the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jun 2026
Evaluation of WRF-Chem (version 4.7.1) aeolian dust emission and land surface models over the dust belt
Abstract. Aeolian dust is a key component of the Earth system, influencing biogeochemical cycles, cloud microphysics, and the radiative energy budget and atmospheric dynamics, while also degrading air quality around major source regions. The representation of mineral dust remains a major challenge for regional and global atmospheric models due to uncertainties in emission processes, land-surface interactions, together with limited availability of observations. In this study, we present the first comprehensive, year-long evaluation of the simulated dust with the WRF-Chem model (v4.7.1) over the dust belt spanning North Africa, the Middle East, and Central Asia. We evaluate an ensemble of six simulations using three widely applied dust emission schemes (GOCART, GOCART-AFWA, and University of Cologne – UoC) combined with two advanced land surface models (LSM): Noah-MP and CLM4. The model performance is evaluated through a set of observations, including the MODIS-derived MIDAS dust optical depth product, AERONET aerosol optical depth (AOD), ERA5-Land surface soil moisture and wind speed, and EMEP coarse particulate matter (PM10-PM2.5) measurements. We find that among the dust emission schemes, GOCART provides the most robust agreement with MIDAS and AERONET, closely followed by AFWA but with a wider spread, while UoC systematically diverges from observations failing to represent realistic column dust optical depth. Evaluation of surface drivers reveals that land-surface representation exerts a strong influence on dust emission magnitude and spatial distribution, with Noah-MP yielding systematically better agreement with observed meteorology and AOD, whereas CLM4 introduces more pronounced regional discrepancies. UoC exhibits improved alignment with coarse particulate matter measurements at the EMEP stations compared to GOCART and AFWA. Finally, we provide empirical scaling factors derived for each emissions mechanism–LSM pairing, applicable for WRF-Chem dust simulations, offering guidance for improved dust and air quality, and climate modelling applications.
- Preprint
(15436 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 19 Aug 2026)
- RC1: 'Comment on egusphere-2026-1900', Anonymous Referee #1, 21 Jul 2026 reply
-
RC2: 'Comment on egusphere-2026-1900', Anonymous Referee #2, 17 Aug 2026
reply
General comments:
This study evaluates the performance of the WRF-Chem model in simulating dust aerosols over the Northern Hemisphere dust belt with different combinations of three dust emission schemes (GOCART, GOCART-AFWA, UoC) and two land surface models (LSMs; Noah-MP and CLM4). By comparing with observational datasets such as AERONET and EMEP and reanalysis, the study evaluates modeled aerosol optical depth (AOD), dust AOD (DOD), surface concentrations of coarse particulate matter (PM10-PM2.5), and surface emission drivers (soil moisture and surface winds) across the six ensemble members. It is found that overall the GOCART and AFWA emission schemes show better agreement with AERONET AOD and MIDAS DOD, while the UoC scheme better captures the observed features of coarse particulate matter measurements. Empirical scaling factors for each combination of dust emission scheme and LSM are also provided for future similar studies. Overall, the paper is well written with well-prepared figures. Including an analysis of the representation of dust seasonality, providing statistics at regional levels, and expanding the discussion of uncertainties associated with the simulation settings would further strengthen the paper.
Specific comments:
- The dust belt is quite a large and geographically diverse area. While the paper aims to provide an overall evaluation over the entire region, it would also be valuable to identify which regions are better represented by the model than others. As discussed in the text and shown in figures (e.g., Figs. 2, 5), model performance shows large spatial variability. It would therefore be helpful to extend the statistical analysis (e.g., Table 1) into several sub-regions, such as North Africa, the Arabian Peninsula, and Central Asia. This would provide a more quantitative assessment of regional differences in model performance in addition to the domain-wide evaluation.
- Section 4.2 examines the model representation of key surface drivers of dust emissions, e.g., soil moisture and surface winds. It would also be helpful to provide some regional-level statistics when compared to the ERA5-Land reanalysis. Also, relating the model representation of these surface variables to model performance in simulating AOD would help understand model capability.
- Since the simulations were conducted over the entire year, it would be useful to provide some evaluation of how well the seasonality of dust aerosols is represented. It would be interesting to know whether the seasonal features are well captured by each combination of dust emission scheme-LSM, and if not, whether the biases in surface emission drivers contribute to the discrepancies.
