the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 May 2026
ODEM v1.0: an offline dust emission model for reanalysis-driven source estimation
Abstract. Global dust emission estimates remain uncertain by a factor of two to three across models, with uncertainty arising from both the emission physics and the choice of meteorological forcing. Disentangling these two sources of uncertainty requires running the same emission scheme with different meteorological inputs, which is not possible in online models where the emission physics and the forcing are coupled within a single atmospheric model.
This paper presents ODEM v1.0 (Offline Dust Emission Model), a standalone Python implementation of the brittle fragmentation dust emission parameterization of Kok et al. (2014), following the implementation of Leung et al. (2023). ODEM accepts either ERA5 or MERRA-2 as meteorological forcing and produces gridded dust emission flux fields at the native spatiotemporal resolution of each reanalysis (0.25°, 1-hourly for ERA5; 0.5° × 0.625°, 1-hourly for MERRA-2). The model applies process-based emission physics – including soil particle size, moisture inhibition, aerodynamic drag partition, and turbulent intermittency corrections – to every land grid cell independently in a single forward pass with no spin-up requirement.
For the year 2006, ODEM driven by ERA5 at 1-hourly resolution produces a global PM20 emission of 15 539 Tg yr−1 and ODEM driven by MERRA-2 produces 12 747 Tg yr−1. These values exceed the observationally constrained PM20 budget of 5000±1600Tg yr−1 by factors of 3.1 and 2.5, respectively. This overshoot is a known property of the Kok et al. (2014) emission equation at its default calibration: Leung et al. (2023) report a comparable factor of 2.3 for their unnormalised scheme using the same equation, indicating that current emission physics cannot constrain the absolute magnitude from first principles. ERA5 produces 22 % more emission than MERRA-2, consistent with known differences in reanalysis friction velocity fields, particularly over North Africa. A sensitivity experiment using ERA5 at 3-hourly resolution yields 15 521 Tg yr−1 (0.1 % lower), confirming that the turbulent intermittency correction effectively accounts for sub-timestep wind variability. Both experiments produce roughly an order of magnitude higher emission than the MERRA-2 online GOCART scheme (1564 Tg yr−1), consistent with the known underestimation by empirical schemes relative to observational constraints.
- Preprint
(2963 KB) - Metadata XML
-
Supplement
(193 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-1905', Anonymous Referee #1, 18 Jun 2026
-
AC1: 'Reply on RC1', Metin Baykara, 09 Sep 2026
We thank the referee for the report. Its overall assessment names two concerns, novelty and validity, and our position on each is stated here first. On novelty: the emission physics is deliberately not new, and the contribution claimed is the released, reproducible implementation and the controlled comparisons it enables (point 1), broadened in the revision to the multi-scheme test-bed proposed by Referee 2 (point 6). On validity: we accept the substantive criticisms. The calibration does not constrain the absolute budget, and this response states the size of that discrepancy plainly (point 3); the submitted ERA5 to MERRA-2 difference conflates forcing physics with spatial resolution (point 5) and, as our own verification found, with temporal sampling convention; the revision replaces the single-number comparison with a decomposition in which each term is measured separately. The report's six general comments are unnumbered in the original; we number them 1 to 6 in the order given.
Before the point-by-point response we report the results of a verification pass we ran on the released code and archived outputs after the reviews arrived, because two findings correct published numbers and all of them bear on points below.
First, we built a regression harness that recomputes every published ODEM global total from the archived output fields. Two of the three totals reproduce exactly (15,539 Tg/yr for ERA5 1-hourly and 12,747 Tg/yr for MERRA-2). The third was misreported: the manuscript prints the ERA5 3-hourly total as 15,521 Tg/yr, while the archived output gives 15,532 Tg/yr, a transcription error of 0.07 percent. Two derived quantities follow it: the normalised value in Table 6 becomes 6213 rather than 6208, and the 1-hourly to 3-hourly difference is 0.05 percent rather than 0.1 percent. Every other ODEM total and derived quantity in the manuscript reproduces exactly from the archive, so the error is confined to one run's row and the two quantities derived from it, and the conclusion drawn from that sensitivity run is unchanged.
