Global Surface Ocean Total Alkalinity Estimation: The Machine Learning Approach

Exploring the intricacies of ocean acidification (OA), this study conducts a comprehensive investigation of global surface ocean total alkalinity (TA) estimation using advanced machine learning (ML) models. The global surface ocean quality-controlled in situ measurements, such as sea surface salinity (SSS), sea surface temperature (SST), and nitrate (Ni), acquired from the Global Ocean Data Analysis Project (GLODAP) were utilized for the study. We assess the performance of feed-forward neural networks (FFNNs), random forests (RFs), and the tabular net (TabNet) model. The findings specifically highlight TabNet as the most efficient model, outperforming both RFs, which exhibit similar but not superior performance, and FFNN, which demonstrate the least effectiveness. TabNet emerges as a robust model, illustrating superior adaptability to complex patterns, surpassing the comparable performance of RF model. These findings improve our understanding of global TA variations and highlight the significance of advanced ML in improving predictions for protecting the marine ecosystem.