Validation of the LH is very difficult because
there are very few direct measurements. Validation
data are largely computed through the bulk parameterization method, with its inherent uncertainties. The mean parameters from merchant ships
have poor quality and distribution. The accuracy
of LH estimated from instantaneous spacebased
data is about 30–40 W m
92
, and rms errors are half
of that for monthly mean over 2° areas (e.g. Schulz
et al., 1997).
3.4.4.2 Short-wave radiation
Global surface net short-wave radiation (SR) has
been computed from the Earth Radiation Budget
Experiment (ERBE) data from 1985 to 1989 (Li
and Leighton, 1993). The variability of SR is
largely controlled by the variability of clouds, and
most of the computations of SR over the ocean
make use of the high resolution and high sampling
of the diurnal cycle by data from geostationary
satellites. The International Satellite Cloud Climatology Project (ISCCP) (Rossow and Schiffer, 1991)
has provided calibrated and standardized cloud
data from four geostationary satellites operated by
the US, Japanese and European space agencies.
A number of methods (e.g. Gautier et al., 1980;
Bishop and Rossow, 1991; Pinker and Laszlo,
1992) have been used to compute net SR at daily
time scales and 2°2.5° resolution, including the
operational effort by the Surface Radiation Budget
Program at Langley Research Center. The availability of ISCCP data is expected well into the
future, but the data record has not yet been
extended to recent years. Higher spatial resolution
is also desirable and is being pursued by the
above-mentioned project.
The SR estimates for the ocean are simplified
compared to those over land by the known
reflectance of the sea surface. Variations in
reflectance due to surface roughness and white
caps could be included in the formulation as a
function of u (e.g. Katsaros, 1990). The multifrequency data of the new generation geostationary
satellites will allow discrimination of atmospheric
aerosol scattering and surface reflection, so that it
will be possible to take these second-order effects
into account. The role of Saharan and other mineral dust and atmospheric pollution (as from wild
fires and volcanoes) is being addressed by several
research groups and will improve the accuracy of
net surface radiative flux estimates.
High correlation between the SR and T s tendency (time differential of T s ) is demonstrated in
Fig. 3.4.4 (see Plate 3.4.4, p. 172). This correlation
is consistent with the notion that solar heating is
the main driver for the seasonal changes of T s
away from the equatorial wave guide, where ocean
dynamics may be more important. The relative
roles of solar heating and evaporative cooling in
changing the annual cycle of T s and T s changes
during El Niño and the Southern Oscillation
(ENSO) have been studied (e.g. Liu and Gautier,
1990; Liu et al., 1994).
3.4.4.3 Sensible and net long-wave heat
fluxes
It is much more difficult to estimate SH and surface net long-wave radiation (LR) from satellite
data than SR or LE. The magnitude and variability
of SH are relatively small over much of the open
ocean. The weighting functions (contributions of
radiance as a function of height) of space-borne
atmospheric sounders are too broad to help retrieval
of near-surface air temperature with sufficient accuracy for the bulk aerodynamic formula (equation
3.4.2). A relationship similar to the W versus q
derived from microwave radiometers has not been
found. Unlike humidity, the vertical variation of
temperature is not coherent through the entire
atmosphere. As suggested by Liu and Niiler (1990)
and Liu (1990), SH can be estimated from LE if
the Bowen ratio or the relative humidity is known.
This is equivalent to deriving T from q. The accuracy of a prescribed formula for Bowen ratio and
relative humidity is, however, quite uncertain (Liu
and Niiler, 1990) and would vary regionally. There
have been a number of attempts (e.g. Thadathil
et al., 1993; Kubota and Shikauchi, 1995; Konda
et al., 1996) to derive T (SH) from q retrieved with
the method of Liu (1986). Recently, Jones et al.
(1999b) attempted to produce T and q using a
neural network approach. Currently there is nothing better to offer, but future numerical models
well constrained with satellite data should be able
to resolve this difficulty.
