the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Jun 2026
Constraining the Hydrogen Soil Sink and Photochemical Source: Insights from Atmospheric H2 Inversions (2003–2023)
Abstract. Hydrogen (H2) is expected to become an increasingly important energy carrier during the energy transition, likely leading to higher atmospheric H2 levels due to losses during production, transport, storage and usage of hydrogen. Multiple studies have shown this could impact atmospheric composition through interactions with the hydroxyl radical. However, the magnitude of this impact remains uncertain due to large uncertainties in the global H2 budget, particularly in the soil sink and photochemical source. To address this, we present a spatiotemporally resolved H2 budget derived using atmospheric inversions with the TM5 chemical transport model. With this approach, we infer a global mean soil sink of 52.8 [47.8–56.7] Tg yr−1 and a photochemical source of 34.6 [29.2–38.2] Tg yr−1 over 2003–2023. Relative to Ouyang et al. (2025), we estimate a soil sink that is 45 % and 35 % weaker in the Middle East and Oceania, and 45 % and 70 % stronger in South America and Russia, respectively. Our results further suggest that variability in the observed H2 growth rate between 2003–2023 was primarily driven by changes in the photochemical source from CH4 oxidation, together with a declining global soil sink at a mean rate of 0.23 Tg yr−2 . Finally, we infer a sensitivity of the soil sink to the El Niño–Southern Oscillation, strongest over diffusion-limited soils in tropical South America, with increased uptake during drier El Niño conditions and reduced uptake during wetter La Niña conditions.
- Preprint
(5035 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2026-3077', Alexander Archibald, 16 Jun 2026
- CC2: 'Reply on CC1', Firmin Stroo, 09 Jul 2026
-
RC1: 'Comment on egusphere-2026-3077', Alexander Tardito Chaudhri, 02 Jul 2026
This useful work is an great contribution that helps to deepen our understanding of atmospheric hydrogen by constraining its soil uptake and chemical source.
In this you have used a model inversion to better constrain the atmospheric hydrogen photochemical source and soil uptake. The model uses 100 ensemble members with Kalman filtering to follow observed atmospheric hydrogen over 20 years to 2023.
You base prior assumptions of soil uptake from implementations of two commonly used moisture-temperature-diffusivity limited soil uptake schemes. Different implementations of these schemes have appeared in various studies so I am glad to see that you directly compare them here.
I particularly liked how you were able to identify particular years and regions where your observation-based method disagrees with the theoretical prior.
Specific comments:These first comments concern experimental configurations that could be commented on and would strengthen these work, and possible substantial conclusions that may have been overlooked in this article:
- Paulot et al. (2024) and Tardito Chaudhri and Stevenson (2025) both identify constraints in hydrogen uptake from its observed seasonality. Presumably your setup is ideal to explore this. I see in Figs E5 g and h that the posterior uptake schemes require distinct seasonality compared to the assumptions based on physical parameterisations. What do your results suggest about the seasonality of soil uptake?
- You find faster soil sinks in South America and Russia than the synthesis (Ouyang et al. 2025) – how sensitive are these to the choice of scaled ocean H2 emission from nitrogen fixing?
- The time evolution of posterior fluxes from non-isoprene photochemical source (Fig 7d) and the soil sink (Fig 7e) are confusing. These require a ~10% increase in the non-isoprene source in 2016-2024 versus 2008-2016, and a 0.23 Tg/yr decrease in soil uptake over the analysis period. The priors had <5% increase in non-isoprene source and an increase in soil uptake.
- Please can you give an explanation for how credible the posterior non-isoprene source variability is (fig.7d)? From line 209, does your choice of the low state vector uncertainty in the methane source imply that this should be well constrained already?
- The prior and intuition (specific comment for Line 317 included) would suggest that the soil uptake should have increased over the period. Given that value chain emissions estimates of up to 20% are provided by both Trapani (2025) and Esquivel-Elizondo et al. (2023, Front. Energy Res), and the range of bottom-up and top-down studies (Table 2) that have higher H2 fossil fuel source – would it be feasible to explore configurations with a varied industrial source to estimate how likely changes here are to explain the trend? The approach of Paulot et al. 2024 ACP also includes industrial H2 in a revision CEDS v20210421. A substantial result might explore whether the recent trend is better described by an unaccounted H_2 emission, or a downturn in the soil uptake rate.
