Growth rate is a crucial constituent parameter for determining basal entrainment rates in debris flow numerical modeling frameworks. This study presents the development of a multiple-regression model to estimate site-specific average growth rates in advance of debris flow event occurrences. A range of geomorphological and geotechnical variables influencing entrainment was selected through an in-depth literature review. A comprehensive dataset from 35 debris flow events in the Gyeonggi province of South Korea was employed to identify statistically significant variables related to the average growth rate. Through correlation analysis, three variables were ultimately applied as predictors for the multiple regression analysis: channel path length, basin area with slope inclinations of 30 % or greater, and median of soil grain-size distribution. The proposed model was statistically validated by demonstrating high coefficients of determination and meeting key assumptions of no-multicollinearity, no-autocorrelation, normality, and homoscedasticity. In a case study of a historical debris flow event that occurred during the 2011 Mt. Woomyeon landslide disaster, the practical relevance of the developed model was tested by comparing two parallel simulation results using the DAN3D numerical model. Different average growth rates were applied to the two parallel simulations, respectively; the target average growth rate calculated using field-measurement data and the predicted average growth rate derived by applying the multiple regression model. The results revealed that the proposed model provided reasonable predictions of debris flow velocities and heights, with only minor discrepancies between the target and predicted average growth rates. Notably, measured average growth rates of immediately adjacent debris flow events significantly differed from that of the case study event. This highlights that adopting entrainment data from neighboring events for debris flow simulations can result in significantly erroneous debris flow hazard predictions. Finally, the limitations and contributions of this study and future research directions are discussed.