Due to the leak in the gappy western equatorial Pacific, sea level (SL) is highly correlated with El Niño all along the western Australian coast. According to standard theory, this coastal interannual (IA) signal should propagate westward as Rossby waves with large zonal scale. High-resolution satellite SL estimates show that along the shelf edge south of 23°S the IA SL signal does not have the expected large spatial scale as it decreases rapidly seaward from the shelf edge. The drop in IA SL amplitude coincides with the mean position of the Leeuwin Current (LC). Theory shows how a nearly meridional mean flow, as in the case of the LC, can induce this fall in IA signal amplitude. The associated IA shelf-edge flow tends to strengthen the LC during La Niña, weaken it during El Niño and may profoundly affect the recruitment of the western rock lobster.
 During an El Niño anomalous westerly equatorial winds push the surface equatorial layer of water eastward, lowering SL in the western equatorial Pacific. This lowered SL signal ‘leaks’ through the gap in the western equatorial Pacific ‘boundary’, and, consistent with geostrophic balance and no flow normal to the coast, SL is anomalously low and spatially constant all along the western Australian (WA) coast. Conversely, during La Niña, anomalous easterly winds push the surface equatorial layer of water westward, raising SL in the western Pacific and along the WA coast. This dynamics, proposed by Clarke , was used to explain the IA Australian coastal SL observations reported by Pariwono et al. . We verified this dynamics with more recent Jan 1986–Dec 2002 monthly SL at the 6 WA coastal stations Darwin, Wyndham, Broome, Port Hedland, Carnarvon and Fremantle. The first empirical orthogonal function (EOF) of these sea levels filtered with an IA filter [Trenberth, 1984] explains nearly all (94%) of the variance. As expected the spatial structure function is nearly constant along the coast (with an amplitude of about 8 cm) and the principal component time series is highly correlated (r = −0.81) with the interannually filtered El Niño index NINO3.4.
 Based on low-frequency theory [Schopf et al., 1981; Cane and Moore, 1981; Clarke, 1983], this WA SL signal should propagate westward as Rossby waves with wavelengths ∼ several hundred kilometers or more. In the LC region, within about 100 km of the shelf edge and south of 20S, the IA signal therefore should hardly change seaward from the shelf edge. We can test this gentle slope prediction and calculate the corresponding geostrophic flow using coastal tide gauge and TOPEX/POSEIDON (T/P) satellite SL estimates.
2. Interannual Shelf-Edge Flow
 The T/P satellite estimates SL height every 6–7 km, repeating the measurement every 10 days along tracks about 250–300 km apart. The along-track estimates are usually spatially averaged to reduce the noise, but at a cost of decreased spatial resolution. Here we need high spatial resolution to estimate accurately SL gradients and the associated geostrophic flow. Since we are interested in low-frequency flow we can remove noise by averaging in time, first forming monthly T/P SL heights from the 10-day estimates and then filtering using the same IA filter [Trenberth, 1984] as for the coastal time series. Although the raw monthly satellite records are nearly a decade long, beginning in Jan 1993 and ending in Apr 2002, they are still short on an IA time scale. However our results are consistent with theory and a recent 50-yr long hydrographic analysis at 32°S [Feng et al., 2003].
Figure 1 shows the T/P satellite tracks off Western Australia. Along the red and blue tracks we have plotted (Figure 2) correlation and regression coefficients of the satellite-estimated SL with the nearest of the 6 IA coastal tide gauge time series. Raw satellite SL height estimates are subject to increased error near the coast and points closer to the coast than about 50 km along-track were not used. The very high correlation of the satellite SL estimates we did use near the coast suggests that our IA satellite SL estimates are accurate.
 The correlation coefficient is nearly one for all points along the red track and the regression coefficient decreases only very slowly from unity from the coast, consistent with the in-phase constant amplitude coastal SL relationship and the expected huge westward scales of Rossby waves propagating westward from the coast. In marked contrast, correlation and regression coefficients along the blue track do not behave at all like that expected for Rossby wave dynamics, dropping sharply seaward of the shelf edge due to a decrease in the IA SL signal. The red track behavior is typical for tracks crossing the shelf edge north of 23°S and the blue track behavior for tracks crossing the shelf edge south of 23°S. By geostrophic balance, the coherent fall in SL amplitude seaward of the shelf edge for the blue track corresponds to an anomalous flow perpendicular to the track. This is only one component of the (vector) flow which, given the proximity of the shelf edge, is probably parallel to the shelf edge.
