The wind erosion process on agricultural soils is being modeled as the time-dependent conservation of mass transport of soil moving as saltation and creep. Emission of loose soil and abrasion of clods and crust act as sources, whereas trapping and suspension act as sinks for the moving soil. In this study, an expression for the abrasion source term was derived. Abrasion flux from aggregates or crust was shown to be the product of three variables - fraction of saltation impacting the target, an abrasion coefficient, and saltation discharge. Various aspects of the proposed abrasion source term were then investigated in three wind tunnel studies. First, crusted trays were abraded using a range of windspeeds and sand abrader rates. Regression analysis showed there was no significant relationship between crust abrasion coefficients and fraction of abrader moving below 0.1 m (i.e. abrader trajectories). This result shows practical abrasion coefficients can be developed which depend only on the properties of the target and abrader. Second, a relationship was developed to predict fraction of saltation impacting surface aggregates (or intervening crust) as a function of surface aggregate cover and roughness. The relationship was tested in the tunnel by abrading crusted trays partially covered (0 to 30%) with non-abradable aggregates. Regression analysis showed there was good agreement (R2 = 0.97) between observed and predicted fraction of abrader impacting aggregates. In the third experiment, trays were filled with various mixtures of large- and saltation-size aggregates. The trays were abraded in the wind tunnel by a low saltation discharge from a narrow upwind aggregate bed. The results showed that, because the wind transport capacity significantly exceeded the saltation discharge, the surface tended to armor with large aggregates. In this case, the fraction of abrader impacting large aggregates was not significantly different from one for a wide range of aggregate mixtures.