Stream flow is an important factor in reservoir utilization management. Using stochastic methods, simulation and forecasting of streamflow can be performed at a yearly scale. In a monthly scale, the number of parameters in these methods is so highly increased that will cause a considerable error of simulation. A new approach for stream flow simulation was presented by Sharma et al. (1997). In this method, stream flow is assumed to be a random variable, previous time dependent characterized by a joint probability density function. Kernel method is used to estimate this function and simulation is done using the conditional probability density function derived from the kernel estimate of the joint density function. In this research, a computer code has been developed based on the above mentioned method. To evaluate this model to a real world conditions, Karoun reservoir monthly inflow data have been used. Method for choosing the best simulation is based on the evaluation of mean absolute error (MAE) factor. A comparison between the observed and synthetic stream flow data shows a satisfactory and good relation. With these results obtained, this approach can be proposed to simulate the monthly stream flow at other reservoir dam sites.
)