Abstract

The atmospheric component of the United Kingdom’s new High-resolution Global Environmental Model (HiGEM) has been run with interactive aerosol schemes that include biomass burning and mineral dust. Dust emission, transport, and deposition are parameterized within the model using six particle size divisions, which are treated independently. The biomass is modeled in three nonindependent modes, and emissions are prescribed from an external dataset. The model is shown to produce realistic horizontal and vertical distributions of these aerosols for each season when compared with available satellite- and ground-based observations and with other models. Combined aerosol optical depths off the coast of North Africa exceed 0.5 both in boreal winter, when biomass is the main contributor, and also in summer, when the dust dominates. The model is capable of resolving smaller-scale features, such as dust storms emanating from the Bodélé and Saharan regions of North Africa and the wintertime Bodélé low-level jet. This is illustrated by February and July case studies, in which the diurnal cycles of model variables in relation to dust emission and transport are examined. The top-of-atmosphere annual mean radiative forcing of the dust is calculated and found to be globally quite small but locally very large, exceeding 20 W m−2 over the Sahara, where inclusion of dust aerosol is shown to improve the model radiative balance. This work extends previous aerosol studies by combining complexity with increased global resolution and represents a step toward the next generation of models to investigate aerosol–climate interactions.

1. Introduction

Accurate modeling of mineral dust is known to be important because of its radiative impact in both numerical weather prediction models (Milton et al. 2008; Haywood et al. 2005) and climate models (Miller et al. 2004; Yoshioka et al. 2007). The emission, transport, and deposition of dust depend on physical processes on a wide range of scales; representation of the dust cycle is particularly difficult in atmospheric general circulation models (AGCMs), whose resolution may be too coarse to produce the small-scale wind flows that give rise to much of the world’s dust production (Miller et al. 2006; Engelstaedter and Washington 2007). In particular, observation campaigns over Africa (the world’s main dust source) have shown that the orography is crucial in funneling the surface winds around the Hoggar, Tibesti, and Aïr mountain ranges to produce dust from the Bodélé Depression in boreal winter and spring (Washington et al. 2006; Washington and Todd 2005) and from Mali and Algeria in summer (Flamant et al. 2007). Washington et al. (2006) comment that the representation of the observed Bodélé low-level jet is a particular challenge for climate models.

AGCMs with embedded mineral dust cycles are computationally expensive to run because of the complexity of the aerosol processes, the storage overheads of modeling several size bins (typically at least 4), and the necessity for many of the models to run for many years or decades to capture the interannual variability of the dust production. As a consequence, some groups have used high-resolution regional models to carry out in-depth case studies of particular dust events, other groups have used offline dust transport models driven by global model wind fields or reanalysis data, whereas others have used aerosol models embedded within relatively low-resolution AGCMs. Each of these approaches has its advantages and can provide valuable information about modeling strategies and methods of validation, but they also have clear limitations. For example, Tegen et al. (2006) presented regional model simulations of dust emission events during the Bodélé Dust Experiment (BoDEX; March 2005). They found that events could be reasonably well reproduced, though peak model wind speeds were lower than those observed, even with a horizontal resolution of 7 km. Lunt and Valdes (2002) used an offline dust model, driven by a slab ocean version of the Hadley Centre GCM (horizontal resolution 3.75° × 2.5°) to test the validity of its global large-scale dust simulations, and by the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) fields (resolution 2.8° × 2.8°) to test the model on Asian and African dust plume case studies. They found that Southern Hemisphere and Asian dust sources were well simulated, but that western Saharan sources and dust transport across the North Atlantic were not well handled. This was attributed partly to the subgrid-scale nature of dust “hot spots,” such as the Bodélé, and also to the inability of the model to produce sufficiently strong surface winds, despite the inclusion of a parameterization of wind gustiness. Miller et al. (2006) obtained an improved agreement of dust distributions with observations when using an embedded dust aerosol scheme in the Goddard Institute for Space Studies (GISS) AGCM, rather than an offline tracer transport model. However, with a horizontal resolution of 4° × 5°, the introduction of a subgrid distribution of surface wind speed was necessary to improve the summertime dust emissions over hot deserts associated with dry convective eddies.

The aim of this work is to attempt to circumvent some of these problems by embedding an interactive mineral dust scheme in a high-resolution AGCM, thus narrowing the gap between high-resolution regional models and lower-resolution global climate models. We investigate the ability of the model to represent the large-scale transport of dust and the smaller-scale features such as dust storms, and we evaluate the direct radiative forcing resulting from the dust, because these are important tests of the model’s credibility for use in climate impact studies. Although the validation of dust models has hitherto been difficult because of the high spatial and temporal variability of dust and paucity of observations (Duce 1995), some recent campaigns have provided information about the composition and horizontal and vertical distributions of dust and biomass aerosols over Africa [e.g., BoDEX, see Washington et al. (2006), and the Dust and Biomass Experiment (DABEX), see Johnson et al. (2008)]. Satellite data have also provided improved estimates of outgoing longwave radiation (OLR; Comer et al. 2007), dust aerosol sources (Schepanski et al. 2007), and aerosol optical thickness (AOT; Hsu et al. 2004) at increasingly high spatiotemporal resolution, which can be compared with equivalent model diagnostics.

The structure of this paper is as follows: in section 2, brief descriptions of high-resolution global atmospheric model (HiGAM) and the dust model are provided and the experimental design is described. Section 3 presents the results, including validation of the model dust and biomass aerosol burdens and spatial distributions, detailed analysis of the representation of February and July Saharan dust storms in the model, evaluation of the model over Africa against satellite observations, and discussion of the radiative forcing resulting from dust. Section 4 provides a summary and discusses future work.

2. Model description and experiment design

High-resolution Global Environmental Model (HiGEM) is a higher-resolution version of the Hadley Centre Global Environmental Model (HadGEM), which was built and run by the HiGEM consortium and is fully described in Shaffrey et al. (2009). In this work, the atmospheric component was run (HiGAM), forced by time-varying sea surface temperatures (SSTs) taken from the Atmospheric Model Intercomparison Project version II (AMIP-II) dataset (available online at http://www-pcmdi.llnl.gov/projects/amip), with the addition of an adapted version (see later in the study) of the interactive dust model described in Woodward (2001). HiGAM is a nonhydrostatic grid point model with horizontal resolution of 1.25° longitude × 0.83° latitude and 38 levels in the vertical. It has a semi-Lagrangian advection scheme, prognostic cloud physics, and shallow and deep convection parameterizations. There is a land surface exchange scheme with boundary layer mixing of surface fluxes, and the radiation scheme is the two-stream Edwards–Slingo code (Edwards and Slingo 1996). In addition to the dust component added in this work, sulfate, black carbon, and biomass burning aerosols are modeled interactively in HiGEM. These are subject to advection, mixing by turbulence in the boundary layer and by convection, and dry and wet deposition, as well as their own aerosol “chemistry” (which models processes that cause changes in particle sizes). These aerosols feed back on the model via their direct and indirect radiative effects, by which they scatter and absorb radiation and also affect cloud albedo and lifetimes (Forster et al. 2007).

The dust model used here was designed for the lower-resolution Hadley Centre Atmosphere Model, version 3 (HadAM3; 3.75° × 2.5°, 19 levels in the vertical), and it required some tuning to produce realistic global dust burdens at a higher resolution (as was found by Milton et al. 2008). This is to be expected because the dust is produced internally by the model, and the variables on which the dust emission depends (chiefly surface wind and soil moisture) are likely to have more extreme values when grid box sizes are smaller [see, e.g., the method of Werner et al. (2002) for systematically increasing the surface wind speeds in their lower-resolution model]. The dust scheme used here has six particle size bins in the range of 0.03–30-μm radius, covering clay (0–1 μm), silt (1–25 μm), and sand (25–1000 μm; note that larger sand particles are too big to be mobilized as dust aerosol). A dust parent soil ancillary file (interpolated from Wilson and Henderson-Sellers 1985) determines the fractions and relative masses of clay, silt, and sand available at the surface. Dust is mobilized when the friction velocity exceeds a threshold value U*t, which is a function of the representative particle diameter Drep (m) and soil moisture content W (kg m−2, for top 10-cm soil layer) given by

