In this manuscript, Xie et al. run an offline mineral dust emission model forced by a significant suite of climate model simulations that look at the broad stripes of climate change across the Phanerozoic. Based on the results of these simulations, it is argued that the principal driver of dust emission across the Phanerozoic is the degree of continentality, i.e., times when there are big continents across the subtropics. Greenhouse gas levels are of minimal import. The manuscript compares its results with the distribution of evaporite records and finds some agreement (“fair” used in the abstract may be overly generous), which is considered sufficient validation of the model. I will note that the results about the broad movements in Phanerozoic climate align well with the results of Hu et al. (2023, https://doi.org/10.1038/s41561-023-01288-y) on monsoon strength, possibly because the underlying model experiments use similar approaches.

The work presented in this manuscript is a valuable contribution to our understanding of the geological record of dust by providing a sort of null hypothesis testable against that record: that dust emission (and the level of global deposition) are driven mainly by variability in hydroclimate (and factors directly related to hydroclimate such as vegetation cover).

The experiments presented in the manuscript are suitable to provide this null hypothesis. The manuscript is clearly written and well-organised. However, the manuscript takes little account of research on the geological record of dust, either in the recent or deep past. The outcome is that evaporite distribution is used as a proxy for dust emission. However, this type of validation is circular. Evaporite deposition is, of course, more demonstrably controlled by hydroclimate and continentality. So, of course, a model that has a response strongly controlled by hydroclimate and continentality is going to match the distribution of evaporites quite well.

The alternate hypothesis that these simulations should allow to be tested once further work is done to characterise the geological record of dust through time is that dust emission through time is a balance between two, sometimes competing factors.

First, dust is mostly produced by physical weathering in glacial margin environments rather than “just there” or forming from the collision of desert sands. Second, dust is most easily emitted in dry, but not too dry environments, where vegetation cover is minimal but crusting together of small particles by evaporite salts is also minimal or reduced by occasional re-wetting.

The upshot of this hypothesis is that dust emission increases gradually in icehouse climates as sediment availability increases, peaks in the early part of greenhouse climates (or in states of higher global continentality) following an icehouse climate, and then wanes through the course of long eras of greenhouse climate, as the dust (technically silt-sized sediment) reaches long-term sediment sinks, like the ocean.

Because glacial grinding is a contested process, it is therefore very helpful to have a modelling study that removes the sediment from the equation and looking at change across the Phanerozoic. But it would be best if the manuscript gave the reader better information about the alternate hypothesis and the observables in the geological record.

I therefore recommend that this manuscript be published after revision to better review the deep time geological record of dust and consider the inconsistencies the modelling approach in the manuscript might have with even the relatively recent record of dust.

I will focus first on how the modelling in this manuscript looks in the light of geological evidence from the Holocene and Pleistocene, provide some useful background on ongoing work in reconstructing the deep time record of dust, and then provide some minor comments on the manuscript line by line.

The model in this manuscript is one of atmospheric dust emission. By tuning to the AMIP models, the manuscript is strictly speaking only treating the dust sizes (0.1-10 μm) typically represented in the AMIP models on the grounds of their relevance to atmospheric radiation/clouds and presence in dust deposition records distal from emission (such as ice cores). See Mahowald et al.  (2006 , https://doi.org/10.1029/2005JD006653) or Mahowald et al. (2014, https://doi.org/10.1016/j.aeolia.2013.09.002) for further discussion of these points.

However, the best proxies for the geographic distribution of dust emission are proximal deposits of dust, where particle sizes are bigger (and much of the mass is concentrated geologically). These are known as loesses and can be 10s or sometimes 100s of m thick. Loesses typically have particle sizes ~50 μm. Thus, if the approximate atmospheric dry deposition lifetime of 5-10 μm is 1.7 days (Mahowald et al., 2006), the lifetime of loess particles is ~ 1 hour. Remobilisation complicates the picture, and the nature of sedimentary geology is for sinks to be few at any one time. But the upshot is that the geological record of dust emission is strongly concentrated within the bounds of wherever the wind can blow dust in a few hours (~ 100 km).

So when faced with a figure showing the world’s loess deposition in recent sediments, any model that connects dust emission to hydroclimate etc. rather than sediment availability is faced with a problem. (See Fig. 2 of https://doi.org/10.1016/j.earscirev.2019.102947) for a good example. Loess is abundant in mid-latitude and some high-latitude areas across North America and Eurasia, South America and parts of Australasia that include New Zealand. And it is relative rare along the southern margin of the Sahara. The model presented in the manuscript might explain the rarity of Saharan loess for the most recent time slice. (But where are the Pliocene loesses in the Sahara contemporary with the Eurasian ones?) But North American loess, European loess, and New Zealand loesses are inexplicable. This observation is the motivation for the idea that dust emission is shaped by sediment availability, with sediment primarily being generated by physical weathering at glacial margins (see discussion in Mahowald et al., 2006 and Prospero et al., 2002, https://doi.org/10.1016/j.earscirev.2019.102947).

