Comparison of the performance of stochastic models in the generation of synthetic monthly flows data: A case study on Marun river

One of the most important issues in planning and managing water resources is the accurate estimation of monthly input discharge of the reservoirs in the future years, which is always associated with uncertainty. To cover these uncertainties, synthetic stream flow data generation models have been used by various researchers to generate stochastic time series. The computational basis of different stochastic models for generating monthly data has been different and this can have a significant effect on their performance. Therefore, selection of the best model of stochastic data generation for accurate planning and management of a water resource system is one of the major concerns of water resources specialists. In this research, the performance of parametric models of synthetic stream flow generation including Thomas-Fiering, Fragment and ARMA (1,1) and ARMA (1,2) combined with Valencia-Schaake and Mejia and Rousselle models were compared and evaluated. For this purpose, 30 years data of monthly discharge of Marun river in Khuzestan province were used and 900 synthetic monthly flow time series were generated using each of the models mentioned above. Based on the obtained results, the ARMA (1,2) model combined with the Valencia-Schaake model was recognized as the best model, considering the very desired performance in preserving the statistical parameters of historical data and generating maximum and minimum discharges related to wet and dry periods in different probabilities. This model can be used with greater confidence to analyze river systems and reservoirs, manage drought and apply water rationing rules in future drought conditions.