the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Impact of the variability of O2 and CO2 biosphere exchange on modeled O2-based fossil fuel CO2 emission estimates
Abstract. Atmospheric oxygen (O2) can be used to partition components of the carbon budget at global and regional scales. These methods generally assume a constant net biospheric O2:CO2 exchange ratio (ERnet), although the exchange ratios of the gross fluxes that form ERnet, assimilation (ERa) and respiration (ERr), vary across ecosystems and time, suggesting ERnet may also vary spatially and temporally. To investigate this, we implemented O2 fluxes in the Simple Biosphere Model (SiB4) by assigning plant functional type- and carbon pool-specific ERa and ERr values based on limited observations. This produces the first global O2 biosphere flux dataset with spatially and temporally varying fluxes (0.5° × 0.5°, 3-hourly). Results show seasonal and spatial variability in ERnet, ranging from about 1.1 in winter to 0.8 in summer, with regional summer values differing by more than 0.1 across Europe. Temporal variability is primarily driven by changes in the relative contributions of photosynthesis and respiration, whereas spatial variability reflects differences in dominant plant functional types. The modeled ERnet range exceeds the currently recognized terrestrial O2:CO2 variability and may effect O2-based carbon partitioning methods. Neglecting ERnet variability can bias regional estimates of fossil fuel CO2 signals derived from atmospheric potential oxygen. At a strongly biosphere-influenced location, a modeled O2-derived fossil fuel CO2 signal of 1.5 ppm is overestimated by nearly 12 ppm when ERnet variability is ignored. These results highlight the importance of accounting for biospheric contributions in O2 partitioning methods.
- Preprint
(10867 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4491', Anonymous Referee #1, 06 Oct 2026
-
RC2: 'Comment on egusphere-2026-4491', Anonymous Referee #2, 30 Oct 2026
Review of “Impact of the variability of O2 and CO2 biosphere exchange on modeled O2-based fossil fuel CO2 emission estimates” by Faassen et al.
In this paper, the authors implemented O2 fluxes in the Simple Biosphere Model (SiB4) by assigning plant functional type- and carbon pool-specific ERa and ERr values based on limited observations and calculated spatiotemporal variations in ERnet, which represents net biospheric O2:CO2 exchange ratio. ERnet were observed by only a few studies at forest and urban sites (Faassen et al., 2022; Ishidoya et al., 2015, 2020), and I believe that this is the first model-based study to evaluate its spatiotemporal variability. Clarifying the mechanisms underlying ERnet variations is necessary for applying it not only to estimates of local CO2 budgets in forests and urban areas, as demonstrated in previous studies by Faassen et al. and Ishidoya et al., but also to improve the separation of the contribution from fossil fuel consumption and land biospheric activities, as discussed in this paper. I recognize the importance of the authors’ approach; however, based on the following major concerns, I believe that the manuscript requires major revision.
Major concerns
I could not find any validation of the simulated ERnet values. A model that is not validated against observations runs the risk of becoming a purely theoretical exercise. In fact, the authors calculated ERnet of 0.42 from the global average O2 and CO2 fluxes for 2021, and they consider this value to be unrealistic for the long-term mean biosphere exchange. This raises the question of whether the calculation of large-scale ERnet based on the values in Table 1 is realistic. In other words, given the very limited observational evidence supporting the significant difference between ERatm and ERnet, further effort should be devoted to validating the existence of such a difference through observations. The authors repeatedly discuss how the model might be validated but then point to the difficulties of doing so. However, scientific progress requires making at least some attempt at observational validation, even if only a limited one is currently feasible. Therefore, I encourage the authors to consider the following approach to validation.
1) The ff_CO2[APO] calculated using the variable alpha_B differs substantially from that calculated using the alpha_B of 1.1 at BIK in summer (Fig. 7j). In order to evaluate which ff_CO2[APO] value is more reliable, I consider it is useful to compare the ff_CO2[APO] with ff_CO2 simulated using an atmospheric transport model and European fossil fuel CO2 emission inventories such as EDGAR and ODIAC. If it is difficult for the authors to carry out such simulations, comparison with ff_CO2 values already simulated in previous studies (e.g., Liu et al., Atmos. Chem. Phys., 17, 14145–14169, 2017) may be useful as an alternative approach. Other approaches would also be welcome if available. In any case, the ff_CO2[APO] difference at BIK can potentially be validated, and I encourage the authors to make further efforts to validate it.
2) The authors did not compare the simulated ERnet with the observations. However, I understand that the authors have ERnet (ERforest) data observed at a boreal forest site in Finland. Could these observational data be compared with the model results of this study? In addition, there are several forest sites in Europe where O2 is measured at two heights, as the authors described. Especially, O2 measurements at two heights are presented in the figures by Popa et al. (2010) and Thompson et al. (2009). If the data can be available by contacting the authors of these papers, additional ERnet could be derived and compared with the model results of this study.
