Assessing carbon storage capacity and saturation across six central US grasslands using data-model integration

Wilcox, Kevin R.; Collins, Scott L.; Knapp, Alan K.; Pockman, William; Shi, Zheng; Smith, Melinda; Luo, Yiqi

doi:https://doi.org/10.5194/bg-2022-164

Preprints

https://doi.org/10.5194/bg-2022-164

Preprints

18 Aug 2022

| 18 Aug 2022

Status: a revised version of this preprint is currently under review for the journal BG.

Assessing carbon storage capacity and saturation across six central US grasslands using data-model integration

Kevin R. Wilcox, Scott L. Collins, Alan K. Knapp, William Pockman, Zheng Shi, Melinda Smith, and Yiqi Luo

Abstract. Future global changes will impact carbon (C) fluxes and pools in most terrestrial ecosystems and the feedback of terrestrial carbon cycling to atmospheric CO₂. Determining the vulnerability of ecosystems to future changes in C is thus vital for targeted land management and policy. The C capacity of an ecosystem (X_C) is a function of its C inputs (e.g., net primary productivity – NPP) and how long C remains in the system before being respired back to the atmosphere (ecosystem C residence time – τ_E). The proportion of X_C currently stored by an ecosystem (i.e., its C saturation – C_SAT) provides information about the potential for long-term C pools to be altered by environmental and land management regimes. We estimated X_C, C_SAT, NPP, and τ_E in six US grasslands spanning temperature and precipitation gradients by integrating high temporal resolution C pool and flux data with a process-based C model. As expected, NPP across grasslands was strongly correlated with mean annual precipitation (MAP), while τ_E was primarily a function of mean annual temperature (MAT). We link soil temperature, soil moisture, and inherent C turnover rates (potentially due to differences in microbial function) as determinants of τ_E. Overall, we found that intermediates between extremes in moisture and temperature had low CSAT, indicating that ecosystem C in these systems may be buffered against global change impacts on X_C. Hot and dry grasslands had greatest C_SAT due to both small C inputs through NPP and high C turnover rates during periods of favorable soil conditions. C_SATalso was high in tallgrass prairie due to frequent fire that reduced inputs of aboveground plant material. Accordingly, we suggest that both hot, dry ecosystems and those frequently disturbed should be subject to careful land management and policy decisions to prevent losses of C stored in these systems.

Received: 05 Aug 2022 – Discussion started: 18 Aug 2022

Kevin R. Wilcox et al.

Status: final response (author comments only)

RC1:
'Comment on bg-2022-164', Anonymous Referee #1, 27 Sep 2022

This is an interesting paper, examining the potential carbon storage capacity of grasslands. Overall, it is well-written, and provides interesting and important results. That said, there are many limitations.

First, the number of abbreviations, acronyms, and symbols make it almost unreadable. I understand that the authors are working to present efficiently, but it's too much. The paper would benefit from a major re-writing, using real words instead of writing sentences and even research questions and interpretations about CSAT, Xp, XC, TE, etc. Most of us know NPP, MAT, and MAT, but still, these others in long complicated sentences make the paper impossible to really understand without having the glossary at hand.

The list of "questions and associated hypotheses" is sloppy. The first one has been tested many times over (and they are very short on citations about this); the second is not even close to a hypothesis (they should either use questions or hypotheses instead of garbling them up together); the third doesn't have a causal explanation, necessary for a hypothesis (a prediction with a causal explanation). As I understand it Csat is the proportion of current carbon storage relative to the potential carbon storage, and the explanation doesn't relate to this. That third one is an example of an idea that a) doesn't make sense, and b) is impossible to interpret given the density of acronyms/symbols.

It's quite difficult to tell from the Methods which of the values were modeled and which were using the measured data from the sites. The explanation of "formal model validation" states that they validated the vegetation components. Notably, vegetation does not comprise the major pools of C in these systems. There are some problems (line 195) with this method of estimating carbon storage with depth, as this varies so much by soil profile and location - it's not terrible, but a caveat should express the limitaitons of the approach. I don't understand the "normalization" nor which slopes the authors are referring to (line 210).

For a number of these systems, a large proportion of the carbon stored is in recalcitrant soil pools. IThere needs to be MUCH more citation and analysis - and probably reconsideration of these residence times. Previous authors have shown that a good portion of the ecosystem carbon in these grasslands turns over on thousand year time scales, not time scales of 20 or so to 50 years. This alone gives me a great deal of concern about the paper. The paper should at least note that their estimates are orders of magnitude less than others have published.

