46
A.D. Del Genio
a bias rather than a real imbalance. Seasonal and interannual variations in fluxes can be
determined, but the diurnal cycle is probably useless. Cloud forcing derived from ERBE has
larger errors, associated with cloud detection and temperature/humidity differences between
clear and cloudy skies. Cloud forcing estimates are unreliable over snow and ice.
-Surface radiation budget.
Direct information is not available globally; TOA information is combined with radiation models
to derive surface radiation estimates with 10-20 Wm- 2 accuracy (d. Zhang et al., 1995). Surface
longwave fluxes may not be sufficiently accurate at high latitudes, where humidity is low enough
for cloud base errors to have an impact. Surface incident shortwave fluxes had been thought
to be fairly reliable, but recent controversy over 25 Wm- 2 of apparent unexplained shortwave
absorption by the atmosphere (Cess et al., 1995) has temporarily called this into question.
Surface albedo errors in GCMs may well be significant, but few GCMs have attempted to
validate their surface radiation budget.
-Turbulent surface fluxes.
Evaporation algorithms for the oceans only, based on microwave precipitable water and SST
estimates (Liu, 1988), provide crude estimates on monthly and longer time scales and allow
interannual variability to be documented, but evapotranspiration over land is not observed
globally. Sensible heat flux estimates are lacking over most of the globe but are important only
at higher latitudes. Precipitation estimates on monthly and longer time scales are acceptable
over some land areas but existing satellite algorithms for ocean are validated only regionally
and of unknown quality globally. The TRMM satellite mission will address this problem for
the tropics and subtropics beginning in 1997 (Del Genio, 1996b, this volume, chapter 20).
Although precipitation is much larger in the tropics and of first-order importance for the general
circulation there, the more demanding requirement driven by climate modeling is the accuracy
of 0.5-1.0 mm d- 1 required to ensure accurate driving of the thermohaline circulation in the
North Atlantic (Zaucker et al., 1994).
-Cloud cover, top height, optical thickness.
ISCCP (Rossow and Schiffer, 1991) has provided a useful climatology of these cloud properties
spanning almost a decade, with information on the mean and all three important time scales of
variability. Like ERBE, it is extremely uncertain over snow and ice. Inter-satellite calibration
problems cause some spurious interannual signal. ISCCP misses or misidentifies very thin
cirrus, and there are errors due to the use of Mie phase functions and similar particle size
assumptions for all clouds (Minnis et al., 1993; Liao et al., 1995), but a re-analysis of the data
set is in the process of rectifying this situation. The 5 km pixel size causes errors in cloud
cover of no more than 10%, and much less in many places (Wielicki and Parker, 1992); subpixel
cloudiness effects on optical thickness estimates have not been quantified. Climate modelers
have underutilized these data, particularly the optical thickness information. Satellite data
cannot uniquely determine low and middle cloud amounts due to shielding by higher clouds. The
former can be obtained from the atlas of surface observations (Warren et al., 1986, 1988), but
coverage varies over the globe. The surface data set also performs separation into frequencies
and amounts-when-present, and attempts to distinguish cloud types morphologically. SAGE II
(Liao et al., 1995) provides a multi-year record of cirrus frequency and/or cover.
-Cloud base height.
A global data set does not exist, although crude global patterns have recently been derived from
radiosonde profiles (Wang and Rossow, 1995). This parameter is almost never examined by
climate modelers, but accuracy sufficient to produce surface longwave flux errors of 10 Wm- 2
or less is desireable for surface energy budget studies. For some climate process problems (see
discussion below), accuracy higher than that attainable globally may be required.
A.D. Del Genio
a bias rather than a real imbalance. Seasonal and interannual variations in fluxes can be
determined, but the diurnal cycle is probably useless. Cloud forcing derived from ERBE has
larger errors, associated with cloud detection and temperature/humidity differences between
clear and cloudy skies. Cloud forcing estimates are unreliable over snow and ice.
-Surface radiation budget.
Direct information is not available globally; TOA information is combined with radiation models
to derive surface radiation estimates with 10-20 Wm- 2 accuracy (d. Zhang et al., 1995). Surface
longwave fluxes may not be sufficiently accurate at high latitudes, where humidity is low enough
for cloud base errors to have an impact. Surface incident shortwave fluxes had been thought
to be fairly reliable, but recent controversy over 25 Wm- 2 of apparent unexplained shortwave
absorption by the atmosphere (Cess et al., 1995) has temporarily called this into question.
Surface albedo errors in GCMs may well be significant, but few GCMs have attempted to
validate their surface radiation budget.
-Turbulent surface fluxes.
Evaporation algorithms for the oceans only, based on microwave precipitable water and SST
estimates (Liu, 1988), provide crude estimates on monthly and longer time scales and allow
interannual variability to be documented, but evapotranspiration over land is not observed
globally. Sensible heat flux estimates are lacking over most of the globe but are important only
at higher latitudes. Precipitation estimates on monthly and longer time scales are acceptable
over some land areas but existing satellite algorithms for ocean are validated only regionally
and of unknown quality globally. The TRMM satellite mission will address this problem for
the tropics and subtropics beginning in 1997 (Del Genio, 1996b, this volume, chapter 20).
Although precipitation is much larger in the tropics and of first-order importance for the general
circulation there, the more demanding requirement driven by climate modeling is the accuracy
of 0.5-1.0 mm d- 1 required to ensure accurate driving of the thermohaline circulation in the
North Atlantic (Zaucker et al., 1994).
-Cloud cover, top height, optical thickness.
ISCCP (Rossow and Schiffer, 1991) has provided a useful climatology of these cloud properties
spanning almost a decade, with information on the mean and all three important time scales of
variability. Like ERBE, it is extremely uncertain over snow and ice. Inter-satellite calibration
problems cause some spurious interannual signal. ISCCP misses or misidentifies very thin
cirrus, and there are errors due to the use of Mie phase functions and similar particle size
assumptions for all clouds (Minnis et al., 1993; Liao et al., 1995), but a re-analysis of the data
set is in the process of rectifying this situation. The 5 km pixel size causes errors in cloud
cover of no more than 10%, and much less in many places (Wielicki and Parker, 1992); subpixel
cloudiness effects on optical thickness estimates have not been quantified. Climate modelers
have underutilized these data, particularly the optical thickness information. Satellite data
cannot uniquely determine low and middle cloud amounts due to shielding by higher clouds. The
former can be obtained from the atlas of surface observations (Warren et al., 1986, 1988), but
coverage varies over the globe. The surface data set also performs separation into frequencies
and amounts-when-present, and attempts to distinguish cloud types morphologically. SAGE II
(Liao et al., 1995) provides a multi-year record of cirrus frequency and/or cover.
-Cloud base height.
A global data set does not exist, although crude global patterns have recently been derived from
radiosonde profiles (Wang and Rossow, 1995). This parameter is almost never examined by
climate modelers, but accuracy sufficient to produce surface longwave flux errors of 10 Wm- 2
or less is desireable for surface energy budget studies. For some climate process problems (see
discussion below), accuracy higher than that attainable globally may be required.