Operational use of machine learning models for sea-level modeling

Abstract

Intense activity offshore warrants a temporal and accurate prediction of sea-level variability. Besides, the sea-level plays an important role in the groundwater level and quality of coastal aquifer. Climate change influences considerable change in all the hydrological parameters and apparently affects sea-level variability. For prediction, highly complex numerical models are usually generated. To address these challenges, the study proposes the use of machine learning (ML) models with the climate change predictands and sea-level predictors. Three ML models are employed in this study, viz., Regression Vector Machine (RVM), Extreme Learning Machine (ELM), and Gaussian Process Regression (GPR). The performance of the developed models is evaluated by visual comparison of predicted and observed datasets. Regression error curve plots, frequency of forecasting errors and Taylor diagram, along with statistical performance metrics were developed. Overall, it is found that the operational use of the selected ML algorithms was quite appealing for modeling studies. Among the three ML models, GPR performed slightly better than ELM and RVM.