A fuzzy rule-based system, designed with genetic algorithm (GA), is developed for estimation of model parameter values of a distributed rainfall-runoff model, which consists of a number of physically influenced model parameters. Input cell-based physical characteristics of the watershed and output estimated model parameter values are expressed by different fuzzy classes, such as 'high', 'medium', 'low', etc. These classes are described by membership functions, and their effects on parameter values are defined by fuzzy rules. The cell-based model parameters are estimated from an optimized fuzzy system, where GA searches for optimal membership functions. To estimate the model parameters, the minimum summation of square errors is used as an objective function. The estimated river discharges using proposed model parameter estimation system at both calibration and verification periods show a good matching with observed discharges, with the model efficiency criterion R_2 of 96.67 and 94.57%, respectively. The index of volumetric fit V_f and the mean sum of square error E are obtained within well-acceptable ranges. The effects of various cell characteristics are also reflected very clearly in the estimated model parameters. The developed system has great potential to serve as a tool to estimate model parameters and to improve the model performance.