1

Title 1

Improved dust representation in the Community Atmosphere Model 2

3

Authors 4

S. Albani [s.albani@cornell.edu] [1] [2] *corresponding author 5

N. M. Mahowald [mahowald@cornell.edu] [1] 6

A. T. Perry [atp42@cornell.edu] [1] 7

R. A. Scanza [ras486@cornell.edu] [1] 8

C. S. Zender [zender@uci.edu] [3] 9

N. G. Heavens [nicholas.heavens@hampton.edu] [4] 10

V. Maggi [valter.maggi@unimib.it] [2] 11

J. F. Kok [jasperkok@cornell.edu] [1] 12

B. L. Otto-Bliesner [ottobli@ucar.edu] [5] 13

14

[1] Department of Earth and Atmospheric Sciences, Cornell University, Ithaca NY, USA 15

[2] Department of Environmental Sciences, University of Milano-Bicocca, Milano, Italy 16

[3] Department of Earth System Science, University of California, Irvine, Irvine CA, USA 17

[4] Department of Atmospheric and Planetary Sciences, Hampton University, Hampton VA, USA 18

[5] National Center for Atmospheric Research, Boulder CO, USA 19

20

*corresponding author: 1123 Bradfield Hall, 14853 Ithaca NY, USA 21

22

23

24

25

26

2

Key points 27

Refined physical parameterizations of dust in the Community Atmosphere Model 28

Improved soil erodibility, size distributions, wet deposition and optics 29

Better representation of dust cycle, size distributions and radiative forcing 30

Abstract 31

Aerosol-climate interactions constitute one of the major sources of uncertainty in assessing 32

anthropogenic and glacial radiative forcing. Here we focus on improving the representation of 33

mineral dust in the Community Atmosphere Model and assessing the impacts of the improvements 34

in terms of direct effects on the radiative balance of the atmosphere. We simulated the dust cycle 35

while using different parameterization sets for dust emission, size distribution, and optical 36

properties. Comparing the results of these simulations with observations of concentration, 37

deposition, and aerosol optical depth allow us to refine the representation of the dust cycle and its 38

climate impacts. We find that the magnitude of the dust cycle is sensitive to the observational 39

datasets and size distribution chosen, and that the direct radiative forcing of dust is strongly 40

sensitive to the optical properties and size distribution used. Our results from simulations applying 41

the refined parameterization set indicate a net top of atmosphere direct dust radiative forcing of -42

0.22 ± 0.12 W/m2 for present day and -0.33 ± 0.18 W/m

2 at the Last Glacial Maximum. These 43

estimates are smaller than previous model simulations due to changes in size distribution, modeled 44

spatial distribution and optical parameters. 45

46

Index terms 47

Earth system modeling; Aerosols and particles; Radiation: transmission and scattering; Glacial. 48

49

Keywords 50

Mineral dust; dust size; dust optical properties; radiative forcing; last glacial maximum. 51

3

1. Introduction 52

Mineral dust is entrained into the atmosphere by the action of wind stress on the land surface. Most 53

sources of dust aerosols are arid or semiarid areas with low vegetation cover and easily erodible 54

soils or fine-grained loose surface deposits, [Gillette et al., 1998; Prospero et al., 2002]. With 55

typical atmospheric lifetimes of a few days [Alfaro and Gomes, 2001; Forster et al., 2007; An and 56

Zender, 2010], dust aerosols can travel long-distances [Prospero and Nees, 1986] and interact with 57

the climate system, both directly by absorbing and scattering short- (SW) and long-wave (LW) 58

radiation and indirectly through interactions with other aerosols and clouds [Miller and Tegen, 59

1998; Forster et al., 2007]. In addition, dust deposition to the surface can alter surface albedo 60

[Conway et al., 1996] and impact biogeochemical cycles [Martin et al. 1990; Mahowald, 2011]. 61

Sedimentary records from ice cores [e.g. Petit et al., 1999], marine [e.g. Rea, 1994] and terrestrial 62

deposits [e.g. Pye, 1995] showed that dust deposition varied greatly in a broad range of time scales, 63

most remarkably on the glacial-interglacial time scales [e.g. Lambert et al., 2008; Albani et al., 64

2012a], but also within interglacials [deMenocal et al., 2000] and in recent centuries/decades 65

[McConnell et al., 2007; Mulitza et al., 2010]. 66

The role of mineral aerosols in the climate system has motivated their inclusion in climate models 67

[e.g. Tegen and Fung, 1994; Schultz et al., 1998; Mahowald et al., 1999; Ginoux et al., 2001] and 68

monitoring of the present-day dust cycle, thanks to long-term in situ [e.g. Prospero and Lamb, 69

2003] or remote sensing observations [e.g. Holben et al., 1998; Kaufman et al., 2002], and by 70

intensive field campaigns [e.g. Heintzemberg, 2009]. In addition, extensive compilations of past 71

dust deposition rates from paleoarchives [Kohfeld and Harrison, 2001] served to test climate 72

simulations [Tegen et al., 2002], especially for the glacial-interglacial comparison [Mahowald et al., 73

2006a; Albani et al., 2012b]. 74

Aerosols interactions with climate still constitute one of the major uncertainties in assessing the 75

global average radiative forcing (RF) [Forster et al., 2007], and uncertainties in modeling the dust 76

4

cycle can be still quite large [Huneeus et al., 2011]. Besides the difficulties in reproducing dust 77

emissions and consequently the magnitude of the dust cycle [e.g. Cakmur et al., 2006], major 78

sources of uncertainty include the size distribution and optical properties of dust [Tegen, 2003]. 79

Despite the early recognition of the importance of dust size distributions in modeling [Schultz et al., 80

1998; Tegen and Lacis, 1996], the most recent reviews still place emphasis on devoting more 81

attention to this feature [Huneeus et al., 2011; Formenti et al., 2011; Maher et al., 2010], although 82

observational analyses emphasize the difficulties of accurately measuring size distribution [e.g. 83

Reid et al., 2003]. 84

Previous work with dust in the Community Climate System Model version 3 (CCSM3) [Mahowald 85

et al., 2006a] was extensively used and tested for a variety of experiments [e.g. Mahowald et al., 86

2006b; 2011] and on different time scales [Mahowald et al., 2010; Albani et al., 2012b]. 87

In this work we used the Community Earth System Model Version 1 (CESM1). More specifically 88

we used two release versions of the model: (1) the Community Atmosphere Model version 4 89

(CAM4) [Neale et al., 2010b; Gent et al., 2011] – in the framework of the Community Climate 90

System Model version 4 (CCSM4), which is also enclosed in CESM1 and (2) the Community 91

Atmosphere Model version 5 (CAM5) [Neale et al., 2010a; Liu et al., 2012]. We performed a series 92

of tests to optimize some key physics parameterizations related to dust (i.e. soil erodibility, dust 93

emission size distributions, wet deposition and optical properties). We show some relevant 94

improvements compared to the release versions, by extensively comparing with observations, 95

especially in terms of magnitude of the dust cycle, size distributions and optical properties. We also 96

show the CCSM4 case for the Last Glacial Maximum (LGM), corresponding to ~21,000 years 97

Before Present (21 ka BP). The effects of the improved dust parameterization on dust radiative 98

forcing and the sensitivity to specific changes are then described and compared to observations 99

(Section 3). 100

101

5

2. Methods 102

103

2.1 Model description 104

The CCSM4 is a general circulation climate model with atmosphere, land, ocean and sea ice 105

components linked by a flux coupler. The atmospheric component (CAM4) uses a finite-volume 106

dynamical core, rather than the CCSM3 spectral core [Gent et al. 2011]. The CCSM4 model is part 107

of the Paleoclimate Modelling Intercomparison Project Phase 3 (PMIP3) [Otto-Bliesner et al., 108

2009] and the Coupled Model Intercomparison Project Phase 5 (CMIP5) [Taylor et al., 2012] 109

experiments. 110

The CESM1 also includes a new version of the atmospheric model (CAM5). With respect to 111

CAM4, enhancements in physical parameterizations allow, in particular, the simulation of full 112

aerosol cloud interactions including cloud droplet activation by aerosols, precipitation processes due 113

to particle size dependent behavior and explicit radiative interaction of cloud particles, in turn 114

making it possible to simulate the cloud-aerosol indirect radiative effects [Neale et al., 2010a; Liu et 115

al., 2012]. CAM5 also participated in the CMIP5 experiments. In this study, we consider the dust in 116

both CAM4 and CAM5 versions of the model. 117

The dust model consists of three major components: (1) emission, (2) vertical transport, wet and dry 118

deposition and (3) optical scheme. Prognostic dust is available in CESM1, and the default dust 119

model follows the same treatment from CCSM3 [Mahowald et al., 2006a; Yoshioka et al., 2007], 120

but with different optical parameters [Hess et al., 1998], and a higher threshold Leaf Area Index 121

(0.3 instead of 0.1) based on observations of dust emission [Okin, 2008]. Dust emissions are 122

modeled using the source entrainment mechanism from Zender et al. [2003a], based on a saltation-123

sandblasting process dependent on modeled wind friction velocity, soil moisture and 124

vegetation/snow cover [Mahowald et al., 2006a]. Note that some of the dependence on soil 125

6

moisture relative to Zender et al. [2003a] was reduced in the CAM3 release, which is maintained in 126

this version. 127

CAM4 has a Bulk Aerosol Model (BAM) parameterization of the dust size distribution [Neale et 128

al., 2010b], where emission fluxes have a fixed size distribution partitioning for four size bins (bin1 129

= 0.1-1.0 μm; bin2 = 1.0-2.5 μm; bin3 = 2.5-5.0 μm; bin4 = 5.0-10.0 μm) [Mahowald et al., 2006a]. 130

In addition, differences in soils’ susceptibility to erosion (i.e. related to soil grains size and textures) 131

are summarized in a multiplicative parameter for the dust flux - the geomorphic soil erodibility 132

[Zender et al., 2003b] - based on the concept of preferential sources [Ginoux et al., 2001]. Dust 133

transport is controlled by the CAM4 tracer advection scheme [Neale et al., 2010b]. Modeled dry 134

deposition for dust includes gravitational and turbulent deposition processes [Zender et al., 2003a]. 135

Wet deposition of dust results from both convective and large scale rain and snow precipitation 136

simulated in CAM4, and is dependent on prescribed solubility and parameterized scavenging 137

coefficients [Mahowald et al., 2006a; Neale et al., 2010b]. Dust size distribution (i.e. the relative 138

proportions of mass in each of the size bins) evolves in time in response to transport and deposition 139

processes [Mahowald et al., 2006a; Albani et al., 2012b]. The dust optics is derived from Mie 140

calculations for the size distribution represented by each size bin [Neale et al., 2010b]. 141

For the LGM dust simulation we used initial conditions from a fully-coupled climate equilibrium 142

run with CAM4 [Brady et al., 2013], and an annual cycle of vegetation prescribed from BIOME4 143

[Kaplan et al., 2003] simulations, an approach similar to Mahowald et al. [2006a]. 144

