the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Jun 2026
Impacts of mesh refinement on the simulation of a long-range transported extreme dust storm over East Asia in the global variable-resolution model (iAMAS v2.6.3)
Abstract. Long-range transported dust storms challenge regional high-resolution models because lateral boundary conditions constrain the consistent representation of dust emission, transport, and deposition. Here we use the integrated Atmospheric Model Across Scales (iAMAS v2.6.3), a global variable-resolution physics-chemistry coupled model, to simulate the extreme East Asian dust storm from 13 to 18 March 2021. Three experiments adopt a globally quasi-uniform 50 km mesh (U50 km) and two source-refined meshes at 16–50 km (V16 km) and 4–50 km (V4 km). Evaluated against reanalysis and ground-based observations, the simulations reproduce the large-scale synoptic evolution, while mesh refinement captures more detailed near-surface dynamical features and terrain-channeled winds through improved topographic representation. The model reasonably represents 10-m wind speed, surface PM10 concentrations, and aerosol optical depth. Relative to U50 km, total dust emissions over East Asia increase by 35.24 % in V16 km and 54.98 % in V4 km, as refined meshes resolve localized friction-velocity enhancement and nonlinear saltation-threshold exceedances. By altering emission evolution and its interaction with atmospheric circulation, mesh refinement also affects the spatial distribution of transported dust. The contrast between V16 km and V4 km reveals regional heterogeneity in downwind dust mass loading, with V4 km capturing localized wet-scavenging enhancement not evident in V16 km and reducing dust loading over the Yangtze River Delta by 30.49 % relative to U50 km. This research emphasizes the importance of mesh refinement for representing long-range transported dust storms and provides a basis for future evaluations of multiple dust emission schemes in global variable-resolution models.
- Preprint
(4990 KB) - Metadata XML
-
Supplement
(8989 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3149', Anonymous Referee #1, 23 Jul 2026
-
RC2: 'Comment on egusphere-2026-3149', Anonymous Referee #2, 10 Aug 2026
This manuscript investigates the impacts of mesh refinement on the simulation of an extreme East Asian dust storm using a global variable-resolution model. The study is interesting and provides useful insights into how increasing resolution affects near-surface winds, dust emission, and long-range transport. I have several comments that may help improve the manuscript.
Major comments
1. Contributions of dust-emitting area and emission intensity
The authors have already discussed the changes in saltation occurrence frequency and emission intensity, which helps explain the enhanced dust emission at finer resolution. However, it would be useful to further quantify the contributions from active dust-emitting area and emission intensity per unit area, and directly compare these quantities among U50 km, V16 km, and V4 km. This would provide a clearer explanation for the substantial increase in total dust emission with mesh refinement.
2. PDF of friction velocity
The manuscript suggests that the increase in dust emission is mainly associated with stronger extreme friction velocity rather than more frequent saltation. I suggest providing the probability density distributions (PDFs) of friction velocity for the three resolutions, with the threshold friction velocity indicated, to clearly show how mesh refinement modifies the high-u∗ tail.
3. Does the resolution effect converge?
Dust emission continuously increases from U50 km to V16 km and V4 km. This raises an important question: would dust emission continue to increase if the model resolution were further refined? The authors do not necessarily need to conduct additional simulations, but the possible resolution limit or convergence of dust emission should be discussed.
4. Reduced dust loading around the Qaidam Basin
Figure 12 shows decreased dust mass loading around the Qaidam Basin in the refined simulations. The reason for this regional decrease is unclear. Please clarify whether it is related to changes in local emission, terrain-modulated transport, or deposition/removal processes.
Minor comments
1. Figure 4
Figure 4 contains too many wind-field panels, making detailed comparison difficult. I suggest selecting one or two representative days and enlarging the panels so that the differences among observations and the three simulations can be more clearly examined.
