🌊 Flood Simulation Accuracy Assessment – Yamba, Australia Flood (March 2022)
How fast can we simulate a real flood without losing accuracy?
To find out, I investigated the Yamba flood event (NSW, Australia, February–March 2022) to evaluate the accuracy and efficiency of fast flood simulation modules in STE compared with HEC-RAS and Sentinel-1 satellite observations.
Using STE’s GEE Script Writer, I accurately extracted the flood extent from Sentinel-1 SAR imagery through an advanced flood-mapping algorithm developed for precise flood detection.
🌧️ Stage–hydrograph data were obtained from the Bureau of Meteorology
🗺️ Terrain and land-cover data were processed entirely in STEGIS:
• Topography: Copernicus 30 m DEM (via STE’s OpenTopography plugin)
• Land Cover: ESA Land Cover dataset (via STE plugin)
• Manning’s n Raster: Automatically generated based on ESA land-cover for high spatial precision
The 2D flood domain was identical for all models — 2869 km² with a reach length of 91 km. Both STE SWELIA and HEC-RAS SWE-LIA used a 115 m grid resolution with sub-grid variability activated.
The simulation covered 500 hours (28 Feb 2022 00:00 – 15 Mar 2022 23:00) under identical hardware conditions.
💻 Results:
HEC-RAS SWE-LIA: 135 min (stopped at 60% completion, Simulated till 07 Mar 2022) → 94.54 % accuracy vs. Sentinel-1
STE SWELIA: 17 min total → 94.53 % accuracy vs. Sentinel-1
STE Real-Time Flood Mapper: 0.1 min (≈ 6 seconds) → 93.55 % accuracy vs. Sentinel-1
💡 Conclusion:
The results confirm that STE’s highly optimized flood simulation modules can achieve near-satellite accuracy while drastically reducing computation time. This makes real-time or rapid-response flood mapping highly feasible — without compromising scientific precision.
STE’s flood simulation modules achieved near-satellite accuracy while being up to 800× faster, demonstrating the potential for real-time flood mapping and rapid decision support in emergency management.
📊 More details and the stage–hydrograph data are provided in the accompanying images.