Several modifications to the code were made to improve the model in the “new” versions (Table 1). 145

First, the distribution of dust in the 4 bins (or in the accumulation/coarse in the CAM5), was 146

changed according to Kok [2011]. In addition, in order to better match the observations (see Section 147

3.1), the wet deposition in CAM4 was changed by increasing the solubility for all dust particles 148

from 0.15 to 0.30, similar to that in CAM5 [Liu et al., 2012; Ghan et al., 2012], and using a larger 149

below cloud scavenging coefficient for large particles (Table 1) [e.g. Andronache, 2003]. 150

7

151

2.2 Optical parameters 152

For the release version of the model, SW optics from Hess et al. [1998] were accidentally included, 153

instead of the more accurate values from CCSM3 [Yoshioka et al., 2007; Mahowald et al., 2010] for 154

both the CAM4 and CAM5 optics. The release version optics are largely based on d’Almeida et al. 155

[1991]. Old measurements of the dust refractive index [d’Almeida et al., 1991 and Patterson et al., 156

1977 alike] yield optical properties that give dust the tendency to be too absorbing compared to 157

observations [Kaufman et al., 2001; Colarco et al., 2002; Sinyuk et al., 2003; Yoshioka et al., 2007; 158

Balkanski et al., 2007]. In addition, LW interactions, based on Volz [1973] refractive indices, were 159

turned off in the CAM4 default version, although they were included in the CAM5 release. 160

For this study, dust optics are calculated assuming Maxwell-Garnett theory for mixing optical 161

properties, assuming volume fractions of 47.6% quartz, 25% illlite, 25% montmorillonite, 2% 162

limestone, and 0.4% hematite. The assumed densities are 2660 kg/m3 for quartz, 2750 kg/m3 for 163

illite, 2350 kg/m3 for montmorillonite, 2710 kg/m3 for limestone, and 5260 kg/m3 for hematite. 164

The implied mass fractions of hematite and of iron are 0.80% and 0.56%, respectively. A 165

comparison of the assumed optical properties with available observations in regions 166

thought to be solely mineral aerosol is shown in Figure 1. 167

168

2.3 Observational datasets & diagnostics 169

Evaluating the model’s ability to reproduce different features of the dust cycle requires an adequate 170

set of observations and a protocol that makes comparisons consistent [Cakmur et al., 2006; 171

Huneeus et al., 2011]. The goal of the comparison is not to obtain an exact match with a specific 172

datum, but rather to enhance the ability to reproduce the overall features, given the relative 173

peculiarities, uncertainties and errors in both models and observations [e.g. Cakmur et al., 2006; 174

Albani et al., 2012b]. 175

8

In this work we focused on columnar (Aerosol Optical Depth (AOD) / column load) and surface 176

(atmospheric concentration and deposition flux) properties of the global dust cycle, by considering 177

their magnitude, spatial distribution, size distribution, seasonality and relation with the dust sources 178

(Table 2; Text S1; Table S1). 179

Observational constraints for columnar dust properties are based on the AERONET network 180

[Holben et al., 1998], from which we considered monthly averages for the AOD, and the average of 181

normalized size distributions [Dubovik and King, 2000], rebinned to match the model size scheme 182

(Text S1). We focus only on regions where dust is predicted to dominate (>50%) the aerosol optical 183

depth, as done in Mahowald [2007]. 184

Surface concentration measurements from high-volume filter collectors, shown as monthly 185

averages, were taken from the University of Miami Ocean Aerosols Network [e.g. Prospero and 186

Nees, 1986; Prospero et al., 1989]. In addition, annual averages from compiled datasets for station 187

data were also used [Mahowald et al., 2009], although sites in industrialized areas were not 188

included in the optimization described later. 189

Dust deposition flux data tables were compiled, based on merging and revising pre-existing datasets 190

(Text S1; Tables S2, S3), for modern climate [Ginoux et al., 2001; Tegen et al., 2002; Mahowald et 191

al., 2009; Lawrence and Neff, 2009], the LGM [Kohfeld and Harrison, 2001; Maher et al., 2010], 192

and interglacial climate [Kohfeld and Harrison, 2001; Maher et al., 2010]. In particular, we 193

included information or assumptions (Text S1; Table S2) on the dust flux size range in order to be 194

consistent with the model. 195

In addition, size distributions from mid-latitudes [Wu et al., 2009] and polar ice cores [Steffensen, 196

1997; Delmonte et al., 2004; McConnell et al., 2007] were reported at the model’s size bins for 197

comparison. 198

9

We also reviewed available information of dust provenance, for both present/interglacial and LGM 199

climates (Text S1; Table S1), in order to help refine the soil erodibility maps to represent the 200

relative intensity of different source areas. 201

202

2.4 Model optimization 203

To account for unrepresented source differences, soil erodibility maps were optimized for each 204

model configuration [e.g. Mahowald et al., 2006a; Albani et al., 2012b], by applying a set of scale 205

factors for different macro-areas, broadly corresponding to continents or subcontinents. 206

The soil erodibility map for current climate was first objectively optimized (similar to Mahowald et 207

al. [2006a]), but then modified to account for dust provenance information (Table S1; Figure S1). 208

The scale factors for the LGM soil erodibility maps, including explicitly glaciogenic sources 209

[Mahowald et al., 2006a], were based on a three criteria: matching (1) the observed LGM 210

deposition and (2) the LGM/interglacial deposition ratio (dataset largely based on DIRTMAP3 211

[Maher et al., 2010] – see Text S1 and Figure S2), as well as (3) considering the information on 212

dust provenance in the LGM (Table S1; Figure S3). Using LGM/interglacial observation and model 213

deposition ratios requires knowledge of the preindustrial dust flux, about which we have less 214

information than for either current or LGM [e.g. Mahowald et al., 2011]. We considered as a rough 215

approximation for interglacial dust deposition as half of that of current climate [Mulitza et al., 2008; 216

McConnell et al., 2007; Mahowald et al., 2010 and references therein]. 217

For the optimization, the logarithm in errors in AOD, concentration and deposition are all squared 218

and weighted equally, except that ice cores are weighted 10x (as believed to be more accurate), as 219

in Mahowald et al. [2006a]. We use the logarithm of the values, since we want to capture variability 220

in deposition and concentration that varies over 4 orders of magnitude. Without using a logarithm, 221

the optimization would be heavily weighted towards measurements with the highest values. 222

10

Because of errors in both modeling and observations, we only expect to match the data to about a 223

factor of 10, especially for deposition [e.g. Mahowald et al. 2011]. 224

The evaluation of optimization tests (Table 3) was based on extensively comparing the simulated 225

output with observations (Table 2). In Table 4 we summarize the performance of different cases. 226

We show the model’s ability to reproduce the overall magnitude of the dust cycle (median of the 227

modeled/observed ratio) and its variability range (correlation), by analyzing dust AOD, surface 228

concentration and deposition. We also report the evaluation of the seasonal cycle for dust AOD and 229

surface concentration (correlation). Finally, we report a metric (correlation of modeled and 230

observed size bin pairs) that allows comparing the modeled and observed shapes of dust size 231

distributions for column load and deposition at selected sites for the super-micron size range. 232

233

3. Results 234

235

3.1 Current climate dust 236

Several versions of the model were evaluated in this study (Table 2), but for simplicity, we show 237

only detailed results for one version (C4fn). The preliminary tests on the new dust size distribution 238

(C4sr-k) and wet deposition scavenging (C4sr-wk) parameterizations alone show an improved fit to 239

the observations far from the source areas for both the size distribution and the magnitude (Table 4; 240

Figure S4). This suggests that new theories on size distribution better match available observations 241

[e.g. Kok, 2011; Tanré et al., 2001]. 242

In general the release model’s ability to reproduce the magnitude of the dust cycle, both annual 243

mean and seasonal cycle, is quite variable. In general, using reanalysis winds results in better 244

reproduction of the annual cycle (Figures 2-7; Table 4). The modifications included here generally 245

improved the ability of the model to simulate the dust cycle, although some degradation at specific 246

sites is also seen (Figures 2-4; Table 4). The case with CAM4 using the new parameterizations 247

11

(C4fn) shows that the model does a good job of capturing the spatial variability of annual mean 248

deposition across 6 orders of magnitude (Figure 2). 249

Deposition is greatly overestimated at 2 sites (Colle del Lys, Colle Gnifetti), probably because of 250

unresolved orographic control on dust deposition on the Alps with the current model spatial 251

resolution. Underestimation of deposition in the Argentinean Pampa is surprising, because the 252

model matches well with offshore observations in the same region (James Ross Island), and is 253

possibly related to large uncertainties in the estimates from terrestrial sediments or to a slightly 254

inaccurate positioning of the relative dust source. 255

The order of magnitude of surface concentration of dust is reproduced quite well in the model 256

(Figure 3). 257

The model slightly underestimates observed AOD at model dust-dominated AERONET (Figure 4), 258

which is expected since we are not considering the full aerosol load that is recorded by the Sun-259

photometers. While South African observations of AOD are underestimated by the model, the 260

observations and model show similar values for surface concentrations. It is unlikely that the model 261

is underestimating dust emission from the region, because increasing South African emissions 262

would cause too much dust from the region compared to the general understanding based on global-263

scale satellite information [Prospero et al., 2002]. 264

In summary, the effects on the magnitude of the dust cycle and its spatial variability of the new 265

parameterization set compared to the default one are shown in Figure 5, which considers both the 266

cases with model (C4fr, C4fn) and reanalysis winds for CAM4 (Fr, Fn) and CAM5 (C5wr, C5wn). 267

While there is notable improvement over the default version for some variables, there is no 268

parameterization that uniformly improves the fit to all three variables. This is likely related to the 269

large uncertainties that still exist in constraining dust emissions, as well as the model’s meteorology 270

[e.g. Huneeus et al., 2011], and is consistent with previous general circulation model based dust 271

modeling [Tegen and Fung, 1994; Mahowald et al., 2006a]. It is noteworthy that for all cases 272

12

surface concentrations are biased high, while AOD and deposition are generally underestimated in 273

the model. At least part of the explanation for the disparate match to the observations for different 274

variables (see also Section 4.2) is due to the larger variability of the annual cycle of surface 275

concentration, with the tendency to over-predict the peak season at many sites (Figures 6 and 7). 276

This implies that vertical mixing in the model may not be accurate. Note that this is true whether the 277

model is driven by online winds, or reanalysis winds, and does not improve with the CAM5 versus 278

CAM4, which have quite different physics. This result is also seen in other cross-model 279

comparisons [e.g. Huneeus et al., 2011], suggesting issues with either measurement methods or 280

mixing in all the models. 281

The seasonal cycle of dust AOD (Figure 6) shows a reasonable agreement with AERONET 282

observations at most sites for the different model configurations (Table 4), although the magnitude 283

is not properly captured at all sites. The observed latitudinal shift in the positioning of the North 284

African dust plume with the season [Huneeus et al., 2011] is captured by our model, as shown by 285

the opposite seasonality displayed also for transatlantic transport by Northern stations such as Cape 286

