Role of the Deglacial Buildup of the Great Barrier Reef for the Global Carbon Cycle

The carbon isotope 13C is commonly used to attribute the last deglacial atmospheric CO2 rise to various processes. Here we show that the growth of the world's largest reef system, the Great Barrier Reef (GBR), is marked by a pronounced decrease in δ13C in absolutely dated fossil coral skeletons between 12.8 and 11.7 ka, which coincides with a prominent minimum in atmospheric δ13CO2 and the Younger Dryas. The event follows the flooding of a large shelf platform and initiation of an extensive barrier reef system at 13 ka. Carbon cycle simulations show the coral δ13C decrease was mainly caused by the combination of isotopic fractionation during reef carbonate production and the decomposition of organic land carbon on the newly flooded shallow‐water platform. The impacts of these processes on atmospheric CO2 and δ13CO2, however, are marginal. Thus, the GBR was not contributing to the last deglacial δ13CO2 minimum at ∼12.4 ka.

Overall, the late deglacial second minimum in atmospheric δ 13 CO 2 values has gained less attention compared to that observed during the early deglaciation, and its comprehensive explanation considering the complex interactions of all components of the past carbon cycle is still lacking. In this context, the development of shallow-water carbonate reefs, although acknowledged to be of relevance for the deglacial CO 2 rise (Brovkin et al., 2012;Ridgwell et al., 2003;Vecsei & Berger, 2004;W. H. Berger, 1982), has been usually neglected (Bauska et al., 2016;Schmitt et al., 2012). Carbon isotopic fractionation during calcium carbonate (CaCO 3 ) production is relatively small (Al-Rousan & Felis, 2013;Linsley et al., 2019;Swart et al., 2010) compared to fractionation during photosynthesis and, subsequently, little change in atmospheric δ 13 CO 2 is expected. However, to rigorously test this, precisely dated and well constrained reef carbonate records that span the last deglacial period are necessary, but such records are generally not available (Montaggioni, 2005).
A uranium-thorium (U-Th) dated record of coral skeletal δ 13 C values from the Great Barrier Reef GBR), the world's largest contiguous coral reef system off northeastern Australia (Figure 2), shows a pronounced drop of 1.25‰ synchronous to the second minimum in atmospheric δ 13 CO 2 values at ∼12.4 ka ( Figure 1c) suggesting processes related to the growth of the "proto-GBR" that commenced at this time ) might be at least partially responsible for the observed carbon cycle changes. Here, we combine these coral δ 13 C values with other data from Integrated Ocean Drilling Program (IODP) Expedition 325  reconstructing local sea level rise, shelf flooding, and reef carbonate accumulation for the last deglaciation Yokoyama et al., 2018;Hinestrosa et al., 2019Hinestrosa et al., , 2022 in a carbon cycle modeling framework to investigate if and how both minima in δ 13 C values are causally related to each other.

Last Deglacial GBR Coral δ 13 C Record
The GBR record of coral δ 13 C values ( Figure 1c) is a stacked record from IODP drill sites at two locations, Noggin Pass (NOG) and Hydrographer's Passage (HYD; Figure 2). It is based on measurements of δ 13 C values in the carbonate skeleton of U-Th dated Isopora palifera/cuneata colonies . Twenty-five well-preserved fossil corals were recovered at depths between 56 and 126 m below present sea level in the central GBR (Figure 2), where they grew in shallow-water (0-10 m), high-energy reef crest environments during time intervals of the last deglaciation ; Text S1 in Supporting Information S1). In modern reef environments, skeletal δ 13 C values of corals reflect the combined effects of various parameters including the availability of light for photosynthesis of the coral's algal symbionts, the coral's skeletal extension rate, and the autotrophy-heterotrophy ratio of the coral's diet (Al-Rousan & Felis, 2013;Felis et al., 1998;Gagan et al., 2015;Linsley et al., 2019;Suzuki et al., 2005;Swart, 1983;Swart et al., 1996cSwart et al., , 2005Swart et al., , 2010. These effects can result in a large range of skeletal δ 13 C values, or intercolony variability, in corals derived from different parts of the same reef, as revealed by modern Isopora palifera/cuneata colonies from Heron Island (HER) in the southern GBR (Figures 1c and 2) Swart & Coleman, 1980). In addition to variations in skeletal δ 13 C values that are mainly caused by metabolic responses of the corals to reef-scale differences in environmental conditions, basin-scale secular changes in surface ocean δ 13 C DIC are documented. The most compelling evidence for coral δ 13 C values as a proxy for seawater δ 13 C DIC is provided by the imprint of the 13 C Suess effect (Keeling, 1979), the impact of anthropogenic CO 2 emissions with lower amounts of 13 C, on the δ 13 C DIC of the surface ocean (Al-Rousan & Felis, 2013;Dassié et al., 2013;Linsley et al., 2019;Swart et al., 2010). The magnitude of this decrease in surface water δ 13 C DIC can be modulated at a regional scale by input of terrestrial organic material with lower amounts of 13 C through runoff (Swart et al., 1996a(Swart et al., , 1996b and the complex interplay of photosynthesis and respiration during diurnal carbon cycling on shallow carbonate shelves (Geyman & Maloof, 2019; Text S1 in Supporting Information S1).
