The authors assess the effect of soil erosion on soil Carbon fluxes at different spatial and time scales, based on a literature review and relatively simple modelling. The work is highly relevant, original, and of interest to the readers of Biogeosciences. Moreover, the work has large societal relevance in light of sustainable development goals with regard to land degradation neutrality and climate change.

// Thank you very much for this positive assessment.

My main reservation with regard to the work is that the literature reveals large uncertainties in the parameters that govern the C fluxes in (parts of) the total system at different timescales. Yet, in the Table 2 (summarizing parameters) and in the modelling that is reported in Fig. 4 the authors only report and use the estimates derived from a non-linear regression, without uncertainties. Hence the uncertainty is not shown in the final modelling result, which is a pity and a shortcoming of the work. I would encourage the authors to include uncertainties in the table and model, and represent these uncertainties in shading in the resulting figure 4. Such a representation would provide a much better image of the state of knowledge on this subject, including which parts of the system are least well understood.

// We fully agree and will certainly add the uncertainties in Fig 4.

In addition, I suggest that some additional effort is needed to improve layout and clarity of the figures, including legends and captions. Specific recommendations with regard to figures and text are added to the annotated PDF of the manuscript.

// Thank you very much for the suggestions, we will improve the figures based on your comments in the revised version.

Finally, I would encourage the authors to relate their findings to the present challenges with regard to land degradation neutrality and climate change. Their figure 4 shows that soil erosion is a net Carbon source at decadal timescales. This is exactly the timescale at which reducing atmospheric CO2 is most needed to reach Paris climate agreement targets. Thus, while (pre-)historic soil erosion may be a C sink in coming decades, present-day erosion will provide a C source in that same time period. This implies that preventing soil erosion contributes not only to food security, but also to climate change mitigation in coming decades.

// We will discuss our findings in the light of land degradation and climate change in the short-term and the long-term. However, Fig 4 refers to the time since agricultural conversion. Based on our literature review and meta-analysis, we suggest that recently converted land may provide a net source. However, most agricultural land has been converted for more than several decades and only recently converted land represents a source. As a result, the suggestion that present-day erosion represents a source is not consistent with our findings. We will describe this duality more clearly in the discussion.

General comments

The article attempts to do exactly as the title describes, to reconcile potentially competing perspectives on soil carbon erosion. The reconciliation is undertaken within the framework of scale which is used by the authors to demonstrate how these competing perspectives can exist at the same time and hence explain the paradox. I think this work contrasts markedly from the vast majority on this topic and many others in environmental science. The characteristics of that majority is typically atomised, perhaps even siloed, with a single perspective which is much easier to write, much easier for reviewers to understand and therefore readily published. Consequently, I congratulate the authors on this sophisticated integrated approach which is difficult to undertake and explain. The benefits of such a sophisticated approach are evident in the work, we have a proposal for understanding difference in perspective which enables the potential for the soil erosion community to re-gather momentum around the idea. I think the work is valid, straight-forward and effective which from my perspective equates to the work being incisive. On these bases I think the work should be published to act as a catalyst for further discussion on the topic. I have included below in the next section some specific comments which could form the basis for that discussion, would need clarification in the manuscript, but which I feel do not preclude the publication of this work.  

// We thank the reviewer for these very positive comments and are grateful for his view on the topic that is in line with ours. We agree that a single perspective is most often chosen because it is easier, while an integrated approach, like we present, is needed.