Specific comments line-by-line:
Line 12, 296, 360: El Nino ~ wet western but dry equatorial-eastern (looking at Cai et al. 2020)?
Line 22 and 35: It would be helpful to point to the existing constraints for these summarised in Table 2.
Patterson et al. 2025 also used new observations to find constraints on how sources and deposition might have evolved in the last century.
Line 24: The distribution of ocean and soil emissions might be poorly constrained. Does the calculation with seawater CO concentration match with REVISED Paulot et al. 2024 ACP supplementary figure 3 where SH ocean H_2 emission is reduced versus GFED4s?
Line 40: Northern Hemisphere subtropics + midlatitudes expected to have comparable or higher uptake (Paulot et al. 2024 ACP and Tardito Chaudhri and Stevenson 2025 ACP).
Line 96: Here you take the pattern of ocean emissions from CO, then introduce a scaling. My understanding is that we don’t necessarily have good constraints on ocean H2 emissions, and inventories for these add a source mostly south of the 25N land bisector (line 189). Increasing this source requires increased uptake south of this divide – previous comment, the revision in Paulot et al. reduces higher-latitude SH ocean emissions to meet observations.
Considering this, how robust is your South America posterior uptake to this choice of scaling?
Line 171 - 184: A linearisation is made to the uptake scheme, a posterior scheme is then found and the average deviation from linear uptake is then presented. Is it worth checking there aren’t particular regions where this linearisation has performed better and worse? This would reassure the reader that your conclusions are sound at the particular locations with the biggest changes in updated fluxes.
Line 213-221: Another possible error comes from flask observations typically being made at similar times of day at each site, Forster et al. 2012 record ~10ppb variations in hydrogen from diurnal cycles, with notable drops during night-time inversions.
Line 247-248: Please can this sentence be made clearer? ‘Due to the larger freedom to adjust sources in sinks …’
Line 267: Would this suggest that the prior configuration would lead to too-high H_2 in the southern hemisphere versus Northern Hemisphere? Would it be clearer to also include the latitudinal distribution that the prior would have achieved and what the posterior achieves?
This might be beyond scope, but would a configuration with higher anthropogenic H_2 emission (more NH) see increased but similar pattern of H_2 uptake as the prior?
Line 317 and Fig 7e: This result is counter-intuitive. If we assume rates of microbial activity increase up to about 36C and that global patterns of soil moisture have not generally changed over 2003-2023 then intuition would suggest that it would be more likely that soil uptake would have increased rather than decreased?
Line 325: However, this could be a useful result. Is it easy to estimate HD depletion rates that would be expected for the constraints found in this work versus an emission change explanation?
References:
Patterson et al. 2025, https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/2025JD043662
Paulot et al. 2024, ACP, https://acp.copernicus.org/articles/24/4217/2024/
Esquivel-Elizondo et al. 2023, Front. Energy Res, https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2023.1207208/full
Cai et al 2020, https://www.nature.com/articles/s43017-020-0040-3
Patterson et al. 2025, https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/2025JD043662
Forster et al. 2012, https://www.tandfonline.com/doi/full/10.3402/tellusb.v64i0.17771
Citation: https://doi.org/10.5194/egusphere-2026-3077-RC1 -
RC2: 'Comment on egusphere-2026-3077', Anonymous Referee #2, 22 Jul 2026
Review of “Constraining the Hydrogen Soil Sink and Photochemical Source: Insights from Atmospheric H2 inversions (2003-2023)”, by Firmin T. Stroo et al.
General Comments
This is an interesting study that seeks insights into uncertain terms in the global H2 budget using atmospheric observations. It is a bit unclear on a few aspects (see specific comments below) – this may be partly related to my fairly rudimentary understanding of the process of running atmospheric inversions – but I think several points need clarifying. If these points can be satisfactorily addressed, I will be happy to recommend publication.
Specific Comments
L53 “…Bousquet et al. (2011) used observations from 1991-2004, despite nearly two decades of additional observations being available since then.” This comment seems a bit unfair, since Bousquet et al. could hardly have used observations from the future in 2011! I guess you mean something slightly different.
