Predicting sea level variations for early warning using SARIMA model and deep learning techniques in the northwest Arabian Gulf

Abstract Predicting sea level variations (SLV) in coastal areas is essential for flood warnings, environmental protection, and infrastructure management. While sea level forecasting has been studied in various parts of the Arabian Gulf, limited research exists on water level prediction for Iraqi coastal waters. This study applies traditional statistical models, including the Autoregressive Integrated Moving Average (ARIMA) and Advanced Seasonal ARIMA (SARIMA), alongside advanced deep learning techniques, such as Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and a hybrid CNN-LSTM model, to predict sea level variations in the northwest Arabian Gulf. The results demonstrate that ARIMA provides poor predictions, while SARIMA offers better forecasts with MSE = 0.0265, RMSE = 0.1626, and MAE = 0.1288. However, deep learning models, particularly CNN-LSTM, significantly outperform traditional models, with CNN achieving MSE = 0.0191, RMSE = 0.1384, and MAE = 0.1126, LSTM achieving MSE = 0.0172, RMSE = 0.1311, and MAE = 0.1055, and CNN-LSTM achieving MSE = 0.0165, RMSE = 0.1282, and MAE = 0.1015. This research highlights the potential of deep learning techniques for more reliable sea level predictions and improved flood risk management, emphasizing their advantage over conventional statistical methods.