 
formula

where A = −0.2, and B and C are constants determined empirically (by the trial and error method). The quantity A log10(Drep) + C may be considered as the dry soil threshold friction velocity and it is a straight-line fit to the relevant part of the U*t versus particle diameter curve from Bagnold (1941), with the C term also including a correction to account for the difference between point observations and grid box mean values. The soil moisture factor (B) includes a correction to account for the difference between soil moisture at the surface and in the top 10-cm soil layer of the model (see Woodward 2001). The magnitude of the dust flux is then a function of the cube of the friction and threshold friction velocities; therefore, the dust burden in the model is very sensitive to the choice of the values of B and C in Eq. (1). The values used in Woodward (2001) for the lower-resolution HadAM3 (B = 0.5, C = −1.2) produced very high dust burdens when used in HiGAM; therefore, some short (3 year) sensitivity experiments were carried out where it was found that setting B = 0.15 and C = −0.7 gave more realistic dust loadings when compared with the available data and other models. These changes act to increase the threshold friction velocity, which is consistent with HiGAM having higher grid box mean surface wind speeds than the lower-resolution HadAM3. The following aspects of the dust distribution were taken into account when selecting these parameters: 1) the seasonal and annual mean global dust loadings and distributions, 2) the regional horizontal and vertical distributions over desert areas and “background” loading in remote areas, and 3) the relative proportions of total dust mass in each size bin and the associated aerosol optical thickness distribution. Of course, tuning in this way affects only the amount of dust emitted at the surface and does not change other model processes such as the dust deposition and transport; however, this is justified because other aerosols in the model (sulfate, black carbon, and biomass burning) share these processes and are found to be robust to the increase in resolution, as detailed in Shaffrey et al. (2009). The change to a less negative value of C is consistent with Milton et al. (2008), who found that a value of C = −0.15 was appropriate when using a modified version of this dust scheme in the Met Office global numerical weather prediction (NWP) model, which has an even higher resolution (0.5625° longitude × 0.375° latitude) than HiGAM.

In contrast to the dust, biomass emissions were monthly mean values externally imposed from a dataset supplied by T. Nozawa (Met Office 2003, personal communication), intended to represent both forest fires and agricultural burning for the present day (year 2000). They are injected into the model at the surface (agricultural emissions) and between levels 7 and 10 (1–2 km) to represent hot lofted emissions from forest fires. Biomass aerosol is represented in three modes: freshly emitted, aged (which is more hygroscopic), and in cloud (incorporated into cloud droplets). The modes are subject to transport and deposition, and the scheme is unchanged from that described in Martin et al. (2006) for HadGEM.

Initially, two HiGAM experiments were started from the model year 1982 (from a previously spunup initial field): one with dust acting as a passive tracer (called PasD1) and the other (ActD1) with dust feeding back on the model via its direct radiative effect (scattering and/or absorbing solar and thermal radiation). The radiative properties of dust are modeled by assuming that the particles are spherical with refractive indices derived from various sources representative of dust from different locations (Woodward 2001). Recent evidence (e.g., Balkanski et al. 2007) suggests that these values may be too high in the visible for Saharan dust, causing it to be too absorbing. Although extensive experimentation with the spectral properties was outside the scope of this work, some experiments using revised values from Balkanski et al. (2007) in the lower-resolution HadGEM2 showed that this produced a reduction of 14% in the global mean AOT and 12% over the Sahara. Although this is not negligible, these reductions are fairly small relative to other uncertainties in dust modeling. The single scattering albedos and extinction coefficients at the 550-nm wavelength, used in the model for dust and dry biomass aerosol, are shown in Table 1.

Table 1.

Particle sizes, single scattering albedos, and extinction coefficients for dust and biomass (assuming dry aerosol) in HiGAM.

Particle sizes, single scattering albedos, and extinction coefficients for dust and biomass (assuming dry aerosol) in HiGAM.
Particle sizes, single scattering albedos, and extinction coefficients for dust and biomass (assuming dry aerosol) in HiGAM.

When the first two experiments had reached model year 2001, a second pair was started using initial conditions from near the end of the first experiment pair, and was run from model years 1983 to 2000 (PasD2 and ActD2). The purpose of this strategy was to create a small ensemble of experiments, which although too small for statistical analysis might give a more robust indication of the variability of the dust and its impact in the model. Although space does not allow for a full investigation of the feedback effects in this work, the results of all of the experiments are included for completeness. In the passive dust experiments, the “double-radiation call” method was used to calculate the dust radiative forcing (as in Woodward 2001), in which the shortwave (SW) and longwave (LW) radiative increments resulting from the dust are calculated but are not used when the model is advanced (thus eliminating feedbacks). In all of the experiments, the biomass aerosol was radiatively active via its direct and indirect effects.

Accurate representation of orography is crucial to the modeling of surface winds that determine dust aerosol emissions, and the underlying surface albedo strongly affects the shortwave radiative forcing effect (with dust appearing bright over dark surfaces, e.g., oceans and vegetation, and dark over bright surfaces, such as clouds and deserts). The soil albedo and orography over North Africa used in HiGAM are shown in Fig. 1. The soil albedo file [based on analysis of Moderate Resolution Imaging Spectroradiometer (MODIS) data, see Shaffrey et al. (2009) for details] was demonstrated in Milton et al. (2008) to improve the performance of the Met Office NWP model, and the mean albedo over the Sahara (20°–30°N, 15°W–20°E) is 0.34. This is comparable with other modeling studies, for example, Yoshioka et al. (2007) have 0.27 and report that Miller et al. (2004) have 0.33. The model orography [derived from Global Land One-kilometer Base Elevation (GLOBE) dataset, see Shaffrey et al. (2009)], allows the Atlas, Hoggar, Tibesti, and Aïr Mountains to be well resolved.

Fig. 1.

HiGAM soil albedo (grayscale) and orography (open contours; km) over North Africa. Locations of orographic features and stations referred to in the text are also shown.

Fig. 1.

HiGAM soil albedo (grayscale) and orography (open contours; km) over North Africa. Locations of orographic features and stations referred to in the text are also shown.

3. Results

a. Global mean dust and biomass aerosol loadings

Details of the four experiments with their annual mean biomass and dust loadings are given in Table 2. Although the biomass loading is virtually the same for all years and all experiments, the dust shows much higher variability both between and within the experiments. This is demonstrated by the range of annual mean loadings and their standard deviations, and also by the time series of annual mean dust loadings for each year of the four HiGAM experiments shown in Fig. 2. The dust burdens, ranging from 31.73 to 43.05 Tg for each complete run mean and from 20.23 to 72.98 Tg for individual years, are not unrealistic, although these values are at the higher end of those reported in other papers. For example, Zender et al. (2004) summarize estimates from models and observations as lying in the range of 8–36 Tg, whereas Ginoux et al. (2001) obtain 31–40 Tg for individual years from the Goddard Chemistry Aerosol Radiation and Transport (GOCART) model. This may be at least partially explained by the wider particle size range (from 0.03- to 30-μm radius) used in this work; the largest particles (>10 μm) contribute up to 1.94 Tg (∼5%) to the total dust load. Also shown in Table 2 are the mass fractions of the mean dust load with particle radius below and above 1 μm (with size bins 1–3 and 4–6, respectively); for the active dust experiments the ratio is 1:4 and for the passive dust experiments it is 1:3. These proportions are similar to Woodward (2001) and Miller et al. (2006); however, other studies have a higher proportion of smaller particles, for example, Ginoux et al. (2001) have a ratio of 2:3 with a particle size range of 0.1–6 μm. Others have a lower proportion, for example, Yoshioka et al. (2007) have a ratio of 1:7 with modeled particle sizes of 0.1–10 μm. This is an important aspect of the dust modeling because it impacts the lifetime of the dust in the atmosphere (larger particles fall out more quickly than smaller ones) and also the radiative forcing, as a result of the different spectral properties of the different size modes.

Table 2.

Details of the four HiGAM dust experiments with global annual mean biomass and dust loadings (Tg) (for all dust particle sizes and for those <10 μm), the relative mass fractions of dust below and above 1-μm radius (%), and the standard deviations of the annual mean dust loading (Tg with % in parentheses) within each experiment.

Details of the four HiGAM dust experiments with global annual mean biomass and dust loadings (Tg) (for all dust particle sizes and for those <10 μm), the relative mass fractions of dust below and above 1-μm radius (%), and the standard deviations of the annual mean dust loading (Tg with % in parentheses) within each experiment.
Details of the four HiGAM dust experiments with global annual mean biomass and dust loadings (Tg) (for all dust particle sizes and for those <10 μm), the relative mass fractions of dust below and above 1-μm radius (%), and the standard deviations of the annual mean dust loading (Tg with % in parentheses) within each experiment.
Fig. 2.

Annual mean dust loadings for the two active (ActD1 and ActD2) and two passive (PasD1 and PasD2) dust experiments. Note that loadings for years 1982–85 for ActD1 and PasD1 are affected by a soil moisture error in the model (see text).

Fig. 2.

Annual mean dust loadings for the two active (ActD1 and ActD2) and two passive (PasD1 and PasD2) dust experiments. Note that loadings for years 1982–85 for ActD1 and PasD1 are affected by a soil moisture error in the model (see text).