L92 I am slightly unclear on the fossil fuel related H2 emissions derived from scaling CO emissions. If I understand correctly, your simulations cover 2003-2023 (it would be useful to explicitly state this near the start of Section 2.2), and use CO emissions from CAMS and H2:CO emission ratios from CEDS. So presumably there is a significant trend of these H2 emissions over 2003-2023? Later (e.g., Figure 7), you discuss the growth rate of H2 and relate this to trends in the photochemical sources of H2, the soil sink, and biomass burning emissions. But what about the trend in anthropogenic emissions? This doesn’t seem to be documented anywhere in the paper. Don’t we also need to know the contribution of that to trends in atmospheric H2? Isn’t there also some (quite large?) uncertainty in the magnitude and trend of this source of H2? Or are you assuming this is small and insignificant? If so, say so.
L99 The biogenic source of H2 is also unclear. You say you use the biogenic CO fluxes from MEGAN, scaled to 3 Tg/yr (I guess you mean 3 Tg(H2)/yr?). Is this fixed over the whole period, with no inter-annual variation? Or is it scaled to 3 Tg/yr in the first year, then allowed to vary inter-annually, based on MEGAN? How do you represent the biogenic source of H2 that comes from isoprene (and other biogenic VOCs)? (Is the biogenic CO somehow approximating the source from BVOCs?) Presumably isoprene emissions are included explicitly in the full chemistry runs (L112)? (You say you derive the photochemical H2 source from isoprene oxidation, but don’t say anything about isoprene emissions). Similar to the previous comment, why don’t you show the trend and variability in the biogenic H2 source in Figure 7? If you are assuming it is small and insignificant, explicitly say so.
L129 It is perhaps worth noting that the seven ecosystems in the Sanderson et al. (2003) scheme do not very obviously relate to different soil types, whereas the Bertagni et al. (2021) scheme does this more explicitly.
L139 Thanks for referencing the source for the H2 + OH reaction rate (Stuhl and Niki, 1972). But it would also be useful to know if this is also the recommendation from more up-to-date compilations of reaction rates (i.e., IUPAC, JPL), or perhaps they do not document this reaction?
L207 You use “statevector”, “state-vector” and “state vector” at various points in the text. You should probably decide which one is correct and consistently use it.
L213 Please explicitly define what you mean by “model-data mismatch”. Is this the maximum difference between model and observations that you allow in the inversions?
L244 The caption for Figure 3 doesn’t define the cross and triangle points (you may just want to say “defined in the text”).
L278 The caption for Figure 6 says “Photochemical source…”, but the figure only seems to show soil sink results? (Are you referring to Figure E4?) The text discussing Figure 6 (l272-274) also seems to imply the figure shows results about the photochemical source. Please clarify.
L280 Suggest add “…shows the observed annual H2 growth rate at 2 sites (SPO and MLO)…”
L280-285 Suggest say Figure 7a, 7b, 7c, 7d rather than “panel a” etc.
L291 Figure 7 – See my earlier comments: what happens to other H2 emissions and the isoprene-related source of H2?
L295 Suggest change to “…in most regions, including tropical South America, the soil sink is dominantly diffusion-limited…”
L300 Figure 8 Does the Oceanic Nino Index have units of K, or is it dimensionless?
L301 Section 4 – suggest add some discussion about trends in anthropogenic H2 emissions (based on earlier comments).
L328 In addition to the isotopic signature of fossil fuel emissions changing with engine technology, I suspect the H2:CO ratio of the emissions may also have changed.
L354 There has been some work on the HCHO photolysis rate by Holland et al. (https://meetingorganizer.copernicus.org/EGU24/EGU24-3142.html). I am unsure if there is an associated publication, rather than just a conference abstract.
P21 Figure A2 caption. Are these data based on all ATom flights (N=35 mentioned in the text, L395)? This seems inconsistent with the 48 ATom “Stations/Flights” noted in Table E1?
L404 I understand that the underestimation of HCHO ”…is likely related to an overestimation of the photolysis rate of CH2O…” But if this is the case, and HCHO was higher, but the flux through HCHO + h.nu was less, would the H2 source be larger or smaller?
L418 I don’t really understand what a “5-year value” (or “monthly scaling” (L416)) means here. Do you mean you use a 5-year running mean (with a moving window) (for each month)? What does “multi-year correlations” in the caption for Table B1 refer to? Please clarify.
