Traditional vegetation techniques for the control of concentrated flow erosion are widely recognized, whereas only a few studies have experimentally investigated the impacts of belowground roots on the erodibility of topsoils in semi-arid areas. To quantify the effects of root architectures on soil erodibility and its relevant structural properties, simulated flow experiments were conducted at six-week intervals from 18 July to 20 October in 2012 in the hilly Loess Plateau. Five treatments were: 1) bare(control), 2) purple alfalfa(Medicago sativa), representing tap roots(T), 3) switchgrass(Panicum virgatum), representing fibrous roots(F), 4) purple alfalfa and switchgrass, representing both tap and fibrous roots(T + F), and 5) natural recovery(N). For each treatment, soil structural properties and root characteristics were measured at an interval of six weeks. Soil anti-scouribility was calculated. Results showed that grass planting slightly reduced soil bulk density, but increased soil aggregate content by 19.1%, 10.6%, 28.5%, and 41.2% in the treatments T, F, T + F, and N, respectively. Soil shear strength(cohesion and angle of internal friction(φ)) significantly increased after the grass was planted. As roots grew, soil cohesion increased by 115.2%–135.5%, while soil disintegration rate decreased by 39.0%–58.1% in the 21 th week compared with the recorded value in the 9th week. Meanwhile, root density and root surface area density increased by 64.0%–104.7% and 75.9%–157.1%, respectively. No significant differences in soil anti-scouribility were observed between the treatments of T and F or of T + F and N, but the treatments of T + F and N performed more effectively than T or F treatment alone in retarding concentrated flow. Soil aggregation and root surface-area density explained the observed soil anti-scouribility during concentrated flow well for the different treatments. This result proved that the restoration of natural vegetation might be the most appropriate strategy in soil reinforcement in the hilly Loess Plateau. Traditional vegetation techniques for the control of concentrated flow erosion are widely recognized, and only only a few studies have experimentally investigated the impacts of belowground roots on the erodibility of topsoils in semi-arid areas. To quantify the effects of root architectures on soil erodibility and its relevant structural properties, simulated flow experiments were conducted at six-week intervals from 18 July to 20 October in 2012 in the hilly Loess Plateau. Five treatments were: 1) bare (control), 2) purple alfa alfa (Medicago sativa), representing tap 3) switchgrass (Panicum virgatum), representing fibrous roots (F), 4) purple alfalfa and switchgrass, both both and fibrous roots (T + F), and 5) natural each treatment, soil structural properties and root characteristics were measured at an interval of six weeks. Soil was screened slightly reduced soil bulk density, but increased soi Soil aggregates by 19.1%, 10.6%, 28.5% and 41.2% in the treatments T, F, T + F, and N, respectively. Soil shear strength (cohesion and angle of internal friction (φ)) significantly increased after the grass was planted. As roots grew, soil cohesion increased by 115.2% -135.5%, while soil disintegration rate decreased by 39.0% -58.1% in the 21th week compared with the recorded value in the 9th week. Meanwhile, root density and root Surface area density increased by 64.0% -104.7% and 75.9% -157.1%, respectively. No significant differences in soil anti-scouribility were observed between the treatments of T and F or T + F and N, but the treatments of T + F and N performed more effectively than T or F treatment alone in retarding concentrated flow. Soil aggregation and root surface-area density explained the observed soil anti-scouribility during concentrated flow well for the different treatments. This result proved that the restoration of natural vegetation might be the most a ppropriate strategy in soil reinforcement in the hilly Loess Plateau.