the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Feb 2026
LOAC-OCB v1.0: A model to explore terrestrial organic carbon burial along the land-to-ocean aquatic continuum
Abstract. Various aquatic environments along the land-to-ocean aquatic continuum (LOAC) retain and potentially bury significant amounts of organic carbon (OC). However, the total amount of buried OC, the relative importance of different ecosystems in this process, and the hierarchical influence of upstream systems on downstream burial dynamics remain uncertain. A major limitation in quantifying these processes is the absence of an integrative, process-based modeling framework operating at Earth system scales. Here, we present the LOAC-OCB model, the first global tool for simulating the transport and burial of particulate terrestrial OC along the LOAC. Using openly available data products, this steady-state model provides spatially explicit organic carbon burial (OCB) estimates at 0.0625 x 0.0625° spatial scale, incorporating 170,997 lakes, 6,000 reservoirs, 3,515 floodplains, and 377 coastal ecosystems worldwide. The model was evaluated through a multi-faceted validation using independent global datasets and previously published estimates. Our results indicate that the LOAC buries approximately 52.1 % of the particulate OC imported from terrestrial ecosystems, with reservoirs and coastal ecosystems showing the highest median OCB rates (94.3 ± 3.8 and 53.1 ± 14.1 g C m-2 y-1, respectively). Additionally, floodplains and reservoirs exert the greatest influence on global OCB fluxes, contributing 0.97 and 0.72 Pg C y-1, respectively. LOAC-OCB also enables further exploration of the interactions among aquatic ecosystems, shedding light on their interconnected roles in the global distribution of OCB and the relevance of burial processes in modulating terrestrial-to-ocean OC fluxes.
- Preprint
(7535 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-5359', Anonymous Referee #1, 12 May 2026
-
RC2: 'Comment on egusphere-2025-5359', Anonymous Referee #2, 03 Jun 2026
Henry-Pinilla and colleagues introduce the LOAC-OCB model, a framework designed to simulate the transport and burial of particulate terrestrial organic carbon along the land-to-ocean aquatic continuum (LOAC) on a global scale. The model integrates rivers, lakes, reservoirs, floodplains, and coastal ecosystems to generate spatially explicit global estimates of organic carbon burial at high resolution. Using openly available datasets, the authors assess the model's performance by comparing simulated river discharge, sediment export, sedimentation rates, and organic carbon burial rates with global observational compilations. Their results provide a new framework for quantifying the relative contributions of different aquatic ecosystems to global organic carbon burial and for investigating how upstream processes influence carbon sequestration downstream. While I fully agree on the importance of improving quantitative estimates of OC transport and burial along the LOAC, and I appreciate the conceptual design and ambition of the model, the current implementation relies heavily on empirically derived parameterizations based on a relatively limited set of localized observations. I recognize that such compromises are often unavoidable in the development of global-scale models and that some degree of simplification is necessary. However, a more thorough evaluation of model sensitivity and uncertainty is needed to establish confidence in the results, clarify the conditions under which the model can be applied reliably, and identify priorities for future model development.
Given the limitations outlined below, particularly the lack of a quantitative sensitivity and uncertainty analysis, I am unfortunately unable to recommend publication of the manuscript in its current form in GMD. In my view, the manuscript would require at least a major revision before it could be considered for publication. I hope that the comments provided below will be helpful to the authors in strengthening the study.
Some specific methodological examples that I believe require either improvement or a more thorough discussion and evaluation of their associated uncertainties are:
1) The preservation/burial of organic carbon (OC) is simply calculated based on a burial efficiency which is a function of linear sedimentation rate. The underlying sigmoidal relationship is derived from 30 data points from only 13 lakes and reservoirs. Given the importance of this parameterization to the model, is there really no larger dataset available on which it could be based?
This relationship is then adapted for marine coastal ecosystems (here mangroves, saltmarshes, seagrass) by reproducing a similar relationship reported in Katsev & Crowe (2015). However, while this study may report a relationship for the marine environment more generally, it is not obvious that it is directly applicable to the specific coastal ecosystems considered here. The implications and uncertainties associated with this assumption should therefore be discussed in more detail.