Second, running the released v1.0 on a single month exposed a defect: the model computes correctly and writes its monthly output file, then exits with an error in the run summary, because the period-mean flux field the summary reports was assigned only for multi-month runs. Both quick-start examples in the README trigger it. The fix changes no computed value; the corrected release is published at the code archive alongside the reviewed v1.0, which remains archived unchanged under the version DOI the manuscript cites.
Third, as a check on the archive itself, a fresh run of released v1.0 on the archived January 2006 MERRA-2 forcing reproduces our archived January output, from which the published fields were derived, bitwise at all 154,704,384 values.
Point 1. Novelty of a released implementation
Referee: "My biggest concern is that such python scripts are already developed alongside the code to be implemented into a model. People usually first develop high-level scripts to test and then implement it into Fortran-based chemical transport model or climate model code. For example, Leung et al (2023) posted their scripts in R language to calculate the emission flux. Using different meteorological fields is just changing the input source and variables into the calculation equation."
We accept the premise and contest the conclusion.
The premise is that the emission physics in ODEM is not new. It is not, the manuscript says so, and keeping it identical to the published scheme is the design: the paper's object of study is the sensitivity of computed emission to the meteorological forcing, which requires the physics to be held fixed.
The conclusion we contest is that a released implementation therefore adds nothing beyond the scripts accompanying the original papers. We examined that premise directly: we downloaded the archive in question (https://doi.org/10.5281/zenodo.7811426) and traced what it contains and what it requires. It contains twelve R scripts and a licence file, 86.9 kB in total: no data, no documentation, and no README. Every script begins by setting its working directory to a personal cloud-storage drive, and such paths appear throughout the code. Executing the scripts requires more than thirty preprocessed .RData input files that are in neither the archive nor any other public deposit; preprocessing code is included for one of the four model grids, and for that grid it reads pre-regridded meteorology whose generating code is likewise not included. The scripts therefore cannot be run from the archive without first re-deriving those intermediate inputs from their public sources. None of this is unusual for research code released in support of a paper, and it did not prevent us from reimplementing the scheme from the paper's equations. But it is the referee's premise, not ours, that these scripts constitute an available implementation. What they document is the method; what they do not provide is a runnable one.
ODEM v1.0 is what an available implementation looks like. MIT licence, a user manual, a physics verification suite of 38 independently computed cases, archived code and output (https://doi.org/10.5281/zenodo.19390605, https://doi.org/10.5281/zenodo.19320862), no hard-coded paths, and every input drawn from publicly available datasets, with sources and preprocessing steps documented. The contribution claimed here is the released, reproducible implementation and the controlled comparisons it enables, not the physics.
On "using different meteorological fields is just changing the input source": The two reanalyses do not even share a sampling convention for friction velocity (ERA5 archives an instantaneous field, MERRA-2 an hourly mean), and the effect of that difference on the global budget is not negligible in principle; we bound it in our response to point 3. Making two forcings genuinely comparable, in space, in time, and in convention, and treating forcing substitution as trivial is precisely how confounded comparisons arise. The revised manuscript treats forcing equivalence as an explicit part of the method rather than an input detail.
Finally, the revision broadens the contribution in the direction Referee 2 proposes: version two implements additional emission schemes behind the same forcing layer, so that the tool compares schemes as well as forcings. The scope is stated in our response to Referee 2.
Point 2. Offline motivation
Referee: "The author argued that using online dust emission scheme will have combined effects from the dust emission scheme and their radiative forcing, but there are offline chemical transport models using offline dust emission scheme to drive simulations. Online chemical transport models or climate models are not limited to online dust scheme as well. Offline dust emission scheme can be prescribed should the user choose. For example, an offline chemical transport model already had the Leung 2023 dust emission scheme implemented(Zhang et al., 2025)."
We accept the factual core of this point and withdraw the motivating claim it targets.
It is true that offline chemical transport models driven by prescribed emission exist, that online models can prescribe offline emission should the user choose, and that Zhang et al. (2025) implemented the Leung et al. (2023) scheme for GEOS-Chem, with the emission computed offline at the native resolution of its meteorology. The manuscript states in the abstract and in the conclusions that forcing-controlled comparison is "not possible in online models". That is too strong, and the revised manuscript withdraws it in favour of the narrower statement its own introduction already makes: within a coupled model, where the atmosphere generates the forcing fields internally, the meteorological forcing cannot be replaced wholesale while everything else is held fixed. The revision motivates the tool by what it makes convenient and reproducible rather than by what it asserts to be impossible elsewhere.
