1. Graduate School of Oceanography, University of Rhode Island, S. Ferry Road, Narragansett, Rhode Island 02882, USA

Correspondence to: B. Mete Uz Correspondence and requests for materials should be addressed to B.M.U. (e-mail: Email: m.uz@gso.uri.edu).

Much of the motion in the open ocean is baroclinic: at the surface, geostrophic flow balances the pressure gradient caused by a sloping sea surface. The geostrophic current is typically strongest near the surface, and decreases with depth. Because the horizontal pressure gradient must similarly decrease with depth in order to maintain geostrophic balance, constant-density surfaces must slope in the opposite direction to the sea surface. As a result, a baroclinic disturbance that displaces the constant-density surfaces upwards has a negative sea surface height (SSH) anomaly, which can be measured by altimeters. These disturbances propagate westward—as planetary waves or eddies—mostly due to the variation of the Coriolis parameter with latitude. Westward-propagating first-mode baroclinic features dominate the propagating components of the SSH satellite-sensed record from the TOPEX/POSEIDON mission^{7, 8, 9}. They have also been observed in images of sea surface temperature from the satellite-borne Advanced Very High Resolution Radiometer^{10, 11, 12}.

If the displacement of constant-density surfaces by baroclinic disturbances brought new nutrients into the euphotic zone, the concentration of phytoplankton chlorophyll would be expected to increase above the disturbance. As the disturbance propagates westward, there would be enhanced production near the leading edge and fall-out and grazing of the biomass anomaly near the trailing edge, so that the chlorophyll anomaly would also propagate with the disturbance. If this were to happen, chlorophyll variability would have a component that was coherent with the SSH variations.

To look for this signature in the satellite imagery, we sought large-scale anomalies in SeaWiFS¹³ remotely sensed oceanic chlorophyll concentration that propagate westward in coherence with SSH anomalies. Even though the eddy pumping mechanism was originally intended for mesoscale eddies, our study focuses on the larger scales of Rossby waves; this larger scale enables us to do more spatial smoothing and compositing (replacing cloud-masked pixels of one image with corresponding pixels of a consecutive image) over time. The compression of the upper layers of the ocean by a Rossby wave is very similar dynamically to the compression caused by a small-amplitude eddy, and the same mechanism of nutrient enhancement should apply. With a nonlinear eddy that transports water, the pumping would occur only once—during the spin-up of the eddy.

Propagating anomalies were readily observed as slanted features in time–longitude diagrams (Fig. 1) once data were bandpass-filtered to remove noise and variability at small scales as well as seasonal and other very-large-scale variations. These anomalies persisted for a long time. For example, a high-chlorophyll-concentration anomaly at 30° N which was at longitude -200° (160° E) around day 150 (150 days after 7 September 1997) was still visible on day 550, by which time it had moved to longitude -225° (135° E).

Figure 1: Time–longitude diagrams of the bandpass-filtered SeaWiFS chlorophyll concentration data.

The figure shows zonal propagation of anomalies as slanted features. The phase speeds of linear non-dispersive first-mode baroclinic Rossby waves are shown by a solid line at each latitude (a, 30° N; b, 20° N; c, 10° N). The variation with latitude of the propagation speed of chlorophyll anomalies may be seen as an animation sequence at http://po.gso.uri.edu/~uz/chl_anomaly.htm.

High resolution image and legend (233K)

We used spectral analysis to investigate the coherence between chlorophyll and SSH features in order to separate different scales of variability and propagation directions. We calculated the power spectral density of—and coherence between—chlorophyll and SSH for individual ocean basins. At most latitudes, westward-propagating features made up almost all of the spectral density of SSH, whereas chlorophyll spectral density showed more spread. Part of the spread in the chlorophyll spectral density may be due to non-propagating phenomena such as the expansion of blooms in both directions. High coherence between chlorophyll and SSH was mostly in the westward-propagation quadrants (Fig. 2).

Figure 2: Wavenumber–frequency spectra of coherence between log(chlorophyll concentration) and sea surface height.

Most of the spectral density is in the second and fourth quadrants, where propagation is westward. Data shown are from the Pacific Ocean at latitudes of 30° N (a), 20° N (b) and 10° N (c). Dotted lines bracketing the axes show cut-off points for the highpass filters. The linear dispersion relation is shown by a solid line.

High resolution image and legend (125K)

Phase speeds of planetary waves measured with the TOPEX/POSEIDON altimeter have been seen to vary between 0.5 and 3 times the phase speed from the linear dispersion⁷, though a recent study reports that with different methods the linear phase speed is within the error bars¹⁴. Nevertheless, the theoretical phase speed of baroclinic planetary waves may not be adequate for a precise comparison with measurements. Qualitatively, the propagation speeds of chlorophyll anomalies that we saw in the time–longitude diagrams were similar to the theoretical propagation speed of first-mode baroclinic planetary waves, and (more importantly) they varied with latitude in the same way.

