Soil dust aerosols created by wind erosion are typically assigned globally uniform physical and chemical properties within Earth system models, despite known regional variations in the mineral content of the parent soil. Mineral composition of the aerosol particles is important to their interaction with climate, including shortwave absorption and radiative forcing, nucleation of cloud droplets and ice crystals, heterogeneous formation of sulfates and nitrates, and atmospheric processing of iron into bioavailable forms that increase the productivity of marine phytoplankton. Here, aerosol mineral composition is derived by extending a method that provides the composition of a wet-sieved soil. The extension accounts for measurements showing significant differences between the mineral fractions of the wet-sieved soil and the emitted aerosol concentration. For example, some phyllosilicate aerosols are more prevalent at silt sizes, even though they are nearly absent at these diameters in a soil whose aggregates are dispersed by wet sieving. We calculate the emitted mass of each mineral with respect to size by accounting for the disintegration of soil aggregates during wet sieving. These aggregates are emitted during mobilization and fragmentation of the original undispersed soil that is subject to wind erosion. The emitted aggregates are carried far downwind from their parent soil. The soil mineral fractions used to calculate the aggregates also include larger particles that are suspended only in the vicinity of the source. We calculate the emitted size distribution of these particles using a normalized distribution derived from aerosol measurements. In addition, a method is proposed for mixing minerals with small impurities composed of iron oxides. These mixtures are important for transporting iron far from the dust source, because pure iron oxides are more dense and vulnerable to gravitational removal than most minerals comprising dust aerosols. A limited comparison to measurements from North Africa shows that the model extensions result in better agreement, consistent with a more extensive comparison to global observations as well as measurements of elemental composition downwind of the Sahara, as described in companion articles.
Climate perturbations by soil dust aerosols created by wind erosion depend
fundamentally upon the physical and chemical properties of the aerosol
particles. However, Earth system models typically assume that soil dust
aerosols have globally uniform composition, despite known regional variations
in the mineral composition of the parent soil. Perturbations by dust to the
energy and water cycles depend upon aerosol radiative forcing
Deriving aerosol mineral composition requires global knowledge of soil
mineral content.
The challenge remains to derive mineral fractions of the emitted dust based
upon their fractions measured in wet-sieved soils. Previous attempts to
predict the aerosol mineral composition have generally neglected the effects
of wet sieving
In this article, we propose a model of dust mineral composition to
address these challenges. Some of the extensions of our model have
been introduced previously
Our model extensions are motivated by observations. In Sect. 4, we
show that our new model is in better agreement with aerosol
measurements at a site in North Africa after accounting for the
effects of wet sieving. Detailed comparison of the model to a broader
array of observations is deferred to companion articles. In
Our aim is to predict regional variations of aerosol mineral
composition as a function of particle size. For comparison, ModelE2
currently predicts the size distribution of dust aerosols, but assumes
a globally uniform mineral content
Volume distribution of minerals with respect to particle
diameter, calculated as described in the Supplement, using
size-resolved dust number and volume fraction measured by
Differences between the size distribution of the soil after wet sieving and
during emission are potentially large. Figure
Figure
The presence of significant clay mass at silt diameters argues that
aggregates in the original soil subject to wind erosion are
significantly dispersed by wet sieving. The alteration of the
carbonate size distribution during emission
Direct entrainment by the wind of the smaller soil particles that
travel thousands of kilometers downwind from their source (whose
diameters are generally below 20
Most of the smaller particles that are transported globally are
entrained into the atmosphere during the fragmentation of aggregates
that are bombarded by larger particles, or else are large enough to be
lifted directly by the wind and disintegrated through repeated
collisions
An additional modeling challenge is that different minerals may have
different size distributions in the soil and may not be equally
susceptible to disaggregation and fragmentation during wet sieving and
emission, respectively. The size distribution of each mineral in
Fig.
In Sect.