Long-wave radiation is strongly affected by
atmospheric properties below cloud base, which
are hidden from space-borne sensors. Methods
that combine cloud information with atmospheric soundings and numerical models have
been attempted (e.g. Frouin et al., 1988; Gupta
et al., 1992). Classical LR climatologies
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
178
there are very few direct measurements. Validation
data are largely computed through the bulk parameterization method, with its inherent uncertainties. The mean parameters from merchant ships
have poor quality and distribution. The accuracy
of LH estimated from instantaneous spacebased
data is about 30–40 W m
92
, and rms errors are half
of that for monthly mean over 2° areas (e.g. Schulz
et al., 1997).
3.4.4.2 Short-wave radiation
Global surface net short-wave radiation (SR) has
been computed from the Earth Radiation Budget
Experiment (ERBE) data from 1985 to 1989 (Li
and Leighton, 1993). The variability of SR is
largely controlled by the variability of clouds, and
most of the computations of SR over the ocean
make use of the high resolution and high sampling
of the diurnal cycle by data from geostationary
satellites. The International Satellite Cloud Climatology Project (ISCCP) (Rossow and Schiffer, 1991)
has provided calibrated and standardized cloud
data from four geostationary satellites operated by
the US, Japanese and European space agencies.
A number of methods (e.g. Gautier et al., 1980;
Bishop and Rossow, 1991; Pinker and Laszlo,
1992) have been used to compute net SR at daily
time scales and 2°2.5° resolution, including the
operational effort by the Surface Radiation Budget
Program at Langley Research Center. The availability of ISCCP data is expected well into the
future, but the data record has not yet been
extended to recent years. Higher spatial resolution
is also desirable and is being pursued by the
above-mentioned project.
The SR estimates for the ocean are simplified
compared to those over land by the known
reflectance of the sea surface. Variations in
reflectance due to surface roughness and white
caps could be included in the formulation as a
function of u (e.g. Katsaros, 1990). The multifrequency data of the new generation geostationary
satellites will allow discrimination of atmospheric
aerosol scattering and surface reflection, so that it
will be possible to take these second-order effects
into account. The role of Saharan and other mineral dust and atmospheric pollution (as from wild
fires and volcanoes) is being addressed by several
research groups and will improve the accuracy of
net surface radiative flux estimates.
High correlation between the SR and T s tendency (time differential of T s ) is demonstrated in
Fig. 3.4.4 (see Plate 3.4.4, p. 172). This correlation
is consistent with the notion that solar heating is
the main driver for the seasonal changes of T s
away from the equatorial wave guide, where ocean
dynamics may be more important. The relative
roles of solar heating and evaporative cooling in
changing the annual cycle of T s and T s changes
during El Niño and the Southern Oscillation
(ENSO) have been studied (e.g. Liu and Gautier,
1990; Liu et al., 1994).
3.4.4.3 Sensible and net long-wave heat
fluxes
It is much more difficult to estimate SH and surface net long-wave radiation (LR) from satellite
data than SR or LE. The magnitude and variability
of SH are relatively small over much of the open
ocean. The weighting functions (contributions of
radiance as a function of height) of space-borne
atmospheric sounders are too broad to help retrieval
of near-surface air temperature with sufficient accuracy for the bulk aerodynamic formula (equation
3.4.2). A relationship similar to the W versus q
derived from microwave radiometers has not been
found. Unlike humidity, the vertical variation of
temperature is not coherent through the entire
atmosphere. As suggested by Liu and Niiler (1990)
and Liu (1990), SH can be estimated from LE if
the Bowen ratio or the relative humidity is known.
This is equivalent to deriving T from q. The accuracy of a prescribed formula for Bowen ratio and
relative humidity is, however, quite uncertain (Liu
and Niiler, 1990) and would vary regionally. There
have been a number of attempts (e.g. Thadathil
et al., 1993; Kubota and Shikauchi, 1995; Konda
et al., 1996) to derive T (SH) from q retrieved with
the method of Liu (1986). Recently, Jones et al.
(1999b) attempted to produce T and q using a
neural network approach. Currently there is nothing better to offer, but future numerical models
well constrained with satellite data should be able
to resolve this difficulty.
Long-wave radiation is strongly affected by
atmospheric properties below cloud base, which
are hidden from space-borne sensors. Methods
that combine cloud information with atmospheric soundings and numerical models have
been attempted (e.g. Frouin et al., 1988; Gupta
et al., 1992). Classical LR climatologies
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
178