An assessment of dust emission schemes in modeling east ...€¦ · An assessment of dust emission schemes in modeling east Asian dust storms T. L. Zhao,1,2 S. L. Gong,1,2 X. Y. Zhang,2

An assessment of dust emission schemes in modeling east Asian

dust storms

T. L. Zhao,1,2 S. L. Gong,1,2 X. Y. Zhang,2 A. Abdel-Mawgoud,3 and Y. P. Shao4

Received 25 December 2004; revised 28 July 2005; accepted 8 November 2005; published 15 March 2006.

[1] By implementing dust emission schemes developed by Marticorena and Bergametti(1995), Alfaro et al. (1997), Alfaro and Gomes (2001) (hereinafter referred to as MBA)and Shao (2001, 2004) into a regional climate model with a size-distributed active aerosolalgorithm, NARCM (Northern Aerosol Regional Climate Model), an assessment ofdust emission schemes in the simulation of east Asian dust storms for March 2002 wascarried out. Sensitivity of the parameters used for both the MBA and Shao schemes is firstanalyzed with a box version of the NARCM, where the wind erosion threshold frictionvelocities for both schemes are in good agreement for soil grain size range in diameterfrom 40 mm to 400 mm but differ for other size ranges. Although the impacts of clay,silt, loam and sand contents on vertical dust fluxes show a similar trend, their dependenceson friction velocity vary substantially as the correction factors in each scheme to thethreshold friction velocity, soil moisture and vegetation cover present a different degree ofimpact on vertical dust fluxes with wind friction velocity. One specific parameter, soilplastic pressure p, required by the Shao scheme varies between 103 Pa for loose surfacesand 105 Pa for hard crusted surfaces, which controls significantly emission flux. On thebasis of the comparison of dust emission with the MBA scheme in the box model, thesoil plastic pressure p applicable to Asian deserts for the Shao scheme is set to be 1000 Pafor sandy, 5000 Pa for loamy and silty and 10,000 Pa for clay soil in March 2002. In 3-Dsimulations, both schemes captured the dust mobilization episodes during this period ineast Asia and produced the similar spatial distributions of Asian dust column loading.Compared with the MBA scheme, the Shao scheme predicted much lower dust emissionand surface concentration in eastern Mongolia and eastern and central north China andslightly higher with some additional dust emission sources in north western China, easternKazakhstan and western Mongolia. The key parameters responsible for the differencesbetween the MBA and Shao emission schemes are the surface and soil-related factorsincluding soil moisture and vegetation coverage.

Citation: Zhao, T. L., S. L. Gong, X. Y. Zhang, A. Abdel-Mawgoud, and Y. P. Shao (2006), An assessment of dust emission schemes

in modeling east Asian dust storms, J. Geophys. Res., 111, D05S90, doi:10.1029/2004JD005746.

1. Introduction

[2] Because of its climatic, environmental and geochem-ical importance, many attempts have been made to simulatethe dust aerosol at a regional and global scale by usingmicrophysical, radiative transfer, chemical transport, weatherforecasting, and climate models [Ginoux et al., 2001; Gonget al., 2003b; Liu et al., 2003; Marticorena and Bergametti,

1995; Shao et al., 2003; Tegen and Fung, 1994; Uno et al.,2003; Zender et al., 2003]. A challenge in dust aerosolmodeling is to accurately parameterize the emission rate ofdust particles in all size ranges for natural surfaces on basisof the current understanding on the physical processesinvolved in this wind forced movement of soil dustparticles. Efforts have been made to develop dust emissionschemes such as by Alfaro et al. [1997], Alfaro and Gomes[2001], and Marticorena and Bergametti [1995] (hereinafterreferred to as MBA) and by Shao [2001, 2004]. It is widelyconsidered that the main mechanism for dust emission issaltation bombardment and aggregates disintegration con-trolled by two factors: surface wind speed and soil surfaceproperties [Shao, 2000]. The MBA and Shao emissionschemes adapted a different parameterization for themechanism and the influence factors and hence yieldeddifferent simulation results from their integrated modelingsystem. Although the recent achievements in soil dustmodeling are significant, especially for Asian dust stormsduring ACE-Asia (Aerosol Characterization Experiment),

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D05S90, doi:10.1029/2004JD005746, 2006

1Air Quality Research Branch, Meteorological Service of Canada,Toronto, Ontario, Canada.

2Centre for Atmosphere Watch and Services, Chinese Academy ofMeteorological Sciences, China Meteorological Administration, Beijing,China.

3Department of Chemistry, Atmospheric Science Group, University ofGothenburg, Gothenburg, Sweden.

4Department of Physics and Materials Science, City University of HongKong, Kowloon, Hong Kong.

Copyright 2006 by the American Geophysical Union.0148-0227/06/2004JD005746$09.00

D05S90 1 of 11

the capability for dust aerosol modeling and quantitativeprediction of dust storms remains quite limited [Sokolik etal., 2001]. During ACE-Asia, various models performedwell in terms of predicting the frequency and generallocation of the dust emissions associated with Asian duststorms. However, there were large differences in dustemissions simulated by the various models [Huebert etal., 2003]. To investigate and isolate the impacts of emissionschemes, the MBA and Shao dust emission schemes wereimplemented into a regional climate model with a size-distributed active aerosol algorithm: NARCM (NorthernAerosol Regional Climate Model) [Gong et al., 2003a,2003b] to simulate east Asian dust storms in March 2002.Focus in this paper is on how the different parameterizationsin the two schemes affect the simulations on Asian dustaerosols. The objectives are to improve the understandingon the main mechanism and the controlling factors forAsian dust emission and their parameterizations in Asiandust modeling and prediction.

