The number of deaths due to flash floods in China has increased sharply in recent decades. Therefore, a quantitative assessment of loss of life at the potentially affected areas is highly recommendable, which can improve the strategies of disaster risk mitigation and reduction. Numerous papers published to date have focused on human vulnerability to river floods but seldom to flash floods because of the lack of data on loss of life or flood intensity matched with loss of life. This research aims at reconstructing the July 2016 flash flood event by combining hydro-hydraulic models with post-flood surveys to obtain the data of flood intensity and then developing human vulnerability curves for flash floods through statistical functions. The hydrological model HEC-HMS was used to obtain the hydrograph of the July 2016 flash flood, which was evaluated using the flow data obtained from references. Subsequently, the hydraulic model FLO-2D was used to reconstruct the July 2016 flash flood event to obtain the flood characteristics (e.g., water depth, flow velocity, depth-velocity product, and impact force). The evaluation of FLO-2D was carried out using post-flood surveys defining maximum flooded extent and flood depths. Then, statistical functions were used to fit the relationship between the flood characteristics and mortality calculated through post-flood surveys. The results show that the simulated peak discharge (548 m3/s) of the July 2016 flash flood for Taitou catchment was close to the estimated peak discharge (540 m3/s) by the existing literature. Likewise, the FLO-2D model showed a good performance in reconstructing the flash flood event (FitA = 80.89%, RMSE = 0.21, NSE = 0.89), indicating that the flood characteristics outputting from FLO-2D model can be able to develop the human vulnerability curves for flash floods. Furthermore, four different human vulnerability curves for flash floods were developed, which are all power functions of water depth, flow velocity, depth-velocity product, and impact force, respectively. The mortality caused by flash floods increased with all four flood characteristics; the growth rate of the velocity curve was the fastest. However, the fitting effect of the velocity curve was the poorest (R2 = 0.87, SSE = 0.155, RMSE = 0.18), while that of the water depth curve was the best (R2 = 0.97, SSE = 0.003, RMSE = 0.03). Our findings provide a quantitative way to quickly assess loss of life due to flash floods, which is more likely to have a good performance in arid and semi-humid regions of northern China.