Verde and Barbados compared to the southern stations such as Banizoumbou (Figure 6) and 287

Surinam (not shown). 288

Observations of the surface concentration (Figure 7) generally show a less smooth annual cycle than 289

the AOD, and the correlation with the model is quite poor in most cases (Table 4). The better 290

performance in simulating the climatology of vertically integrated parameters such as AOD 291

compared to surface concentration is a feature common to most climate models [Huneeus et al., 292

2011]. Note how in remote Pacific stations the new parameterization, in particular the wet 293

deposition changes, reduces the overestimation in the peak season. 294

Observationally-derived size distributions of dust in the atmospheric column from the dust-295

dominated AERONET sites provide a first-order constraint for this parameter close to the source 296

areas. The size distribution resulting from the new parameterization set undoubtedly shows a much 297

13

improved fit to the data (Figure 8). This holds true when using either modeled or reanalysis winds, 298

and for both stations located very close to the source area (e.g. Cairo University) and stations 299

located at a relatively small distance from the main dust sources (e.g. Cape Verde), as suggested 300

also by different shapes of the size distribution for the two groups of cases. 301

The evolution of dust size distribution deriving from medium- to long-range transport is contrasted 302

to observations from ice cores, by comparing dust deposition (Figure 9). In the case of medium 303

range transport (Asian ice cores) again there is a general improvement in representing the actual 304

size distribution. For long-range transport (polar ice cores) the situation is more complicated. While 305

some improvement may be arguably conveyed by the new parameterization set alone, model-data 306

comparisons suggest that both an accurate representation of dust size (new parameterization) and 307

meteorology (reanalysis winds) are necessary to provide a good fit to the observations. In general 308

we can state that the new parameterization set indeed provides a qualitatively more accurate 309

representation of the dust cycle. 310

The “new” cases tend to show a higher load and a shorter lifetime compared to the Aerocom 311

median estimates of 15.8 Tg and 4.6 days respectively [Huneeus et al., 2011], consistent with our 312

shift of mass away from the smallest size bin. Noteworthy are the differences in dust lifetimes 313

among different cases. The new wet deposition parameterization in CAM4 causes a shortening of 314

lifetimes, which in fact does not show up between the two cases using CAM5 nor among the CAM4 315

cases with changes either in size or optics (Table 5). In addition, lifetimes are longer for the C4wn 316

case compared to both C4fn and C5wn. For the latter, the short aerosol lifetimes are attributed to the 317

large wet removal rate, also due to internal mixing with more hygroscopic species [Liu et al., 2012]. 318

Differences between C4fn and C4wn are likely related to differences between model and reanalysis 319

winds. 320

321

3.2 Last Glacial Maximum dust 322

14

In the case of the LGM simulation (C4fn-lgm), the magnitude of modeled dust deposition is 323

representative of the paleodust observations in most cases, over the 4 orders of magnitude spanned 324

by the data (Figure 10). Dust emissions are 6500 Tg/y, about 2.3 times the current estimate with 325

C4fn (Table 5). About 20% of the emissions come from the glaciogenic sources, which is less than 326

the 1/3 in the CAM3 simulations [Mahowald et al., 2006a]. The LGM/current ratio for dust load 327

(1.7) is smaller than the emissions ratio, because of a shorter lifetime, which happens to be 328

comparable but higher (2.4 vs 2.0 days) than Mahowald et al. [2006a]. Although the dust load is 329

much lower (42 vs 77 Tg/y), the LGM/current ratios for CAM4 are very similar to CAM3 330

[Mahowald et al., 2006a] for emissions, load and lifetimes. The magnitude of the dust cycle in 331

terms of dust load is also higher than other model estimates, that indicate respectively 23 ± 14 Tg 332

[Werner et al., 2002] and 30.84 Tg [Takemura et al., 2009], with larger LGM vs 333

current/preindustrial ratios (2.39 and 2.8 respectively, compared to our 1.7). This is in line with 334

opposite trends for dust lifetimes, which are decreasing in the LGM in our simulations but increase 335

in Werner et al. [2002]. The decrease in lifetimes in our simulation is related to a reduced lofting, as 336

suggested for work with CAM3 [Mahowald et al., 2006a; Albani et al., 2012b]. In particular that is 337

related to the shift of the relative importance of dust sources to higher (northern) latitudes, where 338

the planetary boundary layer height is lower and decreases in glacial climate (Figure S5).Compared 339

to Lunt and Valdes [2002] our LGM vs current increase in dust load is larger (1.7 vs 1.39), although 340

in their simulations lifetimes are almost unchanged with climate and are much longer than ours (5.5 341

days). In terms of AOD our simulation indicates a much smaller value (0.04 vs 0.14) than Claquin 342

et al. [2003], as for the LGM/current ratio (1.7 vs 2.8). 343

344

3.3 Dust Radiative Forcing 345

The refined parameterization set described in the previous sections conveyed changes in dust 346

Radiative Forcing (RF) in the “new” cases simulations compared to the respective release versions 347

of the model. We compare both pairs with estimates, based on satellite-derived observations, of 348

15

clear-sky dust Top Of the Atmosphere (TOA) RF over the North Atlantic [Li et al., 2004] and the 349

Sahara [Zhang and Christopher, 2003; Patadia et al., 2008] (Table 6). It is evident how the new 350

cases are in much better agreement with observations, not only as obviously expected for the LW 351

(which was ignored for the release version of CAM4), but also for the SW RF. For our comparison 352

we ignore the estimates over the North Atlantic for the ”low” dust season (DJF), which are probably 353

contaminated by biomass burning aerosols as indicated by the low single scattering albedo [Li et al., 354

2004]. 355

An accurate partitioning of solar absorption between atmosphere and the surface, often not achieved 356

by climate models, is also important [Miller et al., 2004; Wild, 2008]. Estimates of surface RF 357

efficiency (-65 W/m2) over the North Atlantic by Li et al. [2004] are in good agreement with C4fn 358

(-68 W/m2), which improves over C4fr (-92 W/m2). In addition, observations of the clear-sky 359

surface incident radiation at a dust-dominated location, Tamanrasset, Algeria (299 W/m2) [Wild et 360

al., 2006], are also in better agreement with C4fn (299 W/m2) than C4fr (292 W/m2), indicating 361

that both the vertical structure of dust direct impacts on climate and its surface impacts of the new 362

parameterization are realistic. 363

The combination of absorption and reflection of solar radiation caused by the dust burden leads to a 364

negative SW RF at the surface, centered over and downwind of the major (and high albedo or 365

“brighter”) desert (North Africa and Arabian peninsula) dust sources (Figure 11), with absorption 366

causing a positive atmospheric column SW RF. The balance at the TOA is a generalized slightly 367

positive SW RF right above the major dust sources and a negative SW RF elsewhere (Figure 11a). 368

The radiation absorption and re-emission by dust in the LW spectrum causes a positive RF at 369

surface, associated with a negative atmospheric LW RF. At the TOA this yields a slightly positive 370

LW RF (Figure 11b). Resulting from both SW and LW contributions, the net balance shows a 371

generalized positive net RF in the atmosphere, whereas a somewhat similar structure at the surface 372

and at the TOA shows respectively a null and a slightly positive net RF over the major desert dust 373

16

sources and a negative net RF everywhere else (Figure 11c). This spatial pattern is similar to a study 374

using similar optical properties [Balkanski et al., 2007]. 375

A roughly similar spatial pattern is also shown by the cases using reanalysis winds and/or a 376

different version of the model, as also indicated by similarity in the features of the zonally averaged 377

net RF (Figure 12a-c). Nonetheless, the area characterized by a RF markedly different from zero is 378

less pronounced (both spatially and in magnitude) in both the C4wn and C5wn cases than in C4fn 379

(Figure 12a-c). Globally, this results in a more negative TOA RF for C4fn (-0.22 W/m2) compared 380

to the other 2 cases (Table 7). This feature stays true even if scaled by unit AOD, indicating 381

different RF efficiencies (Table 7). Compared to the new versions of the model, dust in CCSM3 382

[Yoshioka et al., 2007] had a larger TOA net RF (-0.61 W/m2), mostly due to the offsetting effects 383

of the SW and LW over land vs ocean surface [Yoshioka et al., 2007], with a strong negative 384

balance in the atmosphere over North African and Asian dust sources, from a shift in the sizes 385

(Figure 12d-f, solid lines). 386

Overall our net dust TOA RF is within the range of estimates from other models (Table 8), showing 387

a clear negative but relatively smaller in magnitude RF. 388

In the LGM simulation (C4fn-lgm) the zonal spatial pattern of surface RF, similar to C3-lgm 389

[Mahowald et al., 2006a], resembles the geographical distribution of the main dust sources, 390

including the mid/high latitude glaciogenic sources in both hemispheres, but shows a slight decrease 391

compared to C4fn over the desert areas centered at 15°N (Figure 12f). The net atmosphere RF is 392

positive and, as for current climate, does not show a contrasted net RF balance depending upon the 393

underlying surface being land or ocean, as was the case for CAM3 (Figure 12h). The balance at the 394

TOA resembles the surface features, except it shows a positive net RF at Northern high latitudes, 395

similar to C3-lgm (Figure 12d). Those areas seem to correspond to bright surfaces with high albedo, 396

such as the sea-ice covered Arctic Ocean and the southern margins of the Laurentide and Fenno-397

Scandian Ice Sheets, close to the major North American and European dust sources (Figure S6). 398

17

The average global net TOA RF for C4fn-lgm is -0.33 W/m2, which is lower than the C3-lgm 399

estimate (-1.61 W/m2), as a result of an almost halved dust load, and possibly to a lower RF 400

efficiency. This value is much lower than other model-based estimates (3.2 and 3.3 ± 0.8 W/m2 401

respectively) by Claquin et al. [2003] and Chylek and Lohmann [2008], but is higher than the 402

Takemura et al. [2009] estimate of -0.02 W/m2 (Table 8). Simpler estimates extrapolated from ice 403

cores data suggest an intermediate value of -1.9 ± 0.9 W/m2 [Köhler et al., 2010]. The 404

LGM/current ratio for the net TOA RF is 1.5, which is lower compared to the 2.7 from the previous 405

estimates from Yoshioka et al. [2007] and Claquin et al. [2003]. 406

407

4. Discussion 408

409

4.1 Model Sensitivity 410

In this section we discuss the model’s sensitivity to changes in dust size and optics, disentangling at 411

the same time the impacts that each one has on RF in the new version (C4fn) compared to the 412

release (C4fr) (e.g. similar to Perlwitz et al. [2001]). Comparing the net RF for C4fn and C4fr 413

shows that at TOA the net global balance is not too different for the two cases (-0.22 vs -0.27 414

W/m2), which also have the same RF efficiency (Table 7). Nonetheless, while the spatial patterns of 415

net TOA RF look rather similar around the main dust sources, C4fr shows a positive RF over the 416