Specific comments

There is an implicit assumption by many researchers working on soil erosion that the processes are dominated by water erosion. This is of course not the case in the vast nearly 50% of the Earth’s land surface dominated by drylands where magnitude and frequency of wind erosion and dust emission very likely outweigh the influence of water erosion. Consequently, I would like to see improved clarification of the specific processes that are being considered throughout this manuscript. For example, starting with the title, should it read something like: “Reconciling the paradox of soil carbon erosion by water”. The first sentence of the abstract perhaps should more precisely be “The acceleration of erosion, transport and burial of soil organic carbon (C) by water in response to….”. Clarifications of this type throughout the manuscript, I think will serve to remind readers that much of the current thinking about SOC erosion is dominated by humid / temperate experience and measurements. Whilst the processes may be universal (notwithstanding a difference in fluid viscosity) the outcomes may be very different in relatively dryland regions. The authors might even like to include in their manuscript a statement that the paradox is only understood to occur in humid-temperate regions because there is far less work / understanding on this topic in dryland regions. The point I raise is perhaps best exemplified at Line 77 “On eroding hillslopes, soils are truncated, and C depleted subsoil material is brought to the surface layers.” In drylands, I think soils may not be truncated and the subsoil may not be C depleted. The implication of this difference is that in drylands, soil erosion may be a limiting factor in the balance between SOC decomposition and SOC redistribution. This thinking is already included in the Section on C recovery and evident in the text around Lines 100-110. However, it is not clear how or indeed whether drylands are included in the universal nature of the description, whether wind erosion and dust emission are a special case, or are not included. I have no problem with the authors simply clarifying the scope of the manuscript and not extending in to these larger issues, unless of course they are already included and just not explicit. In which case, I think there is a need to clarify on that basis.  

// The reviewer raises a very valid point. We will clarify the scope of our study (I.e. focus on water erosion with insights derived mainly from humid/temperate settings) in the title and throughout the manuscript in the revised version. The reviewer also brings up an interesting hypothesis about soil erosion being a limiting factor in the balance between SOC decomposition and redistribution in drylands. However, addressing this in our manuscript would be out of scope because of our focus on water erosion in humid/temperate settings.

The points above about soil carbon erosion in drylands raise the need to consider an additional clarification. There is only one mention (in the abstract) of the word organic linked to the words carbon erosion. I think the focus on soil organic carbon (SOC) erosion should be made clear (like the point above about water erosion), in the title and throughout the manuscript as appropriate. I think this is important so that the focus on SOC erosion is distinguished from soil inorganic carbon (SIC) erosion. The SIC cycling and erosion processes are prevalent in dryland regions but not widely recognised / connected in the literature on soil erosion. Consequently, it is not clear from the manuscript whether / how SIC processes should be considered in the paradox.

// We fully agree and will more clearly highlight that we focus on organic carbon erosion and not on SIC erosion.

The geography of SOC erosion demonstrates the overlap particularly in semi-arid regions of wind and water erosion processes. The significance of that interplay between wind and water erosion is its redistribution and difference in the sink of SOC. Wind erosion and particularly dust emission releases SOC in to the atmosphere and may transport SOC large distances from source, potentially influencing ocean carbon cycling. The main focus in the manuscript and the paradox, is the redistribution of SOC by water which is for a given erosion event relatively localised. Furthermore, there appears to be an implicit assumption that water erosion is dominant even in regions well-known to be influenced by wind erosion and dust emission. The question remains in my mind whether these differences influence the source-sink paradox. I recognise that this issue is beyond the scope of this manuscript. As in the previous paragraphs, I think there is a need in this manuscript to clarify the scope of the SOC erosion paradox described and perhaps even include a statement that defines clearly the focus. The impact of these clarifications will I think be the broader recognition that the geomorphic conveyor is beyond water and consequently there may then be much broader recognition of the source-sink across domains. I note that some clarity already exists e.g., Section 2 is entitled Transport in runoff and rivers. However, the preceding section is written in a way which gives me the impression that the commentary is universally applicable. However, I think we are a little way from that knowledge and understanding from across wind and water erosion and from drylands being combined.

// We fully accept this criticism but feel that including wind and dust emissions into our manuscript will make it less focused. We think that a clear definition of the scope (i.e. water erosion in humid/temperate settings) in the revised manuscript will address this comment.

The next point I have to make is a little tricky since it is not directly evident in the literature. Nevertheless, it is relatively easy to appreciate even if one does not accept it. Some of the C recovery section in the manuscript is based on the relation between net primary productivity (NPP) and SOC erosion. Whilst NPP is an important concept, it is grounded / implemented by the use of leaf area index underpinned by reflectance-based vegetation indices. The vegetation indices describe greenness which is due but cannot readily be assigned to dual signals of plant health and / or plant coverage. Consequently, if e.g., plant coverage changes as is partially evident in satellite measurements of global ‘greening’, then it is very difficult to distinguish plant coverage from plant productivity. Incorrect attribution of greening to one or the other will introduce Type I and II errors incrreasing uncertainty about the relation between NPP and C erosion. Although the duality of information contained in NDVI is well known, it has not generally been troubling because of the endemic assumption of stationarity and in modelling which is intrinsically steady state. However, a changing climate or other underlying changes, now confound the ability to understand plant productivity. So the relevance to this manuscript is that over long time periods underlying change may cause a difference in the response between SOC erosion and plant productivity, where that productivity is assumed stationary by using a contemporary vegetation index framework.

