Machine learning algorithm reveals surface deoxygenation in the Agulhas Current due to warming

Abstract

The Agulhas Current is the strongest western boundary current (WBC) in the Southern Hemisphere with a significant impact on the global climate. Through the Agulhas leakage, it transports warm, saline waters into the South Atlantic Ocean. In recent years, a warming trend has been highlighted in the Agulhas Current, with possible implications for dissolved oxygen ([O2]) due to the link between warming and reduced solubility – dynamics that remain relatively unknown in the region. To address this knowledge gap, we use the random forest regression algorithm to predict near–surface [O2] from multiple predictors in the Agulhas Current, presenting the first analysis of these dynamics. The Agulhas–RFR algorithm predicts [O2] exceptionally well, with permutation importance from the ensemble indicating that sea surface temperature (SST) is the highest–ranking predictor. Seasonal changes in solubility, wind, and productivity drive [O2] and the [O2] flux in the Agulhas Current. The seasonal [O2] flux to the atmosphere reaches –1.84 mol m−2 yr−1 during the austral winter across the Agulhas Current. A significant decreasing [O2] trend of up to –7 µmol kg−1 yr−1, attributed to warming, is revealed for the period from 2000 to 2023. Strengthening westerlies and cooling contribute to [O2] drawdown towards the Indian Ocean gyre. The Agulhas–RFR algorithm reveals a declining [O2] trend of –2.29 ± 0.61 µmol kg−1 yr−1 across the Agulhas Current for the study period, representing a 1.4% deoxygenation rate, which is slightly higher than global estimates.