The experiments including dust radiative effects (ActD1 and ActD2) have annual mean dust burdens exceeding those of the passive dust experiments (PasD1 and PasD2) by approximately 18%. The authors recognize that this overall positive feedback on the dust production is unusual [e.g., Miller et al. (2004), and others, find negative feedbacks], and it was a strong motivation for carrying out the second pair of experiments. Investigation shows that the feedback mechanisms are complex and highly heterogeneous in space and time, with some regions having increased and some having decreased dust burdens at different times of year. A full explanation and discussion of the processes at work are outside the scope of this paper, and will be discussed, together with other feedback effects, in a later publication. It is also evident that the dust burdens in the second experiment pair (ActD2 and PasD2) are higher than those of the first experiment pair (ActD1 and PasD1). This was found to be because of a problem with the parameterization of soil moisture in the model, which introduced a systematic moist bias over bare silt and clay surfaces in the first experiment pair; this suppressed dust emissions in the first few model years, especially over the western Sahara. This problem also affected the initial dust-tuning experiments, which were too short (3 yr) for the problem to be noticed, and so dust burdens in the longer runs were higher than expected. Although not ideal, this is not a major problem in the present paper, which is focused on the physical processes that are improved with resolution, rather than the overall dust mass loadings.

b. Dust and biomass aerosol distributions

The 20-yr mean global dust distributions from the experiment ActD1 are shown for each season in Fig. 3 [the means are denoted for December–February (DJF), March–May (MAM), June–August (JJA), and September–November (SON)]. These agree well qualitatively with other models, for example, Lunt and Valdes (2002), Ginoux et al. (2001) (which shows less seasonal variation), and Miller et al. (2006). North Africa is the strongest source throughout the year, with significant contributions from Saudi Arabia, China, and North America for March–November, and from Australia and South America for September–February. The corresponding vertical cross sections of zonal mean dust concentrations can be seen in Fig. 4. In all seasons, North Africa is the main source (15°–30°N), however, smaller sources at 10°N from Somalia in JJA and at 35°–40°N from China in JJA and SON can also be seen. In the Southern Hemisphere, dust sources in Australia, South Africa, and the Atacama in South America are apparent at 20°–30°S in SON and DJF, and dust from Patagonia at 45°S can be seen in DJF.

Fig. 3.

The 20-yr seasonal mean dust from HiGAM experiment ActD1: vertically integrated column loadings (mg m−2) for DJF, MAM, JJA, and SON.

Fig. 3.

The 20-yr seasonal mean dust from HiGAM experiment ActD1: vertically integrated column loadings (mg m−2) for DJF, MAM, JJA, and SON.

Fig. 4.

The 20-yr seasonal mean dust from HiGAM experiment ActD1: zonal mean mass mixing ratios (μg kg−1) vs model level for DJF, MAM, JJA, and SON. Vertical axes show (left) model level with (right) equivalent pressures (hPa) for lhs plots and equivalent height (m) for rhs plots.

Fig. 4.

The 20-yr seasonal mean dust from HiGAM experiment ActD1: zonal mean mass mixing ratios (μg kg−1) vs model level for DJF, MAM, JJA, and SON. Vertical axes show (left) model level with (right) equivalent pressures (hPa) for lhs plots and equivalent height (m) for rhs plots.

The 20-yr annual and seasonal mean biomass emissions and atmospheric loadings in experiment ActD1 are shown in Table 3. The annual total emission of 80 Tg is in fairly good agreement with the best estimated flux (90 Tg yr−1) in Kiehl and Rodhe (1995). The emissions have two peaks: one in boreal winter (DJF), mainly from the Sahel and equatorial region of Africa, and the other in austral winter (JJA), mainly from southern Africa and Brazil. Also shown are aerosol optical depths (AODs) at 550-nm wavelength of the biomass and dust, which have been calculated offline from their respective aerosol loadings using the spectral properties used in the model radiation code, assuming dry aerosol for simplicity. The AOD of the biomass has an annual mean of 0.016, with a maximum of 0.026 in DJF and a minimum of 0.007 in MAM. The AOD of the dust has an annual mean of 0.038 (double that of biomass), peaking in JJA at 0.057 and falling to 0.016 in DJF. The dust AOD dominates that of biomass in all seasons except DJF.

Table 3.

Global annual and seasonal mean dust and biomass loadings (Tg) and their AODs from the active dust experiment ActD1 for DJF, MAM, JJA, and SON.

Global annual and seasonal mean dust and biomass loadings (Tg) and their AODs from the active dust experiment ActD1 for DJF, MAM, JJA, and SON.
Global annual and seasonal mean dust and biomass loadings (Tg) and their AODs from the active dust experiment ActD1 for DJF, MAM, JJA, and SON.

The spatial distributions of the combined dust and biomass burning aerosol loadings and optical depths in DJF and JJA are shown in Fig. 5a from experiment ActD1 (note that biomass-only loadings are overplotted in open contours, and dust-only loadings are shown in Fig. 3). It can be seen that in DJF the biomass burning is the major contributor to the AOD, with the maximum value being located over the Sahel region of Africa. In JJA, the biomass burning dominates in the Southern Hemisphere with maxima in the AOD over southern Africa and central South America; however, the dust dominates the AOD in the Northern Hemisphere with maxima over the Sahara and Taklimakan (central Asia). Comparing with aerosol optical thicknesses over oceans from the Advanced Very High Resolution Radiometer (AVHRR) data published in Husar et al. (1997; not shown), the seasonal values over the North and South Atlantic agree well, having a maximum in DJF in excess of 0.5 in the Gulf of Guinea and decreasing westward across the Atlantic to 0.2 off the coast of northern South America. In JJA, there is a local maximum of 0.5 off the coast of Mauritania, decreasing to 0.2 in the Caribbean. Note that the AOD for other aerosol species in the model (sulfate, black carbon, and sea salt) are not included in Fig. 5a but are included in plate 1 of Husar et al. (1997). Figure 5b shows the annual mean combined biomass and dust AOD from model ActD1 compared with the 2000–04 annual mean AOD at 550 nm from MODIS (data available online at http://modis-atmos.gsfc.nasa.gov). Again, the MODIS product contains contributions from other aerosols; however, over the tropical North Atlantic, where dust and biomass dominate, it compares well with the model and both have AOD values of 0.3–0.4 off the west coast of North Africa.

Fig. 5

a. DJF and (bottom) JJA column-integrated combined biomass and dust aerosol mass loadings (colored-filled contours) overplotted with biomass only (open black contours; mg m−2). (middle) Corresponding DJF and JJA optical depths at 550 nm for combined dust and biomass aerosol. All results are 20-yr means from model ActD1.

Fig. 5

a. DJF and (bottom) JJA column-integrated combined biomass and dust aerosol mass loadings (colored-filled contours) overplotted with biomass only (open black contours; mg m−2). (middle) Corresponding DJF and JJA optical depths at 550 nm for combined dust and biomass aerosol. All results are 20-yr means from model ActD1.

Fig. 5

b. The 20-yr annual mean optical depths at 550 nm for combined dust and biomass aerosol from model ActD1 and (bottom) annual mean AOD for years 2000–04 from MODIS.

Fig. 5

b. The 20-yr annual mean optical depths at 550 nm for combined dust and biomass aerosol from model ActD1 and (bottom) annual mean AOD for years 2000–04 from MODIS.

c. Vertical profiles of dust and biomass aerosol

Next, we examine vertical profiles of dust and biomass aerosol in the model and how they change as the aerosols are advected downstream of the source regions in North Africa. Figure 6 shows latitude–vertical cross sections of the dust and biomass concentrations, averaged between 20°W and 30°E for the winter (DJF) and summer (JJA) seasonal means from ActD1. In DJF, the dust layer remains lower in the atmosphere and is concentrated farther south than in JJA, whereas the biomass is farther north in DJF than in JJA. Also, the biomass is concentrated higher in the atmosphere than the dust in DJF, although they occupy similar latitudes, whereas in JJA the dust and biomass are separate in both latitude and height. This layered structure of aerosols over Africa has been noted in observation campaigns (e.g., Johnson et al. 2008). Looking in more detail at individual locations, Fig. 7 shows vertical profiles of dust and biomass aerosol concentrations from the model for locations representing the western Sahara (Zouerate; 23°N, 13°W), the Bodélé region (17°N, 18°E), and far downstream across the Atlantic at Bermuda and Barbados. Note that the Bodélé and Zouerate locations are just south of the maxima in dust loadings for DJF and JJA, respectively, shown in Fig. 6. Profiles for DJF and JJA are shown in Fig. 7, and the horizontal scale is logarithmic in order to accommodate both types of aerosol on the same plot for comparison. In all four locations, and at all levels, and in both summer and winter there is more dust aerosol than biomass, with the exception of the Bodélé; here, the biomass slightly exceeds the dust at 600 hPa in the wintertime. In all except the Bodélé location at low levels, the dust concentration in JJA exceeds that in DJF, whereas the biomass concentration in DJF is greater than that in JJA. At low levels in the Bodélé this situation is reversed, with wintertime dust and summertime biomass dominating. The wintertime dust, which is generated locally in the Bodélé Depression, reduces rapidly with height because the dust is advected away from the area and the heavier particles fall out of the air column. By contrast, in the summer, the dust concentration is almost unchanged with a height up to about 600 hPa, with a slight increase at the top of this layer, indicating a deeper, well-mixed boundary layer with advection of dust into the region from more distant sources. Farther north and west at Zouerate, the dust concentrations at the surface are very high in both seasons, decreasing steadily with height; however, this occurs more rapidly in winter than in summer (as can be seen in Fig. 6). The biomass in DJF has a similar profile to that in the Bodélé; however, it is much smaller in JJA, where the more northern location remains clear of the central African biomass sources.

