Photo-realistic flood scenes

Media such as television news weather segments has been simulating natural disasters like floods to enhance visual impact for years, but technology like FloodGen makes this job much faster and cheaper. FloodGen is an advocacy tool that uses generative AI to create photorealistic scenes of predicted flooding that evoke a sense of urgency about risks and impacts. Users begin with a citywide view of flood-prone neighborhoods and can zoom in to more than 30 high-risk, high-vulnerability locations in 10 case study sites. These sites were selected based on coastal and stormwater flood hazards, vulnerability of environmental justice areas and hurricane evacuation zones, and proximity to transit, public housing, hospitals, commercial areas, schools, and libraries. The images are transformed using CycleGAN, a generative adversarial network (GAN) model that reprocesses normal street view images to depict flooding scenarios.

This signal highlights how community-based resilience advocates can use generative AI tools to illustrate localized impacts of climate change to mobilize support for resilience efforts.