We calculated propagation speeds from the chlorophyll and chlorophyll–SSH coherence spectra, and compared these to the non-dispersive linear phase speed calculated using climatological density profiles¹⁵ (Fig. 3). Between 10° and 25° in both hemispheres, estimates from chlorophyll, SSH and chlorophyll–SSH coherence were all similar and within a factor of two of the theoretical phase speed. At higher latitudes (>25°), propagation speeds from chlorophyll spectra were significantly higher than the rest. This deviation must be caused by phenomena not related to SSH, because the propagation speed of the SSH coherent features (the estimate from chlorophyll–SSH coherence) remained close to the phase speed from SSH and the theoretical phase speed.

Figure 3: Propagation speeds of Chl and SSH anomalies.

The figure shows that propagation speeds of chlorophyll anomalies associated with baroclinic features vary with latitude in nearly the same way as do the theoretical phase speeds of Rossby waves (dotted lines). The propagation speeds are estimated from the wavenumber spectra of chlorophyll (dash-dot line), sea surface height (SSH; dashed line) and chlorophyll–SSH coherence (solid lines) in the Pacific Ocean. The theoretical speeds are based on baroclinic radii of deformation from ref. 15.

High resolution image and legend (29K)

The confinement of high chlorophyll–SSH coherence in a localized area of the wavenumber–frequency domain allowed us to construct a filter to isolate chlorophyll features which were associated with the SSH anomalies. The variance of the coherent, westward-propagating signal was around 5–20% of the variance of the input (Fig. 4). These fractions depend on the choice of the threshold coherence, and are provided here as only a rough scale for the magnitude of the effect. This fraction varied with latitude, and between basins at any given latitude. There were local minima at latitudes corresponding to the middle of the subtropical gyres. The coherence estimate is sensitive to the relative magnitude of error in the data. If each data series is modelled as the sum of an ideal signal and random noise, wherever the amplitudes of the signals are reduced, the coherence estimate also goes down even though the coherence between the two ideal signals remains the same. This effect may be what we are seeing in the middle of the gyres, where the mean as well as the variability of chlorophyll is low and therefore the noise figure is high. The local minima at these locations may then be mostly in the estimate, and might not occur in the estimates if more accurate data were available. Nevertheless, in all basins there were large areas where coherence was observed between chlorophyll and SSH. We note that our analysis used the entire time record; for chlorophyll and SSH features to be coherent, they must not only coincide in space at one point in time, but they must propagate together. A phase offset may exist between chlorophyll and SSH features without reducing the coherence as long as the offset remains constant.

Figure 4: Features associated with westward-propagating baroclinic disturbances make up 10% of the variance of the log(chlorophyll concentration) data.

These chlorophyll data were filtered so as to eliminate variability at length scales larger than 25° and timescales longer than 260 d.

High resolution image and legend (34K)

The mechanism of enhanced primary production proposed in ref. 3 may explain how the chlorophyll anomaly can follow the baroclinic disturbance detected by the SSH anomaly. In this mechanism, the disturbance introduces nutrients into the euphotic zone; primary production is therefore enhanced, and eventually chlorophyll concentration increases in the surface water over the nutrient-enriched area. The nutrient enhancement may occur through the compression of near-surface layers, so that deeper layers richer in nutrients extend into the euphotic zone¹⁶. Alternatively, it may occur through isopycnal mixing within a tilted constant-density layer that protrudes into the euphotic layer over the disturbance¹⁷. There may also be a physiological delay between the introduction of the nutrients and the enhancement in chlorophyll concentration. For large-scale disturbances that have periods on the order of months, the phase offset caused by such delays should be very small, and unlikely to change much even as the anomaly propagates through areas of different planktonic community structures. The chlorophyll anomaly induced through this mechanism by a SSH anomaly should be easily detected with our methods.

The coherence between SSH and chlorophyll that we observe is consistent with the eddy pumping mechanism, but does not yet prove it: there are means other than enhanced primary production by which the coherence may be created. One of these is the advection of a horizontal background chlorophyll gradient by the geostrophic velocity associated with the baroclinic anomaly. In this case, chlorophyll acts as a passive tracer and the contrast visible in the imagery will be determined by the magnitude of the gradient and its displacement by the disturbance. Another possible mechanism is the shoaling of a deep chlorophyll maximum by the disturbance from an original depth too deep to be seen by the sensor. In the interior of the gyres, where the horizontal chlorophyll gradients are weak and the deep chlorophyll maximum occurs at greater depths, neither of these alternative mechanisms would be very effective. The presence of propagating chlorophyll anomalies in these areas supports the hypothesis of enhanced primary productivity.

References

References

Acknowledgements

We thank the SeaWiFS team for chlorophyll data and AVISO for SSH data; we also thank D. Siegel, P. Polito and S. Schollaert for discussions and recommendations. This work was supported by NASA.