2. Figure 5
The current design makes it difficult to distinguish station observations from the background PM10 shading, particularly over regions with high concentrations. Please redesign the figure using clearer station symbols, outlines, or another approach to improve the distinction between observations and simulations.Citation: https://doi.org/10.5194/egusphere-2026-3149-RC2 -
RC3: 'Comment on egusphere-2026-3149', Anonymous Referee #3, 21 Aug 2026
General Comments
This manuscript investigates the impacts of mesh refinement on the simulation of an extreme East Asian dust storm using the variable-resolution iAMAS model. While the topic is relevant and the study presents some interesting results, I have some concerns about the current manuscript. Important aspects of the model implementation and experimental design are not sufficiently documented, and the mechanistic interpretation is not adequately supported by the analyses presented. In several places, diagnostic or statistical relationships are interpreted as physical mechanisms without clear process-level attribution. Moreover, the conclusions are largely drawn from a single event, raising concerns about their robustness and general applicability.
In its current form, I do not think the manuscript provides sufficient evidence to support several of its main conclusions. Substantial revisions are required before the manuscript can be considered for publication. In particular, the authors need to clearly address the methodological uncertainties, strengthen the mechanistic attribution, and better establish the robustness and novelty of the results. The major issues listed below should be fully addressed in a revised manuscript.
Major Comments
- The resolution dependence of dust emissions and its novelty need to be better addressed (Sects. 2.2 and 3.2.2): The nonlinear and threshold dependence of dust emission on surface wind speed, and the resulting sensitivity of dust emissions to model resolution, have been well documented in previous studies (e.g., Ridley et al., 2013; Meng et al., 2021). In particular, Ridley et al. (2013) demonstrated strong resolution dependence of dust emissions in GEOS-Chem due to the nonlinear response of dust emission to resolved surface winds, while Meng et al. (2021) developed grid-independent high-resolution dust emissions to reduce this artificial resolution dependence. These studies are directly relevant to the central issue of this manuscript but are not discussed. The authors should compare the magnitude and behavior of the resolution sensitivity found here with previous studies and clarify what new physical insight is provided by the variable-resolution framework, rather than presenting the resolution dependence itself as a new mechanism.
- The opposite resolution response compared with Feng et al. (2023) deserves careful explanation (Introduction and Sect. 3.2.2): Feng et al. (2023), using a closely related global variable-resolution modeling framework, reported decreased dust emissions over East Asia with mesh refinement, attributed to reduce near-surface winds associated with better-resolved complex terrain. In contrast, the present study finds a substantial increase in dust emissions with increasing resolution and attributes this to stronger local wind/friction-velocity extremes. Why does mesh refinement produce opposite responses in these two studies? Is the difference related to synoptic conditions, season, source regions, resolution range, or other aspects of the model configuration? This apparent contradiction should be explicitly discussed and may provide important insight into when and why mesh refinement increases or decreases dust emissions.
- S04 Scheme, land surface properties, and dust source function details: Since this study focuses primarily on dust simulation, a more detailed explanation of the S04 dust emission scheme and its key differences from S11 should be provided in the text. Furthermore, while the physical emission processes are described, the manuscript should elaborate on how land surface properties (e.g., soil moisture, vegetation cover, snow cover) modulate dust emissions. Lastly, please clarify the definition of the "dust source function" used in the model. Does it specifically represent soil erodibility?
- The aerosol spin-up period appears too short (Sect. 2.4, around Lines 304–315): The simulation starts on 8 March, providing only about four days of aerosol spin-up before the meteorological reinitialization on 12 March. Four days may be insufficient, particularly for the smaller dust size bins with relatively long atmospheric lifetimes. The authors should either use a longer spin-up period (e.g., 10 days to two weeks), provide a sensitivity test, or provide sufficient evidence demonstrating that a four-day spin-up is adequate for the aerosol fields considered in this study.
- The meteorological reinitialization procedure needs further clarification and justification (Sect. 2.4, around Lines 308–315): The meteorological fields are reinitialized on 12 March while the chemical tracers are retained. The authors explain that this approach is intended to reduce meteorological drift while maintaining continuity of the chemical fields. However, retaining the aerosol fields while replacing the meteorological state may introduce inconsistency between the two. The authors should clarify how the reinitialization is technically implemented and, more importantly, demonstrate or provide evidence that this procedure does not introduce significant discontinuities or affect the subsequent dust emission, transport, and deposition during the analyzed event.
