The authors present a manuscript describing changes in the dust cycle between pre-industrial and LGM climate conditions simulated by a state of the art atmospheric and aerosol model, where surface boundary conditions rely on a dynamic vegetation model that influences surface properties linked to the generation of dust emissions, also in different climates. This study is indeed a welcome contribution to the dust and paleoclimate research field, in particular providing elements to discuss how the dust cycle is influenced by climate conditions.

The manuscript is generally well written  and figures are clear, the design of the study is well conceived, and the methodology and results are overall nicely described. However, I found some aspects that should be clarified and/or improved; in particular I would recommend that the authors provide a clearer and enhanced description of specific aspect of the methodology and of the comparisons with observations.

Specific comments

30-33: More precisely, dust scatters and absorbs both SW and LW radiation, although scattering prevails in SW (still, the single scattering albedo of dust is not equal to 1, e.g. Balkanski et al., 2007) and absorption in the LW (although scattering may be important too, e.g. Dufresne et al., 2002).

63-65: “We compare present-day simulation results to model results …”. Please rephrase.

165-169: It’s not very clear to me what these regional correction factors are exactly, and how they are applied to the present study, to maximize the match with which observations and how, or what are they values. Please clarify the procedure in more detail.

176: “Since our simulation periods are comparably short”… compared to what? I do not understand this passage. I gather you use an atmosphere only model coupled to land surface scheme and consider prescribed SST for the ocean surface. Okay, so how does this sentence fit into that? Please rephrase.

213-214: This statement is essentially based on a set of global metrics compared to Huneeus et al. (2011). It is true that the dust scheme is described in more detail Stanelle et al. 2014, and there validated against a wide set of observations of other features of interest for the representation of the dust cycle; however I would expect to see some comparison here too, with the current version of the ECHAM model setup, also because it appears that some tuning was done, and I found no reference to another paper describing it. The spatial patterns of dust emissions indeed appear to show some difference with Stanelle et al. 2014, also concerning the Southern Hemisphere. Please add some more information in this respect or an appropriate reference if that exists already.  

261: Among the model factors affecting dust emissions surely there is also the vegetation cover, here simulated thanks to a dynamic vegetation model. I would suggest adding a panel showing a map of the vegetation fraction, or anyway a vegetation-related variable that closely resembles the way vegetation affects dust emissions in the model.  

283: The observational data used for figure 3 do not appear to correspond to the original DIRTMAP dataset (i.e. Figure 8 in Kohfeld and Harrison, 2001). Please make sure that you add a reference corresponding to the actual version of the dataset you used, and specify whether additional data were included.

283-315: Several data points in the Southern Ocean appear to be south of the Polar front, which should raise a flag about non-aeolian contributions to the terrigenous fraction of the sediment, and therefore the opportunity to use these data for a robust estimation of dust mass accumulation rates (e.g. Kohfeld and Harrison, 2001).

352-356: There is a substantial difference in the experimental design of Albani et al. (2012 and 2014) and this work; here it appears that the amount and proportions of dust from different sources result only from the model itself (and indirectly the regional tuning on dust emissions made on present day conditions, apparently), whereas the cited work explicitly used regional tuning also for the LGM, in a data-assimilation fashion, in order to obtain a match on dust amounts, LGM/interglacial ratio, as well as source mix based on geochemical fingerprinting on Antarctic ice core samples (e.g. Delmonte et al., 2010). In other words, one could say that the CAM3 results that you mention indicate a dominance of South American dust because ice core data suggest just that, of course under the assumption that simulated transport and deposition can be considered reasonable.

352-368: Based on my previous comment, I would recommend that a more thorough discussion is carried out considering also the available data on dust provenance. It is indeed very important that you explain your results based on the modeled processes, as you did, but I believe that they should also be put more in perspective by comparing them to observational evidence, also for this particular aspect (which by the way you mention later on while discussing the matter of size, and you also acknowledge in the conclusions).

412-414: Is there a variability on size distributions at the stage of dust emissions in your model formulation? I don’t think so, so I’m a bit confused, why would you expect that?

472-474: Where does this come from? This aspect is not shown or discussed anywhere in the text.

478-479: I would suggest adding two lines bracketing the +/- 1 order of magnitude in the scatterplots of Figure 3, for a clearer reading.

500-504: I would recommend that these considerations on the chosen boundary conditions are also reported in the methods and/or results sections, as appropriate.

466-504: I would suggest enriching a bit the conclusion section with references to the literature, where appropriate.