Arctic (Figure 12g), and the 2 cases also show a quite different TOA vs surface heating (Figure 417

12g-i). The anomalous RF at Northern high latitudes in C4fr is likely due to the stronger long-range 418

transport to the Arctic of dust aerosols compared to the other cases. 419

The overlapping effects of changing dust optics and size distribution operate together throughout 420

the atmospheric column, but some of the effects emerge clearly if we differentiate the TOA, 421

atmospheric, and surface forcing. In the atmospheric column (Figure 12h) it is shown how the new 422

optics give a smaller positive RF because of the new extinction coefficients (C4fr/C4fn-ro vs C4fn-423

18

SW/C4fn-rs-SW), as expected (see Section 2.2). On the other hand the comparison among some 424

cases for the surface RF (Figure 12i), highlights a reduced negative RF (C4fr/C4fn-rs-SW vs C4fn-425

ro/C4fn-SW) as a consequence of the smaller SSA (Figure 1) arising from the new size partitioning 426

(Table 9). 427

The TOA vs surface forcing of the RF can be analyzed in comparison to other models that use 428

comparable optics [e.g. Sinyuk et al., 2003]. The net TOA/surface ratio for C4fn is 0.41, which is 429

similar to Miller et al. [2006] and Balkanski et al. [2007], respectively 0.48 and 0.42. Interestingly, 430

if we focus on the SW component alone, the TOA/surface ratio for C4fn is 0.31, which is 431

significantly lower than the 0.47 ratio in Miller et al. [2006]. The different loading size distributions 432

seems to be explanatory: 80% of mass is < 2 μm diameter in Miller et al. [2006], compared to 28% 433

< 2.5 μm in C4fn (Table 9) – for C4fn-rs the ratio is 0.41 with 44% < 2.5 μm – and highlights the 434

balancing role of the LW with the size shift, given the TOA net balance. 435

Based on these considerations, and on a possible underestimation of the submicron fraction 436

compared to observations for long-range transport (Table 10; see also Section 4.2), we also 437

conducted a sensitivity test (C4fn-s2) with a slight variation on Kok [2011] size distribution, in 438

order to double the emission relative contribution of the smaller, more reflective particles, which are 439

poorly constrained by observations (bin1: 0.1-1 µm). The spatial distribution pattern of net RF 440

looks similar to C4fn (Figure 12g-i), with a slightly higher negative global net TOA RF (-0.28 441

W/m2). This is associated with a higher RF efficiency than C4fn, but much lower than that of the 442

release size distribution (C4fn-rs) (Table 7), in line with the relative proportions of small particles 443

(Table 9). The size sensitivity case (C4fn-lgm-s2) shows a RF spatial pattern (Figure S7) similar to 444

C4fn-lgm (not shown), whereas the global net TOA RF is respectively -0.33 and -0.41 W/m2, again 445

indicating a higher but not huge RF efficiency for the case “enriched” in finer dust (Table 7), 446

suggesting that the radiative forcing is sensitive to the poorly constrained contribution of submicron 447

dust to the dust load. 448

19

In general, the spread of the RF plots from Figure 12g-i shows how variations in individual aspects 449

of the modeled dust cycle have a large impact on RF (Figure 12d-f). 450

451

4.2 Discussion of the uncertainties 452

The major source of uncertainty in dust models is represented by the magnitude and location of the 453

emissions [e.g. Cakmur et al., 2006; Shao et al., 2011; Huneeus et al., 2011]. To overcome this 454

limitation some tuning of model emissions is still necessary. As already described, we achieved this 455

by optimizing the soil erodibility maps to give best fit to the observations. 456

To estimate the uncertainty associated with the particular choice of the observational dataset, we 457

compare two cases where just the optimization algorithm was used [Mahowald et al., 2006a], 458

without further refining. In the first case the bulk value of Mass Accumulation Rates was used, in 459

the second case we just considered the fraction < 10 µm, consistent with our model size range 460

(Figure S7). According to Table 8, failing to match the observational and model’s size range can 461

yield a difference of ~20% in AOD estimates globally. 462

More in general, large uncertainties still exist on both sides of the dust size range we consider in our 463

model. Our parameterization assumes that dust is emitted only through the mechanism described by 464

Kok [2011], and within that framework we chose to limit our size to particles with diameters 465

smaller than 10 µm. Airborne dust particles larger than 10 µm, despite the uncertainties in the 466

measurement techniques [Reid et al., 2003], have indeed been observed both over the major dust 467

sources [Reid et al., 2003; Ryder et al., 2013] and over the sea in the North African continental 468

margin [Stuut et al., 2005], but they are known to get scavenged across the Atlantic Ocean [Maring 469

et al., 2003]. On the other hand, on the lower side of our size range, our sensitivity study 470

highlighted the potential for amplification of uncertainties in the submicron size range when it 471

comes to RF. Observations in this size range are challenging [Reid et al., 2003], but in general our 472

model simulations tend to underestimate the submicron fraction (Table 10), despite the fact that the 473

20

match with the very same set of observations in the 1-10 µm (where the observations are more 474

reliable) is good (Figure 9), at least with reanalysis winds (C4wn). 475

Another interesting outcome is the indication that the case using reanalysis winds (C4wn) has a 476

better ability to reproduce the evolution of size distribution with a shift towards finer particles with 477

long-range transport (Figure 9 and Table 1). This may be connected to the spatial features of RF, 478

which seems to suggest that the model winds tend to be more conducive to mass transport (and 479

associated RF) far from the source areas then the reanalysis winds (C4wr, C5wr) (Figure 12a-c), 480

likely through the larger dust bins (Figure 9). This would explain the lower RF efficiency of C4wr 481

and C5wr then C4fr (Table 7). 482

This variable behavior may be related to problems with the model vertical mixing, which is also 483

suggested by the inconsistency between the high bias of modeled surface concentrations compared 484

to the low bias of deposition and AOD (Figure 5). Some anomaly may be suggested also by looking 485

at the seasonal cycle simulated at Barbados (Figure 6 and Figure 7). The column-integrated 486

seasonal cycle (AOD) is well reproduced by the model (Figure 6). On the other hand the C4fn (but 487

not C4wr or C4wn) fails to reproduce the observations of dust concentration at the surface (Figure 488

7), but the seasonal cycle does mimic the surface observations at higher levels i.e. 2-3 vertical levels 489

atop (not shown). 490

In addition, as discussed (Section 3.1), the magnitude of the dust cycle may be slightly 491

underestimated in our model if we consider the AOD and deposition data compared to the surface 492

concentration data. (Figure 5). 493

494

4.3 Quantification of the uncertainties 495

We consider three synthetic parameters to give a quantitative estimate of the uncertainty associated 496

to dust RF with our model. 497

21

First, the impact of the choice of the observational size range for constraining the emissions was 498

already estimated as 20% (Section 4.2). Second, the uncertainty associated with the magnitude of 499

the dust cycle and the spatial distribution of dust in response to the meteorology is estimated as the 500

coefficient of variation (standard deviation/mean*100) of the net TOA RF from the C4fn, C4wn, 501

and C5wn group of runs, which is 48%. Third we consider the uncertainty related to the size 502

distribution and optics, again calculating the coefficient of variation of the couple C4fn and C4fn-s2 503

for the net TOA RF (17%). For the LGM we assume the uncertainties for the first two are the same 504

as for the current climate, but we consider the C4fn-lgm, C4fn-lgm-s2 couple for the estimation of 505

the size/optics uncertainty (15%). 506

The total uncertainty is derived as the root of the quadrature sum of the three elements described, 507

resulting in an overall uncertainty of 55%. If we apply this to the reference estimate with C4fn and 508

C4fn-lgm, this results in a net TOA RF for dust of -0.22 ± 0.12 W/m2 for the current climate and -509

0.33 ± 0.18 W/m2 for the LGM. 510

These estimates of uncertainty are related just to the “internal” variability of the model used, and do 511

not account for biases/errors related to processes not considered by the model and the 512

parameterizations used. Those additional uncertainties may be accounted for by comparing other 513

models, although there may also be biases across all models for these estimates. 514

515

5. Conclusions 516

In this work we have refined the dust parameterization set included in CAM4 and CAM5, parts of 517

the CESM model. We restored new generation dust optics [e.g. Yoshioka et al., 2007] based on 518

realistic absorption coefficients [e.g. Kaufmann et al., 2001; Sinyuk et al., 2003], compared to the 519

obsolete optical properties that were erroneously put in the release versions of CESM. In addition, 520

we refined the CAM4 wet deposition parameterizations [e.g. Andronache, 2003] and we updated 521

the soil erodibility maps [Mahowald et al., 2006a] in order to give a better representation of the 522

22

magnitude and spatial distribution of the dust cycle, also accounting for the limited available 523

information of dust provenance at remote sites. Notably, for scaling the soil erodibility maps and for 524

model-observations comparisons we considered observations consistent with the model’s size range 525

[e.g. Cakmur et al., 2006], and estimated a 20% error if ignoring this constraint. Finally, the most 526

novel change we adopted is a new size distribution for dust emissions, based on Kok [2011]. 527

Our results show how the model is able to capture the overall magnitude and spatial variability of 528

dust cycle, but at the same time highlight the difficulty in improving with respect to previous work. 529

As widely recognized, accurately estimating dust emissions in terms of magnitude, timing and 530

geographical location still remains a major challenge [e.g. Huneeus et al., 2011]. A significant 531

improvement over previous work is due to the new size distribution, which shows a good agreement 532

with observations especially for the super-micron size range over a wide range of spatial scales, 533

from the dust sources to remote sites. 534

The analysis of the effects that the refined parameterization set developed in this work has on dust 535

RF was analyzed in its components and the uncertainties internal to the model processes were 536

evaluated. The comparison with observational-based estimates of dust RF efficiency shows that our 537

model is able to reproduce these features realistically, both over North Africa [Zhang and 538

Christopher, 2003; Patadia et al., 2008] and the North Atlantic Ocean [Li et al., 2004], and 539

improves the simulated incident solar radiation balance at surface in desert areas [Wild et al., 2006] 540

compared to the release version. 541

Uncertainties in dust optical properties [Perlwitz et al., 2001], also related to the size distribution of 542

dust [Tegen and Lacis, 1996; Sokolik and Toon, 1998], especially for the smaller but also the upper 543

edges of the model (0.1-10 µm) size distributions [Reid et al., 2003], will combine with 544

uncertainties in the magnitude and spatial distribution of dust to render the balance between dust 545

LW and SW interactions very sensitive, determining the wide range of estimates [Forster et al., 546

2007] for dust RF at the TOA. 547

23

After considering the uncertainties in our work we indicate a value of -0.22 ± 0.12 W/m2 as an 548

estimate for the dust global net TOA RF for current climate, and -0.33 ± 0.18 W/m2 for the LGM 549

climate in our model. The next step will be to evaluate the climate impacts of dust at the 550

equilibrium under the new parameterization set. 551

24

552

Acknowledgements 553

We acknowledge the support of NSF-0932946 and 1003509, 0745961, 1137716, and doeDE-554

