Satellite Remote Sensing of Evaporation over Sea
453
16.4 Water Vapour Over Sea
The near-surface specific humidity, q, that is desired for the parameterisation of the evaporation
over sea cannot be remotely sensed because remote sensing techniques rely on the emission or
attenuation of radiation from optically thick layers. Therefore, a proxy parameter must be found
that can be remotely sensed and that is closely related to q. Liu and Niiler (1984) have chosen
monthly mean total columnar water-vapour contents, w, as derived from SMMR measurements
and related it to q by using a polynom of fifth degree. This method is restricted to average
atmospheric situations and fails in single cases as well in monthly mean situations which deviate
from normal cases. In particular, when the planetary boundary layer is decoupled from the
higher troposphere by temperature inversions one can expect a lack of determinism between q
and w. This is elucidated in figure 16.10 for about 9400 atmospheric situations collected in the
time period July 1987 through June 1988. A detailed analysis of the data shows that in single
situations q can be derived from W with an accuracy of 2.2 g/kg only when using a multilinear
regression on powers of W up to five. Therefore, it is indicated to look for thinner layers in
the lower troposphere that can be sensed by radiometry and that has a closer connection to
the near-surface specific air humidity. Schulz et al. (1993) have shown that the water-vapour
column of the lower 500 m, WI, can be retrieved from SSM/I measurements independently from
the total columnar water vapour. As seen in figure 16.10, q and WI are indeed closely related
to each other. However, outliers from this close relationship can be expected in coastal areas
when dry air is advected over the ocean and the marine boundary layer is of shallow depth as
indicated by the outliers in figure 16.10 that show high values of q and rather low water-vapour
columns. From SSM/I measurements at 19, 22, and 37 GHz WI can be retrieved by
W, = -5.934 + 0.03697T19v - 0.02390T19h + 0.01559T22v - 0.00497T37v
(16.29)
where the brightness temperatures have to be entered in K and WI results in g/cm 2 or cm. The
main information about WI comes from the channel at 19 GHz, vertical polarization, while T22
is used to suppress the effect of water vapour above the bottom layer. The information from
T19h and T37v is used to account for wind speed and liquid-water effects, respectively. Equation
(16.29) allows the retrieval of w, with an accuracy of 0.07 g/cm 2 as long as no high liquid-water
contents hamper the transmisison of radiation from the boundary layer to the satellite. Cases
with (T37v - T37h ) > 50 K or T19h > 165 K are excluded from the computation of W,. The
specific humidity can then be obtained via
q = -0.17 + 20.00wI
(16.30)
where w, is entered in g/cm 2 and q is retrieved in g/kg with an accuracy of 1.2 g/kg. The
success of this two-step method in retrieving q has justified to go a step further in directly
deriving q from microwave measurements of the SSM/I. This minimizes the propagation of
retrieval errors and leads to lower systematic errors (Schliissel et al., 1995):
q = -80.23 + 0.6295T19v - 0.1655T19h + 0.1495T22v - 0.1553T37v - 0.06695T37h (16.31 )
Here, information from the vertically polarized channel at 37 GHz is included in order to
better account for the wind-induced surface roughness. This channel did not add significant
information to the retrieval model (16.29) and was ignored in that case. The performance of
equation (16.31) is slighly better than the combination (16.29)-(16.30), the estimated retrieval
error is 1.1 g/kg. An intercomparison of the different methods for retrieving q with ship
measurements in the time period July 1987 to June 1988 is shown in figure 16.11, covering
4245 globally distributed match-ups. Besides the methods (16.29)-(16.30) and (16.31) the
453
16.4 Water Vapour Over Sea
The near-surface specific humidity, q, that is desired for the parameterisation of the evaporation
over sea cannot be remotely sensed because remote sensing techniques rely on the emission or
attenuation of radiation from optically thick layers. Therefore, a proxy parameter must be found
that can be remotely sensed and that is closely related to q. Liu and Niiler (1984) have chosen
monthly mean total columnar water-vapour contents, w, as derived from SMMR measurements
and related it to q by using a polynom of fifth degree. This method is restricted to average
atmospheric situations and fails in single cases as well in monthly mean situations which deviate
from normal cases. In particular, when the planetary boundary layer is decoupled from the
higher troposphere by temperature inversions one can expect a lack of determinism between q
and w. This is elucidated in figure 16.10 for about 9400 atmospheric situations collected in the
time period July 1987 through June 1988. A detailed analysis of the data shows that in single
situations q can be derived from W with an accuracy of 2.2 g/kg only when using a multilinear
regression on powers of W up to five. Therefore, it is indicated to look for thinner layers in
the lower troposphere that can be sensed by radiometry and that has a closer connection to
the near-surface specific air humidity. Schulz et al. (1993) have shown that the water-vapour
column of the lower 500 m, WI, can be retrieved from SSM/I measurements independently from
the total columnar water vapour. As seen in figure 16.10, q and WI are indeed closely related
to each other. However, outliers from this close relationship can be expected in coastal areas
when dry air is advected over the ocean and the marine boundary layer is of shallow depth as
indicated by the outliers in figure 16.10 that show high values of q and rather low water-vapour
columns. From SSM/I measurements at 19, 22, and 37 GHz WI can be retrieved by
W, = -5.934 + 0.03697T19v - 0.02390T19h + 0.01559T22v - 0.00497T37v
(16.29)
where the brightness temperatures have to be entered in K and WI results in g/cm 2 or cm. The
main information about WI comes from the channel at 19 GHz, vertical polarization, while T22
is used to suppress the effect of water vapour above the bottom layer. The information from
T19h and T37v is used to account for wind speed and liquid-water effects, respectively. Equation
(16.29) allows the retrieval of w, with an accuracy of 0.07 g/cm 2 as long as no high liquid-water
contents hamper the transmisison of radiation from the boundary layer to the satellite. Cases
with (T37v - T37h ) > 50 K or T19h > 165 K are excluded from the computation of W,. The
specific humidity can then be obtained via
q = -0.17 + 20.00wI
(16.30)
where w, is entered in g/cm 2 and q is retrieved in g/kg with an accuracy of 1.2 g/kg. The
success of this two-step method in retrieving q has justified to go a step further in directly
deriving q from microwave measurements of the SSM/I. This minimizes the propagation of
retrieval errors and leads to lower systematic errors (Schliissel et al., 1995):
q = -80.23 + 0.6295T19v - 0.1655T19h + 0.1495T22v - 0.1553T37v - 0.06695T37h (16.31 )
Here, information from the vertically polarized channel at 37 GHz is included in order to
better account for the wind-induced surface roughness. This channel did not add significant
information to the retrieval model (16.29) and was ignored in that case. The performance of
equation (16.31) is slighly better than the combination (16.29)-(16.30), the estimated retrieval
error is 1.1 g/kg. An intercomparison of the different methods for retrieving q with ship
measurements in the time period July 1987 to June 1988 is shown in figure 16.11, covering
4245 globally distributed match-ups. Besides the methods (16.29)-(16.30) and (16.31) the