Rising atmospheric CO2 concentration is one of the major drivers of climate change. To provide effective mitigation policies to curb these emissions, a thorough understanding of past changes in the carbon cycle is required. Decades of research on understanding carbon cycle changes during the last glacial cycle have put forward several processes impacting the concentration of atmospheric CO2. One of these processes is changes in aeolian iron flux into the Southern Ocean. Marine plankton fix dissolved inorganic carbon (DIC) during photosynthesis and transfer the fixed carbon to the deep ocean. DIC removal from the surface lowers the surface ocean partial pressure of CO2, which leads to carbon drawdown from the atmosphere. As the Southern Ocean is a high-nutrient-low-chlorophyll region, the increase in iron input can impact Southern Ocean marine ecosystems, by increasing export production, and therefore decreasing surface DIC. This thesis aims to investigate the responses of Southern Ocean marine ecosystems to changes in iron flux, and their impact on ocean biogeochemistry and atmospheric CO2 during the last glacial period. For this, I use a recently developed complex ecosystem model, which includes four different classes of phytoplankton functional types. Chapter 2 of this thesis is the first study to use this complex ecosystem model and document the competitive dynamics between different plankton species for light and nutrient availability under Last Glacial Maximum (LGM) climate boundary conditions (∼21 thousand years ago, 21 ka). Chapter 2 further assesses the impact of enhanced aeolian iron input on ecosystems. This study shows that lower sea surface temperatures and greater sea ice cover during the LGM causes a 2.4% reduction in Southern Ocean export production. However, a 78% increase in iron supply with a weaker ventilation in the Weddell Sea, increases diatoms and coccolithophores in the Southern Ocean, leading to a 4.4% higher carbon export at the LGM compared to pre-industrial (PI). Proxy records indicate a ∼32 ppm decrease in CO2 around ∼70 ka. Previous modelling studies have indicated a possible decline of 5 to 28 ppm in atmospheric CO2 driven by enhanced iron fertilization under PI and LGM boundary conditions. I constrain this contribution in chapter 3, by performing a series of sensitivity experiments under 70 ka climate boundary conditions taking into account the uncertainty associated with iron solubility in the ocean. I find that the CO2 change follows an exponential decay relationship with increasing iron flux due to saturation of biological pump at high iron values. Based on this, I suggest that enhanced iron input at 70 ka most likely led to a 9 to 11 ppm CO2 decrease with a maximum decrease of 21 ppm. Iron fertilisation could thus provide a 28 to 34% contribution to the total observed CO2 decline at 70 ka. Finally, in chapter 4, I include a unique approach to understand the processes leading to the abrupt 15 to 20 ppm increase in atmospheric CO2 during Heinrich Stadials, which are associated with a near collapse of the Atlantic Meridional Overturning Circulation (AMOC), a sudden decrease in Greenland temperature and warming in the Southern Ocean. Previous modelling studies have investigated the role of the ocean circulation in driving changes in atmospheric CO2 concentration during these abrupt events, while the role of reduced aeolian iron input during Heinrich stadials remained poorly constrained. I find that reduced iron fertilization combined with an AMOC shutdown could lead to a 7 ppm CO2 increase, 6 ppm of which is due to iron fertilisation. The research presented in this thesis improves our understanding of the impact of iron fertilization on Southern Ocean ecosystems, and on the global carbon cycle, particularly in the context of the last glacial period. This work also elucidates the importance of including changes in iron input to the ocean when investigating changes in atmospheric CO2 during abrupt climate change.