399
carried out by Goldenberg and O'Brien (1981) the constant PuCD was not included and the data
provided by the FSU data base are given as pseudo-stresses, i.e. vectors of magnitude UIUI.
The data in our forced experiment will be the FSU pseudo-stresses. This data base is constituted
by monthly means over the 33-year period 1961-1993 on a 2° x2° grid covering the region between
124°E-700W and 29°N-29°S. The analysis of wind data over all this 33-year period are based
on WM data set for the years 1961-65, while the analysis for the years 1966-80 is based on the
COADS data set. From 1981 to present, the analysis are based on the National Climatic Data
Center (NCDC) surface marine observation data (TD-1l29).·
The value for the drag coefficient is calculated by making the FSU climatology compatible with
that of Hellerman and Rosentein (1983) (HR hereafter) climatology used conventionally. To
do so the following condition is imposed: the magnitude of the zonal wind annual mean over
all the ocean points between lOON and 10
0
S must be the same for both FSU and HR wind
fields. Taking Pu = 1.2 Kg m- 3 , the value of 1.535 x 10- 3 is obtained for the drag coefficient
which is similar to that used by WM ( 1.5 x 10- 3 ). With this parametrization the FSU wind
stresses are in good agreement with those of HR concerning the stress magnitude (Wyrtki and
Meyers, 1975a) but indicate a shift in the northeast trade wind direction toward a more zonal
orientation, probably due to the different time periods covered by the data and the well known
variations over long periods of the northeast trade wind direction (Wentworth, 1949).
In addition to the wind climatologies already cited (WM, GO and HR) other available wind
stress data sets for different periods are Wright (1988), Harrison (1989) and ECMWF analyses
(Trenberth et 0.1., 1989). Wright (1988) is constituted as a global data set with a 2° resolution,
and is based on COADS data over the period 1950-1979 .. The Ha.rrison (1989) global data
set with 1° resolution is based on the NCDC HSST data over the period 1859-1979, although
only 1950-1979 were available for the Pacific. Also ECMWF analyses (Trenberth et al., 1989)
for the period 1980-86 constitute another possible source of wind data for driving ocean general
circulation models. Among all these possible choices, the FSU data set was used since it provides
a long period of observed data as is needed for low frequency variability studies.
It has been pointed out by several authors (Large a.nd Pond, 1981; Harrison, 1989) that the
value of the drag coefficient used to construct HR climatology is too large. Although there
exists incertitude about the value of this magnitude, the majority of ocean models use a drag
reduction coefficient /1- equal to 0.75 (Stor.kda.le et aL, 1993; Da.ndin, 1993). Balma.seda (1993)
used a /1- value adjusted in a simulation of the seasonal cycle using the FSU climatology, being
initially equals to 0.75 and 0.5 thereafter. The necessity of such a large reduction factor in this
study may be due to the use of an ocean model of intermediate complexity. In this study, /1- = 1
will be used and therefore no reduction in the FSU wind stress will be considered.
3.1.2 Detrending.
Inhomogeneous sampling and changes in the measurements techniques cause artificial trends in
the wind stress data set. Therefore some statistical preprocessing is needed.
Merchant ship observations appear to indicate an increase in the strength of the winds in the
tropical Pacific and elsewhere in recent decades (Cardone et 0.1., 1990; Ramage, 1984; Whysa.ll
et a.l., 1987; Wyrtki a.nd Meyers, 1976 for the Pacific). The same general trend is apparent in
wind analysis carried out at Florida State University (FSU winds) by Goldenberg and O'Brien
(1981) .
Ramage (1984) noted that the observed variables from which heat flux is calculated, wind speed
and SST have undergone indeterminate spurious changes due to modifications in estimating
and measuring methods. Whysall et aL (1987) suggested that the trends in the surface winds
carried out by Goldenberg and O'Brien (1981) the constant PuCD was not included and the data
provided by the FSU data base are given as pseudo-stresses, i.e. vectors of magnitude UIUI.
The data in our forced experiment will be the FSU pseudo-stresses. This data base is constituted
by monthly means over the 33-year period 1961-1993 on a 2° x2° grid covering the region between
124°E-700W and 29°N-29°S. The analysis of wind data over all this 33-year period are based
on WM data set for the years 1961-65, while the analysis for the years 1966-80 is based on the
COADS data set. From 1981 to present, the analysis are based on the National Climatic Data
Center (NCDC) surface marine observation data (TD-1l29).·
The value for the drag coefficient is calculated by making the FSU climatology compatible with
that of Hellerman and Rosentein (1983) (HR hereafter) climatology used conventionally. To
do so the following condition is imposed: the magnitude of the zonal wind annual mean over
all the ocean points between lOON and 10
0
S must be the same for both FSU and HR wind
fields. Taking Pu = 1.2 Kg m- 3 , the value of 1.535 x 10- 3 is obtained for the drag coefficient
which is similar to that used by WM ( 1.5 x 10- 3 ). With this parametrization the FSU wind
stresses are in good agreement with those of HR concerning the stress magnitude (Wyrtki and
Meyers, 1975a) but indicate a shift in the northeast trade wind direction toward a more zonal
orientation, probably due to the different time periods covered by the data and the well known
variations over long periods of the northeast trade wind direction (Wentworth, 1949).
In addition to the wind climatologies already cited (WM, GO and HR) other available wind
stress data sets for different periods are Wright (1988), Harrison (1989) and ECMWF analyses
(Trenberth et 0.1., 1989). Wright (1988) is constituted as a global data set with a 2° resolution,
and is based on COADS data over the period 1950-1979 .. The Ha.rrison (1989) global data
set with 1° resolution is based on the NCDC HSST data over the period 1859-1979, although
only 1950-1979 were available for the Pacific. Also ECMWF analyses (Trenberth et al., 1989)
for the period 1980-86 constitute another possible source of wind data for driving ocean general
circulation models. Among all these possible choices, the FSU data set was used since it provides
a long period of observed data as is needed for low frequency variability studies.
It has been pointed out by several authors (Large a.nd Pond, 1981; Harrison, 1989) that the
value of the drag coefficient used to construct HR climatology is too large. Although there
exists incertitude about the value of this magnitude, the majority of ocean models use a drag
reduction coefficient /1- equal to 0.75 (Stor.kda.le et aL, 1993; Da.ndin, 1993). Balma.seda (1993)
used a /1- value adjusted in a simulation of the seasonal cycle using the FSU climatology, being
initially equals to 0.75 and 0.5 thereafter. The necessity of such a large reduction factor in this
study may be due to the use of an ocean model of intermediate complexity. In this study, /1- = 1
will be used and therefore no reduction in the FSU wind stress will be considered.
3.1.2 Detrending.
Inhomogeneous sampling and changes in the measurements techniques cause artificial trends in
the wind stress data set. Therefore some statistical preprocessing is needed.
Merchant ship observations appear to indicate an increase in the strength of the winds in the
tropical Pacific and elsewhere in recent decades (Cardone et 0.1., 1990; Ramage, 1984; Whysa.ll
et a.l., 1987; Wyrtki a.nd Meyers, 1976 for the Pacific). The same general trend is apparent in
wind analysis carried out at Florida State University (FSU winds) by Goldenberg and O'Brien
(1981) .
Ramage (1984) noted that the observed variables from which heat flux is calculated, wind speed
and SST have undergone indeterminate spurious changes due to modifications in estimating
and measuring methods. Whysall et aL (1987) suggested that the trends in the surface winds