This paper summarizes an important work on improving the dust emission representation in CESM2. Overall the paper is well written and presents sufficient analysis of the dust model performance. I have some comments and questions, described as below.

Given the complexity of the model, consider adding a table to list all the mathematical symbols and abbreviations defined in the paper, and/or a flowchart of the model components.

Section 3.3:

Clarify whether Feff in Eq. 7 represents the fraction of wind drag available for dust lifting or consumed by non-erodible materials, to avoid confusion of the readers.

Section 3.4:

What is the physical rationale of representing the intermittent dust emission simply as a scaling factor on the saltation-driven dust emission? Turbulent/convective and saltation-driven dust emissions are two separate physical processes, and have their own separate forms of model parameterizations.

Also, I understand the authors' goal of incorporating as many physical processes as possible, which however may not always improve the model. I am curious whether the importance of dust emission intermittency is evaluated, e.g., via sensitivity analysis? Fig. S2 shows the intermittency factor is much like a global erodible fraction map with high values over subtropical regions, and low values elsewhere.

Section 3.5:

The purpose is upscaling correction map (Kc) is explained, but confusing. If the authors intend to capture the finer-scale variability in wind and dust emission, why not just increase the model resolution? I assume computational resources is not a limiting factor since it's not mentioned in the paper.

A standalone experiment is performed to calculate Kc. What's the time period of the standalone experiment? Does Kc vary significantly in time, e.g., between seasons and from year to year? If yes, explain why applying a constant Kc is acceptable for the multi-year simulation.

Section 5.2

Since MIDAS relies on MERRA2 for deriving DAOD, it's subject to MERRA2 aerosol model biases in addition to MODIS sensor/algorithm errors. What's the implication for the comparison with CESM2 model performance?

Section 5.3 Line 785, any proof that the DAOD over Taklamakan has mixed signals from Karakum/Kyzylkum? Those are two desert regions separated by high mountains - there is a very small chance the dust from either one is transported to the other.

How does the dust emission amount affect the temporal (daily) correlation of DAOD? Could meteorology be important?

I find the authors' explanation of the model underperformance over China speculative at best.