// Detecting the relation between SOC erosion and plant productivity based on remote sensing methods is indeed difficult. However, most of the studies used in our literature review are small-scale case-studies based on process measurements or space for time substitutions and do not rely on remote sensing. The issue highlighted by the reviewer is therefore a methodological issue that is relevant when upscaling or performing global scale monitoring of SOC erosion. Our work provides a perspective that should fuel further discussion on the topic and we feel adding this issue would dilute the concepts and main message of our study.

I’m not a great fan of merging a Discussion and a Conclusion. I wonder if what is provided in the labelling of that section of the manuscript is strictly neither of those, but is something more akin to ‘The implications of….’. I think many of the clarifications and issues raised here could usefully be included in that section to encourage workers to consider the implications from various perspectives.

// We agree and will provide an ‘implications of’ section in the revised manuscript.

Again, congratulations on putting together such a sophisticated and well-considered commentary. I believe and hope that it will act as an important catalyst for broad considerations of the C erosion paradox.

// Thank you! We really appreciate the constructive and thoughtful review provided by you!

Best wishes,

Adrian Chappell

The manuscript submitted by Kristopf van Oost attempts to be a review of the state of the research on the role of soil erosion for the global Carbon cycle. Depending on the study, erosion is seen as either a source or a sink of organic Carbon. Kristof van Oost and Johan Six argue, as in most of their previous work in the past 20 years, that soil erosion moves Carbon from the atmosphere into long-term geologic sinks. 

I have reviewed a manuscript by the two authors with the same title for another journal approximately a year ago. Apparently, the manuscript has been rejected by that journal. Comparing the two manuscripts reveals no major changes in both argument and literature.

// We strongly disagree with this statement. This study was indeed submitted to another journal, but the decision was a major revision, not a rejection. The editor of that journal also suggested to rework our paper as an original research paper. As also identified by reviewer 1 and 2, we see our work as a perspective that reconciles the opposing views and should serve as a starting point for future discussions on this topic. Hence, in the end, we felt that the BG Letters format was a much better outlet for our perspective than the original research paper format requested by the editor of the other journal. We also like to highlight that the paper has been substantially revised (based on the reviews received from the previous journal) for BG letters with much more observational data (colluvial and alluvial) and a conceptual framework that links the different space and time scales. We  are of the opinion that these major changes have substantially improved the manuscript and are in that sense grateful for the reviewers comments received from the original journal.

The key conclusion of this manuscript, as in the other publications by the authors on the topic is that the uptake, or dynamic replacement, of atmospheric Carbon at sites of erosion compensates for a part of the Carbon loss caused by erosion. In additon, eroded Carbon is deposited in long-term permanent sinks, leaving a negative net balance for atmospheric Carbon caused by erosion. Since many field scale studies show a major negatie impact of erosion on soil Carbon, the sink caused by dynamic and deposition in long-term sinks has been questioned. Kristof van Oost and Johan Six argue that the negative impact observed in field-scale and process studies does is balanced when taking a large-scale, long-term perspectve.  There are three key problems with this argument.

First, soil erosion rates are poorly constrained on a global scale. In their contribution to Nature Communications, Borelli et al. (2017, DOI: 10.1038/s41467-017-02142-7) showed that an increase of the resolution in their global scale erosion model by reducing raster cell sizes to 250 m reduced the estimated global erosion rate approximately by half. This would imply that also only half of the soil Carbon is eroded than previous models suggested, which in turn significantly reduces the potential for Carbon uptake at the sites of erosion. The number of studies currently published on improving the representation of topography in erosion (e.g. Panagos et al. 2015 10.3390/geosciences5020117, Schmidt et al. 2019 doi.org/10.1016/j.mex.2019.01.004) supports the position that the quality of Carbon flux modelling for regional to global scales currently is still poor.

