Surface roughness and canopy cover are important factors in preventing soil erosion. There is limited information on how soil surface roughness changes as a function of natural rainfall erosivity and canopy cover by plants. We hypothesized that canopy cover, tillage systems, and cumulative rainfall erosivity (CRE) would have unique effects on roughness. We tested this hypothesis on a Miami silt loam soil (fine-silty, mixed, mesic Typic Hapludalf) using a portable laser microtopographer. Tillage treatments of conventional (moldboard plowing + disking), chisel plowing, and chisel plowing + dragging a chain produced three initial roughness levels. Surface cover was none (fallow) or soybean [Glycine max (L.) Merr.]. Random roughness (RR), standard deviation (SD), tortuosity (T), and fractal roughness functions, expressed by the fractal index (D) and the crossover length (l), were calculated from microtopography data. Chisel tillage had the greatest initial values of surface roughness, followed by chisel + chain and conventional tillage, as measured by the l index. All indices but D generally decreased with CRE. The RR and SD indices decreased quadratically with CRE, with decreases of 38 and 36%, respectively, from initial values after 200 units of CRE, while the T and l indices decreased exponentially, with decreases of 40 and 60%, respectively, from initial values after 200 units of CRE. Soybean cover lowered soil surface roughness 7% less than fallow, as measured by the l index. The l index was 50, 71, and 205% more sensitive to changes in CRE than RR, SD, and T indices, respectively. The fractal roughness functions, with D and l indices calculated, were the best approaches to characterize surface roughness at small scales, such as existing plant rows, mainly due to l index sensitivity to changes in CRE.