Filling the gaps: anthropogenic CO₂ and the changing biological carbon pump through integrated observations and machine learning

Since the Industrial Revolution, atmospheric carbon dioxide (CO₂) concentrations have surged from 278 ppm in 1750 to over 420 ppm in 2023, driven by fossil fuel use and land-use changes. This rise alters the global carbon cycle, increasing the ocean's role as a carbon sink. Oceans absorb 25–30% of anthropogenic CO₂ emissions, mitigating climate change but inducing long-term changes like acidification. While vital for climate regulation, uncertainties persist about the ocean's future CO₂ sequestration capacity. Two mechanisms govern oceanic CO₂ uptake: the solubility pump and the biological carbon pump (BCP). The solubility pump transfers CO₂ into the deep ocean via thermohaline circulation, while the BCP sequesters CO₂ by converting it into organic matter through photosynthesis. This organic material sinks, reducing atmospheric CO₂ by around 200 ppm. Although the solubility pump is well understood, the BCP's variability and sensitivity to climate change remain less clear. Addressing these gaps is critical for predicting oceanic responses to future CO₂ emissions. This thesis addresses gaps in the marine carbonate system by examining the BCP and anthropogenic CO₂ uptake through in-situ measurements, sensor technologies, and machine learning models. Spatial and temporal data limitations, particularly in regions like the Southern Ocean and tropical Pacific, restrict understanding of seasonal and interannual ocean carbon storage variations. The thesis highlights the need to enhance data coverage and knowledge of BCP evolution under rising CO₂ and climate pressures. Chapter 2 investigates basin-scale changes in the BCP, focusing on the South Atlantic's South Subtropical Convergence (SSTC) at 40°S. Using data from GLODAPv2.2022 and Biogeochemical Argo (BGC-Argo) floats, the study extends observations by a decade, revealing regional BCP strengthening. This increase is driven by higher dissolved inorganic carbon (DIC) and interactions between physical and biological processes. Results highlight the region-specific nature of BCP responses to climate change. Chapter 3 examines surface seawater pH variability, an uncertain factor in the carbonate system. High-resolution pH measurements from PyroScience pH optodes during North Atlantic and South Pacific cruises capture fine-scale fluctuations driven by temperature, biology, and water mass interactions. Despite local variability, data align with basin-scale observations, supporting the reliability of global pH monitoring. Building on these insights, Chapter 4 introduces SOCA-CO2, a four-dimensional model integrating ship-based, BGC-Argo, and satellite data to reconstruct DIC at fine resolutions. This model enhances predictive accuracy in ocean carbonate chemistry, providing weekly updates at a 0.25° spatial scale. SOCA-CO2 addresses DIC prediction gaps, especially in under-sampled regions. Chapter 5 scales findings globally by reconstructing DIC from 2004 to 2022 using CANYON-B and CONTENT algorithms. Over 90% of surface DIC increases are driven by anthropogenic CO₂ uptake. However, regional variations in natural carbon cycling persist, with increases in the North Pacific and Southern Ocean and decreases in the Indian Ocean and coastal zones. This thesis highlights the value of integrating diverse observational platforms and machine learning models to fill knowledge gaps in ocean carbon cycling. Addressing spatial and temporal blind spots enhances understanding of the BCP and carbonate chemistry, improving predictions of the ocean's role in climate change mitigation.