Drivers of decadal trends of the ocean carbon sink in the past, present, and future in Earth system models

Abstract. The land biosphere and the ocean are the two major sinks of anthropogenic carbon at present. When anthropogenic carbon emissions become zero and temperatures stabilizes, the ocean is projected to become the dominant and only global natural sink of carbon. Despite the ocean’s importance for the carbon cycle and hence the climate, observing the ocean carbon sink and detecting anthropogenic changes over time remain challenging because uncertainties of the decadal variability of this carbon sink and the underlying drivers of this decadal variability remain large. The main tools that are used to provide annually resolved estimates of the ocean carbon sink over the last decades are global observation-based pCO2 products that extrapolate sparse pCO2 observations in space and time and global ocean biogeochemical models forced with atmospheric reanalysis data. However, these tools (i) are limited in time over the last 3 to 7 decades, which hinders statistical analyses of the drivers of decadal trends, (ii) are all based on the same internal climate state, which makes it impossible to separate externally and internally forced contributions to decadal trends, and (iii) cannot assess the robustness of the drivers in the future, especially when carbon emissions decline or cease entirely. Here, I use an ensemble of 12 Earth System Models (ESMs) from phase 6 of the Coupled Model Intercomparison Project (CMIP6) to understand drivers of decadal trends of the past, present and future ocean carbon sink. The simulations by these ESMs span the period from 1850 to 2100 and include 4 different future Shared Socioeconomic Pathways (SSPs), from low emissions and high mitigation to high emissions and low mitigation. Using this ensemble, I show that 80 % of decadal trends in the multi-model mean ocean carbon sink can be explained by changes in decadal trends of atmospheric CO2 as long as the ocean carbon sink remains smaller than 4.5 Pg C yr-1. The remaining 20 % are due to internal climate variability and ocean heat uptake, which results in a loss of carbon from the ocean. When the carbon sink exceeds 4.5 Pg C yr-1, which only occurs in the high emission SSP3-7.0 and SSP5-8.5, atmospheric CO2 rises faster, climate change accelerates, the ocean overturning and the chemical capacity to take up carbon from the atmosphere reduce, so that decadal trends in the ocean carbon sink become substantially smaller than estimated based on changes in atmospheric CO2 trends. The breakdown of this relationship in both high emission pathways also implies that the decadal increase in the ocean carbon sink is effectively limited to be ~1 Pg C yr-1 dec-1 in these pathways, even if the trend in atmospheric CO2 continues to increase. Previously proposed drivers, such as the atmospheric CO2 or the growth rate of atmospheric CO2 can explain trends in the ocean carbon sink for specific time periods, for example during exponential atmospheric CO2 growth, but fail when emissions start to decrease again. The robust relationship over a large ESM ensemble also suggests that very large positive and negative decadal trends of the ocean carbon sink by some pCO2 products are highly unlikely, and that the change in the decadal trends of the ocean carbon sink around 2000 is likely substantially smaller than estimated by these pCO2 products.