Soil erosion by rainfall is a process that demands management, both for the prevention of excessive soil erosion and for the protection of the quality of freshwater bodies. Erosion coefficients (K-factors) of the universal soil loss equation (USLE)/revised USLE (RUSLE) model were assigned to 99 mapped Uruguayan soil types at 1: 1,000,000 scale. This work developed a linear mixed model (LMM) with 79 soils with assigned K-factors, in which the following variables were considered: soil taxonomy, chemical composition, and parent material. The developed LMM had an R-2=0.86, in which the soil taxonomy (p<0.0001), parent material (p=0.0174), clay (p=0.0005) and sand (p=0.017) contents had significant statistical effects. The prediction capacity of this model was assessed with 10 soils not previously used in development of the LMM with assigned K-factors. The prediction assessment had an R-2=0.84 and a mean error of 9.08% of the mean K-factor value. The LMM developed was used for the allocation of K-factors to soils mapped at a 1: 20,000-resolution. Thus, the use of LMM increased the soil area with assigned K-factors from 111,822 km(2) (at a scale of 1: 1,000,000) to 174,132 km(2) (1: 20,000).