The manuscript certainly demonstrates limited knowledge of dust records through geological time. At line 49, it says, “similar orbital-driven dust (or aridity) records are also identified much further back in time, as early as the Late Cretaceous 50 (Niedermeyer et al., 2010; Vallé et al., 2017; Zhang et al., 2019).” In fact, the most widespread evidence for orbitally-driven dust deposition is from the Late Paleozoic. A good place to start would be a recent review by Soreghan et al. (https://doi.org/10.1144/SP535-2022-208) and probably too recent for the authors to see (https://doi.org/10.2110/jsr.2023.122).  The work by Lynn Soreghan suggests loess deposition was widespread, even at low latitudes in late Carboniferous and early Permian Pangaea. Loess also seems widespread in much of the Permian (where we can see), not just in places Soreghan and her collaborators have looked but in Russia (Mouraviev et al., 2015, doi:  https://doi.org/10.1134/S0031030115110064). Loess deposition certainly continues deeper into the long greenhouse interval (see Chan et al., 1999, https://doi.org/10.2110/jsr.69.477),  but the Triassic peak in dust emission by the model seems inconsistent with the record, where the places we would expect to see loesses are mostly sandstones and evaporites.

I still do agree with the general thrust of the manuscript that the deep time record of dust is poorly understood, but this should not be done without emphasising the importance of loess as a proxy for dust emission, some more expanded references to work in this area, and more nuance when comparing the model results with evaporite distributions.

Minor Comments

Line 47: This is probably where sediment availability and glacial grinding should appear as a hypothesis. Additional possible references are: Reader et al. 1999 (https://doi.org/10.1029/1999JD900033) and Sugden et al. (2009,  https://doi.org/10.1038/ngeo474)

Line 100: Omitting the Zender et al. formulation is justifiable if you are clear that you are trying to help test between hydroclimate/meteorology and sediment supply hypotheses.

Figure 1: The notation of units in the legend/colorbar is confusing. Normally, you don’t superscript the exponential notation, so there’s a clear distinction between e-09=10^-9 vs. e^-9

Line 339: Desert has become dessert

Fig. A4 Part 3: 213 Ma looks like a plotting error. Please check.

Xie et al. develop and tune a new offline dust emission model, DUSTY, based on which and the climate simulations they have carried out previously, a continuous model-derived timeseries of global dust emissions over the whole Phanerozoic is established. The simulation results provide quantitative insights into significant fluctuations in dust emissions during this Era. Notably, the study highlights the dominant influence of non-vegetated areas on dust emissions. The authors further investigate the mechanisms underlying the hydrological variations, governs the aridity the distribution non-vegetated areas. The findings demonstrate that paleogeography serves as the primary driving force behind dust emission variations, while the role of CO₂ is found to be marginal. The comparison between the simulated region of dust emission and the distribution of evaporite sediment records is not bad given the uncertainties in so many boundary conditions and model parameters. The manuscript provides new knowledge to both geologists and paleoclimatologists and clearly written in general, I recommend publication after a relatively minor revision.

Major comments

1. More detailed comparison should be made here of the 15_MMM and 0Ma simulations, which are crucial for tuning and evaluating the model. Although the model parameters of the DUSTY model have been optimized by maximizing the Arcsin Mielke score and the global mean emissions are the same, significant differences still exist between Figures 1a and 1b. For example, the dust emission in the middle of Eurasia and South America is seriously underestimated while that over Australia is highly overestimated. Are they due to biases in the simulated climate, vegetation or dust parameterization? The implication to the simulated distribution of dust emission in other periods should also be discussed.
2. The absence of land plant colonisation before 410Ma implies that the inclusion of vegetation cover during 541-410Ma in this study may have resulted in an underestimation of dust emission during this period. The influence of this effect on the major conclusions of their study should also be discussed.

 Specific Comments

Line 45-48: May add "glaciogenic rock powder" or other equivalent expressions.

Line 188-190: is the global mean the average over all the land or over the regions with nonzero dust emission?

Line 189: 'C1' should be changed to 'C2'. This correction is also required in Table 1.

Line 230: It should be added here that the results are from the S2 series of experiments.

Line 282-285: How was the value of 0.8 mm/day chosen here? The bracket at the end of Equation 10 needs to be removed.