L420 Here you refer to “the isoprene source” – so, related to my earlier comment about biogenic emissions, are you using MEGAN for isoprene emissions? (Unclear in Section 2.3).
L424 I am surprised the inversions sometimes generate “negative photochemical sources”. Should the inversion be allowed to do this, since that would be physically impossible? Is it not feasible to force the inversion to not do this? (On the other hand, a “positive soil sink” – i.e., a soil source, is not out of the question – occasionally chamber measurements of soils show a source rather than a sink, although, again, if you feel this is physically unreasonable, I don’t understand why you can’t force your inversion to not allow this).
P24 Table B2 footnotes – all the numbers should have units of ppb, i.e., 1.5 ppb MDM, etc.
P25 Figure B3 caption – what does “innovation” chi-squared mean? Please clarify.
P29 Figure E1 – suggest add a line at 25N to separate the two regions.
P30/31 Figure E2/E3. These maps do not show global mean values! (They vary spatially!)
P32 Figure E4. Please clarify that “non-isoprene source” and “isoprene source” refer to photochemical sources of H2, and not H2 emissions. (I assume).
P34 Table E1. See earlier comment – Are there 35 or 48 ATom flights?
L498 neither -> none?
Citation: https://doi.org/10.5194/egusphere-2026-3077-RC2 -
RC3: 'Comment on egusphere-2026-3077', Anonymous Referee #2, 23 Jul 2026
PS I wanted to add a couple of further comments:
L224 The choice of the four observational metrics is not well justified and feels a bit arbitrary. The choices are not unreasonable, but could you better justify: Why only five stations were chosen for the mean RMSE metric (and why those five?); Why MLO for the growth rate (why not more/all stations?); Why not all/more high-latitude sites for the interpolar difference? And Why only use Mace Head for the seasonal amplitude – why not more/all sites?
I also wasn’t entirely clear on how these metrics were used to identify the best-performing configurations in Figure 3. Is this a quantitative process? E.g., is each metric weighted equally? Given the relatively arbitrary choice of metrics, this process felt only semi-quantitative at best.
Maybe the point is these metrics are only used to loosely identify global scalings that are refined in the detailed inversions, so they don't need to be very rigorous. Is that the correct way to think about it? This process could perhaps be more clearly explained.
l247 Should this read "Due to the larger freedom to adjust sources and sinks..."?
Citation: https://doi.org/10.5194/egusphere-2026-3077-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 93 | 46 | 11 | 150 | 10 | 13 |
- HTML: 93
- PDF: 46
- XML: 11
- Total: 150
- BibTeX: 10
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Stroo et al. present an important study on the use of inverse modelling to constrain two of the major uncertainties in the hydrogen budget: the magnitudes of the photochemical source and the soil sink.
This is a nice study and adds to our currently expanding understanding of the atmospheric chemistry of hydrogen.
I had a couple of small point I thought worth discussing, regarding the section on the photochemical source of hydrogen (4.2).
1. Photolysis rate versus photolysis rate constant (or frequency).
Please check throughout section 4.2 as to the use of photolysis rate (J*[HCHO]) versus photolysis rate constant (or frequency; J). On reading this section I was confused if the rate of HCHO photolysis was being talked about and or the rate constant for photolysis was being discussed. My guess was that most of this section is discussing the rate constant for photolysis of HCHO to form H2 (some times called J-molecular).
2. Evaluation against ATom data.
I found the evaluation of HCHO mixing ratios from the model and ATom to be useful and help clarify the point that the reduction in photochemical source (in the posterior) could be coming from an overestimate in J-molecular. This can and could be investigated. I would encourage the authors to consider comparing the modelled J values with the ATom determined J values or potentially equally useful, combining the ATom HCHO and J values to create an observed photochemical H2 source to compare against the model. Both I think would help clarify whether the issue with the model is coming from incorrect treatment of HCHO photolysis or something else.
References:
Hall et al. (2018), "Cloud impacts on photochemistry: building a climatology of photolysis rates from the Atmospheric Tomography mission," Atmos. Chem. Phys. 18:16809–16828, doi:10.5194/acp-18-16809-2018