SC00006735. S. Albani acknowledges funding from ‘‘Dote ricercatori’’: FSE, Regione Lombardia. 555

The AERONET data were retrieved online at http://aeronet.gsfc.nasa.gov. We thank the Principal 556

Investigators and their staff for establishing and maintaining the sites used in this work. We 557

gratefully acknowledge Joseph Prospero for providing dust surface concentration data from in situ 558

measurements from the University of Miami Ocean Aerosol Network. These simulations were 559

conducted at the National Center for Atmospheric Research’s Computation Information Systems 560

Laboratory, an NSF funded facility. 561

562

References 563

564

Albani S., B. Delmonte, V. Maggi, C. Baroni, J.R. Petit, B. Stenni, C. Mazzola, and M. Frezzotti 565

(2012a). Interpreting last glacial to Holocene dust changes at Talos Dome (East Antarctica): 566

implications for atmospheric variations from regional to hemispheric scales. Clim. Past, 8, 741-750, 567

doi: 10.5194/cp-8-741-2012. 568

569

Albani S., N. M. Mahowald, B. Delmonte, V. Maggi, and G. Winckler (2012b). Comparing 570

modeled and observed changes in mineral dust transport and deposition to Antarctica between the 571

Last Glacial Maximum and current climates. Clim. Dyn., 38, 9, 1731-1755, doi: 10.1007/s00382-572

011-1139-5. 573

574

Alfaro S.C., and L. Gomes (2001). Modeling mineral aerosol production by wind erosion: Emission 575

intensities and aerosol size distributions in source areas. J. Geophys. Res., 106, D16, 18,075-18,084. 576

577

25

Andronache, C. (2003). Estimated variability of below-cloud aerosol removal by rainfall for 578

observed aerosol size distributions. Atmos. Chem. Phys., 3, 131–143. 579

580

Balkanski, Y., M. Schultz, T. Claquin, and S. Guibert (2007). Reevaluation of mineral aerosol 581

radiative forcings suggests a better agreement with satellite and AERONET data. Atmos. Chem. 582

Phys., 7, 81-95. 583

584

Brady, E. C., B. L. Otto-Bliesner, J. E. Kay, and N. Rosenbloom (2012). Sensitivity to glacial 585

forcing in the CCSM4. J. Climate, doi:10.1175/JCLI-D-11-00416.1, in press. 586

587

Cakmur, R. V., Miller, R. L., Perlwitz, J., Geogdzhayev, I. V., Ginoux, P., Koch, D., Kohfeld, K. 588

E., Tegen, I., and Zender, C. S. (2006). Constraining the magnitude of the global dust cycle by 589

minimizing the difference between a model and observations. J. Geophys. Res., 111, D06207, 590

doi:10.1029/2005JD005791. 591

592

Chylek, P., and U. Lohmann (2008). Aerosol radiative forcing and climate sensitivity deduced from 593

the Last Glacial Maximum to Holocene transition. Geophys. Res. Lett., 35, L04804, 594

doi:10.1029/2007GL032759. 595

596

Claquin, T., C. Roelandt, K. E. Kohfeld, S. P. Harrison, I. Tegen, I. C. Prentice, Y. Balkanski, G. 597

Bergametti, M. Hansson, N. Mahowald, H. Rodhe, and M. Schultz (2003). Radiative forcing of 598

climate by ice-age atmospheric dust. Clim. Dyn., 20, 193-202. 599

600

Colarco, P. R., O. B. Toon, O. Torres, and P. J. Rasch (2002). Determining the UV imaginary index 601

of refraction of Saharan dust particles from Total Ozone Monitoring data using a three-dimensional 602

model of dust transport. J. Geophys. Res., 107, D16, 4289. 603

26

604

Conway, H., A. Gades, and C. F. Raymond (1996). Albedo of dirty snow during conditions of 605

meltwater. Water Res. Research, 32, 6, 1713–1718. 606

607

d’Almeida, G. A., P. Koepke, and P. E. Shettle (1991). Atmospheric Aerosols: Global Climatology 608

and Radiative Chracteristics. A. Deepak Publishing, 561 pp. 609

610

Delmonte, B., J.-R. Petit, K. K. Andersen, I. Basile-Doelsch, V. Maggi, and V. Y. Lipenkov (2004). 611

Dust size evidence for opposite regional atmospheric circulation changes over east Antarctica 612

during the last climatic transition. Clim. Dyn., 23,427–438. doi:10.1007/s00382-004-0450-9. 613

614

deMenocal, P., J. Ortiz, T. Guilderson, J. Adkins, M. Sarnthein, L. Baker, and M. Yarusinsky 615

(2000). Abrupt onset and termination of the African Humid Period:: rapid climate responses to 616

gradual insolation forcing Original Research Article. Quat. Sci. Rev., 19, 1–5, 347-361. 617

618

Dubovik, O., and M. D. King (2000). A flexible inversion algorithm for retrieval of aerosol optical 619

properties from Sun and sky radiance measurements. J. Geophys. Res., 105, D16, 20,673-20,696. 620

621

Formenti, P., L. Schütz, Y. Balkanski, K. Desboeufs, M. Ebert, K. Kandler, A. Petzold, 622

D. Scheuvens, S. Weinbruch, and D. Zhang (2011). Recent progress in understanding physical and 623

chemical properties of African and Asian mineral dust. Atmos. Chem. Phys., 11, 8231-8256. 624

625

Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D. W. Fahey, J. Haywood, J. Lean, D. 626

C. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz, R. V., and R. V. Dorland (2007). 627

Changes in Atmospheric Constituents and in Radiative Forcing, in: Climate Change 2007: The 628

Physical Science Basis, Contribution of Working Group I to the Fourth Assessment Report of the 629

27

Intergovernmental Panel on Climate Change, eds. Solomon, S., Qin, D., Manning, M., Chen, Z., 630

Marquis, M., Averyt, K.B., Tignor, M., and Miller, H.L., Cambridge University Press, Cambridge, 631

UK and New York, NY, USA, 130–234, 2007. 632

633

Gent, P. R., G. Danabasoglu, L. J. Donner, M. M. Holland, E. C. Hunke, S. R. Jayne, D. M. 634

Lawrence, R. B. Neale, P. J. Rasch, M. Vertenstein, P. H. Worley, Z.-L. Yang, and M. Zhang 635

(2011). The Community Climate System Model Version 4. J. Climate, 24, 19, 4973-4991. 636

637

Ghan, S. J., X. Liu, R. C. Easter, R. Zaveri, P. J. Rasch, and J.-H. Yoon (2012). Toward a minimal 638

representation of aerosols in climate models: comparative decomposition of aerosol direct, 639

semidirect, and indirect radiative forcing. J. Climate, 25, 6461-6476. 640

641

Ginoux P., M. Chin, I. Tegen, J. Prospero, B. Holben, O. Dubovik, S.-J. Lin (2001). Sources and 642

distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res., 106(D17), 643

20255-20273. 644

645

Gillette, D. A., B. Marticorena, and G. Bergametti (1998). Change in the aerodynamic roughness 646

height by saltating grains: Experimental assessment, test of theory, and operational 647

parameterization. J. Geophys. Res., 103(D6), 6203-6209. 648

649

Han, Q. and C. S. Zender (2010). Desert dust aerosol age characterized by mass-age tracking of 650

tracers. J. Geophys. Res. 115, D22201, doi:10.1029/2010JD014155. 651

652

Heintzenberg, J. (2009). The SAMUM-1 experiment over Southern Morocco: overview and 653

introduction. Tellus B, 61, 2–11. 654

655

28

Hess, M., P. Koepke, I. Schult, 1998: Optical Properties of Aerosols and Clouds: The Software 656

Package OPAC. Bull. Amer. Meteor. Soc., 79, 831–844. 657

658

Holben B. N., T. F. Eck, I. Slutsker, D. Tanré, J. P. Buis, A. Setzer, E. Vermote, J. A. Reagan, Y. J. 659

Kaufman, T. Nakajima, F. Lavenu, I. Jankowiak, and A. Smirnov (1998). AERONET - A Federated 660

Instrument Network and Data Archive for Aerosol Characterization. Remote Sens. Env., 66, 1-16. 661

662

Huneeus, N., M. Schulz, Y. Balkanski, J. Griesfeller, J. Prospero, S. Kinne, S. Bauer, O. Boucher, 663

M. Chin, F. Dentener, T. Diehl, R. Easter, D. Fillmore, S. Ghan, P. Ginoux, A. Grini, L. Horowitz, 664

D. Koch, M. C. Krol, W. Landing, X. Liu, N. Mahowald, R. Miller, J.-J. Morcrette, G. Myhre, J. 665

Penner, J. Perlwitz, P. Stier, T. Takemura, and C. S. Zender (2011). Global dust model 666

intercomparison in AeroCom phase I. Atmos. Chem. Phys., 11, 7781-7816, 2011 667

668

Jacobson, M. Z. (2001). Global direct radiative forcing due to multicomponent anthropogenic and 669

natural aerosols. J. Geophys. Res., 106(D2), 1551–1568. 670

671

Kaplan, J. O., N. H. Bigelow, I. C. Prentice, S. P. Harrison, P. J. Bartlein, T. R. Christensen, W. 672

Cramer, N. V. Matveyeva, A. D. McGuire, D. F. Murray, V. Y. Razzhivin, B. Smith, D. A. Walker, 673

P. M. Anderson, A. A. Andreev, L. B. Brubaker, M. E. Edwards, and A. V. Lozhkin (2003). 674

Climate change and Arctic ecosystems: 2. Modeling, paleodata-model comparisons, and future 675

projections, J. Geophys. Res., 108, 8171, doi:10.1029/2002JD002559, D19. 676

677

Kaufmann, Y. J., D. Tanré, O. Dubovik, A. Karnieli, L. A. Remer (2001). Absorption of sunlight by 678

dust as inferred from satellite and ground-based remote sensing. Geophys. Res. Lett., 28, 1479-679

1482. 680

681

29

Kohfeld, K. E., and S. P. Harrison (2001). DIRTMAP: the geological record of dust. Earth-Sci. 682

Rev., 54, 81–114. 683

684

Köhler, P., R. Bintanja, H. Fischer, F. Joos, R. Knutti, G. Lohmann, and V. Masson-Delmotte 685

(2010). What caused Earth's temperature variations during the last 800,000 years? Data-based 686

evidence on radiative forcing and constraints on climate sensitivity. Quat. Sci. Rev., 29, 1–2, 129-687

145. 688

689

Kok, J. F. (2011). A scaling theory for the size distribution of emitted dust aerosols suggests climate 690

models underestimate the size of the global dust cycle. Proc. Natl. Acad. Sci. U.S.A., 108, 1016-691

1021, doi:10.1073/pnas.1014798108. 692

693

Lambert F., B. Delmonte, J.-R. Petit, M. Bigler, P.R. Kaufmann, M.A. Hutterli, T.F. Stocker, U. 694