One clarification, because the wording of this point and of point 3 attributes to the paper an aim it does not have: ODEM computes emission only, and the manuscript makes no claim to disentangle radiative effects. The separation at issue throughout the paper is between meteorological forcing and emission physics.
We thank the referee for the pointer to Zhang et al. (2025) (https://doi.org/10.5194/gmd-18-6767-2025), which our submission should have engaged and the revision will. Its aim is complementary to ours: it drives the scheme with a single meteorological product chosen for its resolution (GEOS-FP), and a cross-reanalysis forcing comparison of the kind attempted here is not part of that study.
Point 3. Calibration constant and the charge of circularity
Referee: "Although the Leung 2023 dust emission scheme is process-based, it still needs empirical scaling factor to be calibrated to observational constraints. The choosing of different empirical parameters is subject to fine tuning to different meteorological fields. Using the same dust emission scheme equation is not desirable to disentangle the effects from emission scheme and their radiative effects. When the emission scheme is not tuned to generate reasonable dust emission flux, the disentangled radiative effects would be useless as well."
We accept the first and second of these, refer the third to the clarification under point 2, and contest only the inference that calibration makes the comparisons circular. We take them in that order.
The absolute magnitude is not constrained by the model, and we do not claim it is. C_tune is fitted, not derived. At C_tune = 0.05, the value carried over from Leung et al. (2023), our archived annual totals stand at 2.5 times the Kok et al. (2021) PM20 constraint of 5000 Tg/yr for the MERRA-2 configuration and 3.1 times it for ERA5, and the manuscript should state that as a limitation of the calibration rather than reporting it alongside the constraint and leaving the reader to infer the discrepancy. The referee is right that a reader cannot take the absolute totals as a prediction.
The referee is also right that the constant does not transfer between forcing datasets. A value fitted against one reanalysis has no claim to be correct for another, and the revised manuscript will say so explicitly rather than carrying a single constant across both configurations without comment.
On disentangling radiative effects: ODEM computes no radiative quantity and the manuscript claims no radiative disentangling; the clarification is under point 2 and we do not repeat it here.
What does not follow is that the comparisons are circular. The charge is checkable rather than arguable, and it fails on two independent grounds.
First, by construction. C_tune enters the emission calculation at exactly one place, as a scalar multiplier on the computed flux. It appears in no threshold friction velocity, in no drag partition, in no moisture or clay correction, and in no intermittency term. The flux is therefore exactly proportional to it, and any ratio taken between two runs that share a value of C_tune is exactly independent of that value. This is a property of where the constant sits in the equations, not an approximation that holds over some range.
Second, and more usefully for a reader who would rather not take our word for the algebra, it is demonstrated by output that is archived alongside the paper and reported in the submitted manuscript itself. Sect. 5.3 and Table 6 present both calibrations, the table caption states that every entry in the normalised column equals the default column times 0.4 with all ratios preserved exactly, and the data deposit (https://doi.org/10.5281/zenodo.19320862) contains all six annual fields. The C_tune = 0.020 runs are exact 0.4 scalings of their 0.05 counterparts to floating-point precision, everywhere in the field and not merely in the global total. The linearity is therefore a verified property of the released model rather than a claim about it.
So the relative comparison between two forcing datasets is unaffected by the choice of calibration constant, including by a badly chosen one. What the calibration limits is the absolute budget, which we concede without reservation. Two consequences follow for how the comparison should be read, and the revision states both. A separate calibration per forcing would make the global budgets agree by construction and remove the ratio from view; the shared-constant ratio is kept because it is the quantity a modeller needs when carrying a constant from one reanalysis to another, namely how far the untuned physics moves under the forcing swap. And the regional, seasonal and spatial differences between the two runs, which carry most of the comparison in Sects. 4.2 to 4.4, are independent of any global constant, shared or separate.