 Flow direction and magnitude near the shelf edge can be checked at the points A and B in Figure 1 where satellite tracks cross and thus give us two velocity components at a given point. The first EOF of satellite SL heights near the intersecting tracks explains nearly all of the variance in each case (89% near A and 97% near B) and gradients of the spatial EOF functions show that the IA flow is indeed directed alongshore with typical IA velocity amplitudes of 4.5 cm/sec at A and 1.4 cm/sec at B. The velocity vectors at A and B are highly positively correlated with the IA coastal SL record (r = 0.87 and 0.97 respectively), suggesting that when the coastal SL is higher than normal (typically during La Niñas), the along shelf-edge flow anomaly is southward and when the coastal SL is lower than normal (typical during El Niño), the along shelf-edge flow anomaly is northward.
 If we make the assumption that the current near the shelf edge is parallel to the shelf edge, then we can estimate the anomalous geostrophic along-shelf edge IA flow anywhere a satellite track crosses the shelf edge. The results of such calculations are shown in Figure 3. Consistent with the sharp drop in IA SL amplitude seaward of the shelf edge only occurring south of about 23°S, by geostrophy the anomalous shelf-edge IA flow is much stronger south of 23°S.
 The sudden drop in IA SL amplitude just seaward of the shelf edge and the associated along shelf-edge flow occurs as the IA signal tries to cross the LC, a strong narrow southward flow running along the shelf edge south of about 23°S. It seems that the mean LC somehow distorts the large-scale Rossby wave IA signal to produce an IA shelf edge flow.
 While the influence of zonal mean flow on low-frequency ocean variability has been the subject of several observational and theoretical studies, little attention has been given to the influence of non zonal flows like the LC. To examine this influence, we consider a simple ocean model consisting of two layers of water of constant density, the upper layer, about 250 m thick, being much thinner than the lower layer, greater than 2 km deep. The upper layer flow is governed by the potential vorticity equation
where h is the thickness of the upper layer, u is the horizontal velocity, ∇ the horizontal gradient operator, t the time, ζ the relative vorticity due to the movement of the water and f is the Coriolis parameter.
 Writing u = + u′, ζ = + ζ′ and h = + h′ where refers to the mean plus annual cycle and ( )′ the IA deviation, the linearized anomalous version of (1) for small IA perturbations is
where q is the potential vorticity and and linearized q′ are given as
Equation (2) can be considerably simplified. First, because and u′ are quasigeostrophic, with g* as the reduced gravity
For the realistic values ∣u′∣ ∼4 cm/sec, ∣∣ ∼40 cm/sec, h′ ∼10% of , ∣f∣ ∼7 × 10−5 s−1, β northward gradient of f = 2 × 10−11 m−1 s−1, ∂/∂t ∼ ω = 2π/3 yrs, across current length scale ∼100 km and an along current length scale ∼1000 km, the dominant balance in equation (6) is
where the large-scale mean potential vorticity qLS is f/ and the large-scale perturbation potential vorticity is q′LS = −fh′/()2. Physically, equations (7) and (8) show that because mean meridional flow can advect planetary vorticity ( · f ≠ 0), to conserve potential vorticity the southward advection of the anomalous large-scale potential vorticity by the mean LC ( · ∇q′LS ≠ 0) must be balanced by an anomalous flow advecting mean potential vorticity (u′ · LS).
 In the region of interest near the shelf edge, and u′ are parallel to the shelf edge. Therefore if es is a unit vector parallel to the shelf edge, u′ = v′es and = es. Consequently equation (7) reduces to
Since anomalous upper layer thickness h′ is in phase with anomalous SL, equation (9) is consistent with observations; the anomalous flow v′ is in the same direction as the LC when the SL is high (usually during La Niña) and opposes it when the SL is low (usually during El Niño). Since the IA variations of h′ are about 10% of , from equation (9)v′ should have a typical amplitude of about 10% of , i.e., about 2–5 cm/sec as observed.
 The above observations and theory for IA shelf edge flow differ markedly from that expected when there is no mean flow. Near an eastern ocean boundary with = 0, IA flow has large spatial scales and ζ is negligible in equation (3). Therefore, for a coastline that does not run east–west, equation (2) reduces to the Rossby wave balance
where θ is the angle that the coastline makes with due north [Clarke, 1983]. Since h′ is proportional to coastal SL, the alongshore flow in (10) is proportional to the time derivative of SL rather than SL itself as in the observations.