2) The approaches to estimate sediment trapping efficiency (TE) appear relatively simplistic and should be evaluated through a quantitative uncertainty analysis to assess their influence on model results. For lakes & reservoirs, it is unclear how many lakes were used to derive these relationships? For floodplains, a uniform TE efficiency is assumed everywhere, whereas for coastal ecosystems TE is also related to particle velocity and water depth. Is it reasonable to use these three rather different approaches together in a unified framework? It would be helpful to evaluate their performance against available observations and, where possible, compare the resulting TE estimates across environments. Such an analysis would provide greater confidence that the model predictions are not overly sensitive to the choice of parameterization.
3) More information is needed on the runoff that is used in “Input step 2”. I was unable to find an article describing the output. In particular, it woluld be helpful to know whether the dataset represents surface runoff only, or also subsurface flow and groundwater discharge which together define the total river flow. Given the central role of runoff in transporting sediments and OC through the LOAC, it would also be useful to assess the sensitivity of the model results to this input.
4) Given the ambitious global scope of the model, it would be helpful to first demonstrate its performance in a well-studied regional setting where (ideally) a relatively dense observational dataset is available. Such an application could provide additional confidence in the model and help identify its strengths and limitations.
5) I don’t find the comparison between model results and observations in Fig. 6 fully convincing. The number of observations is often orders of magnitude smaller than the number of modeled systems and may also come from only a limited number of locations. Therefore, I am not sure if this comparison provides a robust evaluation of model performance.
6) 3.3 OCBE: This part would benefit from a comparison of the modeled results with existing theoretical expectations and observational constraints. Such a comparison would help readers assess the realism of the predicted OCBEs across the different environments.
7) I am not sure that comparing median OC mineralization rates simulated by the model with CO₂/CH₄ emission rates from specific lakes or published estimates for coastal ecosystems is particularly informative. First, as the authors note, terrestrial OC is neither the sole nor always the dominant source of CO₂ and CH₄ emissions, especially in productive lakes and coastal ecosystems. Second, given the large variability in emission rates among lakes worldwide, observations from individual lakes are not directly comparable to median emission rates predicted by a global model. It would therefore be helpful to discuss the limitations of this comparison and its implications for model evaluation, or (if possible) compare specific lake estimates with the results of the model for the respective grids.
8) Are there specific areas where the model shows large discrepancies with other models (e.g., compared to Li et al. (2022) for the reservoirs)?
9) Figure 6: Is this truly an independent evaluation? My understanding is that the LSR–OCBE parameterization shown in Fig. 3 is also derived, at least in part, from the dataset of Henry et al. (2024). If so, the degree of independence between model calibration and evaluation should be clarified.
In addition, the simulated LSR (and, consequently, OCB) appears to be systematically lower than the observations. It would be helpful to discuss possible reasons for this bias. One potential explanation could be that the model represents only clay, silt, and sand fractions, while other particle classes that contribute to sediment accumulation (such as plant debris, algal biomass, biogenic silica, carbonates, or organic-mineral aggregates) are not explicitly represented. Could the omission of such materials contribute to the discrepancy?
Some detailed/technical comments:
line 186: Please provide a reference for this equation and/or the parameters used. Are they specifically for terrestrial OC?
Line 204: why is the depth of the river needed
Line 209 – 211: reference needed
line 218: “that depends on the flow direction”: unclear
Eq. (14): how is SEDbur different from Ocbur-ws in eq. (2)?
Eq. (15) OCws is only mentioned here
Section 2.5 Input data: I suggest changing the titles of the subsubsections (2.5.1 – 2.5.3) to something more descriptive
line 311 – 313: unclear
line 320 – 324: please rewrite, I found this a little unclear
line 338 – 339: unclear
line 354: GloSEM, is this a typo? I couldn’t find the abbreviation in the paper.
