• Alfaro, S. C., and L. Gomes, 2001: Modeling mineral aerosol production by wind erosion: Emission intensities and aerosol size distribution in source areas. J. Geophys. Res., 106, 18 07518 084, https://doi.org/10.1029/2000JD900339.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Babiker, A., I. Eltayeb, and M. Hassan, 1987: A statistical model for horizontal mass flux of erodible soil. J. Geophys. Res., 92, 14 84514 849, https://doi.org/10.1029/JD092iD12p14845.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Beegum, S., I. Gherboudj, N. Chaouch, F. Couvidat, L. Menut, and H. Ghedira, 2016: Simulating aerosols over Arabian Peninsula with CHIMERE: Sensitivity to soil, surface parameters and anthropogenic emission inventories. Atmos. Environ., 128, 185197, https://doi.org/10.1016/j.atmosenv.2016.01.010.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Briant, R., P. Tuccella, A. Deroubaix, D. Khvorostyanov, L. Menut, S. Mailler, and S. Turquety, 2017: Aerosol–radiation interaction modelling using online coupling between the WRF 3.7.1 meteorological model and the CHIMERE 2016 chemistry-transport model, through the OASIS3-MCT coupler. Geosci. Model Dev., 10, 927944, https://doi.org/10.5194/gmd-10-927-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Christofferson, R., and D. Gillette, 1987: A simple estimator of the shape factor of the two-parameter Weibull distribution. J. Climate Appl. Meteor., 26, 323325, https://doi.org/10.1175/1520-0450(1987)026<0323:ASEOTS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cuevas, E., and et al. , 2015: The MACC-II 2007–2008 reanalysis: Atmospheric dust evaluation and characterization over northern Africa and the Middle East. Atmos. Chem. Phys., 15, 39914024, https://doi.org/10.5194/acp-15-3991-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dubovik, O., and M. D. King, 2000: A flexible inversion algorithm for retrieval of aerosol optical properties from Sun and sky radiance measurements. J. Geophys. Res., 105, 20 67320 696, https://doi.org/10.1029/2000JD900282.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gillette, D., and R. Passi, 1988: Modeling dust emissions caused by wind erosion. J. Geophys. Res., 93, 14 23314 242, https://doi.org/10.1029/JD093iD11p14233.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ginoux, P., M. Chin, I. Tegen, J. M. Prospero, B. Holben, O. Dubovik, and S. J. Lin, 2001: Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res., 106, 20 25520 273, https://doi.org/10.1029/2000JD000053.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gómez-Navarro, J. J., C. C. Raible, and S. Dierer, 2015: Sensitivity of the WRF model to PBL parametrisations and nesting techniques: Evaluation of wind storms over complex terrain. Geosci. Model Dev., 8, 33493363, https://doi.org/10.5194/gmd-8-3349-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grini, A., and C. S. Zender, 2004: Roles of saltation, sandblasting, and wind speed variability on mineral dust aerosol size distribution during the Puerto Rican Dust Experiment (PRIDE). J. Geophys. Res., 109, D07202, https://doi.org/10.1029/2003JD004233.

    • Search Google Scholar
    • Export Citation
  • Gryning, S.-E., R. Floors, A. Peña, E. Batchvarova, and B. Brümmer, 2016: Weibull wind-speed distribution parameters derived from a combination of wind-lidar and tall-mast measurements over land, coastal and marine sites. Bound.-Layer Meteor., 159, 329348, https://doi.org/10.1007/s10546-015-0113-x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • He, Y., A. H. Monahan, C. G. Jones, A. Dai, S. Biner, D. Caya, and K. Winger, 2010: Probability distributions of land surface wind speeds over North America. J. Geophys. Res., 115, D04103, https://doi.org/10.1029/2008JD010708.

    • Search Google Scholar
    • Export Citation
  • Holben, B., and et al. , 2001: An emerging ground-based aerosol climatology: Aerosol optical depth from AERONET. J. Geophys. Res., 106, 12 06712 097, https://doi.org/10.1029/2001JD900014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huneeus, N., and et al. , 2011: Global dust model intercomparison in AeroCom phase I. Atmos. Chem. Phys., 11, 77817816, https://doi.org/10.5194/acp-11-7781-2011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Justus, C., W. Hargraves, and A. Mikhail, and D. Graber, 1978: Methods for estimating wind speed frequency distributions. J. Appl. Meteor., 17, 350353, https://doi.org/10.1175/1520-0450(1978)017<0350:MFEWSF>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kelly, M., I. Troen, and H. Jorgensen, 2014: Weibull-k revisited: “Tall” profiles and height variation of wind statistics. Bound.-Layer Meteor., 152, 107124, https://doi.org/10.1007/s10546-014-9915-5.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kok, J. F., E. J. R. Parteli, T. I. Michaels, and D. Bou Karam, 2012: The physics of wind-blown sand and dust. Rep. Prog. Phys., 75, 106901, https://doi.org/10.1088/0034-4885/75/10/106901.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Loveland, T. R., B. C. Reed, J. F. Brown, D. O. Ohlen, Z. Zhu, L. Yang, and J. W. Merchant, 2000: Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. Int. J. Remote Sens., 21, 13031330, https://doi.org/10.1080/014311600210191.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mailler, S., and et al. , 2017: CHIMERE-2017: From urban to hemispheric chemistry-transport modeling. Geosci. Model Dev., 10, 23972423, https://doi.org/10.5194/gmd-10-2397-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marticorena, B., G. Bergametti, B. Aumont, Y. Callot, C. N’Doume, and M. Legrand, 1997: Modeling the atmospheric dust cycle: 2. Simulations of Saharan dust sources. J. Geophys. Res., 102, 43874404, https://doi.org/10.1029/96JD02964.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marticorena, B., and et al. , 2010: Temporal variability of mineral dust concentrations over West Africa: Analyses of a pluriannual monitoring from the AMMA Sahelian Dust Transect. Atmos. Chem. Phys., 10, 88998915, https://doi.org/10.5194/acp-10-8899-2010.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., 2008: Sensitivity of hourly Saharan dust emissions to NCEP and ECMWF modelled wind speed. J. Geophys Res., 113, D16201, https://doi.org/10.1029/2007JD009522.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., C. Schmechtig, and B. Marticorena, 2005: Sensitivity of the sandblasting fluxes calculations to the soil size distribution accuracy. J. Atmos. Oceanic Technol., 22, 18751884, https://doi.org/10.1175/JTECH1825.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., and et al. , 2013a: CHIMERE 2013: A model for regional atmospheric composition modelling. Geosci. Model Dev., 6, 9811028, https://doi.org/10.5194/gmd-6-981-2013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., C. Pérez, K. Haustein, B. Bessagnet, C. Prigent, and S. Alfaro, 2013b: Relative impact of roughness and soil texture on mineral dust emission fluxes modeling. J. Geophys. Res. Atmos., 118, 65056520, https://doi.org/10.1002/jgrd.50313.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., G. Siour, S. Mailler, F. Couvidat, and B. Bessagnet, 2016: Observations and regional modeling of aerosol optical properties, speciation and size distribution over Northern Africa and western Europe. Atmos. Chem. Phys., 16, 12 96112 982, https://doi.org/10.5194/acp-16-12961-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., S. Mailler, B. Bessagnet, G. Siour, A. Colette, F. Couvidat, and F. Meleux, 2017: An alternative way to evaluate chemistry-transport model variability. Geosci. Model Dev., 10, 11991208, https://doi.org/10.5194/gmd-10-1199-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nachtergaele, F., H. van Velthuizen, and L. Verelst, 2009: Harmonized World Soil Database. Version 1.1, FAO, IIASA, ISRIC, ISS-CAS, JRC, 38 pp.

