Determining the causes of atmospheric CO2 changes during the last glacial-interglacial cycle: a model-data study

• Main Author: O'Neill, Cameron
• Format: Other/Unknown Material
• Language: English
• Published: The Australian National University 2021
• Subjects: Antarctic; Southern Ocean; Antarc*; Sea ice
• Online Access: https://dx.doi.org/10.25911/467a-gw62; https://openresearch-repository.anu.edu.au/handle/1885/229785

The Earth’s surface carbon cycle naturally distributes carbon among its main reservoirs: the ocean, atmosphere, terrestrial biosphere, marine and continental sediments. This natural cycling of carbon regulates atmospheric CO2 on long and short timeframes. Now that the carbon cycle is being perturbed by anthropogenic CO2 emissions, it is important to understand how it works and may change in the future. The CO2 released to the atmosphere via industrial emissions and land clearing is passed into the Earth’s carbon cycle reservoirs on varying timeframes, with positive and negative feedbacks. The recent and distant geological past provide evidence for how the carbon cycle responds to natural environmental fluctuations, and how the carbon cycle may respond in future. The glacial-interglacial (G-IG) cycles of CO2 in particular exhibit multiple changes in the carbon cycle on short and long time scales, and are littered with clues preserved in the geological record. Chief among these changes are ∼80-90 ppm oscillations in atmospheric CO2 concentration during G-IG progressions. The definitive causes of these large changes in atmospheric CO2 concentration have been keenly debated for the last 40 years, yet remain unresolved, despite substantial progress in analytical and modelling techniques and the growing bodies of proxy data as evidence. The unresolved problem of atmospheric CO2 variations has been dubbed "the holy grail" of G-IG research. This thesis describes the development of a simple modelling tool to analyse the carbon cycle, that uses proxy data directly to constrain the modelling results, and explains its use in a model-data study of the last G-IG cycle of CO2 spanning 130 thousand years ago to the present. The thesis aims to address what caused the G-IG variation in atmospheric CO2. A review of G-IG CO2 research literature reveals important insights with which to undertake a model-data study. For example, substantial debate exists over the relative role of ocean circulation and mixing and other physical processes, versus biological and biogeochemical processes, in driving the G-IG fluctuations in atmospheric CO2 concentration. Ideas have of20 ten been presented in the literature as competing, or mutually exclusive explanations for the entire G-IG CO2 pattern. The literature review reveals that there are several processes that influence the carbon cycle, individually varying in importance for driving atmospheric CO2 concentration, yet collectively could provide the answer. Key among them are ocean circulation, sea surface temperatures, air-sea gas exchange between the ocean and atmosphere, marine biology and biogeochemistry, and the terrestrial biosphere. Some are both well understood and quantified for the last G-IG cycle of CO2, such as sea surface temperature (SST), relative sea level (RSL) and the terrestrial biosphere. Others could strongly influence the carbon cycle and atmospheric CO2, but are less well-understood and are poorly constrained. The latter includes ocean circulation and marine biological export productivity, which are both key subjects of the "holy grail" debates within the G-IG research literature. The proxy data supporting either of these processes as drivers of G-IG CO2 can be interpreted in other ways to support another hypotheses, leaving the debate over G-IG CO2 open. Other important carbon cycle processes include volcanism, continental weathering, marine sediment burial and weathering, and cosmic radiocarbon fluxes. Review of the literature of paleooceanographic model-data inversions and optimisations reveals there are difficulties applying orthodox inversion approaches to the paleo- age ocean. This difficulty arises mainly due to the relative sparsity and quality of data in the past versus the modern ocean. Inversions may struggle to incorporate proxy data such as carbon isotopes - which undergo numerous biogeochemical transformations during carbon cycle fluxes, and present in model-data-unfriendly formats, such as ratios of two elements versus the ratio of a geological reference standard. Furthermore, review of the use of complex carbon cycle models for model-data studies reveals a tendency towards experimenting over a narrow range of potential outcomes, which is only a limited extension of the well-established approach of hypothesis testing. This can expose the research findings to confirmation bias because only a subset of possible outcomes are considered. The field of model-data analysis in paleaonography is re-emergent after a period of stasis, and is recently re-invigorated by the steadily growing body of proxy data, freely available software and improving personal computer power. The "Simple Carbon Project Model" (SCP-M) box model was constructed for this thesis. The model was built to incorporate elements and proxies that are relatively data-rich and also to represent the processes which influence their concentrations in the carbon cycle. Furthermore, SCP-M was built to enable exhaustive model-data experiments to explore a wide range of possible parameter values and combinations. SCP-M incorporates phosphate, alkalinity, dissolved inorganic carbon (DIC), alkalinity, isotopes of carbon (δ13C and ∆14C) and the carbonate ion. SCP-M contains a quantitative representation of carbon cycle fluxes including ocean circulation and mixing, marine biological export productivity, marine carbonate production and dissolution, air-sea gas exchange, the terrestrial biosphere, volcanism and continental weathering. Some of the natural fluxes are relatively small in terms of carbon, yet they can be important for the fluxes of the isotopes of carbon, and other proxies, which are rich in data observations and provide the all important "clues" for changes in atmospheric CO2. Additional fluxes included in SCP-M for the purposes of calibration and testing are industrial era CO2 emissions, and fluxes of radiocarbon to the atmosphere from nuclear weapons testing in the 1950’s and 1960’s. SCP-M is initially constructed as a seven box plus atmosphere model, and is shown to provide an accurate representation of the modern ocean and atmosphere geochemistry, under the influence of anthropogenic emissions and bomb radiocarbon. It is tested against model predictions to the year 2100 for atmospheric CO2, and air-sea gas fluxes of carbon, from the Coupled Model Intercomparison Project (CMIP) models for the IPCC’s representative concentration pathways (RCPs). SCP-M provides a reasonable approximation to the CMIP RCP scenarios, with best matching for the lower CO2 emissions pathways. However, the SCP-M model cannot claim to replicate even a fraction of the complexities and detail captured by the CMIP models. The seven-box version of SCP-M ("v1.0") is demonstrated with a model-data optimisation for the last glacial maximum (LGM) and Holocene periods, using data for atmospheric CO2, ocean and atmosphere carbon isotopes and the oceanic carbonate ion proxy. The model-data optimisation experiment includes forcings for the carbon cycle mechanisms that are identified by literature review to be relatively well understood and constrained in the last G-IG cycle: SST, salinity, RSL and atmospheric 14C production. The optimisation solves for the values of global and Atlantic overturning circulation, deep ocean mixing, and marine biological productivity in the LGM and Holocene. The experiments highlight the important role for global overturning and Atlantic meridional overturning circulation in driving the increase in atmospheric CO2 from the LGM to the late Holocene period, alongside changes in SST. A second model-data study is undertaken in this thesis using an enhanced version of SCP-M with 12 ocean boxes plus atmosphere ("v2.2"). The analysis is expanded from the LGM-Holocene to cover the last G-IG cycle from 130 thousand years to the present, with model-data experiments undertaken in time-slices at each marine isotope stage (MIS). The model is forced with proxy data for SST, salinity, RSL, Antarctic sea-ice cover, coral reef carbonate production and dissolution, and atmospheric 14C production. The model-data optimisation solves for model values of global and Atlantic overturning circulation and Southern Ocean biological export productivity. The experiment results show that sequential changes in ocean circulation and biological export productivity took place over the last G-IG cycle. Early in the G-IG cycle, during MIS 5a-e, with cooling SST and increased sea-ice cover, the global ocean circulation began to slow (-7 Sv relative to the penultimate interglacial MIS 5e), with reduced upwelling of abyssal waters in the Southern Ocean, and slower Antarctic Bottom Water formation. Later in the G-IG cycle, at MIS 4, Atlantic Meridional overturning circulation also slowed (-5 Sv relative to MIS 5e), leading to a further drop in atmospheric CO2. The LGM model-data experiment showed that both ocean circulation mechanisms remained subdued (GOC -13 Sv and AMOC -5 Sv), and were accompanied by an increase in Southern Ocean biological export productivity (+∼2 mol C m−2 a−1), to achieve the LGM CO2 drawdown. By the Holocene period, both ocean circulation and marine biological productivity returned to modern-type values, that were also similar to their settings in the penultimate interglacial period (MIS 5e). Importantly, the mechanisms solved in the optimisation, the so-called "holy grail" of G-IG CO2 research - ocean circulation and marine biological export productivity, only account for part of the full glacial CO2 drawdown (-55 ppm). An attribution analysis for the model-data results on atmospheric CO2 shows that cooling SST (-25 ppm) was also a major contributor, and to a lesser extent, coral reef carbonates (-8 ppm). The model-data results corroborate hypotheses that feature multiple processes to deliver G-IG CO2. While these processes unwound dramatically at the glacial termination, they took place sequentially through the glacial lead up to the LGM over a period of ∼100 thousand years. Future work includes the continued effort to integrate model inversion or data optimisation procedures into more complex carbon cycle models, and to incorporate the range of emerging proxies (e.g. δ15N) into these models to allow additional multiproxy constraints on hypotheses for G-IG cycles of atmospheric CO2. Furthermore, attempts to undertake model-data studies of the last G-IG cycle (and prior cycles) will be enhanced by refinement and improvement of data constraints such as SST, RSL, salinity, sea-ice cover and coral reef accumulation and dissolution volumes. Finally, the expansion of ocean and atmosphere proxy databases will continue to contribute significantly to attempts to constrain paleoceanographic hypotheses with models and data.