2. Assessment of Parameterizations in DustEmission Schemes

[3] A dust emission scheme comprises three key compo-nents: (1) the threshold friction velocity u*t at which winderosion is initiated, (2) the horizontal (streamwise) andvertical dust emission flux and (3) the surface and soil-related factors influencing either the threshold frictionvelocity or the dust fluxes. With coupling the MBA andShao dust emission schemes into a box version of NARCM,an assessment of the parameterizations for the three com-ponents in each scheme is first carried out.

2.1. Threshold Friction Velocity

[4] It is generally recognized that soil dust particles aremobilized only for wind speed greater than a thresholdvalue [Marticorena and Bergametti, 1995; Shao and Lu,2000]. This threshold of wind speed depends on thethreshold friction velocity, u*t, at which the wind erosionis initiated. For a smooth surface, the threshold frictionvelocity in the Greeley-Iversen expression [Greeley andIversen, 1985] is introduced byMarticorena and Bergametti[1995] in the MBA scheme:

ut* dð Þ ¼0:129K

1:928Re0:092 � 1� �0:5 0:03 < Re � 10

0:129K 1� 0:0858 exp �0:0617 Re� 10ð Þ½ �f g Re > 10

8><>:

ð1Þ

where

Re ¼ a dð ÞXþ b a ¼ 1331 cm�X b ¼ 0:38 X ¼ 1:56

K ¼rpgdra

� �0:5

1þ 0:006

rpg dð Þ2:5

!0:5

with the gravity parameter g = 9.81 m.s–2, rp and ra beingparticle and air density.[5] Shao and Lu [2000] presented a simple expression for

u*t for spherical particles loosely spread over a dry bare

surface. The expression in Shao’s emission scheme forcalculating u*t is

ut* dð Þ ¼ AN dpgd þ g

rad

� �� �0:5

ð2Þ

with AN being around 0.0123, g being around 1.65*10�4kgs�2 and dp = rp/ra.[6] Figure 1 shows the comparison of the threshold

friction velocity u*t in the MBA and Shao schemes. Forthe soil grain size range 40 < d < 400 mm, u*t in bothschemes is in good agreement with the same minimal valuesbut differ somewhat for the other particle size ranges. Itshould be noted, that the differences of u*t between theMBA and Shao schemes increase as the particle sizedecreases.

2.2. Dust Emission Mechanism

[7] The main mechanism for dust emission is widelyconsidered to be saltation bombardment and aggregatesdisintegration. Both the MBA and Shao emission schemeshave taken into account of these mechanisms with theparameterizations for the calculation of horizontal (stream-wise) saltation flux Q and vertical dust emission rate F forvarious particle sizes.[8] The horizontal dust mass flux or streamwise saltation

flux Q describes the intensity of saltation. Q is calculated inboth schemes using the White [1979] sand flux equation:

Q dð Þ ¼ cragu*3 1þ Rð Þ 1� R2

� �ð3Þ

where Q is expressed as a function of friction velocity u*and threshold friction velocity u*t, c is a constant ofproportionality with a value of 2.6 and R = u*t /u*.In theMBA scheme an addition of u*, above the nonsaltatingwind friction velocity caused by saltating sand grains is alsoconsidered. The dust emission rate is usually defined to bevertical mass flux of dust particle at the surface. Theparameterization on dust emission rate is to establish arelationship between the vertical dust flux F and thestreamwise saltation flux Q with the energy-based approach

Figure 1. Relationship of threshold friction velocity toparticle diameter in the MBA and Shao emission schemes.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

2 of 11

D05S90

in the MBA scheme and the volume-removal basedapproach in the Shao scheme [Shao, 2000].[9] The energy-based parameterization in the MBA

scheme is derived on the basis of the energy balance of asaltating particle during the particle and surface collisionbetween the kinetic and binding energy. In the saltation andsandblasting process, the fine particles released either fromsaltating aggregates or from the surface depend on theindividual kinetic energy [Alfaro et al., 1998]. The massmedian diameters (di), standard deviations (si) for the lognormally size distributions and binding energies (ei) of threeaerosol populations (i = 1, 2, 3) are introduced in NARCMby considering the soil features in Asia, especially in China,and source region dust size distribution measurements[Gong et al., 2003b]. The kinetic energy flux dFkin(d) ofsaltating aggregates with diameter from d to d + Dd isproportional to the corresponding horizontal mass fluxdQ(d) [Alfaro et al., 1997; Alfaro and Gomes, 2001]:

dFkin dð Þ ¼ bdQ dð Þ

with b = 16,300 cm s�2. By defining the fraction pi = pi(d)from the binding energies (ei) and the individual kineticenergy of an aggregate [Alfaro et al., 1997], the fraction ofdFkin(d) that is available to release particles of the ithaerosol population is pibdQ(d) and its particle number fluxdNi(d) = bdQ(d)pi/ei. Consequently, the vertical massflux for the ith aerosol population with the aerosol particlesize di is

