[Background] The surface runoff is one of the most significant components of hydrologic process. It's also the main force of soil erosion and transportation. As the process-based data for runoff is poorly accessible in comparison with that of rainfall, as well as runoff volume, the calculating of runoff rates based on accessible data turned out to be the key point for relative researches such as the evaluation of hydrologic progress and assessment on soil erosion and loss. [Methods] Aiming at acquiring event-based runoff rates as accurately as possible given rainfall progress and runoff volume which were relatively easy in terms of accessibility, three simply-structured infiltration models: known as Model I-constant infiltration capacity, Model II-constant runoff coefficient and Model III-spatial variable infiltration rates, were examined in this paper. Each of 3 models has only one parameter, with constant infiltration capacity known as index for Model I, self-explanatory runoff coefficient for Model II, and spatially average maximum infiltration rate for model III. The models were evaluated and compared based on 107 rainfall-runoff site events from 3 plots of Shejiagou watershed in Tuanshangou, Zizhou experiment station of the Yellow River Basin. To test the model efficiency in different time scales, process-based rainfall data was resampled at time intervals of 1 min, 6 min and 15 min separately, and consequently the outcome of runoff rates were obtained depending on the pattern of import data. [Results] The results,based on moderately 6-min data, showed that the model with constant infiltration capacity performed best with higher model efficiency (0.84) and lower mean absolute error (5 mm/ h) in predicting of peak runoff rates, compared with Model II(0.65, 6 mm/ h) and Model III(0.82, 6 mm/ h). The same pattern occurred to the calculation of effective runoff rates. Results for the distribution of relative errors in peak runoff rates and effective runoff rates by Model I were also acceptable as the error mainly distributed around zero with about 60% out of 107 site-events in the range from - 20% to 20%.According to the responses to different time intervals of data collecting, the estimation accuracy of Model 玉in simulating peak runoff rates was obviously improved as the time interval increased from 1 to 15 min,with model efficiency increasing from 0.69 to 0.92, and a decrease of mean absolute error from 17 mm/h to 2 mm/ h. And consequences of effective runoff rates showed a similar pattern, which might indicate that the application of the model was much appropriate in situation without large-density processed data.[Conclusions] The results can serve as providing efficient method of calculating runoff rates, and furthermore peak or effective runoff rates for areas lack of process-based runoff data, and are also conductive to the quantitative description and simulation of hydrologic process, soil erosion and transportation process.