Keynote Talk: Improvements on flow forecasting using precipitation nowcasting based in radar advection techniques: Assessment of predictability and uncertainty propagation

In hydrological warning systems, precipitation nowcast is a very important issue in order to extend the time series of precipitation inputs of a runoff forecasting model in real-time. In this framework, radar information is very useful because of its high resolution both in time and space, which provides a good description of the rainfall field evolution. Advection techniques based on radar have shown some skill in extending the anticipation with which flow forecasts may be simulated through a distributed rainfall-runoff model. However, this improvement has been found to be very dependent on the nature of the event: the quality of flow forecasts decays rapidly for fast evolving precipitation patterns (for example, when short-lived convective cells affect the studied basin). The main purpose of the present work consists of exploring the relation between the lifetime of rainfall fields and the predictability of the response of a given hydrological basin (that is, to the anticipation of flow forecasts of good quality). On the other hand, firsts attempts of coupling ensemble rainfall forecasts generated using a heuristic stochastic extrapolation technique with a distributed rainfall-runoff forecasting model are also presented in order to quantify the uncertainty induced in flow forecasts due to the uncertainty related to the different precipitation scenarios generated with the nowcasting technique.