Other Specific comments
Line 257: “Proposed by (Pickers et al., 2022)” should be revised to “proposed by Pickers et al. (2022).”
Line 259 (equation (5)): I think “APO = O2 – alpha_B CO2” should be revised to “APO = O2 + alpha_B CO2” because alpha_B is assumed to be positive (1.1).
Line 275 (equation (7)): “APO = ff_O2 + bio_O2 – alpha_B ff_CO2 + bio_CO2” should be revised to “APO = ff_O2 + bio_O2 + alpha_B (ff_CO2 + bio_CO2)”.
Lines 414-420 and Figure 7: The distinction among ffCO2, ffCO2[APO], ffCO2 [APO(αB = 1.1)], and ffCO2[true] is unclear and confusing. The author wrote “ffCO2[APO] (Figure 7i and 7j)” and “ffCO2 [true], not shown”, however, I consider ffCO2 [true] are shown in Figures 7i and 7j by blue lines because they described “We analyze the difference between the fossil fuel atmospheric signal obtained with APO using either a value for alpha_B of 1.1 (ffCO2 [APO(αB = 1.1)]) or using the variable alpha_B values from SiB4 (ffCO2 [true]) (Lines 315-316)”. Please check the consistency in the use of these terms throughout the manuscript. Unlike in Fig. 7, the ffCO2 [APO(αB = 1.1)] and ffCO2[true] are explicitly show in Fig. A3.
Citation: https://doi.org/10.5194/egusphere-2026-4491-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 385 | 179 | 78 | 642 | 83 | 82 |
- HTML: 385
- PDF: 179
- XML: 78
- Total: 642
- BibTeX: 83
- EndNote: 82
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The study “Impact of the variability of O2 and CO2 biosphere exchange on modeled O2-based fossil fuel CO2 emission estimates” by Faassen and colleagues assesses variability in the O2:CO2 exchange ratio (ERnet) of the land biosphere. It quantifies this variability from measurements upscaled by the SiB4 model, and tries to assess its impact on quantification of fossil fuel emissions from atmospheric O2 data. This is an interesting and timely contribution, given the strong anthropogenic disturbance of the carbon cycle. The implementation of O2 into SiB4 is well described and sound (within some limits clearly described and discussed). In notable contrast, the assessment of impacts on fossil fuel quantification contains serious flaws; the description is partially convoluted, and the equations use inconsistent notation and inconsistent sign conventions. Nevertheless, I feel that the main points can still be made; see below for a suggestion how to avoid the flaws and unnecessary complications. I feel it is worthwhile to publish this study after corresponding revisions.
Major comments:
Problem 1:
In contrast to the well-behaved ERa and ERr for the gross fluxes, the ERnet of the net flux can become unboundedly large or undefined. The Authors do discuss this in the context of Fig 3. However, they then nevertheless use ERnet as if it was a well-defined quantity on all time scales. For the time scale of seasons (discussion of Fig 5 and 6), it may be possible to define ERnet in a meaningful way; though the exact definition would need to be made explicit in the text (ratio of seasonally averaged fluxes?). For any faster time scale (as in Fig 7 (e),(f)), ERnet does not seem to be meaningful anymore, and therefore should not be used. In particular, as I will explain in my suggestion below, ERnet is not actually needed to get fossil-fuel CO2 enhancements, and therefore should be removed there.
Problem 2:
Unless I misunderstand the mathematical symbols completely, Eq (11) is wrong. One cannot just divide Eq (10) by Eq (9) and neglect the fact that F is a space and time dependent quantity that is convolved with the footprints (as alluded to by the summation in Eqs (9) and (10) which I assume is meant to run over the spatial pixels of your domain as well as backward time). Such a simple division would only be possible under special conditions, e.g. if F was constant in space and time such that it could be taken in front of the sum in Eqs (9) and (10), but that is obviously not the case for fossil fuel emissions being heterogenous.
I would expect that the fossil fuel flux errors implied by the 1.1 assumption can be obtained by a suitably designed atmospheric inversion. I fully understand that this would be beyond the scope of this paper. Nevertheless, without that, the flux maps shown in Fig 8 calculated from Eq (11) do not say anything about realistic estimation errors. In addition to Eq (11) being faulty, the large flux difference in panel (e), for example, would only represent a short moment in time; thus even if the numbers were correct, they would not be relevant for fossil-fuel estimation at any time scales of practical interest.
Suggestion:
Given these severe problems, I suggest to entirely remove Eq (11) (including its derivation) and Fig 8, and to discuss the impact of deviations from a 1.1 emission ratio in terms of the fossil-fuel CO2 enhancement only (Fig 7 (i,j)). In addition (also to circumvent Problem 1), the derivation (and the description of the STILT runs) could be simplified as follows:
— As in Pickers et al (2022), only considering tracer enhancements (=differences from the background). In the model world, this actually makes things much easier as you can just use a zero background in STILT, w/o calculating any non-local contribution from a global model.
