5.4.1.4 Comparison of modelled and satellite-derived aerosol optical depth
Comparison of model results with remote surface observations of the major components
making up the composition of the atmospheric aerosol provide a test of whether
the models treat transport and removal of individual species adequately. But
as noted above, the difference between the vertical distribution of species
among the models is significant (i.e. a difference of more than a factor of
2 in the upper troposphere) and the global average abundance of individual components
varies significantly (more than a factor of 2) between the models, especially
for components such as dust and sea salt. Two methods have been used to try
to understand whether the models adequately treat the total aerosol abundance.
The first, comparison of total optical depth with satellite measurements, was
employed by Tegen et al. (1997), while the second, comparison of total reflected
short-wave radiation with satellite observations, was employed by Haywood et
al. (1999). We examined both measures in an effort to understand whether the
model-predicted forcing associated with aerosols is adequate.
Figure 5.11 shows the zonal average
optical depth deduced from AVHRR data for 1990 by Nakajima and Higurashi (Nakajima
et al., 1999) and for an average of the time period February 1985 to October
1988 by Mishchenko et al. (1999) and by Stowe et al. (1997). The two results
from Nakajima et al. (1999) demonstrate the sensitivity of the retrieved optical
depth to the assumed particle size distribution. Results from the models which
participated in the intercomparison workshop are also included. Because the
GISS, CCM1, ECHAM/GRANTOUR and ULAQ models all used the same sources, the differences
between these models are due to model parametrization procedures. The GOCART
(GSFC) model used a source distribution for sea salt that was derived from daily
varying special sensor microwave imager (SSM/I; Atlas et al., 1996) winds for
1990 and was, on average, 55% larger than the baseline sea salt source specified
for the model workshop. The MPI/Dalhousie model used monthly average dust and
sea salt distributions from prior CCM1 model simulations (cf., Lohmann et al.,
1999b,c).
Workshop participants were asked to report their derived optical depths. However,
these varied widely and were often much smaller than that derived here. Therefore,
we constructed the optical depths shown in Figure 5.11
from the frequency distribution of relative humidity from the T21 version of
the ECHAM 3.6 general circulation model, together with the reported monthly
average distributions of aerosol mixing ratios. The model optical depths were
derived using extinction coefficients at 0.55 mm for dry sea salt of 3.45 m2g-1,
0.69 m2g-1, and 0.20 m2g-1 for diameters
in the size range from 0.2 to 2 mm, 2.0 to 8 mm and 8 to 20 mm, respectively,
and an extinction coefficient of 9.94 m2g-1 for sulphate.
The humidity dependence of the extinction for sea salt and dust was determined
using the model described in Penner et al. (1999a). The dust extinction coefficients
were from Tegen et al. (1997) and organic and black carbon extinction coefficients
for 80% relative humidity were from Haywood et al. (1999). We also examined
the sensitivity of the modelled optical depths to a factor of two increase in
the DMS flux and to the use of monthly average sea salt fluxes derived using
the SSM/I wind fields. The OC and BC extinction coefficients were also varied,
adding the humidity dependence determined by Penner et al. (1998b). Finally,
the extinction coefficient for sulphate was altered to that calculated for an
assumed ratio of total NH3 and HNO3 to H2SO4
of four. The line in the graph labelled “summed sensitivity study”
shows the results for the ECHAM/GRANTOUR model when these para-meters were varied.
Most of the difference between the summed sensitivity case and that for the
baseline ECHAM/GRANTOUR model is due to the use of larger DMS and sea salt fluxes.
The satellite-derived optical depths from Stowe et al. (1997) are lower on
average by 0.05 and by 0.03 than those from Mischenko et al. (1999) and result
2 from Nakajima et al. (1999), respectively. The latter two retrievals make
use of a two-wavelength technique which is thought to be more accurate than
the one-wavelenth technique of Stowe et al. (1997). However, it is worth bearing
in mind that most of the difference in retrieved aerosol optical depth may be
related to cloud screening techniques (Mishchenko et al., 1999) or to assumed
size distribution (Nakajima et al., 1999).
Modelled optical depths north of 30°N are sometimes higher and sometimes
lower than those of the retrieved AVHRR optical depths. For example, there is
an average difference of 0.13 in July for the ULAQ model in comparison with
result 2 for Nakajima et al. (1999) while the average difference is -0.09 in
January for the ECHAM/GRANTOUR model in comparison with the retrieved optical
depths from Mishchenko et al. (1999). The modelled optical depths in the latitude
band from 30°N and 40°N are systematically too high in July. For example,
the average of the modelled optical depths is larger than the satellite-derived
optical depth of Nakajima, Mishchenko, and Stowe on average by 0.06, 0.05 and
0.04, respectively. We note that sulphate and dust provide the largest components
of optical depth in this region with sea salt providing the third most important
component. Since the sources represent the year 2000, while the measured optical
depths refer to an average of the years 1985 to 1988, some of the overprediction
of optical depth may be associated with higher sources than the time period
of the measurements. The black dashed line shows the estimated optical depths
from the ECHAM/GRANTOUR model with the larger sea salt fluxes deduced from the
SSM/I winds, with doubled DMS flux, and with optical properties for an assumed
ratio of total NO3 to H2SO4 of 4:1. Comparison
of these results with those of the retrieved optical depths shows that the uncertainties
in these parameters lead to changes in optical depth that are of the order of
0.05 or more.
Continues on next page
