Evaluation and uncertainty assessment of a real-time regional ensemble flow forecasting system in Catalonia

Abstract

A real-time flow forecasting system with significant predictability is an effective tool for facing the impacts caused by extreme weather events in the context of climate change. However, the forecasting skill is limited by the uncertainty of the different model components. The ensemble flow forecast has been frequently used to enhance the accuracy of the hydro-meteorological forecast system and describe forecast uncertainty. In this case, the assessment and enhancement of the system's performance after a long-term application and the source of uncertainty is fundamental. This study presents an approach to evaluate the accuracy of a hydrological forecasting system that has been running in real time at a regional scale in Catalonia (NE Spain) since 2020, and generates flow forecasts in all the gauging stations of the Catalan Water Agency. The rainfall observation and precipitation forecast, which is compos the European Centre for Medium-range Weather Forecasts have been used in the modified rainfaled of 52 members, produced by the rainfall-runoff HBV model to generate probabilistic flow forecasts. The study analyzes the sources of uncertainty, and proposes guidelines to enhance the accuracy of the forecasting. The quality of the forecasts has been analyzed based on the most significant rainfall events for a period of 2 years as a function of the lead time. The observed flow information and the simulated result (0-h lead time forecast) have been compared with the forecasted flow. Both deterministic and probabilistic scores have been used in the comparison and the no-rainfall forecast has been used as a reference. The relationships between forecast accuracy and both catchment size and rainfall type have been discussed and the sources of uncertainty in the system have been explored.