This document discusses modeling the mineral composition of dust aerosols and its implications. It begins by explaining why mineral composition is important to model, as properties like single scattering albedo and ice nucleation depend on composition. It then describes how previous models assumed uniform composition but that composition actually varies spatially. It outlines the key data needed - maps of soil types and their mineral contents. It also explains that soil analyses destroy aggregates, while emitted dust is aggregated; a new approach is needed. This new approach models dust as brittle fragments, combining fragmentation theory with size-dependent mineral distributions from measurements. Preliminary results show the modeled size-dependent mineral fractions of emitted dust.

1. Predicting the Mineral Composition of Dust
Aerosols and Implications for Ice Forming Nuclei
Jan P. Perlwitz1,2, Carlos Pérez García-Pando1,2, and Ron L. Miller2,1
IFN: Ann M. Fridlind2 and Daniel A. Knopf3
1
Department of Applied Physics and Applied Mathematics, Columbia University,
New York, NY, USA, Contact: jan.p.perlwitz@columbia.edu
2
NASA Goddard Institute for Space Studies, New York, NY, USA
3
Institute for Terrestrial and Planetary Atmospheres / School of Marine and
Atmospheric Sciences, Stony Brook University, NY, USA
Image: Terra MODIS, September 28, 2010. NASA Image by Jeff Schmaltz, MODIS Rapid Response Team at NASA GSFC
Acknowledgements: The work has been supported by the Department of Energy,
NASA, NSF, and Ministry Economy and Competitiveness of Spain

2. Outline
1. Soil dust in the Earth system
2. Why is the mineral composition of dust important?
3. How to model the mineral composition of dust aerosols?
4. Evaluation of simulated mineral fractions
5. Preliminary results on my first attempt on ice nucleation
6. The Grand Plan

3. Soil Dust Aerosols in the Climate System
●
Major aerosol in the atmosphere (global emission 1000 to >6500
Tg/a; total mass load in atmosphere approximately 15-35 Tg)
●
High spatial and temporal variability; dust particle size range 0.1 to
500 µm particle diameter
●
Absorbs and reflects radiation => impact on radiation balance,
warming or cooling effect depending on the single scattering albedo
of the dust particles
●
Cloud condensation nuclei (CCN), ice forming nuclei (IFN)
● Atmospheric chemistry (e.g., uptake of SO2, H2SO4, HNO3, N2O5,
NO3, NO2, O3, H2O2, OH, HO2) => formation of coatings, like sulfates,
nitrates, and/or effect on trace gas budgets, and indirectly on the load
and fluxes of other aerosols
●
Carrier of nutrients like iron => Fertilization of phytoplankton =>
Carbon cycle; or phosphorus for terrestrial biosphere (e.g., Amazon)

4. Dust Particles
(Alastuey et al., 2005)

5. Global Dust Sources
(Muhs et al. 2014)
Global Dust Emission:
1000 - >6500 Tg/a
(Cakmur et al. 2006,
Evan et al. 2014)
Northern Africa:
515 Tg/a
(Miller et al. 2004)
to 6500 Tg/a
(Evan et al., 2014)
East Asia:
54 Tg/a (Luo et al., 2003)
to 460 Tg/year
(Laurent et al. 2006)
Arabian Peninsula:
43 Tg/a
(Miller et al., 2004)
to 496 Tg/a
(Ginoux et al., 2004)
Australia:
37 Tg/a
(Zender et al., 2003)
to 148 Tg/a
(Miller et al., 2004)

6. Dust source Example: Nile River Basin (15 Feb 2016)

7. Desert with Dry Stream Valleys Next to Nile River (15 Feb 2016)

8. Why is the mineral composition of dust important?

9. 
Traditionally, global dust models have
used globally uniform dust properties
Limitations for following reasons:

Single scattering albedo (SSA) of dust
particles depends on mineral
composition, particularly on the mass
fraction of hematite or goethite

In turn, aerosol forcing and the response
of clouds and atmospheric circulation to
the forcing depend on the SSA

