To quantitatively simulate the particle size distribution (PSD) of eroded sediment on grass and root slopes, simulated rainfall experiments were implemented on grass slopes with intact plants and root slopes with the aboveground parts removed. There were four coverage degrees (Cr) (0, 20, 40, and 60%), two rainfall intensities (R) (60 and 90 mm h–1) and two slope gradients (J) (15° and 25°). The content of sand‐sized particles was greatest in the eroded sediment from the grass slope, whereas the content of fine silt‐sized particles was greatest in the eroded sediment from the root slope. The PSD of eroded sediment was influenced by Cr, R, J, and hydrodynamic characteristics [flow velocity (u), flow depth (h), Reynolds number (Re), Froude number (Fr), resistance coefficient (f), runoff shear force (t), runoff power (w), unit runoff power (P)]. Through grey relational analysis, the relational degree (γ) of these impact factors to the PSD of eroded sediment was ranked γR > γRe > γJ > γh > γw > γu > γt > γP > γFr > γCr > γf on the grass slope and γR > γRe > γJ > γh > γt > γw > γu > γFr > γp > γf > γCr > on the root slope. The radial basis function neural network performed well at simulating the PSD of eroded sediment on both the grass and root slopes. These results lay the foundation for the establishment of a physics‐ based soil erosion model and a chemical transport model.