The paper misses alot of literature about carbon storage, NPP, and decomposition across the region - it is almost shocking. There are very solid papers on the trends in soil carbon storage of grasslands vs. croplands (disturbed systems) across the central grasslands region, and on the trends in NPP and decomposition (k values) tested against mean annual precipitation and mean annual temperature that are never cited, in addition to other papers addressing mechanisms of C storage across the grasslands gradients in the region, in large scale databases and in a very original and key modeling paper for the region - this latter paper seems like a seriously important progenitor and the gap in citing it is pretty egregious. The paper does not really address the key issues about the effects of most disturbances (ie the distance between Xp and Xc, ack) on soil carbon storage, or what really happens that reduces C storage. It's quite theoretical, which is ok, but ungrounded in other literature on carbon storage either from empirical work or from modeling work in these grasslands. . There's no citation of where the effects of temperature on decomposition came from, from the empirical literature, used in the modeling (fig. A4 showing the modeled relationship between temperature and decomposition ) - it's as though the authors made up these relationships and values out of their heads, instead of from empirical data or others' work.

The conclusions are broad, and don't really present new insights. One reason that the hot and dry grasslands may have more C than they think is "their capacity" (the gap between the Xp and Xc) ould be that the model is not representing the systems well - that should be clearly state - it may not actually be a reflection of system dynamics. Finally, the conclusion spends a lot of time on burning, a relatively rare disturbance in some of the systems studied, and consideration of other disturbances should be included.

Citation: https://doi.org/10.5194/bg-2022-164-RC1
- AC1: 'Reply on RC1', Kevin Wilcox, 15 Nov 2022
  
  This is an interesting paper, examining the potential carbon storage capacity of grasslands. Overall, it is well-written, and provides interesting and important results. That said, there are many limitations.
  Thanks. And we feel confident in our ability to address the limitations of the manuscript outlined below.
  First, the number of abbreviations, acronyms, and symbols make it almost unreadable. I understand that the authors are working to present efficiently, but it's too much. The paper would benefit from a major re-writing, using real words instead of writing sentences and even research questions and interpretations about CSAT, Xp, XC, TE, etc. Most of us know NPP, MAT, and MAT, but still, these others in long complicated sentences make the paper impossible to really understand without having the glossary at hand.
  In the manuscript, we tried to balance efficiency with readability. However, based on the comment here and that from the other reviewer, it seems like we leaned too far to the efficiency end. If invited to submit a revision, we will address this issue.
  The list of "questions and associated hypotheses" is sloppy. The first one has been tested many times over (and they are very short on citations about this); the second is not even close to a hypothesis (they should either use questions or hypotheses instead of garbling them up together); the third doesn't have a causal explanation, necessary for a hypothesis (a prediction with a causal explanation). As I understand it Csat is the proportion of current carbon storage relative to the potential carbon storage, and the explanation doesn't relate to this. That third one is an example of an idea that a) doesn't make sense, and b) is impossible to interpret given the density of acronyms/symbols.
  This is another instance where brevity was our goal. Instead of having separate sections for questions and hypotheses, we decided to combine them. This sometimes resulted in questions without associated hypotheses (e.g., Q2) when we felt the current understanding was not sufficient to posit meaningful hypotheses. We agree that Q1 about relationships of NPP, 𝜏𝐸, and 𝑋𝑃 with MAP and MAT has been the focused on in previous studies. Yet, we argue that there is value to having both questions that test current ideas along with questions that test entirely new ideas, thus grounding novel work with more established patterns. This being said, we agree that we could execute the questions, hypotheses, and predictions better and will do so if invited to provide a revised manuscript.
  It's quite difficult to tell from the Methods which of the values were modeled and which were using the measured data from the sites.
  We apologize that the Methods were unclear about the sources of the values used in the paper. Wherever possible, we used empirical data to inform our analyses. However, much of the time, empirical data did not capture the fullness of information necessary for calculating holistic variables, such as carbon potential. In a revised version, we will be more explicit about where each piece of information came from. For reference, we summarize this here. GPP and NPP were obtained from the model (L118-120), which was forced using environmental data from the sites (L123-126) and vegetation components were calibrated using observations of ANPP and BNPP, as well as abiotic characteristics such as soil texture and soil moisture dynamics (L128-131 and Appendix D).
  The explanation of "formal model validation" states that they validated the vegetation components. Notably, vegetation does not comprise the major pools of C in these systems.
  We agree. This, along with difficulties present in measuring C cycling parameters in the field (e.g. slow C turnover), is why we used the data assimilation approach to estimate the processes that drive soil C pools. We used high temporal resolution to estimate 15 parameters associated with the C submodel (Fig. A1-4, Table B1), which we feel is in itself a valuable contribution to the field. We argue that this approach is a much more robust method of estimating C cycling pools and processes than simple model calibration and validation, since there could be many ways to achieve the observed carbon flux values (e.g. fast inherent turnover times, high temperature sensitivity).
  There are some problems (line 195) with this method of estimating carbon storage with depth, as this varies so much by soil profile and location - it's not terrible, but a caveat should express the limitaitons of the approach.
  The reviewer makes a good point about difficulties estimating carbon storage by depth. However, we do think it is important to estimate soil C from 0-20 cm to ensure that simulated and observed C pools match in the layers of soil they are estimated in. We will include caveats to these estimates in the text if we are invited to resubmit.
  I don't understand the "normalization" nor which slopes the authors are referring to (line 210).
  We mean that the values were scaled by the standard deviation and centered around the mean, which we will explain better in the text.
  
