AbstractThe Hydrologic Engineering Center – Hydrologic Modelling System (HEC‐HMS) model has been widely applied for estimating hydrological variables at event scale. The choice of the most suitable infiltration method simplifies model's applicability under different environmental conditions. A proper case study for evaluating the runoff prediction capability of HEC‐HMS by the available infiltration methods is the semi‐arid torrents typical of Southern Italy as they are small and intermittent water courses, often subject to high‐magnitude flash floods and erosive events. HEC‐HMS performance of the ‘SCS‐CN’, ‘Green‐Ampt’ and ‘Initial and Constant’ infiltration methods in predicting runoff volume and peak flow was evaluated at the outlet of the Mésima torrent, Calabria, Southern Italy. Fourteen rainfall–runoff events were simulated by HEC‐HMS and compared with the corresponding observations by a quantitative approach. A good accuracy in predicting runoff volume was achieved using the ‘SCS‐CN’ method after calibration of the initial curve numbers. Peak flow was better estimated using the ‘Initial and Constant’ method, also thanks to calibration of the ‘constant rate’ parameter. The calibrated hydrographs were very similar to the observations for both ‘SCS‐CN’ and ‘Initial and Constant’ methods. Adopting the ‘Green‐Ampt’ equations, however, showed low reliability. The evaluation of the time to flood peak was in some cases inadequate. Copyright © 2016 John Wiley & Sons, Ltd.