Fig. 6.

The 20-yr mean seasonal vertical cross sections of dust (colored filled contours) and biomass burning (black open contours) aerosol concentrations over Africa for (top) DJF and (bottom) JJA. Sections are averaged between 20°W and 30°E from experiment ActD1 (μg kg−1). Vertical axes show (left) model level with (right) equivalent pressures (hPa).

Fig. 6.

The 20-yr mean seasonal vertical cross sections of dust (colored filled contours) and biomass burning (black open contours) aerosol concentrations over Africa for (top) DJF and (bottom) JJA. Sections are averaged between 20°W and 30°E from experiment ActD1 (μg kg−1). Vertical axes show (left) model level with (right) equivalent pressures (hPa).

Fig. 7.

Vertical profiles of dust and biomass aerosol concentrations (μg m−3) from experiment ActD1 for DJF and JJA seasons at four locations. The rhs is representative of (top) the western Sahara and (bottom) the Bodélé region. The lhs shows locations downstream of the aerosol plumes at (top) Bermuda and (bottom) Barbados. Solid lines represent dust and dashed lines represent biomass aerosol (JJA, gray lines; DJF; black lines).

Fig. 7.

Vertical profiles of dust and biomass aerosol concentrations (μg m−3) from experiment ActD1 for DJF and JJA seasons at four locations. The rhs is representative of (top) the western Sahara and (bottom) the Bodélé region. The lhs shows locations downstream of the aerosol plumes at (top) Bermuda and (bottom) Barbados. Solid lines represent dust and dashed lines represent biomass aerosol (JJA, gray lines; DJF; black lines).

Looking downstream of the main dust and biomass sources, at Barbados, dust concentrations are of the same order of magnitude in winter and summer near the surface; however, they differ aloft, decreasing with height in DJF but increasing in JJA up to around 500 hPa before decreasing. The dust, here, has been transported across the Atlantic from Africa, from sources in the Bodélé in DJF and the Sahara in JJA. The biomass, however, originates from the Sahel in DJF and is an order of magnitude higher at the surface than in JJA, when it originates from northern and central South America. At Bermuda, the DJF and JJA vertical profiles of dust are similar in shape, being almost constant with height up to about 150 hPa, though an order of magnitude higher in JJA than in DJF. Here, the source of dust is again Africa, although much less reaches Bermuda than Barbados in DJF because of its more northern location. Biomass concentrations at Bermuda are very low in both seasons because of its remoteness from African and American sources.

d. Dust storm case studies

We now examine individual dust outbreaks to provide a more focused evaluation of the physical processes in the model and the level of detail that it can represent. HiGAM cannot reproduce real events as in other work, where either real-time or reanalysis data are used to drive the dust production; however, the following “case studies” demonstrate that HiGAM is capable of modeling dust storms with realistic time and space scales, generated by synoptic-scale events. We examine how the model simulates dust storms emanating from North Africa, because this is the major global producer of dust throughout the year. In DJF the intertropical convergence zone (ITCZ) is well to the south of the region, at approximately 5°N, and high pressure systems drifting across North Africa from the Atlantic bring strong northeasterly winds through the dry valleys between the Hoggar, Tibesti, Aïr, and Darfur highlands (see Fig. 1). These deflate dust from the Bodélé and southeastern Sahara regions, which is then transported southwestward across Niamey and out to the Atlantic coast. By contrast, in JJA, the ITCZ is positioned just north of the Sahel region, at about 20°N; moist southwesterly winds from the Atlantic lie at its southern edge, whereas low pressure regions (heat lows) develop over the hot, dry deserts to the north. The main dust production areas are farther north and west than in the winter, in the central Sahara (∼27°N, 2°E) and coastal region of the western Sahara, and the deflated dust is lofted higher than in the winter months and carried out westward across the Atlantic by the easterly jet and easterly waves (at 600–700 hPa). We now consider how well HiGAM can simulate these seasonal processes by analyzing individual dust storms generated within the model during February and July.

1) February dust outbreak

Considering the February case first, Fig. 8 shows dust loadings with 925-hPa winds and mean sea level pressure (MSLP) with grid box mean rainfall rates from the model for 1200 UTC 2–7 February. To understand the model evolution, it is helpful to examine this sequence together with Fig. 9, which shows the time series of 3-hourly data for the same but slightly extended period (1–10 February) for a location in the western Sahara (Zouerate) and in the Bodélé region of Chad (see Fig. 1 for locations). The dust storm was initiated by a ridge of high pressure drifting eastward across Algeria, similar to the one observed in the BoDEX 2005 campaign (reported by Washington et al. 2006). In the model, the latitudinal pressure gradient across North Africa increases until 5 February, bringing stronger surface winds that cause the dust to be emitted; first northwest of Zouerate from the western slopes of the Hoggar (on 3 February), then north of the Bodélé from the western slopes of the Tibesti (on 4 February), and then from the Bodélé itself (on 5 February). Dust emissions from all three sources continue until 7 February, after which the pressure gradient relaxes and surface winds and emissions diminish. The dust that is already lofted spreads south and west, forming an aerosol cloud that covers most of northwest Africa up to 30°N, moving westward over the coast and out across the Atlantic. The model has two areas of precipitation, with one linked to the low pressure region over the Mediterranean and the other just north of the equator, associated with the ITCZ.

Fig. 8.

(left to right) (top) Dust loadings (color shading, mg m−2) with 925-hPa wind vectors (m s−1) and (next row) mean sea level pressure contours (hPa) with grid-box mean precipitation rates (color shading, mm day−1) for 1200 UTC 2–4 Feb from experiment ActD2. (bottom six panels) The same but for 1200 UTC 5–7 Feb.

Fig. 8.

(left to right) (top) Dust loadings (color shading, mg m−2) with 925-hPa wind vectors (m s−1) and (next row) mean sea level pressure contours (hPa) with grid-box mean precipitation rates (color shading, mm day−1) for 1200 UTC 2–4 Feb from experiment ActD2. (bottom six panels) The same but for 1200 UTC 5–7 Feb.

Fig. 9.

Time series of 3-hourly data from the model for locations in (top) the western Sahara and (bottom) the Bodélé. (left) Surface temperature (°C, red), wind speed at 925 hPa (m s−1, dark blue), wind speed at 10 m (m s−1, cyan), and log10 dust emissions (μg m−2 s−1, green). (right) OLR (W m−2, red), mean sea level pressure (decaPa – 10000, black), cloud cover (%, blue), and column dust load (g m−2, green). Vertical lines denote 1200 UTC on each day for the 1–10 Feb period.

Fig. 9.

Time series of 3-hourly data from the model for locations in (top) the western Sahara and (bottom) the Bodélé. (left) Surface temperature (°C, red), wind speed at 925 hPa (m s−1, dark blue), wind speed at 10 m (m s−1, cyan), and log10 dust emissions (μg m−2 s−1, green). (right) OLR (W m−2, red), mean sea level pressure (decaPa – 10000, black), cloud cover (%, blue), and column dust load (g m−2, green). Vertical lines denote 1200 UTC on each day for the 1–10 Feb period.