Our second extension is to account for mixtures of minerals that are
often observed within a single aerosol particle
For minerals removed from the atmosphere at the same rate, their
combination can be represented as an external mixture, requiring no
additional prognostic variables. In this case, the mineral fraction
at a particular size is interpreted as the fractional mass of that
mineral that is present either in pure form or as an aggregate. In
the latter case, it is the diameter of the aggregate that is used to
assign a size category. As an example, a
The rate of aerosol removal is distinguished in part by particle
density that controls the speed of gravitational settling. Many
minerals commonly observed to comprise dust particles have similar
densities, suggesting that external mixing is a reasonable
idealization that is attractive for its computational
simplicity. Mineral fractions also evolve during wet scavenging as
a result of contrasts in mineral solubility and during gravitational
settling through contrasts in particle shape. We neglect both sources
of complexity in the present study, although
In contrast, iron oxides like hematite and goethite have densities
that are twice that of the other minerals, and would thus be removed
by gravitational settling within roughly half the distance from their
source. This density contrast means that mineral combinations
containing iron oxides cannot be represented implicitly as external
mixtures like combinations of other minerals. Iron oxide mixtures
must be treated explicitly as separate prognostic variables that are
distinct from pure crystalline forms of this mineral. In general,
impurities of iron oxides are only a small fraction of the total
particle mass, and only slightly perturb the particle density that is
determined primarily by the host mineral
To our knowledge, measurements of mineral mixtures within individual aerosol
particles are mostly anecdotal and provide only limited guidance about the
combinations that need to be represented by a global model. In this study,
the only combinations we represent explicitly are internal mixtures of iron
oxides with another mineral, following
Our extensions to
The transformation of the particle size distribution of the
(undispersed) parent soil into the emitted size distribution is
a complicated process that depends upon wind speed and the physical
properties of the soil and land surface
The theory of brittle fragmentation has been invoked to suggest that
this invariance is robust, despite limited measurements of
size-resolved emission
For soil aggregates,
The emitted number concentration in Eq. (
Size distribution of emitted dust (black line) derived from
Eq. (
Equation (
The dotted curve in Fig.
From left to right: Distribution of dust volume calculated
by summing minerals in Fig.
The process of brittle fragmentation that leads to the emitted size
distribution in Fig.
By apportioning silt emission with measurements of the volume fraction after
transport to a single location, we are making at least two approximations.
First, we are assuming that the distribution at Tinfou is representative of
other sources. The increase of the emitted silt fraction with increasing
particle size (Fig.
Data sets used in this study.
Minerals represented in ModelE2. Closed circles (
Here, we describe our calculation of the emitted fraction of each
mineral and its particle size distribution. We treat the dust
particles as an external mixture of minerals, each corresponding to
a separate prognostic variable. We create additional prognostic
variables for mixtures of each mineral with iron oxides, where the
latter is assumed to be a small fraction of the total particle
mass. Calculation of iron oxide mixtures is described separately in
Sect.
We first derive the mineral composition of the fully dispersed soil following
Soil texture classes with sand, silt, and clay percentages,
and clay (
To calculate the mineral fractions of the dispersed soil at each
location, we specify the fraction of each size category present,
provided by the soil texture class
We have derived Eq. (
Let
We prescribe the mass fraction of the emitted clay-sized particles
using brittle fragmentation theory, as described in
Sect.
Because of Eq. (
We also assume that the emitted mass fraction of each mineral
As a consequence of Eqs. (
Fractional distribution of volume within the ModelE2 size
bins for the minerals in Table
Equation (
Because the total fractional silt emission is assumed to be fixed
according to Eq. (
Feldspar and gypsum are observed as aerosols at both clay and silt
sizes (Fig.
Size categories for dust transported in ModelE2.
List of symbols used to represent mass fractions of soil
and emitted minerals (Sect.
List of symbols used to represent mixtures of iron oxide
and other minerals (Sect.
