With the increase in span, the flexibility and sensitivity to aerodynamic damage of bridges significantly increase. However, current research studies primarily focus on analyzing the single environment, leaving the safety and durability of bridges under extreme weather relatively understudied, such as snowdrift. This study selects four typical box girder bridges as test models to conduct wind tunnel tests. By analyzing similarity principles and particle properties, polyethylene particles are chosen as the test medium, and their simulation accuracy is validated using a step-type flat roof model. The study then explores wind-induced snow redistribution on the surfaces of the four box girder bridges under different horizontal wind speeds, initial particle heights, and wind attack angles. The results demonstrate that as the surface configurations of the test models become more complex, the particle redistribution becomes more chaotic. Among the three test conditions, the wind attack angle exerts the greatest influence, followed by horizontal wind speed and initial particle height. Notably, the dimensionless redistribution coefficients of particles on the surface of the large-span highway box girder bridge show the largest differences under these test conditions, with average differences and maximum differences reaching 50.3% and 63.5%, respectively, for different wind attack angles. These findings provide data support for the investigation of the safety and durability of real bridges under extreme weather conditions. [ABSTRACT FROM AUTHOR]