From Fig. 9, it can be seen that there is a strong diurnal cycle in all of the variables shown, and dust emissions are initiated when surface winds exceed 7 m s−1. However, these winds need to be sustained in order for significant amounts of the dust to be lofted into the atmospheric column, as can be seen by comparing the emissions plots on the left of Fig. 9, with the dust column loadings on the right. Increasing wind strength and the presence of dust aerosol both act to reduce the amplitude of the diurnal cycle in surface temperature, from around 30° on 2 February to about 5° on 5 February, at Zouerate, for example. The diurnal cycle in the OLR is affected by both dust and cloud, as can be seen from the reduction in amplitude between 5 and 7 February at the Bodélé location (because of dust), and on 3 February at Zouerate (because of clouds). Comparing the surface (10 m) and 925-hPa winds at the Bodélé location, it can be seen that between 1 and 4 February there is a marked diurnal cycle in both winds. However, there is a phase difference of ∼9 h, as the 925-hPa wind peaks during the night and reaches a minimum during the early afternoon, whereas the surface wind peaks at midday and is weakest during the night. Washington et al. (2006) found a similar phenomenon in their analysis of data from the BoDEX field experiment, in which they identified a low-level jet in the Bodélé region at ∼925 hPa that was responsible for raising dust plumes. They found maximum surface wind speeds of 10 m s−1 on dust-free days and in excess of 16 m s−1 on dusty days, whereas in the jet core at 925 hPa, speeds in excess of 24 m s−1 were recorded, although measurements could not be made on the dustiest days. They commented that ERA-40 was not able to resolve such mesoscale features and that it represented a challenge for climate models. The model, here, appears to simulate these low-level winds quite realistically, reaching a maximum of ∼17 m s−1 at the surface and ∼27 m s−1 at 925 hPa between 5 and 6 February (see Fig. 9), after which time surface winds diminish and dust emissions are reduced. The main reason for this is the relaxation of the large-scale pressure gradient, causing a reduction in surface wind speeds; however, on a local scale it could also be a result of the negative feedback mechanism (discussed in Miller et al. 2004), by which the dust lofted by these winds causes a reduction in the surface temperature, thus stabilizing the boundary layer and reducing the surface wind speeds and subsequent dust emissions. This could be inferred from the time series in Fig. 9 between 5 and 8 February, in the Bodélé region. Although the meteorological situation that initiated this event differs from the one in March 2006 studied by Slingo et al. (2006), in other respects the model reproduces much of the observed large-scale behavior, notably the interaction between the large-scale flow, the topography, and the generation and propagation of the dust.

2) July dust outbreak

Turning now to the July case, Fig. 10 shows dust loadings with 925-hPa winds and mean sea level pressure with grid box mean rainfall rates from the model for 1200 UTC 24–29 July. Figure 11 shows the time series of 3-hourly data for the same, slightly extended period (21–30 July), for the same locations as those in Fig. 9. In this case, it is the strengthening cyclonic winds around a developing heat low over the western Sahara that cause the dust uplift. In contrast to the February case, the Bodélé is not a major dust source region and emissions arise from the Sahara at various sources between 20° and 30°N. Comparing the surface and 925-hPa winds for the Bodélé region in Figs. 9 and 11, it can be seen that the low-level jet, which drives the dust emissions in February, is much weaker in July, as noted by Washington et al. (2006). At Zouerate, the westward passage of the developing low pressure region and associated dust cloud can be seen on 27 July; however, from the time series in Fig. 11 it can be seen that most of the dust has been advected in rather than raised locally, because emissions are not particularly high on 27 July. Suppression of the diurnal cycle in OLR and surface temperature can be seen at Zouerate resulting from the dust, and also at the Bodélé location resulting from the cloud associated with the ITCZ, which is marked by the band of precipitation centered at ∼10°N on Fig. 10. The precipitation band is stronger and farther north in July than in February because of the more northerly location of the ITCZ, and the West African monsoon winds can be seen in bringing moist air from the Gulf of Guinea to the Sahel. As the low pressure region over Zouerate on 27 July moves westward out to the Atlantic, a second low can be seen forming at 30°E over the Darfur highlands, which deepens and moves westward, reaching 15°E on 29 July. This is the classical signature of an African easterly wave, described by Burpee (1972) and others, and it is recognized that their westward propagation is associated with the transport of large amounts of Saharan dust across the Atlantic in the Saharan air layer during boreal summer (e.g., Carlson and Prospero 1972).

Fig. 10.

As in Fig. 8, but for 1200 UTC 24–26 July and 1200 UTC 27–29 Jul.

Fig. 10.

As in Fig. 8, but for 1200 UTC 24–26 July and 1200 UTC 27–29 Jul.

Fig. 11.

As in Fig. 9, but for 21–30 Jul.

Fig. 11.

As in Fig. 9, but for 21–30 Jul.

Comparing the structure and movement of the aerosol clouds in the February and July case study sequences, it is clear that the dust generation and transport mechanisms are quite different in the two seasons and that the model has sufficient resolution and complexity to capture them both. This is important because although the dust storms are relatively small-scale, short-lived phenomena, they modify the earth’s radiation budget both regionally, during the course of the storm, and globally, via the release of small, long-lived particles that are carried high into the atmosphere and far from the original source region.

e. Comparison with observations

Although quantitative verification of global dust distributions can be problematic (as noted in the introduction), it is nonetheless important to make comparisons with observations where possible. Figure 12 shows scatterplots comparing the active dust model (ActD2) and the passive dust model (PasD2) against the observations. The top-left panel compares the 18-yr annual mean combined dust and biomass AOD from the models with observed annual mean values from the Aerosol Robotic Network Program (AERONET) stations, selected as those most affected by these aerosols and with more than 2-yr recorded data (available online at http://aeronet.gsfc.nasa.gov). The stations were the following: Barbados; Bermuda; Dry-Tortugas (three stations that are fairly remote from main dust and biomass sources); Cape Verde; Dahkla, Morocco; Izana, Canary Islands (all of which are close to African dust sources); Ouagadougou, Burkina Faso (affected by African dust and biomass burning); Ilorin, Nigeria (affected mainly by African biomass burning); Sede-Boker, Israel; and Solar Village (both of which are close to Arabian dust sources). Model AOD values depend on the column loading and the spectral properties of the aerosol; ideally it is desirable for the model to fit the observations to within a factor of 2, as is the case for some of the stations shown. Locations where the model AOD is lowest relative to the observations are those remote from dust and biomass sources, where other aerosols (such as sulfate and sea salt) may be influencing observed values. The location where the model AOD is highest relative to the observation is Izana, where the model dust burdens are too high (see later in the study).

Fig. 12.

Scatterplot showing model vs observed values of (top left) AOD and (top right and bottom) surface dust concentrations (μg m−3). The dashed diagonal line indicates where model = observed values, the dotted diagonals indicate one order of magnitude difference. Black and gray symbols denote 18-yr annual mean values from experiments ActD2 and PasD2, respectively. Observed AOD values are from selected AERONET sites (see text). Surface dust concentrations are from Miami University sites (J. Prospero and D. L. Savoie, personal communication, where M_Uni_VG denotes high-confidence observations, M_Uni_OK are those occasionally affected by local dust sources, and elev are those from elevated stations) and Duce (1995). Note that scales are logarithmic.

Fig. 12.

Scatterplot showing model vs observed values of (top left) AOD and (top right and bottom) surface dust concentrations (μg m−3). The dashed diagonal line indicates where model = observed values, the dotted diagonals indicate one order of magnitude difference. Black and gray symbols denote 18-yr annual mean values from experiments ActD2 and PasD2, respectively. Observed AOD values are from selected AERONET sites (see text). Surface dust concentrations are from Miami University sites (J. Prospero and D. L. Savoie, personal communication, where M_Uni_VG denotes high-confidence observations, M_Uni_OK are those occasionally affected by local dust sources, and elev are those from elevated stations) and Duce (1995). Note that scales are logarithmic.

Scatterplots of the annual mean surface concentrations of dust against observations from the University of Miami dataset (J. Prospero and D. L. Savoie, personal communication; see Woodward 2001 for further details) and from Duce (1995) are also shown in Fig. 12. All of the observations with the highest confidence levels (denoted M_Uni_VG; upper-right panel) are fitted to within an order of magnitude and there is no obvious bias in the models at high or low dust concentrations. Although greater accuracy is desirable, surface concentrations are highly variable and represent a strong test of the model’s emission, transport, and deposition processes (e.g., Todd et al. 2008), and our results are of similar accuracy to other models (e.g., Stier et al. 2005; Miller et al. 2006). Some of the larger discrepancies are at sites where there is less confidence in the observed values because they are likely to be influenced by local sources (denoted by M_Uni_OK); for most of these, the model surface concentrations are lower than those observed. Large differences are also seen for elevated locations (denoted by elev), where the model values are generally higher than those observed. This is likely to be because of differences between the model and actual orographic height, even though dust concentrations at the model level closest to the altitude of the observation have been plotted. In general, there is little difference between the active and passive dust experiments, in terms of goodness of fit to the surface observations.

We now examine selected stations in more detail, comparing the seasonal mean surface dust concentrations from all of the model experiments, with observations at four stations in the University of Miami aerosol network that have high confidence levels and are located downstream of the dust plumes emanating from North Africa. The locations chosen are Bermuda; Miami, Florida; Barbados; and Izana, and the comparisons are shown in Fig. 13a. Although the model dust concentrations are higher than those observed at all stations shown, at Bermuda and Miami, the values are correct to within a factor of 5 and the seasonality is well captured by all the model experiments. At Barbados, there is more variation between the models; however, the seasonality is correct in all except ActD2, which peaks in MAM instead of JJA. This is because of strong Saharan dust storm activity in that season in ActD2 (which has the highest annual mean dust burden of all the model experiments). At Izana, very close to the Saharan dust source, the model experiments show the most variation from the observations, being too high by a factor of 5–100; however, concentration gradients are very steep in this area, and small positional differences can produce large changes in dust concentrations. Observations at the same location (Tenerife, Canary Islands) reported in Duce (1995; from Arimoto et al. 1993) give a range of 0.06–380 μg m−3, within which our model annual mean values lie.