Importantly, the GBR record of coral δ 13 C values does not reveal the typical w-shaped evolution observed in the last deglacial record of atmospheric δ 13 CO 2 values ( Figure 1c). The record of coral δ 13 C values does not document the first prominent minimum in atmospheric δ 13 CO 2 at ∼16.0 ka as a result of the lack of suitable corals from this interval, but is largely consistent with the broad atmospheric δ 13 CO 2 minimum that commences after ∼17 ka. The rather gentle decrease in coral δ 13 C by about 0.5‰ between the ∼21-17 ka interval and a cluster of corals at ∼15 ka can be interpreted as an imprint of the atmospheric δ 13 CO 2 fall of similar magnitude on global 10.1029/2021GL096495 4 of 9 surface ocean δ 13 C DIC via air-sea gas exchange (Lynch-Stieglitz et al., 2019;Shao et al., 2021). This process is similar to the 13 C Suess effect (Keeling, 1979) and its imprint on surface ocean δ 13 C DIC and coral δ 13 C today (Al-Rousan & Felis, 2013;Dassié et al., 2013;Linsley et al., 2019;Swart et al., 2010) and gives support to the use of GBR coral δ 13 C values as proxy for surface ocean δ 13 C DIC changes during the last deglaciation. After another decrease of about 0.4‰ between ∼15 and ∼13 ka the most striking feature in the GBR record of coral δ 13 C values is a pronounced decrease of up to 1.25‰ that occurred between 12.8 and 11.7 ka. The magnitude of this decrease in coral δ 13 C is by far larger than the synchronous decline in atmospheric δ 13 CO 2 by 0.1-0.2‰ into its second prominent minimum centered at ∼12.4 ka (Bauska et al., 2016;Schmitt et al., 2012), and points to processes intrinsic to the GBR (Figure 1c). Importantly, a potential influence of site-specific effects, intercolony variability, and changes in light availability, due to orbital insolation changes or a larger water depth of coral growth and/ or increased reef water turbidity during the course of the last deglaciation, on the overall temporal evolution of the GBR record of coral δ 13 C values, especially after 13.0 ka, can be largely ruled out (Text S1 in Supporting Information S1). While the overall trend in the record is considered as relatively robust given the uncertainty in Flooded areas corresponding to sea levels between 45 and 75 m below present indicated (green) on present-day GBR bathymetry (Beaman, 2010). 45-75 m depth band shows extent of flooded terrestrial areas following 13.0 ka shelf platform flooding encompassing maximum range (depth and time) at which growth of "proto-GBR" (Reef 4) took place . Reef growth predominantly occurred along shelf edge seaward of modern GBR (Hinestrosa et al., 2019) (white areas). Not all flooded areas (green) were in situ reefs, but rather a variety of relatively shallow shelf settings. Locations of Integrated Ocean Drilling Program (IODP) Expedition 325 drilling sites at Noggin Pass (NOG) and Hydrographer's Passage (HYD) at central GBR shelf edge ) (red squares). Modern coral sites at Heron Island (HER) (southern GBR) and Myrmidon Reef (MYR) and Magnetic Island (MAG) (central GBR) (yellow squares). Inset: Regional close-up of IODP site at NOG.