Heterogenous chemistry (e.g., uptake
rates) of dust particles depends on
mineralogical and chemical composition

Hygroscopicity of dust particles, the
ability to act as cloud condensation
nuclei, depends on the mineralogical
composition

Ice nucleating properties are mineral
dependent (K-Feldspar)

Fertilization of phytoplankton in oceans is
linked to availability of soluble iron, i.e., to
the mineral types of dust
Why is the Mineral Composition of Soil Dust Important?
Perlwitz and Miller, JGR (2010),
Zonal Average 20º – 85º E in JJA:
More reflecting dust More absorbing dust
Moosmüller et al., JGR (2012)
Hematite Mass Fraction → Single
Scattering Albedo
Shaded: Blue: Less Upward Red: More Upward

10. 
Traditionally, global dust models have
used globally uniform dust properties
Limitations for following reasons:

Single scattering albedo (SSA) of dust
particles depends on mineral
composition, particularly on the mass
fraction of hematite or goethite

In turn, aerosol forcing and the response
of clouds and atmospheric circulation to
the forcing depend on the SSA

Heterogenous chemistry (e.g., uptake
rates) of dust particles depends on
mineralogical and chemical composition

Hygroscopicity of dust particles, the
ability to act as cloud condensation
nuclei, depends on the mineralogical
composition

Ice nucleating properties are mineral
dependent (K-Feldspar)

Fertilization of phytoplankton in oceans is
linked to availability of soluble iron, i.e., to
the mineral types of dust
Why is the Mineral Composition of Soil Dust Important?
Perlwitz and Miller, JGR (2010),
Zonal Average 20º – 85º E in JJA:
More reflecting dust More absorbing dust
Moosmüller et al., JGR (2012)
Hematite Mass Fraction → Single
Scattering Albedo
Shaded: Blue: Less Upward Red: More Upward

11. 
Traditionally, global dust models have
used globally uniform dust properties
Limitations for following reasons:

Single scattering albedo (SSA) of dust
particles depends on mineral
composition, particularly on the mass
fraction of hematite or goethite

In turn, aerosol forcing and the response
of clouds and atmospheric circulation to
the forcing depend on the SSA

Heterogenous chemistry (e.g., uptake
rates) of dust particles depends on
mineralogical and chemical composition

Hygroscopicity of dust particles, the
ability to act as cloud condensation
nuclei, depends on the mineralogical
composition

Ice nucleating properties are mineral
dependent (K-Feldspar)

Fertilization of phytoplankton in oceans is
linked to availability of soluble iron, i.e., to
the mineral types of dust
Why is the Mineral Composition of Soil Dust Important?
Perlwitz and Miller, JGR (2010),
Zonal Average 20º – 85º E in JJA:
More reflecting dust More absorbing dust
Moosmüller et al., JGR (2012)
Hematite Mass Fraction → Single
Scattering Albedo
Shaded: Blue: Less Upward Red: More Upward

12. 
Traditionally, global dust models have
used globally uniform dust properties
Limitations for following reasons:

Single scattering albedo (SSA) of dust
particles depends on mineral
composition, particularly on the mass
fraction of hematite or goethite

In turn, aerosol forcing and the response
of clouds and atmospheric circulation to
the forcing depend on the SSA

Heterogenous chemistry (e.g., uptake
rates) of dust particles depends on
mineralogical and chemical composition

Hygroscopicity of dust particles, the
ability to act as cloud condensation
nuclei, depends on the mineralogical
composition

Ice nucleating properties are mineral
dependent (K-Feldspar)

Fertilization of phytoplankton in oceans is
linked to availability of soluble iron, i.e., to
the mineral types of dust
Why is the Mineral Composition of Soil Dust Important?
Perlwitz and Miller, JGR (2010),
Zonal Average 20º – 85º E in JJA:
More reflecting dust More absorbing dust
Moosmüller et al., JGR (2012)
Hematite Mass Fraction → Single
Scattering Albedo
Shaded: Blue: Less Upward Red: More Upward

