Prediction of Daily Water Level Variations in Thurmond Reservoir Using Classical and Artificial Intelligence Methods

DOI: 10.22161/ijeab.114.9

Abstract:
Accurate prediction of reservoir water levels is essential for sustainable water resources management, reservoir operation, and water resources planning under increasing water demand and climate variability. In this study, the daily water level of Thurmond Reservoir, located on the Savannah River between South Carolina and Georgia, USA, was predicted using daily average air temperature (T), relative humidity (RH), precipitation (P), and one-day lagged reservoir water level [RWL(t−1)]. A total of 1608 daily observations collected between 2017 and 2023 were used, with 80% of the dataset allocated for training and the remaining 20% for testing. Multiple Linear Regression (MLR), Interaction Regression (IR), Quadratic Regression (QR), and Long Short-Term Memory (LSTM) models were employed to estimate daily reservoir water levels. Model performances were evaluated using the Coefficient of Determination (R²), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The results demonstrated that all developed models achieved satisfactory performance for daily reservoir water-level prediction. Among the evaluated models, the QR model achieved the highest prediction accuracy and the lowest error. The results obtained show that the proposed models can be effectively used for reservoir operation, water resources planning, and sustainable reservoir management.

Keywords:
Reservoir water level, daily water level prediction, multiple linear regression, LSTM, water resources management

Article Info:
Received: 18 Jun 2026; Received in revised form: 12 Jul 2026; Accepted: 16 Jul 2026; Available online: 22 Jul 2026

Cite this Article: