A hybrid decomposition-based Machine Learning approach for predicting subsurface temperature in the Arabian Sea

Ocean warming has emerged as a pivotal research topic in the field of climate science. With increasing global warming the surface as well as the Subsurface Temperature (ST) of the global oceans have exhibited a steady increasing trend over the last few decades. This paper presents a hybrid method to...

Full description

Saved in:
Bibliographic Details
Published in:Modeling earth systems and environment 2024-12, Vol.10 (6), p.7295-7314
Main Authors: Malavika, A. R., Pai, Maya L., Johny, Kavya
Format: Article
Language:English
Subjects:
Chemistry and Earth Sciences
Climate change
Climate prediction
Climate science
Computer Science
Correlation
Correlation analysis
Datasets
Decomposition
Earth and Environmental Science
Earth Sciences
Earth System Sciences
Ecosystems
Environment
Global warming
Heat
Heat flux
Heat transfer
Humidity
Learning algorithms
Machine learning
Math. Appl. in Environmental Science
Mathematical Applications in the Physical Sciences
Ocean temperature
Ocean warming
Oceanic analysis
Oceans
Original Article
Parameters
Physics
Radiation
Rain
Regions
Residues
Sea level
Sea level pressure
Sea surface temperature
Statistics for Engineering
Trends
Weather
Wind
Wind speed
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Ocean warming has emerged as a pivotal research topic in the field of climate science. With increasing global warming the surface as well as the Subsurface Temperature (ST) of the global oceans have exhibited a steady increasing trend over the last few decades. This paper presents a hybrid method to predict the ST of the Arabian Sea by following a decomposition-based Machine Learning (ML) regression. The study considers Sea Level Pressure, humidity, wind speed, Sea Surface Temperature and heat fluxes which impact the ST. The proposed methodology first uses the Akaike Information Criterion to determine the most significant parameters influencing the ST. Next, Empirical Mode Decomposition is applied to each of the selected parameters, forming a dataset of corresponding Intrinsic Mode Functions (IMFs) and residues. Redundant IMFs and residues are filtered out by Spearman’s correlation function. The newly formed compact dataset then undergoes ML regression. This hybrid methodology has demonstrated improved accuracy in predicting the ST at the depths of 5 m and 25 m.
ISSN:2363-6203
2363-6211
DOI:10.1007/s40808-024-02167-0