Development and evaluation of an ensemble forecast/hindcast system for storm surges in the Rio de la Plata Estuary

The development and evaluation of ensemble hindcasting and forecasting systems (EHS and EFS, respectively) for storm surges in the Rio de la Plata Estuary (RdP) is presented. The models were forced by atmospheric sea level pressure and 10 m winds. The ensemble forcing for the EHS was generated by temporal-spatial shifting of the operational global control ERA5 reanalysis provided by the European Centre of Medium-Range Weather Forecast (ECMWF), because the ERA5 associated ensemble shows too little dispersion in this area. EFS was instead based on a lead time of 4 days and forced with the 50-member high-resolution ensemble prediction system of the ECMWF. EHS was evaluated over a long period (the 2000-2010 decade), whereas EFS was evaluated for the ten most extreme surges that occurred during that period: five positive (which caused coastal flooding) and five negative (which affected navigation and drinking water supply) events. Based on traditional statistics (area under the ROC curve and Brier scores), both systems were assessed from a probabilistic point of view. Results show that both EHS and EFS can incorporate more than 90% of the observations in the uncertainty range. They also showed good skill in hindcasting and forecasting surges, particularly extreme events, EHS being about 20% better than the control model and EFS up to 55% better, in agreement with state-of-the-art models developed for other parts of the world. Results also showed that EFS can predict most of the surge peaks with 95% confidence, with a range of uncertainty of about +/- 0.90 m and +/- 9 hr. Therefore, results encourage the implementation of EHS and EFS as useful and robust tools for future climate studies, decision makers and the general public, to improve the quality of risk management decisions by quantifying forecast uncertainty.

Automated summary

The study presented the development and evaluation of ensemble hindcasting and forecasting systems for storm surges in the Rio de la Plata Estuary. Results showed that both systems performed well in predicting extreme events, with EFS performing the best in predicting surge peaks.