  • Pérez, I. A., M. L. Sánchez, and M. A. García, 2007:Weibull wind speed distribution: Numerical considerations and use with sodar data. J. Geophys. Res., 112, D20112, https://doi.org/10.1029/2006JD008278.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Prigent, C., C. Jiménez, and J. Catherinot, 2012: Comparison of satellite microwave backscattering (ASCAT) and visible/near-infrared reflectances (PARASOL) for the estimation of aeolian aerodynamic roughness length in arid and semi-arid regions. Atmos. Meas. Tech., 5, 27032712, https://doi.org/10.5194/amt-5-2703-2012.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pryor, S. C., J. T. Schoof, and R. J. Barthelmie, 2005: Empirical downscaling of wind speed probability distributions. J. Geophys. Res., 110, D19109, https://doi.org/10.1029/2005JD005899.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ridley, D. A., C. L. Heald, J. R. Pierce, and M. J. Evans, 2013: Toward resolution-independent dust emissions in global models: Impacts on the seasonal and spatial distribution of dust. Geophys. Res. Lett., 40, 28732877, https://doi.org/10.1002/grl.50409.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schepanski, K., I. Tegen, B. Laurent, B. Heinold, and A. Macke, 2007: A new Saharan dust source activation frequency map derived from MSG-SEVIRI IR-channels. Geophys. Res. Lett., 34, L18803, https://doi.org/10.1029/2007GL030168.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shao, Y., 2001: A model for mineral dust erosion. J. Geophys. Res., 106, 20 23920 254, https://doi.org/10.1029/2001JD900171.

  • Shao, Y., and I. Lu, 2000: A simple expression for wind erosion threshold friction velocity. J. Geophys. Res., 105, 22 43722 443, https://doi.org/10.1029/2000JD900304.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shao, Y., M. R. Raupach, and P. A. Findlater, 1993: Effect of saltation bombardment on the entrainment of dust by wind. J. Geophys. Res., 98, 12 71912 726, https://doi.org/10.1029/93JD00396.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Skamarock, W., J. Klemp, J. Dudhia, D. Gill, D. Barker, W. Wang, and J. Powers, 2007: A description of the Advanced Research WRF version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp.

  • Su, L. and O. B. Toon, 2009: Numerical simulations of Asian dust storms using a coupled climate-aerosol microphysical model. J. Geophys. Res., 114, D14202, https://doi.org/10.1029/2008JD010956.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tegen, I., and I. Fung, 1994: Modeling of mineral dust in the atmosphere: Sources, transport, and optical thickness. J. Geophys. Res., 99, 22 89722 914, https://doi.org/10.1029/94JD01928.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tegen, I., B. Heinold, M. Todd, J. Helmert, R. Washington, and O. Dubovik, 2006: Modelling soil dust aerosol in the Bodélé depression during the BoDEx campaign. Atmos. Chem. Phys., 6, 43454359, https://doi.org/10.5194/acp-6-4345-2006.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weibull, W., 1951: A statistical distribution function of wide applicability. J. Appl. Mech., 18, 293297.

  • White, B. R., 1986: Particle dynamics in two-phase flows. Encyclopedia of Fluid Mechanics: Slurry Flow Technology. N. P. Cheremisinoff, Ed., Gulf Publishing, 239–282.

  • Wild, O., X. Zhu, and M. J. Prather, 2000: Fast-J: Accurate simulation of in- and below-cloud photolysis in tropospheric chemical models. J. Atmos. Chem., 37, 245282, https://doi.org/10.1023/A:1006415919030.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, K., C. Zhao, H. Wan, Y. Qian, R. C. Easter, S. J. Ghan, K. Sakaguchi, and X. Liu, 2016: Quantifying the impact of sub-grid surface wind variability on sea salt and dust emissions in CAM5. Geosci. Model Dev., 9, 607632, https://doi.org/10.5194/gmd-9-607-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    The domain defined for the real test case. (top) Map of the erodibility (0:1). (bottom) Map of the orography variance (m2).

  • View in gallery

    Example of Weibull wind speed distribution for a mean value of = 8 (black) and 14 (blue) m s−1. The three curves correspond to k = 3, k = 4, and the parameterization in J78, respectively. With J78, the corresponding k values are k = 2.66 for = 8 m s−1 and k = 3.52 for = 14 m s−1.

  • View in gallery

    The factor variability as a function of the orography variance.

  • View in gallery

    (top) Values of the k parameter values for the J78 parameterization and include the orographic weight factor proposed in this study. The two curves represent the extrema of the change, with and corresponding to the lowest and largest orography variances, respectively. (middle) Weibull distribution with and without taking into account orography variance. (bottom) Sensitivity of the mineral dust emission as a function of the impact of the orography.

  • View in gallery

    Distribution (%) of the orography variance, SSO variance (m2), over the whole modeled domain and for the erodible cells only.

  • View in gallery

    The k parameter values using the orographic weight factor proposed in this study, and for = 10 m s−1. Values are estimated for cells with an erodibility of 10% at least. The frames indicate the main mineral dust sources in Africa. The surface stations are indicated with their names.

  • View in gallery

    Hourly PM10 surface concentrations time series from 1 Jan to 30 Apr 2012, for the model (with the two simulations without and with the orography variance) and the SDT sites Banizoumbou, Cinzana, and Dakar.

  • View in gallery

    AOD (λ = 675 nm) time series from 1 Jan to 30 Apr 2012, for the model and the AERONET sites. Observations are represented by symbols, and the two plain curves represent the model simulation without and with orographic variability of the k parameter.

  • View in gallery

    AOD (λ = 600 nm) maps for 1200 UTC on 15 Jan, 15 Feb, 15 Mar, and 15 Apr 2012. (left) Absolute values of AOD for the J78 + Var_oro simulation and (right) AOD differences for J78 + Var_oro − J78.

  • View in gallery

    ASD compared between the AERONET inversions and the CHIMERE modeled concentrations.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 57 57 13
PDF Downloads 43 43 5

Modeling of Mineral Dust Emissions with a Weibull Wind Speed Distribution Including Subgrid-Scale Orography Variance

View More View Less
  • 1 Laboratoire de Météorologie Dynamique, Ecole Polytechnique, IPSL Research University, Ecole Normale Supérieure, Université Paris-Saclay, Sorbonne Universités, UPMC Univ. Paris 06, CNRS, Route de Saclay, Palaiseau, France
© Get Permissions
Full access

Abstract

The modeling of mineral dust emissions requires an extensive knowledge of the wind speed close to the surface. In regional and global models, Weibull distributions are often used to better represent the subgrid-scale variability of the wind speed. This distribution mainly depends on a k parameter, itself currently parameterized as a function of the wind speed value. In this study we propose to add the potential impact of the orography variance in the wind speed distribution by changing the k parameter value. Academic test cases are designed to estimate the parameters of the scheme. A realistic test case is performed over a large domain encompassing the northern part of Africa and Europe and for the period 1 January–1 May 2012. The results of the simulations are compared to particulate matter (PM10) surface concentrations and Aerosol Robotic Network (AERONET) aerosol optical depth and aerosol size distribution. We show that with the orography variance, the simulation results are closer to the ones without variance, showing that this additional variability is not the main driver of possible errors in mineral dust modeling.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Laurent Menut, menut@lmd.polytechnique.fr

