Influence of the equatorial deep jets on the north equatorial countercurrent

Abstract

An ocean circulation model is run using two different idealized equatorial basin configurations under steady wind forcing. Both model versions produce bands of vertically alternating zonal flow at depth, similar to observed Equatorial Deep Jets (EDJs) and with a time scale corresponding to that of the gravest equatorial basin mode for the dominant baroclinic vertical normal mode. Both model runs show evidence for enhanced variability in the surface signature of the North Equatorial Counter Current (NECC) with the same time scale. We also find the same link between the observed NECC and the EDJs in the Atlantic by comparing the signature of the EDJ in moored zonal velocity data at 23° W on the equator with the signature of the NECC in geostrophic velocities from altimeter data. We argue that the presence of a peak in variability in the NECC associated with the EDJ basin mode period is evidence that the influenceatthis time scale is upward, from the EDJ to the NECC.

Appendix

The spectra in Figs. 5b, 8a, and 9a are derived first by computing the modulus of the Fourier transform, $\hat {u}_{\omega ,x,y,z}$, of the zonal velocity, u _t,x,y,z , for each frequency ω and at each grid point inside the respective boxes; these boxes are a surface box (referred to in the figure captions) and a deep box extending from 450 to 550 m depth at a location on the equator near the center of the basin (45° from the western boundary for model A and 30° for model B). The spectra shown in the figures are then computed by averaging these moduli over the respective boxes.

To estimate the significance of the peaks in the spectra, we first extract a subsample of the time series in u _t,x,y,z . The reason for using a subsample is to take account of the reduction in the number of degrees of freedom arising from the autocorrelation of the time series. We then construct random permutations of the subsampled time series using the same permutation at each grid point within each box. Using the same permutation at each grid box preserves the covariance in the x, y, and z directions within each box. For each realization the mean spectrum in the boxes is then computed in the same way as for the original time series. This ensures that the randomly generated spectra have the same number of degrees of freedom in the boxes as the original spectra. To arrive at a significance level, the original spectra are compared to the 90^th percentile of the randomly generated spectra. Then, 10,000 different permutations are used and for each frequency, the amplitude at which 9000 realizations have a lower amplitude defines the 90^th percentile, where amplitude here refers to the average of the moduli of the Fourier transforms within each box. A peak in the original spectrum is then significant at the 90 % level if the amplitude at that frequency exceeds the 90^th percent percentile.