- Aerosol–meteorology interactions need to be clarified and separated from the resolution effect (Sects. 2.4 and 3.2.4): It is unclear whether the direct and indirect effects of aerosols, particularly dust, are activated in these experiments. This needs to be explicitly stated. If aerosol–radiation and/or aerosol–cloud interactions are included, the interpretation becomes considerably more complicated because mesh refinement changes dust emissions and distributions, which can in turn modify radiation, clouds, precipitation, and other meteorological fields. The meteorological differences among U50, V16, and V4 may therefore reflect both (1) the direct effect of mesh refinement on meteorology and (2) aerosol–meteorology feedbacks induced by resolution-dependent dust emissions and distributions. These contributions need to be distinguished. If aerosol feedbacks are active, sensitivity experiments with these feedbacks turned off, or other quantitative analyses, would be needed to determine their relative contributions.
- The mechanistic attribution remains incomplete (Sect. 3.2): The diagnostic analyses of terrain, surface winds, friction velocity, dust emissions, transport, and wet scavenging are useful, but they do not fully isolate the contributions of individual processes. Since these processes change simultaneously with mesh refinement, their concurrent changes alone are not sufficient to establish causality or quantify their relative contributions. Additional process-level analyses or sensitivity experiments would strengthen the mechanistic attribution. For example, using the same dust emissions across different-resolution simulations could help separate emission-related effects from those of transport and removal. A quantitative dust budget would also help distinguish the contributions of emission, transport, dry deposition, and wet deposition. If such analyses are not feasible, the mechanistic conclusions should be appropriately qualified.
- The conclusions rely on a single dust event and their robustness needs to be demonstrated (Sects. 3.2 and 4): The proposed mechanisms are derived entirely from the March 2021 event. However, the response of surface winds and dust emissions to mesh refinement may depend strongly on synoptic conditions, source regions, season, and background meteorology. This concern is particularly relevant given the opposite resolution response reported by Feng et al. (2023). I strongly recommend adding several dust events under different meteorological conditions to test whether the proposed mechanisms are robust. If additional cases are not feasible, the conclusions should be substantially narrowed and explicitly presented as findings specific to this event.
- The observational constraints remain limited for a single-case study (Sects. 2.5 and 3.1): The authors acknowledge the limited AERONET coverage (23 sites). However, the evaluation relies largely on event-mean AOD, which provides limited information on the timing and evolution of the dust plume. PM10 is also not a direct measurement of dust. Given that the conclusions are based on a single event, stronger observational constraints would be helpful. The authors should consider evaluating the temporal evolution of the event and, if available, include satellite observations or measurements of the vertical dust distribution to better constrain the simulated dust transport. As highlighted in the Introduction, a key strength of the iAMAS global variable-resolution model is its seamless global-to-regional multi-scale capability. However, the current manuscript focuses mainly on source and near-source regions (e.g., Gobi and Taklimakan). To fully evaluate the model's capability in capturing long-range transboundary transport, additional validation over downwind remote regions should be included. Incorporating available surface PM10 networks and AERONET AOD data in South Korea and Japan would significantly strengthen the manuscript.
- Reconciling PM10 and AOD validations across mesh resolutions: There is an apparent discrepancy in model performance between surface PM10 and column AOD. Why does U50 km show the lowest AOD values while significantly overestimating surface PM10? Conversely, V4 km shows good agreement for both metrics. Did the different horizontal resolutions affect the vertical mixing or size distribution of dust particles? Please provide a more detailed explanation connecting the spatial plume structure, vertical distribution, and AOD/ PM10 mechanics. Additionally, if spatial plume distribution strongly drives AOD accuracy, AERONET sites nearer to the source regions should be included in the analysis if available.
- Improved agreement with PM10 and AOD does not necessarily imply more realistic dust emissions (Sects. 3.1.3–3.1.4 and 3.2.1): The manuscript should avoid implying that improved PM10/AOD agreement validates the larger dust emission flux. Better agreement of V4 with observed PM10 and AOD does not directly demonstrate that the larger dust emission flux is more realistic, because PM10 and AOD are also controlled by transport, vertical mixing, and dry and wet deposition. The authors should therefore avoid using improved PM10/AOD performance as direct validation of the increased dust emission flux. The discussion should clearly distinguish among the resolution sensitivity of dust emissions, improvement in simulated atmospheric dust concentrations/loading, and accuracy of the emission flux itself.
