Abstract: Conventional approaches of predicting debris flow runout distance are predominantly dependent on establishing empirical formulas through static analysis. Such approaches can not explain why debris flows with identical volumes and initial elevations may exhibit markedly different runout distances, as they largely neglect the influence of dynamic parameters on runout. To address this limitation, a series of flume experiments were conducted. Under fixed initial elevation and flume slope, a controlled-variable methodology was adopted to investigate how solid volume fraction, particle size, and total volume affect the dynamic parameters of debris flows (flow velocity and flow depth). The effects of velocity and depth on runout distance were then examined, and a machine-learning-based predictive model for debris flow runout distance was developed based on velocity and depth. The results show that: ① Flow velocity is negatively correlated with both solid volume fraction and median particle size (D₅₀), with correlation coefficients of -0.68 and -0.23, respectively; the influence of debris-flow volume on velocity is limited. ② Flow depth is positively correlated with volume, solid volume fraction, and D₅₀, with correlation coefficients of 0.51, 0.61, and 0.31, respectively. ③ Increasing solid volume fraction promotes a transition from collisional and frictional forces to viscous resistance, whereas increasing D₅₀ enhances internal particle collisions and generates dispersive stress, thereby suppressing velocity and increasing flow depth. These findings demonstrate that debris-flow runout distance is jointly governed by flow velocity and flow depth, with velocity exerting a more pronounced effect, which can improve the accuracy of runout distance estimation.