F dið Þ ¼ prpd3i =6

� �Ni ð4Þ

[10] In contrast to the energy based parameterization inthe MBA scheme, the Shao scheme proposed the volumeremoval–based parameterization, which estimates dustemission rate on the basis of volume removal caused bysaltating particles as they impact the surface [Lu and Shao,1999]. A simplification of dust emission equation [Shao,2004] in the Shao scheme is expressed as

F di; dð Þ ¼ cyhfi 1� gð Þ þ gsp �

1þ smð ÞgQ=u*2 ð5Þ

where cy is a dimensionless coefficient and g is a functionspecified as g = exp [�(u* � u*t )]; Q is the dust streamwisesaltation (horizontal mass) flux of aggregate size d; sp isfree dust to aggregated dust ratio and sm is the ratio betweenmass of impacting particle and mass ejected by bombard-ment. sm can be also written as

sm ¼ 12u*2rbp

1þ 14u*

ffiffiffiffirbp

r� �ð6Þ

where p is soil plastic pressure, and soil bulk density rb =1000 kg m�3.[11] The physical basis for both schemes is that vertical

dust emission rate F is proportional to saltation flux Q.However, the proportionality depends on the dust particle-binding energy ei in the MBA scheme and soil plasticpressure p in the Shao scheme. Uncertainties of dustemission rate were involved in specifying ei or p. With

the estimated binding energy by considering the soil fea-tures in the arid and semiarid regions in China [Zhang et al.,2003] NARCM reasonably modeled Asian soil dust duringACE-Asia 2001 and its 44-year climatology [Gong et al.,2003b; Zhang et al., 2003; Zhao et al., 2003]. The soilplastic pressure p, required by the Shao scheme variesbetween 103 Pa for loose sandy soils and 105 Pa for hardcrusted clay soils [Shao, 2004], which controls significantlydust emission flux (Figure 2). According to the comparisonof dust emission simulation from the binding energy-basedMBA scheme, the soil plastic pressure p applicable to Asiandeserts for the Shao scheme is set to be 1000 Pa for sandy,5000 Pa for loamy, silty and 10,000 Pa for clay soils in theNARCM modeling on east Asian dust storms (Figure 3).These p values are within the range of soil plastic pressuresuggested by the comparisons with several data sets pub-lished in the literature [Shao, 2004], so that both schemescould produce the comparable dust emission fluxes beforethe effects of surface and soil-related factors are considered(Figure 3).

2.3. Surface and Soil-Related Factors

[12] The surface and soil-related factors that stronglyaffect dust emission mainly include soil moisture, soiltexture and the presence of surface roughness elements withvegetation covers. A pragmatic approach to account fortheir impact on surface dust emission is through the cor-rection of the threshold friction velocity u*t (d), with thefollowing form

ut* d;l;wð Þ ¼ ut* dð Þfl lð Þfw wð Þ ð7Þ

The MBA and Shao schemes adapted different approachesto parameterize the roughness correction function fl for thefraction l of area covered by surface roughness elementsand the soil moisture correction function fw for the soilmoisture w.[13] Taking into account the effects of nonerodible ele-

ments in the grid by using a roughness length parameteri-zation, the MBA scheme estimates the roughness correction

Figure 2. Variation of dust emission flux with the soilplastic pressure required in the Shao scheme.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

3 of 11

D05S90

function with the following drag partitioning parameteriza-tion [Marticorena and Bergametti, 1995]:

fl ¼ 1� ln Zm=z0Sð Þln 0:35 10=z0Sð Þ0:8h i

24

35

8<:

9=;

�1

ð8Þ

where Zm (cm) is the initial roughness length of hetero-geneous land covers and z0S (�10�3 cm) is the localroughness length of the uncovered surface. The effectiveroughness length [Taylor, 1987] is introduced into the MBAscheme to define the reasonable overall roughness length Zmof heterogeneous land covers (roughness elements plusunderlying, uncovered surface) over a model grid inNARCM. An effective roughness length Zm from a spatialaverage of the logarithm of the local micrometeorologicalroughness length over a model grid square is given by

ln Zm ¼ ln Z0

where lnZ0 is a grid square average over all vegetationcovers and erodible deserts with their fractions. Givenroughness length Zo for each land use category of allvegetations according to the definition of Gong et al.[2003b], fl in the MBA scheme is a function of the coverfractions of vegetations and erodible deserts. The range ofestimated roughness length for east Asian deserts in theMBA scheme with the magnitude between 10�4 and 10�2 cmis consistent with the estimation of the roughness lengthmap derived from the satellite POLDER/ADEOS for the

same regions [Laurent et al., 2005] and also comparable tothe roughness length experimentally determined for aninterdunal area and desert flats in the Namibia desert (4 �10�3 cm and 4.2 � 10�2 cm) [Greeley et al., 1997].[14] In Shao’s scheme an alternative approach to drag

partitioning is used to represent the effect of roughnesselements in terms of frontal area index, which quantifies thedensity of roughness elements on the surface. The rough-ness correction function fl in Shao’s scheme is based on thefollowing expression [Raupach, 1992]:

fl lð Þ ¼ 1� mrsrlð Þ0:5 1þ mrbrlð Þ0:5 ð9Þ

where sr, the basal element area to frontal area ratio, is 1,mr, a tuning parameter to account for nonuniformity in thesurface stress, is 0.5 and br = 90 [Raupach et al., 1993]. Thefrontal area index l is estimated from the vegetation coverfraction ac in l = �Cl ln (1 � ac) with the empiricalcoefficient Cl = 0.35 for the roughness of stubble [Shao,2000].[15] The influences of above-parameterized fl in the

