Abstract:
Machine learning has become a useful tool to interpolate ship measurements of pCO2 to a gridded map using satellite data. In this study we use an ensemble of three machine learning methods: Support Vector Regression (SVR. and Random Forest Regression (RFR. from Gregor et al. (2017.; and the SOM-FFN method from Landschützer et al. (2016.. The interpolated data were separated into nine regions defined by basin (Indian, Pacific and Atlantic. and functional biomes. The regional approach showed a seasonal decoupling of the modes for summer and winter interannual variability. Winter interannual variability had a longer mode of variability compared to summer, which varied on a 4–6-year time scale. To understand this variability of ?pCO2, we investigated changes in summer and winter ?pCO2 and the drivers thereof. The dominant winter changes are driven by wind stress variability. This is consistent with the temporal and spatial characteristics of the Southern Annular Mode (SAM., which has a decadal mode of variability (Lovenduski et al., 2008; Landschützer et al., 2016.. Interannual trends in summer variability of ?pCO2 are consistent with chlorophyll-a variability where the latter had high mean seasonal concentrations. In regions of low chlorophyll-a concentrations, wind stress and sea surface temperature emerged as stronger drivers of ?pCO2. In summary we propose that sub-decadal variability is explained by summer drivers, while winter variability contributes to the long-term changes associated with the SAM. These findings reemphasise the importance of the long-term changes in wind stress over the Southern Ocean. We propose that more work is required in understanding the variability of long term wind stress. - Abstract as displayed in the - Abstract booklet. The presentation on the day may differ from the - Abstract.