This article presents a field-based method to assess site- and rainfall-specific runoff coefficients to be expected for a given period of the year. The method is applied to recognize soil uses/covers leading to reduced runoff water supply of gullies in Kinshasa. The computation of the runoff coefficient needs an infiltration envelope, established on site during a period of interest, and a local pluviogram decomposed in pluviophases. Rainfall simulation is carried out in 35 representative urban sites located in gully runon areas to establish a site-specific infiltration envelope. The runoff coefficient of the 35 sites is calculated for 25 geomorphologically active rains recorded between 1975 and 2012. The results show that several site-specific characteristics control runoff coefficient. The first factor is the over-compaction of the soil. Earthen roads show a runoff coefficient of 96.0%. The second factor is the presence of a lichen seal. Bare loose soil only colonized by a lichen seal shows a runoff coefficient of 40.7%. For the other sites, the runoff coefficient is inversely proportional to the percentage of vegetation soil cover, a normally compacted bare soil having a runoff coefficient of up to 30%, parcels with high grass or cultures providing complete coverage showing no runoff at all. However, mowed lawns develop an impervious root mat close to the surface and, therefore, do not follow this rule: they quickly produce runoff similar to the bare and compacted surfaces. Finally the factor slope gradient is involved. The differences due to vegetation cover disappear gradually with decreasing slope. Below a slope gradient of 0.08 m m-1 the runoff coefficient is null on a bare surface. Currently the critical rainfall for gullying in the high town of Kinshasa is 24.9 mm with a mean intensity of 21.8 mm h-1. Roads generate by far most runoff, and therefore, are considered as the primary reason for gullying. The other soil uses lead most of the time to much smaller runoff coefficients but their relative contribution to the supply of gullies grows with rainfall increase in height and intensity. The results provide material for gully management and adaptation strategies and open perspectives for the development of an early warning system in the region of Kinshasa. The method shows potential for being applied in other urbanized environments.