Flash floods in unconfined rivers have become more frequent and severe, causing
significant damage to infrastructure and inducing widespread evacuations. Understanding the inundation area, water depths, and flow velocities of such events
is critical for identifying flood-prone areas and developing effective communication
strategies to support vulnerable populations and river managers. Traditional
communication methods, such as reports and maps, may not be efficient or accessible to broader audiences. Digital twins, highly detailed and accurate virtual
representations of real-world flooding events often utilizing Virtual Reality (VR)
experiences, are promising communication tools. This research explores the potential of advanced visualization techniques and hydraulic models to enhance flood hazard communication and management.
To achieve this, the technical requirements and applications of visualizing hydraulic
models with computer graphic suites were investigated. This led to the development
of a framework that integrates advanced fluid solvers within these suites to visualize local flooding. To ensure a realistic visualization of flash floods, understanding
these phenomena and identifying the most suitable and efficient modeling
techniques is key. The methods were demonstrated on the 2017 flash flood in the
Norwegian mountain river Storelva in Utvik, following preliminary testing and
validation on this and other case studies. Data mining played an essential role in
retrieving calibration and validation data for a numerical model of the 2017 flash
flood in Utvik through soft data and remote sensing techniques. Utilizing on-site
and post-event information, the reach affected by the 2017 flash flood was reconstructed, revealing critical locations, main flow paths, and maximum wetted, and dry areas. The steep slope and loose material composition of the valley and riverbed contributed significantly to sediment transport during this extreme event. The dataset served as calibration and validation for the 2D IberPLUS model. Graphics Processing Unit (GPU)-based computing and sensitivity analysis contributed to the development of a benchmark 2D numerical model in IberPLUS using the dataset obtained from data mining. GPU-computing proved to be notably faster than standard and parallelized CPU-computing, resulting in up to 99.5% reduction in computing time. Models with fewer than 150,000 cells demonstrated the best accuracy-speed balance. The flood model was optimized following the computing speed-accuracy trade-off guidelines resulting from the sensitivity analysis and further coupled into a VR-based framework for its visualization.
The VR-based framework allows for integration of data from various meshbased
Eulerian numerical models, enabling users to interact with and explore geophysical
flows in a VR environment. This versatility was demonstrated through three case studies in Norway: a virtual snow avalanche in Skogshorn, Hemsedal, the 2020 Gjerdrum quick clay landslide, and the 2017 Utvik flash flood. Intuitive menus and interactions in the VR environment allow users to read flow depth and velocity values, fostering a direct connection between numerical data and visual
representation. A flood VR prototype was further expanded into a digital twin utilizing serious gaming. To assess the effectiveness of the digital twin and its impact on users, tests were carried out to gauge emotional responses to flooding scenarios compared to safe scenarios. Findings suggest that identity priming, i.e., providing cues to influence the users’ perception of their identity and social role, and emotional state, i.e., fear or sense of safety, significantly impacted user involvement
and experienced realism in the VR environment. Additionally, identity and fear
considerably affected cognitive workload and frustration when users were tasked
with locating a safe spot in the flooding scenario.
Steep rivers respond quickly to extreme hydrologic conditions, requiring thorough
monitoring, data collection, and modeling tools to recreate flash floods. Utilizing
GPU-equipped PCs enables near real-time hydrodynamic information, making
flood data more accessible and facilitating efficient hazard management. Integrating
advanced fluid solvers within computer graphics suites and serious gaming
through a VR-based framework offers a powerful solution for achieving realistic
visualizations of reliable hydrodynamic simulations. This combination of technologies
encourages collaboration and resource sharing among different disciplines and audiences, enhancing the effective study and communication of natural hazards. Testing a flood prototype on users revealed the impact of factors such as identity and emotions when visualizing natural hazards in VR. These findings emphasize the importance of designing hazard communication tools that are usercentered, engaging, and effective.