Using a Deep Learning Approach in Nutritional Index Analysis (Case Study: Karkheh Dam)

Bozorgi, Parisa; Javid, Amir Hossein; Allahyaribeik, Sara; Rahmati, Seyedeh Hoda

doi:10.22059/ije.2026.416051.1917

Using a Deep Learning Approach in Nutritional Index Analysis (Case Study: Karkheh Dam)

Articles in Press

Document Type : Research Article

Authors

Parisa Bozorgi ¹

Amir Hossein Javid ¹

Sara Allahyaribeik ²

Seyedeh Hoda Rahmati ¹

¹ Department of Environmental Engineering, SR.C., Islamic Azad University, Tehran, Iran

² Department of Physics, SR.C, Islamic Azad University ,Tehran, Iran

10.22059/ije.2026.416051.1917

Abstract

Subject: Nutrient orientation is one of the most important quality challenges in the exploitation of dam reservoirs, which can cause algae growth, reduced water clarity, decreased dissolved oxygen, and create restrictions on the use of water resources.
Objective: Simulation of the nutritive orientation index in the Karkheh Dam reservoir is a deep learning model including Recurrent Neural Network (RNN) and Long short-term memory (LSTM) in combination with two meta-heuristic optimization algorithms horse herd optimization (HOA) and Fire Hawk Optimizer (FHO).
Research method: The Carlson index was calculated at 6 monitoring stations and 4 different depths, and then its simulation was performed on a monthly scale using 10 different scenarios of input variables. The data used were from a five-year period, four years of which were considered for training and one year for testing the models.
Findings: Zoning results showed that the value of the nutritive orientation index decreased with increasing depth, and in the summer, due to increasing water temperature, stability of thermal stratification, and increased residence time, the value of the index increased. Also, stations with a greater distance from the dam construction site and the outflow showed higher values of the index due to greater water retention and the possibility of nutrient accumulation. The simulation results indicated the superiority of the LSTM-FHO model in estimating the eutrophication index. This superiority can be attributed to the ability of LSTM to learn long-term time dependencies and the ability of FHO to optimize model parameters. The study of input scenarios also showed that the parameters of dissolved oxygen, water temperature, surface flow entering the reservoir and water level are the most important effective variables in simulating the Carlson index.
Conclusion: Combining deep learning models with optimization algorithms can be an efficient approach for monitoring, simulating and managing eutrophication in dam reservoirs.

Keywords

Eutrophication

Carlson index

dam reservoir

deep learning

LSTM method

Subjects