Abstract

The modeling of mineral dust emissions requires an extensive knowledge of the wind speed close to the surface. In regional and global models, Weibull distributions are often used to better represent the subgrid-scale variability of the wind speed. This distribution mainly depends on a k parameter, itself currently parameterized as a function of the wind speed value. In this study we propose to add the potential impact of the orography variance in the wind speed distribution by changing the k parameter value. Academic test cases are designed to estimate the parameters of the scheme. A realistic test case is performed over a large domain encompassing the northern part of Africa and Europe and for the period 1 January–1 May 2012. The results of the simulations are compared to particulate matter (PM10) surface concentrations and Aerosol Robotic Network (AERONET) aerosol optical depth and aerosol size distribution. We show that with the orography variance, the simulation results are closer to the ones without variance, showing that this additional variability is not the main driver of possible errors in mineral dust modeling.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Laurent Menut, menut@lmd.polytechnique.fr

1. Introduction

For many geophysical studies, the wind speed close to the surface is a key parameter. This includes the analysis and modeling of mineral dust emissions. Primarily, the understanding of these emissions was performed using in situ or wind tunnel experiments (Gillette and Passi 1988; Shao et al. 1993). This means that the deduced parameterizations were first mainly representative of a local spatial area. Using these experimental studies, dust production models (DPM) were designed and later implemented in regional and global transport models. The transition between the local-scale and large-scale modeled fluxes is ensured by using tuning parameters. For example, Tegen and Fung (1994) uses a simple relationship for the vertical flux estimation, including a scaling factor; Marticorena et al. (1997) linked the horizontal (saltation) flux Fh to the vertical (sandblasting) flux Fυ using constant α factors, (with a dependency on the soil characteristics only). Other well-known and used DPMs, such as Alfaro and Gomes (2001), Shao (2001), and Kok et al. (2012), among others, are more realistic and calculate the vertical dust flux using more parameters. But even if they are more complex, they are also using tuning factors. Once the DPM scheme is included in transport models, simulations are most often validated with a derived budget but not direct measurements of emissions (as this is not possible to measure emissions out of a wind tunnel). The most employed proxy is the aerosol optical depth (AOD) derived from satellite data or photometers from the Aerosol Robotic Network (AERONET; Holben et al. 2001). More recently, other derived parameters are used, such as SEVIRI to estimate the hourly frequency of emissions (Schepanski et al. 2007). These AOD measurements are the results of vertically integrated concentrations, converted using uncertain optical properties of aerosols. It is therefore difficult to directly use these data to better understand or to constrain the surface emissions process. This explains why the DPM may vary a lot between models, leading to a large variability in the mineral dust emissions estimates (Huneeus et al. 2011; Cuevas et al. 2015).

To fill the gap between locally estimated emissions and a regional scale model, recent studies explored the way to describe the DPM’s input parameters with distributions in place of single mean values (Grini and Zender 2004; Pryor et al. 2005; Menut 2008). This corresponds to the way to represent variables also with their subgrid-scale variability. This variability may be taken into account for many parameters in the DPM: recently, Menut et al. (2013b) use spatially high-resolution databases of roughness lengths, soil, and surface parameters to represent the existing variability in a model grid cell representing several tens of squared kilometers.

This is also essential for the wind speed used in the emissions calculation. The mean wind speed used in a model often represents averaged values both in the temporal and spatial dimensions. But, as the mineral dust emissions is a threshold problem, the use of a single mean value may induce large modeling errors. For example, for particular soil conditions, emissions of dust may appear only up to 8.5 m s−1: if the “real” mean wind speed is 8 m s−1, then there is no emission at all. But if the “modeled” wind speed is 9 m s−1, then emissions are calculated. This model error is possible, since the wind speed model uncertainty is 1 m s1 in regional models (see Gómez-Navarro et al. 2015; among others). To avoid these possible large errors caused by the conjunction of wind speed uncertainty and threshold calculation using this wind speed, the DPM uses wind speed distributions, such as the Weibull distribution (Weibull 1951). These distributions represent the fact that the mean wind speed may be biased or uncertain, but it is also a way to represent the fact that during one hour and for a large grid cell, the instantaneous wind speed may vary a lot.

In this study the main goal is to improve the subgrid-scale variability of the mean wind speed by adding additional information in the k parameter estimation. This additional information is the orography variance of a model grid cell. The choice to add this specific variability was done because this information is available but not used in the model and because it seems logical that the surface wind speed is not spatially constant over surfaces of tens of squared kilometers. Then, in a model, when using a “mean” wind speed, it is clear that this value does not reflect completely all possible values observed in a grid cell, where the orography may vary.

By adding a relationship between k and the orography, we add realism to the simulation and we also expect to reduce the uncertainty linked to the model resolution (the more realistic subgrid-scale variability is taken into account, the less the results depend on the horizontal resolution). Using the WRF and CHIMERE models, simulations are done with and without the modification of k. Using comparisons to measurements, we quantify whether the modeled results are improved.

First, the data and tools used in this study are presented with the observations in section 2 and the models in section 3. The Weibull distribution is presented in section 4, and the proposed modified formulation of the shape parameter k is presented in section 5. Results with an academic test case are presented in section 6. Then, in section 7, the impact of this additional term is evaluated on a realistic test case representing the simulation of the period January–April 2012 over a large domain encompassing the northern Africa (including Sahara and Sahel) and the Mediterranean regions. Conclusions are presented in section 8.

2. The observations

In this study an academic and a real test case are presented. For the real test case, several observations are used: (i) the surface concentrations of particulate matter (PM10) as measured by the Sahelian Dust Transect (SDT); and (ii) the AERONET photometers’ measurements for the AOD, the angstrom exponent (Ae), and the aerosol size distribution (ASD). Finally, the statistical scores used to quantify the results are presented in this section.

a. Surface concentrations of PM10

For the surface concentrations of PM10 we use the SDT measurements. These measurements are performed in Banizoumbou (Niger), Cinzana (Mali), and M’Bour and Bambey (Senegal) as described in Marticorena et al. (2010). The concentrations are measured using a tapered element oscillating microbalance with a PM10 inlet. These microbalances are at the same place as the photometers of the AERONET.

b. Aerosol optical depth

The stations are selected to be representative of different locations and different proximities to mineral dust sources. Banizoumbou and Cinzana are located in the center of western Africa and also correspond to stations of the SDT. They are considered as locations close to the sources. Dakar, Senegal, is located close to the coast on the western side of West Africa and Izana (Tenerife Island) is located on the Canary Islands. Dakar and Izana are more representative of stations “under mineral dust plumes” after long-range transport.

The aerosol optical properties are compared between observations and model using the AERONET measurements (Holben et al. 2001). First, the comparison is done using the AOD and a wavelength of λ = 550 nm. Second, the comparison is performed using the ASD estimated after inversion of the photometers’ data as described in Dubovik and King (2000). The level 2 data are used. For each AERONET station used in this study and listed in Table 1, the inversion algorithm provides a volume particle size distribution for 15 bins, logarithmically distributed for a radius between 0.05 and 15 µm.

Table 1.

Names and locations of the AERONET stations used for model comparisons to AOD and ASD data. The stations are ordered from south to north. The altitude MSL is not presented, the measurements being representative of the vertically integrated atmospheric column above ground level (AGL).

Table 1.

c. Statistical scores

The statistics are calculated for temporal correlation (Rt), root-mean-square error (RMSE), and bias. The correlation used in this study is the Pearson’s correlation. Each correlation provides specific information on the quality of the simulation. The temporal correlation Rtis estimated station by station and uses daily averaged data in order to have homogeneous comparisons between all variables. This correlation is directly related to the variability from day to day, for each station. The observed value and the modeled value are at time t and for the station i, for a total of T days and I stations. The mean time-averaged value is defined as
e1
The temporal correlation for each station i is calculated as
e2
The mean used in this study is thus
e3
with I as the total number of stations. The normalized root-mean-square error is expressed as
e4
for all stations i and all times t.