- It is mentioned that model wind components, temperature, and humidity were nudged toward reanalysis (line 87). Was the nudge applied to the whole atmosphere at every pressure level, or only above the boundary layer, or at the surface level?
- Different empirical scaling factors for dust emissions were used for each combination of dust emission scheme and LSM to achieve better agreement with the MIDAS dataset. While the information is quite helpful for future similar studies, it should be noted that, because different scaling factors were applied for each ensemble member, the discrepancies among the ensemble members may reflect three factors: dust emission scheme, LSM, and empirical scale factors. The uncertainties associated with this should be discussed in Section 5.
- In Section 2.1, it would be helpful to clarify whether the threshold wind velocity (Ut) in the GOCART model and the threshold friction velocity (U*t) in the AFWA scheme were prescribed or calculated.
- Line 230, please consider including more information regarding the soil moisture from the ERA5-Land reanalysis. Are there any observations assimilated into the dataset?
- Line 252, “derived from MIDAS dust AOD observations”, MIDAS dust AOD is not a purely observational dataset, as the ratio of DOD to total AOD is derived from the MERRA-2 reanalysis. Uncertainties regarding the dataset should be discussed.
- Line 256, since this is the first table mentioned in the text, it should be numbered as Table 1.
- Fig. 1, it would be helpful to add latitude and longitude labels.
- Fig. 3, please consider adding the same statistics as shown in Fig. 2 (e.g., SF, R2, RMSE).
Technical corrections:
Line 179, “… and surface fiels”, do you mean “surface fields”?
Fig. 2, top panel, the figure title shows “MIDAS (AOD)” while the figure caption indicates dust AOD, please clarify which variable was plotted.
Citation: https://doi.org/10.5194/egusphere-2026-1900-RC2 -
CC1: 'Comments on "Evaluation of WRF-Chem (version 4.7.1) aeolian dust emission and land surface models over the dust belt" by S. Deb et al. These comments should be addressed before the publication.', Alexander Ukhov, 18 Aug 2026
reply
1. The literature review appears incomplete regarding previous WRF-Chem/GOCART dust studies over the Middle East. Ukhov et al. (2020, ACP) performed a multi-year, high-resolution WRF-Chem evaluation using AERONET, MODIS, MAIAC, MERRA-2, and CAMS, including AOD, aerosol size distributions, and surface PM. Ukhov et al. (2021, GMD) further documented important corrections and improvements to the WRF-Chem/GOCART dust implementation. These studies are directly relevant to the present model configuration and evaluation and should be referenced and discussed. Their omission is surprising given the strong overlap with the present work.
2. Authors state that "A comprehensive evaluation of the dust scheme in WRF-Chem over extended periods is lacking in the literature." This statement should be qualified. Previous multi-year WRF-Chem dust evaluations exist, including over the Middle East (see Comment 1). The broader dust-belt domain and systematic dust-scheme/LSM comparison may be novel, but the current statement understates previous work.
3. Use of MIDAS is reasonable because it provides dust-specific optical depth rather than total MODIS AOD. However, MIDAS is not purely observational because its dust fraction is derived from MERRA-2. The authors should discuss how this dependence may affect the evaluation and the derived scaling factors. Comparison with an additional independent product would strengthen the analysis.
4. Authors define (s_p) as "the fraction for each size section" but do not provide the numerical values used. These values should be explicitly reported for reproducibility. Dust size partitioning can substantially affect AOD, transport, deposition, and surface PM, and therefore may also influence the derived scaling factors.
5. The manuscript sometimes appears to equate better AOD agreement with a better dust emission scheme. However, AOD depends not only on dust emission flux but also on particle-size distribution, optical properties, transport, vertical distribution, and deposition. Better agreement with observed AOD therefore does not necessarily imply a more physically accurate emission flux. This distinction should be made clearer.
6. The 50 km horizontal resolution appears too coarse for evaluating regional dust-emission schemes, particularly over heterogeneous source regions such as the Arabian Peninsula. This concern is supported by 10km sensitivity experiment, where the resolution-related AOD variability reaches local values exceeding 60–80% over parts of the Middle East, and Arabian Peninsula. However, the 10km experiment covers only 24 days and uses only one LSM. Therefore, this limited sensitivity experiment does not seem sufficient to support the broad conclusion that the derived scaling factors are resolution-independent and transferable to higher-resolution simulations. This conclusion should be qualified.
