Investigation of Mixed Layer Depth through the lens of Artificial Intelligence

Abstract

Predicting the various oceanic parameters responsible for air-sea coupling is crucial to understanding how the climate and weather systems can affect the ecosphere. One of the most important among these oceanic parameters is the Mixed Layer. In this study, the convolutional long short-term memory(ConvLSTM) based Neural Network(NN) architecture is used for monthly forecasting of Mixed Layer Depth(MLD) in the Bay of Bengal(BOB) region. The study uses multi variables corresponding to other prominent ocean surface phenomena as input and the AI model is used to learn and understand the link between these input variables and the output variable MLD. This study forecasts the MLD with a correlation better than the operational dynamical Hindcast model and the ablation study suggests a decline in performance when any of these 5 input variables were removed from the training. The study not only deciphers the relationship between these variables and the MLD but also opens an interesting field to explore the forecasting of other ocean phenomena which directly or indirectly depend on the MLD.