Research Information System
Natural Hazards, 2025 (SCI-Expanded)
In recent decades, climate change has emerged as a significant issue, causing ongoing increases in ocean and atmospheric temperatures. This trend indicates that sea levels are expected to rise at faster rates in the future compared to the present sea level. Ongoing increases in sea levels could potentially trigger catastrophic natural disasters worldwide. So that reason, predicting sea level rise (SLR) is crucial for future planning in areas such as human living conditions, flood prevention, and coastal development. This study focuses on evaluating the ability of conventional and deep learning time series methods such as seasonal autoregressive integrated moving average (SARIMA), long short-term memory (LSTM) neural network, and gated recurrent unit (GRU) in estimating the current and future global mean SLR. The models were trained and tested using monthly SLR data collected between 1993 and 2023 and then future predictions were made until 2050. A total of 366 monthly SLR data were used where 288 SLR data (78%) from January 1993 to December 2016 were utilized in the training phase and 78 SLR data (22%) from January 2017 to June 2023 were utilized in the testing phase. The findings demonstrate that although the SLR values estimated with the all developed models are closely compromised with real SLR values in the testing phase, the LSTM model provides more precise predictions than the others. While MAPE, MAE and RMSE parameters for the prediction of SLR data with the LSTM model are observed as 0.0631%, 0.0058 cm and 0.0073 cm, respectively, they are detected as 0.0899%, 0.0084 cm and 0.0109 cm with GRU model. On the other hand, in the testing stage, the worst estimation is performed with the SARIMA model with a MAPE of 0.1335%, MAE of 0.0123 cm and RMSE of 0.0155 cm. Furthermore, the LSTM algorithm which is observed as the best accurate model, predicts the SLR values as 17.218 cm by July 2040 and 21.236 cm by July 2050 when the global average sea level at the beginning of 1993 is taken as reference. Therefore, the developed algorithms show potential as effective tools for modeling and predicting global mean sea level rise in the future and this study will be highly valuable for decision-makers in creating mitigation strategies for sea level rise associated with climate change, utilizing the developed models.