One further qualification, which we raise ourselves because it bears on the same comparison. The referee's fifth point, that our resolution comparison conflates resolution with forcing physics, prompted us to examine what else the ERA5 against MERRA-2 difference contains. It contains a third term that neither of us identified at review time. The two products do not share a sampling convention: ERA5 friction velocity is an instantaneous value at the whole hour, while the MERRA-2 field is an hourly time average. Emission is strongly convex in friction velocity, so time-averaged forcing yields systematically less emission than instantaneous forcing drawn from the same winds. We have measured the size of this effect directly, by re-running a month of the 3-hourly ERA5 forcing with the friction velocity smoothed and holding every other input fixed: a two-sample (6 hour) mean costs 10.0 percent of the global budget and a three-sample (9 hour) mean costs 17.9 percent. The 6 hour window is the mildest constructible from 3-hourly data, and both are wider than the MERRA-2 averaging interval of one hour, so they bound the effect rather than quantify it, but the direction is unambiguous.
That does not overturn the C_tune argument above, which is exact and independent of it. It does mean that the difference we reported between the two reanalyses is not attributable to forcing physics alone, and we will not present it as though it were. The revision decomposes it into spatial resolution, sampling convention, and scheme physics, each measured separately, which is the form in which the comparison answers the referee's point 5 rather than merely surviving it.
Point 4. Comparison with the online GOCART scheme
Referee: "There are too many differences between the Leung 2023 scheme and the online GOCART scheme. I do not see the rational to compare the two."
We accept this. The DUEM row was included as context, being the emission actually used in the MERRA-2 aerosol reanalysis whose meteorology drives one of our runs, but the manuscript presents it in the main budget table and comparison figures, which invites reading it as a meaningful scheme comparison. The revised manuscript removes DUEM from the budget table and the comparison figures and retains at most a single clearly labelled contextual sentence.
Point 5. Spatial resolution and the order of regridding
Referee: "Many prior studies have shown the essential role of spatial resolution of meteorological fields on dust emission flux. The effects are not just regridding the native dust emission flux to a lower resolution, but using regridding the meteorological fields first and then calculate the dust emission flux, because that is what we used actually when conducting the simulation. The differences between ERA5 and MERRA2 on dust emissions are not only from the physical meteorological scheme embedded in the product, but also the different spatial resolutions."
We accept this point in full, and it is the most constructive comment in the report.
The referee is right that regridding the output flux is not equivalent to regridding the meteorological fields and recomputing, because emission is strongly nonlinear in friction velocity, and therefore right that our ERA5 to MERRA-2 difference conflates forcing physics with spatial resolution. The HEMCO description makes a closely related point, that nonlinear emission computations depend on the grid on which they are evaluated (Lin et al., 2021, https://doi.org/10.5194/gmd-14-5487-2021), and we should have engaged it.
The revision runs the controlled experiment the referee describes: ERA5 meteorological fields regridded to the MERRA-2 grid before the emission calculation, with static inputs held identical, so that resolution separates from forcing physics. Our post-submission analysis adds a third term to the same decomposition, the sampling convention described in our response to point 3. The revised comparison therefore reports the inter-reanalysis difference decomposed into spatial resolution, temporal sampling convention, and forcing physics at matched grid, each measured separately, and no single-number attribution of the 22 percent difference to reanalysis physics will be retained.
Point 6. Marginal benefit relative to HEMCO
Referee: "The benefit of using python scripts to calculate dust emissions without running a chemical transport model is marginal as people can turn off all other atmospheric processes but only emissions to generate offline dust emissions. For example, the emission model called HEMCO(Lin et al., 2021) can achieve this purpose gracefully without running all other atmospheric processes in GEOS-Chem."
We accept the premise here as well: standalone offline emission computation is not in itself new, HEMCO's standalone mode is exactly that, and our submission should have cited Lin et al. (2021). The revision corrects this.
The distinction is in what each tool computes and requires. The standalone mode of HEMCO computes dust with the DEAD and Ginoux extensions, and adding a scheme means writing a new extension within its Fortran framework; the Leung et al. (2023) scheme has since been run offline within the GEOS-Chem system by Zhang et al. (2025), under one meteorological product. ODEM is a dependency-light Python implementation that ingests reanalysis fields from more than one product directly, without the transport-model infrastructure. These are different tools for different users rather than competing implementations of the same capability, and the revised manuscript positions ODEM against HEMCO and GEOS-Chem explicitly rather than leaving the comparison to be inferred.