4. Interannual Flow and Marine Life
 Variations in the strength of the LC profoundly affect marine life. To see how this may happen, consider Australia's most lucrative single species fishery, that of the western rock lobster (Panulirus cygnus). Western rock lobsters spawn on the outer continental shelf between Jul–Sep and the larvae hatch in the southern hemisphere summer from Dec to Mar. They rise to the surface, are carried rapidly offshore in the thin wind-driven surface Ekman layer and are widely dispersed over the SE Indian Ocean. The greatest concentrations are 400–1000 km from the coast. By the southern winter and spring the larvae have reached mid and late larval stages, remaining at depths of 50–120 m during the day. At this depth they are below the winter and spring weakly wind-driven surface Ekman layer flow and right in the heart of the eastward geostrophic flow that supplies the narrow LC running southward along the shelf edge. The eastward flow carries the larvae toward the shelf edge where they metamorphose to the puerulus stage. Unlike the larvae, the pueruli can swim effectively and they head toward the coast across the continental shelf in spring and summer, settling in the shallow reefs. After about 4 yrs of growth in these reefs, most juvenile lobsters migrate offshore to depths of 30–200 m. In about another one to two years the juveniles have become adults and can spawn and begin the life cycle again.
Pearce and Phillips  showed that the logarithm of an index of the number of pueruli settling near the WA coast is positively linearly correlated with variations in annual Fremantle SL and hence LC variability. This approximate statistical relationship may be expressed as
where n is the number of pueruli settling at the coast, μ is a constant independent of n, η is the IA coastal SL anomaly and n0 is the number of pueruli settling under normal conditions (η = 0). Equation (11) shows that the number of settling pueruli depends exponentially on η and so can vary greatly, often halving or doubling from one year to the next.
Pearce and Phillips  hypothesized that a higher SL would correspond to a stronger LC ‘trapped on the shelf and upper slope’ and that this somehow increases the puerulus coastal settlement, perhaps [Phillips and Pearce, 1997] by increased mixing of the shelf and slope waters by the stronger LC. Figure 1, Figure 2 and our theory suggest a different current structure, one in which the IA flow on the shelf is negligible but is strong in the region of the LC. Figures 1 and 4 also suggest another possible mechanism for the linkage of variable coastal puerulus settlement and IA SL. They show that 100 km seaward of the shelf break the IA SL amplitude drops alongshore by about 6 cm over 600 km from location 6 (22°S) to location 8 (27.5°S). So when the SL is anomalously high (low), there is a geostrophic onshore (offshore) flow anomaly of about 1.5 cm/sec, feeding into the mean LC and strengthening (weakening) it. This anomalous feeding of the LC is consistent with Figure 3 showing increased flow anomalies south of 26°S. Since the mean flow toward the coast is about 5 cm/sec and the onshore-offshore current anomaly has a 1.5 cm/sec amplitude, the flow speed toward the coast varies interannually from about 6.5 cm/sec (high SL) to 3.5 cm/sec (low SL). Increased (decreased) flow toward the coast during high (low) SL may result in increased (decreased) advection of larvae toward the shelf edge and consequent increased (decreased) puerulus settlement.
Griffin et al.  examined larval advection using a biological model embedded in flow largely estimated from satellite SL heights from Sep 1992 till the end of 1998. The model currents used in the calculations were highly resolved in time (every 5 days) with 20 km spatial resolution. Surprisingly, the model predicted that the number of larvae metamorphosing offshore should be essentially constant from year to year, in disagreement with the large variability seen in the observations. A likely reason for the discrepancy is that Griffin et al. had larvae in the model hatch between 27°–34°20′S, mostly south of the main region 22°–27.5°S of IA onshore flow. Although the lobsters range from about 22°–34°S [Phillips, 1981], Griffin et al. chose the southern region because more lobsters spawn there. However, spawning in the northern region should not be ignored because particles in the northern onshore flow are, on average, advected closer to the shelf edge and traverse a longer section of shelf edge than those in the southern part of the onshore flow. Proximity to the shelf edge is crucial because food is more abundant there and late-stage larvae need that food to metamorphose to non-feeding pueruli which swim to the coast [Phillips and Pearce, 1997].
 While other factors may contribute to variation in coastal puerulus settlement [Caputi et al., 2001], it seems likely that year-to-year changes in the strength of the LC have a major influence. By this mechanism, we expect the coastal pueruli settlement to increase (decrease) during high (low) SL. But why should the settlement depend exponentially on the SL anomaly (see equation (11))? Suppose n is the number of pueruli settling at the coast when the mean flow toward the coast is u. We postulate that increasing the speed of the flow toward the coast by a small amount δu will result in a small percentage increase in pueruli coastal settlement that is proportional to δu, i.e., 100 δn/n is proportional to δu. But since δu is proportional to the increased flow of the LC and hence to the SL increase δη, we may write
for some constant μ. Integrating from the neutral conditions η = 0, n = n0 to general η and n gives the empirical relationship (11).
 Although we have focused on IA flow and rock lobsters off Western Australia, high-resolution along-track satellite data is global and so can be used to analyse and understand IA flows near most of the world's coasts. Much progress in our understanding of the large year-to-year fluctuations in many species of coastal marine life will be possible when these IA flows are documented and better understood.