MBA and Shao schemes on dust emission rate F wereshown in Figure 4, considering the natural situations inAsian arid and semiarid regions sparsely covered with grass.The friction velocity u* of 30, 45 and 60 cm s�1 in Figure 4corresponds respectively with the measured surface windspeed between 9 and 15 m s�1 for dust storms and themaximal wind speed 22 m s�1 for the severe dust storms ineast Asia in March 2002 [Shao et al., 2003]. The differenceof F between both schemes is more obvious over the areas

Figure 3. Comparisons of vertical fluxes with change of friction velocity in (a) sandy, (b) loamy, (c) siltyand (d) clay soil between the MBA and Shao schemes.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

4 of 11

D05S90

with the partly vegetated cover of erodible surface at lowerwind speed (Figures 4a and 4b), indicating that within therange of surface wind speed for Asian dust storm the impactof the vegetation covers on dust emission is more sensitiveto the cover fractions of nonerodible surface in the MBAscheme. In these ranges of conditions, the MBA schememay produce little dust flux while the Shao scheme stillgenerates substantial amount of dust flux. This may nothave a great impact on the dust emission from the totallyerodible surface over the deserts, but it certainly willproduce the differences in the emission rate over the partlyvegetated area of deserts between the two schemes, whichwill result in a difference in the spatial distributions of dustemission sources in the model domain.

[16] The soil moisture correction function fw [Fecan etal., 1999] in the MBA scheme can be expressed as afunction of soil moisture and clay contents:

fw wð Þ ¼1 for w < w0

1þ 1:21 w� w0ð Þ0:68h i0:5

for w > w0

8><>: ð10Þ

where w and w0 are the ambient and threshold gravimetricsoil moisture with

w0 ¼ 0:0014 %clayð Þ2 þ 0:17 %clayð Þ:

Figure 4. Impacts of the cover fraction of erodible desertson vertical fluxes at the friction velocity of (a) 30 cm s�1,(b) 45 cm s�1 and (c) 60 cm s�1 in the MBA and Shaoemission schemes.

Figure 5. Impacts of soil moisture on vertical dustemission fluxes at the friction velocity of (a) 30 cm s�1,(b) 45 cm s�1 and (c) 60 cm s�1 in the MBA and Shaoemission schemes.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

5 of 11

D05S90

[17] Alternatively, fw in the Shao scheme is an empiricalfunction based on the observations during the wind tunnelexperiments [Shao et al., 1996] and can be written as

fw wð Þ ¼exp 22:7 *wð Þ for w � 0:03

exp 95:3 *w� 2:029ð Þ for w > 0:03

8<: ð11Þ

here fw is only a function of volumetric soil moisture w.Knowing the soil bulk density rb, the volumetric soilmoisture in (11) can be calculated from the gravimetric soilmoisture in (10). Assuming rb = 1000 kg m�3, the values ofgravimetric and volumetric soil moisture are identical.[18] To evaluate the sensitivity of the soil moisture cor-

rection function fw to dust emission, the comparisons weremade on the simulations of vertical dust flux with the changeof soil moisture w in Figure 5. The soil clay content asadditional parameter in the MBA scheme is set to zero in thedeserts for the comparison. As shown in Figure 5, it isphysically reasonable to have no dust fluxes for a baresurface with both schemes even for a very high wind friction

velocity due to the parameterization of soil moisture correc-tion function fw of equations (10) and (11) for the thresholdfriction velocity u*t (equation (7)). The high soil moisturesbring a large fw and a strong u*t . When a threshold frictionvelocity u*t corrected by equation (7) exceeds the actual windfriction velocity u*, no wind erosion is initiated for anysurfaces. Though both parameterizations presented thedescended trends in dust emission with the high soil mois-ture, the trends profoundly differ with the clear gaps in themaximal (threshold) soil moistures, below which winderosion is initiated, especially at the strong wind speeds(Figures 5b and 5c). This implies that in the most cases ofAsian dust storms, the Shao scheme would produce muchless dust emission than the MBA scheme because of thedifference in parameterizing the impact of soil moisture.

