Comments on the manuscript

Dynamic vegetation reveals unavoidable climate feedbacks and their dependence on climate mean state

by Braconnot, Viory, Marti

Pascale Braconnot and coauthors present a study on vegetation-climate feedbacks. They use the new IPSLCM6 with four different parameter sets with respect to the representation of photosynthesis, bare soil evaporation and two parameters defining vegetation competition and distribution. They thoroughly analyze the effect of changing the parameter sets on the simulated global climate and vegetation patterns for preindustrial and for mid-Holocene climate. I am convinced that the specific results of their study strongly depend on the climate system model used and on the specific choice of the parameter sets. Other climate system models and other variation of other parameters would presumably yield other results. And that should explicitly be stated. However, I am also convinced that modeling groups can learn from the paper when addressing the general issue of tuning and exploring the sensitivity of vegetation-climate feedbacks on the choice of vegetation and soil parameters. Thus, I recommend publication of the study with minor corrections, but I strongly suggest skipping the term “unavoidable” in the title and in the text. This term is not explained in the text. It just sounds alarmistic, as if the model would have a choice to “avoid” any negative consequences of model tuning. Quite the contrary, model tuning is done to improve the model performance.

When reading the paper, I see that the new IPSLCM6 yields a much greener mid-Holocene Sahara than the former IPSLCM5 did. That is an exciting result. The authors highlight this achievement in one sentence (line 278/9) and a half-sentence (line 284/285). It would deserve more appreciation in the conclusions. Perhaps, a figure with a zoom on African biomes, using the biomization tool by Dallmeyer et al (2019), for example, which has already been applied to ORCIDEE PFTs, would be useful for a better comparison with other ESM simulations. But I leave this to the authors to decide. I do not like reviewers who suggest completely re-writing a paper or writing a new story. Hence, I only would like to encourage the authors considering an additional study with a perhaps even more interesting focus on the effects resulting from the upgrade from IPSLCM5 to IPSLCM6 versus changing the parameterization of bare soil and vegetation.

The sensitivity study is clearly written. Perhaps a more formal analysis using the Alpert-Stein factor separation would yield a better understanding of feedbacks and synergies between feedback. But, again, I do not insist on doing a new analysis which would require 2⁴ simulations. It would be sufficient to mention that the present study does not differentiate between the pure contributions triggered by a new parameterization and the possible synergies emerging from combining new parameterizations.

Finally, I suggest skipping trivial common places like the very last paragraph (lines 630 – 634). It is completely true that dynamical vegetation is an important factor which should be considered in ESMs. But this study is not the first one to point to the importance of dynamic vegetation. We (including the authors) have convincingly addressed this topic by numerous studies over the last roughly 30 years.  

Minor comments:

Line 116 and following: It would be useful to learn something about the interaction with the C-cycle. Into which carbon pools of the plants and the soil is the carbon gain by photosynthesis fed? Or does this issue do not play any role here?

Line 242, Fig. 3: The abbreviations in the title lines (dtas, dpr) are not defined in the caption. Why not put a \Delta T_s or \Delta P_r in the title lines and in the caption?

Line 257, Fig. 4: The new parameterizations increase the simulated annual mean precipitation in WA, but still, the simulated precip amounts to only a factor of 0.4 of the reconstructed precip. How is the aggregation of data points and comparison with grid box results done? Using any area-mean? (Would be sensible to only consider grid boxes for which reconstructions are available.)

Line 283 ff: “It results from vegetation feedbacks amplified by synergy with ocean feedbacks …” surely, it does. But without differentiation between feedback and synergy, it remains a trivial statement and could be skipped – in contrast to the second half of the sentence which likely is the real reason and would deserve more attention.

Line 287, Fig. 5: The labels on the colorbars are partly hidden behind the colorbars. Please shift. The global maps, specifically for the differences in lai, are too small to see any details outside the tropics. Please enlarge the figures to the size of the other global maps in the other figures.

