Using the Suess effect on the stable carbon isotope to distinguish the future from the past in radiocarbon

In the RCP8.5 emission scenario mitigation efforts start late leading to anthropogenic emission rates of up to nearly 30 Pg C yr⁻¹ around year 2100 with an assumed linear reduction between 2150 and 2200 to a constant emission rate of 1.5 Pg C yr⁻¹ until year 2500. (figure 2(A)). These emissions would result in a rise in atmospheric CO₂ concentration from present day 400 ppmv to ∼2000 ppmv after year 2200 in both the CMIP5 scenarios (Meinshausen et al 2011) and my carbon cycle simulations (figure 2(C)). The global warming and ocean acidification connected with such a rise in the most important anthropogenic greenhouse gas would be severe leading in my simulations to a temperature rise of 5–6 K (figure S3) and a drop in mean surface ocean pH by 0.8 units (from 8.2 to 7.4) (figure 2(F) inlet).

Zoom In Zoom Out Reset image size

Within the hypothetical CDR scenarios investigated here, net emissions are reduced even faster than in the other RCP emission scenarios assuming negative net emissions from year 2040 onward (figure 2(A)) and therefore broaden the range of possible future scenarios. The carbon extraction achieved in these CDR simulations might be unrealistically high, however, my interest here lies in showing potential maximum impacts on the carbon isotopes and not to investigate the most plausible scenario.

The cumulative airborne fraction (AF) of the anthropogenic emissions E (Pg C yr⁻¹), here defined as the ratio in the difference in atmospheric carbon pool (with respect to the pre-industrial values in year 1765) over the cumulative sum of E, stays in my simulations around 0.6 (figure 2(B)). Cumulative AF calculated from emission driven CMIP5 data are before year 1830 larger than 1, probably due to carbon cycle internal variability not driven by the yet small anthropogenic emissions. In the 21st and 22nd centuries they are slightly smaller than in my simulations. This difference is explained with the passive (=constant) terrestrial carbon pools in my simulations which neglects the terrestrial carbon sink found in the historical data (Le Quéré et al 2015). I refrain from showing results with active (=variable) terrestrial carbon cycle, since for atmospheric CO₂ concentrations well above 500 ppmv, the CO₂ fertilization implemented in my simple model is much too large, when compared with CMIP5 models, leading, due to the massive buildup of terrestrial carbon, to unrealistically low atmospheric CO₂ concentration (Köhler et al 2015). I here restrict simulation results to those obtained with an atmosphere-ocean only setup of the the carbon cycle, which on the long run agree in the atmospheric carbon pools with those of the CMIP5 results, although the still existing uncertainty in the land carbon cycle, partly due to an overestimation of the CO₂ fertilization (Smith et al 2016a), or due to uncertainties in the nitrogen cycle (Meyerholt et al 2016) might indicate that CMIP5 results are also not perfect. On the long run the cumulative AF and atmospheric CO₂ of my simulations converge with those based on CMIP5, indicating a small long-term influence of the terrestrial carbon sink in models contributing to CMIP5 (figures 1(C) and 2(B)). Simulations including terrestrial carbon storage changes would result in smaller simulated atmospheric CO₂, smaller AFs, and less depleted atmospheric ${\delta }^{13}{\rm{C}}$. Therefore, the historical ¹³C Suess effect is better matched by using an active terrestrial carbon cycle (figure S2B), while the effect on the historical ¹⁴C Suess effect reduces the offset between model and data, but has negligible impact on the ¹⁴C dynamic (figure S2C).

All CDR methods have a permanent impact on atmospheric CO₂ concentrations and on surface ocean pH (figures 2(C), (F)). Even in the scenarios BECCSs, DACs and EWs, in which CDR is stopped after some decades (here in year 2070) the simulated CO₂ concentrations (and surface ocean pH) do not reach the values obtained without CDR. The assumed CDR scenarios would eventually lead to a cumulative AF of zero, implying that an amount of CO₂ identical to the sum of all anthropogenic CO₂ emissions has been extracted from the carbon cycle again and atmospheric CO₂ concentration starts to fall below pre-industrial values.

The carbon isotopes of atmospheric CO₂ are both depleted by the massive injection of anthropogenic emissions, since fossil fuels are ¹⁴C-free and contain with about −24 to −29‰ a ${\delta }^{13}{\rm{C}}$ signature (Andres et al 2000, 2015) that is 19‰ lighter than the ${\delta }^{13}{\rm{C}}$ signature of the atmospheric CO₂ itself (figure S1B). Additionally, the radiocarbon cycle is penetrated by the bomb-¹⁴C emissions in the second half of the last century (Naegler and Levin 2006) leading around 1965 to atmospheric ${{\rm{\Delta }}}^{14}{\rm{C}}$ values of up to +700 ± 200‰ in the data (Hua et al 2013) and of +900‰ in my simulations (figure 2(E)) (see supplementary material for further details).