  For a number of these systems, a large proportion of the carbon stored is in recalcitrant soil pools. IThere needs to be MUCH more citation and analysis - and probably reconsideration of these residence times. Previous authors have shown that a good portion of the ecosystem carbon in these grasslands turns over on thousand year time scales, not time scales of 20 or so to 50 years. This alone gives me a great deal of concern about the paper. The paper should at least note that their estimates are orders of magnitude less than others have published.
  We appreciate this comment and will incorporate more citations on this subject (e.g. Conant et al 2011 GCB). We would like to note, however, that these carbon residence time estimates incorporate all carbon substrates including both those that turn over quickly and those that turn over much more slowly. If we only focus on the recalcitrant substrates from our study here, we see inherent turnover rates more in line (167-194 years) with the comment here by the reviewer, albeit not quite thousands of years. The discrepancy here may be due to our focus on relatively shallow soil layers. We propose to include two additional exercises that we think will provide a solution to the issue brought up by the reviewer. First, we will run a simulation where we manually set the inherent turnover time of the passive pool to some of the slower rates found in the literature and report what effects this has on carbon residence times. Not only will this provide insights to how much control these passive turnover pools have on total residence time, but it may also be incentive to better represent these processes into models. Second, we will expand the current section discussing some of these issues (L402-409), which will qualify our findings to the less recalcitrant soil substrates.
  
  The paper misses alot of literature about carbon storage, NPP, and decomposition across the region - it is almost shocking. There are very solid papers on the trends in soil carbon storage of grasslands vs. croplands (disturbed systems) across the central grasslands region, and on the trends in NPP and decomposition (k values) tested against mean annual precipitation and mean annual temperature that are never cited, in addition to other papers addressing mechanisms of C storage across the grasslands gradients in the region, in large scale databases and in a very original and key modeling paper for the region - this latter paper seems like a seriously important progenitor and the gap in citing it is pretty egregious.
  Originally, we had not included text (or associated references) for comparisons of croplands versus non-tilled grasslands since soil processes differ substantially in tilled versus untilled soil and by crop type. However, we recognize the value in this since much of the historical extent of grasslands is now cropland. If invited to revise, we will include discussion and literature on this topic such as Chambers et al 2016 J of soil and water cons., Smith et al. 2005 GCB, 2004 Eur J of Ag.
  We do include articles associate with the mechanisms and gradients of decomposition rates (Brandt et al 2010, Garcia-Palacios et al 2016), and NPP (Sala et al. 2012, Huxman et al. ,2004). However, we will expand these citations to include additional studies such as Zhou et al 2009 Ecosystems, Bontti et al 2009 GCB, Yahdjian et al 2006, Burke et al 1997 Ecology, Parton et al 1993 GBC, and others.
  With regards to addressing previous literature stemming from modeling results and large scale databases, we assume that the reviewer is referring to the LIDET data set and associated papers (e.g., Bonan et al 2012 GCB, Adair et al 2008 GCB, Bontti et al 2009 GCB, Harmon et al 2009 GCB, Parton et al 2007 Science). This was an oversight on our part – we appreciate the reviewer pointing this out and we will include these citations in a revised version.
  The paper does not really address the key issues about the effects of most disturbances (ie the distance between Xp and Xc, ack) on soil carbon storage, or what really happens that reduces C storage. It's quite theoretical, which is ok, but ungrounded in other literature on carbon storage either from empirical work or from modeling work in these grasslands.
  In the simulation of soil carbon capacity, the effect of disturbance is primarily through the loss of NPP inputs due to losses during disturbance, which we could expand on in the text (currently L394-400). A couple of references we could add would be Ojima et al Biogeo 1994, Lorenz and Lal C Seq in Ag Eco 2018).
  