3. The models

For the results calculations, several models are used. First, we describe the dust production model. Second, and for the real test case, we describe the Weather Research and Forecasting (WRF) Model, which calculates the meteorological fields; and the CHIMERE chemistry-transport model, which calculates the concentrations of mineral dust in the troposphere.

a. The dust production model

The mineral dust flux is estimated following two steps. First, the saltation flux is calculated with the scheme of White (1986):
e5
where is the “saltation” friction velocity, that is, the friction velocity calculated using the roughness length . The roughness length is estimated using the global 6-km horizontal resolution Global Aeolian Roughness Lengths from ASCAT and PARASOL (Polarization and Anisotropy of Reflectances for Atmospheric Sciences Coupled with Observations from a Lidar) dataset (Prigent et al. 2012). The acceleration of gravity, where g = 9.81 m s−2, and is the air density (depending on the meteorological model).
The threshold friction velocity depends on the soil particle diameter size and , and is calculated following the parameterization of Shao and Lu (2000):
e6
where the constant parameters are and γ = 300 kg m−2. The particle density = 2.65 × 103 kg m−3 is chosen to be representative of quartz grains clay minerals. The saltation flux is estimated only when for a given . One saltation flux is calculated for each grain size of the soil distribution and then all fluxes are integrated.
Second, the sandblasting flux is calculated following the Alfaro and Gomes (2001) scheme, which was modified in Menut et al. (2005). The calculation is based on the partitioning between the dust cohesion energy and the kinetic energy of each saltating aggregate. The emitted mineral dust is split into three aerosol modes (fine, coarse, and big). These modes are represented using lognormal distributions with diameters = 1.5 × 10−6 m, = 6.7 × 10−6 m, and = 14.2 × 10−6 m, and their associated standard deviations, = 1.7, = 1.6, and = 1.5, respectively. For each mode a kinetic energy is defined, and the flux is calculated as
e7
where is the number of intervals discretizing the soil size distribution in the range ; is the mean mass diameter; β is a constant as β = 16 300; and are factors depending on the soil size, after Alfaro and Gomes (2001). One specificity of this scheme is that, in addition to the mean flux value, the emission size distribution depends on the wind speed. The higher the wind speed, the finer the emitted particles.

The surface and soil databases are described in Menut et al. (2013b). The soil characteristics are from the State Soil Geographic database of the FAO (STATSGO-FAO; Nachtergaele et al. 2009), a global dataset with a native resolution of 30′ × 30′. For the surface characteristics, the NCAR USGS database is used, with the same resolution as the soil database (Loveland et al. 2000). For each model grid cell, we thus have the relative percentage of all soil and surface characteristics. For each of these percentages, a sandblasting flux is calculated and then cumulated to have the total sandblasting flux in each model grid cell.

Complementary to the soil and surface characteristics, an erobility factor is also necessary to estimate the relative part of the grid cell that could emit mineral dust. A global database was derived using MODIS satellite data to estimate this factor, as described in Beegum et al. (2016). A map of erodibility is presented in Fig. 1.

Fig. 1.
Fig. 1.

The domain defined for the real test case. (top) Map of the erodibility (0:1). (bottom) Map of the orography variance (m2).

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

This figure also shows the orography variance (m2) provided by the WRF Model database. This variance varies between 0 and 15 000 m2 on the domain. The most important variability is observed over mountainous areas, when the resolved topography corresponds to an average of very different altitudes above sea level.

b. The WRF and CHIMERE models

The WRF and CHIMERE models are used to calculate the concentration of aerosols in the troposphere. The system is an “offline” coupling model, meaning that the meteorological fields are first calculated with WRF and then these meteorological fields are used by CHIMERE for the transport.

WRF, version 3.6.1, is used (Skamarock et al. 2007). This regional modeling uses NCEP global meteorological fields using a spectral nudging to ensure that the large-scale circulation is well taken into account. The complete setup for this model is presented in Menut et al. (2016).

CHIMERE is used in its version described in Menut et al. (2013a) and includes the last updates presented in Mailler et al. (2017). For this study the gaseous and aerosol chemistry is not used. The only aerosol taken into account is mineral dust: emissions, transport, mixing, and wet and dry deposition. For all other parameters, we also used exactly the same configuration as in Menut et al. (2016). The modeled AOD is calculated by FastJX for several wavelengths (Wild et al. 2000). The boundary conditions for mineral dust are those of the climatology provided by GOCART (Ginoux et al. 2001) and described in Menut et al. (2013a).

To quantify the impact of our changes during a long period, the simulations range from 1 January to 1 May 2012. These four months correspond to the dry period in western Africa, when major mineral dust events occur. The modeled domain has a horizontal resolution of 60 km × 60 km, and represents a large domain encompassing the northern part of Africa and the Mediterranean Sea, as presented in Fig. 1. This domain is strictly the same for the WRF and CHIMERE models. The horizontal resolution was selected for different reasons: (i) this study is about subgrid-scale variability: with this resolution of tens of squared kilometers, the variability is nonnegligible; and (ii) for mineral dust studies, it is important to have a large modeled domain, encompassing all possible sources and long-range transport. With these simulations ranging over 4 months and a large domain, this resolution makes the computational cost manageable. Note that this resolution was used in previous studies, such as Menut et al. (2016), Briant et al. (2017), and Menut et al. (2017), and shows satisfactorily results compared to observations.

The main goal of this study is to change the wind speed. This wind speed is calculated by WRF and is then used by CHIMERE for several processes: the transport, the calculation of the friction velocity (for the dry deposition), and the mineral dust emissions. The use of the Weibull distribution is made only in CHIMERE and only for the mineral dust emissions. For all other processes, the mean wind speed remains unchanged and as it is provided by WRF.

4. The Weibull distribution

This section presents the main concept of the Weibull distribution, the parameters to define, and the way to estimate them.

a. The Weibull distribution formulation

The Weibull distribution is expressed as the following probability density function, following Babiker et al. (1987):
e8
where is the subgrid-scale wind speed value, is its probability density function, k is a dimensionless shape parameter, and λ (m s−1) is a scale parameter related to the mean of the distribution (in general, the mean wind speed value ).
The subgrid-scale wind speed is estimated from the grid-scale mean wind speed as
e9
where i ranges from 1 to , which is the number of subgrid-scale wind speed values defined to estimate accurately the distribution.
The λ value is estimated with the k parameter, which enables reducing the Weibull distribution uncertainty to the k parameter only. The expression of λ is
e10
where is the gamma function. The main problem for an accurate calculation of the Weibull distribution is thus focused on the accuracy of the k parameter.

b. Current estimate of k

The sensitivity to k is a well-known problem. Numerous studies were dedicated to its study and we present here only a few examples. Using measurements made at one site, Babiker et al. (1987) made a sensitivity study to quantify the impact of the mineral dust emissions fluxes to several k values ranging from 1 to 3. Also using measurements (30 years and 1432 different measurements sites in the United States), Gillette and Passi (1988) estimated k values between 0.5 and 3.86. Christofferson and Gillette (1987) proposed a methodology to estimate the k value from numerous wind speed observations. Pérez et al. (2007) used sodar data to retrieve wind speed values close to the surface: in this case, several k values are tested and compared to the observations, in a range from 1 to 4. Kelly et al. (2014) studied the potential variability of k and presented results with values ranging between 1.5 and 3. Gryning et al. (2016) presented a similar technique, but used mast and lidar measurements and concluded that k varies between 1.8 and 3.1 (depending on the altitude and the sites’ localization). He et al. (2010) used wind data from 720 stations (from 1979 to 1999) and showed that the k parameter could depend on the atmospheric stability: the Weibull distribution is narrower during the night, because of intermittent turbulence. Statistically, they showed that the k value could depend on the wind speed value with a larger distribution for weak winds (high wind variability) and narrower for strong winds (well-established wind speed regimes). The first studies showed that a constant value of k is not realistic and that this parameter has to be variable in space and time.

