Hydrological management in our country in the framework of the EC Flood Directive (2007) requires putting in place Flood Early Warning Systems, specially to anticipate the consequences of floods in small and medium torrential catchments. In this context, precise hydrological modeling of the hydrological response requires of data that are not always available (e.g. runoff measurements), a challenge in the estimation and fitting of the parameters of rainfall-runoff parameters over larger domains at high resolution and short computation times. The main goal of this project is the development and evaluation of a system for flood hazard forecasting up to 2 days in advance, covering a large domain (Spain) at high resolution (1 km and 15 minutes), and adapted to real-time operation. The proposed approach was first developed by the research team in the Spanish project FFGRad (CGL2009-13139) and the European project IMPRINTS (FP7-ENV-2008- 1-226555) and has been so far implemented in two pilot basins in Spain. The idea is that, in intense rainfall events with return periods over 10 years, the rainfall integrated at catchment scale is a valid precursor to identify flood emergency situations. Therefore, rainfall observations and forecasts can be aggregated over the catchment upstream of each point of the drainage network (with a resolution of 1 km) to obtain the variable used to estimate the flash flood hazard. The rainfall inputs to the system are: (i) radar rainfall fields from the radar network of the Spanish Agency of Meteorology (AEMET), (ii) rainfall forecasts obtained by extrapolation of radar observations (up to 6 hours), and (iii) the simulations of the rainfall field obtained with a Numerical Weather Prediction model (up to 2 days). The precipitation aggregated upstream of each point (each cell of 1 km) is used as a proxy to estimate the forecasted flood hazard at the point. The flood hazard will be characterized through the probability of exceedance (or return period) of the aggregated rain forecasted at each subbasin. With this objective, the project will analyze different alternatives to characterize the relationship between the variables rainfall intensity-duration-frequency (IDF curves for the aggregated rain) at each point of the drainage network. The results will be analyzed from two different points of view: (i) assessment of the skill in forecasting floods for different recent cases at different scales (from very local to regional events), and (ii) by comparing the forecasted flood hazard with flow observations and forecasts in the catchments of the rivers Besòs (1020 km2) and Llobregat (5040 km2), where the distributed rainfall-runoff model DiCHiTop (Corral et al. 2009) has been implemented.