Effects of soil surface roughness on runoff and sediment production have not been clearly quantified, mostly due to the lack of a logical separation between geometric (i.e., surface microtopography) and process (i.e., runoff generaion, soil detachment by raindrop and runoff) scales.   In this research, we separated soil roughness, based on its functional form, into depressions and mounds due to their converging and diverging nature and satudied their effects under laboratory rainfall simulation.  Our results showed that the only significant roughness effect is the delayed runoff initiation from surface depresions.  At steady state, surface with depresions produced slightly greater runoff, possibly due to more water on the surface and sealing of the depressions.   At steady state runoff, there is no difference in sediment produciton between surfaces with depressions or mounds and those without.   Current erosion prediction models, i.e, RUSLE and WEPP, predicted a reduced soil erosion with an increased soil roughness.  This predicition is only partially correct when rainfall is not sufficient enough to produce full runoff.