// The suggestion that initial estimates of global soil erosion are most likely overestimates has been around for more than a decade now (eg Quinton et al NGS 2010). Our paper focusses on processes and the reconciliation of the opposing views in the context of space-time scales. As such, the absolute magnitude of agricultural soil erosion is not the topic of our paper. When accepting the revised global estimates (of c. 20-40 Pg of soil), organic carbon fluxes associated with soil erosion are still very high and of relevance for the global C budget. Furthermore, the papers cited by the reviewer only consider interrill and to some extent rill erosion because they are based on the RUSLE model. This implies that other erosion processes such as gullying, tillage erosion, harvest erosion etc are not considered yet and thus most likely underestimate erosion rates.

 A second problem arises from the lack of a geographically comprehensive data set on the actual impact of erosion on soil Carbon. The lack of reliable data on soil Carbon, especially from rangelands, has been acknowledged in many studies, including a 2014 paper in Nature that was co-authored by Johan Six (Pittelcow et al. doi.org/10.1038/nature13809) where the authors admit that the data on soil Carbon and from large parts of the planet are poor, mostly concentrated on European and American cropland, and thus the assessment of impacts of farming practices on soil organic matter are highly uncertain for most of Earth’s agricultural land. The final major uncertainty in the argument for an erosion-iduced C sink is the lack of data on the past soil and sediment organic matter content. Kristof van Oost and Johan Six argue that over long periods of time and large spatial scales, the sink effect dominates. To my knowledge, there is no source-to-sink study on a higher order catchment scale that traces eroded soil organic from slope to ocean, nor has this been attempted for the past. Individual sink reconstructions exist, but they lack information on original soil C source which has been eroded or at non-eroding sites, been modified by land use. This leaves the balance Kristof van Oost and Johan Six want to solve with more than one unknown.

// We are caught between a rock and a hard place. Our paper is the first study to collate and synthesize all available data. In our discussion, we also identify that there is a bias with an underrepresentation of tropical and dryland regions. We strongly believe that our assessment and concepts, although it may not be fully representative due to the lack of data, is informative and inform future discussions. Furthermore, we will include, based on the comment of reviewer 1, an uncertainty analysis in our revised manuscript. At the end, this is the main objective of our work. Secondly, there are source-to-sink studies that also include higher order catchments and these are included in our study (eg. Stallard 1998, Dymond 2010, Worrall 2016, Wang 2017 …).

In the light of these uncertainties in the data on Carbon erosion and depostion in space and time, the conclusions drawn by Kristof van Oost and Johan Six appear biased towards the Carbon sink argument. It is also not new, Sandermann and Berhe already made a similar argument in 2017 in their paper on The soil carbon erosion paradox in Nature Geoscience (10.1038/nclimate3281), also referring to Wang et al. (2017) and Chapell et al. (2016). This leaves the key statements of the manuscript presented by Kristof van Oost neither novel nor substantiated by new or more reliable data. 

// The main point of our work is that the source vs sink behavior can be reconciled, and not that erosion represents a sink or a source. We feel that reconciliation, rather than reiterating the paradox as done in other studies, is novel. The latter and is  also emphasizeded by reviewer 1 and especially reviewer 2).

Furthermore, the small size of the potential C-sink induced by soil erosion has been accepted in the scientific literature for about 15 years (e.g. Berhe et al. 2007 doi.org/10.1641/B570408) and the IPCC has followed this argument in its reports on climate change. This leaves the discussion on the impact of erosion on the global Carbon cycle with a small effect, but a large uncertainty. A review should therefore in my mind point out the uncertainties and identify the research needs, rather than developing a conclusion.

// We agree, we will revise our conclusion by avoiding to identify erosion as a source or sink and focus on the implications of our reconciliation framework.

I therefore cannot help but think that this review, in particular the submission of a previously rejected manuscript to a different journal, is an attempt to preserve the legacy of the previous research of the authors rather than being open to the arguments made by reviewers. I therefore suggest to reject the paper.

// As indicated above, this assessment is ungrounded: Firstly, the manuscript was not rejected. Secondly, we have revised our manuscript based on the constructive comments of reviewer input (new data and a conceptual apporach that provides a framework to align past research across spatial and temporal scales). Again, we strongly feel that the BG letter format is much more appropriate for a perspective paper as opposed to an original research paper format that was suggested by the editor of that previous journal.