Line 356 – 358: unclear what is meant by “A separate sediment and OC mass balance are performed for each particle fraction”
Citation: https://doi.org/10.5194/egusphere-2025-5359-RC2
Data sets
Model Inputs/Outputs Daniela Henry-Pinilla https://zenodo.org/records/17476679
Model code and software
Model scripts Daniela Henry-Pinilla https://zenodo.org/records/17476679
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,124 | 646 | 82 | 1,852 | 280 | 353 |
- HTML: 1,124
- PDF: 646
- XML: 82
- Total: 1,852
- BibTeX: 280
- EndNote: 353
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript by Henry-Pinilla et al. presents a global model of terrestrial organic carbon (OC) routing through the inland-water network, including a simplistic representation of OC losses through burial and mineralization in rivers, lakes, reservoirs, floodplains, and coastal wetlands, including mangroves, salt marshes, and seagrass meadows (together defined as the land–ocean aquatic continuum, LOAC). The model is driven by a large number of input datasets (including published data on soil erosion rates, topsoil texture and OC content, the distribution of inland waters and wetlands, flow directions, and simulated runoff) to estimate OC burial and mineralization at an annual time step and at high spatial resolution (3.75 arcmin, approximately 7 km at the equator).
The representation of processes is very simplistic, partly borrowing from concepts developed in much older simple models (for example, Brune’s 1953 concept of trapping efficiency), and relying on assumptions that are in part questionable. For instance, the model assumes that coastal ecosystems such as mangroves, salt marshes, and seagrass meadows receive sediments primarily from rivers, while neglecting sediment inputs from the sea driven by coastal currents and tides. Similarly, it assumes that OC deposited on floodplains or other wetlands, which are periodically dry and more aerated, decomposes at rates comparable to those in permanently inundated benthic sediments, where anaerobic conditions are more likely to develop. Most importantly, the study appears to assume that gross hillslope erosion is equivalent to sediment inputs to inland waters, thereby neglecting the important role of footslopes, which intercept and store much of this material before it reaches the inland-water network.
The limited model evaluation against observations is not very convincing and indicates underestimation of burial rates by up to two orders of magnitude, although this may partly reflect biases in the observational database. Model results are compared mainly with previous global estimates, but here I identified several misinterpretations and inappropriate comparisons.
Overall, I find the model too simplistic and insufficiently evaluated against observations for this to be an acceptable model development or model evaluation study. I recognize that, given the limited observational data and the complexity of the underlying processes, developing a robust model for these dynamics is extremely challenging. I can accept that a simplistic conceptual model such as the one presented here could serve as a basis for re-evaluating and discussing the importance of OC burial in inland waters and associated wetlands at the global scale. However, that would make it better suited to a journal such as ESD or Biogeosciences rather than GMD. Even for such a venue, however, both the study and the manuscript would require substantial revision. While I therefore have to recommend rejection of the manuscript for GMD, I provide some comments below that may help improve it for submission to another journal with a less specific focus on model development.
L7: Maybe it’s better to report the resolution in arc-minutes and to give the approximate size in km in brackets.
L11-12: Please, report uncertainties for all numbers that you report as your results.
L22-24: Drake et al.'s budget doesn't account for burial in coastal sediments nor in floodplains. Also, only a non-determined fraction of these terrestrial carbon inputs is in form of OC. Following Battin et al. (2023), a major proportion of these carbon inputs may be inorganic carbon (IC). Moreover, this value of 5.1 Pg C/yr represents an upper bound estimate, which is based on an erroneous CO2 emission estimate from the lower Amazon river by Sawakuchi et al. (2017) that has since then been corrected with an erratum (estimate of CO2 emission from lower Amazon revised downward by a factor of 10). To my knowledge, the Drake et al. study has not been revised after that.
Drake et al. also report a flux of carbon to the coast, on which their budget calculation of terrestrial carbon inputs is based. To make a consistent estimate of the C lost in the LOAC, you should use that value instead the independent estimate by Liu et al.. In addition, note that Liu et al. also report the amount of dissolved and particulate OC exported to the coast. Finally, the fluvial exports reported in Liu et al. and Drake et al. are exports to the coast, while coastal ecosystems follow downstream in the LOAC.
Battin et al., 2023, https://doi.org/10.1038/s41586-022-05500-8
Sawakuchi et al., 2017, https://doi.org/10.3389/fmars.2017.00076
L36: Why would DOC settle? It would first need to adsorb or flocculate and so become POC, before it could settle!
L38: Here you should explain a bit what allochthonous and autochthonous POC represent. I guess it should be POC here, instead of simply OC.
