Roger Lukas*,
Peter Hacker*, Ming Mao*, Eric Firing*,

Adriana Huyer, and P. Michael Kosro

* School of Ocean and Earth Science and Technology,

University of Hawaii, Honolulu, HI 96822

College of Oceanic and Atmospheric Sciences,

Oregon State University, Corvallis, OR 97331

Velocity observations were made with a 150 kHz, RD Instruments acoustic Doppler current profiler. Dual GPS receivers were used to determine ship heading and to correct for gyro heading errors. The nominal depth range of observations is 20-300 m. Vertical resolution is 8 m; horizontal resolution is less than 1 km. The velocity observations are accurate to better than 2 cm s¯ ¹ on horizontal scales greater than 10 km. Repeat transport calculations around the closed survey segments indicate that current biases are less than 1 cm s¯ ¹. Observations were made almost continuously along the standard butterfly track within the IFA during the periods 13 Nov. to 2 Dec. 1992 (Leg 1), 18 Dec. 1992 to 9 Jan. 1993 (Leg 2), and 27 Jan. to 15 Feb. 1993 (Leg 3).

Our initial analysis focused on the horizontal and vertical structure of velocity along individual segments of the butterfly surveys, and on the mean spatial structure over the three legs. We early realized that the space-time sampling along individual velocity sections, which took about 6 and 10 hours to complete, convolved the unusually strong internal semidiurnal tidal signals, making estimates of horizontal gradients problematic. In order to remove the effect of tides and other higher frequency and short spatial scale features, we implemented a modification of the Candela *et al.* (1992) method.

Our purpose in this analysis is to determine the spatial and temporal gradients of the velocity field on time scales longer than a day, in order to evaluate the various velocity terms in the momentum and scalar property budgets. Preliminary results from our analysis are presented here.

The method consists of representing the observations for each depth and time interval as a sum of terms of the form:

where u(x,y,t) is the observed eastward current component at a given time and position; the are the frequencies of the i

For each mapping period, the coefficients are determined by imposing a least squares requirement on the residuals. We used a three-day period for each analysis, which is long enough to allow for 2 complete circuits around the survey pattern, allowing us to estimate the temporal change at each point. A longer analysis period would have smoothed the rapid temporal changes observed during the IOP. The three-day analysis window was moved temporally through the data in 12-hour steps, to yield a timecentered velocity field estimate every 12 hours.

The velocity terms in the following figures represent estimates at the center point of the survey pattern. Estimates can be generated at all points within the survey domain.

A measure of the robustness of our estimates can be obtained from the number of velocity samples in each analysis, the condition number of the matrix, and the covariance matrix of the coefficients.

Our studies indicate that:

- Performing the analysis over the entire survey domain for 3-day intervals gives robust results.
- Performing the analysis over subdomains such as the individual triangles occasionally gives problematic results when there are too few samples within the 3-day window to estimate the sharper spatial gradients.
- The particular form of the model we used did not completely separate the semidiurnal tide from other high frequency and small-scale variability contained in the residuals, however this does not affect the results for low frequency, large-scale variability. The residual term in the analysis is relatively large (about three times larger than the tidal component) suggesting that energetic small scale features such as internal waves and eddies are present in the data and are not well mapped by the survey ship alone.

While the basic structure of the zonal flow did not change markedly during the IOP, the meridional flow varied considerably. During Leg 1, the flow was broadly southward, except at the beginning and end of the cruise. Vertical coherence was high. During Leg 2, meridional flow was energetic and highly variable in time and depth; there is a suggestion of vertical phase delay in alternating bands of southward and northward flow. Leg 3 showed a variable and vertically-layered structure with weak vertical coherence. Distinctly different time-depth structures of velocity gradients were observed among the three cruises.

Time-dependent velocity, and velocity gradient (not shown), vertical profiles derived from the model were combined to estimate the advective terms in the momentum balance. During Leg 1, the nonlinear terms were generally an order of magnitude smaller than the local acceleration, Leg 2 had significant nonlinear terms, and Leg 3 had nonlinear accelerations as large as the local acceleration.

The relative vorticity and divergence were computed from the velocity gradients; continuity was used to also estimate the vertical velocity component from the divergence (Fig. 3). Upwelling dominated during Leg 1, while alternating periods of upwelling and downwelling were characteristic of Legs 2 and 3. The magnitude of the relative vorticity () was often comparable to local f (=-4.7 x 10¯^{6}s¯ ¹), and considerably larger during Leg 3. The vertical structure of was similar for Legs 1 and 2, with being the same sign as f in the upper 100 m, and in a layer centered near 200 m. was positive in the layers centered near 120 and 300 m. During Leg 3, however, the upper 80 m and the layer between 220 and 300 m had positive , with the layer between 100-220 m having opposite sign with absolute vorticity between 2*f and 4*f. Nonlinear dynamics were clearly important in the momentum balance within the IFA during the TOGA COARE IOP, but most especially near the end of Leg 2 and during Leg 3.