398
associated with an Asselin filter: a time centered scheme is used for the advection terms, while a
forward scheme is used for the turbulent horizontal diffusion terms and a backward one is used
for the vertical diffusion terms. The horizontal domain is 40 0 S/48°N - 120 0 E/700W. At 48°N
and 40 0 S the ocean is closed by artificial coastlines. The time step is one hour and the grid on
which equations are discretized is the one used by Dandin (1993). It varies smoothly in latitude:
at the equator the resolution is 0.33 degrees and reaches 1.5 degrees at 48°N. In latitude, the
resolution is of 0.75 degrees. 28 vertical levels are chosen with 17 levels in the upper 400 meters.
A gridded bathymetric data set of 5' x 5' (ETOP5 1986) resolution has been interpolated onto
the model grid. The abyssal trenches have been cut off at 4950 meters and islands submerged
to 700 meters depth.
3 The Forcing.
3.1 Surface wind stress forcing.
The momentum flux associated with the surface wind stress is the most important forcing for
the upper ocean circulation. The aim of this section is to explain this forcing.
3.1.1 Wind stress data.
Many attempts have been made to investigate specific aspects of the time and space variability of
the wind stress over the tropical Pacific Ocean. Early studies of the winds have made use of ship
observations to study the mean monthly and annual wind distributions (McDonald, 1938; Crowe,
1951a,b; U.S. Navy Hydrographic Office, 1956; Hellerman, 1967). More recently Wyrtki and
Meyers (1975a,b; 1976) (WM hereafter) have processed 5 million ship wind observations between
30 0 N and 30 0 S over the Pacific Ocean for the years 1900-1973. The data were used to create
maps of average monthly winds and wind stresses for 2° x 10° longitude-latitude quadrangles.
Wyrtki and Meyers (1976) used the data to study both seasonal and interannual variability of
the Pacific Ocean trade wind field. Barnett (1977) has used the data to investigate the time
and space variability of the trade winds. Reiter (1978a,b) utilized the data to investigate the
interannual changes in the ocean-atmosphere system. This wind data for the lO-year period
1961-70 were subjectively analyzed and digitized by Goldenb~rg a.nd O'Brien (1981) (GO) on a
2° x 2° longitude-latitude grid. The needs for creating such a revised dataset were that errors
existed in the original data set, a finer grid was necessary to drive coupled atmospheric oceans
models (the original grid was 2° x 10°) and the existence for many months of regions entirely void
of data or 2° x 10° boxes containing so few observations that the representativeness of average
monthly values was in question.
Therefore, the revised Florida State University (FSU) wind data of Goldenberg and O'Brien
(1981), consisted in monthly-averaged wind stress values for the tropical Pacific Ocean on a
2°x2° longitude-latitude grid for the lO-year period January 1961-December 1970. This period
was chosen as having the most consistent data coverage for the years 1947-73. The stresses of
WM were computed from individual ship wind observations according to the bulk formulae:
where T is the wind stress vector, Pa is the air density, CD is the drag coefficient and (j is the
wind velocity. Constant values were assumed for the drag coefficient, CD = 1.5 X 10- 3 , and for
the air density, Pa = 1.2 Kg m- 3 . The justification for these constant values comes from the fact
that it has been demonstrated that the linear stress law (i.e., C D=cte) is adequate for spectral
analysis involving periods longer than 10 days (Willebmnd, 1978). Therefore, in the analysis
associated with an Asselin filter: a time centered scheme is used for the advection terms, while a
forward scheme is used for the turbulent horizontal diffusion terms and a backward one is used
for the vertical diffusion terms. The horizontal domain is 40 0 S/48°N - 120 0 E/700W. At 48°N
and 40 0 S the ocean is closed by artificial coastlines. The time step is one hour and the grid on
which equations are discretized is the one used by Dandin (1993). It varies smoothly in latitude:
at the equator the resolution is 0.33 degrees and reaches 1.5 degrees at 48°N. In latitude, the
resolution is of 0.75 degrees. 28 vertical levels are chosen with 17 levels in the upper 400 meters.
A gridded bathymetric data set of 5' x 5' (ETOP5 1986) resolution has been interpolated onto
the model grid. The abyssal trenches have been cut off at 4950 meters and islands submerged
to 700 meters depth.
3 The Forcing.
3.1 Surface wind stress forcing.
The momentum flux associated with the surface wind stress is the most important forcing for
the upper ocean circulation. The aim of this section is to explain this forcing.
3.1.1 Wind stress data.
Many attempts have been made to investigate specific aspects of the time and space variability of
the wind stress over the tropical Pacific Ocean. Early studies of the winds have made use of ship
observations to study the mean monthly and annual wind distributions (McDonald, 1938; Crowe,
1951a,b; U.S. Navy Hydrographic Office, 1956; Hellerman, 1967). More recently Wyrtki and
Meyers (1975a,b; 1976) (WM hereafter) have processed 5 million ship wind observations between
30 0 N and 30 0 S over the Pacific Ocean for the years 1900-1973. The data were used to create
maps of average monthly winds and wind stresses for 2° x 10° longitude-latitude quadrangles.
Wyrtki and Meyers (1976) used the data to study both seasonal and interannual variability of
the Pacific Ocean trade wind field. Barnett (1977) has used the data to investigate the time
and space variability of the trade winds. Reiter (1978a,b) utilized the data to investigate the
interannual changes in the ocean-atmosphere system. This wind data for the lO-year period
1961-70 were subjectively analyzed and digitized by Goldenb~rg a.nd O'Brien (1981) (GO) on a
2° x 2° longitude-latitude grid. The needs for creating such a revised dataset were that errors
existed in the original data set, a finer grid was necessary to drive coupled atmospheric oceans
models (the original grid was 2° x 10°) and the existence for many months of regions entirely void
of data or 2° x 10° boxes containing so few observations that the representativeness of average
monthly values was in question.
Therefore, the revised Florida State University (FSU) wind data of Goldenberg and O'Brien
(1981), consisted in monthly-averaged wind stress values for the tropical Pacific Ocean on a
2°x2° longitude-latitude grid for the lO-year period January 1961-December 1970. This period
was chosen as having the most consistent data coverage for the years 1947-73. The stresses of
WM were computed from individual ship wind observations according to the bulk formulae:
where T is the wind stress vector, Pa is the air density, CD is the drag coefficient and (j is the
wind velocity. Constant values were assumed for the drag coefficient, CD = 1.5 X 10- 3 , and for
the air density, Pa = 1.2 Kg m- 3 . The justification for these constant values comes from the fact
that it has been demonstrated that the linear stress law (i.e., C D=cte) is adequate for spectral
analysis involving periods longer than 10 days (Willebmnd, 1978). Therefore, in the analysis