The stronger answer to the question behind this point, why the tool should exist, is the direction Referee 2 proposes and the revision adopts: a scheme test-bed, several emission schemes behind one forcing layer, run across more than one reanalysis, with the inter-forcing difference decomposed into resolution, sampling convention and physics. Two-scheme comparison under a single product already exists within GEOS-Chem, where Zhang et al. (2025) compare the Leung et al. (2023) scheme with the default DEAD-based scheme under GEOS-FP; the combination of several schemes with several forcings and a measured decomposition, in a standalone tool, is what none of the systems named in this report provides. The scope is stated in our response to Referee 2.
A revision along these lines involves new experiments and a substantial restructuring rather than textual amendments. So that both referees see one consistent commitment, the full revision scope, including the multi-scheme extension, is stated once, in our response to Referee 2.
References:
Kok, J. F., Adebiyi, A. A., Albani, S., Balkanski, Y., Checa-Garcia, R., Chin, M., Colarco, P. R., Hamilton, D. S., Huang, Y., Ito, A., Klose, M., Li, L., Mahowald, N. M., Miller, R. L., Obiso, V., Pérez García-Pando, C., Rocha-Lima, A., and Wan, J. S.: Contribution of the world's main dust source regions to the global cycle of desert dust, Atmos. Chem. Phys., 21, 8169–8193, https://doi.org/10.5194/acp-21-8169-2021, 2021.
Leung, D. M., Kok, J. F., Li, L., Okin, G. S., Prigent, C., Klose, M., Pérez García-Pando, C., Menut, L., Mahowald, N. M., Lawrence, D. M., and Chamecki, M.: A new process-based and scale-aware desert dust emission scheme for global climate models – Part 1: Description and evaluation against inverse modeling emissions, Atmos. Chem. Phys., 23, 6487–6523, https://doi.org/10.5194/acp-23-6487-2023, 2023.
Lin, H., Jacob, D. J., Lundgren, E. W., Sulprizio, M. P., Keller, C. A., Fritz, T. M., Eastham, S. D., Emmons, L. K., Campbell, P. C., Baker, B., Saylor, R. D., and Montuoro, R.: Harmonized Emissions Component (HEMCO) 3.0 as a versatile emissions component for atmospheric models: application in the GEOS-Chem, NASA GEOS, WRF-GC, CESM2, NOAA GEFS-Aerosol, and NOAA UFS models, Geosci. Model Dev., 14, 5487–5506, https://doi.org/10.5194/gmd-14-5487-2021, 2021.
Zhang, D., Martin, R. V., Liu, X., van Donkelaar, A., Oxford, C. R., Li, Y., Meng, J., Leung, D. M., Kok, J. F., Li, L., Zhu, H., Turner, J. R., Yan, Y., Brauer, M., Rudich, Y., and Windwer, E.: Improving annual fine mineral dust representation from the surface to the column in GEOS-Chem 14.4.1, Geosci. Model Dev., 18, 6767–6803, https://doi.org/10.5194/gmd-18-6767-2025, 2025.
Citation: https://doi.org/10.5194/egusphere-2026-1905-AC1
-
AC1: 'Reply on RC1', Metin Baykara, 09 Sep 2026
-
RC2: 'Comment on egusphere-2026-1905', Anonymous Referee #2, 13 Aug 2026
This model development paper presents an offline Python tool for calculating mineral dust emission fluxes based on different, freely selectable meteorological input data using the ‘brittle fragmentation’ approach. The tool is intended to help assess uncertainties related to meteorological forcing and emission physics and can be used to investigate global dust emission patterns and budgets.
The manuscript is well written and thoroughly prepared, with good-quality figures, and is of solid technical quality overall. However, I have concerns about the novelty and relevance of the concept in its current form, and cannot therefore recommend its publication.
The tool would be of greater scientific value if it allowed not only the selection of different meteorological inputs but also different dust emission schemes, with several approaches already implemented and compared in the paper. Such a test-bed tool would certainly be of scientific interest. But as it is now, the question arises as to whether the code for the underlying and already cited emission scheme is not already available in another form.