Ruth, J.P. Steffensen, V. Maggi (2008). Dust-climate couplings over the past 800,000 years from 695

the EPICA Dome C ice core. Nature, 452, doi:10.1038/nature06763. 696

697

Lawrence, C. R., and J. C. Neff (2009). The contemporary physical and chemical flux of Aeolian 698

dust: A synthesis of direct measurements of dust deposition. Chem. Geol., 257, 46-63. 699

700

Li, F., A. M. Vogelmann, and V. Ramanathan (2004).Saharan dust aerosol radiative forcing 701

measured from space. J. Climate, 17, 2558-2571. 702

703

Liao, H., J.H. Seinfeld, P.J. Adams, and L.J. Mickley (2004). Global radiative forcing of coupled 704

tropospheric ozone and aerosols in a unified general circulation model. J. Geophys. Res., 109, 705

D16207, doi:10.1029/2003JD004456. 706

707

30

Liu, X., R. C. Easter, S. J. Ghan, R. Zaveri, P. J. Rasch, X. Shi, J.-F. Lamarque, A. Gettelman, H. 708

Morrison, F. Vitt, A. Conley, S. Park, R. Neale, C. Hannay, A. M. L. Ekman, P. Hess, N. 709

Mahowald, W. Collins, M. J. Iacono, C. S. Bretherton, M. G. Flanner, and D. Mitchell, D. (2012). 710

Toward a minimal representation of aerosols in climate models: description and evaluation in the 711

Community Atmosphere Model CAM5. Geosci. Model Dev., 5, 709-739, doi:10.5194/gmd-5-709-712

2012. 713

714

Lunt, D. J., and P. J. Valdes (2002). Dust deposition and provenance at the Last Glacial Maximum 715

and present day. Geophys. Res. Lett., 29, 22, 2085. 716

717

Maher, B, A., J. M. Prospero, D. Mackie, D. Gaiero, P. P. Hesse, and Y. Balkanski (2010). Global 718

connections between aeolian dust, climate and ocean biogeochemistry at the present day and at the 719

last glacial maximum. Earth Sci. Rev., 99, 61-97. 720

721

Mahowald, N. (2011). Aerosol indirect effects on biogeochemistry and climate. Science, 334, 794, 722

doi:10.1126/science.1207374. 723

724

Mahowald, N., S. Albani, S. Engelstaedter, G. Winckler, and M. Goman (2011). Model insight into 725

glacial-interglacial paleodust records. Quat. Sci. Rev., 30(7–8), 832–854. 726

727

Mahowald, N. M., S. Engelstaedter, C. Luo, A. Sealy, P. Artaxo, C. Benitez-Nelson, S. Bonnet, Y. 728

Chen, P. Y. Chuang, D. D. Cohen, F. Dulac, B. Herut, A. M. Johansen, N. Kubilay, R. Losno, W. 729

Maenhaut, A. Paytan, J. M. Prospero, L. M. Shank, and R. L. Siefert (2009). Atmospheric Iron 730

Deposition: Global Distribution, Variability, and Human Perturbations. Annu. Rev. Mar. Sci., 1, 731

245–278. 732

733

31

Mahowald N., K. Kohfeld, M. Hansson, Y. Balkanski, S. P. Harrison, I. C. Prentice, M. Schulz, H. 734

Rodhe (1999). Dust sources and deposition during the last glacial maximum and current climate: A 735

comparison of model results with paleodata from ice cores and marine sediments. J. Geophys. Res., 736

104, D13, 15895-15916. 737

738

Mahowald, N. M., D. R. Muhs, S. Levis, P. J. Rasch, M. Yoshioka, C. S. Zender, and C. Luo 739

(2006a). Change in atmospheric mineral aerosols in response to climate: Last glacial period, 740

preindustrial, modern, and doubled carbon dioxide climates. J. Geophys. Res., 111, D10202. 741

742

Mahowald, N., M. Yoshioka, W. Collins, A. Conley, D. Fillmore, and D. Coleman (2006b). 743

Climate response and radiative forcing from mineral aerosols during the glacial maximum, pre-744

industrial, current and doubled-carbon dioxide climates. Geophys. Res. Lett., 33, L20705, 745

doi:10.1029/2006GL026126. 746

747

Mahowald N. M., S. Kloster, S. Engelstaedter, J. K. Moore, S. Mukhopadhyay, J. McConnell, S. 748

Albani, S. Doney, A. Bhattacharya, M. A. J. Curran, M. G. Flanner, F. M. Hoffman, D. M. 749

Lawrence, K. Lindsay, P. A. Mayewski, J. Neff, D. Rothenberg, E. Thomas, P. E. Thornton, and C. 750

S. Zender (2010). Observed 20th century desert dust variability: impact on climate and 751

biogeochemistry. Atmos. Chem.Phys., 10, 22, 10875-10893. 752

753

Maring, H., D. L. Savoie, M. A. Izaguirre, and L. Custals (2003). Mineral dust aerosols size 754

distribution change during atmospheric transport. J. Geophys. Res., 108, D19, 8592. 755

756

Martin, J. H., R. M. Gordon, and S. E. Fitzwater (1990). Iron in Antarctic waters. Nature, 353, 757

6340, 156-158. 758

759

32

McConnell, J. R., A. J. Aristarain, J. R. Banta, P. R. Edwards, and J. C. Simoes (2007). 20th-760

Century doubling in dust archived in an Antarctic Peninsula ice core parallels climate change and 761

desertification in South America. Proc. Natl. Acad. Sci. U.S.A., 104, 14, 5743–5748. 762

763

Miller, R. L., J. P. Perlwitz, and I. Tegen (2004). Feedback upon dust emission by dust radiative 764

forcing through the planetary boundary layer. J. Geophys. Res., 109, D24209, 765

doi:10.1029/2004JD004912. 766

767

Miller, R. L., R. V. Cakmur, J. Perlwitz, I. V. Geogdzhayev, P. Ginoux, D. Koch, K. E. Kohfeld, C. 768

Prigent, R. Ruedy, G. A. Schmidt, and I. Tegen (2006). Mineral dust aerosols in the NASA 769

Goddard Institute for Space Sciences ModelE atmospheric general circulation model. J. Geophys. 770

Res., 111, D06208. 771

772

Mulitza, S., D. Heslop, D. Pittauerova, H. W. Fischer, I. Meyer, J.-B. Stuut, M. Zabel, G. 773

Mollenhauer, J. A. Collins, H. Kuhnert, and M. Schulz (2010). Increase in African dust flux at the 774

onset of commercial agriculture in the Sahel region. Nature, 466, 226–228, 775

doi:10.1038/nature09213. 776

777

Myhre, G., and F. Stordal (2001). Global sensitivity experiments of the radiative forcing due to 778

mineral aerosols. J. Geophys. Res., 106, 18193–18204. 779

780

Neale, R. B., C.-C. Chen, A. Gettelman, P. H. Lauritzen, S. Park, D. L. Williamson, A. J. Conley, 781

R. Garcia, D. Kinnison, J.-F. Lamarque, D. Marsh, M. Mills, A. K. Smith, S. Tilmes, F. Vitt, H. 782

Morrison, P. Cameron-Smith, W. D. Collins, M. J. Iacono, R. C. Easter, S. J. Ghan, X. Liu, P. J. 783

Rasch, and M. A. Taylor (2010a). Description of the NCAR Community Atmosphere Model (CAM 784

5.0). NCAR TECHNICAL NOTE TN-486, pp. 268. 785

33

786

Neale, R. B., J. H. Richter, A. J. Conley, S. Park, P. H. Lauritzen, A. Gettelman, D. L. Williamson, 787

P. J. Rasch, S. J. Vavrus, M. A. Taylor, W. D. Collins, M. Zhang, and S.-J.Lin (2010b). Description 788

of the NCAR Community Atmosphere Model (CAM 4.0). NCAR Tech. Note TN-485, pp.212. 789

790

Okin, G. S. (2008). A new model of wind erosion in the presence of vegetation. J.Geophys. Res., 791

Earth Surface, 113, F02S10, doi:10.1029/2007JF000758. 792

793

Otto-Bliesner, B. L, S. Joussaume, S. P. Harrison, A. Abe-Ouchi, and P. Braconnot (2009). 794

Modeling and Data Syntheses of Past Climates. Eos, Vol. 90, No. 11. 795

796

Patadia, F., E.-S. Yang, and S. A. Christopher (2008). Does dust change the clear sky top of the 797

atmosphere shortwave flux over high surface reflectance regions? Geophys. Res. Lett., 36, L15825. 798

799

Patterson, E. M., D. A. Gillette, and B. H. Stockton (1977). Complex index of refraction between 800

300 and 700 nm of Saharan dust aerosols. J. Geophys. Res., 82, 16,595-16,969. 801

802

Perlwitz,, J., I. Tegen, and R. L. Miller (2001). Interactive soil dust aerosol model in the GISS 803

GCM 1: Sensitivity of the soil dust cycle to radiative properties of soil dust aerosols. J. Geophys. 804

Res., doi: 10.1029/2000JD900668 805

806

Petit, J.-R., J. Jouzel, D. Raynaud, N.I. Barkov, J.-M. Barnola, I. Basile, M. Bender, J. Chappellaz, 807

M. Davisk, G. Delaygue, M. Delmotte, V.M. Kotlyakov, M. Legrand, V.Y. Lipenkov, C. Lorius, L. 808

Pepin, C. Ritz, E. Saltzmank, and M. Stievenard (1999). Climate and atmospheric history of the 809

past 420,000 years from the Vostok ice core, Antarctica. Nature, 399, 429-436. 810

811

34

Prospero, J. M., M. Uematsu, and D. L. Savoie (1989). Mineral aerosol transport to the Pacific 812

Ocean, in Chemical Oceanography, vol. 10, edited by J. P. Riley, R. Chester and R. A. Duce, pp. 813

188–218, Academic, San Diego, Calif., 1989. 814

815

Prospero J.M., P. Ginoux, O. Torres, S. E. Nicholson, and T. E. Gill (2002). Environmental 816

characterization of global sources of atmospheric soil dust identified with the Nimbus 7 Total 817

Ozone Mapping Spectrometer (TOMS) Absorbing Aerosol Product. Rev. Geophys., 40(1), 1002, 818

doi:10.1029/2000RG000095. 819

820

Prospero, J. M., and R. T. Nees (1986). Impact of the North African drought and El Niño on 821

mineral dust in the Barbados trade winds. Nature, 320, 735–738. 822

823

Prospero, J.M., and P.J. Lamb (2003). African Droughts and Dust Transport to the Caribbean: 824

Climate Change Implications. Science, 302, 5647, 1024-1027, DOI: 10.1126/science.1089915. 825

826

Pye, K. (1995). The origin, nature and accumulation of loess. Quat. Sci. Rev., 14, 7-8, 653-667. 827

