| Summary: | Understanding how sinks of atmospheric CO2 are evolving is essential to ensure that solutions to climate change can be defined and implemented. The ocean is a considerable sink of atmospheric CO2, however, observational estimates and model-based projections of the contemporary and future sink remain uncertain. This thesis aims to reduce these uncertainties by improving understanding of the marine carbon cycle and its temporal evolution. This is achieved in three ways: 1) by evaluating and validating observations of surface ocean carbon from Biogeochemical Argo floats. 2) Through introducing a machine learning-based approach, capable of producing the first purely observational, temporally resolved estimate of total added carbon from ocean interior observations. 3) By using an offline model set up to identify the key processes required for effectively simulating alkalinity; one of the foundational components of oceanic carbon modelling. Several key outcomes emerge from each of the areas of interest.1) Biogeochemical Argo floats produce reliable pCO2 estimates, without systematic biases relative to ship-based observations. Float-based measurements can constrain the mixed layer carbon budget in a biologically important region of the Southern Ocean, demonstrating that autonomous platforms can be used to define mixed layer carbon dynamics in under-sampled regions. 2) The machine learning approach can accurately reconstruct the cumulative global total added carbon inventory between 1994 and 2018. Analysis demonstrates that this method can act as an independent validation of pCO2 based flux estimates into the ocean when considered over sufficient temporal and spatial scales. 3) Alkalinity modelling demonstrates that many alkalinity-altering processes commonly excluded or over-simplified in earth systems models can considerably alter oceanic carbon inventories by changing the surface ocean buffering capacity. When such processes are excluded, model projections of the ocean’s future sink capacity will likely contain ...
|
|---|