As one of the most hazardous natural events, flash flood is frequently responsible for loss of life and severe damage to infrastructure and public health. The Rainfall Threshold (RT) is the main warning index of flash flood that is used in China. However, the amount of parameters in RT calculation is significantly large, and the determination of some parameter values is affected by subjective factors substantially. In this paper, a Morris-Sobol Two-layer Progressive Model (M-STPM) is established to identify sensitive parameters in RT calculation. Firstly, the Morris screening method is applied to qualitatively analyse the sensitivity of parameters, while the insensitive parameters are filtered. Then the Sobol method is employed to quantify the sensitivity. The sensitivity of parameters in RT calculation is determined through a case study in Duli Village of Anyang County in Henan Province. The results indicate that mean of 1-h annual maximum point rainfall is the most influential parameter. It is found that the parameters in water level-discharge relation calculation have a large influence on the RT, which shows that the proposed model can provide a reference for determining the value of parameters and can lay a theoretical foundation for further error analysis in RT calculation. Palabras clave: inundación repentina, umbral de precipitación, parámetro de sensibilidad, modelo progresivo Morris-Sobol de dos capas, método inverso del nivel de descarga de agua RESUMEN Las inundaciones repentinas, uno de los eventos naturales más peligrosos, son con frecuencia responsables de la pérdida de vidas humanas y de daños a la infraestructura y a la salud pública. El umbral de precipitación (RT, por sus siglas en inglés) es el principal índice de alerta que se utiliza en China. Sin embargo, la cantidad de parámetros para el cálculo del RT es muy grande y la determinación de los valores de algunos de