Soil detachment capacity (Dc) is an important parameter used to determine erosion intensity in physical-processbased erosion models. Freeze-thaw affects soil detachment processes by altering the mechanical properties of soil; however, due to the compound action of freeze-thaw and runoff on D-c, quantifying the impact of seasonal freeze-thaw on D-c remains challenging. A series of experiments with six freeze-thaw cycles (FTC), six initial soil moisture contents (SMC), three slope gradients, and five flow discharges were conducted to investigate the effect of freeze-thaw and hydrodynamic characteristics on D-c. The results showed that soil shear strength (tau(m)), cohesion (Coh), and internal friction angle (phi) gradually tended to become stable with increasing FTC, indicating that repeated FTC had a cumulative impact on soil mechanical properties, and there was a critical FTC between 5 and 7. When FTC rose from 1 to 15, the reduction in tau m, Coh, and phi was 0.03-23.96%, 2.63-75.21%, and - 5.70-19.24%, respectively, which increased with an increasing SMC, suggesting that the deterioration effect of FTC on soil mechanical properties was promoted by increasing SMC. During alternating FTC, the relative range and variation coefficient of D-c were 2.21-2.43 and 67.87-75.72%, respectively, indicating that D-c was highly sensitive to FTC. Furthermore, D-c increased by 2.37-71.22% after 15 FTC. Alternating freeze-thaw weakened the soil resistance to detachment. Moreover, the promoting effect of FTC on D-c intensified with an increasing SMC, indicating that the variation in D-c was strongly affected by SMC during FTC. A prediction model (R-2=0.955, RRMSE=14.99%) was established to quantify the influence of freeze-thaw and hydrodynamic characteristics on D-c. The explanation rate of variables in the D-c prediction equation was quantitated: the explanation rate of stream power (64.3%) was higher than that of FTC (10.02%) and SMC (3.92%), suggesting that the impact of freeze-thaw on D-c was covered by hydrodynamic characteristics. Further validation is required for the prediction equations when applied beyond the range of construction conditions.