Fig. 13.

(a) Seasonal mean surface dust concentrations (μg m−3) from HiGAM for some of the University of Miami sites and (b) seasonal mean AOD values from some of the AERONET sites. Seasonal means from ActD1 (black solid lines), PasD1 (gray solid lines), ActD2 (black-dashed lines), and PasD2 (gray-dashed lines). Black-dotted lines indicate observations (monthly values have been averaged to give seasonal means).

Fig. 13.

(a) Seasonal mean surface dust concentrations (μg m−3) from HiGAM for some of the University of Miami sites and (b) seasonal mean AOD values from some of the AERONET sites. Seasonal means from ActD1 (black solid lines), PasD1 (gray solid lines), ActD2 (black-dashed lines), and PasD2 (gray-dashed lines). Black-dotted lines indicate observations (monthly values have been averaged to give seasonal means).

Figure 13b shows the seasonal AOD of the dust and biomass burning from all of the model experiments, compared with observations from four of the AERONET stations. Barbados and Izana are chosen for comparison with the surface dust concentrations shown in Fig. 13a, together with Cape Verde (17°N, 23°W) and Ilorin (8°N, 4°E). At Barbados, the model AOD values correctly peak in JJA; however, all are lower than the observed values, despite the surface dust concentration values being too high. As noted previously, this is likely a result of other aerosols affecting the observed AOD values. At Izana, the model AOD values are too high, as are the surface concentrations; however, AOD values are fitted to within a factor of 5 (better than the surface concentrations). At Cape Verde, all of models fit the observed AODs well except for ActD2, which incorrectly peaks in MAM. This feature can also be seen in the AOD and surface concentrations at Barbados, as noted above. At Ilorin, where the AOD is influenced mainly by biomass burning, all the models correctly have a maximum in DJF and a minimum in JJA but are too low in MAM. This may be a result of the (prescribed) biomass burning emissions being too low in that season.

f. Evaluation using GERB data

The Geostationary Earth Radiation Budget (GERB; Harries et al. 2005) instrument aboard the second generation of the Meteosat satellites enables estimates of reflected shortwave and OLR to be made at high temporal resolution. These data are ideal for diurnal cycle (Comer et al. 2007) and model validation studies (e.g., Allan et al. 2005). Here, a comparison is made between the diurnal cycle of OLR, simulated by special HiGAM runs over a single month (representing July 1985), with that estimated using edition 1 GERB OLR data (Dewitte et al. 2008) from July 2006. In these special HiGAM runs, calls to the radiation scheme were made at every model time step (20 min), instead of the usual 3-hourly calls, to resolve the diurnal cycle on a similar time scale to that of the GERB data (15 min). Figure 14 shows the amplitude of the mean July diurnal cycle in OLR over North Africa from GERB data and from HiGAM experiments, excluding and including dust radiative effects. Also shown is the July mean dust loading from the special model run. Although this does not attempt to match the exact dust conditions for the GERB data, it is not very different spatially from the 18-yr JJA mean shown in Fig. 3, which has been shown to be reasonably realistic. The amplitude in the model is clearly higher than that in the observations, however, with a much smaller difference when the dust is included (consistent with the behavior shown in Figs. 9 and 11). To illustrate this improvement further, Fig. 15 shows the mean diurnal cycle of OLR over two areas of the Sahara for the GERB observations and the two model runs. The upper panel shows the diurnal cycles for a region close to the middle of the desert, whereas the lower panel shows the diurnal cycle for a region farther west that has high levels of simulated dust in the model. Without the dust scheme, the model produces very high OLR over the desert compared with observations, particularly in the dustier western region (lower panel). Inclusion of the dust scheme in the model greatly reduces this problem and brings the simulated OLR much closer to the observations. This result is in agreement with Haywood et al. (2005), who compared averaged July 2003 OLR simulations from the Met Office NWP model with OLR estimated from Meteosat-7 radiances. They also found that the simulated OLR over the African desert regions was too high, and they concluded that the most likely explanation was that the model did not simulate the high loadings of mineral dust aerosol found in the area.

Fig. 14.

Twice the amplitude (peak to peak) of the mean July diurnal cycle of the outgoing thermal radiation over North Africa from (a) GERB data and HiGAM, (b) excluding, and (c) including dust radiative effects. (d) The July mean dust loading (mg m−2) from the same HiGAM run.

Fig. 14.

Twice the amplitude (peak to peak) of the mean July diurnal cycle of the outgoing thermal radiation over North Africa from (a) GERB data and HiGAM, (b) excluding, and (c) including dust radiative effects. (d) The July mean dust loading (mg m−2) from the same HiGAM run.

Fig. 15.

Mean July diurnal cycle in OLR from GERB data and HiGAM: (top) the mean for the area 20°–30°N, 10°–20°E and (bottom) the mean for the area 20°–30°N, 0°–10°W.

Fig. 15.

Mean July diurnal cycle in OLR from GERB data and HiGAM: (top) the mean for the area 20°–30°N, 10°–20°E and (bottom) the mean for the area 20°–30°N, 0°–10°W.

g. Radiative forcing

Radiative forcing is a useful measure of the likely impact of an aerosol on the climate system, although by definition it does not represent the full nonlinear response (see Forster et al. 2007). The global annual and seasonal mean dust loading and radiative forcing at the surface (SFC) and top of the atmosphere (TOA) diagnosed from the passive dust experiment (PasD1) are given in Table 4. Forcings are calculated from the differences of the radiation increments when the contribution from the dust is included and excluded. The TOA forcing is positive in MAM and JJA, zero in DJF, and negative in SON, giving an annual mean value very close to zero; whereas the surface net forcing is negative in all seasons, giving an annual mean value of −1.14 W m−2. The global mean atmospheric column heating because of the dust (calculated as the difference between the TOA and surface forcings) is positive in all seasons, with a maximum of 1.84 W m−2 in JJA and a minimum of 0.42 W m−2 in DJF. Miller et al. (2004) experimented with three different sets of spectral properties for dust aerosols, using values representing more absorbing, more scattering, and moderately absorbing and scattering aerosol. The net atmospheric heating ranged from 3.26 (for the more absorbing case) to 0.25 W m−2 (more scattering case), with a value of 1.46 W m−2 for the moderate case, which agrees fairly well with our annual mean value of 1.15 W m−2.

Table 4.

Global annual and seasonal mean dust loading (Tg) and net radiative forcing (W m−2) at SFC and TOA diagnosed from the passive dust experiment PasD1. Forcings are calculated from the differences between the radiation increments when the contribution from the dust is included and excluded. Positive (negative) values of forcing indicate warming (cooling). Atmospheric heating is calculated as the difference between the TOA and SFC values.

Global annual and seasonal mean dust loading (Tg) and net radiative forcing (W m−2) at SFC and TOA diagnosed from the passive dust experiment PasD1. Forcings are calculated from the differences between the radiation increments when the contribution from the dust is included and excluded. Positive (negative) values of forcing indicate warming (cooling). Atmospheric heating is calculated as the difference between the TOA and SFC values.
Global annual and seasonal mean dust loading (Tg) and net radiative forcing (W m−2) at SFC and TOA diagnosed from the passive dust experiment PasD1. Forcings are calculated from the differences between the radiation increments when the contribution from the dust is included and excluded. Positive (negative) values of forcing indicate warming (cooling). Atmospheric heating is calculated as the difference between the TOA and SFC values.

A comparison of the annual mean shortwave, longwave, and net radiative forcing at the TOA and SFC from various models is shown in Table 5. Compared to the results of Yoshioka et al. (2007), their SW surface forcing value is close to ours, but their TOA value is much more negative; this may be partially because of their lower mean surface albedo over the Sahara. In the LW, their surface value is much more positive than in this work, but at TOA it is similar to ours; this can be accounted for by the higher proportion of larger particles in their model (as noted previously). Compared with Miller et al. (2006), LW forcings at SFC and TOA are comparable with this work and with Woodward (2001), as are SFC SW forcings. However, their SW TOA values are much more negative, despite having a similar ratio of small to large particles as found in this work. Miller et al. (2006) comment that this is a result of the increase in scattering properties of their dust aerosol, and it has already been noted (see section 2) that the dust in this work may be too highly absorbing over the Sahara, which could explain the differences. For the net forcings, the present work has the weakest TOA forcing (0.01 W m−2) and the strongest SFC forcing (−1.14 W m−2); this is influenced by the magnitude of the dust burdens, the particle size distributions, the surface albedo underlying the aerosol, and the cloud distributions in the models relative to the aerosol.

