Observational Requirements for Modeling of Global...
47
-Cloud particle size.
Satellite near-infrared retrievals of droplet effective radius for liquid clouds outside the polar
regions have recently become available (Han et al., 1994). These refer primarily to cloud top
conditions; the vertical profile of droplet size through clouds is unknown except for isolated
in situ measurements. Particle size for ice clouds is much more uncertain; comparable global
surveys are only beginning to be produced, and remote retrievals from field programs typically
differ markedly from in situ ice particle size estimates. Ice scattering properties are not welldefined by assuming "equivalent spheres" for Mie scattering calculations, the standard approach
in GCMs. More realistic ice phase functions have been defined (Macke, 1993); it remains to
be seen whether a universal phase function for all ice particle shapes is adequate for GCM
purposes. Only GCMs with prognostic cloud water budgets can currently make use of such
information.
-Cloud liquid and ice water path.
Microwave retrievals of vertically integrated liquid water path are available over ocean, but not
over land. But the algorithms used by different workers produce dramatically different results,
even disagreeing about whether liquid water path is higher in midlatitudes or the tropics (Lin
and Rossow, 1994). Such data are thus virtually useless for GCM validation, a point missed by
many climate modelers who have nonetheless compared their prognostic models to one of the
published estimates. The problems involve whether clear points are included in the average,
the threshold assumed for detecting cloud water, and errors in the assumed water vapor profile
and cloud temperature. Ice water path has not yet been retrieved from satellite; this is perhaps
the major gap in validation data for GCMs with prognostic cloud water schemes. ISCCP can
indirectly estimate ice water path by combining microwave liquid water path and near-infrared
particle size estimates with its own column optical thickness retrieval (Lin and Rossow, 1994),
but this depends on their assumptions about ice particle size and phase function. A major
issue for GCMs is where the transition from liquid to ice occurs as a function of temperature,
cloud depth and age, and other parameters (cf. Mitchell et al., 1989; Li and LeTreut, 1992; Del
Genio et al., 1996).
-Water vapor.
Rossow (1996, this volume, chapter 8) provides a comprehensive discussion of existing water
vapor observation capabilities. Global precipitable water estimates from a combination of microwave and radiosonde data will be produced by the GEWEX Water Vapor Project, but these
are of limited use for climate models, because the real climate issue is the vertical redistribution
of water by dynamical processes. For the lower troposphere, radiosondes provide a useful if
incomplete global picture, but large differences exist in radiosondes used in different parts of the
world (Elliott and Gaffen, 1991). Upper troposphere water vapor has been observed by SAGE
II (Rind et al., 1991), but its sparse coverage allows only a multi-year average to be obtained
with any useful accuracy; there is almost no tropical water vapor information, and what exists
is biased toward non-convecting regions. The GOES 6.7 /lm channel has been used to create a
decade-long climatology of upper troposphere relative humidity (Soden and Bretherton, 1993).
However, this diagnostic and that derived from SAGE II disagree in the sense of their Northern
midlatitude seasonal cycles (as well as the magnitude of their tropical seasonal cycles), creating
a quandary for modelers wishing to use such data to diagnose feedback processes (cf. Del Genio
et al., 1994). The 183 GHz channel on SSMjT-2 is beginning to provide upper and middle
troposphere water vapor profiling on a variety of time scales over oceans.
-Temperature.
Climatological temperature distributions and interannual variations are available from radiosonde data (Oort, 1983), although some modelers validate against analysis products instead.
TOYS gives some temperature information on shorter time scales. GCMs typically exhibit
47
-Cloud particle size.
Satellite near-infrared retrievals of droplet effective radius for liquid clouds outside the polar
regions have recently become available (Han et al., 1994). These refer primarily to cloud top
conditions; the vertical profile of droplet size through clouds is unknown except for isolated
in situ measurements. Particle size for ice clouds is much more uncertain; comparable global
surveys are only beginning to be produced, and remote retrievals from field programs typically
differ markedly from in situ ice particle size estimates. Ice scattering properties are not welldefined by assuming "equivalent spheres" for Mie scattering calculations, the standard approach
in GCMs. More realistic ice phase functions have been defined (Macke, 1993); it remains to
be seen whether a universal phase function for all ice particle shapes is adequate for GCM
purposes. Only GCMs with prognostic cloud water budgets can currently make use of such
information.
-Cloud liquid and ice water path.
Microwave retrievals of vertically integrated liquid water path are available over ocean, but not
over land. But the algorithms used by different workers produce dramatically different results,
even disagreeing about whether liquid water path is higher in midlatitudes or the tropics (Lin
and Rossow, 1994). Such data are thus virtually useless for GCM validation, a point missed by
many climate modelers who have nonetheless compared their prognostic models to one of the
published estimates. The problems involve whether clear points are included in the average,
the threshold assumed for detecting cloud water, and errors in the assumed water vapor profile
and cloud temperature. Ice water path has not yet been retrieved from satellite; this is perhaps
the major gap in validation data for GCMs with prognostic cloud water schemes. ISCCP can
indirectly estimate ice water path by combining microwave liquid water path and near-infrared
particle size estimates with its own column optical thickness retrieval (Lin and Rossow, 1994),
but this depends on their assumptions about ice particle size and phase function. A major
issue for GCMs is where the transition from liquid to ice occurs as a function of temperature,
cloud depth and age, and other parameters (cf. Mitchell et al., 1989; Li and LeTreut, 1992; Del
Genio et al., 1996).
-Water vapor.
Rossow (1996, this volume, chapter 8) provides a comprehensive discussion of existing water
vapor observation capabilities. Global precipitable water estimates from a combination of microwave and radiosonde data will be produced by the GEWEX Water Vapor Project, but these
are of limited use for climate models, because the real climate issue is the vertical redistribution
of water by dynamical processes. For the lower troposphere, radiosondes provide a useful if
incomplete global picture, but large differences exist in radiosondes used in different parts of the
world (Elliott and Gaffen, 1991). Upper troposphere water vapor has been observed by SAGE
II (Rind et al., 1991), but its sparse coverage allows only a multi-year average to be obtained
with any useful accuracy; there is almost no tropical water vapor information, and what exists
is biased toward non-convecting regions. The GOES 6.7 /lm channel has been used to create a
decade-long climatology of upper troposphere relative humidity (Soden and Bretherton, 1993).
However, this diagnostic and that derived from SAGE II disagree in the sense of their Northern
midlatitude seasonal cycles (as well as the magnitude of their tropical seasonal cycles), creating
a quandary for modelers wishing to use such data to diagnose feedback processes (cf. Del Genio
et al., 1994). The 183 GHz channel on SSMjT-2 is beginning to provide upper and middle
troposphere water vapor profiling on a variety of time scales over oceans.
-Temperature.
Climatological temperature distributions and interannual variations are available from radiosonde data (Oort, 1983), although some modelers validate against analysis products instead.
TOYS gives some temperature information on shorter time scales. GCMs typically exhibit