Ocean subsurface temperature prediction using an improved hybrid model combining ensemble empirical mode decomposition and deep multi-layer perceptron (EEMD-MLP-DL)

Ocean Subsurface Temperature (ST) has emerged as a critical factor in understanding global climate change. The penetration of warming signals from the oceanic surface to the deeper layers of oceans necessitates the development of prompt and effective predictive strategies for climate modelling. Recognizing the critical role of ocean ST in climate across the globe, the paper employs a hybrid approach integrating Ensemble Empirical Mode Decomposition (EEMD) and deep Multi-Layer Perceptron (MLP) to predict the ST at different depths. The study utilizes both ocean and atmospheric parameters like sea surface temperature, humidity, pressure, wind speed and heat fluxes, providing a comprehensive framework for assessing the intricate relationships between ocean layers and the atmosphere. The paper compares two methodologies: EMD with MLP of Single Layer (EMD-MLP-SL) and the proposed model EEMD with MLP of Deep Layers (EEMD-MLP-DL) for predicting the ST in the Arabian Sea for depths ranging from 5m to 967m. The results highlight the improved predictive capabilities of the proposed EEMD-MLP-DL methodology, achieving up to 95 % accuracy at 5m depth and maintaining robust performance at every depth, in contrast to the EMD-MLP-SL model. This paper highlights the importance of multifaceted approaches in oceanographic modelling and emphasizes the inclusion of more oceanic and atmospheric factors in understanding climate variability.