The emitted silt particles have diameters ranging between 2 and 50
Finally, we renormalize the mass fractions
Our model generally resembles that of
In our model, iron oxides can travel either in pure crystalline form
or as an internal mixture with other minerals. (Combinations of the
other minerals excluding iron oxides are treated as external
mixtures.) Our apportionment of iron oxides combines the two limiting
cases considered by
At each location, Eqs. (
To simplify notation, we drop the superscripts in
Eqs. (
We first distinguish between each mineral in its pure and mixed state:
We next assume that iron oxide is mixed with the other minerals in
proportion
Finally, we assume that the iron oxide available for mixing is
distributed among the other minerals in proportion to their mass
fraction:
As a result of Eqs. (
The mass fraction of pure, crystalline iron oxide is given by
According to Eq. (
Our modeling assumptions leading to Eq. (
Our model of the emitted mineral fractions has been incorporated into
the CMIP5 version of the NASA GISS Earth System ModelE2
Emission occurs when the surface wind speed
Globally averaged dust emission, load and lifetime. The number in parentheses is 1 standard deviation (SD) of interannual variability.
Emission of each mineral is the total emission
Each mineral is advected using the Quadratic Upstream Scheme
Mineral densities in
Removal of the mineral tracers from the atmosphere takes place by wet
and dry deposition. Dry deposition includes gravitational settling and
turbulent deposition in the surface layer. Settling speeds are
proportional to mineral density
Wet deposition occurs through scavenging both within clouds and below
where there is precipitating condensate, while aerosols are restored
by re-evaporation of cloud droplets and precipitate
We also defer calculation of dust radiative forcing and its dependence
upon spatial variations in the aerosol mineral composition. This
eliminates feedbacks between dust radiative forcing and emission
resulting, for example, from the perturbed surface wind speed or
precipitation
We evaluate our new approach in comparison to a control or “baseline”
simulation that assumes the emitted mineral fractions are identical to those
of the wet-sieved soil. We conduct a set of simulations with the NASA GISS
ModelE2 covering the years 2002 to 2010. This period coincides with a period
of detailed measurements at Izaña
Our baseline simulation is referred to as the “soil mineral
fraction” (SMF) method, and assumes that the emitted mineral
fractions are identical to those of the wet-sieved soil, given by
Eq. (
Our new approach, hereafter described as the “aerosol mineral
fraction” (AMF) method, is described in Sect.
For the AMF simulation, our reaggregation parameter
We carried out two more experiments to show the effect of different
treatments of iron oxides. The first is the AMF experiment but without
accretions of iron oxides with other minerals (denoted by
“AMF-NoFeAcc”). The second (denoted as “SMF-NoClayFe”) corresponds
to the SMF experiment, but without the extension of iron oxides into
clay sizes proposed by
The distribution of the emitted dust mass fraction at each
ModelE2 size bin for the soil mineral fraction (SMF) method (left)
and the aerosol mineral fraction (AMF) method (right). Within each
size bin, the box plots depicts the distribution with respect to
the combinations of the 12 soil textures and the 28 arid soil
types included in the MMT. For each combination, the sum over all
sizes is one. At each bin, each combination within the
distribution is weighted by the total emission (summed over all
sizes) to emphasize prolific sources. Each box shows the range in
which the central 50 % of the data fall. The box borders show
the first and third quartiles and the crossbar shows the
median. Outliers exceeding the quartile values by more than
a factor of 1.5, the interquartile distance, are marked as points.
Note that only diameters below 32
List of experiments (SD
We first compare the emitted distributions of the AMF and SMF experiments
summed over all minerals. This comparison and its implications for long-range
transport help to understand regional variations of surface concentration for
the individual minerals, presented in Sect.
Figure
For each SMF size bin, emission varies according to the local soil
texture. For the AMF method, the size distribution varies additionally
due to reaggregation of certain clay-sized minerals dispersed in the
wet-sieved soil. For clay-sized particles
(Fig.
(Left panels) Annual-average surface concentration (summed
over all minerals) for the AMF method and (right panels) the ratio
of the AMF and SMF concentrations for
Global annual load (Tg) for the AMF and SMF experiments.
Global annual emission (Tg) for the AMF and SMF experiments.
The annual-mean AMF surface concentration for different particle sizes
is shown in Fig.
The distribution of the emitted mass fraction of each
mineral at each ModelE2 size bin for the soil mineral fraction
(SMF) method (green), the aerosol mineral fraction (AMF) method
(orange), and the AMF method with
Contrasts between the experiments are apparent in the emitted
fractions of the individual minerals, shown in
Fig.
