showed that the accuracy of NSCAT data exceeds
the requirement. The comparison in Fig. 3.4.3
shows that other data also meet the requirement.
Figure 3.4.3 illustrates that the three underlying
wind estimates show small biases of the order of
0–1.25 m s
91 and rms values :2 m s
91 compared
with three sets of buoys – the TOGA-TAO (Tropical Ocean and Global Atmosphere–Tropical
Atmosphere and Ocean) array; the NDBC operational array, mid-latitude US coast; and ODAS, the
European network (Bentamy et al., 2000). Higher
accuracy and spatial resolution, however, are desirable in a number of applications, including the
study of hurricanes (Liu et al., 2000) and eastern
boundary wind stress curl (Milliff et al., 2001).
From these multiple data sets new climatological
wind fields can be derived. A merged wind field
employing data from two SSM/Is, the ERS-2 AMI,
and NSCAT have been calculated on a daily basis
at 1°1° resolution (Bentamy et al., 1999).
3.4.4 Thermal forcing
3.4.4.1 Latent heat flux
The computation of LH by the bulk aerodynamic
method requires u, T s and q as seen in equation
(3.4.3). Over the ocean, u and T s have been
directly retrieved from satellite data, but not q.
A method of estimating q and LH from the ocean
using satellite data was proposed by Liu and Niiler
(1984). It is based on an empirical relation
between the integrated water vapour (W, measured
by space-borne microwave radiometers) and q on a
monthly time scale (Liu, 1986). The physical rationale is that the vertical distribution of water vapour
through the whole depth of the atmosphere is
coherent for periods longer than a week (Liu et al.,
1991). The relation does not work well at synoptic
and shorter time scales and also fails in some
regions during summer (Liu et al., 1992). The relation has also been scrutinized in a number of studies
(e.g. Hsu and Blanchard, 1989; Eymard et al.,
1989; Esbensen et al., 1993; Jourdan and Gautier,
1994). Modification of this method by including
additional geophysical parameters or Empirical
Orthogonal Functions (EOFs) as estimators have
been proposed (e.g. Wagner et al., 1990; Cresswell
et al., 1991; Miller and Katsaros, 1991; Chou et al.,
1995), with some overall improvement, but the
inherent limitation is the lack of information about
the vertical distribution of q near the surface.
Liu (1990) suggested two possible improvements in LH retrieval. One suggestion was to
obtain information on the vertical structures of
humidity distribution and the other was to derive a
direct relation between LH and the brightness temperatures (BT) measured by the radiometer. Liu
et al. (1991) demonstrated that the boundary-layer
water vapour is a better predictor of q. Schulz
et al. (1993) developed an algorithm for direct
retrieval of boundary-layer water vapour from
radiances observed by SSM/I. However, without
using any new measurements that can distinguish
vertical structures, and without the knowledge of
the boundary-layer (or mixed-layer) height, any
significant improvement of the retrieval of LH by
this method remains to be demonstrated. Information from the new microwave humidity sounders
on the operational polar orbiters of DMSP and the
National Oceanic and Atmospheric Administration
(NOAA), as anticipated by Liu (1990), has not
been optimally utilized at present.
Because all the three geophysical parameters, u,
T s and W, can be retrieved from the radiances at the
frequencies measured by the Scanning Multichannel
Microwave Radiometer (SMMR) on Nimbus-7
(similar to SSM/I, but with 10.6 and 6.6 GHz channels as well and no 85 GHz channels), the feasibility of retrieving LH directly from the measured
radiances was also demonstrated by Liu (1990).
Such a method may improve accuracy in bypassing
the uncertainties related to the bulk parameterization method. Liu (1990) used coincident BT values
observed by the SMMR and LH computed from
ship data on monthly time scales. While SMMR
measures at ten channels, only six channels were
identified as significantly useful in estimating LH.
SSM/I, the operational microwave radiometer that
followed SMMR, lacks the low-frequency channels
that are sensitive to T s , making direct retrieval of
LH from BT alone unfeasible. The microwave
imager (TMI) on the Tropical Rainfall Measuring
Mission (TRMM), which was launched in 1998,
includes low-frequency measurements sensitive to
T s . Direct retrieval of LH has received renewed
interest. Bourras and Eymard (1988) attempted to
combine SSM/I radiances and T s measurements to
compute LH, and Liu et al. (1999) tried to do the
same using TMI radiances alone, but more vigorous effort in validation is required to demonstrate
any significant improvement of these more direct
methods over the indirect methods.