  There's no citation of where the effects of temperature on decomposition came from, from the empirical literature, used in the modeling (fig. A4 showing the modeled relationship between temperature and decomposition ) - it's as though the authors made up these relationships and values out of their heads, instead of from empirical data or others' work.
  The temperature effect on decomposition (Q10) is shown as a density distibution obtained from the data assimilation process we conducted in our study. The bounds for the parameter were obtained from density distributions presented in Shi et al. (2015 Ecosphere), but the distributions shown in Fig A4 are obtained from the Markov chain Monte Carlo simulations comparing model output with observations.
  The conclusions are broad, and don't really present new insights. One reason that the hot and dry grasslands may have more C than they think is "their capacity" (the gap between the Xp and Xc) ould be that the model is not representing the systems well - that should be clearly state - it may not actually be a reflection of system dynamics. Finally, the conclusion spends a lot of time on burning, a relatively rare disturbance in some of the systems studied, and consideration of other disturbances should be included.
  We respectfully disagree with the reviewer on this point. We think that this study presents major conclusions that are important for addressing critical issues in the future. Although the model-data fusion approach makes assumptions that may not perfectly mirror the carbon cycle in all ecosystems, it allows for estimating ecosystem attributes that are difficult or impossible to estimate without very long and detailed observational records of carbon pools and fluxes – namely, it allows us to estimate future tendancies of ecosystem to gain or lose carbon. First, the finding here that fast carbon turnover rates in drylands without increased NPP inputs may cause future C loss is important for guiding land management and future research priorities. These findings are corroborated by recent empirical findings in drylands showing increased C losses in arid grasslands linked with wetter and productive years likely through increased microbial activity (Hou et al 2021 Biogeosciences). The effect of fire for carbon potential is another important finding from our study. In many ecosystems fire enhances NPP and is used as a common management tool. There is uncertainty surrounding the balance between increases in NPP and decreases in conversion of ANPP to soil C – our study provides evidence that the losses of ANPP may eventually outweigh the increases in total NPP leading to C loss. All of tthis being said, we would include additional discussion about uncertainties associated with these findings in a revised version.
  
  Citation: https://doi.org/10.5194/bg-2022-164-AC1
RC2:
'Comment on bg-2022-164', Anonymous Referee #2, 04 Oct 2022
General comments

This is an interesting study and a generally well-written manuscript. The authors provide a quantitative assessment of the C capacity and saturation of 6 US grasslands using both measured and modelled data.

The main weak points of the manuscript are (1) the inconsistent and confusing use of terms (2) the limited discussion around the role of soil biogeochemical processes for the C balance/capacity/saturation and (3) the minimal explanation of the data assimilation process i.e. what are the assimilated data about ?

After reading the paper carefully I am not able to explain how many key variables were estimated e.g. potential C. I can understand the key findings due to the very nice graphics. I think this reflects what the manuscript is lacking. All the elements of a good publication are in there but not given to the reader in a clear and coherent manner.

Specific comments

abstract : C can be lost via leaching also

abstract : "The proportion of ðc currently stored by an ecosystem (i.e., its C saturation – CSAT)" -- this is assuming a grassland ecosystem is C saturated, which is almost never the case (can be close to but not at Csat)

Page 2 : C capacity Xc and C content Xp become a source of confusion as there are points in the MS where Xc is presented as present/current C content and Xp as the potential C (e.g. L77)

L83: Csat is presented a "the distance between Xp and Xc" but later referred to as "proportion" and "percentage" which leads to different readers understanding this variable very differently

L101 : What is the land use history (at least the recent one) of the examined sites? Where they always grasslands?

Section 2.5 : I believe that all terms used in the MS should be described in one unique section early on. A table and/or schematic would help a lot
Citation: https://doi.org/10.5194/bg-2022-164-RC2
- AC2:
  'Reply on RC2', Kevin Wilcox, 15 Nov 2022
  This is an interesting study and a generally well-written manuscript. The authors provide a quantitative assessment of the C capacity and saturation of 6 US grasslands using both measured and modelled data.
  Thanks for the kind words.
  The main weak points of the manuscript are (1) the inconsistent and confusing use of terms (2) the limited discussion around the role of soil biogeochemical processes for the C balance/capacity/saturation and (3) the minimal explanation of the data assimilation process i.e. what are the assimilated data about ?
  Here, we respond to each of the three points above:
  We agree with the reviewer and we will rewrite the text to limit the number of acronyms used. We think this will substantially improve readability of the manuscript.
  