L60: There would also be ORCHIDEE-Clateral that also represents POC transport, burial and mineralization. Zhang et al., 2022, https://doi.org/10.1038/s43247-022-00575-7
L64 : POM should be POC, POM was not defined.
L84-85: The mobilization of terrestrial POC into the inland water network is the most critical part. It would be easier for the reader if you could start with that. Even if you can reproduce the fluvial export to the coast, given that you don’t have good observational data to really validate predicted burial and mineralization, a too high or too low mobilization of terrestrial OC will simply be counterbalanced by an over- or underestimation of burial and/or mineralization.
L86: Mean runoff is not adequate for sediment transport. It would be good if you could include some high flows in your modelling framework, maybe represented by some upper percentile.
L89: Here maybe list again which kind of systems that includes. I don’t understand why you don’t include estuaries (incl. deltas and fjords) where a lot of carbon can be buried below the water surface, fed by riverine inputs.
L101: I know it comes much later in your MS, but as a reader I really want to know at this point how sediment generation is quantified. This is the most critical point, and in fact not trivial to do.
L102: Other studies suggest enrichment factors, as erosion removes preferentially finer soil textures and organic carbon (check: https://doi.org/10.1016/j.geosus.2025.100328)
L112: “water, sediment and POC”
L115: Please define riverine and palustrine wetlands! I guess some palustrine wetlands are unlikely to receive sediments (bogs and fens).
L120: “biogeochemical and physical”
L132-134: In coastal ecosystems, such as saltmarshes, seagrasses and mangroves, the burial may be a bit different. A lot of carbon buried here is produced in situ, and a lot of the sediments come in from the sea, for which coastal currents and tides are controlling factors.
L170: Mineralization does not release OC, it turns it into CO2.
Eq. 3 doesn’t make sense. You don’t conserve mass here if OC_input is supposed to be the sum of OC_min and OC_depo.
L204: As far as I know, the empirical models by Leopold and Maddock are based on bank-full discharge, which is often substantially higher than mean discharge.
L207: Please use Q for discharge, q is usually used for runoff.
L209-211: Do you mean bank-full discharge instead of baseflow?
L221-223: How is maximum discharge defined? Why not rather use a percentile, which makes more sense because you have flooding more than only once? Or what about average discharge above bankfull discharge. See this paper for estimating bankfull discharge: https://doi.org/10.1016/j.jhydrol.2011.08.004
L228-229: Do you assume that the eroded and mobilized material is fully disaggregated? Fine textures like clay and fine silt will rather remained bound in aggregates.
L237: You can have notable sediment and POC transport in rivers without overflowing the river banks. Some river systems may have a flood recurrence of > 1 year, so that the floodplain is not inundated every year.
L238-240: Depending on the water levels, only a fraction of the floodplain may be flooded.
L245-247: I think you have totally misrepresented the coastal wetlands. How do you take into account the effects of tides and coastal currents? How do you account for the effects of high productivity of these coastal systems on in-situ carbon production and burial. The blue carbon is not imported from the river, but rather produced in-situ!
L261-264: This is very problematic. Sediments that fall regularly dry and are often better aerated will show higher OC decomposition rates than benthic sediments.
Figure 3: Why exactly did you fit a sigmoidal relationship and not simply a linear relationship between OCBE and ln(LSR)? Also, your fit is strongly affect by the single value by Sobek et al. (2012). How would the fit change if you removed that one value? This makes your fit very uncertain. The values after Mendonça et al. seem to follow a different trend. Such uncertainties and potential biases need to be discussed.
L277-278: you mean “carbon mass balance”?
Eq. 14: the name SED_bur is confusing. It’s OC that is buried, not sediments.
L292: Give a reference for that dataset.
L338-339: Why don’t you calculate that from the WaterGAP data?
L346: You should have given that reference way earlier.
