Flash floods are a leading cause of U.S. weather-related fatalities. Because flash floods happen quickly, timely information, communication, and decision making are critical to saving lives and reducing other negative impacts when a flash flood threatens. This poster discusses results from a project aimed at improving risk communication and warning response decisions during extreme weather events, with a focus on flash floods in Boulder, Colorado. Although Boulder has not experienced significant flooding in decades, its geography and meteorology place it at high risk for flash flooding; there have been devastating flash floods in the area, including the 1976 Big Thompson flood and the 1997 Fort Collins flood. One component of the project examines how members of three Boulder-area expert groups – NWS Weather Forecast Office forecasters, local public officials, and television and radio broadcasters – conceptualize flash flood risk (including exposure, effects, and mitigation), use and create information related to flash flood forecasts and warnings, and perceive how warning information is interpreted and used by other information providers and members of the public. Data for this component is based on a group decision modeling session conducted with NWS forecasters and individual mental models interviews with members of the three groups. Findings are summarized in the form of a decision-focused expert model of the flash flood forecast and warning system, along with an analysis of similarities and differences across the groups. A second component of the project examines how members of the Boulder public conceptualize flash flood risk and use flash flood warning information, based on analysis of data from mental model interviews with members of the public and comparison with the expert model. Results from the project will be used to help identify opportunities to improve warning generation, communication, and response.