounts to over $260 billion in 2022. Storms and floods are among the deadliest disasters and are responsible for the highest toll. Despite committed research efforts in strengthening flood forecasting and making those predictions readily and openly available, much remains to be done to facilitate intervention when locally acting upon those forecasts. This research aims at building an automated tool to forecast flood direct damages with a high spatial resolution and timeliness. Thus, allowing prompt, informed and targeted early action on site before the disaster hits. Moreover, it can serve as a device to unravel criticalities within preparedness plans and guide the adoption of adaptation measures in the long term. The proposed research develops a tool to rapidly link GLOFAS discharge forecasts with the relative inundation map and direct damages caused. The method includes three modules: i) a factual component collecting satellite-derived flood maps of historical events; ii) a probabilistic component based on hydrological modelling and iii) the impact assessment. The past event database comprises 10-meter resolution inundation maps derived from Sentinel-1 SAR imagery with a single-scene automated classification method. The outcome of hydrological modelling is then integrated with the remote sensing database to improve its accuracy and spatial resolution. Lastly, the impact assessment module estimates affected people and the economic damage to buildings. The presented methodology is applied to two case studies: the flooding caused by Tropical Cyclone Idai that made landfall in March 2019 in Mozambique and the country-wide flood event that occurred in Pakistan in the summer of 2022.