From hazard-based to impact-based warnings: using social sensing to refine coastal flood thresholds in Maricá – RJ, Brazil

Conventional coastal warning systems issue alerts based on regional oceanographic thresholds calibrated for navigation safety, generating a recurrent disconnection between physical hazard and documented urban impact. This study addresses that disconnection on the coast of Maricá (Rio de Janeiro, Brazil), a microtidal urban shoreline of reflective beaches exposed to South Atlantic swells. Hourly significant wave height ( Hs ) from the ERA5 reanalysis (2022–2025) was validated against the SiMCosta buoy CDRJ-1 ( R = 0.90; RMSE = 0.32 m under extreme conditions) and cross-referenced with 645 documentary records of storm surge episodes, compiled from institutional sources (Brazilian Navy, Maricá Civil Defense), local news portals, and citizen-generated Instagram content, consolidating 113 distinct events. Forecast skill metrics (POD, FAR, CSI) revealed contrasting source profiles: institutional alerts operate with FAR between 5.0% and 16.1% under a hazard-based threshold ( Hs ≥ 2.5 m), while citizen-generated records reached POD = 85.5% with FAR = 31.5%, capturing impacts that institutional logic excludes. In 2024, Maricá Civil Defense replaced the Brazilian Navy as the primary alert issuer. The analysis identified Hs ≥ 3.5 m as the critical threshold above which structural damage to urban infrastructure becomes recurrent, with damage frequency rising from 6.2% below 3.0 m to 50.0% above 3.5 m. Furthermore, 42.5% of documented events occurred below the operational threshold of 2.5 m, exposing the limitation of purely hazard-based warning logic. These results support the recalibration of local warning systems and the transition from hazard-based to impact-based logics on urbanised coasts.