3.4 Air–Sea Fluxes from Satellite Data
177
Liu and Katsaros
the requirement. The comparison in Fig. 3.4.3
shows that other data also meet the requirement.
Figure 3.4.3 illustrates that the three underlying
wind estimates show small biases of the order of
0–1.25 m s
91 and rms values :2 m s
91 compared
with three sets of buoys – the TOGA-TAO (Tropical Ocean and Global Atmosphere–Tropical
Atmosphere and Ocean) array; the NDBC operational array, mid-latitude US coast; and ODAS, the
European network (Bentamy et al., 2000). Higher
accuracy and spatial resolution, however, are desirable in a number of applications, including the
study of hurricanes (Liu et al., 2000) and eastern
boundary wind stress curl (Milliff et al., 2001).
From these multiple data sets new climatological
wind fields can be derived. A merged wind field
employing data from two SSM/Is, the ERS-2 AMI,
and NSCAT have been calculated on a daily basis
at 1°1° resolution (Bentamy et al., 1999).
3.4.4 Thermal forcing
3.4.4.1 Latent heat flux
The computation of LH by the bulk aerodynamic
method requires u, T s and q as seen in equation
(3.4.3). Over the ocean, u and T s have been
directly retrieved from satellite data, but not q.
A method of estimating q and LH from the ocean
using satellite data was proposed by Liu and Niiler
(1984). It is based on an empirical relation
between the integrated water vapour (W, measured
by space-borne microwave radiometers) and q on a
monthly time scale (Liu, 1986). The physical rationale is that the vertical distribution of water vapour
through the whole depth of the atmosphere is
coherent for periods longer than a week (Liu et al.,
1991). The relation does not work well at synoptic
and shorter time scales and also fails in some
regions during summer (Liu et al., 1992). The relation has also been scrutinized in a number of studies
(e.g. Hsu and Blanchard, 1989; Eymard et al.,
1989; Esbensen et al., 1993; Jourdan and Gautier,
1994). Modification of this method by including
additional geophysical parameters or Empirical
Orthogonal Functions (EOFs) as estimators have
been proposed (e.g. Wagner et al., 1990; Cresswell
et al., 1991; Miller and Katsaros, 1991; Chou et al.,
1995), with some overall improvement, but the
inherent limitation is the lack of information about
the vertical distribution of q near the surface.
Liu (1990) suggested two possible improvements in LH retrieval. One suggestion was to
obtain information on the vertical structures of
humidity distribution and the other was to derive a
direct relation between LH and the brightness temperatures (BT) measured by the radiometer. Liu
et al. (1991) demonstrated that the boundary-layer
water vapour is a better predictor of q. Schulz
et al. (1993) developed an algorithm for direct
retrieval of boundary-layer water vapour from
radiances observed by SSM/I. However, without
using any new measurements that can distinguish
vertical structures, and without the knowledge of
the boundary-layer (or mixed-layer) height, any
significant improvement of the retrieval of LH by
this method remains to be demonstrated. Information from the new microwave humidity sounders
on the operational polar orbiters of DMSP and the
National Oceanic and Atmospheric Administration
(NOAA), as anticipated by Liu (1990), has not
been optimally utilized at present.
Because all the three geophysical parameters, u,
T s and W, can be retrieved from the radiances at the
frequencies measured by the Scanning Multichannel
Microwave Radiometer (SMMR) on Nimbus-7
(similar to SSM/I, but with 10.6 and 6.6 GHz channels as well and no 85 GHz channels), the feasibility of retrieving LH directly from the measured
radiances was also demonstrated by Liu (1990).
Such a method may improve accuracy in bypassing
the uncertainties related to the bulk parameterization method. Liu (1990) used coincident BT values
observed by the SMMR and LH computed from
ship data on monthly time scales. While SMMR
measures at ten channels, only six channels were
identified as significantly useful in estimating LH.
SSM/I, the operational microwave radiometer that
followed SMMR, lacks the low-frequency channels
that are sensitive to T s , making direct retrieval of
LH from BT alone unfeasible. The microwave
imager (TMI) on the Tropical Rainfall Measuring
Mission (TRMM), which was launched in 1998,
includes low-frequency measurements sensitive to
T s . Direct retrieval of LH has received renewed
interest. Bourras and Eymard (1988) attempted to
combine SSM/I radiances and T s measurements to
compute LH, and Liu et al. (1999) tried to do the
same using TMI radiances alone, but more vigorous effort in validation is required to demonstrate
any significant improvement of these more direct
methods over the indirect methods.
3.4 Air–Sea Fluxes from Satellite Data
177
Liu and Katsaros