the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Earth system models overestimate the sensitivity of apparent oxygen utilisation to age change in the deep ocean
Abstract. The biological carbon pump (BCP), involving photosynthesis at the surface and remineralisation at depth, maintains a significant vertical gradient in dissolved inorganic carbon (DIC), promoting the ocean's ability to absorb atmospheric CO2. Remineralised DIC is a good indicator of the strength of the BCP. It can be estimated from apparent oxygen utilisation (AOU) that measures the deficit of oxygen compared to saturation. AOU is projected to increase under climate change due to changes in remineralisation rates and circulation. However, the amplitude of the change is still uncertain. Here, we identify linear relationships between AOU trends and age trends in the deep ocean in simulations of the contemporary (1972–2013) and future (2015–2099) periods from five Earth system models (ESMs). Linear relationships identified within observational data for the contemporary period indicate that ESMs overestimate the sensitivity of AOU to age changes in the deep ocean. The study highlights the stability over time of the AOU sensitivity to age changes, suggesting an overestimation of the BCP strengthening inferred from AOU. Furthermore, our analysis underscores the substantial role of circulation slowdown in increasing remineralised DIC. These insights emphasise the challenges and opportunities to constrain future BCP projections due to circulation uncertainties.
- Preprint
(1491 KB) - Metadata XML
-
Supplement
(5049 KB) - BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2025-2566', Anonymous Referee #1, 08 Sep 2025
reply
Summary
Couespel et al. investigated the relationship between apparent oxygen utilization (AOU) and water mass age using both observational data and five Earth system models (ESMs). They found that ESMs tend to overestimate the sensitivity of deep-ocean AOU to changes in ocean ventilation (water mass age) when compared to observations. This overestimation suggests that the models may be overpredicting the future strengthening of the Biological Carbon Pump (BCP) based on AOU trends. Overall, the manuscript is well-written and presented in a clear, accessible manner. The research question addressed is of significant importance to the scientific community. I appreciate the considerable effort the authors have invested. However, I have major concerns regarding the manuscript in its current form that need to be addressed. I recommend a major revision.
Major concerns
- I have two major concerns regarding the methodology employed in this study. First, the authors calculate S^{ΔAOU}_{Δage} using ideal age and AOU outputs from the models, while for observational data, they use the mean age of IG-TTD and AOU. I assume that the calculations for observations are based on single-tracer constrained IG-TTD with assumptions such as Δ/Γ=1 and 100% saturation? Please clarify this in the Methods section. Nevertheless, a recent study from Guo et al. (2025) suggests that the mean age of IG-TTD derived from single-tracer can be biased towards younger ages due to imperfect assumptions about the shape of the transit time distribution and the short atmospheric history of abiotic transient tracers like CFC-12 and SF6. Additionally, the trend in the mean age of these tracers can be influenced by uncertainties in their mixing ratios, potentially producing spurious trends. I recommend calculating S^{ΔAOU}_{Δage} in the models using both AOU and the simulated CFC-12 (or SF6) to ensure methodological consistency. This approach would also allow testing the robustness of the results. To my knowledge, the model you used includes CFC-12 and SF6 simulations, which could be leveraged for this purpose.
Second, the authors calculated trends of AOU and age in the TS bins. In this case, the trend of age and AOU contains several pieces of information: (i) the mean state of water parcel ventilation timescales and AOU may differ across various geographical locations within the same temperature-salinity (TS) bin; (ii) the actual temporal change of age and AOU over time. Based on your Figures 2 and S5, where many data points indicate age trends over roughly 10 yr yr-1, I suspect that the first factor—the spatial variability within TS bins—is the primary contributor to the observed age trend. Given this, I wonder if it is appropriate to describe this as a “trend” with units of years per year (yr yr-). Although it seems it does not affect the meaning of S^{ΔAOU}_{Δage}, which is somehow like the respiration rate, it would be helpful to clarify this point. - I am pleased to see the analysis of AOU biases in the models (Figure 1), even though this may not be the primary focus of the paper. It is interesting to note that the models tend to underestimate AOU in the Southern Ocean, even below 4000 m. I raise this point because, to my understanding, models often struggle to accurately simulate the formation of Antarctic Bottom Water, which would typically lead to an overestimation of AOU in these regions. While I consider this an important question, I understand that providing a detailed answer may be beyond the scope of the current study, and it is acceptable not to address it here.
- The title of this study is “Earth system models overestimate the sensitivity of apparent oxygen utilization to age change in the deep ocean,” but I find that there is limited discussion and analysis of this specific point in the main manuscript.
- If I understand correctly, the models show a greater increase in AOU per unit water age increase compared to observations. However, this seems to contrast with findings from previous studies, such as Oschlies et al. (2018), which suggest that models tend to systematically underestimate the observed rates of ocean deoxygenation—mainly driven by increases in AOU—particularly in the deep ocean below 1000 m. It would be helpful to add a paragraph discussing this potential discrepancy in the Discussion section.
