Overall this paper disentangles key drivers of soil carbon dynamics in the CMIP6 model suite. This set of models is fundamental to developing the IPCC reports and thus understanding why the models return the results they do are critical to continual improvement and uncertainty quantifications for policy advising. In general, I think many of the key elements of a a strong and insightful analysis are here, but need a bit more connection and stronger caveats.

I would urge the authors to spend more time in their methods section integrating the C4MIP runs with the priming hypothesis. How would we expect false-priming to show or not show up in the various runs and why?

The assertion of false priming as the sole explainer for the correlative increase in NPP and reduction in turnover time is, perhaps, a bit strong. What are alternative explanations for the observed correlations? How are climate drivers delt with in both the NPP and Rh submodels? How does the second order NEP effects integration with this false-priming framework?

See line comments below:

=============

Ln 35: Should probably mention expected limitations on the nutrient fertilization effect and colimitation of water and other factors on the turnover time.

Ln 60: repeat information about the ESMs as needed to understand the results of this study. Citation hunts interrupt reading of the study.

Ln 125: Can you pull these ratios from the model to justify this assumption?

How are the different model runs going to be used in the analysis? How would you expect each scenario to behave given their driving conditions within the framework developed in Eqn 8? I suspect that key to the argument that this is a false-priming effect is going to be the C4MIP runs. Setting this up explicating in the methods section makes a lot of sense.

Ln 241: If this term was non-negligible then I would suggest dropping this framing from the introduction and maybe including a comment like “We thought this would be negatable but were surprised to find it was not.”

This false priming analysis feels very tacked on and needs to be introduced before the discussion section more clearly. How was this three box model parameterized? It appears that you are claiming that because you see similar patterns in this 3 pool model that you confirm that this is what is happening in the CMIP models. Maybe but there are other alternatives.

Varney et al. present results on soil carbon changes in CMIP5 and CMIP6 models under different levels of climate change, and quantify the contribution of productivity and turnover controls on these changes in soil carbon. They find that the spread in soil carbon responses across CMIP6 models was less than that across CMIP5 models, suggesting a potential reduction of uncertainty in 21st century soil carbon projections. The study shows that there are still differences in the relative contributions of controls (e.g., NPP and tau) on soil carbon changes across the models. They also illustrate a linear relationship between the change in carbon from NPP and turnover time, which they connect to the concept of false priming and demonstrate that this relationship is tighter across the CMIP6 models. In all, this is an interesting study with nicely summarized figures. Some findings could be discussed in greater detail with appropriate caveats, and I include specific comments below.

Main comments:

It is an important result that the study finds such differences across the two CMIP generations, and I think more discussion on this point could be helpful. Could the authors provide further details on a few of the soil C models that showed large differences between the CMIP generations? How did these soil C model representations change? 

In particular, did the soil C models change with regards to the number of pools from CMIP5 to CMIP6? I was under the impression that most were based on the the Century or DayCent models, which were developed decades ago. It’d be great if the authors could provide more details on the models here. Otherwise, the last sentence of the abstract about the “inclusion of more soil carbon models with multiple pools in CMIP6” does not have sufficient support currently.

How were the particular ESMs included in the analysis chosen? Line 55 says that this was due to data availability, but it is surprising that certain of the CMIP6 models (e.g., CESM2) did not have the necessary data for CMIP5 as well. Either way, it seems that 5-6 models (CanESM, IPSL, MIROC, MPI, NorESM, and HadGEM/UKESM) are included for both the CMIP6 and CMIP5 model output. These models would be good candidates to explore the earlier point above, regarding changes to the soil C modules and the number of pools in the CMIP5 and CMIP6 representations. 

How many of the models did not report a separate litter carbon pool? (These could be briefly listed on line 88.) It seems that the analysis could be more consistent by using only the soil carbon pool across the models. If not, additional rationale can be provided.

The finding of a turning point from increasing to decreasing soil carbon in Fig. 2 (lines 155-165) is really interesting, and the authors mention that this suggests a potential limit to the Cs increase. Why do you think this could be? And why does this turning point appear later or not at all in some models? Do any geographic regions contribute more to this turning point? Some more discussion here would be great, as this is an interesting finding.

The immediate response in respiration in Fig 10b looks surprising, especially the abruptness and shape of the tau curve. I guess this may be because the Cs1 and Cs2 pools both equilibrate almost instantaneously, with intrinsic turnover times of 1 and 10 years. However, it is difficult to see any of these details associated with the short-term response, because the x-axis spans 500 years. It would be helpful to focus on the first 100 or so years following the perturbation, as in Fig. 10a.

The authors could consider adding some discussion/conclusions on how their results on effective turnover times may connect with radiocarbon-based insights from soil carbon ages (e.g., He et al. Science 2016; Shi et al. Nat Geosci 2020) in data and ESMs. 

The last point in the conclusions (#6) reads as if false priming itself is a mechanism that affects soil carbon storage. However, it is in fact an effective bulk quantity that results from differences in mass-weighted and flux-weighted responses when there are multiple soil components with different residence times (as is the case in most models and in soil itself). It can thus be a useful quantity to further probe and diagnose model responses in response to perturbations. The authors may want to clarify and refine this last point.

Minor comments:

Lines 18-24: This background jumps around a bit, giving a case study for warming and then saying ‘therefore’ with a statement about elevated CO2 importance. Consider reorganizing this intro paragraph. Also, there could be a discussion somewhere here regarding the uncertainty resulting from underestimation of soil C ages in most ESMs (e.g., He et al. Science 2016). This is particularly relevant for the discussion of uncertainties resulting from increasing NPP and elevated CO2.

Line 30: There is only reference to Crowther et al. 2016 here, but just a note that there was conflicting evidence in a follow-up to that paper by van Gestel et al. 2018. 

Line 70: ‘subtracted’ instead of ‘taken away’

Line 120: can ‘be’ expanded

Line 314: Can you elaborate on what you mean by “offsets about 40% of the increase in soil carbon that would arise from the NPP increase alone” here?

Line 315: I’m not sure that it is a ‘disequilibrium’ per se, but rather an apparent quantity that emerges from differences in mass-weighted and flux-weighted responses.