Geoscience Reference
In-Depth Information
basis of another surface physical attribute. One scheme,
which led to a notable improvement in the match between
model dust emission and qualitative satellite values of
dust loadings, was the Ginoux et al. (2001) model. Based
on TOMS AI data, Prospero et al. (2002) and Washing-
ton et al. (2003) identified topographical depressions in
arid regions as key dust sources. Ginoux et al. (2001) also
implemented these ideas in a model, thereby introducing
the preferred source area concept. Similar adjustments to
the emission scheme were then introduced in other mod-
els (Tegen et al. , 2002; Luo, Mahowald and del Corral,,
2003; Zender, Bain and Newman, 2003). There are sev-
eral reasons behind the simplification in dust emissions
as represented in these models. These include: (1) numer-
ical models typically underestimate surface windspeeds,
some by more than 50 % (Washington et al. , 2006a) and
so adjustment to entrainment is more easily made to other
controls, including U t and/or the parameters that control
the mass flux of dust as these need to be prescribed in
the model rather than calculated at each time step. Any
adjustment to parameters that are calculated at each time
step, such as wind, would be much more complicated to
implement. (2) The real values of key controls on dust
emission, such as surface roughness, nonerodible fraction
and particle size distribution, have never been measured
across the vast majority of key source areas such as the
Sahara. Therefore the easiest way to improve model dust
emission is simply to alter these unknown values so that
model emissions appear reasonable on the basis of either
satellite estimates of dust distribution or background mea-
surements such as those from the Miami Aerosol Group.
In contrast, more realistic, physically based and more
complex dust emission schemes are increasingly being
used in numerical models. Examples are the Marticorena
and Bergametti (1995) and Shao, Raupach and Leys
(1996) schemes (see Laurent et al. , 2008 and Darmen-
ova et al. , 2009). In these more complex schemes, the
erosion threshold is typically calculated as a function of
surface roughness (both overall and that of the erodible
surface component), soil moisture and particle size distri-
bution. Saltation is a function of friction velocity, fraction
of erodible to total surface and particle size distribution.
Vertical flux is represented through an empirical relation-
ship linking the ratio of the dust flux to the horizontal
(sandblasting) flux to the soil clay content (Marticorena
and Bergametti, 1995).
Dust emission from models has been shown to dif-
fer by an order of magnitude, even when calculated over
a limited source area and for only a few days (e.g. see
Todd et al. , 2008). Improvement to entrainment schemes
hinges on direct collaboration between geomorphologists
science. Key problems include: (a) representation of dust
entrainment at the model grid box scale (tens of kilo-
metres) when relationships between key parameters have
been derived from either point source field data or ide-
alised from wind tunnel experiments (Iversen and White,
1982); (b) lack of observed data over vast key source re-
gions on parameters such as soil particle size; (c) inclusion
of processes currently not represented in the models, e.g.
supply limitations imposed by surface crusting (Lopez,
1998) and alternatives to the horizontal and vertical sedi-
ment flux distinction such as autoabrasion (Warren et al. ,
2006).
20.1.3.3
Dust transport and deposition in models
In contrast to entrainment, transport schemes in numer-
ical models are a relative strength since there are few
other ways of determining dust pathways. Conversely,
there are also few data sets with which to evaluate model
performance. Tracer advection schemes in the models are
typically used to determine dust transport. These include
vertical motion through convection, turbulent mixing and
gravitation settling. Wet deposition within and below
cloud height is often determined by a precipitation rate
based experimentally on a derived dependence on particle
size (e.g. Woodward, 2001).
20.1.4
Distribution of dust
The release of long-term (1979-1993) data from TOMS
AI (Herman et al. , 1997) allowed the first, albeit qualita-
tive, complete picture of dust loadings in the atmosphere
(Prospero et al. , 2002; Washington et al. , 2003) (Figure
20.7). The largest area with high values is a zone that
extends from the eastern subtropical Atlantic through the
Sahara Desert to Arabia and southwest Asia. In addition,
there is a large zone with high AI values in central Asia,
centred over the Tarim basin and the Taklimakan Desert.
Central Australia has a relatively small zone, located in
the Lake Eyre basin, while southern Africa has two zones,
one centred on the Makgadikgadi basin in Botswana and
the other on the Etosha Pan in Namibia. In Latin America,
there is only one easily identifiable zone. This is in the
Atacama and is in the vicinity of one of the great closed
basins of the Altiplano. North America has only one very
small zone with high values, located in the Great Basin
(the drained Owens Lake). The importance of these differ-
ent dust 'hotspots' can be gauged by looking not only at
their areal extents but also at their relative AI values (Ta-
ble 20.2). This again brings out the very clear dominance
Search WWH ::




Custom Search