Line 304: It would be helpful to note that alpha_p is the surface albedo. Commonly, one would symbolize the planetary albedo with the subscript ‘p’.

Eq.(3) and other places in the text: Sometimes the subscripts appear as subscripts, sometimes as an extension of the variable, for example as in SWsi  vs. SW_{si}  or LWsup vs. LW_{sup}. Please harmonize.

Line 318: gases instead of gazes

Line 348: What are pft 7 and 8? It would help reading, if the names of the PFTs are mentioned here or in a table.

Line 454 and other places: Sahara Sahel or Sahel Sahara sounds a bit cumbersome, because the Sahara and the Sahel (region) are pretty different regions.

Line 457: …, so that the magnitude …

Line 509: “The suite of mid-Holocene  … allow us to dig into the complexity of the Earth’s climate system.” That is a rather generic and bold statement as this study just touches a small subset and very specific aspects of the global climate system.

Line 510: “We insist on the fact …” I do not understand, why you have to ‘insist’ on the fact, instead of highlighting the fact.  

Line 519 ff: I do not quite understand the meaning of this sentence. Perhaps it is just the wording ‘associate to’ … The word ‘fulfil’ should be ‘fulfill’.

Line 527: “We show that dynamical vegetation reveals how ….” I am not sure how dynamical vegetation can reveal anything. The analysis of the climate-vegetation interaction can certainly do, but, again, only with respect to the processes considered, not with respect of the entire complexity of all biospheric processes.

Line 537: Which “step changes between the model version … is (shouldn’t is ‘are’) different from …?

Line 542: Which “model content” … lead(s) to different vegetation cover …?

Line 562 ff. Indeed, the statement that “simulated vegetation is an integrator …” is “trivial”. Perhaps a more modest statement would be sensible. This study is not the very first one to highlight the importance of vegetation dynamics.

Line 565/567. I agree that one cannot infer vegetation feedbacks from studies in which vegetation patterns are kept fixed. In this sense, the titles of early studies (e.g. Kutzbach et al. Nature 1996) are misleading. These studies analyzed impacts rather than feedbacks.

Line 596: This would require (instead of requires)

Line 600: … because land use (not land used) is not

Recommendation: 

I recommend major revisions.

Main comments

1) The main finding is that vegetation feedbacks are incredibly important and that they vary seasonally and by location. These have been quantified for 4 different model configurations. From this I’m left wondering whether this is a way to discern more clearly between the model versions. Even if this is not done here, it could be discussed what observations from either present-day or from the mid-Holocene would be required.

2) The Discussion and Conclusion section lacks focus and covers perhaps too many topics. I think this paper would have much greater impact if these two aspects could be separated and a more concise and clear Conclusions section were to be developed.

3) Editing for grammar, typos and figure presentation is needed. 

Minor corrections:

Title: I find the word “unavoidable” slightly misleading here. It has connotations of committed climate change etc. I recommend rewording throughout with something like robust or parameterisation-independent.

Line 132: This is very similar to the changes to soil moistures stress in transient Holocene simulations by Hopcroft & Valdes, 2021 PNAS.

Figure 4: The cyan (data) points are not easy to see. Could you redraw using thicker lines for the data points?

Figure 7: for clarity could you consider creating a single colour bar for all panels and labelling it with PFT groupings instead of numbers.

Figure 8: similar comment as above - label the y-axis with the PFT names not numbers.

Line 260: “We synthesize the mid-Holocene differences with preindustrial by showing the mean root mean square difference between the two climates in Fig. 5 for leaf area index (lai), snow, and atmospheric water content.”

It’s not clear why this choice is made at this point. It will compress everything to be a positive anomaly which is reducing the information. Is this intended?

Lune 278: “all of these model versions produce a green Sahara”

I’m not sure I agree with this. The precipitation anomaly shown in figure 4 is too small and the LAI anomaly is only covering half of the Sahara?