More specifically for the mineral dust emissions problem, some other studies were already dedicated to the best possible estimate of k. Menut (2008) used a constant value of k, limiting the variability. It was done to ensure a certain stability of the calculation in case of operational forecast calculations (i.e., no day-to-day tuning of the model). This approach has to be replaced in models in order to be more realistic. Ridley et al. (2013) made two simulations for the same domain and the same period: the first one, with a high resolution to explicitly calculate the wind speed variability; and the second one, with a coarse resolution, to test the deduced k parameters used with a Weibull distribution. The approach is realistic but valid mainly for one studied case. A change in the period of the domain requires replaying the two simulations, fine and coarse.

A relationship between k, , and its variance was presented by Justus et al. (1978, hereafter J78). This scheme is the only one we found to express the k value as a function of a meteorological parameter. This is why it was selected for this study. It is expressed as
e11
A direct relationship linking k and is used in Grini and Zender (2004) as follows:
e12

This expression does not take into account subgrid-scale variability, such as the orography. It considers that the Weibull shape mainly depends on the wind speed magnitude. This latter relation is used in many studies, such as Su and Toon (2009) with CAM3.0 applied over China; Zhang et al. (2016) with CAM5.0 and at the global scale (2° resolution); and Tegen et al. (2006) for the comparison to observed wind values in Bodele, north central Africa Note that in these studies, they are using a dust production model for which the size distribution is not dependent on the wind speed (unlike the model used in this study).

An example of wind speed distribution using k constant values (k = 3 and 4) and the expression of J78 is presented in Fig. 2. The distributions are calculated for = 8 and 14 m s−1. These values are just examples of several possible modeled wind speed values.

  • For = 8 m s−1, the distribution ranges from 0 to 15 m s−1. The J78 shape parameter value is k = 2.66.
  • For = 14 m s−1, the distribution ranges from 0 to 28 m s−1. The J78 shape parameter value is k = 3.52.
Fig. 2.
Fig. 2.

Example of Weibull wind speed distribution for a mean value of = 8 (black) and 14 (blue) m s−1. The three curves correspond to k = 3, k = 4, and the parameterization in J78, respectively. With J78, the corresponding k values are k = 2.66 for = 8 m s−1 and k = 3.52 for = 14 m s−1.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

Having a k value evolving with the mean wind speed value, it is interesting to notice that for = 8 m s−1, the J78 distribution has the largest width; when = 14 m s−1, the distribution has values between the distributions calculated with k = 3 and k = 4.

5. Adding the orography variability in k

The factor

The main goal of this study is to estimate a k value, taking into account the orography subgrid variability. This quantity represents a factor directly affecting the surface wind speed and is currently not taken into account in the models using this wind speed. For that, we thus propose to add a variability term to the parameterization of J78. The new k depending on the orography is thus expressed as
e13
where is a factor, using a simple function, respecting the following rules:
  • First, we have to ensure that k will evolve around its mean value and within an acceptable range. We thus define two constants, and , designed to act as limiters. They represent the minimum and maximum possible factors around the mean k value, respectively.
  • We consider that over flat terrain, the wind speed module would have a lower variability that over complex terrain with a changing orography. The value of = 1 represents a mean orography variance in the selected range of possible values of orography. For 1, we have k decreasing, representing a more broad wind speed distribution and thus an orography variance less important than the mean value: over flat terrains, the wind speed is more spatially homogeneous, because there is no orography.
  • The change is designed to act only over erodible surfaces, since this is applied only for mineral dust emissions. It is thus necessary to find the maximum of orography variance until the mineral dust emissions occur. The orography variance may be very large over mountainous areas, but these areas are nonerodible (being mainly constituted of rocks) and are thus not taken into account.
  • We also want to have a nonlinear dependency between k and the orography variance in order to have a smooth change between the minimum and maximum orography variance values.
These constraints lead us to define a simple formulation,
e14
where is the orography variance of the model cell where the dust flux is calculated; represents the maximum of orography variance for which we want to reduce the k value; and the a, b, and c values are just constants dedicated to modify the shape of the function. Here, we want to have a smooth function and we selected a = 20, b = 10, and c = 1. Some sensitivity tests were done, and results were not sensitive to a change in these values. A schematic representation of is presented in Fig. 3. For this study we selected an decrease/increase of ±20%, leading to define = 0.8 and = 1.2. These values were arbitrarily selected but represent “reasonable” variability of the k parameter, based on numerous studies having fitted this parameter using wind speed measurements.
Fig. 3.
Fig. 3.

The factor variability as a function of the orography variance.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

6. Academic test case modeling

To understand the variability of k as a function of the mean wind speed and various values of orography variances, a simple academic test case is performed. For this test case the DPM of Alfaro and Gomes (2001) is used alone, in a 0D context (meaning that there is no time or space, only a calculation using a varying academic wind speed value). To quantify the maximum of the impact of the factor, the Weibull distribution is calculated with three different hypothesis and the results are displayed in Fig. 4:

  • = 1 corresponding to the J78 scheme without any change.
  • with = 0.01 corresponding to a low orography variability. In this case we consider that the orography variance corresponds to 1% of the maximum orography variance.
  • with = corresponding to a maximum of orography variability (100%).
Fig. 4.
Fig. 4.

(top) Values of the k parameter values for the J78 parameterization and include the orographic weight factor proposed in this study. The two curves represent the extrema of the change, with and corresponding to the lowest and largest orography variances, respectively. (middle) Weibull distribution with and without taking into account orography variance. (bottom) Sensitivity of the mineral dust emission as a function of the impact of the orography.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

Results are first presented in Fig. 4 (top) for the evolution of k depending on . For = 1 (as noted by J78), the k value regularly increases when increases. For the two other cases, the increase has a similar trend, but the absolute values of k differ. For example, and for = 10 m s−1, we have k = 3 for the J78 scheme, but k 3.5 and 2.5 when the orography is low and high, respectively (sigma = 1% and sigma = 100%).

For the same three expressions of k, the results are presented as a Weibull distribution in Fig. 4 (middle). The distribution with the J78 scheme without any change is in between the two other distribution when is applied.

Finally, the impact on emissions fluxes are presented in Fig. 4 (bottom). The aerosol size distribution of the emissions flux is presented for = 10 m s−1. The corresponding integrated values are presented in Table 2. The flux calculated with the most important variability (then low k and sigma = 100%) shows the highest values. In Table 2, this corresponds to = 60.16 kg m−1 day−1 when = 53.91 kg m−1 day−1 when there is no variability because of orography. For a higher value of k, the distribution is narrower and this leads to less important fluxes: = 39.19 kg m−1 day−1.

Table 2.

Value of the vertical mineral dust flux calculated with the Alfaro and Gomes (2001) scheme and for = 10 m s−1; is in kg m−1 day−1.

Table 2.

Emissions are diagnosed for three main modes as defined previously. Here, they are expressed in radii, to be later compared with AERONET measurements. The highest impact on the fluxes appear for the fine mode of the distribution, rp = 0.75 µm. This mode corresponds to the optically efficient part of the size distribution. The change in k has thus an impact on the vertical flux but also on the deduced AOD. As presented in Table 2, with no variability, AOD = 0.92. If the orography variability is taken into account, then AOD = 0.54 and 1.10 for an orography variance of 1% and 100%, respectively. This huge impact, both on emissions fluxes and on AOD, shows the high sensitivity of the determination of the k parameter for mineral dust model studies.