L353-354: The estimate of sediment mobilization into the LOAC is the most critical component of the overall estimation framework and should therefore be introduced, named, and described much earlier in the manuscript. However, this dataset does not actually quantify sediment delivery to the LOAC; rather, it represents gross hillslope erosion. A large fraction of this eroded material is likely to be intercepted and stored at the footslope or elsewhere within the catchment before ever reaching the LOAC. Using this estimate directly as a proxy for sediment inputs to the LOAC will therefore almost inevitably lead to an overestimation of sediment mobilization. If the aim is to estimate sediment delivery from a watershed to the aquatic network, MUSLE-based estimates would be more appropriate.
L354-356: HWSD also has information on soil texture and soilgrids (cited here) also has estimates on topsoil OC content. I don’t see why you would use different data sources for these two important soil properties. That introduces unnecessary inconsistencies.
Table 1: What do these WaterGAP simulations take into account? Are those naturalized flows or do they account for instance for irrigation? That would be important to clarify. Is “Runoff max” also derived from WaterGAP?
L369-370: You should use observed discharge from the GRDC (https://grdc.bafg.de/).
Figure 5: You should report here some useful metrics, maybe a combination of NSE, R2, RMSE and MBE (mean bias error). In panels a) and b), you should normalize fluxes by watershed size (like mm/yr for discharge and t/km2/yr for sediments). Otherwise I have the feeling that these correlations are only the result of differences in watershed size, which have nothing to do with the goodness of the model.
L386-380: These bootstrapped uncertainties around these median values have no meaning. For your simulation, they are extremely low due to the huge number of simulated values. You should rather try to quantify the actual uncertainty of your results.
L396-404, Figure 6: The large differences between modelled and observed burial rates may partly reflect biases in the observational dataset, which appears to be skewed towards systems with higher burial rates. Based on the spatial distribution of the observations, you should narrow the regions from which simulated values are extracted and perhaps apply a weighting according to the number of observations in each region, in order to improve the comparability between observations and simulations.
L405-412: How large may be the contribution of that autochthonous OC in the total OC burial? For lakes and reservoirs, it may be small, but what about floodplains and coastal wetlands?
L416: In how far is the Okavango Delta a floodplain? Isn’t it an inland delta and endorheic wetland?
L459-460: How does that really compare to other estimates? The Drake et al. estimate doesn't account for floodplains and coastal systems. There burial in lakes and reservoirs in that budget is only up to 0.6 Pg C/yr. The largest losses in that budget is the emission of CO2 to atmosphere, and it is not clear how much of that comes from OC that is decomposed in the LOAC. For that you could look at the estimate by Battin et al., 2023. How large is your mineralization flux? And, moreover, how large is the mobilization of terrestrial organic C into the LOAC that you get (and which should be the sum of burial, mineralization and export to the coast)? For this, you should also take into account that the estimates of fluvial exports of C (Drake et al., or Liu et al.) are exports to the coast; the coastal wetlands would come further downstream in the LOAC.
Battin et al., 2023, https://doi.org/10.1038/s41586-022-05500-8
L463: That is extremely huge considering that terrestrial C inputs are dominated by IC (see Battin et al., 2023).
L465-467: Again, this is the flux to the coast, the coastal wetlands come further downstream in the LOAC. Also, Liu et al. give estimates of dissolved and particulate OC exports.
L469-472: Regnier et al. include OC exports from coastal wetlands like mangroves and saltmarshes to the sea. I understand that you represent these wetlands rather as sinks than as sources of OC, am I right?
L480-481: I didn't read that you have a representation of floodplain erosion. Isn't there always a net-deposition on floodplains in your model? And do Li et al. represent burial on floodplains themselves? Or do they just represent burial in lakes and reservoirs? Maybe this comparison doesn't add up?
L483-484: Are you sure that Tranvik et al., 2009 used a mass balance approach to estimate OCB?
L485-486: But you do not represent burial on hillslopes, or do you? The burial on hillslopes is like much larger than that on floodplains.
L496: That is the upper range of estimates they synthesized, and this is based on an error published by Sawakuchi et al. for the lower Amazon (see my comment above). Also, only a fraction of that CO2 emissions would be contributed by OC mineralization within the LOAC. See Battin et al., 2023 and my comments above.
L518: That would require a corresponding estimate for soil erosion.
L520-522: You could also vary number of reservoirs depending on dam construction. More importantly, you could keep track of OC stock changes at erosion and deposition sites.