Natural variability of CO₂ and O₂ fluxes: What can we learn from centuries‐long climate models simulations?

Abstract

Ocean carbon uptake and oxygen content estimates over the past decades suggest that the anthropogenic carbon sink has changed and that the oxygen concentration in the ocean interior has decreased. Although these detected changes appear consistent with those expected from anthropogenic forced climate change, large uncertainties remain in the contribution of natural variability. Using century-long simulations (500-1000 years) of unforced natural variability from six Earth System Models (ESMs), we examine the internally driven natural variability of carbon and oxygen fluxes from interannual to multidecadal time scales. The intensity of natural variability differs between the ESMs, in particular, decadal variability locally accounts for 10-50% of the total variance. Although the variability is higher in all regions with strong climate modes (North Atlantic, North Pacific, etc.), we find that only the Southern Ocean and the tropical Pacific significantly modulate the global fluxes. On (multi)decadal time scales, deep convective events along the Antarctic shelf drive the global fluxes variability by transporting deep carbon-rich/oxygen-depleted waters to the surface and by reducing the sea-ice coverage. On interannual time scales, the global flux is modulated by (1) variations of the upwelling of circumpolar deep waters associated with the southern annular mode in the subpolar Southern Ocean and (2) variations of the equatorial/costal upwelling combined with changes in the solubility-driven fluxes in response to El Niño Southern Oscillation (ENSO) in the tropical Pacific. We discuss the challenges of measuring and detecting long-term trends from a few decade-long records influenced by internal variability.

Natural variability in the surface ocean carbonate ion concentration

“…Our results suggest that the detection timescale for anthropogenic trends in pH is shorter than that for aragonite , owing to smaller noise-to-signal ratios and lower autocorrelation in pH. We further characterize the surface [CO ] variability in our model is similar to that reported in other studies (e.g., Friedrich et al, 2012), a recent study of natural carbon uptake variability from centuries-long simulations of six ESMs suggests that we should expect the magnitude of this variability to differ from model to model (Resplandy et al, 2015 ] variability in these high variance regions suggests that it is largely driven by large-scale modes of internal climate variability, such as ENSO, PDO, NAO, and AMO. One should consider the phasing of these modes of climate variability, therefore, when interpreting trends calculated from carbonate chemistry data collected in these regions.…”

“…While hindcast studies can capture the observed chronology of these modes, they cannot capture the full spectrum of internal variability in the climate system. Long model simulations (order 1000 years) can capture multiple realizations of climate variability on decadal and multi-decadal timescales, and have shown to be useful in the study of ocean carbon cycle variability on these timescales Séférian et al, 2013Séférian et al, , 2014Keller et al, 2014;Lehner et al, 2015;Resplandy et al, 2015).…”

Climate and carbon cycle dynamics in a CESM simulation from 850 to 2100 CE

“…We find forced decadal-scale variability in CESM and MPI-ESM in response to major volcanic eruptions in both SAT and upper-ocean temperature, while the response in carbon cycle quantities is less coherent among models (see also Resplandy et al, 2015). Outside volcanically active periods, large parts of the decadal-scale variations cannot be attributed to external forcing, suggesting that internal variability masks external forcing influence.…”

Inconsistent strategies to spin up models in CMIP5: implications for ocean biogeochemical model performance assessment

“…Information on spin-up protocols and model initialization is usually not taken into account in model intercomparison studies (e.g., Andrews et al, 2013;Bopp et al, 2013;Cocco et al, 2013;Frölicher et al, 2014;Gehlen et al, 2014;Keller et al, 2014;Resplandy et al, 2013Resplandy et al, , 2015Rodgers et al, 2015;. This information, if available, can only be found separately in the reference papers of individual models (e.g., Adachi et al, 2013;Arora et al, 2011;Collins et al, 2011;Dunne et al, 2013;Ilyina et al, 2013;Lindsay et al, 2014;Romanou et al, 2013;Séférian et al, 2013Séférian et al, , 2016Tjiputra et al, 2013;Vichi et al, 2011;Volodin et al, 2010;Watanabe et al, 2011;Wu et al, 2013).…”