GBR Shelf Flooding and Barrier Reef Initiation
Along the shelf off northeastern Australia, a substantial increase in marine flooded area and carbonate accumulation by coral reef organisms took place after 13.9 ka (Figure 1d). Between ∼13.9 and 13.0 ka, a rapid sea-level rise ( Figure 1e) and the availability of area for flooding favored coral reef growth along the shelf edge (Hinestrosa et al., 2019;Webster et al., 2018;Yokoyama et al., 2018). At 13.0 ka, the flooding of a large platform along the shelf (Figure 2), under a now slowly rising sea level (Hinestrosa et al., 2019;Webster et al., 2018;Yokoyama et al., 2018), marks a major change from narrow fringing reef systems to barrier-reef-dominated morphologies , which can be traced almost continuously over 2,000 km along the GBR shelf (Abbey et al., 2011;Hinestrosa et al., 2014), and which is immediately followed by the pronounced decrease in coral δ 13 C values between 12.8 and 11.7 ka (Figure 1c). The extensive barrier reef system that initiated its growth along the shelf edge soon after the platform flooding was characterized by a thriving shallow-water (<5-10 m) coral community and is considered as the true "proto-GBR" . This reef represents one of the five distinct reef sequences in the IODP Expedition 325 GBR record of the past ∼30 kyr, and is referred to as Reef 4 Yokoyama et al., 2018). Prior to 13.0 ka, the shelf configuration had favored the formation of narrow fringing reefs . We estimate that between 13.9 and 13.0 ka an area of 25,160 km 2 was flooded and 293 Tmol of carbonate were accumulated mainly in shelf edge areas in the growing GBR (Figure 1d; Text S1 in Supporting Information S1), about 6-fold the amount of area flooded than during the 2 kyr before (15.9-13.9 ka). The total carbonate accumulation rose by a factor of 2.5 between both periods, however the annual carbonate accumulation rate increased from 57 to 326 Gmol yr −1 , while this rate when normalized per whole flooded area stayed with 13 mol yr −1 m −2 nearly constant (Table S1 in Supporting Information S1). From 13.0 to 11.7 ka another 24,590 km 2 was flooded on the wide shallow-water platform, leading to the accumulation of 181 Tmol of carbonate within the GBR reef system, mainly in shelf edge areas. The carbonate accumulation rate per whole flooded area dropped more than two-fold to 5.7 mol yr −1 m −2 after 13.0 ka (Table S1 in Supporting Information S1).

Carbon Cycle Modeling
Processes which might have led to the pronounced decrease in GBR coral δ 13 C values after 13.0 ka (Figure 1c) are processes related to shelf platform flooding and changes in the coral's carbonate accumulation rate. Although largely variable the isotopic fractionation during carbonate production by reef corals (ε cor ) is on average reported to be slightly negative (Al-Rousan & Felis, 2013;Linsley et al., 2019;Swart et al., 2010). In our carbon cycle models, a 3-box model tailored for the specific situation of the GBR ( Figure S1 in Supporting Information S1) and the global carbon cycle model BICYCLE (Köhler & Munhoven, 2020; Figure S2 and Text S1 in Supporting Information S1), we assume an ε cor of −2‰, but test widely how this value and that of other parameters influence our results ( Figure S3 and Text S1 in Supporting Information S1). The carbonate accumulation rate in the GBR normalized per area, if implemented in the 3-box carbon cycle model, is the relevant flux that determines the seawater δ 13 C DIC . This rate was before 13.0 ka with 13 mol yr −1 m −2 and more (Table S1 in Supporting Information S1) responsible for an enrichment of seawater δ 13 C DIC of ∼1.5‰ compared to a run without any coral reef growth (scenarios 100 and 000 in Figure 3a). At 13.0 ka the carbonate accumulation rate per whole flooded area dropped more than two-fold from 13.0 to 5.7 mol yr −1 m −2 (Table S1 in Supporting Information S1) drastically reducing the positive enrichment of regional seawater δ 13 C DIC . This process alone can already explain a drop in δ 13 C values of ∼1.0‰, thus the dominant part of what is recorded in the corals between 12.8 and 11.7 ka (scenario 100 in Figure 3a; Text S1 in Supporting Information S1). When a similar carbonate sink via coral reef growth with identical isotopic fractionation is implemented into the surface equatorial Indo-Pacific box in the BICYCLE model (Text S1) such an enrichment of surface δ 13 C DIC leads to a small rise (nearly indistinguishable from zero) in simulated atmospheric δ 13 CO 2 and of +0.2 ppm in atmospheric CO 2 (Figure 4).
