| 23 Apr 2026
Carbon recovered in degraded soils under conservation management is highly vulnerable to loss
Abstract. The recovery of soil organic carbon (SOC) lost due to human activities provides an important opportunity to restore degraded soils and contribute to climate change mitigation. However, there is much debate about how much and for how long carbon can be stored. In this work we took advantage of a long-term experiment established in 1963 in southern Uruguay to evaluate whether integrated crop-pasture systems can recover soil C in a previously degraded site, while also analyzing the persistence of the newly sequestered C. We designed and calibrated compartmental dynamic models that represent the trajectory of soil C and radiocarbon in a soil quality restoration treatment involving the implementation of a crop-pasture rotation following a prolonged period (20 years) of continuous cropping use. We found that it is possible to recover the SOC losses by prolonged agricultural use through the incorporation of perennial pastures in agricultural rotations. Moreover, this recovery occurred at high rates (∼ 0.65 Mg ha-1 yr-1). However, we found minimal allocation of sequestered SOC to compartments that cycle at low rates and behave as kinetically stable. Instead, most of this new C was incorporated into an intermediate-kinetic-pool that cycles in the order of years to decades. Once management changed (in 2008, after 24 years), this recovered C was susceptible to significant selective loss. The results of this study indicate that, although the compartments associated with greater stability are susceptible to loss over short time scales as a result of agricultural management, their recovery would be a much slower process. The rapid SOC recoveries that can be achieved through conservation management appear to be in forms that are highly susceptible to environmental or management changes.
What a fascinating dataset with these distinct phases. The authors do a good job of explain the analyses they did and constraining their interpretations appropriately.
I’d like to see more discussion of tillage as an activity pushing C into the intermediate pool, possibly? Is there something about the reorganization of soil physically that prompts that intermediate turnover? Plus, this evidence is really contrary to other work suggesting that conventionally tilled systems have more C at depth, and no-till systems should accumulate it in the surface (eg Baker 2007).
It’s not clear whether all phases of the experiment are present in all years? Especially if not, but even if so, one caveat I would introduce is whether interannual variability could be a driver? I.e. during the third stage, is it possible that pasture was present in years that were just less productive, so inputs were lower and C turnover higher due to more microbial activity but less to eat? Not having read the other work referenced, I’m not sure whether the authors have annual productivity data, but the sensitivity analysis could definitely get much more detailed. A 20% reduction in biomass inputs may have less effect in a warm + wet year, for example, with high microbial activity year round, rather than a year where decomposition was limited by water or temperature.
L39-40 are there other important milestones along the way, like perhaps pre-industrialized ag in early 19th century? Also, does this metric include grassland forage or grazing?
L102-3 word missing or mistaken? Should and be “an”?
Fig 1 I think this figure might be easier to read if the boxes stretched over a number of years, or if year number was specifically labeled. It’s particularly confusing to understand whether the sorghum or wheat were seeded with the pasture the previous year, or if a one-year-old pasture was torn out to plant an annual crop again?
Fig 3 This is a nice representation of model fit, particularly the insets
L 307 “The better ability of the three-box model to represent the data (Fig. 4c,d) occurred because most of the SOC stock variations were explained by C stock changes in a pool with intermediate kinetics between the compartments represented in the simpler model”. I’m not following this sentence. Most of the change was in the intermediate pool, right? So how was that represented in the simpler model?
Table 1: is bolding intended to highlight something?
L384- 395. I don’t follow this explanation of the results of the input sensitivity analysis. To my eye, Fig 6 shows worse data fits than Fig 5 at both levels, particularly in stage 3, where observations actually decline while models held steady. How would reducing inputs in stage 3 leave you higher SOC values than maintaining inputs from stage 2? Is there more we need to understand re: rates of loss from each pool as well? Can the authors clarify more?
L406-08: minor, but I’d recommend swapping the word order in the sentence to highlight, “it is possible” before giving all the experimental details
L 491-500 What about soil and climate relevance of these meta analyses to the context of this LTE?
L510-516 It’s unsatisfying to claim we should promote practices that increase SOC in stable pools, without any concrete examples; instead saying that all studies have found it’s difficult to increase SOC in stable pools. Can you say more about this challenge, or acknowledge that we don’t know what those practices are, or discuss what you think they might be? I think the elements highlighted in the next paragraph, i.e. maybe the stage 3 practices didn’t keep the microbial necromass flowing, might be a good place to start.
L535 I’m surprised that no-till is on your list of practices to increase C, given that your evidence showed the no-till stage had decreasing soil C. Can you explain why this still looks like a valuable practice in light of your evidence?