Improving the estimation of simulated river discharge values using state space structural models

Document Type : Research Paper

Authors

1 PhD student in Water Resources, Department of Hydrology and Water Resources, Faculty of Water and Environmental Engineering, Shahid Chamran University of Ahvaz, Ahvaz, Iran

2 Associate Professor Department of Hydrology and Water Resources, Faculty of Water and Environmental Engineering, Shahid Chamran University of Ahvaz, Ahvaz, Iran

3 Department of Hydrology and Water Resources, Faculty of Water and Environmental Engineering, Shahid Chamran University of Ahvaz, Ahvaz, Iran.

4 Department of Statistics, Faculty of Mathematics and Computer Science, Shahid Chamran University of Ahvaz, Ahvaz, Iran.

Abstract

System simulation is done with different structures and using different approaches and algorithms. Algorithms are intelligent methods of data processing in machine learning that can identify unknown factors in a time-dependent phenomenon. In the analysis of random phenomena, including methods that can make decision-making easier; The use of ensemble algorithms. With the help of this method, more accurate data management and more knowledge of the studied system is obtained. Since investigating the trend component can be effective in simulating hydrological phenomena and help in interpreting the relationship between hydrological processes and environmental changes in the study areas; State space models have the advantage of analyzing the system flexibly and dynamically. Therefore, this article aims to improve the efficiency of Kalman Filter, ETS, BATS, and TBATS state space time series models with the help of a ensemble method and by comparing with the Box-Jenkins model, show which of these models has a better capability in simulating the monthly discharge of the river. This comparison has been done in three water measuring stations of Sepiddasht Cesar, TangPanj Bakhtiari and Telezang in Dez catchments located in Khuzestan province since (1386-1399). The results of this study, based on the model evaluation criteria (RMSE, MAE and R2), showed that the state space performed better than the Box-Jenkins model (classical), and among the state space models, the local level model (Kalman filter) performed better. So that in the validation stage, Sepiddasht Cesar water measuring station RMSE = 39.21, R2 = 0.79, in TangPanj Bakhtiari station RMSE = 57.89, R2 = 0.76 and in Telezang station RMSE = 113.41 and R2= 0.73 was obtained.

Keywords

Main Subjects