Precipitation forecasting at very short time ranges is an important point in water management and especially in flood warning. In this framework, radar information may be very useful thanks to its high resolution both in time and space.

The purpose of this study is to assess the goodness of the precipitation fields obtained by a nowcasting technique (that only uses radar data) from a hydrological perspective using a distributed rainfall-runoff model, simulating real-time conditions of flood forecasting.

The nowcasting technique (based on those proposed by Seed and Keenan, 2000 or Germann and Zawadzki, 2002) consists on: a) estimating the precipitation motion field (that is kept constant during the forecast) from the recently measured radar fields, and b) filtering small scales of the precipitation field according to an autoregressive model while the forecast time increases (assuming that the further evolution of these scales is unpredictable).

The observed and forecasted rainfall fields from radar are combined with the data measured by a telemetered network of rain gages to generate the input of a rainfall-runoff model. The integration of the nowcasting technique and the hydrological model will provide up to 2 h of precipitation and up to 4 h of flood forecasts in the framework of the Besòs basin flood forecasting system.