The decomposition (or respiration) of organic land carbon with lower amounts of 13 C on a newly flooded shelf platform, as has been suggested of relevance for other time periods (Köhler et al., 2011), potentially also contributed to the signal in δ 13 C DIC values of regional surface waters in the GBR during the last deglaciation. Implemented in our modeling framework, an assumed land carbon density of 30 kgC m −2 (Köhler et al., 2005b) with a δ 13 C value of −21‰ (Köhler et al., 2005a(Köhler et al., , 2011 released from the flooded areas, in general, shifts our simulation of seawater δ 13 C DIC toward smaller (less enriched) values (scenario 110 vs. 100 in Figure 3a and Figure S4a and S4b in Supporting Information S1). Around 13.9 ka, the about 6-fold increase in flooded area (Table S1 in Supporting Information S1), and therefore the higher decomposition rate of organic land carbon, leads to a drop in δ 13 C DIC by 0.5‰ (scenario 110 in Figure 3a and Figure S4b in Supporting Information S1). However, around 13.0 ka, this land carbon decomposition slightly dampens the drop in δ 13 C DIC caused by the reduced carbonate accumulation in corals by 0.2‰ (scenario 110 vs. 100 in Figure 3a and Figure S4a, S4b and Text S1 in Supporting Information S1). Following this process of land carbon decomposition, atmospheric CO 2 rose by another 0.1 ppm, while δ 13 CO 2 decreased by 0.001‰ in the global carbon cycle model simulations using the BICYCLE model (Figures 4b and 4c). Adding this process slightly reduces the enrichment effect of GBR carbonate accumulation on δ 13 CO 2 , but the resulting anomaly is still positive (Figure 4c), opposite to the ice core reconstruction (Figure 1c).
Prior to 13.0 ka, the narrow fringing reefs that developed along the shelf slope can be considered as relatively well-flushed and exposed to rapid exchange of reef waters with the open ocean. In contrast, after 13.0 ka, the wide expanses of the flooded shallow-water platform bounded at the shelf edge by the growing barrier reef system (Hinestrosa et al., 2019;Webster et al., 2018) can be considered as less prone to rapid mixing with open-ocean waters. If implemented in our 3-box-model, a decrease in water exchange rates between the GBR and the open ocean at 13.0 ka leads to a reduction of the drop in δ 13 C DIC caused by the previously implemented processes (scenario 101 in Figure 3a and Figure S4c and Text S1 in Supporting Information S1).
If all three processes, isotopic fractionation during carbonate production, decomposition of flooded land carbon, and water exchange rates are active together they can potentially explain the observed coral δ 13 C signal (scenario 111 in Figure 3a and Figure S4d in Supporting Information S1). The contribution of water exchange rates is rather minor here, and the explained amplitude seems to be with less than 1‰ still too small (Figure 3a). However, a closer match of the GBR record of coral δ 13 C values with simulations is possible by some adjustments in the chosen parameter values, for example, by assuming that land carbon density was with 60 kgC/m 2 on the shelf platform twice as high as on the slope (best-guess scenarios shown in Figure 3b). This is in line with the substantially warmer conditions (Brenner et al., 2020;Felis et al., 2014a; Figure S5 in Supporting Information S1) and developing mangrove forests (Grindrod et al., 1999) after 13.0 ka in the GBR, and with estimates of combined tree, dead wood and soil carbon density of mangrove forests today (Donato et al., 2011;Sanderman et al., 2018). Furthermore, some fine-scale adjustments in the simulated coral δ 13 C are obtained, when the variability of atmospheric δ 13 CO 2 is also taken into account (scenario "111-best-prescribed" in Figure 3b). This leads in δ 13 C coral to a Two scenarios (prescribed vs. internally calculated atmospheric δ 13 CO 2 ) are shown with optimized (best-guess) parametrization (the amount of respired land carbon on flooded shelf assumed to raise from 30 kgC m −2 to 60 kgC m −2 at 13.0 ka). In the prescribed scenario, atmospheric δ 13 CO 2 is fixed for 15.9-13.0, 13.0-12.0, 12.0-11.7 ka at −6.6, −6.8, −6.6‰ following maximum amplitudes in Antarctic ice cores (Bauska et al., 2016), while in the internal scenario atmospheric δ 13 CO 2 is internally calculated in the model. For comparison, Integrated Ocean Drilling Program Expedition 325 GBR coral δ 13 C data from Noggin Pass (NOG) and Hydrographer's Passage (HYD) are shown  10.1029/2021GL096495 7 of 9 drop by another 0.2‰ after 13 ka, allowing in principle for a smaller carbon release from flooded land than assumed so far.

