CS-UDT flood digital twin
2026
A crowd-sourced urban digital twin for hurricane flooding in Jacksonville, with a zero-install citizen client and a planner dashboard. The demo runs on simulated data.
Datasheet
- Client
- Zero-install browser client
- Pipeline
- LLM parsing of citizen reports and social posts into hazard events
- Verification
- Three-channel convergence rule: official sensor anomaly + citizen report + social post in the same map cell
- Engagement
- When two channels agree, nearby residents get a one-tap verification quest
- Replay
- Simulated replay modeled on Hurricane Irma (Sep 10–12, 2017): first three-channel confirmation 10.8 hours into a 30-hour replay; four confirmed cells and three open quests at the peak
- Data
- Simulated replay and synthetic community content
- Paper
- Coauthored extended abstract, DTRUE 2026 (oral presentation)