7. Real test case

The two simulations, without and with the orography impact on the k parameter value, are called “J78” and “J78 + Var_oro,” respectively. The simulations contain only mineral dust as aerosol, and the results are compared to measurements performed in locations where mineral dust dominates the aerosol composition: the AERONET stations located in western Africa and the surface PM10 measurements of the Sahelian Transect.

a. Determination of the k parameter over the domain

The first step for the real test case is to calculate the parameters involved in Eq. (14). Among all parameters it is necessary to select a correct value for . To do this we calculate the distribution of the subgrid-scale orography (SSO) variance. This is calculated for the whole domain and for the erodible surfaces only. The result is displayed in Fig. 5. The SSO variability between 0 and 500 m2 represents occurrence between 80% and 90%. Between 500 and 1000 m2, it reduces to less than 10% for both cases. There is more low variability for the whole domain, probably because ocean model cells are taken into account. The values higher than 1000 m2 represent a few percent of the distribution. As we want to act mainly on the range where the variability is the most frequent over the erodible areas, we select = 1000 m2.

Fig. 5.
Fig. 5.

Distribution (%) of the orography variance, SSO variance (m2), over the whole modeled domain and for the erodible cells only.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

The results of the k parameter calculation using these criteria is presented in Fig. 6. Values are estimated for cells with an erodibility of 10% at least. For the other cells, the k parameter is defined in the model as described by J78, but it is not used for mineral dust calculation, because the cells are being considered as nonerodible. For a wind speed of 10 m s−1, the J78 scheme gives as result k = 2.97 (as presented in Table 2). Using the orography variance, the k values vary between 2.2 and 3.8. The lowest values are obtained where the orography is high: for example, to the north of the Bodele region and in the Hoggar. The highest values are defined where the orography is low, and these areas represents the majority of the erodible cells for the studied domain.

Fig. 6.
Fig. 6.

The k parameter values using the orographic weight factor proposed in this study, and for = 10 m s−1. Values are estimated for cells with an erodibility of 10% at least. The frames indicate the main mineral dust sources in Africa. The surface stations are indicated with their names.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

Globally, we probably will have mineral dust emissions, which is less important than using only the J78 parameterization. But, we expect to have a different spatial variability of the emissions, since the change in the k parameter depends on varying orography over the whole domain.

b. Results for surface concentrations of PM10

The PM10 surface concentrations are compared between the model and the observations and with an hourly time step. The results are first presented as time series in Fig. 7 and for the sites of Banizoumbou, Cinzana, and Dakar. The observations, represented as symbols, show a large time variability and the values range from 0 to 2000 µg m−3.

Fig. 7.
Fig. 7.

Hourly PM10 surface concentrations time series from 1 Jan to 30 Apr 2012, for the model (with the two simulations without and with the orography variance) and the SDT sites Banizoumbou, Cinzana, and Dakar.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

The results mainly show the high temporal variability of the measurements and the fact that the model is able to reproduce this variability. The J78 + Var_oro configuration provides a PM10 lower than the J78 configuration. This means that over the erodible regions leading to mineral dust measured at these sites, we have an orography variance less important than the mean average value and thus an increased k parameter, leading to a broader Weibull distribution and thus less mineral dust emissions.

The results are quantified as statistical scores in Table 3. The first part of this table corresponds to the model results for J78 and the second part to J78 + Var_oro. For the correlation and the RMSE, the best statistical scores are bolded. For the correlation and the RMSE, the results are better for J78 + Var_oro than for J78 (except for the correlation in Banizoumbou, but the difference is low and only 0.01). The most important benefit with the J78 + Var_oro configuration is for the RMSE. This is mainly because k is increased, leading to a sharper Weibull distribution and thus less variability in the mineral dust emissions. The model error is reduced by at least 50% for each of the three stations. This benefit being true for the three stations, and these three stations being not close but representative of different regions, we can expect that this J78 + Var_oro configuration improves the results in a realistic way.

Table 3.

Scores for the comparisons between observations (Sahelian Transect) and model (CHIMERE) for the surface concentrations of PM10. Results are presented with N (%) as the hourly mean available measurements for the period 1 Jan–1 May 2012, the mean values over the period [observations (Obs) and modeled (Mod) values], Rt, and the bias (model minus observations). Stations are sorted in increasing latitude, from south to north. Values are bolded for the simulation having the highest correlation or the lowest RMSE.

Table 3.

c. Aerosol optical depth time series

Results for the AOD comparisons with the AERONET data and the CHIMERE model outputs are displayed in Fig. 8 as time series, from 1 January to 30 April 2012.

Fig. 8.
Fig. 8.

AOD (λ = 675 nm) time series from 1 Jan to 30 Apr 2012, for the model and the AERONET sites. Observations are represented by symbols, and the two plain curves represent the model simulation without and with orographic variability of the k parameter.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

The time series show that several high peaks of mineral dust occurred during the four studied months. For all these peaks, the model is able to retrieve correctly their timing (the correct day and the duration) and their magnitude. A few peaks are missing (as in January in Banizoumbou or mid-April in Izana), but they are very peaked and probably correspond to local events, which is difficult to catch with the used model resolution, such as convective cold pools or sporadic wind speed extremes. In general, the model is able to estimate the background values and the major dust events. As for the PM10 time series, the J78 + Var_oro configuration simulates less AOD than the J78 configuration.

To quantify these results of AOD, statistical values are presented in Table 4. The simulation with the orography variance improves the model results compared to the observations—the RMSE is better for all stations—showing that the model error is globally reduced. For the correlation, the differences between the two simulations are less important, even if the best results are also mainly for the J78 + Var_oro version. The bias decreases with J78 + Var_oro, highlighting the results showed with the time series: for these stations, the AOD is generally lower than for the J78 version.

Table 4.

Scores for the comparisons between observations (AERONET) and model (CHIMERE) for the AOD. Results are presented with N (%) as the hourly mean available measurements for the period 1 Jan–1 May 2012, the mean values over the period [observations (Obs) and modeled (Mod) values], Rt, RMSE, and the bias (model minus observations). Values are bolded for the simulation having the highest correlation or the lowest RMSE.

Table 4.

d. Aerosol optical depth maps

As a complement to the time series and the statistical scores, maps of AOD values are presented in Fig. 9. The modeled period lasted 4 months and the maps represent 4 days during the period: 1200 UTC 15 January, 1200 UTC 15 February, 1200 UTC 15 March, and 1200 UTC 15 April 2012. Note that other maps were analyzed, and the discussion and conclusions are similar. The left column represents the AOD absolute values calculated with the J78 + Var_oro simulation and the right column corresponds to the difference of AOD calculated with AOD(J78 + Var_oro) − AOD(J78). The AOD highest values are located in different places according to the day. But for the 4 days, J78 + Var_oro corresponds always to a decrease in AOD, with the largest differences being collocated with the highest AOD values. It shows that the use of the orography variance leads to an increase in k and thus a decrease of emissions, thus AOD. It means that the change is mainly applied over erodible areas where the orography variance is low.

Fig. 9.
Fig. 9.

AOD (λ = 600 nm) maps for 1200 UTC on 15 Jan, 15 Feb, 15 Mar, and 15 Apr 2012. (left) Absolute values of AOD for the J78 + Var_oro simulation and (right) AOD differences for J78 + Var_oro − J78.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

For example, for 15 April 2012, two main plumes are visible: the first one in the center of Africa (over Niger and Chad), and the second one in Guinea and Senegal and transported toward the Atlantic sea. In these plumes the AOD values reach 2 at the maximum when the “background values” over Africa are between 0 and 1. For the AOD background values, the differences reach 0.2. In the plumes, where the AOD are maximum, the differences reach 0.8, corresponding to an important change in the AOD estimation.

e. Aerosol size distribution

As presented in Fig. 4, a change in the k parameter acts on the mineral dust emissions flux magnitude but also on its size distribution. To quantify the impact of the k changes on the size distribution, we compare here the model results to the AERONET inversion. In the AERONET database, note that all AOD measurements are not inverted and we had to find available data. Two comparisons are presented but many other size distributions were analyzed and the conclusions are similar. The results presented correspond to available data close to the results showed in Fig. 9, and correspond to 15 March and 15 April 2012. Comparisons are presented in Fig. 10. Note that, here, the distributions are expressed in radii (to be consistent with the raw data provided by AERONET).

