Author(s): Ernando Mendez; Laura Cagigal; Sonia Castanedo; Beatriz Perez Diaz
Linked Author(s): SONIA CASTANEDO
Keywords: Hybrid modelling coastal hazards tropical cyclone BlueMath
Abstract: 1. Introduction One of the main components of the World Bank’s Pacific Resilience Program (PREP) for Pacific Islands is to strengthen Early Warning Systems and Preparedness to natural hazards such as tropical cyclones, coastal/riverine flooding, volcanoes, tsunamis and earthquakes by improving the quality of forecasting and warning services. In particular, coastal flooding is caused by extreme total water levels triggered by the combination of concurrent large-scale oceanographic processes as mean sea level, waves, tides, and storm surges. These processes interact at a range of different spatial and temporal scales, creating the necessity to analyse the multivariate nature of the problem and the dependence between the different variables involved. On top of that, Small Pacific Island Countries as Samoa and Tonga, exposed to recurrent, increasing and devastating flooding events, are in urgent need of decision-support systems adapted to their limited computational resources to assist disaster managers and coastal planners. 2. Methodology A fast and reliable system for predicting flooding extents in coastal areas based on a cascade of hybrid models or metamodels has been built, integrating the multivariate effect of regular day conditions and extreme events (i. e., Tropical Cyclones). The key principle is that the hybrid system relies on libraries of pre-run cases of a variety of numerical hydrodynamic models (e. g., SWAN, DELFT3D, XBeach, LisFlood-FP) for simulating waves, wind, overtopping, riverine flooding and coastal flooding. These libraries, coupled with state-of-the-art statistical techniques (clustering algorithms, non-linear interpolation techniques) allow to efficiently downscale hydrodynamic conditions to shore and ultimately produce inundation maps in a matter of seconds. Figure 1 shows the cascade of five metamodels that comprise the TC early warning system, all of them included in BlueMath (https: //gitlab. com/geoocean/bluemath), an innovative, open-source framework developed in Python and designed to run on Jupyter Notebooks environments. Specifically, the metamodels are: (1) GreenSurge (Perez-Diaz et al., under review), which given the track of a TC, allows obtaining the associated storm surge; (2) SHyTCWaves (van Vloten et al., 2024), which provides directional wave spectrum estimates at a regional scale; (3) BinWaves (Cagigal et al., 2024), which downscales the regional wave spectra to a local near-shore area of interest; (4) HyBeat (Zornoza et al., 2024) which allows further downscaling of local waves and water levels to the coast to account for nonlinear effects that occur at lower depths; and (5) HyFlood which uses the levels at the coast to produce flooding maps (e. g. Figure 2) 3. Conclusions This fast system allows for assessing the risks faced by coastal communities and infrastructures at a range of time horizons. When applied to the past, the system allows to probabilistically study the areas more prone to flooding, while when applied to forecast conditions, both at a seasonal scale (~months) and short-term scale (~days), it improves the preparedness and minimizes the risk by the implementation of Early Warning Systems, in all cases at a very low computational effort. Integrating the hazard maps with exposure layers and damage functions by means of the software RiskScape, the system provides impact maps (economical lossess) due to coastal flooding, rainfall flooding and wind damage.
Year: 2025