— Getting the simulated “traditional” APO enhancement via Eq (7) (corrected, see below) from the simulated O2 and CO2 enhancements of fossil-fuel and biosphere runs, using alpha_B=1.1.
— Getting “error-bearing” fossil fuel CO2 enhancements via Eq (6) from “traditional” APO (and simply with APO_BL=0), again using alpha_B=1.1.
— For the “true” fossil fuel CO2 enhancements, simply using the result of the simulation directly, without re-computing it via Eq (8). This avoids a calculation involving badly defined ERnet values which later cancel out anyway. As the simulated CO2 enhancement does represent the “truth” in your synthetic calculation, I do not see a conceptual reason either for not using this truth directly.
Of course, even though this allows a simplified and shortened description without explicit ERnet, the numerical results will still be the same as in Fig 7 (i,j). This means that the problem of how to interpret the short-term peaks of overestimation (like the 12 ppm peaks at BIK) remains. In my opinion, however, it would be entirely OK to leave this issue open, to be addressed in a follow-on study. Even without solving this issue quantitatively, the present study would still make the important point that deviations from a 1.1 behaviour need attention, and it provides the SiB O2 fluxes as an important step towards that.
Minor comments:
P1 L13: In light of the problems discussed above, citing a 12 ppm error in the abstract is misleading, as it sounds like a sustained error over a longer time scale. I feel that the evidence shown only allows a qualitative statement to be given in the abstract.
P2 1st paragraph: The later Sect 2.1 clarifying the relationship between OR and ER is very useful, thank you for putting it! It would be good to refer to this section here.
P2 L26-37: The paragraph frequently cites Friedlingstein, which may give the wrong expression that this is the original reference for the partition method. I feel it would be good to make the actual intellectual ownership more clear.
Eqs (1) and (2): maybe briefly introduce these relationships as representing net fluxes of carbon and oxygen
P3 L63-65: Better exchange these 2 sentences. Also, maybe reformulate into “we neglect fires and disturbances and therefore their influence on ERnet”
P4 L104: The sentence may be simplified and shortened as “In Sect. 4, we discuss the potential….”
Sect 2.1: As said above, this section is very helpful! You could also use it to clarify that you define ER always as positive (ie., it is actually the -O2:CO2 ratio)
Fig 1: What does the double arrow (“<—>”) mean?
After Eq (4): Maybe add “The CO2 gross fluxes…”
P7 L191: I’m not clear what “growth-stage processes” are, and why they can be neglected.
Tab 1: There is ERa = ERr(auto) throughout. Thus, why do you actually need the complex middle part of this table? Couldn’t it be removed?
Tab 1: I first read the footnotes as “squared” etc. Maybe put a space between number and exponent to avoid that
Eq (5): As you have defined alpha_B as a positive number (i.e., -O2:CO2), the sign is wrong here
P10 L262: The term “following definition” seems to be wrong in this context
P10 L262: Again, if alpha_F is defined as a positive number, it has the wrong sign in the inline equation
Eq (6) is confusing, which seems to arise from an inconsistent use of mathematical symbols (also in later equations). It seems that the symbol “ff_CO2” should rather be CO2 with subscript “ff”. If the authors aimed to avoid double subscripts, they may choose to put “ff”, “bio”, and “wo” as superscripts (consistently for CO2, O2, and APO).
Eq (7): Parentheses are missing around the term overbraced as “TotalCO2”, in addition to the wrong sign inherited from Eq (5)
P11, L297: I was stumbling on the formulation “together with the alpha_B …”. It seems that the simulation does not actually involve any explicit alpha_B, but rather that the effect of alpha_B is contained implicitly in the O2 fluxes from SiB given to STILT. It would be important to formulate this more clearly.
P11, L300: I assume that the 1.1 case did not actually use the O2 flux from SiB (and in effect not even the SiB CO2 fluxes, as the biosphere fluxes cancel out in APO in this case), but that remains unclear. To make it clear, add something like “…calculated for alpha_B=1.1 implying that the biospheric fluxes cancel out.”
P12, L316 (and paragraph before): If you only look at this difference, then the background is cancelling out, isn’t it? If so, why do you need to model it at all?
Eqs (9) and (10): The subscript “s” does not seem to distinguish these quantities from anything else. For the sake of reducing notational complexity, I’d suggest to remove this subscript.
P14 L337: Is “elevated” the right word when referring to fluxes? Don’t you mean “O2 source fluxes”?
P17 L407: Something seems to be missing in the sentence “Spatial and … fluxes.”
P17 L409, P18 L416: Make clear that what you transport is the fluxes, not ER. In the moment, this is a bit confusing, in particular the formulation at P17 L409.
Typos:
P1 L10: “affect”
P7 L191: Spurious “and”
P18 L416: Spurious “the” before “generate”
P23 L548: Spurious “a”
P24 L586: “phenological”
After Eqs (3), (6), (7): use a lower-case “w” as the sentence is being continued