Table 5.

Comparison of annual mean SW, LW, and net radiative forcing (W m−2) at the TOA and SFC from various models. Mean albedo over Sahara (20°–30°N, 15°W–20°E) for each model is also given.

Comparison of annual mean SW, LW, and net radiative forcing (W m−2) at the TOA and SFC from various models. Mean albedo over Sahara (20°–30°N, 15°W–20°E) for each model is also given.
Comparison of annual mean SW, LW, and net radiative forcing (W m−2) at the TOA and SFC from various models. Mean albedo over Sahara (20°–30°N, 15°W–20°E) for each model is also given.

Figure 16 shows the spatial distribution of the SW, LW, and net (SW + LW) radiative forcing of the dust as diagnosed from PasD1. The global annual mean radiative forcing is −1.14 W m−2 at the surface and zero at the TOA; however, the TOA value varies seasonally from 0.47 in MAM to −0.57 in SON, and it can locally be very high, up to 20 W m−2 over the northwest Sahara in the 20-yr JJA seasonal mean (not shown). Compared to the results of Woodward (2001; in which a similar dust scheme was used in HadAM3), our longwave SFC and TOA forcings are similar; however, our shortwave forcings are more negative, by 28% at the surface and 75% at the TOA. Comparing the SW TOA forcing in clear-sky conditions, our distribution is very similar to that in Woodward (2001), with positive forcing over bright regions such as deserts, snow, and ice, although we have a weaker negative forcing off the west coast of Australia resulting from our smaller Australian dust burdens. However, there are some big differences in the SW TOA for all-sky conditions, most noticeably over the Atlantic off the northwest African coast, where there is strong positive forcing in our model (HiGAM) where dust lies above the marine cloud in boreal summertime and strong negative forcing in HadAM3, indicating a lack of cloud in the region. The improvement in cloud representation in HiGEM, especially the increase in marine stratocumulus, has been noted in Shaffrey et al. (2009).

Fig. 16.

The 20-yr annual mean dust radiative forcing from experiment PasD1. (top) (left) net (SW + LW) surface forcing, (middle) net TOA forcing, and (right) net atmospheric heating as a result of the forcing. (middle) (left) The SW at the surface, (middle) the TOA under all-sky conditions, and (right) the TOA under clear-sky conditions. (bottom) (left) The LW at the surface and (right) the TOA under all-sky conditions. Units are W m−2.

Fig. 16.

The 20-yr annual mean dust radiative forcing from experiment PasD1. (top) (left) net (SW + LW) surface forcing, (middle) net TOA forcing, and (right) net atmospheric heating as a result of the forcing. (middle) (left) The SW at the surface, (middle) the TOA under all-sky conditions, and (right) the TOA under clear-sky conditions. (bottom) (left) The LW at the surface and (right) the TOA under all-sky conditions. Units are W m−2.

4. Summary and discussion

The Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report recognized the progress that has been made in aerosol modeling over recent years, noting that several global atmospheric models now have resolution better than 2° × 2° in the horizontal and 30 levels in the vertical and include most of the important anthropogenic and natural aerosol species (Forster et al. 2007). At the same time, advances in satellite- and ground-based instrumentation have provided more high-quality measurements of aerosol properties against which model output parameters, such as aerosol optical depth, can be compared. It is, thus, timely both to investigate the benefits of further increases in model resolution and to capitalize on the increase in the breadth of data now available for evaluation. In this work, we have investigated the behavior of a tuned version of the mineral dust scheme described by Woodward (2001) in the atmosphere-only version of the high-resolution global climate model U.K. HiGEM. The model has been run in paired experiments with dust treated either as a passive tracer or actively feeding back on the model via its direct radiative effect. Two active and two passive dust experiments were run, starting from different initial conditions, to obtain some measure of the variability of the system and to assess the significance of the results. Although the dust burdens obtained are high compared with other work, this will ultimately be helpful in identifying differences between active and passive dust cases to investigate climate impacts and feedback mechanisms, which will be the subject of future work. The focus of the present paper is to see how well the dust is handled by a global high-resolution model, which we expect to be particularly important and beneficial for dust modeling because of its reliance on the model itself for producing realistic emissions, unlike other aerosols that have externally prescribed emissions distributions.

We have evaluated the model globally and also focused on Africa because it is the world’s largest source of dust aerosol. We also included biomass aerosol in the analysis because it is the second most abundant aerosol over North Africa and observations of aerosol optical depth (from satellite-, aircraft-, and ground-based data) cannot easily distinguish between aerosol types; therefore, a better comparison is achieved by including both in our model analyses. We have demonstrated the importance of the detailed representation of orography and soil albedo over Africa in the February and July dust storm case studies and have shown that the model is capable of simulating the synoptic situations that typically initiate them. There are, of course, some smaller and perhaps equally important dust-lofting mechanisms, which are not explicitly represented in the model, such as subgrid-scale gustiness and surface abrasion processes. The effects of these, in the form of higher dust burdens, have been subsumed (albeit rather crudely) in the tuning process. The model is also critically dependent on the accuracy of the dust parent soil ancillary file for the production of realistic dust fields, and the sensitivity of the regional and global dust burdens and radiative forcing to this may be the subject of future work. Nevertheless, the results show that significant progress has been made by combining a global high-resolution atmospheric GCM with a credible global and local dust cycle, which can be used to investigate climate impacts such as the radiative forcing, feedback effects, and interaction between dust aerosol and easterly waves and tropical cyclones. We believe this represents an important step toward the next generation of dust models that will combine higher horizontal and vertical resolution with increased levels of complexity to investigate aerosol–climate interactions.

Acknowledgments

The advice and support of Prof. Tony Slingo, who sadly passed away in October 2008 and without whom this work would not have been carried out, is gratefully acknowledged. Thanks are also due to two anonymous reviewers whose helpful suggestions improved this work. The support of NERC through the UK-HiGEM project is acknowledged.

This work made use of the computational facilities of HPCx, the U.K. Research Councils’ national high-performance computing service, which is provided by Edinburgh Parallel Computing Centre at the University of Edinburgh and by the Council for the Central Laboratory of the Research Councils Daresbury Laboratory, and is funded by the Office of Science and Technology through the Engineering and Physical Sciences Research Council’s High-End Computing Programme.

