Predicting sea levels using ML algorithms in selected locations along coastal Malaysia

Area of Science:

• Environmental Science
• Data Science
• Climate Science

Background:

• Sea level rise is a significant consequence of climate change.
• Accurate daily sea level prediction is crucial for coastal management and mitigation strategies.
• Traditional methods may not fully capture the complex dynamics of sea level changes.

Purpose of the Study:

• To evaluate the performance of six distinct machine learning (ML) algorithms for daily sea level prediction.
• To identify the optimal ML model for specific locations in Malaysia.
• To determine the impact of data lag on prediction accuracy.

Main Methods:

• Utilized sea level data from 1985 to 2018 for model training and testing.
• Applied six different ML algorithms, including linear regression, interactions linear regression, and Gaussian process regression.
• Assessed model performance using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R-square metrics.

Main Results:

• Different ML models demonstrated superior performance at various study locations in Malaysia (e.g., Pulau Langkawi, Geting, Pulau Pinang, Sandakan).
• The use of a 7-day lag in sea level data consistently enhanced prediction accuracy across all tested models.
• Models utilizing less than a 7-day lag showed reduced accuracy, indicating the importance of temporal data patterns.

Conclusions:

• Machine learning models, when properly trained and tested, offer reliable tools for predicting daily sea levels.
• The optimal ML model is location-specific, necessitating tailored approaches.
• Incorporating a 7-day lag in sea level data is critical for improving the accuracy and reliability of ML-based predictions for climate change adaptation.