828

Rea, D. K. (1994). The paleoclimatic record provided by eolian deposition in the deep sea: the 829

geologic history of wind. Rev. Geophys., 32, 2, 159-195. 830

831

Reddy, M.S., O. Boucher, Y. Balanski, and M. Schulz (2005). Aerosol optical depths and direct 832

radiative perturbations by species and source type. Geophys. Res. Lett., 32, L12803, 833

doi:10.1029/2004GL021743. 834

835

Reid, J. S., H. H. Jonsson, H. B. Maring, A. Smirnov, D. L. Savoie, S. S. Cliff, E. A. Reid, J. M. 836

Livingston, M. M. Meier, O. Dubovik, and S.-C. Tsay (2003). Comparison of size and 837

35

morphological measurements of coarse mode dust particles from Africa. J. Geophys. Res., 108, 838

D19, 8593, doi:10.1029/2002JD002485. 839

840

Ryder, C. L., E. J. Highwood, P. D. Rosenberg, J. Trembath, J. K. Brooke, M. Bart, A. Dean, J. 841

Croisier, J. Dorsey, H. Brindley, J. Banks, J. H. Marsham, J. B. McQuaid, H. Sodemann, and R. 842

Washington (2013). Optical properties of Saharan dust aerosol and contribution from the coarse 843

mode as measured during the Fennec 2011 aircraft campaign. Atm. Chem. Phys., 13, 303-325. 844

845

Shao, Y., K.-H. Wyrwoll, A. Chappell, J. Huang, Z. Lin, G. H. McTainsh, M. Mikami, T. Y. 846

Tanaka, X. Wang, and S. Yoon (2011). Dust cycle: An emerging core theme in Earth system 847

science. Aeol. Res., 2, 181-204. 848

849

Schulz, M., Y. J. Balkanski, W. Guelle, and F. Dulac (1998). Role of aerosol size distribution and 850

source location in a three-dimensional simulation of a Saharan dust episode tested against satellite-851

derived optical thickness. J. Geophys. Res., 103, D9, 10,579-10,592, 1998, doi:10.1029/97JD02779. 852

853

Sinyuk, A., O. Torres, and O. Dubovik (2003). Combined use of satellite and surface observations 854

to infer the imaginary part of refractive index of Saharan dust. Geophys. Res. Lett., 30, 2, 1081, 855

doi:10.1029/2002GL016189. 856

857

Sokolik, I. N., and O. B. Toon (1996). Direct radiative forcing by anthropogenic airborne mineral 858

aerosols. Nature, 381, 681-683. 859

860

Steffensen, J. P. (1997). The size distribution of microparticles from selected segments of the 861

Greenland Ice Core Project ice core representing different climatic periods. J. Geophys. Res., 102, 862

C12, 26,755-26,763. 863

36

864

Stuut, J.B., M. Zabel, V. Ratmeier, P. Helmke, E. Schefuss, G. Lavik, and R. Schneide (2005). 865

Provenance of present day eolian dust collected off NW Africa. J. Geophys. Res., 110, D4, 866

10.1029/2004JD005161 867

868

Takemura, T., M. Egashira, K. Matsuzawa, H. Ichijo, R. O’ishi, and A. Abe-Ouchi (2009). A 869

simulation of the global distribution and radiative forcing of soil dust aerosols at the Last Glacial 870

Maximum. Atmos. Chem. Phys., 9, 3061-3073. 871

872

Taylor, K. E., R. J. Stouffer, and G. A. Meehl (2012). An overview of CMIP5 and the experiment 873

design. Bull. Am. Met. Soc., April 2012, 485-498, DOI:10.1175/BAMS-D-11-00094.1 874

875

Tegen I. (2003). Modeling the mineral dust aerosol cycle in the climate system. Quat. Sci. Rev., 22, 876

1821–1834. 877

878

Tegen, I., S.P. Harrison, K. Kohfeld, I.C. Prentice, M. Coe, and M. Heimann (2002). Impact of 879

vegetation and preferential source areas on global dust aerosol: Results from a model study. J. 880

Geophys. Res., 107, D21, 4576, doi:10.1029/2001JD000963. 881

882

Tegen I., and I. Fung (1994). Modeling of mineral dust in the atmosphere - sources, transport, and 883

optical thickness. J. Geophys. Res., 99, D11, 22,897-22,914. 884

885

Tegen, I., and A. A. Lacis (1996). Modeling of particle size distribution and its influence on the 886

radiative properties of mineral dust aerosol. J. Geophys. Res., 101, D14, 19,237-19,244. 887

888

37

Volz, F. E. (1973). Infrared optical constants of ammonium sulfate, Sahara dust, volcanic pumice, 889

and flyash. Appl. Opt., 12, 564-568. 890

891

Werner, M., I. Tegen, S. P. Harrison, K. E. Kohfeld, I. C. Prentice, Y. Balkanski, H. Rodhe, and C. 892

Roelandt (2002). Seasonal and interannual variability of the mineral dust cycle under present and 893

glacial climate conditions. J. Geophys. Res., 107, D24, 4744. 894

895

Wild, M. (2008). Short-wave and long-wave surface radiation budgets in GCMs: a review based on 896

the IPCC-AR4/CMIP3 models. Tellus A, 60, 932-945. 897

898

Wild, M., C. N. Long, and A. Ohmura (2006). Evaluation of clear-sky solar fluxes in GCMs 899

participating in AMIP and IPCC-AR4 from a surface perspective. J. Geophys. Res., 111, D01104. 900

901

Wu, G., T. Yao, B. Xu, L. Tian, C. Zhang, and X. Zhang (2009). Volume-size distribution of 902

microparticles in ice cores from the Tibetan Plateau. J. Glaciol., 55, 193, 859-868. 903

904

Yoshioka M., N. Mahowald, A. Conley, W. Collins, D. Fillmore, C. Zender, and D. Coleman 905

(2007). Impact of Desert Dust Radiative Forcing on Sahel Precipitation: Relative importance of dust 906

compared to sea surface temperature variations, vegetation changes and greenhouse gas warming. J. 907

Climate, 20, doi:10.1175/JCLI4056.1,1445-1467. 908

909

Zender C.S., H. Bian, and D. Newman (2003a). Mineral Dust Entrainment and Deposition (DEAD) 910

model: Description and 1990s dust climatology. J. Geophys. Res., 108, D14, 4416, 911

doi:10.1029/2002JD002775. 912

913

38

Zender, C., D. Newman, and O. Torres (2003b). Spatial heterogeneity in Aeolian erodibility: 914

Uniform, topographic, geomorphic, and hydrologic hypotheses. J. Geophys. Res., 108(D17), 4543, 915

doi:10.1029/2002JD003039. 916

917

Zhang, J., and S. A. Christopher (2003). Longwave radiative forcing of Saharan dust aerosols 918

estimated from MODIS, MISR, and CERES observations on Terra. Geophys. Res. Lett., 30, 23, 919

2188. 920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

39

Tables 940

941

Table 1. Schematic of default versus new set of parameterizations 942

Physical process Release parameterization New parameterization

Soil erodibility Mahowald et al. [2006] New tuning, this work

Emissions size

distribution

(0.038, 0.11, 0.17, 0.67) [Mahowald

et al., 2006]

(0.011, 0.087, 0.277, 0.625)

[Kok, 2011]

Wet deposition Dust solubility = 0.15 ; Scavenging

coeff = (0.1,0.1,0.1,0.1)

Dust solubility = 0.3 ;

Scavenging coeff = (0.1,0.1,0.3,0.3)

Dust optics CAM4/CAM5 default This work

943

944

Table 2. Observational datasets used in this work and related dust properties for comparison with the two models. 945

Property Feature Dataset Reference Metric Code CAM4-

BAM

CAM5-

MAM

AOD Magnitude AERONET Holben et al.,

1998

AOD, annual average O1 x x

AOD Seasonality AERONET Holben et al.,

1998

AOD, monthly

average

O2 x x

Column

load

Size

distribution

AERONET Dubovik and

King, 2000

Correlation for 1-10

μm range

O3 x

Surface

conc.

Magnitude U. Miami Prospero et al.,

1998

Concentration (μg/m3),

annual average

O4 x x

Surface

conc.

Seasonality U. Miami Prospero et al.,

1998

Concentration (μg/m3),

monthly average

O5 x x

Deposition Magnitude This work This work (Text

S1)

Modern flux

(mg/m2*y)

O6 x x

Deposition Magnitude This work This work (Text

S1)

LGM flux (mg/m2*y) O7 x

Deposition Magnitude This work This work (Text

S1)

Interglacial flux

(mg/m2*y)

O8 x

Deposition Size

distribution

This work This work (Text

S1)

Correlation for 1-10

μm range

O9 x

Deposition Provenance This work This work (Text

S1, Table S1)

Source apportionment O10 x x

946

947

948

949

950

951

40

Table 3. Description of the model simulations in this work 952

case Setup Group Run

length

(years)

Soil

erodibility

Emissions size

distribution

Wet

deposition

Dust

optics

C4sr CAM4 slab

ocean 2x2

current

prelimary

tests

3 Release Release Release Release

C4sr-k CAM4 slab

ocean 2x2

current

prelimary

tests

3 Release Kok, 2011 Release Release

C4sr-kw CAM4 slab

ocean 2x2

current

prelimary

tests

3 Release Kok, 2011 This work Release

C4fr CAM4 fully-

coupled 1x1

current

main cases

current

climate

3 Release Release Release Release

C4fn CAM4 fully-

coupled 1x1

current

main cases

current

climate

3 This work Kok, 2011 This work This

work

C4wr CAM4 2x2

MERRA

winds

main cases

current

climate

3 Release Release Release Release

C4wn CAM4 2x2

MERRA

winds

main cases

current

climate

3 This work Kok, 2011 This work This

work

C5wr CAM5.1 2x2

GEOS-05

winds

main cases

current

climate

3 Release Release Release Release

C5wn CAM5.1 2x2

GEOS-05

winds

main cases

current

climate

3 This work Kok, 2011 Release This

work

C4fn-rs CAM4 fully-

coupled 1x1

current

sensitivity

studies

3 This work Release This work This

work

C4fn-ro CAM4 fully-

coupled 1x1

current

sensitivity

studies

3 This work Kok, 2011 This work Release

C4fn-x2 CAM4 fully-

coupled 1x1

current

sensitivity

studies

3 This work Kok, 2011 This work This

work

C4fn-s2 CAM4 fully-

coupled 1x1

current

sensitivity

studies

3 This work [0.02,0.09,0.27,0.62] This work This

work

C4fn-lgm CCSM4 1x1

LGM

main case

LGM

3 This work Kok, 2011 This work This

work

C4fn-lgm-s2 CCSM4 1x1

LGM

sensitivity

study LGM

3 This work [0.02,0.09,0.27,0.62] This work This

work

953

954

955

956

41

Table 4. Model performance compared to observations (fraction <10 µm). 957

C4sr C4sr-k C4sr-kw C4fr C4fn C4wr C4wn C5wr C5wn

Deposition corr. (O6) 0.73 0.73 0.7 0.78 0.8 0.92 0.9 0.33 0.27

Dep. model/obs. ratio median

(O6)