The SMF method emits clay-sized dust aerosols that are comprised
mostly of phyllosilicates with median values of 0.40 for illite, 0.22
for smectite and 0.19 for kaolinite
(Fig.
Annual-average fraction of emission of
Fractional emission as in Fig.
In the AMF method, the emitted silt mass of each mineral is
distributed empirically across the model size categories using the
normalized volume fraction derived for each individual mineral based
upon measurements at Tinfou (cf. Eq.
Figure
The regional variations of emitted mineral fractions are displayed for
illite and kaolinite (Fig.
Fractional emission as in Fig.
The AMF global fraction of emitted illite is 33 % at clay sizes
(Fig.
Illite, kaolinite, and smectite (the latter not shown) are absent at
silt sizes in the SMF. The AMF experiment extends these
phyllosilicates into the silt size range (the size at which the
prescribed fraction of emission is largest according to
Fig.
The AMF global fraction of emitted quartz is roughly 7 % in the
clay size range (Fig.
The AMF global fraction of emitted carbonate is 5 and 6 % at clay
and silt sizes, respectively (Fig.
Annual-average fraction of surface concentration for
Fractional surface concentration as in
Fig.
Annual-average column mass fraction of
The AMF global fraction of emitted feldspar is roughly 13 % at
both silt and all sizes (Fig.
Emission of gypsum and iron oxides is comparatively small, with local
fractions never exceeding a few percent. The global emitted fraction
of iron oxides is nearly identical in the AMF and SMF experiments
(Fig.
Figures
Attribution of the mineral fractions to contrasts between the AMF and
SMF methods is challenging because the fractional surface
concentration depends upon the interaction of numerous processes
including the size dependence of emission and removal, along with the
proximity of sources enriched or depleted in different minerals.
Nonetheless, the figures illustrate the effect of some physical
assumptions underlying the methods. For example, the AMF kaolinite and
iron oxide fractions are large downwind of the Sahel and southern
Africa (Figs.
Differences between the AMF and SMF methods are also illustrated. Far
from source regions, clay-sized particles dominate the concentration.
Thus, the difference of the clay-sized fractions of emission
determines approximately whether the concentration of a particular
mineral in remote regions is larger according to one method. (This
relation is approximate, because the smallest silt particles have
lifetimes that are only slightly shorter.) The clay-sized fraction of
phyllosilicates is smaller in the AMF experiment
(Figs.
Immediately downwind from a source region enriched in phyllosilicates
and iron oxides, the ratio of the AMF and SMF fractional concentration
decreases (Figs.
The fraction of the iron oxide in accreted form is shown in
Fig.
Regions where the soil is enriched in iron oxides correspond to
a maximum of accreted iron oxide mass relative to the total dust mass
(Fig.
The fraction of quartz and phyllosilicates containing iron oxide
accretions compared to the total dust mass is shown in
Fig.
We carried out additional experiments to illustrate the effect of our
model assumptions for iron oxides and its mixtures
(Table
The effect of accretions is shown by contrasting the AMF and AMF-NoFeAcc experiments. The iron oxide mass at clay sizes is nearly identical in the two experiments because removal of this particle size is dominated by wet deposition that is independent of particle density. However, at larger silt sizes, whose concentration is more vulnerable to removal by gravitational settling, the AMF experiment with accretions has a global iron oxide mass that is larger by 40 %.
Mineral fractions of surface concentration relative to
total dust concentration. Values at Tinfou, Morocco, are
calculated from volume fractions of minerals and number of total
dust particles provided by
Same as Fig.
An extensive comparison of the model to observations is deferred to
a companion article
Figure
Feldspar is the exceptional mineral where the SMF fraction is more realistic at all silt sizes. The AMF experiment underestimates the measured feldspar fraction, although it predicts a non-zero fraction at clay sizes in contrast to the SMF experiment. In the companion article, we show that the AMF feldspar fraction is generally in better agreement at other locations. Both methods underestimate the iron oxide fraction, and the discrepancy of the SMF value increases with particle diameter. The relatively large density of the pure crystalline form enhances gravitational removal, reducing the particle lifetime as diameter increases. In contrast, the internal mixtures present only in the AMF experiment are removed more slowly, reproducing the measured weak dependence of the iron oxide fraction upon particle size.