Fig. 10.
Fig. 10.

ASD compared between the AERONET inversions and the CHIMERE modeled concentrations.

Citation: Journal of Atmospheric and Oceanic Technology 35, 6; 10.1175/JTECH-D-17-0173.1

For the 4 days and the four sites, the most important mode corresponds to the “coarse” mode with rp 3.35 µm. Less important, concentrations are visible for the “fine” mode, rp = 0.8 µm, with the model but not in the observations. The simulation using J78 + Var_oro is closer to the observations than the J78 simulation. If the coarse mode dominates the mass, then the fine mode explains the most important differences between the two simulations. And for the calculation of AOD as compared with AERONET (with λ = 550 nm), this is the most sensitive part of the size distribution. These figures show that the variability of k acts on the correct part of the size distribution and proves that the RMSE is highly improved for the correct reasons.

8. Conclusions

This study examined the impact of the wind speed Weibull distribution on the calculation of mineral dust emissions. This distribution depends mainly on one parameter, k, parameterized as a function of the mean wind speed. In this study we consider that for regional models, having a grid cell of tens of kilometers, the subgrid-scale orography may also be an important parameter to take into account in the estimation of k, since the orography has a direct impact on the value of the mean wind speed close to the surface.

Using the information of subgrid orography variance, we proposed an adjustment of the estimation of the k parameter. We first test this change with academic cases in order to estimate realistic boundaries to the potential evolution of the k value. Second, we applied this change in the framework of a realistic modeling: 4 months—from 1 January to 1 May 2012—and over a large domain, encompassing western Africa and Europe. The modeled results were compared to surface PM10 concentrations and AERONET aerosol optical depth and aerosol size distribution. Correlation, root-mean-square error, and bias were also calculated. We showed that for a large part of the comparisons between observation and model results, the correlation is slightly improved ( +0.01) and the RMSE is greatly improved ( −30%) both for mass and AOD. Over the studied period, a systematic decrease of AOD and PM10 is modeled with the changed k shape parameter. It means that, over erodible surfaces, k was increased. Considering the proposed function, it means we are in the range of “low” variability of orography.

Even if the RMSE is better, the correlation is not really improved. It showed that the subgrid-scale variability of orography is not the main driver for the mineral dust emissions. This is surprising because a significant impact was expected: the DPM schemes are very sensitive to the mean wind speed and the mean wind speed is sensitive to the orography close to the surface. The fact to have quantified this impact remains useful even if the impact intensity is lower than expected. For processes such as mineral dust emissions, the subgrid-scale variability remains a way to link a local phenomenon and an averaged model grid cell. It is a better approach than the model tuning, sometimes efficient, but not able to explain the physics and often limited to a specific model, with its own parameterizations and resolution. The model tuning is also systematic where the change proposed in this study varies in each model grid cell, adding realism to the system. A potential future direction could be the convective cold pools, which are known to increase mineral dust emissions very locally and are certainly missing in many regional models.

Acknowledgments

For the Sahelian Dust Transect data, we thank Bernadette Chatenet, the technical PI of the Sahelian stations from 2006 to 2012; Béatrice Marticorena and Jean-Louis Rajot, the scientific co-PIs; and the African technicians who manage the stations. We thank the principal investigators and their staff for establishing and maintaining the AERONET sites used in this study.

REFERENCES

  • Alfaro, S. C., and L. Gomes, 2001: Modeling mineral aerosol production by wind erosion: Emission intensities and aerosol size distribution in source areas. J. Geophys. Res., 106, 18 07518 084, https://doi.org/10.1029/2000JD900339.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Babiker, A., I. Eltayeb, and M. Hassan, 1987: A statistical model for horizontal mass flux of erodible soil. J. Geophys. Res., 92, 14 84514 849, https://doi.org/10.1029/JD092iD12p14845.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Beegum, S., I. Gherboudj, N. Chaouch, F. Couvidat, L. Menut, and H. Ghedira, 2016: Simulating aerosols over Arabian Peninsula with CHIMERE: Sensitivity to soil, surface parameters and anthropogenic emission inventories. Atmos. Environ., 128, 185197, https://doi.org/10.1016/j.atmosenv.2016.01.010.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Briant, R., P. Tuccella, A. Deroubaix, D. Khvorostyanov, L. Menut, S. Mailler, and S. Turquety, 2017: Aerosol–radiation interaction modelling using online coupling between the WRF 3.7.1 meteorological model and the CHIMERE 2016 chemistry-transport model, through the OASIS3-MCT coupler. Geosci. Model Dev., 10, 927944, https://doi.org/10.5194/gmd-10-927-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Christofferson, R., and D. Gillette, 1987: A simple estimator of the shape factor of the two-parameter Weibull distribution. J. Climate Appl. Meteor., 26, 323325, https://doi.org/10.1175/1520-0450(1987)026<0323:ASEOTS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cuevas, E., and et al. , 2015: The MACC-II 2007–2008 reanalysis: Atmospheric dust evaluation and characterization over northern Africa and the Middle East. Atmos. Chem. Phys., 15, 39914024, https://doi.org/10.5194/acp-15-3991-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dubovik, O., and M. D. King, 2000: A flexible inversion algorithm for retrieval of aerosol optical properties from Sun and sky radiance measurements. J. Geophys. Res., 105, 20 67320 696, https://doi.org/10.1029/2000JD900282.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gillette, D., and R. Passi, 1988: Modeling dust emissions caused by wind erosion. J. Geophys. Res., 93, 14 23314 242, https://doi.org/10.1029/JD093iD11p14233.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ginoux, P., M. Chin, I. Tegen, J. M. Prospero, B. Holben, O. Dubovik, and S. J. Lin, 2001: Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res., 106, 20 25520 273, https://doi.org/10.1029/2000JD000053.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gómez-Navarro, J. J., C. C. Raible, and S. Dierer, 2015: Sensitivity of the WRF model to PBL parametrisations and nesting techniques: Evaluation of wind storms over complex terrain. Geosci. Model Dev., 8, 33493363, https://doi.org/10.5194/gmd-8-3349-2015.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grini, A., and C. S. Zender, 2004: Roles of saltation, sandblasting, and wind speed variability on mineral dust aerosol size distribution during the Puerto Rican Dust Experiment (PRIDE). J. Geophys. Res., 109, D07202, https://doi.org/10.1029/2003JD004233.

    • Search Google Scholar
    • Export Citation
  • Gryning, S.-E., R. Floors, A. Peña, E. Batchvarova, and B. Brümmer, 2016: Weibull wind-speed distribution parameters derived from a combination of wind-lidar and tall-mast measurements over land, coastal and marine sites. Bound.-Layer Meteor., 159, 329348, https://doi.org/10.1007/s10546-015-0113-x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • He, Y., A. H. Monahan, C. G. Jones, A. Dai, S. Biner, D. Caya, and K. Winger, 2010: Probability distributions of land surface wind speeds over North America. J. Geophys. Res., 115, D04103, https://doi.org/10.1029/2008JD010708.