In addition, the manuscript’s intended conclusions regarding spatial and temporal dependencies are not examined in a sufficiently rigorous manner, partly because the available input data do not allow for this. While the need to further fine-tune the emission scheme is highlighted, the unrealistically high emission values produced by the tool are not really addressed. Instead, the authors mainly point out that similar overestimation has already been reported for this approach in previous studies. This leaves some uncertainty regarding the validity of the tool and, in particular, the reliability of its results. At the same time, potentially interesting applications, such as comparing meteorological driving data with and without moist convection-resolving information or comparing different emission schemes across various dust source regions, remain largely unexplored
Citation: https://doi.org/10.5194/egusphere-2026-1905-RC2 -
AC2: 'Reply on RC2', Metin Baykara, 09 Sep 2026
We thank the referee for the report, and we accept its central suggestion as the direction of the revision.
Referee (the report's opening summary paragraph is omitted): "The manuscript is well written and thoroughly prepared, with good-quality figures, and is of solid technical quality overall. However, I have concerns about the novelty and relevance of the concept in its current form, and cannot therefore recommend its publication.
"The tool would be of greater scientific value if it allowed not only the selection of different meteorological inputs but also different dust emission schemes, with several approaches already implemented and compared in the paper. Such a test-bed tool would certainly be of scientific interest. But as it is now, the question arises as to whether the code for the underlying and already cited emission scheme is not already available in another form.
"In addition, the manuscript’s intended conclusions regarding spatial and temporal dependencies are not examined in a sufficiently rigorous manner, partly because the available input data do not allow for this. While the need to further fine-tune the emission scheme is highlighted, the unrealistically high emission values produced by the tool are not really addressed. Instead, the authors mainly point out that similar overestimation has already been reported for this approach in previous studies. This leaves some uncertainty regarding the validity of the tool and, in particular, the reliability of its results. At the same time, potentially interesting applications, such as comparing meteorological driving data with and without moist convection-resolving information or comparing different emission schemes across various dust source regions, remain largely unexplored"
The revision scope. Version two restructures the paper around a multi-scheme comparison. The emission physics in v1.0 is already isolated behind a single interface, one routine with no internal state evaluated once per timestep, and a second scheme, an albedo-based drag partition formulation after Chappell and Webb (2016), has been implemented and run for a test month in the development version. The revision extends this to at least three scheme families, adding the DEAD saltation-sandblasting scheme (Zender et al., 2003a) as configured in GEOS-Chem, run in turn under the two source functions compared by Leung et al. (2023), those of Ginoux et al. (2001) and Zender et al. (2003b), so the compared configurations outnumber the scheme families. All schemes are run against identical forcing from more than one reanalysis through the same input layer, each at its published constants; the untuned comparison is the primary result, with budget-normalised fields carried only for pattern comparison, across the major dust source regions. Alongside the scheme comparison, the revision adds the controlled resolution experiment requested by Referee 1 (meteorological fields regridded to a common grid before the emission calculation) and the measured treatment of forcing sampling conventions described below.
On whether the underlying scheme's code is already available in another form. The question is fair, and we checked it directly rather than answering from impression. As documentation of the method, yes; as a runnable implementation, no. The archive accompanying Leung et al. (2023) (https://doi.org/10.5281/zenodo.7811426) contains twelve R scripts and a licence, 86.9 kB, with no data and no documentation; executing the scripts requires more than thirty preprocessed input files that exist in no public deposit, and preprocessing code is included for one of the four model grids. The scripts cannot be run from the archive without first re-deriving those inputs. The full trace is in our response to Referee 1, point 1. ODEM v1.0 is released with a user manual, a physics verification suite, archived code and outputs, and inputs drawn from public sources throughout. Under the restructuring above, the question also changes character: the revised tool's value lies in the comparison capability, not in any single scheme's availability.