1.77 1.72 1.42 1.33 1.49 0.98 0.92 0.5 0.55

Surf. conc. corr. (O4) 0.66 0.66 0.62 0.65 0.59 0.63 0.57 0.64 0.61

Surf. conc. model/obs. ratio

median (O4)

3.98 4.04 1.64 2.87 1.3 1.89 0.6 0.65 0.81

AOD corr. (O1) 0.78 0.77 0.76 0.7 0.67 0.72 0.7 0.53 0.55

AOD model/obs. ratio median

(O1)

0.9 0.65 0.58 1.07 0.65 0.51 0.35 0.17 0.68

AOD seas. corr. (O2) 0.46 0.46 0.41 0.6 0.41 0.79 0.69 0.57 0.67

Surf. conc. seas. corr. (O5) 0.24 0.25 0.29 0.3 0.31 0.5 0.45 0.31 0.49

Column size ditsr. corr. (O3) -0.23 0.86 0.82 -0.36 0.77 0.26 0.81

Dep. size distr. corr. (O9) 0.1 0.18 0.41 0.11 0.15 0.5 0.87

958

959

Table 5. Diagnostics of the global dust cycle for different model cases. 960

case Emissisons (Tg/y) load (Tg) Lifetime (d) AOD

C4fr 2395 29.4 4.5 0.032

C4fn 2783 24.6 3.2 0.023

C4wr 2018 34.8 6.3 0.028

C4wn 1901 24.9 4.8 0.018

C5wr 3834 25.4 2.4 0.008

C5wn 4966 31.2 2.3 0.028

C4fn-rs 2855 26.8 3.4 0.036

C4fn-ro 2563 23.2 3.3 0.018

C4fn-x2 5812 52.2 3.3 0.050

C4fn-s2 2743 26.3 3.5 0.026

C4fn-lgm 6523 42.0 2.4 0.040

C4fn-lgm-s2 6705 42.4 2.3 0.045

961

962

963

964

965

966

42

Table 6. Comparison of observed and modeled net TOA clear-sky RF (W/m2). 967

Parameter SW TOA

JJA

LW TOA

Sept

SW TOA JJA

Domain N. Atlantic

15-25 N,

45-15 W

Sahara

15-35 N,

18W-40E

Sahara 15-

35 N, 18W-

40E

Reference Li et al.,

2004

Zhang and

Christopher,

2003

Patadia et. al.,

2009

Obs. -35 ± 3 15 ~0 for Albedo

of 0.40

C4fr -31.52 - 13.73

C4fn -34.25 10.79 3.63

C4wr -25.14 - 16.28

C4wn -28.91 9.57 3.86

C5wr -19.01 - 17.41

C5wn -28.48 8.85 -0.18

C4fn-ro -24.36 - 30.05

C4fn-rs -32.82 7.52 -0.56

C4fn-s2 -33.63 9.39 2.33

968

969

Table 7. Globally averaged dust loading (Tg), Radiative Forcing at TOA, atmosphere and surface (W/m2), and RF 970

efficiency (W/m2*Tg). 971

case load

(Tg)

net RF

TOA

net RF

atm.

net RF

surf.

SW RF

TOA

SW RF

atm.

SW RF

surf.

net RF TOA

/ load

SW RF TOA

/ load

C4fr 29.4 -0.27 1.82 -2.09 -0.27 1.82 -2.09 -0.009 -0.009

C4fn 24.6 -0.22 0.32 -0.54 -0.36 0.79 -1.15 -0.009 -0.015

C4wr 34.8 0.02 1.69 -1.67 0.02 1.69 -1.67 0.001 0.001

C4wn 24.9 -0.08 0.26 -0.34 -0.19 0.63 -0.82 -0.003 -0.008

C5wr 25.4 0.12 1.13 -1.01 0.01 1.54 -1.52 0.005 0.001

C5wn 31.2 -0.13 0.25 -0.38 -0.32 0.80 -1.52 -0.004 -0.010

C4fn-rs 26.8 -0.45 0.34 -0.79 -0.61 0.88 -1.49 -0.017 -0.023

C4fn-ro 23.2 0.02 1.38 -1.36 0.02 1.38 -1.36 0.001 0.001

C4fn-x2 52.2 -0.42 0.71 -1.13 -0.72 1.61 -2.33 -0.008 -0.014

C4fn-s2 26.3 -0.28 0.34 -0.62 -0.44 0.85 -1.28 -0.011 -0.017

C4fn-

lgm

42.0 -0.33 0.53 -0.86 -0.57 1.29 -1.86 -0.008 -0.014

C4fn-

lgm-s2

42.4 -0.41 0.54 -0.95 -0.65 1.32 -1.97 -0.010 -0.015

972

973

974

975

976

43

Table 8. Comparison of dust TOA RF estimates from the literature (* are reported from IPCC AR4). 977

Reference Climate SW TOA LW TOA net TOA

Liao et al., 2004* Current -0.21 0.31 0.1

Reddy et al., 2005* Current -0.28 0.14 -0.14

Jacobson, 2001* Current -0.2 0.07 -0.13

Myhre and Stordal, 2001* Current -0.53 0.13 -0.4

GISS (Aerocom)* Current -0.75 0.19 -0.56

UIO-CTM (Aerocom)* Current -0.56 0.19 -0.37

LSCE (Aerocom)* Current -0.6 0.3 -0.3

UMI (Aerocom)* Current -0.54 0.19 -0.35

Miller et al., 2006 Current -0.62 0.23 -0.39

Balkanski et al., 2007 Current -0.68 0.29 -0.39

Takemura et al., 2009 Pre-industrial -0.1 0.09 -0.01

Yoshioka et al., 2007 Current -0.92 0.31 -0.61

Claquin et al., 2003 Current -1.2

C4fn, this work Current -0.36 0.14 -0.22

C4wn, this work Current -0.19 0.11 -0.08

C5wn, this work Current -0.32 0.19 -0.13

Takemura et al., 2009 LGM -0.24 0.22 -0.02

Chylek and Lohman, 2008 LGM -3.3

Claquin et al., 2003 LGM -3.2

Kohler et al., 2008 LGM -1.88

C4fn-lgm, this work LGM -0.57 0.24 -0.33

978

979

Table 9. Relative proportions (%) of dust emissions, load, and AOD in the 4 model bins. 980

C4fr C4fn C4wr C4wn C4fn-rs C4fn-ro C4fn-x2 C4fn-s2 C4fn-lgm

emission b1 3.8 1.1 3.8 1.1 3.8 1.1 1.1 2.0 1.1

emission b2 11 9 11 9 11 9 9 9 9

emission b3 17 28 17 28 17 28 28 27 28

emission b4 67 63 67 63 67 63 63 62 63

load b1 12 3 10 3 12 3 3 7 4

load b2 32 25 28 23 32 25 26 28 25

load b3 26 40 28 43 24 40 41 44 37

load b4 31 31 34 30 32 31 30 35 34

AOD b1 41 18 39 17 43 16 18 28 18

AOD b2 36 41 34 39 36 40 41 36 41

AOD b3 14 30 17 33 13 31 30 26 28

AOD b4 9 11 10 11 8 12 11 10 12

981

982

983

984

44

Table 10. Comparison of observed and modeled relative (%) amount of dust in model bin1 (0.1-1 µm). 985

Site Reference Observations C4fn C4fn-s2 C4wn

EDC Delmonte et al., 2004 7.4 7.1 2.9 9

JRI McConnell et al., 2007 21.9 6.7 4.2 5.9

GRIP Steffensen, 1997 15.3 8.3 4.3 5.7

Dunde Wu et al., 2009 0.2 2.3 1.3 1.9

Everest Wu et al., 2009 1 5.9 3.2 3

Dasuopu Wu et al., 2009 0.2 5.1 3.9 2.2

Muztagata(MA7010) Wu et al., 2009 0.6 3.4 2 2

986

987

988

Table 11. Comparison of simulations where with soil erodibility constrained by observations of bulk deposition (full 989

size range) versus the fraction <10 µm. 990

Emission (Tg/y) Burden

(Tg)

AOD

Full size range 2751 22.1 0.020

< 10 m 2321 18.2 0.017

991

992

993

Figure captions 994

995

Figure 1. Wavelength dependence of dust optical properties in the visible part of the spectrum. Left 996

panel: Imaginary refractive index of bulk dust and individual mineral components in our new 997

model, compared with remote sensing observations. Right panel: Single Scattering Albedo for bulk 998

dust and individual size bins in our model (based on annual average simulation at Solar Village), 999

compared to AERONET observations. 1000

1001

Figure 2. Comparison of simulated dust deposition (mg/m2*y) for the C4fn case, compared to 1002

observations of modern dust fluxes (Table 2). Upper row: observations. Middle row: model. Bottom 1003

45

row: model vs observations scatterplot. Locations of observational sites are clustered in the 1004

scatterplots based on their geographical location.mfk 1005

1006

Figure 3. Same as figure 2, for dust surface concentrations (μg/m3). 1007

1008

Figure 4. Same as figure 2, for dust AOD. 1009

1010

Figure 5. Taylor diagram showing the synthetic comparison of different model cases (see Table 3) 1011

to observations of dust deposition, surface concentration and AOD for current climate, shown in 1012

detail for the C4fn case in Figures 2-4 (see Table 2). 1013

1014

Figure 6. Comparison of model and observation for the annual cycle for AOD. Observations are 1015

from AERONET stations [Holben et al., 1998]. 1016

1017

Figure 7. Comparison of model and observation for the annual cycle for dust surface concentration 1018

(μg/m3). Observations are from Prospero et al. [1989]. 1019

1020

Figure 8. Comparison of model and observation for dust size distributions in the 1-10 μm range. 1021

Obervations are from AERONET stations [Dubovik and King, 2000]. Correlation coefficients 1022

between model and observations across the three size bins are reposted for each case. 1023

1024

Figure 9. Same as figure 8 for dust deposition. Observations are from ice cores [Delmonte et al., 1025

2004; Steffensen, 1997; McConnell et al., 2007; Wu et al., 2009]. 1026

1027

Figure 10. Same as Figure 2, for the Last Glacial Maximum. 1028

1029

46

Figure 11. Dust radiative forcing (W/m2) for the C4fn case. Left panel: SW RF. Central panel: LW 1030

RF. Right panel: net RF. Top row: TOA. Middle row: atmospheric column. Bottom row: surface. 1031

Right panels associated to each map represent the zonal averages of dust net RF. 1032

1033

Figure 12. Zonal averages of dust radiative forcing (W/m2) for different model cases. Left panel 1034

(a,b,c): compares C4fn, C4wn and C5wn cases for net RF. Central panel: compares C4fn, C4fn-1035

lgm, C3 and C3-lgm for net RF. Right panel: compares C4fn, C4fr, C4fn-rs, C4fn-ro, C4fn-s2 cases 1036

for net RF and C4fn-rs (C4fn-rs-SW) and C4fn (C4fn-SW) cases for the SW component of RF. Top 1037

row: TOA. Middle panel: atmospheric column. Bottom panel: surface. 1038