Numerical Dust Models - Mineral Dust: A Key Player in the Earth System - page 211

Experiment A, as reported by Textor et al. ( 2006 ), is 1,640 Tg/a for global annual
dust emissions, 20.5 Tg for global dust burden and 4.04 days for lifetime, with large
variations for the individual models. A revised analysis of AeroCom dust including
CAM and ECMWF model results, but excluding models with fixed dust emission
fields, suggests slightly lower median emissions of 1,123 Tg/a and 4.6 days lifetime
(Table 9.1 ). Large variations are documented not only for emissions but also for
lifetime and mass extinction efficiencies, making it difficult to attribute performance
differences to differences in emission process modelling only.
In Textor et al. ( 2006 ), the model diversities • are defined as the standard
deviation of the model results normalized with the average of all models, expressed
in percentage. The authors find that for dust emissions in Experiment A, the model
diversity ı is 49 %, which is considerably higher than for the other aerosol types
except sea salt aerosol. Similarly, the model diversities are 40 % and 43 % for
burden and residence time, respectively. These are lower than the emission diversity,
probably because global dust burdens are dominated by fine particles. Transported
dust clouds, with larger fractions of fine dust particles, can be better constrained due
to numerous observations of optical depth by satellite. By contrast, few observations
exist to constrain the simulation of coarse dust, which may dominate emission
fluxes.
The results for Experiment B with harmonized emissions, though distributed in
particle size differently, cannot be compared directly to these results, since a lower
number of models contributed. However, the diversities from the model results for
Experiment B were at a similar level as in Experiment A for the atmospheric burdens
and residence times (Textor et al. 2007 ), even though (almost) identical emission
fields were used for dust mass. This confirms the crucial role of dust size but also
that of wet deposition fluxes computed by the different models. The latter may be
due to a lack of understanding of the hygroscopic behaviour of dust, which in turn
influences the washout and rainout efficiencies of the dust particles. The globally
averaged split between wet and dry deposition for the different models is shown
in Fig. 9.4 . While dry deposition is very important in all models, wet deposition
contributesupto65%insomemodels.
The emitted dust mass fluxes are divided into different size fractions, which are
then further transported and deposited in the models. Unfortunately, the diagnostics
of the models collected at the time were not very precise with respect to the actual
size distribution used. An average redistribution of the aggregated size bins for dust
aerosol is shown in Fig. 9.5 for the models that contributed to AeroCom Experiment
B. The differences in the size distribution lead to high diversities in the modelled
deposition rates and atmospheric lifetimes, even for Experiment B, which uses the
same emission mass fluxes for all models.
Focussing on dust aerosol simulations, Huneeus et al. ( 2011 ) evaluated the
results of 14 different models that contributed to the AeroCom project with
detailed comparisons of observations for aerosol optical thicknesses, near-surface
concentrations and deposition fluxes (Fig. 9.6 shows the median dust aerosol optical
thickness distribution for the AeroCom model ensemble). The authors find that the
averages and seasonal variability of vertically integrated mineral dust parameters