Dust Deposition - Mineral Dust: A Key Player in the Earth System - page 186

1.9

Iso-log (geometric mean diameter)

1.8

Iso-gradient

1.7

1.6

1.5

1.4

1.3

1.2

1.1

1

0.9

0.8

468 10 12 14 16 18 20 22 24 26 28 30

number of size bins

Fig. 8.2 Mass error ratios using an iso-log size bin scheme ( open squares ) and an iso-gradient
scheme ( filled circles ) plotted against the number of size bins used. The error ratio in each case is
the ratio of the mass of airborne dust aerosol remaining after two days of dry deposition calculated
using 1,000 size bins (reference) or a reduced number of size bins (from 4 to 30 here). This is
a measure of the error made when using a limited number of bins for representing the dust size
distribution and the subsequent dry deposition. The initial mass size distribution is that shown in
Fig. 8.1 (Adapted from Forêt et al. 2006 )

controls the model accuracy and concludes that a minimum of 12 iso-log bins are
necessary to correctly simulate both the number and the mass size distributions of
tropospheric aerosols.
To limit the number of bins but keeping a good accuracy, Forêt et al. ( 2006 )
proposed a new approach for binning the size distribution into a limited number
of size intervals. Their method, called iso-gradient scheme, assumes that it is more
efficient to have more size bins for the size domain where the removal processes
are strongly size dependent rather than having size bins equally distributed over the
whole domain. Thus, they split the size distribution into bins having an equal range
in V d , i.e., following the size dependence in dry deposition velocity by satisfying an
“iso-gradient” condition in V d . Figure 8.2 shows a comparison of the accuracy of
both approaches when compared to a highly size resolved simulation using 1,000
size bins. The new approach largely reduces the discretization errors for low bin
numbers (Fig. 8.2 ). This has also been confirmed in case of full 3D dust simulations
(Menut et al. 2007 ).

8.3 Dust Deposition Measurements

Measurements of dry deposition fluxes are hard to perform, since it is difficult to
develop a sampler that correctly mimics natural surfaces (Fowler et al. 2009 ). As
the deposition in the surface layer is highly dependent on the properties of that
surface, such differences between the sampler and the surface generate significant