Line 284: “and from atmospheric physics and land surface improvement between the IPSLCM5 and IPSLCM6 versions of the IPSL model (Boucher et al., 2020; Hourdin et al., 2020).”

I’m not sure this is very well supported. can you either explain in more detail ?

Lines 333-336: “The snow albedo effect is amplified

334 when grass is replaced by forest in the mid-Holocene simulation, which occurs over a large area in Eurasia for V2

335 and V3 compared to V1 where grass is dominant or V4 where a larger fraction of forest is still present in the

336 preindustrial simulation (Fig. 7).”

Should this be the other way around or could you clarify? Grass being replaced with trees would result in lower albedo  overall because trees are lower albedo than grasses and trees cannot be covered as efficiently by snow as can grass?

Line 362: “It appears to be a critical model aspect contributing to a better representation of boreal forest.”

Again I’m not sure I agree as the difference in the boreal forest pft 7 seems very small between V1 and V2.

Technical corrections

Overall there are a lot of minor typos, grammatical errors and cropped edges of figures. Some of these are included below.

Line 9: “with the IPSL climate models for which dynamic vegetation is switch on.”

This should be: switched on

Lines 16-17: “which are needed to fulfill the global energy conservation constraint of the climate system.”

I don’t really understand what this means in this context.

Lines 18:

control -> controls

Line 18:

“Photosynthesis parameterization ..” 

should be

“The photosynthesis parameterization …”

Line 25:

“The Green Sahara”

Line 41: “The increase *in the* number of “

Line 42: “has emphasize”  -> “has emphasized”

Line 129-131: “This adjustment in the bare soil evaporation parameterization was not incorporated into IPSLCM6A-LR due to the fact that it induces a surface warming that was not fully understood to be used in the whole suite of CMIP6 simulations (Cheruy et al., 2020).”

This is a grammatical error in this sentence.

Line 896: “the vcmax curves are plotted toe a mean temperature”

typo

Line 152-155: “Another important difference is that in PhotoCM6, the response to temperature is adapted to the local long term (i.e. 10 years) temperature of each pixel whereas in PhotoCM6, the temperature dependence is fixed for the whole pft.”

This does not make sense to me.

Line 198: “It guaranties the entire consistency between the simulated climate and the simulated vegetation.”

This doesn’t really make sense to me.

Line 217: “A conclusion from Fig. 1 is that 300 years of”

this should be figure 2.

Line 244: |standard IPSL model without dynamical vegetation”

which model configuration is that  - state here please.

Line 302: “atmospheric diffusion”

do you mean scattering and absorption?

Line 296: “Positive values (negative) indicate that the feedback brings more (less) energy to the climate system in V4”

These double meaning sentences in brackets are in my opinion extremely hard to read and should be avoided. e.g. https://doi.org/10.1029/2010EO450004

Line 510-511: “We insist on the fact that climate-vegetation interactions induce seasonal feed backs that trigger unavoidable first order albedo and water vapor radiative feedbacks”

This use of “insist” and “unavoidable” comes across a little odd. Could you clarify e.g. “We find” instead of we insist, and instead of unavoidable use a word like robust or parameterisation-independent?

Lines 517-521: “The LW radiative feedback is less discussed when the role of vegetation is inferred from vegetation alone simulations or simulations where the sea surface temperature and sea-ice cover are prescribed. It is a first order effect associate to the change in temperature and fulfil the convective radiative equilibrium which serves as a basis for the reasoning on climate sensitivity  (Dufresne and Bony, 2008; Manabe and Wetherald, 1975; Sherwood et al., 2020)”

This isn’t clear at all.

Figure 4: it’s really not clear which circle is what in this figure. Please improve the legend.

Line 934” Not that” should be “Note that”

FIgure 9: make panel titles in English not in model variable codenames please. e.g. total soil moisture instead of mrso.

Figure 13: consider connecting the same-coloured dots with lines for clarity?

Figure 14: this and other figures have edges of the figure cropped.