Atmospheric ${{\rm{\Delta }}}^{14}{\rm{C}}$ then drops around 2150 to −300‰ in RCP8.5 and to −415‰ in all CDR approaches. This most depleted ${{\rm{\Delta }}}^{14}{\rm{C}}$ signature of −415‰ is identical to that of a 4300 year old carbon sample (figure 3(A)). Depending on the assumed CDR method ${\delta }^{13}{\rm{C}}$ of atmospheric CO₂ drops at the same time to values of (RCP8.5) −13.3‰, (EW) −12.6‰, or (DAC) −16.6‰ (figure 2(D)). For BECCS ${\delta }^{13}{\rm{C}}$ of atmospheric CO₂ returns to its pre-industrial value of −6.5‰ in year 2150 and rises thereafter to values up to −2‰. Here, the difference of how the CDR methods modify the carbon cycle has a significant impact on the resulting atmospheric ${\delta }^{13}{\rm{C}}$ signature: BECCS operates as negative land use change, therefore reversing the ¹³C Suess effect. In scenario EW alkalinity is added to the ocean. The isotopic fractionation within the dissolved inorganic carbon (DIC) in the ocean and therefore of the ocean-atmosphere gas exchange depends directly on the concentration of ${\mathrm{HCO}}_{3}^{-}$ and ${\mathrm{CO}}_{3}^{2-}$, two of the chemical species of DIC. However, the concentrations of these species change with a rise in alkalinity to allow a larger oceanic CO₂ storage. Therefore, the isotopic fractionation during gas exchange indirectly depends on the surface ocean alkalinity (Zeebe and Wolf-Gladrow 2001) and is in detail implemented in BICYCLE similarly as in other models (Ridgwell 2001).

Zoom In Zoom Out Reset image size

When ${{\rm{\Delta }}}^{14}{\rm{C}}$ and ${\delta }^{13}{\rm{C}}$ are plotted against each other it clearly becomes evident that the Suess effects on both isotopes will in the future bring the isotopic carbon cycle into a regime in which it has not been during at least the last 50 000 years. The historical Suess effect before 1950 (−0.7‰ in ${\delta }^{13}{\rm{C}}$ and −20‰ in ${{\rm{\Delta }}}^{14}{\rm{C}}$) already shifted the atmospheric variables away from its natural state (figure 3(A)). The atmospheric ${{\rm{\Delta }}}^{14}{\rm{C}}$ simulated in response to the bomb-¹⁴C injection led to 0 to +900‰, slightly larger than the range of −25-to- +575‰ that has been reconstructed for the pre-industrial 50 000 years from various archives (Köhler et al 2006, Reimer et al 2013). Already the historical emissions from 1950 onward including the foreseeable emissions until 2020 shift the atmospheric ${\delta }^{13}{\rm{C}}$ by another −2‰. In most scenarios a further depletion in both carbon isotopes takes place in the near future. At the extreme, values of ${{\rm{\Delta }}}^{14}{\rm{C}}=-415\textperthousand$ and ${\delta }^{13}{\rm{C}}=-16.6\textperthousand$ are reached in the atmospheric carbon reservoir. The exceptions to this rule are scenarios in which BECCS plays a dominant role, also implying that RCP2.6 has a different dynamic in the carbon isotopes than the other RCP scenarios. EW would first lead to a small rise in ${\delta }^{13}{\rm{C}}$ but on the long run also to a depletion. In BECCS the simulated ${\delta }^{13}{\rm{C}}$ on the long run is higher than what is known from the paleo record. Most scenarios might, after having a maximum depletion in the isotopic phase space, return to less extreme anomalies in both isotopes, only RCP2.6 returns in the ${{\rm{\Delta }}}^{14}{\rm{C}}\mbox{--}{\delta }^{13}{\rm{C}}$-scatter plot back to conditions seen in pre-industrial times or found in the paleo simulations or reconstructions.

To analyze how the carbon isotopes in the ocean might change due to the Suess effects I focus on the two end-member in the oceanic carbon cycle: (a) North Atlantic surface waters, where North Atlantic Deep Water formation occurs and a dominant part of deep ocean water masses have last contact with the atmosphere and (b) the deep Indo-Pacific, in which the oldest, most ${{\rm{\Delta }}}^{14}{\rm{C}}$-depleted water masses are found. A similar pattern as found in the atmosphere emerges in the North Atlantic surface waters, although with smaller amplitude (figure 3(B)): the bomb-¹⁴C spike is found with slightly more than +100‰, the ¹³C Suess effect leads until 2020 to a reduction in ${\delta }^{13}{\rm{C}}$ by nearly −1.5‰, and all scenarios but RCP2.6 enter uncharted waters in the ${{\rm{\Delta }}}^{14}{\rm{C}}\mbox{--}{\delta }^{13}{\rm{C}}$ phase space. Clearly seen is also that the rising ocean alkalinity in the EW CDR method leads to a more depleted surface ocean ${\delta }^{13}{\rm{C}}$, explaining the lower isotopic fractionation (less depletion) in the atmospheric ${\delta }^{13}{\rm{C}}$ record and the special dynamics for BECCS leading to ${\delta }^{13}{\rm{C}}$ of nearly +3‰. An overlap of the historical and future simulations with the data range spanned by paleo data (Reimer et al 2013, Peterson et al 2014) and paleo simulations (Köhler et al 2006) covering the last 50 000 years is only obtained for the bomb-¹⁴C spike. Also note, that these paleo simulations, performed with a previous version of the same model, were imperfect, since they were not able to explain the full decline in atmospheric ${{\rm{\Delta }}}^{14}{\rm{C}}$ found in the paleo reconstructions (Reimer et al 2013).