Conclusions
Our modeling framework was able to identify the processes that were likely responsible for the last deglacial dynamics in δ 13 C values observed in the GBR corals. The combination of isotopic fractionation during carbonate production and decomposition of flooded land carbon contributed to the dominant features of the record of coral δ 13 C values, including the drop by 1.25‰ after 13.0 ka. Small variation of up to 0.2‰ from the imprint of changes in atmospheric δ 13 CO 2 values via air-sea gas exchange and a reduction in the simulated drop in δ 13 C DIC by reduced water exchange rates after 13.0 ka are only minor contributing processes. The nonlinear effect of the marine carbonate system, which via changing alkalinity and DIC concentrations alters the dissolved CO 2 concentration and thus the gas exchange rates, is essential for an understanding of these processes (Zeebe & Wolf-Gladrow, 2001). Therefore, this process identification can only be performed within a carbon cycle modeling framework. We note that extending the GBR record of Isopora coral δ 13 C values to younger than 11.7 ka, supported by carbon cycle modeling, could provide further insights into the processes identified for the last deglaciation and their relative importance under changing Holocene boundary conditions. The combined impact of both carbonate production and land carbon release from flooded areas in the GBR explains a deglacial rise in atmospheric CO 2 of 0.3 ppm only, and in atmospheric δ 13 CO 2 values of less than 0.001‰. Thus, the processes identified to be responsible for the large −1.25‰ anomaly in GBR coral δ 13 C values between 12.8-11.7 ka did not contribute to the second prominent minimum in atmospheric δ 13 CO 2 values during the last deglaciation. The synchronicity of the δ 13 C minima in both atmospheric CO 2 and GBR corals without one being largely responsible for the other should caution us to call for causal relationships without naming and quantifying a specific process.

Data Availability Statement
The coral data for this paper (Felis et al., 2014b) are available at https://doi.org/10.1594/PANGAEA.833408, and the Matlab/Octave code of the 3-box carbon cycle model and simulation results (Köhler et al., 2021), including shelf flooding and reef carbonate accumulation data as integrated in Figure 4a, are available at https://doi.pangaea.de/10.1594/PANGAEA.930114, at the World Data Center PANGAEA.  (Köhler & Munhoven, 2020) was used for the simulations. (a) Model forcing by the prescribed flooding of GBR shelf area and of coral carbonate (CaCO 3 ) production. Changes in atmospheric (b) CO 2 and (c) δ 13 CO 2 following three scenarios: (1) the prescribed accumulation of GBR coral carbonate with an isotopic fractionation during carbonate production of −2‰; (2) shelf flooding and release of CO 2 with δ 13 C = −21‰ from respired land carbon into the atmosphere assuming 30 kgC m −2 (15.9-13.0 ka) or 60 kgC m −2 (13.0-11.7 ka); (3) a combination of both.

Acknowledgments
We acknowledge the International Ocean Discovery Program (IODP) and the European Consortium for Ocean Research Drilling (ECORD) for drilling the Great Barrier Reef (IODP Expedition 325 -Great Barrier Reef Environmental Changes). We thank H. Fischer, M. K. Gagan and H. Kuhnert for early discussions, reviewers for comments, and Expedition 325 scientists for contributions to previous publications. This work received funding from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) -project number 180346848, through Priority Programme 527 "IODP". Financial support was also provided by the Australian Research Council (grant no. DP1094001). Open access funding enabled and organized by Projekt DEAL.