- The attribution of the YRD dust reduction to enhanced precipitation and wet scavenging needs stronger evidence (Sect. 3.2.4): The manuscript attributes the reduced dust loading over the YRD in V4 primarily to enhanced precipitation and wet scavenging. However, dust emissions, dynamical transport, PBL mixing, convection, gravitational settling, and dry deposition all change simultaneously among the experiments. These processes are also closely linked to the meteorological fields, which themselves are affected by the change in model resolution. Therefore, the concurrent increase in wet scavenging and decrease in dust loading does not, by itself, establish a causal relationship. A quantitative regional dust budget separating emission, transport convergence/divergence, dry deposition, and wet deposition would provide much stronger support for this conclusion. The authors should quantify the relative contributions of these processes to the dust-loading differences over the YRD rather than inferring the dominant mechanism mainly from their spatial and temporal correspondence.
Minor Comments
- Section 2.4: Please provide the total number of grid cells and, if available, the computational cost for U50, V16, and V4.
- Section 2.5 and 3.1: Please clarify how the surface wind, PM10, and AERONET observations are collocated with the model grids, particularly given the different grid spacing among the three experiments.
- Section 3.2.2: Please clarify the temporal resolution used to calculate (u_*), (u_{*t}), the threshold-exceedance frequency, and (S_3).
- Section 2.5 and 3.1.3: PM10 is used as a proxy for dust. Please briefly discuss the possible contribution of non-dust aerosols to the observed PM10 during the event.
- Section 3.1: Please report the number of observations used in each statistical comparison, particularly for the regional PM10 and surface-wind evaluations.
- Lines 60–62: The sentence describing regional model limitations is oversimplified. Please expand it to provide a more precise description, such as highlighting the "limited representation of lateral boundary conditions in regional models."
- Lines 76–78: The description of regional model constraints is too simplified. Regional models are limited in dust simulations not only due to their strong dependency on global driver models, but also because running large domains at high resolution to cover long-range transport pathways incurs substantial computational costs.
- Lines 121–123: Please provide a clearer rationale for selecting the March 2021 event. How severe was this dust storm compared to other major East Asian dust events or observation campaign periods?
- Lines 424–450: Have you evaluated the model against hourly surface wind speed observations? In high and complex terrain, comparing daily mean wind speed is insufficient because the model does not use daily winds dynamically, and sub-daily peak winds drive threshold exceedance. How do the normalized mean biases (NMBs) in wind speed quantitatively impact dust emission amounts across the different experiments? Providing a quantitative assessment of emission variability relative to hourly wind speed biases would help readers better understand the underlying sensitivity.
- Please use terms such as “improvement” with caution, as the response to increasing resolution is not always monotonic across U50, V16, and V4.
Citation: https://doi.org/10.5194/egusphere-2026-3149-RC3
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 68 | 35 | 16 | 119 | 22 | 11 | 8 |
- HTML: 68
- PDF: 35
- XML: 16
- Total: 119
- Supplement: 22
- BibTeX: 11
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
In this study, titled “Impacts of mesh refinement on the simulation of a long-range transported extreme dust storm over East Asia in the global variable-resolution model (iAMAS v2.6.3)” by Xue et al., the authors conducted three experiments, with the highest model resolution increased from 50 km to 16 km to 4 km, to simulate an extreme dust storm event in East Asia. The authors analyzed the impact of model resolution on simulated dust aerosol properties and processes.
1.In section 2.2, Dust emission scheme, the authors could highlight the factors affected by model resolution.
2.Line 179-180: Add more information about the dust source function (e.g., definition, range, physical meaning) and model resolution in Figure 1. Same for Figure 1 title. Do variable-resolution models use the same dust source function, with 50 km resolution?
3.Line 304-306: Since the resolution of ERA5 data is 0.25-degree and lower than that of variable-resolution models, how does the interpolation, from low to high resolution, affect the simulation results? Also, “Meteorological initial conditions and surface boundary fields are obtained from the ERA5 reanalysis at 6-hourly intervals”, does the model read initial conditions every 6 hour?
