5.4.4 Model Validation of Indirect Effects
Validation of the simulations from global models is an essential component
of estimating and reducing the uncertainties in the indirect forcing. Comparisons
of observations and modelled concentrations of chemical species have been discussed
in Section 5.4.1.3 while comparisons of modelled and satellite-derived
aerosol optical depths were discussed in Section 5.4.1.4.
Here, comparisons with observations of several other model products important
for indirect forcing are examined. Unfortunately, there is only a very small
set of observations of the physical, chemical, and radiative properties of clouds
from in situ methods available. Thus, validations with these types of datasets
are left to limited temporal and spatial scales and to comparing relationships
among various quantities. Lohmann et al. (2001), for example, compared prognostic
simulations with observations of the relationships between particulate sulphate
and total particle mass, between particle number concentration and sulphate
mass, and between Nd and sulphate mass. The relationships between
sulphate mass and total particle number concentrations was larger than observations
in the case of internal mixtures but was smaller than observations in the case
of external mixtures. Ghan et al. (2001a) found that the results of their determination
of Nd using their mechanistic parametrization were comparable to
the results using the empirical parametrization of Boucher and Lohmann (1995).
Such tests are important for large-scale model parametrizations because comparisons
of absolute concentrations on the scale of the model grid size are difficult.
Satellites offer observations over large temporal and spatial scales; however,
for the derived parameters of interest, they are much less accurate than in
situ observations. Han et al. (1994) retrieved an reff for liquid-water
clouds from ISCCP satellite data that showed a significant land/sea contrast.
Smaller droplets were found over the continents, and there was a systematic
difference between the two Hemispheres with larger droplets in the Southern
Hemisphere clouds. The reff calculated by different models and the
observations from Han et al. are shown in Table 5.12.
Since the reff tends to increase with increasing height above cloud
base and the satellite observations of reff are weighted for cloud
top, the satellite observations will tend to overestimate the overall reff
compared with that determined from in situ studies. In situ data sets against
which to make absolute comparisons are few in number (e.g. Boers and Kummel,
1998). However, for now model evaluations are better done using the contrasts
in reff between the land and ocean and between the Southern Hemisphere
and the Northern Hemisphere. Most of the models listed in Table
5.12, with the exception of Roelofs et al. (1998), approximate the difference
between reff over the Southern Hemisphere ocean vs the Northern Hemisphere
ocean. Over land, the Southern Hemisphere vs Northern Hemisphere difference
from Roelofs et al. (1998) is closest to the observed difference. For reff
over Southern Hemisphere land vs Southern Hemisphere ocean, several models are
relatively close to the difference from the observations (Boucher and Lohmann,
1995; Chuang et al., 1997; Roelofs et al., 1998; Lohmann et al., 1999b,c). Some
models compare with observations better than others, but there is no model that
is able to reproduce all the observed differences. The reff calculated
with the same parametrization but using different GCM meteorologies are quite
different (compare Jones and Slingo (1996) vs Boucher and Lohmann (1995)). As
noted above, Jones and Slingo determined the “cloud top” reff
by assuming a LWC profile that increased with height from cloud base to cloud
top. Such a profile is more similar to observed profiles and might be expected
to produce a better comparison with the observations. While the Jones and Slingo
(1996) model does reasonably well in terms of hemispheric contrasts, their results
indicate a land-ocean contrast in the opposite direction to that from the other
models and the observations. We note that many factors may affect the results
of this type of comparison. For example, Roelofs et al. (1998) estimate the
sensitivity of the reff calculations to uncertainties in the sulphate
concentration field, cloud cover and cloud liquid-water content to be of the
order of a few micrometres. Moreover, the satellite determination of reff
is probably no more accurate than a few micrometres (Han et al., 1994).
| Table 5.12: Cloud droplet effective radius of warm
clouds (in µm). |
 |
| All results for 45° S to 45°N |
Ocean Southern Hemisphere
|
Ocean Northern Hemisphere
|
Land Southern Hemisphere
|
Land Northern Hemisphere
|
Total
|
|
| Han et al. (1994) |
11.9
|
11.1
|
9.0
|
7.4
|
10.7
|
| Boucher and Lohmann (1995) |
8.9 to 10.1
|
8.3 to 9.3
|
5.4 to 8.7
|
4.9 to 8.0
|
|
| Jones and Slingo (1996) |
9.6 to 10.8
|
9.0 to 10.4
|
10.2 to 11.8
|
9.9 to 10.8
|
9.5 to 11.1
|
| Roelofs et al. (1998) |
12.2
|
10.3
|
8.8
|
6.9
|
10.4
|
| Chuang et al. (1997) |
11.6 to 12.0
|
10.7 to 11.4
|
8.8 to 9.1
|
8.6 to 9.0
|
10.7 to 11.2
|
| Lohmann et al. (1999b,c) |
10.7
|
10.2
|
8.3
|
4.9
|
|
| Rotstayn (1999) |
11.2
|
10.9
|
9.8
|
9.5
|
10.7
|
| Ghan et al. (2001a,b) |
|
|
|
|
11.0 to 11.7
|
|
Continues on next page
