Observational Requirements for Modeling of Global...
45
and Twomey, 1994). Thus, the indirect effect may not have the same spatial pattern as the
direct effect. A recent GCM estimate suggests that the global indirect radiative effect may be
comparable in magnitude to the direct effect (Jones et aI., 1994). The magnitude of the indirect
microphysical effect is an even greater unknown. A final complicating factor is that biogenic
sources of sulfate in the oceans may contribute their own indirect effect far from industrial
regions (Charlson et aI., 1987), but climate changes in plankton emissions into the atmosphere
are completely unknown.
One possible hint of a regional indirect climate effect is the observation that the diurnal temperature range of surface air has decreased by about 0.5°C over the past four decades over much
of the Earth's land area (Karl et a\., 1993) with most of the change having occurred at night.
Greenhouse gas increases and the direct effect of aerosols seem to be insufficient to explain the
observed narrowing. Hansen et al. (1995a) use GCM simulations to argue that either low or
midlevel cloud increases over this time are the only plausible candidates to produce the observed
magnitude of the effect. They suggest that anthropogenic aerosol increases might have provided
the required stimulus for a low cloud cover increase. Indeed, cloud cover has increased over
both the United States and Europe during the past 50 years (Henderson-Sellers, 1992). But
whether this can be traced to aerosols is as yet unknown; the GISS GCM predicts increasing
cloud cover over midlatitude land in a warming climate without anthropogenic aerosol climate
forcing (Fig. 2.7), and climatic increases in the vigor of the hydrologic cycle might also explain
the observed weakening of the diurnal cycle (Dai, 1995).
2.4 Observational Strategies for Improving Climate
Model Predictions
With this background, we can now assess the suitability of existing observations and suggest
strategies for future data acquisition that would maximize the impact on climate model performance. Needed data fall into one of three categories, according to how they are to be used
by the climate model. First, there are validation data required to determine how well GCMs
simulate the mean state and variability of parameters in the current climate. Second, we need
physical process data to determine why the GCM performs incorrectly in a given situation
and thus to provide insight into parameterization improvements. Third, we require monitoring
data to ultimately detect the elements of decadal climate change. We discuss specific data
requirements for each class below.
a) Validation data sets.
The basic requirement for validation data are that they be global and document both the mean
state and variability on diurnal, seasonaL and interannual time scales. A non-rigorous but
practically useful accuracy standard for such data sets is that uncertainties be significantly less
than the range of differences in a given parameter among various climate models. Since almost
every GCM in the world is participating in the AMIP intercomparison, required accuracy
estimates of this kind will be available in the near future. A sample of the range of model
predictions for several parameters from an earlier intercomparison is given in Table 2.1. The
status of data sets for particular parameters follows.
-TO A radiation budget.
ERBE (Barkstrom, 1984) has become the standard of comparison for many GCMs, but the
data set covers only 4 complete years, with a current gap until CERES flies on the TRMM
satellite in 1997. The angle and diurnal models used to derive the fluxes are crude, but errors
are probably < 10 Wm- 2 regionally. The net annual imbalance of 6 Wm- 2 presumably reflects
45
and Twomey, 1994). Thus, the indirect effect may not have the same spatial pattern as the
direct effect. A recent GCM estimate suggests that the global indirect radiative effect may be
comparable in magnitude to the direct effect (Jones et aI., 1994). The magnitude of the indirect
microphysical effect is an even greater unknown. A final complicating factor is that biogenic
sources of sulfate in the oceans may contribute their own indirect effect far from industrial
regions (Charlson et aI., 1987), but climate changes in plankton emissions into the atmosphere
are completely unknown.
One possible hint of a regional indirect climate effect is the observation that the diurnal temperature range of surface air has decreased by about 0.5°C over the past four decades over much
of the Earth's land area (Karl et a\., 1993) with most of the change having occurred at night.
Greenhouse gas increases and the direct effect of aerosols seem to be insufficient to explain the
observed narrowing. Hansen et al. (1995a) use GCM simulations to argue that either low or
midlevel cloud increases over this time are the only plausible candidates to produce the observed
magnitude of the effect. They suggest that anthropogenic aerosol increases might have provided
the required stimulus for a low cloud cover increase. Indeed, cloud cover has increased over
both the United States and Europe during the past 50 years (Henderson-Sellers, 1992). But
whether this can be traced to aerosols is as yet unknown; the GISS GCM predicts increasing
cloud cover over midlatitude land in a warming climate without anthropogenic aerosol climate
forcing (Fig. 2.7), and climatic increases in the vigor of the hydrologic cycle might also explain
the observed weakening of the diurnal cycle (Dai, 1995).
2.4 Observational Strategies for Improving Climate
Model Predictions
With this background, we can now assess the suitability of existing observations and suggest
strategies for future data acquisition that would maximize the impact on climate model performance. Needed data fall into one of three categories, according to how they are to be used
by the climate model. First, there are validation data required to determine how well GCMs
simulate the mean state and variability of parameters in the current climate. Second, we need
physical process data to determine why the GCM performs incorrectly in a given situation
and thus to provide insight into parameterization improvements. Third, we require monitoring
data to ultimately detect the elements of decadal climate change. We discuss specific data
requirements for each class below.
a) Validation data sets.
The basic requirement for validation data are that they be global and document both the mean
state and variability on diurnal, seasonaL and interannual time scales. A non-rigorous but
practically useful accuracy standard for such data sets is that uncertainties be significantly less
than the range of differences in a given parameter among various climate models. Since almost
every GCM in the world is participating in the AMIP intercomparison, required accuracy
estimates of this kind will be available in the near future. A sample of the range of model
predictions for several parameters from an earlier intercomparison is given in Table 2.1. The
status of data sets for particular parameters follows.
-TO A radiation budget.
ERBE (Barkstrom, 1984) has become the standard of comparison for many GCMs, but the
data set covers only 4 complete years, with a current gap until CERES flies on the TRMM
satellite in 1997. The angle and diurnal models used to derive the fluxes are crude, but errors
are probably < 10 Wm- 2 regionally. The net annual imbalance of 6 Wm- 2 presumably reflects