400
might be indicative of a real climate change. Ramage (1987) cautioned that such variations
might be an artifact of the measurement technique. Cardone et al. (1990) undertook a direct
investigation and in their compilation of surface winds from ship reports they observed a trend
toward strengthening winds over the past three decades. Posmentier et al. (1989) took an
indirect approach, investigating the trends in tropical Pacific sea surface temperature and sea
level which are closely related to the winds. They found an inconsistency between the analyzed
data and the apparent wind change. Again this supported the hypothesis that the apparent
wind changes are an artifact introduced by changes in measurement technique which may be
related to the use of higher ships. The trend is most apparent in the 1960s data where the wind
stress appears to be too weak.
Vautard et 01. (1992) proposed that the SSA (Singular Spectral Analysis) technique could be
used to detect and eliminate trends in time series. This technique has been used by Allen (1992)
to eliminate the trend in FSU winds. To allow the wind data to be easily compared with recent
observations, it is supposed that the trend in actual and future data is zero. For this hypothesis
to be valid it is necessary that future measurements are taken with the same procedure used
now. Balmaseda (1993) performed a comparison among the different detrending approaches and
concluded that the SSA technique appeared as the most robust method to overcome the trend.
3.1.3 From FSU data to the OPA ocean model grid.
The domain covered by the FSU winds is between 124°E-700W and 29°N-29°S. However, the
OGCM is defined over 1200E-69°W and 49°N-41°S.
Three different extrapolation approaches have been tested. Linear and sinusoidal extrapolation
(i.e. with a sine or cosine functions) for which smoothing to zero at the border of the extended
domain is considered and finally the "curl" extrapolation. From these three extrapolations the
"curl" one was retained. The reason for this choice comes from the observation of general
structure of the zonal component of the wind stress and the curl of the wind stress over tropical
and extratropical regions. Such a structure for the zonal component Tx of the wind stress can
be schematically represented by the expression:
(10)
where A is constant to be fixed. For y = 30° and 60° Tx = 0, having with this function an
easterly maximum at 15° and a westerly maximum at 45°. This is an empirical fit to the general
pattern observed for Tx. Only the constant A remains to be adjusted. To obtain the value for
this constant we take the meridional derivative
8Tx
1r
(
)
-=-Acos (y-30)1rj30
8y
30
(11)
from which we find that:
308TX
A = --I,}=300
7r 8y .
(12)
The zonal wind component in latitudes beyond y = 30° have been computed considering this
"curl" extrapolation, following expression (1) where A is given by (3). For the meridional wind
component a linear extrapolation smoothing to zero at the border was used. This method
helps provide a smooth wind curl over the domain hence leading to realistic ocean gyres in the
subtropics.
The wind data were interpolated from the FSU extended mesh (regular 2° x 2°) to the irregular
OPA grid. Classical bilinear and bicubic interpolation have been tested. A continuous first
derivative is obtained in the bicubic case and hence this interpolation was adopted.
might be indicative of a real climate change. Ramage (1987) cautioned that such variations
might be an artifact of the measurement technique. Cardone et al. (1990) undertook a direct
investigation and in their compilation of surface winds from ship reports they observed a trend
toward strengthening winds over the past three decades. Posmentier et al. (1989) took an
indirect approach, investigating the trends in tropical Pacific sea surface temperature and sea
level which are closely related to the winds. They found an inconsistency between the analyzed
data and the apparent wind change. Again this supported the hypothesis that the apparent
wind changes are an artifact introduced by changes in measurement technique which may be
related to the use of higher ships. The trend is most apparent in the 1960s data where the wind
stress appears to be too weak.
Vautard et 01. (1992) proposed that the SSA (Singular Spectral Analysis) technique could be
used to detect and eliminate trends in time series. This technique has been used by Allen (1992)
to eliminate the trend in FSU winds. To allow the wind data to be easily compared with recent
observations, it is supposed that the trend in actual and future data is zero. For this hypothesis
to be valid it is necessary that future measurements are taken with the same procedure used
now. Balmaseda (1993) performed a comparison among the different detrending approaches and
concluded that the SSA technique appeared as the most robust method to overcome the trend.
3.1.3 From FSU data to the OPA ocean model grid.
The domain covered by the FSU winds is between 124°E-700W and 29°N-29°S. However, the
OGCM is defined over 1200E-69°W and 49°N-41°S.
Three different extrapolation approaches have been tested. Linear and sinusoidal extrapolation
(i.e. with a sine or cosine functions) for which smoothing to zero at the border of the extended
domain is considered and finally the "curl" extrapolation. From these three extrapolations the
"curl" one was retained. The reason for this choice comes from the observation of general
structure of the zonal component of the wind stress and the curl of the wind stress over tropical
and extratropical regions. Such a structure for the zonal component Tx of the wind stress can
be schematically represented by the expression:
(10)
where A is constant to be fixed. For y = 30° and 60° Tx = 0, having with this function an
easterly maximum at 15° and a westerly maximum at 45°. This is an empirical fit to the general
pattern observed for Tx. Only the constant A remains to be adjusted. To obtain the value for
this constant we take the meridional derivative
8Tx
1r
(
)
-=-Acos (y-30)1rj30
8y
30
(11)
from which we find that:
308TX
A = --I,}=300
7r 8y .
(12)
The zonal wind component in latitudes beyond y = 30° have been computed considering this
"curl" extrapolation, following expression (1) where A is given by (3). For the meridional wind
component a linear extrapolation smoothing to zero at the border was used. This method
helps provide a smooth wind curl over the domain hence leading to realistic ocean gyres in the
subtropics.
The wind data were interpolated from the FSU extended mesh (regular 2° x 2°) to the irregular
OPA grid. Classical bilinear and bicubic interpolation have been tested. A continuous first
derivative is obtained in the bicubic case and hence this interpolation was adopted.