  This is another good point raised by the reviewer. We have identified a number of literature sources to include in the discussion to address specific biogeochemical processes responsible for the biogeochemical metrics presented in this study. See response to Reviewer 1 for some of these.
  
  Originally, we had minimized the text describing the data assimilation process, but based on this comment and some confusion surrounding Fig. 2-4, we will bring in additional text to better describe the data assimilation process. Briefly though, we used NPP inputs, soil temperature and soil moisture to drive the carbon turnover submodel. Then, we compared model output to observations of ANPP, root standing crop, plant litter, soil carbon, and surface CO2 efflux measurements through time to optimize C turnover, transfer, and sensitivity parameters. These parameters were then used to calculate carbon capacity.
  
  After reading the paper carefully I am not able to explain how many key variables were estimated e.g. potential C. I can understand the key findings due to the very nice graphics. I think this reflects what the manuscript is lacking. All the elements of a good publication are in there but not given to the reader in a clear and coherent manner.
  This is a good point. We think that the revisions suggested above by both reviewers will result in a manuscript that is more transparent.
  Specific comments
  abstract : C can be lost via leaching also
  
  We agree and will incorporate this into a revised version.
  abstract : "The proportion of ðc currently stored by an ecosystem (i.e., its C saturation – CSAT)" -- this is assuming a grassland ecosystem is C saturated, which is almost never the case (can be close to but not atCsat)
  
  We agree. We tried to make this point in Figure 1, where the C capacity changes through time. We will make this point more clear.
  Page 2 : C capacity Xc and C content Xp become a source of confusion as there are points in the MS where Xc is presented as present/current C content and Xp as the potential C (e.g. L77)
  
  We think that removing the symbols and using consistent terminology, based on both reviewer comments, should alleviate this issue.
  L83: Csat is presented a "the distance between Xp and Xc" but later referred to as "proportion" and "percentage" which leads to different readers understanding this variable very differently
  
  Same as previous response – this should be fixed during the revision process.
  L101 : What is the land use history (at least the recent one) of the examined sites? Where they always grasslands?
  
  This is an important point. Many of the sites were abandoned grazing land so there is a legacy of that management that may still be present. We will include this information in a revision.
  Section 2.5 : I believe that all terms used in the MS should be described in one unique section early on. A table and/or schematic would help a lot.
  
  We think this is a great idea, and would implement in a revision.
  
  Citation: https://doi.org/10.5194/bg-2022-164-AC2

Kevin R. Wilcox et al.

Viewed

Total article views: 559 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
396	148	15	559	1	10

HTML: 396
PDF: 148
XML: 15
Total: 559
BibTeX: 1
EndNote: 10

Views and downloads (calculated since 18 Aug 2022)

Month	HTML	PDF	XML	Total
Aug 2022	174	65	4	243
Sep 2022	81	28	3	112
Oct 2022	29	17	2	48
Nov 2022	43	10	4	57
Dec 2022	18	4	0	22
Jan 2023	20	7	1	28
Feb 2023	28	15	1	44
Mar 2023	3	2	0	5

Cumulative views and downloads (calculated since 18 Aug 2022)

Month	HTML	PDF	XML	Total
Aug 2022	174	65	4	243
Sep 2022	81	28	3	112
Oct 2022	29	17	2	48
Nov 2022	43	10	4	57
Dec 2022	18	4	0	22
Jan 2023	20	7	1	28
Feb 2023	28	15	1	44
Mar 2023	3	2	0	5

Viewed (geographical distribution)

Total article views: 535 (including HTML, PDF, and XML) Thereof 535 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 05 Mar 2023

Short summary

The capacity for carbon storage (C capacity) is an attribute that determines how ecosystems will store carbon in the future. Here, we employ novel data-model integration techniques to identify the carbon capacity of six grassland sites spanning the US Great Plains. Hot and dry sites had low C capacity due to less plant growth and high turnover of soil C so they may be a C source in the future. Alternately, cooler and wetter ecosystems had high C capacity, so these systems may be a future C sink.


Total:	0
HTML:	0
PDF:	0
XML:	0