    • Search Google Scholar
    • Export Citation
  • Holben, B., and et al. , 2001: An emerging ground-based aerosol climatology: Aerosol optical depth from AERONET. J. Geophys. Res., 106, 12 06712 097, https://doi.org/10.1029/2001JD900014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huneeus, N., and et al. , 2011: Global dust model intercomparison in AeroCom phase I. Atmos. Chem. Phys., 11, 77817816, https://doi.org/10.5194/acp-11-7781-2011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Justus, C., W. Hargraves, and A. Mikhail, and D. Graber, 1978: Methods for estimating wind speed frequency distributions. J. Appl. Meteor., 17, 350353, https://doi.org/10.1175/1520-0450(1978)017<0350:MFEWSF>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kelly, M., I. Troen, and H. Jorgensen, 2014: Weibull-k revisited: “Tall” profiles and height variation of wind statistics. Bound.-Layer Meteor., 152, 107124, https://doi.org/10.1007/s10546-014-9915-5.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kok, J. F., E. J. R. Parteli, T. I. Michaels, and D. Bou Karam, 2012: The physics of wind-blown sand and dust. Rep. Prog. Phys., 75, 106901, https://doi.org/10.1088/0034-4885/75/10/106901.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Loveland, T. R., B. C. Reed, J. F. Brown, D. O. Ohlen, Z. Zhu, L. Yang, and J. W. Merchant, 2000: Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. Int. J. Remote Sens., 21, 13031330, https://doi.org/10.1080/014311600210191.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mailler, S., and et al. , 2017: CHIMERE-2017: From urban to hemispheric chemistry-transport modeling. Geosci. Model Dev., 10, 23972423, https://doi.org/10.5194/gmd-10-2397-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marticorena, B., G. Bergametti, B. Aumont, Y. Callot, C. N’Doume, and M. Legrand, 1997: Modeling the atmospheric dust cycle: 2. Simulations of Saharan dust sources. J. Geophys. Res., 102, 43874404, https://doi.org/10.1029/96JD02964.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marticorena, B., and et al. , 2010: Temporal variability of mineral dust concentrations over West Africa: Analyses of a pluriannual monitoring from the AMMA Sahelian Dust Transect. Atmos. Chem. Phys., 10, 88998915, https://doi.org/10.5194/acp-10-8899-2010.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., 2008: Sensitivity of hourly Saharan dust emissions to NCEP and ECMWF modelled wind speed. J. Geophys Res., 113, D16201, https://doi.org/10.1029/2007JD009522.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., C. Schmechtig, and B. Marticorena, 2005: Sensitivity of the sandblasting fluxes calculations to the soil size distribution accuracy. J. Atmos. Oceanic Technol., 22, 18751884, https://doi.org/10.1175/JTECH1825.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., and et al. , 2013a: CHIMERE 2013: A model for regional atmospheric composition modelling. Geosci. Model Dev., 6, 9811028, https://doi.org/10.5194/gmd-6-981-2013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., C. Pérez, K. Haustein, B. Bessagnet, C. Prigent, and S. Alfaro, 2013b: Relative impact of roughness and soil texture on mineral dust emission fluxes modeling. J. Geophys. Res. Atmos., 118, 65056520, https://doi.org/10.1002/jgrd.50313.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., G. Siour, S. Mailler, F. Couvidat, and B. Bessagnet, 2016: Observations and regional modeling of aerosol optical properties, speciation and size distribution over Northern Africa and western Europe. Atmos. Chem. Phys., 16, 12 96112 982, https://doi.org/10.5194/acp-16-12961-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Menut, L., S. Mailler, B. Bessagnet, G. Siour, A. Colette, F. Couvidat, and F. Meleux, 2017: An alternative way to evaluate chemistry-transport model variability. Geosci. Model Dev., 10, 11991208, https://doi.org/10.5194/gmd-10-1199-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nachtergaele, F., H. van Velthuizen, and L. Verelst, 2009: Harmonized World Soil Database. Version 1.1, FAO, IIASA, ISRIC, ISS-CAS, JRC, 38 pp.

  • Pérez, I. A., M. L. Sánchez, and M. A. García, 2007:Weibull wind speed distribution: Numerical considerations and use with sodar data. J. Geophys. Res., 112, D20112, https://doi.org/10.1029/2006JD008278.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Prigent, C., C. Jiménez, and J. Catherinot, 2012: Comparison of satellite microwave backscattering (ASCAT) and visible/near-infrared reflectances (PARASOL) for the estimation of aeolian aerodynamic roughness length in arid and semi-arid regions. Atmos. Meas. Tech., 5, 27032712, https://doi.org/10.5194/amt-5-2703-2012.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pryor, S. C., J. T. Schoof, and R. J. Barthelmie, 2005: Empirical downscaling of wind speed probability distributions. J. Geophys. Res., 110, D19109, https://doi.org/10.1029/2005JD005899.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ridley, D. A., C. L. Heald, J. R. Pierce, and M. J. Evans, 2013: Toward resolution-independent dust emissions in global models: Impacts on the seasonal and spatial distribution of dust. Geophys. Res. Lett., 40, 28732877, https://doi.org/10.1002/grl.50409.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schepanski, K., I. Tegen, B. Laurent, B. Heinold, and A. Macke, 2007: A new Saharan dust source activation frequency map derived from MSG-SEVIRI IR-channels. Geophys. Res. Lett., 34, L18803, https://doi.org/10.1029/2007GL030168.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shao, Y., 2001: A model for mineral dust erosion. J. Geophys. Res., 106, 20 23920 254, https://doi.org/10.1029/2001JD900171.

  • Shao, Y., and I. Lu, 2000: A simple expression for wind erosion threshold friction velocity. J. Geophys. Res., 105, 22 43722 443, https://doi.org/10.1029/2000JD900304.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shao, Y., M. R. Raupach, and P. A. Findlater, 1993: Effect of saltation bombardment on the entrainment of dust by wind. J. Geophys. Res., 98, 12 71912 726, https://doi.org/10.1029/93JD00396.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Skamarock, W., J. Klemp, J. Dudhia, D. Gill, D. Barker, W. Wang, and J. Powers, 2007: A description of the Advanced Research WRF version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp.

  • Su, L. and O. B. Toon, 2009: Numerical simulations of Asian dust storms using a coupled climate-aerosol microphysical model. J. Geophys. Res., 114, D14202, https://doi.org/10.1029/2008JD010956.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tegen, I., and I. Fung, 1994: Modeling of mineral dust in the atmosphere: Sources, transport, and optical thickness. J. Geophys. Res., 99, 22 89722 914, https://doi.org/10.1029/94JD01928.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tegen, I., B. Heinold, M. Todd, J. Helmert, R. Washington, and O. Dubovik, 2006: Modelling soil dust aerosol in the Bodélé depression during the BoDEx campaign. Atmos. Chem. Phys., 6, 43454359, https://doi.org/10.5194/acp-6-4345-2006.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weibull, W., 1951: A statistical distribution function of wide applicability. J. Appl. Mech., 18, 293297.

  • White, B. R., 1986: Particle dynamics in two-phase flows. Encyclopedia of Fluid Mechanics: Slurry Flow Technology. N. P. Cheremisinoff, Ed., Gulf Publishing, 239–282.

  • Wild, O., X. Zhu, and M. J. Prather, 2000: Fast-J: Accurate simulation of in- and below-cloud photolysis in tropospheric chemical models. J. Atmos. Chem., 37, 245282, https://doi.org/10.1023/A:1006415919030.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, K., C. Zhao, H. Wan, Y. Qian, R. C. Easter, S. J. Ghan, K. Sakaguchi, and X. Liu, 2016: Quantifying the impact of sub-grid surface wind variability on sea salt and dust emissions in CAM5. Geosci. Model Dev., 9, 607632, https://doi.org/10.5194/gmd-9-607-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
Save