Reconstruction of dust emission from Western Africa since 1900

1
Reconstruction of dust emission from Western
Africa since 1900

• Paul Ginoux
• NOAA-GFDL

GEIA Open Conference Paris, Nov 29-Dec 1, 2006
2
Outline

• Why reconstructing dust emission?
• Physical processes involved
• Methods of reconstruction
• Application effects on climate of dust increase
  during the Sahel drought
• Outstanding issues

3
Observed Variability of Dust for the last 100k
years
Laser Light Scattering by dust in Greenland ice
core (Ram et al., 1997)
LGM
Eemian
Holocene
Pleistocene

• Dust varies by a factor 5 or more
• over geological cycles glacial/inter-glacial
• over a surprisingly constant period of 11 years

12k B.P.
11 years
4
Observed Variability of Dust for the last 6000
Years
Dust deposition from sediments SE of lake Chad
(Street-Perrott et al., 2000)
Present value 9.9 10-3g/cm2/yr (McTainsh, 1980)
Severe drought with 2 rainfall minima
Dust deposition rates have increased by a factor
5 from wet (5000 yr BP) to dry (since 1000 yr BP)
periods. Present value has been observed before
(1000 yr BP).
Fall of lake level, Shift of vegetation
Perennial freshwater lake
5
Observed Variability of Dust for the last 50 Years
Dust concentration at Barbados (Prospero and
Lamb, 2003)
Factor 4 increase
Sahel drought
Since 1970ies dust concentration in Caribbean
(Prospero and Lamb, 2003) and dust deposition in
French Alps (De Angelis and Gaudichet, 1991) have
increased by a factor 4-5
Correlation at Barbados (Prospero and Lamb, 2003)
Barbados Dust
Sahel Precipitation Index (previous year)
6
What can we conclude from the observations?

• There is a natural variability of dust on
  different time scales. This variability can be of
  up to a factor of 5.
• The most regular variation is the 11-year cycle
  observed over 150,000 years.
• Drought in Africa have been observed 1000 years
  ago with rate of dust deposition in Sahel similar
  with 1980ies values.
• Since the early 1970ies African dust is 5 times
  higher than previously observed.
• There is a large gap of data between the time
  series of U. of Miami data starting in 1965 and
  ice cores data.

7
Physical Processes Dust uplifting
Field and laboratory studies show that dust
concentration is enhanced by orders of magnitude
in presence of saltation (hopping motion) of
large sand particles. Without saltation impacts a
surface of dust particles is remarkably stable
(Bagnold, 1941).
Vertical Flux of clay and silt
Fa Q
Horizontal flux of Sand
Efficiency factor Marticorena and Bergametti
(1995)
QC sp u2(u-ut)
The threshold velocity ut is a function of
particle radius, surface roughness height,
temperature, and surface wetness
8
Attempt to simulate dust increase during Sahel
drought
Barbados dust with GFDL AM2 observed SST
The model largely overestimate dust concentration
in 50ies and 60ies by considering only surface
wind as controlling factor.
1965
1950
Barbados dust with NCAR MATCH NCEP re-analysis
(Mahowald et al., 2002)
80
80
Dynamic vegetation ( land use changes) better
constrained dust simulation
1965
1984
9
Solution I By-pass the sources characterization
Dust emission as a function of Sahel
precipitation index (SPI) of previous year

• Methodology
• Dust at Barbados is correlated with the SPI of
  the previous year (Prospero and Lamb, 2003).
• Dust at Barbados is correlated with its emission
  from West Africa
• Dust emission f(SPI)
• The scaling factor is assumed constant all year
  long.
• N.B. Lowest obs. dust supposed in 1950ies but not
  data. To be conservative, minimum emission
  (factor 5 reduction) is associated with 1950.

Similar Method Moulin and Chiapello (2006) have
also related Barbados dust with SPI but add a
contribution from NAO index and a continuously
increasing anthropogenic emission.
10
Dust at Barbados using the Sahel Precipitation
Index as a proxy for constraining dust emission
Observations
Simulation
Dust concentration can be easily simulated as
long as the SPI is available.

• Limitations of the method
• Ignores other possible factors changes of land
  use
• Although N. Africa represents 2/3 of global
  emission, it ignores variability of emission over
  other continents

11
Solution 2 Solve dynamically the dust source and
emission
Dust source characteristics (roughness length,
soil moisture, land use changes) are simulated by
the GFDL LM3v model part of the GFDL coupled
models.
Shevliakova et al., 2006
12
Comparison Leaf Area Index (LAI) simulated with
LM3v and retrieved from AVHRR data
AVHRR ISLSCP2
LM3v Model
January 1984
July 1984
13
Time dependent dust sources
Dust emission as a function of LAI, SAI, dead
biomass, soil moisture, and snow as calculated
dynamically by AM2-LM3v
Dust emission from bare ground
14
Dust simulation by coupling dust emission with
AM2-LM3v-observed SST
Simulation
Observations
Dust concentration can be simulated for past,
present and future by constraining dust emission
with land characteristics. The effects of changes
of land use can be evaluated. The feedbacks can
be studied
Potential problems The dust emission will
strongly depends on how realistic is the
surface Characteristics simulated by the land
model
15
Application Effects of Sahel drought
Simulation of climate at equilibrium (100 year
model run) with GFDL AM2mixed layer
ocean Control all aerosols corresponding to
1990 Perturbed all aerosols corresponding to
1990, but dust 1965
Forcing (Control-Perturbed) at surface and
tropopause
16
Difference of surface temperature
Sm2.1 Control-perturbed (mean years 41-100) with
95 confidence level
17
Difference of precipitation
Sm2.1 Control-perturbed (mean years 41-100) with
95 confidence level
Units cm/day
18
Implications of a 5 x emission increase

• Surface forcing of -1W/m2 globally, -15W/m2 over
  N. Africa
• Reduced precipitation over Sahel (positive
  feedback), Amazon, Mediterranean basin 10-30
  mm/month
• Reduced SST over eastern part of North Atlantic
  0.5 degree

19
Outstanding issue Improving dust source inventory
20
Natural Sources Ephemeral Lakes
21
Case of Bodele Depression
22
Anthropogenic Sources Aral sea
Sources associated with human activities
disturbing soils or drying up lake for
irrigation. Estimates Tegen and Fung (1995)
50 Mahowald et al. (2002) 10-50 Moulin and
Chiapello (2006) 30-50
23
Improving dust source inventories in the
framework of AEROCOM
Comparison of Ginoux et al (2001), Tegen et al.
(2002), Zender et al. (2003) inventories with AOD
from AERONET, AVHRR, TOMS, and concentration from
U. of Miami, deposition rates, size distribution
from AERONET (Cakmur et al., 2005)
Error associated with each inventory
24
Conclusions

• Dust natural variability can reach more than a
  factor of 5, as observed over long time records
• The latest major increase of dust load in the
  late sixties has been associated with the Sahel
  drought and some anthropogenic contribution
• Most GCMs climate simulations have not
  considered dust annual variability.
• Two methods have been proposed to simulate dust
  emission beyond the last two decades
• Reconstructing dust emission using the Sahel
  Precipitation Index as a proxy
• Imbed dust emission into dynamic vegetation model
• Both methods have their advantages and
  limitations. The first one is simple and is
  applied to study the effects on climate of
  increased dust during the latest Sahel drought.
  The second one is complex but allows to study
  feedbacks and to evaluate anthropogenic
  contributions.
• Still, dust source inventory needs some
  improvement which could be reached by using MODIS
  data and high resolution model. This could be
  realized in the framework of AEROCOM in
  association with GEIA.