On the rigour of the spatial and temporal conclusions. We accept this, and our own post-submission verification supports the referee's reading with specifics. The submitted comparison of 1-hourly and 3-hourly ERA5 forcing supports a weaker conclusion than the text states: both runs draw on the same hourly product and differ only in which hours were retrieved, so the experiment demonstrates insensitivity to subsampling of one product, not insensitivity to forcing temporal resolution in general. The revised text will claim only what the experiment shows. Beyond that, the two reanalyses do not share a sampling convention: ERA5 archives instantaneous friction velocity, MERRA-2 an hourly mean, and emission is strongly convex in friction velocity. Measured on one month of the 3-hourly ERA5 forcing with every other input held fixed, smoothing friction velocity over two samples (6 hours) lowers the global budget by 10.0 percent and over three samples (9 hours) by 17.9 percent; the 6 hour window is the mildest constructible from 3-hourly data, and both are wider than the MERRA-2 averaging interval of one hour, so they bound the effect rather than quantify it, but the direction is unambiguous. Part of the published 22 percent ERA5 to MERRA-2 difference is therefore sampling convention rather than atmosphere, which is exactly the class of confound the referee points to. The referee is also right that the input data limit what can be cleaned: MERRA-2 publishes no instantaneous friction velocity, so for a friction-velocity-driven scheme the convention mismatch can be bounded but not removed. Wind-driven schemes in the revised test-bed can be run convention-matched using the instantaneous MERRA-2 collection (M2I1NXASM), and the revised manuscript states, per comparison, which fields are convention-matched and which are bounded.
On the emission magnitudes. The referee is right that pointing at comparable overestimates in earlier studies is not an answer, and the revision does not repeat it. Two things replace it. First, per-scheme, per-region budgets are set against the observationally constrained regional PM20 emission of Kok et al. (2021), which the submitted manuscript already uses in Sect. 4.3, so that the overshoot is located by region and by scheme rather than stated once as a global factor. Second, every multiplicative factor in the emission equation (drag partition, moisture inhibition, erodibility, fragmentation exponent, intermittency) is computed per grid cell and can be written as diagnostic output, so the revision reports, for each region, which factors contribute most to the magnitude and how they differ between schemes. Untuned magnitudes are presented through that decomposition, with budget-normalised fields carried for pattern comparison. This turns the magnitude from a reported discrepancy into a located and attributed one.
On the two applications named. The revision covers the cross-region scheme comparison: the test-bed reports per-scheme, per-region budgets and spatial patterns. The comparison of forcing with and without moist-convection-resolving information is out of scope for this revision: the forcing set remains global reanalyses, and a convection-permitting comparison is a study of its own. The revised outlook states this explicitly rather than leaving it implied.
One further disclosure, because the referee raises the reliability of the results. A post-submission verification pass produced numerical corrections that are detailed in our response to Referee 1: a misreported total (15,521 corrected to 15,532 Tg/yr, with two derived values), and a fixed defect in the single-month code path that changes no computed value. One further finding belongs here because it concerns a printed equation. Eq. (10) of the submitted manuscript, the turbulent intermittency factor, is misstated: it writes the un-normalised threshold crossing rate, a quantity with units of inverse time, where a probability belongs, and it anchors the expression to Eq. (5) of Comola et al. (2019), which is a different relation in that paper. The released code implements the correct object: in place of the rate, the weight of Comola et al. (2019), their Eq. (8), which is the normalised ratio of the crossing rates at the two thresholds, dimensionless and bounded between 0 and 1 by construction, entering the occurrence probability exactly as their Eq. (6) prescribes. The prefactor of the crossing rate cancels in that ratio, which is why the Lagrangian timescale quoted in the manuscript text does not appear in the implementation and the boundary-layer height enters only as the presence switch the manuscript already describes, not numerically; the code evaluates the wind statistics at 10 m with the neutral-limit standard deviation sigma_u = 2.3 u* (Panofsky et al., 1977). The computation and all archived results are unchanged. Version two corrects the equation and its citation anchor, and documents the neutral-stability simplification explicitly. We report this as found by our own verification so that the record is complete.
References
Chappell, A. and Webb, N. P.: Using albedo to reform wind erosion modelling, mapping and monitoring, Aeolian Res., 23, 63–78, https://doi.org/10.1016/j.aeolia.2016.09.006, 2016.
Comola, F., Kok, J. F., Chamecki, M., and Martin, R. L.: The intermittency of wind-driven sand transport, Geophys. Res. Lett., 46, 13430–13440, https://doi.org/10.1029/2019GL085739, 2019.
Ginoux, P., Chin, M., Tegen, I., Prospero, J. M., Holben, B., Dubovik, O., and Lin, S.-J.: Sources and distributions of dust aerosols simulated with the GOCART model, J. Geophys. Res., 106, 20255–20273, https://doi.org/10.1029/2000JD000053, 2001.
