The scaling problem: Native precision vs. international attain
Specialised, hyper-local early warning programs have been engineered to deal with flash floods from rainfall in particular city settings, with examples in Florida (US), Barranquilla (Colombia), Manila (Philippines), Nakhon Si Thammarat (Thailand), Mayaguez (Puerto Rico), and Barcelona (Spain). These programs sometimes depend on a community of bodily sensors monitoring variables like direct and radar-inferred precipitation, water ranges and movement velocities. Whereas extremely correct for his or her particular places, they’re troublesome to scale due to the excessive prices of {hardware} deployment, the necessity for site-specific calibration algorithms and engineering experience.
At a broader stage, initiatives such because the WMO’s Flash Flood Steering System (FFGS), the European Runoff Index primarily based on Climatology (ERIC) flash flood indicator, and the US Nationwide Climate Service (NWS) Flash Flood Warnings system present wider protection via distant sensing and numerical climate fashions. These programs, nonetheless, encounter vital hurdles relating to international implementation. A main challenge is their dependency on high-resolution hydrological maps and radar-based climate forecasts, assets which can be largely unavailable throughout the World South. Moreover, the reliance on skilled hydrologists to interpret complicated mannequin information and distribute actionable warnings presents a second main problem.
To attain near-global attain, our mannequin makes use of solely international climate merchandise (NASA IMERG, NOAA CPC) in addition to real-time international climate forecasts from the ECMWF Built-in Forecast System (IFS) Excessive Decision (HRES) atmospheric mannequin and the AI-based medium-range international climate forecasting mannequin by Google DeepMind. The system presently operates at a 20×20 kilometer spatial decision, a constraint primarily pushed by the decision of worldwide out there information sources.