13. 
Traditionally, global dust models have
used globally uniform dust properties
Limitations for following reasons:

Single scattering albedo (SSA) of dust
particles depends on mineral
composition, particularly on the mass
fraction of hematite or goethite

In turn, aerosol forcing and the response
of clouds and atmospheric circulation to
the forcing depend on the SSA

Heterogenous chemistry (e.g., uptake
rates) of dust particles depends on
mineralogical and chemical composition

Hygroscopicity of dust particles, the
ability to act as cloud condensation
nuclei, depends on the mineralogical
composition

Ice nucleating properties are mineral
dependent (K-Feldspar)

Fertilization of phytoplankton in oceans is
linked to availability of soluble iron, i.e., to
the mineral types of dust
Why is the Mineral Composition of Soil Dust Important?
Perlwitz and Miller, JGR (2010),
Zonal Average 20º – 85º E in JJA:
More reflecting dust More absorbing dust
Moosmüller et al., JGR (2012)
Hematite Mass Fraction → Single
Scattering Albedo
Shaded: Blue: Less Upward Red: More Upward

14. 
Traditionally, global dust models have
used globally uniform dust properties
Limitations for following reasons:

Single scattering albedo (SSA) of dust
particles depends on mineral
composition, particularly on the mass
fraction of hematite or goethite

In turn, aerosol forcing and the response
of clouds and atmospheric circulation to
the forcing depend on the SSA

Heterogenous chemistry (e.g., uptake
rates) of dust particles depends on
mineralogical and chemical composition

Hygroscopicity of dust particles, the
ability to act as cloud condensation
nuclei, depends on the mineralogical
composition

Ice nucleating properties are mineral
dependent (K-Feldspar)

Fertilization of phytoplankton in oceans is
linked to availability of soluble iron, i.e., to
the mineral types of dust
Why is the Mineral Composition of Soil Dust Important?
Perlwitz and Miller, JGR (2010),
Zonal Average 20º – 85º E in JJA:
More reflecting dust More absorbing dust
Moosmüller et al., JGR (2012)
Hematite Mass Fraction → Single
Scattering Albedo
Shaded: Blue: Less Upward Red: More Upward

15. How to model the mineral composition of dust
aerosols?

16. The Main Data Sets Needed
1. Mean Mineralogical Table (MMT) by Claquin et al., JGR (1999) + Nickovic et al., ACP (2012)
The mineralogical composition of soils varies with the soil type. The MMT provides this information for
28 arid soil types
An updated table has just been published by Journet et al. ACP (2014), which is not used here.
Lithosols
Calcic
Fluvisols
Eutric
Fluvisols
Sand Dunes
Salt Flats

17. 2. Digital Soil Map of the World (DSMW) (FAO-UNESCO, 2007)
Geographical distribution of dominant top soil types
(5'x5' latitude by longitude)

18. 3. FAO/STATSGO Soil Texture Fractions
Geographical distribution of clay, silt, and sand fraction for soil texture types
(5'x5' latitde by longitude)
Figure source: http://ldas.gsfc.nasa.gov/gldas/GLDASsoils.php
Fully dispersed soils!

19. Mean Mineralogical Table
For Clay and Silt
Digital Soil Map of the World
Geographical Distribution of Mineral Fractions in Soil
for Clay and Silt
Dust Emission Flux at Location
for Clay and Silt
Mineral Emission Flux at Locations for Clay and Silt
x
Soil Texture Fractions of
Clay and Silt
How to Obtain the Emitted Mineral Fractions?
The Simple Approach. Case 1 - Soil Mineral Fraction (SMF)
Method

20. Challenge: Emission of Minerals from Soils
Previous dust models with mineralogy have assumed 1 to 1 translation of
mineral fractions in soil data sets to mineral fractions of dust aerosols
Emitted Dust: Aggregated and Fragmented Dust Particles
Shao et al. (2011)

21. Soil Texture and Mineral Fractions Determined Using Techniques
Leading to Nearly Full Destruction of Aggregates
Source: http://www.fhwa.dot.gov/engineering/geotech/pubs/05037/05a.cfm
Wet Sieved Soil Texture Fractions ≠ Size Distribution of Eroded Soils
Wet Sieved Soil Texture Fractions ≠ Suspended Dust Size Distribution
Wet Sieved Clay/Silt Mineral Fractions ≠ Mineral Fractions of Suspended Dust