3. Comparison of NARCM Simulation onAsian Dust Storm

[19] NARCM is a modeling system in which the Cana-dian Regional Climate Model (RCM) is coupled with the

Figure 6. Total mass of modeled dust emission (ton.km�2) during 11 and 27 March 2002 with the(a) MBA and (b) Shao schemes in NARCM.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

6 of 11

D05S90

Canadian Aerosol Module: CAM [Gong et al., 2003a].RCM includes the physics package from the CanadianGlobal Climate Model [McFarlane et al., 1992], a semi-Lagrangian and semi-implicit transport scheme for dynam-ics and passive tracers [Robert et al., 1985] and CanadianLand Surface Scheme: CLASS [Verseghy, 1991] consider-ing three soil layers, a snow layer where applicable, and avegetative canopy treatment. The averaged volumetric liq-uid and frozen moisture contents are modeled for each soillayer as prognostic variables. The layer depths currentlyused are 0.10, 0.25 and 3.75 m. The soil moisture content inthe first model layer (0.10 m) was used to drive the soil dustemission scheme. NARCM possesses all the atmosphericaerosol processes: production, transport, growth, coagula-tion, dry and wet deposition and an explicit microphysicalcloud module to treat aerosol-cloud interactions. A size-segregated multicomponent aerosol mass conservationequation in CAM is expressed as follows [Gong et al.,2003a]:

@cip

@t¼

@cip

@t

����TRANSPORT

þ@cip

@t

����SOURCES

þ@cip

@t

����CLEAR AIR

þ@cip

@t

����DRY

þ@cip

@t

����IN-CLOUD

þ@cip

@t

����BELOW-CLOUDS

where the rate of change of mixing ratio of dry particle massconstituting p in a size range i has been divided into factorterms (or tendencies) for transport, sources, clear air, drydeposition, in-cloud and below-cloud processes. The trans-port includes resolved motion as well as subgrid turbulent

diffusion and convection. The sources include (1) surfaceemission rate of both natural and anthropogenic aerosols and(2) production of secondary aerosols (i.e., airborne aerosolmass-produced by chemical transformation of their precur-sors). The latter together with particle nucleation, condensa-tion and coagulation contribute to the clear-air processes. Drydeposition of gases and particles affects the ‘‘Dry’’ tendency.Scavenging in in-cloud and below-cloud processes isregarded as wet deposition of gases and particles.[20] To further investigate the sensitivity of dust storm

simulations to dust emission schemes, 3-D simulations withNARCM were conducted for east Asian dust storms inMarch 2002. The meteorological boundary and initial con-ditions for RCM are driven with the 6-hourly NCEPreanalyzed meteorological data for the period. NARCMruns on a stereographic projection with a horizontal resolu-tion of 45 km at 60N and 22 vertical levels on a Gal-Chenterrain following coordinate system from ground to about30 km. The integration time step was 20 min. Twelvediameter classes from 0.01 to 40.96 mm were used torepresent the size distribution of all aerosols. All atmosphericaerosol quantities including dust emission fluxes, concen-trations and deposition were calculated for each size bin. Thesize distributed dust emission schemes from MBA and Shaowere integrated to calculate the dust emission for the sourceterm in NARCM. The same data sets for the desert distribu-tion/texture and satellite derived land use/roughness lengthprovided a coherent input parameter set for both MBA andShao emission scheme arid and semiarid regions in east Asia.[21] The modeled dust emissions for a dust storm in

March 2002 with MBA and Shao emission schemes were

Figure 7. Simulated surface dust concentrations at (a) Beijing, (b) Changchun, (c) Hohhot and (d) Aksuduring the Asian dust storms between 11 and 27 March 2002 with the MBA and Shao emission schemesin NARCM.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

7 of 11

D05S90

compared in Figure 6. The dust emission sources for thisperiod with the MBA scheme distributed over desert areasin eastern Mongolia (source S3), eastern and central northChina (source S1 and S2), where the dust emission strengthswere much higher than those with the Shao scheme,especially in source S2, whose dust emission almost van-ished with the Shao scheme. However, the quantity ofmodeled dust emission with the Shao scheme is higher overthe arid and semiarid regions in north western China (sourceS4 and S5), eastern Kazakhstan and western Mongolia(source S6), where the additional dust emission sourcesS4 and S6 for March 2002 were simulated with the Shaoscheme. Figures 7a, 7b and 7c show the comparison of timeseries of soil dust concentrations simulated with bothschemes and EPM10 at the city stations in northern China.The equivalent PM10 (particulate matter <10 mm) concen-

trations (EPM10) were deduced from the AQI data thatinclude all particulate matter sources. This EPM mayoverestimate the dust concentration during nondust eventbut is a good indication of relative dust concentrationsacross China, especially during dust episodes. NARCMsimulations with both schemes captured the dust stormepisodes with a severe dust storm even around 20 March2002 (Figure 7). However, the simulated surface dustconcentrations with the Shao scheme were much lower ineastern and central north China (Figures 7a, 7b and 7c) andhigher in western China (Figure 7d) than those with theMBA scheme. Compared with the surface measurements ofheavy Asian dust event around 20 March 2002 [Han et al.,2004; Shao et al., 2003; Sugimoto et al., 2003], NARCMsimulation with the MBA scheme reproduced more realisticsurface dust concentrations in north China. The daily

Figure 8. Averaged distribution of atmospheric dust column loading (kg.km�2) on 19 March 2002modeled with the (a) MBA and (b) Shao schemes in NARCM.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

