Flash floods are one of the most significant natural hazards of today. Due to the complexity of flash flood triggering factors, to prevent or mitigate flood triggered losses, numeric model based flood forecasting models are capable tools to predict stream water levels. The main goal of the current research was to reproduce two flow peaks with the HEC-HMS rainfall-runoff model and test the model sensitivity for various input parameters. To obtain sufficient input data, we monitored soil depth, maximum infiltration rate, soil moisture content, rainfall, time of concentration and flow. To obtain input data, parameters were calculated, measured in the Sás Valley experimental watershed (SW Hungary) or optimized with the built in function of the HEC-HMS. Soil moisture was monitored in the 1.7 km2 pilot catchment over the period between September 2008 and September 2009. HEC-HMS had a good performance reproducing the two events, however simulated flow time series are highly influenced by the antecedent soil moisture, infiltration rate and canopy storage. Outflow modeled data were verified for two flood events (June 4, 2008 and July 9, 2009). The HEC-HMS was over-sensitive for input soil moisture and with increasing input rainfall and increasing outflow, larger simulation errors were observed. [ABSTRACT FROM AUTHOR]