22. Normalized Volume Size Distribution of Mineral Fractions in Dust
Derived From Data Provided by Kandler et al. Tellus B (2009)

Illite and kaolinite: Similar
volume size distribution;
most of the volume (mass)
is found in higher particle
size classes, even beyond
silt size range (probably
mostly due to aggregation)

The carbonates and
gypsum peak in the coarse
silt size class

Distinctive size
distribution of quartz with
steep increase in the
volume distribution for
largest particle sizes

23. Normalized Volume Size Distribution of Mineral Fractions in Dust
Derived From Data Provided by Kandler et al. Tellus B (2009)

Illite and kaolinite: Similar
volume size distribution;
most of the volume (mass)
is found in higher particle
size classes, even beyond
silt size range (probably
mostly due to aggregation)

The carbonates and
gypsum peak in the coarse
silt size class

Distinctive size
distribution of quartz with
steep increase in the
volume distribution for
largest particle sizes

24. Normalized Volume Size Distribution of Mineral Fractions in Dust
Derived From Data Provided by Kandler et al. Tellus B (2009)

Illite and kaolinite: Similar
volume size distribution;
most of the volume (mass)
is found in higher particle
size classes, even beyond
silt size range (probably
mostly due to aggregation)

The carbonates and
gypsum peak in the coarse
silt size class

Distinctive size
distribution of quartz with
steep increase in the
volume distribution for
largest particle sizes

25. The New Approach to Simulate Prognostic
Minerals in a Climate Model (AMF Method)
Approach: Combining brittle fragmentation theory (Kok,
PNAS 2011) with an empirical mineral size distribution
(Kandler et al., Tellus B 2009) to derive the size
dependent mineral fractions of the emitted dust aerosol
Brittle fragmentation theory (Kok, 2011)
1. Assumption: Dust aggregates behave in saltation
regime like brittle material. Size distribution due to
fragmentation:
dNf
dln Df
∝Df
−2
exp[−(
Df
λ
)
3
], (x0<Df )
x0 - indivisible length scale λ - crack propagation length

26. 2. Aggregate dust aerosol particles of size Df can only be
composed of soil particles Ds ≤ Df. Hence, following
proportionality can be written:
dNd
d ln Dd
∝∫
0
Dd
Ps(Ds)dDs
3. Assuming log-normal distribution for the fully dispersed
soil Ps(Ds), a size distribution for the emitted dust
aggregates can be derived by combining the two
relations above.

27. Fully dispersed mineral
mass in soil
Reconstructed size-dependent
mineral mass in emitted aerosols
Re-aggregation
Brittle fragmentation theory (Kok, PNAS 2011)
Perlwitz et al., ACP (2015a)
Only for emission by saltation!
BFT size distribution
+ Fragmentation

28. Volume Size Distribution of Emitted Minerals in ModelE2
Perlwitz et al., ACP (2015a)
Derived from measurements at Tinfou, Morocco, provided by Kandler et al. (2009)

29. Modeling Soil Dust in NASA GISS ModelE2
●
ModelE2 - Earth system model: to include all relevant processes, the energy and
mass fluxes within and between the components of the planet (atmosphere,
ocean, land, cryosphere, biosphere) to simulate climate variability in response to
changing boundary conditions by external drivers (e.g., forcing by solar
variability, anthropogenic greenhouse gases, volcanic aerosols, anthropogenic
aerosols) or due to unforced internal variability
●
Resolution of atmospheric model: 2ºx2.5º latitude by longitude (about 250 km;
optional: cubed sphere version C90 (C180) with about 110 (55) km resolution
(1º and 0.5º, respectively), which will be used for the coupled model
intercomparison project CMIP6), 40 vertical layers up to 0.1 hPa, will be
extended to 96 and 102 layers (0.1 and 0.03 hPa, respectively), physics time
step of 30 minutes
●
Aerosols: solid or liquid particles in the atmosphere, which influence the
radiation fluxes, clouds, and atmospheric chemistry
●
Soil dust as a major aerosol is an integral part of ModelE2; 5 size bins covering
0.1 to 32 µm diameter (experimental: additional size bin from 32 to 64 µm)

