The beneficial effects of plant roots in decreasing soil detachment in forest ecosystems exposed to rill erosion are well known. However, these effects vary largely between different plant species. There has been lots of research into the relationship between root-soil systems and rill erodibility with a particular focus on grass species. Conversely, fewer studies are available for tree species, especially in forests of semi-arid or arid environments. Greater knowledge is therefore needed to identify the most effective tree species against rill erosion in these ecosystems, where water availability is the limiting factor for vegetation growth and afforestation is often the only solution to control erosion. To fill this gap, this study evaluates the rill detachment capacity of soils with four tree species (Parrotia persica, Carpinus betulus, Quercus castaneifolia, and Pinus taeda) in a semi-arid forest ecosystem in Northern Iran. These species are typical of these forests, but they also grow in other environmental contexts. The rill detachment was simulated in a laboratory flume at five slope gradients (1% to 5%) and five flow rates (0.22 to 0.69 L m−1 s−1) on soil samples with each of the tree species. The specific goal of the study was to evaluate which tree species with its specific root characteristics is most effective at reducing the rill detachment capacity. Moreover, simple prediction models are proposed to evaluate if it is possible to estimate the rill detachment capacity and rill erodibility (Kr) from the unit stream power for the investigated tree species. The soils with Parrotia persica and Carpinus betulus showed the lowest and the highest rill detachment capacity, respectively. The greater root system biomass of Parrotia persica could have played a binding effect on the soil, thus improving its aggregate stability thanks to the action of plant's root system. Based on these results, Parrotia persica is able to provide a higher soil protection capacity against erosion compared to the other tree species. A logarithmic function was accurate in predicting the rill detachment capacity from the unit stream power at water flow rates over 0.0025 m s−1. By a regression between the rill detachment capacity and the shear stress of the soil, rill erodibility and critical shear stress of soils were estimated for the four tree species; the rill erodibility and critical shear stress are important input parameters for physically-based erosion models. Overall, the results of this study can support land planners in the choice of tree species most indicated for soil conservation as well as in the extensive application of erosion prediction models.