HYDROLOGICAL PROCESSES Hydrol. Process. 25, 2801–2813 (2011) Published online 16 March 2011 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hyp.8042 Distributed rainfall-runoff modelling for flood frequency estimation and flood forecasting L. Brocca,* F. Melone and T. Moramarco Research Institute for Geo-Hydrological Protection, CNR, Via Madonna Alta 126, 06128 Perugia, Italy Abstract: Nowadays, in the scientific literature many rainfall-runoff (RR) models are available ranging from simpler ones, with a limited number of parameters, to highly complex ones, with many parameters. Therefore, the selection of the best structure and parameterisation for a model is not straightforward as it is dependent on a number of factors: climatic conditions, catchment characteristics, temporal and spatial resolution, model objectives, etc. In this study, the structure of a continuous semi-distributed RR model, named MISDc (‘Modello Idrologico Semi-Distribuito in continuo’) developed for flood simulation in the Upper Tiber River (central Italy) is presented. Most notably, the methodology employed to detect the more relevant processes involved in the modelling of high floods, and hence, to build the model structure and its parameters, is developed. For this purpose, an intense activity of monitoring soil moisture and runoff in experimental catchments was carried out allowing to derive a parsimonious and reliable continuous RR model operating at an hourly (or smaller) time scale. Specifically, in order to determine the catchment hydrological response, the important role of the antecedent wetness conditions is emphasized. The application of MISDc both for design flood estimation and for flood forecasting is reported here demonstrating its reliability and also its computational efficiency, another important factor in hydrological practice. As far as the flood forecasting applications are concerned, only the accuracy of the model in reproducing discharge hydrographs by assuming rainfall correctly known throughout the event is investigated indepth. In particular, the MISDc has been implemented in the framework of Civil Protection activities for the Upper Tiber River basin. Copyright  2011 John Wiley & Sons, Ltd. KEY WORDS rainfall-runoff model; flood frequency; flood forecasting; experimental monitoring Received 4 June 2010; Accepted 28 January 2011 INTRODUCTION Due to the large floods that have occurred in recent years in many regions of the world, local, national, and interna- tional authorities have shown an increasing awareness of flood and inundation hazard (WMO, 2004). Two possible strategies can be used to reduce flood losses: (1) non- structural by developing real-time flood forecasting sys- tems that reduce flood risk by issuing warnings (with the complementary strategy of the education of the public on the appropriate response to warnings), and (2) structural by building protection measures such as flood retention basins and/or dikes for the reduction of flood hazard. The prerequisite for many flood protection measures is a good forecast of expected flood levels, whereas the design of permanent measures requires the estimation of flood levels for different exceedance probabilities (Plate, 2009). In both cases, rainfall-runoff (RR) models play a central role. RR models should be a component of a real-time flood forecasting system in small to medium catchments (500–5000 km 2 ) because, for such situations, a sufficient lead time cannot be obtained through the application of a flow routing model alone, i.e. a model allowing to simulate downstream water levels based on * Correspondence to: L. Brocca, Research Institute for Geo-Hydrological Protection, CNR, Via Madonna Alta 126, 06128 Perugia, Italy. E-mail: l.brocca@irpi.cnr.it upstream observations (Moramarco et al., 2005a). Addi- tionally, in absence of long enough discharge time series, for design flood estimation the application of a RR model is needed. In the scientific literature, a plethora of RR models are available, each one characterized by a different level of complexity and data requirement. RR models can be sub- divided as a function of their spatial structure (lumped versus semi-distributed or distributed), time represen- tation (continuous time versus event-based) or process description (physically meaningful versus data driven). A comprehensive compendium of presently available catchment models can be found in Singh and Woolhiser (2002); in addition Kampf and Burges (2007) recently reviewed and compared different spatially distributed RR models. Therefore, the large number of RR models now available makes the discussion on the value and the reli- ability of these models a topic of an increasing scientific interest (e.g. Beven, 2008; Sivakumar, 2008; Andreassian et al., 2009; Savenije, 2009) and for that, three important issues are addressed in this study: 1) Parameter identifiability (uncertainty). By applying RR models, different sets of parameters might pro- duce very similar model performance thus prevent- ing the ‘best’ parameter identification (Beven, 2008). Moreover, the high number of parameters incorpo- rated in some RR models can increase the model Copyright  2011 John Wiley & Sons, Ltd.