30. Modeling Soil Dust in NASA GISS ModelE2
●
ModelE2 - Earth system model: to include all relevant processes, the energy and
mass fluxes within and between the components of the planet (atmosphere,
ocean, land, cryosphere, biosphere) to simulate climate variability in response to
changing boundary conditions by external drivers (e.g., forcing by solar
variability, anthropogenic greenhouse gases, volcanic aerosols, anthropogenic
aerosols) or due to unforced internal variability
●
Resolution of atmospheric model: 2ºx2.5º latitude by longitude (about 250 km;
optional: cubed sphere version C90 (C180) with about 110 (55) km resolution
(1º and 0.5º, respectively), which will be used for the coupled model
intercomparison project CMIP6), 40 vertical layers up to 0.1 hPa, will be
extended to 96 and 102 layers (0.1 and 0.03 hPa, respectively), physics time
step of 30 minutes
●
Aerosols: solid or liquid particles in the atmosphere, which influence the
radiation fluxes, clouds, and atmospheric chemistry
●
Soil dust as a major aerosol is an integral part of ModelE2; 5 size bins covering
0.1 to 32 µm diameter (experimental: additional size bin from 32 to 64 µm)

31. Modeling Soil Dust in NASA GISS ModelE2
●
ModelE2 - Earth system model: to include all relevant processes, the energy and
mass fluxes within and between the components of the planet (atmosphere,
ocean, land, cryosphere, biosphere) to simulate climate variability in response to
changing boundary conditions by external drivers (e.g., forcing by solar
variability, anthropogenic greenhouse gases, volcanic aerosols, anthropogenic
aerosols) or due to unforced internal variability
●
Resolution of atmospheric model: 2ºx2.5º latitude by longitude (about 250 km;
optional: cubed sphere version C90 (C180) with about 110 (55) km resolution
(1º and 0.5º, respectively), which will be used for the coupled model
intercomparison project CMIP6), 40 vertical layers up to 0.1 hPa, will be
extended to 96 and 102 layers (0.1 and 0.03 hPa, respectively), physics time
step of 30 minutes
●
Aerosols: solid or liquid particles in the atmosphere, which influence the
radiation fluxes, clouds, and atmospheric chemistry
●
Soil dust as a major aerosol is an integral part of ModelE2; 5 size bins covering
0.1 to 32 µm diameter (experimental: additional size bin from 32 to 64 µm)

32. Modeling Soil Dust in NASA GISS ModelE2
●
ModelE2 - Earth system model: to include all relevant processes, the energy and
mass fluxes within and between the components of the planet (atmosphere,
ocean, land, cryosphere, biosphere) to simulate climate variability in response to
changing boundary conditions by external drivers (e.g., forcing by solar
variability, anthropogenic greenhouse gases, volcanic aerosols, anthropogenic
aerosols) or due to unforced internal variability
●
Resolution of atmospheric model: 2ºx2.5º latitude by longitude (about 250 km;
optional: cubed sphere version C90 (C180) with about 110 (55) km resolution
(1º and 0.5º, respectively), which will be used for the coupled model
intercomparison project CMIP6), 40 vertical layers up to 0.1 hPa, will be
extended to 96 and 102 layers (0.1 and 0.03 hPa, respectively), physics time
step of 30 minutes
●
Aerosols: solid or liquid particles in the atmosphere, which influence the
radiation fluxes, clouds, and atmospheric chemistry
●
Soil dust as a major aerosol is an integral part of ModelE2; 5 size bins covering
0.1 to 32 µm diameter (experimental: additional size bin from 32 to 64 µm)

