A Comparison among the Performance of the Stochastic Models in Generating the Monthly Streamflow and Rainfall Data

Synthetic data generation models have been recognized as useful tools to predict and generate alternative time series or long-term series throughout the studies conducted in the domain of water resources management. Accordingly, these models have widely been used by different researchers across the world. In the recent decades, these models have been developed to generate annual, monthly, and daily rainfall or river flow data. Among the synthetic data generated, monthly data are of great importance since they are used in the critical and important studies in the field of water resource systems, such as storage tanks and drought monitoring. Accordingly, the utilization of the monthly synthetic data models leads to more detailed analyses about the real performance of such systems. On the other hand, the theoretical basis of different stochastic models is the generation of diverse monthly data and the performance of these models can remarkably be affected by this fact. Therefore, one can argue that selecting an appropriate model is one of the major concerns of water resources experts. As such, this study made use of the Monte Carlo simulation method to compare and assess the performance of four types of non-parametric Bootstrap models as well as parametric models of Valencia-Schaake, Thomas-Fiering, and Fragment in monthly synthetic data generation. To do this, the monthly flow data of Nazluchay, Shaharchay, and Barandozchay rivers, located at the Western Azerbaijan province in the North West of Iran, were analyzed over a 47-year period. Then, 1000 synthetic time series of monthly flows for these rivers were generated and used for each of the desired seven models over a 47-year period thereof. The results indicated that, compared to other models, the Valencia-Schaake distribution model had a very high performance in terms of all well-known assessment statistical indicators.