8 of 11

D05S90

analysis on atmospheric dust loading from 17 to 23 March2002 showed that the dust aerosol was transported fromemission sources eastward across the downwind areas innortheast Asia during this severe dust storm with both theMBA and Shao schemes. Figure 8 presented the dailyaveraged dust aerosol loading on 19 March 2002 over theNARCM model domain. Both schemes simulated the sim-ilar spatial distributions of Asian dust column loading, butthe magnitudes of atmospheric dust loading were also muchhigher with the MBA scheme, caused by the differences inthe modeled dust emissions between the MBA and Shaoschemes in Figure 9. The source S1, S2 and S3 in easternMongolia, eastern and central north China dominated thedifferences of dust emission between the MBA and Shaoschemes (Figure 9) and contributed most of atmospheric dustloading (Figure 8) and surface concentration (Figures 7a,7b and 7c) over the downwind areas for the Asian duststorm. Additionally, the source region S4, S5 and S6 werealso the important emission sources for the dust storm over

the downwind areas (Figure 7d) in the model domain fromthe simulation with the Shao scheme.[22] Key parameters responsible for those differences in

the modeled dust emissions between the MBA and Shaoemission schemes (Figure 6) were explored. As discussed insection 2, both schemes use the different parameterizationsin the calculation of threshold friction velocity, vertical dustemission flux, surface and soil-related factors, thus resultingin the differences in the dust emissions. From the evalua-tions on the impacts of the parameters such as soil texture,fraction of vegetation cover and soil moisture on dustemission with the variation of friction velocity (Figures 3,4 and 5), it is found that the parameterizations on theimpacts of soil moisture and fraction of vegetation coverare the dominant factors in producing the significant differ-ences in dust emission rate between two schemes. Figure 10shows the fraction of erodible desert cover and modeled soilmoisture for March 2002 in east Asia to discuss theirrelationship with the differences of modeled dust emissions.

Figure 9. Differences of averaged dust emission (ton.km�2) between the MBA and Shao schemes for19 and 20 March 2002.

Figure 10. Cover fraction of erodible deserts in east Asia (shaded contour areas) and the simulated soilmoisture averages (100*m

3 m�3) from NARCM for the period from 11 to 27 March 2002 (contour lines).

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

9 of 11

D05S90

The soil moisture exceeded 0.04 m3 m�3 over the most dustsource areas of S1, S2 and S3 in eastern Mongolia, easternand central north China (Figures 6 and 10). Under that soilmoisture the Shao scheme produced very little dust emis-sion, even at the high wind speed (Figures 5b and 5c). Thesurface moisture between 0.07 and 0.1 m3 m–3 in the regionS2 inhibited most of dust emission from the Shao scheme(Figures 6b and 10). The soil moisture and its parameteri-zation are of critical importance to the simulations on Asiandust storms. The vegetation cover and the parameterizationassociated with drag partitioning also play a key role in dustemission modeling. The higher dust emission sources mod-eled with the Shao scheme in north western China andwestern Mongolia (source region S4, S5 and S6 in Figure 6)spread over the partially vegetated areas with a smallercover fraction of deserts, mostly corresponding with therelative lower soil moisture there (Figure 10). This reflectsthe factor that the Shao scheme produces more dust emis-sion from deserts with the vegetation cover than the MBAscheme (Figure 4).

4. Summary

[23] Although the simulations of Asian dust storms fromvarious models coupled with the MBA and Shao schemesshowed the possibility to simulate and forecast Asian duststorms with reasonable success [Gong et al., 2003b; Shao etal., 2003], limitations and uncertainties remained for aquantitative dust modeling and prediction due to complexnature of the processes involved in the parameterizations ofdust emission and transport. An assessment of the twoemission schemes in NARCM found that the parameter-izations of surface and soil-related factors including soilmoisture and vegetation cover played a dominant role incausing the uncertainties and differences for quantitativemodeling of east Asian dust storms between the MBA andShao schemes.[24] It is impossible to accurately determine the magni-

tude of the forces acting on small particles. The uncertaintyin the prediction of threshold friction velocity consequentlybecomes larger as the particle become smaller [Shao andLu, 2000], as shown in the comparison of parameterizationsbetween the MBA and Shao schemes.[25] The MBA and Shao emission schemes for vertical

dust flux are energy-based and volume removal mecha-nisms, respectively. Proprietary parameters such as bindingenergy for dust grains in the MBA scheme and soil plasticpressure for the Shao scheme vary to a large degree, even fora given location and time. The difficulties in estimating thoseparameters have proved to cause the major uncertainties inthe quantitative modeling. The soil plastic pressure applica-ble to Asian deserts for the Shao scheme are chosen with1000 Pa for sandy, 5000 Pa for loamy, silty and 10,000 Pa forclay soil in the NARCM modeling. Observations for theseparameters in the source regions are urgently needed.[26] The drag partitioning approaches have been devel-

oped under large assumptions and described best the windtunnel data for situation of uniform roughness elements.There are nonerodible (vegetated) and erodible lands for thenature surface. It is very difficult to determine the frontalarea index accurately in practice for the drag partitioning inthe Shao scheme. An effective roughness length for drag

partitioning in the MBA scheme considered the differentroughness length for the various land use categories in thenonerodible lands and could form the roughness elementdensity of each category and the overall roughness length ina grid depending on the model resolution. High-resolutionsatellite observations of roughness length for the wholesimulation domain would provide a coherence data set tocalibrate the two parameterizations.[27] Being a quantity governed by atmospheric and