4.Figure 2: The grid lines and differences in grid sizes in (a)-(c) are not clear.
5.When evaluating the simulated large-scale circulation and wind speed, the authors could quantify the model performance using spatial correlation coefficients, scatterplot, root mean square error, etc. This can provide evidence for the descriptions such as “generally well reproduced”, “reasonably represented”, “comparable spatial pattern”, “broadly consistent”, (Line 365-370). For 10-m wind speed, besides temporal correlation coefficients in Table 2, the scatterplot (like Figure 6) could be used to show how well the models capture the extreme wind speed in dust source regions.
6.Line 367: For mean sea level pressure in TKD region, ERA5 shows a low pressure and wind convergence, but models do not capture this, especially during March 15-16. Since TKD is a key source of dust, this bias may affect dust transport. Are the model results regridded to ERA5 grids? If so, the authors could show the differences between simulations and ERA5 in supplementary information.
7.Line 381-382: Figure 3 does not show the information about “terrain transitions”, and the audiences might not be familiar with the definitions of Hexi Corridor/Mongolian Plateau. The authors could overlay the winds on topography and highlight the regions with significant differences in winds if possible.
8.Line 463: What do the authors mean by “more spatially diffuse”? The dust in U50 goes farther from source to downwind regions? On March 16 (Figure 5), over the downwind Shandong Peninsula, V16 and V4 show larger concentrations (>500 ug/m3) than U50 does (200-500 ug/m3). This suggests that dust in V16 and V4 goes farther. Where are “elevated background concentrations” (Line 465-466) located at? For the “positive mean bias presented in Fig. 6” (Line 466-467), most of the observed concentrations are below 400 ug/m3 while simulated concentrations are 200-1300 ug/m3, so, I guess these points are associated with stations in northwestern regions in Figure 6 where all three models show higher bias in concentrations.
9.Line 489-492: How about the effects of dry deposition near dust source regions on dust concentrations/transport?
10.Section 3.1.4: The authors could compare the simulated spatial distribution of AOD with satellite observations/reanalysis data. Does the model output include dust optical depth? If so, the authors could evaluate it using observation/reanalysis data such as MERRA2. What other aerosol species are considered in the model? The biases in anthropogenic aerosols could affect the model performance in simulating AOD in the downwind regions of dust.
11.Line 583-585: What does “the strongest” mean? It refers to magnitude or spatial range? What do the authors mean by “consistent enhancement”?
12.Line 643-645: Both the model resolution and topography can affect the low-level circulation. Therefore, to separate these two factors, the authors may need to use higher-resolution model with coarse-resolution topography. In TKD (40N, 90E), the topography differences are small (<10m, Figure 9 d, e), but the differences in wind friction field are relatively large (-0.2~-0.3, Figure 10 d, e).
13.Figure 12: With mesh refinement, dust burden decreases in the region with large topography gradient (30-40N, 105E). This may be related to higher topography blocking dust transport. In northeastern China (40-45N, 110-130E), V16km shows decreases in dust burden, but V4km shows increase. Which model has model performance regarding this? The authors may compare the simulation results with observations/reanalysis data of dust burden or optical depth.
14.Figure 13: The authors may change the range of colorbar to show values below 10 mg/m2/day, because the current figure shows precipitation decrease (Figure 13c, blue color) but wet scavenging does not (Figure 13f, white color) as the changes are less than 10 mg/m2/day. Changing colorbar could be more consistent with Table 5 which shows wet scavenging in NCP decreases (from 6.733 to 1.109 mg/m2/day) and support the conclusion “precipitation differences … modulate dust removal” (Line 755-757).
15.Line 764-767: What do the authors mean by “more consistent”? How to quantify it? Do you mean the contribution from emission changes to dust burden changes is larger than that from wet removal changes? Also, in these source regions, what are the changes in dust dry deposition as the model resolution increases?
16.Line 776: Besides “precipitation structure”, how about the effects of model resolution on clouds? How does the in-cloud and below-cloud wet scavenging change as model resolution increases?