The simulated changes in the deep Indo-Pacific during the next five centuries are much smaller than for the surface ocean (figure 3(C)). Until 2020 the Suess effects or even the ¹⁴C-bomb spike are not detectable in this reservoir, however the effect of further anthropogenic emissions will over the course of the simulations found its way to this most remote ocean reservoir and both Suess effects will then be visible there. The simulated future trends in the deep Indo-Pacific ${\delta }^{13}{\rm{C}}$ have some overlap with the range of reconstructed ${\delta }^{13}{\rm{C}}$, however, the knowledge on deep ocean ${{\rm{\Delta }}}^{14}{\rm{C}}$ is still limited. While my previous (imperfect) simulations suggest that deep Indo-Pacific ${{\rm{\Delta }}}^{14}{\rm{C}}$ was always higher than −150‰ throughout the last 50,000 years, the limited available deep ocean ${{\rm{\Delta }}}^{14}{\rm{C}}$ reconstructions show a different picture (Ronge et al 2016): ${{\rm{\Delta }}}^{14}{\rm{C}}$-values as low as −200‰ are found in waters above 2000 m and below 4300 m water depth in the South Pacific with some water masses in between (and in intermediate depths of ∼600 m around the Galapagos Islands (Stott et al 2009) having during the last 25 000 years a ${{\rm{\Delta }}}^{14}{\rm{C}}$ signature as low as −600‰. This would imply that for most of the RCP emission scenarios the deep Pacific data in the ${{\rm{\Delta }}}^{14}{\rm{C}}\mbox{--}{\delta }^{13}{\rm{C}}$ phase space might already have been obtained in some form during glacial conditions in the past. These most recent deep Pacific data with low ${{\rm{\Delta }}}^{14}{\rm{C}}$ signature (Ronge et al 2016) are not yet completely understood. It is not yet clear how wide-spread this water mass is and the explaining hypothesis put forward so far suggests the release of ¹⁴C-free CO₂ from hydrothermal activities along mid-ocean ridges during sea-level low stand in glacial times. This would imply that the deep glacial ocean would contain, in addition to the fossil fuel emissions into the atmosphere, another source of ¹⁴C-free carbon. The interpretation of deep ocean carbon isotopic signatures might therefore be not yet straight-forward.

Simulation results for other surface ocean reservoirs are qualitatively similar to the North Atlantic surface end member discussed in detail above (figure S4), allowing in surface reservoirs to use the ¹³C Suess effect to distinguish past from future carbon fluxes. Interestingly, the largest oceanic anomalies in ${\delta }^{13}{\rm{C}}$ are obtained in the surface equatorial Atlantic Ocean (figure S4B) with ${\delta }^{13}{\rm{C}}$ falling down to −13‰ for EW scenarios, probably caused by the way the EW fluxes are prescribed. These fluxes enter the surface ocean only in the equatorial regions, with 50% each routed in the Atlantic and Indo-Pacific. Combined with the smaller size of the Atlantic basin, the effect of EW on the local carbon cycle is more pronounced in the Atlantic than in the Indo-Pacific. Since the prescribed water mass fluxes to the surface North Pacific area are all sourced in deep ocean regions, ${\delta }^{13}{\rm{C}}$ in this area follows in the EW scenarios the dynamics seen in the atmosphere (less depleted than in RCP8.5, figure S4F). Carbon isotopic dynamics in the deep ocean of the Atlantic (figure S4C) and to some extend in the Southern Ocean (figure S4E) depart from known data ranges in the past. My approach to disentangle past from future carbon cycle changes therefore seemed also to be applicable to data from these deep ocean reservoirs. Further regional details are better obtained with spatially higher resolved models.

Fossil fuel fluxes contain also emissions from industrial processes, namely cement production. The ${\delta }^{13}{\rm{C}}$ signature of fossil fuels therefore depends on the source mix and ranges from 0‰ (cement production) to −44‰ (natural gas) (Andres et al 2000). About 6% of the CO₂ emissions summarized as fossil fuels in year 2014 have been from cement production (Le Quéré et al 2015). In my standard scenarios I assume that the source mix (and therefore the ${\delta }^{13}{\rm{C}}$ signature of fossil fuels) remains the same from year 2011 onward. In one scenario (RCP8.5@cement) I test the effect when cement production would slowly become the one and only source of the fossil fuel emissions in year 2250 (evolution of ${\delta }^{13}{\rm{C}}$ of fossil fuels shown in figure S1B). Simulated ${\delta }^{13}{\rm{C}}$ values would then be less depleted than in our standard simulations (figure 2(D)), but isotopic values would still be outside of their ranges known from the past (figure 3), and the overall conclusion would therefore not be affected by such a rise in the relative importance of cement in the source mix of future fossil fuel emissions.