Project Background
Taiwan is frequently affected by typhoons and heavy rainfall, which may cause extensive inundation and result in loss of life and property. To provide governments at all levels with essential information for flood disaster prevention, preparedness, and emergency response, the government has implemented the Potential Inundation Maps Project. The maps simulate potential inundation depths and affected areas under different rainfall scenarios and serve as an important reference for disaster risk management.
Project Objectives
The Third-Generation Potential Inundation Maps, representing the second major update of the maps, were completed between 2014 and 2016. Since then, changes in topography and land conditions, together with advances in inundation modeling techniques and improvements in data quality, have created a need for another comprehensive update.
The Fourth-Generation Potential Inundation Maps incorporate updated basic data reflecting recent changes in terrain and land conditions and employ advanced numerical modeling technologies. The models are calibrated and validated using historical typhoon and flood events. Following a series of technical reviews and evaluations, Potential Inundation Maps and associated datasets are produced for the specified rainfall scenarios.
Main Project Activities
The project is implemented in two major phases.
Phase I: Data Collection and Model Development
This phase includes the collection and analysis of hydrological, topographic, and oceanographic data, together with the development of numerical inundation models.
Phase II: Model Simulation and Results Production
This phase includes model calibration and validation, scenario simulations, local consultation meetings or stakeholder interviews, report preparation, and the production of inundation maps and numerical datasets.
Potential Inundation Maps have continuously evolved in terms of spatial resolution, rainfall scenarios, topographic data, and numerical modeling capabilities. Spatial resolution has improved from approximately 200 meters for the first generation and 40 meters for the second generation to 10 meters or finer in urban planning areas for the fourth generation.
Topographic data have also advanced from photogrammetric base maps in the first generation and Digital Elevation Models (DEMs) in the second generation to the high-resolution Hydraulic Digital Elevation Model (HyDEM) adopted by the fourth generation. Modeling capabilities have progressively expanded to incorporate storm surge and external water levels, sewer systems, spatially distributed rainfall, and the effects of buildings on overland flow.
Table 1. Comparison of the Four Generations of Potential Inundation Maps
Generation | Period | Spatial Resolution | Rainfall Scenarios | Major Features |
|---|
First Generation | 1999–2001 | 200 m × 200 m × 1 m | Design rainfall: 150, 300, 450, and 600 mm/day | Used photogrammetric base-map data surveyed between 1981 and 1989; adopted a two-dimensional zero-inertia wave model. |
Second Generation | 2007–2009 | 40 m × 40 m × 0.1 m | Design rainfall: 200, 350, 450, and 600 mm/day; return periods of 1.1, 2, 5, 10, 20, 25, 50, 100, 200, and 500 years | Used updated DEM data; incorporated detention facilities and hydraulic structures as well as storm surge, wave overtopping, and overflow from rivers and regional drainage systems; adopted two-dimensional zero-inertia and SINOTOPO models. |
Third Generation | 2014–2016 | 40 m × 40 m × 0.1 m or finer | 6 hr: 150, 250, 350 mm; 12 hr: 200, 300, 400 mm; 24 hr: 200, 350, 500, 650 mm | Incorporated sewer systems, regional drainage inundation simulation, and results of flood-prone area management projects; considered flood-flow processes and differences in rainfall distribution between plains and mountainous areas; adopted the SOBEK model with structured grids and single-core computation; results were made available online. |
Fourth Generation | 2024–2026 | 10 m × 10 m × 0.1 m or finer in urban planning areas | 6 hr: 150, 250, 350 mm; 12 hr: 150, 300, 400 mm; 24 hr: 200, 350, 500, 650 mm | Uses HyDEM jointly developed by the Water Resources Agency and the Ministry of the Interior; adopts gridded rainfall data and accounts for building elevations and flow-blocking effects; uses the Multi-Dimensional Service Platform of the National Land Surveying and Mapping Center for visualization; adopts Delft3D FM 1D2D with unstructured meshes and multi-core computation. |
Expected Benefits
Unlike the Third-Generation Potential Inundation Maps, which were developed using the SOBEK model, the fourth generation adopts the Delft3D FM 1D2D modeling system developed by Deltares in the Netherlands.
The new approach allows rainfall to be applied directly to the two-dimensional land surface to simulate overland flow, while dynamically coupling surface runoff with one-dimensional river and drainage-channel processes. This provides a more realistic representation of rainfall–runoff and inundation processes.
Delft3D FM 1D2D uses flexible, unstructured meshes to represent complex computational domains and supports multi-core computation, substantially improving computational efficiency.
The fourth-generation models also use the Hydraulic Digital Elevation Model (HyDEM), jointly developed by the Water Resources Agency and the Ministry of the Interior. In addition to high-resolution elevation information, HyDEM incorporates hydraulic spatial features such as levee overflow lines, land–sea boundaries, and water-body polygons. These features allow one-dimensional and two-dimensional model components to be represented more accurately and improve the ability of simulations to reproduce actual field conditions.
Simulation results can also be dynamically visualized through the Multi-Dimensional Service Platform developed by the National Land Surveying and Mapping Center, Ministry of the Interior, providing both the public and technical professionals with more detailed information on inundation processes and simulation results.
AI-Based Inundation Surrogate Models and Future Development of Real-Time Inundation Forecasting
The nationwide numerical inundation models developed for the Fourth-Generation Potential Inundation Maps can provide a foundation for future applications involving extreme rainfall events and climate-change scenarios.
By conducting simulations under a wide range of rainfall, tidal, and hydrological conditions, a comprehensive inundation simulation database can gradually be established and used as training data for the development of AI-based inundation surrogate models. As more simulation results become available, these surrogate models can be further developed to provide rapid inundation predictions and ultimately support real-time inundation forecasting.
Future applications may integrate meteorological rainfall forecasts, radar-derived rainfall data, IoT-based real-time hydrological monitoring, and digital-twin technologies. Such integration could enable rapid estimation of inundation extent and depth, support disaster-prevention decision-making and emergency response, and further strengthen the Water Resources Agency’s capabilities in smart flood management and integrated watershed governance.