Development of a Flood Early Warning System for Critical Catchments in Southern Chile

Floods are among the most devastating extreme events that affect people worldwide. Their frequency and severity are expected to increase as a result of climate change, which requires timely and informed decisions to protect people and reduce economic losses. In Chile, the current flood warning system compares real-time streamflow observations with flood thresholds calculated by the National Water Directorate (DGA) and issues flood warnings only a few hours in advance. Therefore, this work develops a prototype flood early warning system for Andean catchments, using meteorological ensemble forecasts coupled with a hydrological model that provides streamflow forecasts with lead times of up to 10 days. The methodology implements the Novel Multi-objective Particle Swarm Optimisation (NMPSO) algorithm to calibrate the TUWmodel, a conceptual hydrological model that explicitly accounts for snow processes and rainfall-runoff dynamics. This optimisation framework ensures robust parameter estimation for multiple hydrological objectives, with a focus on better reproducing high streamflows while preserving low-flows dynamics. The calibrated model is then forced with daily mean air temperature and precipitation data from two medium-range meteorological forecast ensembles, namely MSWX-Mid and ECMWF-IFS, comprising 30 and 51 members, respectively. Tobias-correct meteorological forcings, an empirical quantile mapping approach is implemented using the daily 5-km Chilean dataset CR2METv2.5 as reference. The efficiency of the system is evaluated in three snow-influenced Andean catchments in southern Chile, which were affected by severe floods events during 2023 and 2024.The developed prototype is expected to be soon available online to improve medium-range flood forecasting in critical Andean catchments and to provide timely and reliable information to decision makers and the public. We gratefully acknowledge the financial support of ANID-Fondecyt Regular 1212071 and ANID-PCI NSFC 190018, and ANID/FONDAP 1523A0002.