Citation: https://doi.org/10.5194/egusphere-2026-1900-CC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 77 | 43 | 8 | 128 | 12 | 10 |
- HTML: 77
- PDF: 43
- XML: 8
- Total: 128
- BibTeX: 12
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General Comments:
This study presents the evaluation of WRF-Chem simulations in dust events across a broad region covering North Africa, Middle East, Central Asia and Europe during 2021. It focuses on the comparison of the performance of three dust emission schemes (GOCART, AFWA, and UoC) and two land surface models (Noah-MP and CLM). The study has shown some noticeable differences in dust simulations among the three dust emission schemes and between the two land surface models. It also summarizes some key differences and similarities. This study could provide some important information on the applicability of the dust emission schemes and land surface models in the simulation and forecast of dust events, and thus it could potentially serve as a useful reference for the relevant community. However, due to unclear statements and unreasonable interpretation in some key results, I suggest denial of the manuscript for its further improvement.
Major comments:
In fact, in MERRA-2, biases in DOD would lead to the biases in the ratio of DOD to total AOD. This would further lead to the biases in the estimation of DOD in MIDAS, which combines MODIS AOD with MERRA-2 ratio of DOD to total AOD (according to Equation in Line 223). If MERRA-2 underestimates the dust concentrations (especially the peak dust concentrations during strong dust events as some previous studies have shown), MIDAS DOD would be underestimated too. Therefore, compared to MIDAS DOD, WRF-Chem with UoC scheme shows significant overestimation, but this is not necessarily true. Indeed, compared to AERONET observations, WRF-Chem with UoC scheme shows underestimation, which is inconsistent with overestimation when compared to MIDAS DOD. Given the uncertainty in MERRA-2, it is not appropriate to use MIDAS DOD to have a quantitative conclusion. If the authors want to use MIDAS DOD to derive a quantitative conclusion, I think it should be combined with AERONET AOD to generate a new dataset for a consistent conclusion.
Specific comments:
Line 5: year-long: please also mention the year here.
Lines 30-31: within the atmospheric model: deleted?
Line 34: models: not clear.
Line 40: for the GOCART aerosol model: deleted?
Line 42: to model: changed to “to simulate”.
Line 63: in both cases: not clear.
Line 67: the dust scheme: which scheme?
Line 70: land surface: changed to “land surface model”?
Line 73: that captures: changed to “that is affected by”?
Line 76: The sentence is too long.
Line 78: in the broader spatio-temporal domain affected by the dust belt: not clear.
Lines 111-112: based on topographic source identification: Moved to after "Ginoux et al. (2001)"?
Line 123: dust particles: changed to “soil particles”.
Line 133: mesoscale features: What it means is not clear to me.
Line 139: m is the mass: m is the mass of saltating particle.
Line 146: Shao (2004): changed to “(Shao, 2004)”.
Line 148: deleted “, (Shao, 2004)”.
Line 151: where σp represents the mass ratio between free dust particles and aggregated dust: I do not think so. Probably not use σp here, and directly use pm(di)/pf (di) as above.
Line 153: The grammar is not correct.
Line 163: Niu et al. (2011): changed to “(Niu et al., 2011)”.
Lines 166-167: The grammar is not correct.
Line 169: which consists of.
Lines 176-177: It is not clear to me. I think only soil moisture at the top surface layer affects dust emission.
Line 178: Sec. 4.1.2: Sec. 3.1?
Line 179: fiels?
Lines 184-188: It is not clear to me.
Lines 214-215: AOD 550 nm MODerate resolution Imaging Spectroradiometer
215 (MODIS) satellite products?
Lines 235-238: Use the unit of cm as the previous paragraph?
Lines 241-242: What do the differences mean here (difference between different layers)? Please clarify.
Line 256: Table 2: It appears before Table 1.
Line 274: pockets?
Figure 2: What does SF mean?
Figure 2: middle: It is better to use modeled dust AOD minus a reference data.
Figure 4: Table 1: Why are only one value of MBE, MAE, or NME shown in Table 1?
Figure 5: The range of soil moisture is too large for dust emission region.
Line 332: East Asia: changed to “Central Asia”.
Lines 396-398: It is not clear.
Section 6: Conclusions: It is a bit too long. Please shorten it.
Line 469: the coarse dust consistency of surface concentrations: not clear.