surface hydrological processes, soil moisture is difficult tomeasured or modeled with accuracy, especially over largedesert areas. In NARCM the observed and predicted valuesagree to within a factor of two while the spatial distributionsof arid and semiarid regions are reasonably well represented[Gong et al., 2003b]. Furthermore the physical aspectsabout the impact of soil moisture on dust emission are quitedifferently parameterized in the MBA and Shao schemes.Those all suggest that accurate measurement and modelingof the surface soil moisture and the improvements of land-surface parameterization are imperative in soil dust model-ing and prediction.[28] Because of the difference in the parameterizations for

the above mentioned processes, the NARCM simulationwith the MBA emission scheme yielded more reasonabledust surface concentrations for most of the east Asiandomain than with the Shao scheme except for the westernChina for March 2002.[29] Ultimately, any dust emission scheme should be

evaluated and calibrated with filed measurement data. Dustflux measurements of both horizontal and vertical move-ments are critically needed under various meteorologicaland surface conditions. Parameters in various schemes needto be assessed with such closure experimental data. Satelliteobservations of soil moisture and snow cover should beadapted in future dust emission estimates. By analyzing therelevance or the weakness of the different emission schemesthrough the comparisons of the simulations with the com-prehensive observations for a longer time period, theensemble forecast of Asian dust storms could be realizedin further studies.

[30] Acknowledgments. The authors wish to thank the CanadianFoundation for Climate and Atmospheric Sciences (CFCAS) for thefinancial support to carry out this research. The work was also supportedby the funds from MOST (G2000048703), Aerosol and Climate Projects ofCMA, NSF of China (90102017, 40121303, and 49825105) and CAS(KZCX2-305).

ReferencesAlfaro, S. C., and L. Gomes (2001), Modeling mineral aerosol productionby wind erosion: Emission intensities and aerosol size distribution insource areas, J. Geophys. Res., 106(D16), 18,075–18,084.

Alfaro, S. C., A. Gaudichet, L. Gomes, and M. Maille (1997), Modeling thesize distribution of a soil aerosol produced by sandblasting, J. Geophys.Res., 102, 11,239–11,249.

Alfaro, S. C., A. Gaudichet, L. Gomes, and M. Maille (1998), Mineralaerosol production by wind erosion: Aerosol particle sizes and bindingenergies, Geophys. Res. Lett., 25(7), 991–994.

Fecan, F., B. Marticorena, and G. Bergametti (1999), Parameterizationof the invrease of the aeolian erosion threshold wind friction velocitydue to soil moisture for arid and semi-arid areas, Ann. Geophys., 17,149–157.

Ginoux, P., M. Chin, I. Tegen, J. Prospero, B. N. Holben, O. Dubovik, andS.-J. Lin (2001), Sources and distributions of dust aerosols simulatedwith the GOCART model, J. Geophys. Res., 106(D17), 20,255–20,274.

Gong, S. L., et al. (2003a), Canadian Aerosol Module: A size-segregatedsimulation of atmospheric aerosol processes for climate and air quality

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

10 of 11

D05S90

models: 1. Module development, J. Geophys. Res., 108(D1), 4007,doi:10.1029/2001JD002002.

Gong, S. L., X. Y. Zhang, T. L. Zhao, I. G. McKendry, D. A. Jaffe, andN. M. Lu (2003b), Characterization of soil dust distributions in Chinaand its transport during ACE-ASIA: 2. Model simulation and validation,J. Geophys. Res., 108(D9), 4262, doi:10.1029/2002JD002633.

Greeley, R., and J. D. Iversen (1985), Wind as a Geological Process onEarth, Mars, Venus and Titan, Cambridge Univ. Press, New York.

Greeley, R., D. G. Blumberg, J. F. McHone, A. Dobrovolski, J. Iversen,N. Lancaster, K. R. Rasmussen, S. Wall, and B. White (1997), Applica-tions of spaceborne radar laboratory data to the study of aeolian pro-cesses, J. Geophys. Res., 102, 10,971–10,983.

Han, Z. W., H. Ueda, K. Matsuda, R. J. Zhang, K. Arao, Y. Kanai, andH. Hasome (2004), Model study on particle size segregation and deposi-tion during Asian dust events in March 2002, J. Geophys. Res., 109,D19205, doi:10.1029/2004JD004920.

Huebert, B. J., T. Bates, P. B. Russell, G. Shi, Y. J. Kim, K. Kawamura,G. Carmicheal, and T. Nakajima (2003), An overview of ACE-Asia:Strategies for quantifying the relationships between Asian aerosols andtheir climatic impacts, J. Geophys Res., 108(D23), 8633, doi:1029/2003JD003550.

Laurent, B., B. Marticorena, G. Bergametti, P. Chazette, F. Maignan, andC. Schmechtig (2005), Simulation of the mineral dust emission frequen-cies from desert areas of China and Mongolia using an aerodynamicroughness length map derived from the POLDER/ADEOS 1 surfaceproducts, J. Geophys. Res., 110, D18S04, doi:10.1029/2004JD005013.

Liu, M., D. Westphal, S. Wang, A. Shimizu, N. Sugimoto, J. Zhou, andY. Chen (2003), A high-resolution numerical study of the Asian duststorms of April 2001, J. Geophys. Res., 108(D23), 8653, doi:10.1029/2002JD003178.