REFERENCES

REFERENCES
Allan
,
R. P.
,
A.
Slingo
,
S. F.
Milton
, and
I.
Culverwell
,
2005
:
Exploitation of Geostationary Earth Radiation Budget data using simulations from a numerical weather prediction model: Methodology and data validation.
J. Geophys. Res.
,
110
,
D14111
.
doi:10.1029/2004JD005698
.
Arimoto
,
R.
,
R. A.
Duce
,
B. J.
Ray
,
J. D.
Cullen
,
W. C.
Graustein
, and
K. K.
Turekian
,
1993
:
Naturally occurring continental materials in the atmosphere over the North Atlantic Ocean.
First IGAC Scientific Conf., Eilat, Israel, International Global Atmospheric Chemistry
.
Bagnold
,
R. A.
,
1941
:
The Physics of Blown Sand and Desert Dunes.
Methuen and Co., 265 pp
.
Balkanski
,
Y.
,
M.
Schulz
,
T.
Claquin
, and
S.
Guibert
,
2007
:
Reevaluation of mineral aerosol radiative forcings suggests a better agreement with satellite and AERONET data.
Atmos. Chem. Phys.
,
7
,
81
95
.
Burpee
,
R. W.
,
1972
:
The origin and structure of easterly waves in the lower troposphere of North Africa.
J. Atmos. Sci.
,
29
,
77
90
.
Carlson
,
T. N.
, and
J. M.
Prospero
,
1972
:
The large-scale movement of Saharan air outbreaks over the northern equatorial Atlantic.
J. Appl. Meteor.
,
11
,
283
297
.
Comer
,
R. E.
,
A.
Slingo
, and
R. P.
Allan
,
2007
:
Observations of the diurnal cycle of outgoing longwave radiation from the Geostationary Earth Radiation Budget instrument.
Geophys. Res. Lett.
,
34
,
L02823
.
doi:10.1029/2006GL028229
.
Dewitte
,
S.
,
L.
Gonzalez
,
N.
Clerbaux
,
A.
Ipe
,
C.
Bertrand
, and
B.
De Paepe
,
2008
:
The Geostationary Earth Radiation Budget Edition 1 data processing algorithms.
Adv. Space Res.
,
41
,
1906
1913
.
Duce
,
R. A.
,
1995
:
Sources, distributions and fluxes of mineral aerosols and their relationship to climate.
Aerosol Forcing of Climate: Report of the Dahlem Workshop on Aerosol Forcing of Climate, R. J. Charlson and J. Heintzenberg, Eds., John Wiley & Sons, 43–72
.
Edwards
,
J. M.
, and
A.
Slingo
,
1996
:
Studies with a flexible new radiation code. I: Choosing a configuration for a large-scale model.
Quart. J. Roy. Meteor. Soc.
,
122
,
689
719
.
Engelstaedter
,
S.
, and
R.
Washington
,
2007
:
Atmospheric controls on the annual cycle of North African dust.
J. Geophys. Res.
,
112
,
D03103
.
doi:10.1029/2006JD007195
.
Flamant
,
C.
,
J-P.
Chaboureau
,
D. J.
Parker
,
C. M.
Taylor
,
J-P.
Cammas
,
O.
Bock
,
F.
Timouk
, and
J.
Pelon
,
2007
:
Airborne observations of the impact of a convective system on the planetary boundary layer thermodynamics and aerosol distribution in the inter-tropical discontinuity region of the West African Monsoon.
Quart. J. Roy. Meteor. Soc.
,
133
,
1175
1189
.
Forster
,
P.
, and
Coauthors
,
2007
:
Changes in atmospheric constituents and in radiative forcing.
Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 129–217. [Available online at http://www.ipcc.ch/ipccreports/ar4-wg1.htm]
.
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
, (
D17
).
20255
20273
.
Harries
,
J. E.
, and
Coauthors
,
2005
:
The Geostationary Earth Radiation Budget (GERB) project.
Bull. Amer. Meteor. Soc.
,
86
,
945
960
.
Haywood
,
J. M.
,
R. P.
Allan
,
I.
Culverwell
,
T.
Slingo
,
S.
Milton
,
J.
Edwards
, and
N.
Clerbaux
,
2005
:
Can desert dust explain the outgoing longwave radiation anomaly over the Sahara during July 2003?
J. Geophys. Res.
,
110
,
D05105
.
doi:10.1029/2004JD005232
.
Hsu
,
N. C.
,
S. C.
Tsay
,
M. D.
King
, and
J. R.
Herman
,
2004
:
Aerosol properties over bright- reflecting source regions.
IEEE Trans. Geosci. Remote Sens.
,
42
,
557
569
.
Husar
,
R. B.
,
J. M.
Prospero
, and
L. L.
Stowe
,
1997
:
Characterization of tropospheric aerosols over the oceans with the NOAA Advanced very high resolution radiometer optical thickness operational product.
J. Geophys. Res.
,
102
, (
D14
).
16889
16909
.
Johnson
,
B. T.
,
S. R.
Osborne
,
J. M.
Haywood
, and
M. A. J.
Harrison
,
2008
:
Aircraft measurements of biomass burning aerosol over West Africa during DABEX.
J. Geophys. Res.
,
113
,
D00C06
.
doi:10.1029/2007JD009451
.
Kiehl
,
J. T.
, and
H.
Rodhe
,
1995
:
Modelling geographical and seasonal forcing due to aerosols.
Aerosol Forcing of Climate: Report of the Dahlem Workshop on Aerosol Forcing of Climate, R. J. Charlson and J. Heintzenberg, Eds., John Wiley & Sons, 281–296
.
Lunt
,
D. J.
, and
P. J.
Valdes
,
2002
:
The modern dust cycle: Comparison of model results with observations and study of sensitivities.
J. Geophys. Res.
,
107
,
4669
.
doi:10.1029/2002JD002316
.
Martin
,
G. M.
,
M. A.
Ringer
,
V. D.
Pope
,
A.
Jones
,
C.
Dearden
, and
T. J.
Hinton
,
2006
:
The physical properties of the atmosphere in the New Hadley Centre Global Environmental Model (HadGEM1). Part I: Model description and global climatology.
J. Climate
,
19
,
1274
1301
.
Miller
,
R. L.
,
I.
Tegen
, and
J.
Perlwitz
,
2004
:
Surface radiative forcing by soil dust aerosols and the hydrological cycle.
J. Geophys. Res.
,
109
,
D04203
.
doi:10.1029/2003JD004085
.
Miller
,
R. L.
, and
Coauthors
,
2006
:
Mineral dust aerosols in the NASA Goddard Institute for Space Sciences ModelE atmospheric general circulation model.
J. Geophys. Res.
,
111
,
D06208
.
doi:10.1029/2005JD005796
.
Milton
,
S. F.
,
G.
Greed
,
M. E.
Brooks
,
J.
Haywood
,
B.
Johnson
,
R. P.
Allan
,
A.
Slingo
, and
W. M. F.
Grey
,
2008
:
Modeled and observed atmospheric radiation balance during the West African dry season: Role of mineral dust, biomass burning aerosol, and surface albedo.
J. Geophys. Res.
,
113
,
D00C02
.
doi:10.1029/2007JD009741
.
Schepanski
,
K.
,
I.
Tegen
,
B.
Laurent
,
B.
Heinhold
, and
A.
Macke
,
2007
:
A new Saharan dust source activation frequency map derived from MSG-SEVIRI IR-channels.
Geophys. Res. Lett.
,
34
,
L18803
.
doi:10.1029/2007GL030168
.
Shaffrey
,
L.
, and
Coauthors
,
2009
:
U.K. HiGEM: The new U.K. high-resolution global environmental model—Model description and basic evaluation.
J. Climate
,
22
,
1861
1896
.
Slingo
,
A.
, and
Coauthors
,
2006
:
Observations of the impact of a major Saharan dust storm on the atmospheric radiation balance.
Geophys. Res. Lett.
,
33
,
L24817
.
doi:10.1029/2006GL027869
.
Stier
,
P.
, and
Coauthors
,
2005
:
The aerosol-climate model ECHAM5-HAM.
Atmos. Chem. Phys.
,
5
,
1125
1156
.
Tegen
,
I.
,
B.
Reinhold
,
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
,
4345
4359
.
Todd
,
M. C.
, and
Coauthors
,
2008
:
Quantifying uncertainty in estimates of mineral dust flux: An intercomparison of model performance over the Bodélé Depression, northern Chad.
J. Geophys. Res.
,
113
,
D24107
.
doi:10.1029/2008JD010476
.
Washington
,
R.
, and
M. C.
Todd
,
2005
:
Atmospheric controls on mineral dust emission from the Bodélé Depression, Chad: The role of the low level jet.
Geophys. Res. Lett.
,
32
,
L17701
.
doi:10.1029/2005GL023597
.
Washington
,
R.
,
M. C.
Todd
,
S.
Engelstaedter
,
S.
Mbainayel
, and
F.
Mitchell
,
2006
:
Dust and the low-level circulation over the Bodélé Depression, Chad: Observations from BoDEx 2005.
J. Geophys. Res.
,
111
,
D03201
.
doi:10.1029/2005JD006502
.
Werner
,
M.
,
I.
Tegen
,
S. P.
Harrison
,
K. E.
Kohfeld
,
I. C.
Prentice
,
Y.
Balkanski
,
H.
Rodhe
, and
C.
Roelandt
,
2002
:
Seasonal and interannual variability of the mineral dust cycle under present and glacial climate conditions.
J. Geophys. Res.
,
107
,
4744
.
doi:10.1029/2002JD002365
.
Wilson
,
M. F.
, and
A.
Henderson-Sellers
,
1985
:
A global archive of land cover and soils data for use in general circulation climate models.
Int. J. Climatol.
,
5
,
119
143
.
Woodward
,
S.
,
2001
:
Modeling the atmospheric life cycle and radiative impact of mineral dust in the Hadley Centre climate model.
J. Geophys. Res.
,
106
,
18155
18166
.
Yoshioka
,
M.
,
N. M.
Mahowald
,
A. J.
Conley
,
W. D.
Collins
,
D. W.
Fillmore
,
C. S.
Zender
, and
D. B.
Coleman
,
2007
:
Impact of desert dust radiative forcing on Sahel precipitation: Relative importance of dust compared to sea surface temperature variations, vegetation changes, and greenhouse gas warming.
J. Climate
,
20
,
1445
1467
.
Zender
,
C. S.
,
R. L.
Miller
, and
I.
Tegen
,
2004
:
Quantifying mineral dust mass budgets: Systematic terminology, constraints and current estimates.
Eos, Trans. Amer. Geophys. Union
,
85
, (
Fall Meeting Suppl.
).
509
512
. .

Footnotes

* Current affiliation: Met Office, Exeter, United Kingdom

Corresponding author address: Dr. M. J. Woodage, Environmental Systems Science Centre, Harry Pitt Building, 3 Earley Gate, University of Reading, Reading, Berkshire RG6 6AL, United Kingdom. Email: mjw@mail.nerc-essc.ac.uk