33. Surface Concentration of Total Dust

34. Simulated Mineral Fractions of Dust Column Mass

35. Simulated Mineral Fractions of Dust Column Mass

36. Iron Oxide Accretions With Other Minerals
ASSUMPTION: Accretion
Probability: P=f(M)x(1-f(Fe-ox))
Non-accreted mineral fractions at emission
Operations
(done in GCM)
Non-accreted other minerals
Accretions of iron oxides
with each of the other minerals
Non-accreted
iron oxides
Almost all of iron oxide mass accreted About 35% of non-FeOx dust with FeOx impurities

37. ●
Compilation of about 60 references from literature with mineral fraction
measurements (Perlwitz et al., ACP 2015b)
●
Limitations: Mostly campaign data or cruises, small sampling size, possible
biases depending on the methods; How to compare to model climatology?
●
For details on the individual measurements (type, location, time period,
sample size) from the references see supplemental material of Perlwitz et
al., 2015b.
Evaluation of simulated mineral fractions

38. Locations of Measurements from Literature
Perlwitz et al., ACP (2015b)

39. Mineral Fractions Simple (SMF) vs. New (AMF) Method
Perlwitz et al., ACP (2015b)

40. Mineral Fractions Simple (SMF) vs. New (AMF) Method

41. Evaluation - Silt Sized Dust

42. Playing with ice nucleation

Very few attempts to calculate ice forming nuclei (IFN)
abundance from mineral species simulated with a global
model

Kaolinite and illite/montmorillonite: Hoose et al., ERL (2008)

K-feldspar: Atkinson et al., Nature (2013)

Approach: Using directly the mineral fractions in soils for the
mineral fractions of dust aerosols

43. What about IFN numbers by feldspar if we use
our improved dust mineral model instead?
Four experiments:
1.Baseline experiment: Same set up as by Atkinson et al.
(active sites parameterization with nucleation densities at
fixed temperatures), mineral fractions in soil projected onto
AeroCom dust emission.
2.Aerosol mineral fraction (AMF) method: used for minerals in
dust module as described in Perlwitz et al. (2015a,b).
3.AMF AeroCom: Mineral fractions from AMF method,
projected onto AeroCom dust emissions.
4.AMF Feldspar: Sensitivity to a feldspar distribution that is
steeper toward larger particle sizes.

45. Can we reproduce the previous study?

46. Sensitivity to size distribution

47. Current and Planned Work With Dust Aerosols in ModelE
Dust emission =
f(wind speed)
5 size bins (0.1-32 µm)
8 minerals plus FeOx-mineral
accretions
Constraints: soil moisture, vegetation
cover, size dependent mineral
composition of soils
Anthropogenic dust sources
Atmospheric load of
8 minerals plus
FeOx-mineral accretions
Advection
wet deposition,
dry turbulent deposition,
gravitational deposition of
8 minerals plus accretions
with different densities,
different solubilities
Sulfate and nitrate
precursors (gases)
Other Aerosols
Phytoplankton
in oceans
Carbon cycle
Mineral dependent
pools for processing of
bioavailable iron
Radiative
forcing with mineral
dependent refractive
indexes
Climate State
Minerals as CCN
and IFN with mineral
dependent activation
Heterogeneous chemistry
(sulfate, nitrate, ammonium on
dust) with mineral dependent
reaction probabilities

48. Summary

Dust is an important aerosol which is involved in different physical and chemical
processes in the Earth system

Using one type of dust in modeling has provided important knowledge over the last
20 years about the role of dust, but it comes with limitations. To overcome this
limitations the mineral composition of dust needs to be simulated

We have developed an improved approach by applying Claquin's Mean
Mineralogical Table in combination with brittle fragmentation theory and
measurements to derive aerosol size distributions at emission for each mineral type

We simulate emission, transport, and deposition of a mixture of non-accreted
minerals and accretions of iron oxides with each of the other minerals in clay and silt
size bins

The comparison between the SMF method and the AMF method shows robust
improvements for latter, although some deficiencies remain

Claquin et al./Nickovic et al.' s MMT are very useful data. They have to be properly
applied to derive the mineral fractions of dust aerosols

Simulating minerals opens door to many interesting applications, e.g., ice nucleation