Lu, H., and Y. Shao (1999), A new model for dust emission b saltationbombardment, J. Geophys. Res., 104(D14), 16,827–16,842.

Marticorena, B., and G. Bergametti (1995), Modeling the atmospheric dustcycle: 1. Design of a soil-derived dust emission scheme, J. Geophys. Res.,100, 16,415–16,430.

McFarlane, N. A., G. J. Boer, J.-P. Blanchet, and M. Lazare (1992), TheCanadian Climate Centre second-generation general climate model andits equilibrium climate, J. Clim., 5, 1013–1044.

Raupach, M. R. (1992), Drag and drag partition on rough surfaces, Bound-ary Layer Meteorol., 60, 375–395.

Raupach, M. R., D. A. Gillette, and J. F. Leys (1993), The effect ofroughness elements on wind erosion thresholds, J. Geophys. Res., 98,3023–3029.

Robert, A., T. L. Yee, and H. Ritchie (1985), A semi-Lagrangian and semi-implicit numerical integration scheme for multilevel atmospheric models,Mon. Weather Rev., 133, 388–394.

Shao, Y. (2000), Physics and Modelling of Wind Erosion, Springer, NewYork.

Shao, Y. (2001), A model for mineral dust emission, J. Geophys. Res., 106,20,239–20,254.

Shao, Y. (2004), Simplification of a dust emission scheme and comparisonwith data, J. Geophys. Res., 109, D10202, doi:10.1029/2003JD004372.

Shao, Y., and H. Lu (2000), A simple expression for wind ersion thresholdfriction velocity, J. Geophy. Res., 105(D17), 22,437–22,443.

Shao, Y., M. R. Raupach, and J. F. Leys (1996), A model for predictingaeolian sand drift and dust entrainment on scales from paddock to region,Aust. J. Soil Res., 34, 309–342.

Shao, Y., et al. (2003), Northeast Asian dust storms: Real-time numericalprediction and validation, J. Geophys. Res., 108(D22), 4691,doi:10.1029/2003JD003667.

Sokolik, I. N., D. Winker, G. Bergametti, D. Gillette, G. Carmichael,Y. Kaufman, L. Gomes, L. Schuetz, and J. Penner (2001), Introductionto special section: Outstanding problems in quantifying the radiativeimpacts of mineral dust, J. Geophys. Res., 106(D16), 18,015–18,027.

Sugimoto, N., I. Uno, M. Nishikawa, A. Shimizu, I. Matsui, X. H. Dong,Y. Chen, and H. Quan (2003), Record heavy Asian dust in Beijing in2002: Observations and model analysis of recent events, Geophys. Res.Lett., 30(12), 1640, doi:10.1029/2002GL016349.

Taylor, P. (1987), Comments and further analysis on the effective roughnesslength for use in numerical three dimensional models, Boundary LayerMeteorol., 39, 403–419.

Tegen, I., and I. Fung (1994), Modeling of mineral dust in the atmosphere:Sources, transport, and optical thickness, J. Geophys. Res., 99, 22,897–22,914.

Uno, I., et al. (2003), Regional chemical weather forecasting systemCFORS: Model descriptions and analysis of surface observations atJapanese island stations during the ACE-Asia experiment, J. Geophys.Res., 108(D23), 8668, doi:10.1029/2002JD002845.

Verseghy, D. L. (1991), CLASS—A Canadian land surface scheme forGCMS. I: Soil model, Int. J. Climatol., 11(2), 111–133.

White, B. (1979), Soil transport by wind on Mars, J. Geophys. Res., 84,4643–4651.

Zender, C., H. Bian, and D. Newman (2003), Mineral Dust Entrainment andDeposition (DEAD) model: Description and 1990s dust climatology,J. Geophys. Res., 108(D14), 4416, doi:10.1029/2002JD002775.

Zhang, X. Y., S. L. Gong, Z. X. Shen, F. M. Mei, X. X. Xi, L. C. Liu, Z. J.Zhou, D. Wang, Y. Q. Wang, and Y. Cheng (2003), Characterizationof soil dust distributions in China and its transport during ACE-ASIA:1. Network measurements, J. Geophys. Res., 108(D9), 4261,doi:10.1029/2002JD002632.

Zhao, T. L., S. L. Gong, X. Y. Zhang, and I. G. McKendry (2003),Modelled size-segregated wet and dry deposition budgets of soil dustaerosol during ACE-Asia, 2001: Implications for trans-Pacific transport,J. Geophys. Res., 108(D23), 8665, doi:10.1029/2002JD003363.

�����������������������A. Abdel-Mawgoud, Department of Chemistry, Atmospheric Science

Group, University of Gothenburg, SE-412 96 Gothenburg, Sweden.S. L. Gong and T. L. Zhao, Air Quality Research Branch, Meteorological

Service of Canada, 4905 Dufferin Street, Toronto, ON, Canada M3H 5T4.([email protected])Y. P. Shao, Department of Physics and Materials Science, City University

of Hong Kong, Kowloon, Hong Kong.X. Y. Zhang, Centre for Atmosphere Watch and Services, Chinese

Academy of Meteorological Sciences, China Meteorological Administra-tion, Beijing 100081, China.

D05S90 ZHAO ET AL.: ASSESSMENT OF DUST EMISSION SCHEMES

11 of 11

D05S90