Kok, J. F., Adebiyi, A. A., Albani, S., Balkanski, Y., Checa-Garcia, R., Chin, M., Colarco, P. R., Hamilton, D. S., Huang, Y., Ito, A., Klose, M., Li, L., Mahowald, N. M., Miller, R. L., Obiso, V., Pérez García-Pando, C., Rocha-Lima, A., and Wan, J. S.: Contribution of the world's main dust source regions to the global cycle of desert dust, Atmos. Chem. Phys., 21, 8169–8193, https://doi.org/10.5194/acp-21-8169-2021, 2021.
Leung, D. M., Kok, J. F., Li, L., Okin, G. S., Prigent, C., Klose, M., Pérez García-Pando, C., Menut, L., Mahowald, N. M., Lawrence, D. M., and Chamecki, M.: A new process-based and scale-aware desert dust emission scheme for global climate models – Part 1: Description and evaluation against inverse modeling emissions, Atmos. Chem. Phys., 23, 6487–6523, https://doi.org/10.5194/acp-23-6487-2023, 2023.
Panofsky, H. A., Tennekes, H., Lenschow, D. H., and Wyngaard, J. C.: The characteristics of turbulent velocity components in the surface layer under convective conditions, Bound.-Layer Meteorol., 11, 355–361, https://doi.org/10.1007/BF02186086, 1977.
Zender, C. S., Bian, H., and Newman, D.: Mineral Dust Entrainment and Deposition (DEAD) model: Description and 1990s dust climatology, J. Geophys. Res., 108, 4416, https://doi.org/10.1029/2002JD002775, 2003a.
Zender, C. S., Newman, D., and Torres, O.: Spatial heterogeneity in aeolian erodibility: Uniform, topographic, geomorphic, and hydrologic hypotheses, J. Geophys. Res., 108, 4543, https://doi.org/10.1029/2002JD003039, 2003b.
Citation: https://doi.org/10.5194/egusphere-2026-1905-AC2
-
AC2: 'Reply on RC2', Metin Baykara, 09 Sep 2026
Data sets
ODEM v1.0 model output: annual mean dust emission fields (2006) M. Baykara https://doi.org/10.5281/zenodo.19320862
Model code and software
ODEM v1.0: Offline Dust Emission Model M. Baykara https://doi.org/10.5281/zenodo.19390605
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 311 | 103 | 24 | 438 | 47 | 25 | 21 |
- HTML: 311
- PDF: 103
- XML: 24
- Total: 438
- Supplement: 47
- BibTeX: 25
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This paper developed an offline tool to calculate dust emission flux using python with different meteorological inputs based on the brittle fragmentation theory. I am concerned with the novelty and validity of the approach and thus do not recommend publication.
General Comments
References
Leung, D. M., Kok, J. F., Li, L., Okin, G. S., Prigent, C., Klose, M., Pérez García-Pando, C., Menut, L., Mahowald, N. M., Lawrence, D. M., and Chamecki, M.: A new process-based and scale-aware desert dust emission scheme for global climate models – Part I: Description and evaluation against inverse modeling emissions, Atmospheric Chemistry and Physics, 23, 6487–6523, https://doi.org/10.5194/acp-23-6487-2023, 2023.
Lin, H., Jacob, D. J., Lundgren, E. W., Sulprizio, M. P., Keller, C. A., Fritz, T. M., Eastham, S. D., Emmons, L. K., Campbell, P. C., Baker, B., Saylor, R. D., and Montuoro, R.: Harmonized Emissions Component (HEMCO) 3.0 as a versatile emissions component for atmospheric models: application in the GEOS-Chem, NASA GEOS, WRF-GC, CESM2, NOAA GEFS-Aerosol, and NOAA UFS models, Geosci. Model Dev., 14, 5487–5506, https://doi.org/10.5194/gmd-14-5487-2021, 2021.
Zhang, D., Martin, R. V., Liu, X., van Donkelaar, A., Oxford, C. R., Li, Y., Meng, J., Leung, D. M., Kok, J. F., Li, L., Zhu, H., Turner, J. R., Yan, Y., Brauer, M., Rudich, Y., and Windwer, E.: Improving annual fine mineral dust representation from the surface to the column in GEOS-Chem 14.4.1, Geoscientific Model Development, 18, 6767–6803, https://doi.org/10.5194/gmd-18-6767-2025, 2025.