Minor comments
Line 56 - 58. The oxygen and age are hardly affected similarly by transport in the real ocean due to the spatial heterogeneity of respiration on isopycnals, as found for idealized isopycnals with prescribed patterns of respiration (Koeve and Kähler, 2016; Guo et al., 2023). “Sufficient oxygen concentration” is not the precondition that AOU and age are linear; instead, the even distribution of respiration rate across spatial is.
Line 78 - 87. I suggest including a table that summarizes key information about the models used in the study.
Line 88 - 89. Do you mean the model outputs are resampled analogously to the observations?
Line 90. It should be GLODAPv2.2023.
Line 129. Could you provide a reason why you exclude the Arctic Ocean?
Line 131. To clarify, please add a sentence indicating that all longitudes are considered for Southern Ocean water masses.
Line 151. What is the difference between B and Ɛ?
Line 161. Jenkins (1982) and Guo et al. (2023) could be included in the citations.
Line 171-180. See my major concern 1.
Line 195 - 267. I suggest dividing the Results section into several subsections for improved clarity and organization.
Line 219 - 221. AOU versus age across water masses can be the reason, but also the change of local biogeochemical processes?
Figure 2. Are the blue dots representing only values with a significant trend? If so, please clarify this. Otherwise, I am curious why Southern Ocean and Atlantic light/dense water make up around 98% of the deep ocean, yet many points remain grey—are these from the upper 1000 m?
Figure 4. Please improve the visualization of panel (a), as it currently appears quite cluttered, despite its importance. Are the zonal section panels essential, or could they be simplified or omitted to enhance clarity?
Line 279 - 290. This section is overall too descriptive. Additionally, as I mentioned in Major concern 1, I am concerned about potential inconsistencies in the methodology when estimating S^{ΔAOU}_{Δage} in models versus observations.
Line 299 - 302. Guo et al. (2023) could be cited here. These authors proposed two possible explanations for weak connection between age change and AOU change. First, changes in ocean circulation within a warming climate could alter water mass composition, as different water masses with varying biogeochemical histories are recombined over time—potentially introducing younger yet more oxygen-depleted waters into a given region. Second, local biological activity may influence the AOU signal independently of ventilation. For instance, even if ventilation slightly increases, an increase in local respiration rates could cause AOU to rise, reflecting biological consumption rather than changes in physical mixing.
References:
Guo, H., Kriest, I., Oschlies, A., & Koeve, W. (2023). Can Oxygen Utilization Rate Be Used to Track the Long‐Term Changes of Aerobic Respiration in the Mesopelagic Atlantic Ocean?. Geophysical Research Letters, 50(13), e2022GL102645.
Guo, H., Koeve, W., Oschlies, A., He, Y. C., Kemena, T. P., Gerke, L., & Kriest, I. (2025). Dual-tracer constraints on the inverse Gaussian transit time distribution improve the estimation of water mass ages and their temporal trends in the tropical thermocline. Ocean Science, 21(3), 1167-1182.
Jenkins, W. J. (1982). Oxygen utilization rates in North Atlantic subtropical gyre and primary production in oligotrophic systems. Nature, 300(5889), 246-248.
Koeve, W., & Kähler, P. (2016). Oxygen utilization rate (OUR) underestimates ocean respiration: A model study. Global Biogeochemical Cycles, 30(8), 1166-1182.
Oschlies, A., Brandt, P., Stramma, L., & Schmidtko, S. (2018). Drivers and mechanisms of ocean deoxygenation. Nature geoscience, 11(7), 467-473.
Citation: https://doi.org/10.5194/egusphere-2025-2566-RC1 - I have two major concerns regarding the methodology employed in this study. First, the authors calculate S^{ΔAOU}_{Δage} using ideal age and AOU outputs from the models, while for observational data, they use the mean age of IG-TTD and AOU. I assume that the calculations for observations are based on single-tracer constrained IG-TTD with assumptions such as Δ/Γ=1 and 100% saturation? Please clarify this in the Methods section. Nevertheless, a recent study from Guo et al. (2025) suggests that the mean age of IG-TTD derived from single-tracer can be biased towards younger ages due to imperfect assumptions about the shape of the transit time distribution and the short atmospheric history of abiotic transient tracers like CFC-12 and SF6. Additionally, the trend in the mean age of these tracers can be influenced by uncertainties in their mixing ratios, potentially producing spurious trends. I recommend calculating S^{ΔAOU}_{Δage} in the models using both AOU and the simulated CFC-12 (or SF6) to ensure methodological consistency. This approach would also allow testing the robustness of the results. To my knowledge, the model you used includes CFC-12 and SF6 simulations, which could be leveraged for this purpose.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
272 | 30 | 8 | 310 | 10 | 6 | 12 |
- HTML: 272
- PDF: 30
- XML: 8
- Total: 310
- Supplement: 10
- BibTeX: 6
- EndNote: 12
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1