Review of machine learning methods for sea level change modeling and prediction

Area of Science:

• Environmental Science
• Climate Science
• Data Science

Background:

• Sea level change (SLC) is a critical consequence of global climate change, posing significant risks to coastal communities worldwide.
• Effective coastal management and adaptation strategies necessitate accurate and timely forecasting of SLC.

Purpose of the Study:

• To review and assess methodologies for developing robust machine learning (ML) models for predicting and forecasting sea level change.
• To evaluate the performance of different ML techniques and identify key factors influencing model accuracy.

Main Methods:

• Comprehensive literature review of machine learning approaches applied to sea level change prediction and forecasting.
• Analysis of various ML models, including artificial neural networks (ANNs), deep learning, regression, and support vector machines.
• Examination of the impact of input variable selection and data partitioning on model performance.

Main Results:

• Deep learning models and hybrid ANNs demonstrate superior performance in short-term sea level anomaly prediction compared to simpler ML methods.
• Supervised learning is prevalent, while semi-supervised methods show efficacy in short-term projections.
• Model accuracy is significantly influenced by the selection of atmospheric, oceanic, and geological input variables and the training/testing data balance.

Conclusions:

• Machine learning offers powerful tools for sea level change prediction, with advanced ANNs excelling in short-term anomaly forecasting.
• Future research should prioritize integrating physics-based general circulation models (GCMs) with ML techniques for enhanced regional long-term forecasting crucial for coastal resilience.