Figure
The SMF experiment strongly overestimates the mineral fractions at
clay sizes due to its relatively large emission at this size
(Fig.
Feldspar and the phyllosilicates are difficult to distinguish within
single particles using X-ray diffraction (XRD) due to the small sample
mass, so
Aerosol mineral composition depends upon the composition of the parent
soil and its size fractionation during mobilization. Soil mineral
fractions have been estimated by
We have proposed a heuristic method to reconstruct the size
distribution of mineral aggregates that are emitted from the
undisturbed soil that is subject to wind erosion. We assume that some
of the emitted silt-size particles correspond to the clay-sized
fraction of the wet-sieved soil
(Eq.
In addition, we create an additional class of iron oxide aerosol that is a small impurity embedded within other minerals, allowing it to travel farther than in its pure crystalline state. We assume that these impurities are least frequent in soils rich in iron oxides (as a result of the assumed effect of weathering that creates pure iron oxide crystals). Nonetheless, the abundance of iron oxide in these soils means that the absolute value of iron oxides in accretions is large. We assume that iron oxides contribute 5 % to the combined particle mass, with the remainder transported as pure crystals that fall out quickly due to their higher density. The resulting global mass fraction of iron oxides that combines accreted and pure forms is just under 2 %, a value that we suggest is insensitive to the assumed mass fraction of this mineral in accreted form. Our treatment of iron oxides is constrained by relatively few observations and worthy of more precise future treatments. Modeling of aerosol iron is important for its influence upon several climate processes, including aerosol radiative forcing and marine photosynthesis that modulates atmospheric carbon dioxide.
These extensions define our AMF experiment. In contrast, the SMF
experiment serves as a control whose size distribution of the emitted
mineral fractions is taken directly from that of the wet-sieved soil
and excludes iron accretions. For both experiments, we calculate the
regional distribution of minerals using the NASA GISS ModelE2, whose
dust size categories range in diameter from 0.1 to 32
Emission of clay-sized particles is much smaller in the AMF experiment, due to an observational constraint upon the emitted size distribution. This has implications for long-range transport. Both the SMF and AMF have identical global emission (by construction), but the column load and surface concentration are much lower in the latter experiment, because the particles are larger. Nonetheless, the emission of clay minerals (i.e., phyllosilicates) is only slightly smaller in the AMF experiment. This is a consequence of reaggregation of the wet-sieved soil that results in a substantial fraction of phyllosilicate particles at silt sizes.
In companion articles
Both experiments predict comparable fractions of quartz at Tinfou, in
spite of the substantially greater silt-sized emission of the AMF
method compared to the SMF. This agreement is the result of the
reaggregation of clay minerals that reduces the quartz fraction at
silt sizes in the undispersed soil prior to emission. This reduction
occurs because the total silt emission summed over all minerals is
fixed by our empirical constraint
Eq. (
In general, our reconstruction of emitted aggregates from the
wet-sieved soil allows us to shift clay-sized phyllosilicates toward
silt sizes where they are observed and maintain realistic fractions of
quartz, despite the observed size distribution of emission that is
heavily biased toward silt sizes. We have made little effort to find
the optimal amount of reaggregation. Instead, we are developing a more
physically based model of reaggregation and brittle fragmentation that
extends studies by
Here we use the assumptions in Sect.
We have assumed that accreted iron oxides contribute fraction
The pure or unmixed mass fraction of each mineral can be derived from
Eq. (
We thank Paul Ginoux, Konrad Kandler, Natalie Mahowald, Sergio Rodríguez and Rachel Scanza for helpful conversations. This
article was improved by the thoughtful comments of Yves Balkanski,
Jasper Kok and two additional reviewers. This research was
supported by the National Science Foundation (ATM-01-24258), the
Department of Energy (DE-SC0006713), the NASA Modeling, Analysis and
Prediction Program and the Ministry of Economy and Competitiveness
of Spain through the POLLINDUST project (CGL2011-26259). NCEP
Reanalysis winds were provided by the Physical Sciences Division at
the National Oceanic and Atmospheric Administration Earth System
Research Laboratory via