Time Series Modeling of Jiroft Dam Inflow Using Tow Deep Learning models case study (Jiroft and Nesa Dams)

Sayari, Nasrin; Shahrokhi, Armin

doi:10.22059/ije.2025.398750.1879

Time Series Modeling of Jiroft Dam Inflow Using Tow Deep Learning models case study (Jiroft and Nesa Dams)

Articles in Press

Document Type : Research Article

Authors

nasrin sayari ¹
Armin Shahrokhi ²

¹ Assistant Professor Dept. of Irrigation(Water Engineering Group) Faculty of Agriculture Shahid Bahonar Univ. of Kerman Kerman IRAN

² Department of Water Science and Engineering, Faculty of Agriculture, Shahid Bahonar University of Kerman, Kerman, Iran

10.22059/ije.2025.398750.1879

Abstract

Analyzing temporal and spatial changes in precipitation, temperature, and surface runoff is essential in water resources discussions and long-term forecasts for watershed management. Therefore, by estimating the aforementioned variables, appropriate solutions can be provided to deal with future events. The aim of this study was to model the time series of inflows to Jiroft and Nesa dams under different climatic conditions. For this purpose, two LSTM and GRU models, which are deep learning methods, were used in the Python program environment. Based on the results of the output, the LSTM model in Jiroft dam performed more accurately according to the results of model convergence and the output of the evaluation criteria, while the GRU model performed about 45 seconds faster in each execution round, but in Nesa dam, the GRU model performed 56 seconds later. However, the results of the evaluation criteria indicated that this model was better than the LSTM model in time series prediction. As a result, it can be said that the larger the data interval, the LSTM method has a better response, and the smaller the GRU method has a better response. According to the results, it is possible to prevent damage caused by outlet .

Keywords