References

Balkanski, Y., Schulz, M., Claquin, T., & Guibert, S. (2007). Reevaluation of mineral aerosol radiative forcings suggests a better agreement with satellite and AERONET data. Atmospheric Chemistry and Physics, 7, 81–95. https://doi.org/10.5194/acp‐7‐81‐2007

Dufresne, J. L., Gautier, C., Ricchiazzi, P., & Fouquart, Y. (2002). Longwave scattering effects of mineral aerosols. Journal of the Atmospheric Sciences, 59, 1959–1966. https://doi.org/10.1175/1520‐0469(2002)059<1959:LSEOMA>2.0.CO;2

Delmonte, B., Andersson, P., Schöberg, H., Hansson, M., Petit, J. R., Delmas, R., Gaiero, D. M., Maggi, V., and Frezzotti, M.: Geographic provenance of aeolian dust in East Antarctica during Pleistocene glaciations: preliminary results from Talos Dome and comparison with East Antarctic and new Andean ice core data, Quat. Sci. Rev., 29, 256–264, doi:10.1016/j.quascirev.2009.05.010, 2010

Review of Simulating glacial dust changes in the Southern Hemisphere using ECHAM6.3-HAM2.3 by Krätschmer, et al. for Climate of the Past.

The authors of this study use the ECHAM climate model to investigate the global variability of mineral dust emission, transport, and deposition across three different mean climates. (modern, pre-industrial, and LGM), with a particular focus on the Southern Hemisphere.  The land model component includes dynamic vegetation that determines dust source locations, thereby permitting prognostic determination of changing dust source strength and location through time, and their LGM simulation includes enlarged coastlines to include exposed continental shelf among the potential dust sources.  For each of their simulations the authors compare their results to observations and other model results, sometimes finding agreement and sometimes not.  Their simulation of the LGM, and investigation of how it differs from the PI, is noteworthy as there are limited model studies of this time period that include dynamic dust sources.  Of particular note are a series of results regarding the spatial variability of the provenance of dust deposited on Antarctica and the Southern Ocean.  These results are a valuable contribution to the ongoing discussion regarding the relative contributions of South American and Australian sources through time, and thus the appropriate interpretation of archives of paleodust from the region.  As this is a topic with some disagreement, the addition of results from a new model is welcome.

Overall I found the paper very well written and organized, and enjoyable to read.  I do have some unanswered questions as well as some minor suggestions for the authors, so my recommendation is acceptance upon minor revisions.

Primary comments

1. The paper would benefit from more time in the introduction and conclusion spent reviewing what is known and the disagreements regarding dust provenance to Antarctic and the Southern Ocean. I was pleased when the authors brought up many of the studies when discussing their results, but I felt the bigger picture was somewhat overlooked.  Specifically, there are conflicting studies regarding whether South America or Australia is the primary source of dust (and for that matter how much is contributed by Antarctic sources), and the relative role of dust source strength and transport efficiency.  I think the authors have room here to set up and then answer some questions about how these discrepancies can be resolved by considering the time-varying relative strength of sources, the transport efficiency, and the spatial distribution of their influence.  I think much of this information is already contained in the paper, but an explicit consideration of the debate would be valuable.  One additional source to consider is Markle, et al. (2018).
2. What about New Zealand? I was surprised that the LGM simulations don’t seem to include an expanded dust source from the exposed continental shelf around New Zealand, nor any discussion of it as a dust source during that period.  Neff, et al. (2015) and Koffman, et al. (2021) would be relevant to this discussion.
3. The authors cover many of the modeling studies of dust transport to the Antarctic in their discussion of provenance, but there are also studies that take an isotopic approach that should be discussed as well. Wengler et al. (2019), McGee et al. (2016).

Minor comments

1. How is land tiling / vegetation coverage determined for the newly exposed continental shelf? Essentially, is the newly exposed land always a dust source, or can it become vegetated?
2. Since additional land area during the LGM is credited as one of the causes of increased dust emission, I would like some information, similar to the reported wind and precip changes, that tells me how much additional land there is in each region (and possibly some discussion of how much of the dust is being created from this new land).
3. In Figure 3, why are the simulated dust deposition values so stratified? The observations appear continuous across a couple orders of magnitude, while the simulated deposition values appear to form horizontal lines.
4. Continuous colorbars on log plots are difficult to accurately interpret. The maps in Figures 3 and 1 would be much easier to read if the colorbars had discreet steps (while keeping the log scale).  In Figure 1 it didn’t bother me to much because I was more interested in the qualitative pattern than the quantitative values, but Fig. 3e I wanted o know where the one contour was, which is quite difficult to tell.  I would suggest colorbars similar to those in Figs 2 and 5.

References

Markle, et al. (2018) Concomittant variability in high-latitude aerosols, water isotopes, and the hydrologic cycle, Nature Geoscience.

Neff, et al. (2015) Trajectory modeling of modern dust transport to the Southern Ocean and Antarctica, JGR: Atmospheres.

Koffman, et al. (2021) New Zealand as a source of mineral dust to the atmosphere and ocean, Quaternary Science Reviews.

McGee, et al. (2016) Tracking eolian dust with helium and thorium: impacts of grain size and provenance, Geochimica et Cosmochimica Acta

Wengler, et al. (2019) A geochemical approach to reconstruct modern dust fluxes and sources to the South Pacific, Geochimica et Cosmochimica Acta