• Bograd, S. J., , T. K. Chereskin, , and D. Roemmich, 2001: Transport of mass, heat, salt and nutrients in the southern California Current System: Annual cycle and interannual variability. J. Geophys. Res., 106 , (C5). 92559275.

    • Search Google Scholar
    • Export Citation
  • Cane, M. A., , and D. W. Moore, 1981: A note on low-frequency equatorial basin modes. J. Phys. Oceanogr., 11 , 15781584.

  • Chelton, D. B., , and R. E. Davis, 1982: Monthly mean sea level variability along the west coast of North America. J. Phys. Oceanogr., 12 , 757784.

    • Search Google Scholar
    • Export Citation
  • Chelton, D. B., , P. A. Bernal, , and J. A. McGowan, 1982: Large-scale interannual physical and biological interaction in the California Current. J. Mar. Res., 40 , 10951125.

    • Search Google Scholar
    • Export Citation
  • Chhak, K. C., , E. Di Lorenzo, , N. Schneider, , and P. F. Cummins, 2009: Forcing of low-frequency ocean variability in the northeast Pacific. J. Climate, 22 , 12551276.

    • Search Google Scholar
    • Export Citation
  • Clarke, A. J., 1983: The reflection of equatorial waves from oceanic boundaries. J. Phys. Oceanogr., 13 , 11931207.

  • Clarke, A. J., , and C. Shi, 1991: Critical frequencies at ocean boundaries. J. Geophys. Res., 96 , 1073110738.

  • Clarke, A. J., , and A. Lebedev, 1999: Remotely driven decadal and longer changes in the coastal Pacific waters of the Americas. J. Phys. Oceanogr., 29 , 828835.

    • Search Google Scholar
    • Export Citation
  • Clarke, A. J., , and M. Dottori, 2008: Planetary wave propagation off California and its effect on zooplankton. J. Phys. Oceanogr., 38 , 702714.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., , A. J. Miller, , N. Schneider, , and J. C. McWilliams, 2005: The warming of the California Current: Dynamics and ecosystem implications. J. Phys. Oceanogr., 35 , 336362.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., and Coauthors, 2008: North Pacific Gyre Oscillation links ocean climate and ecosystem change. Geophys. Res. Lett., 35 , L08607. doi:10.1029/2007GL032838.

    • Search Google Scholar
    • Export Citation
  • Ebisuzaki, W., 1997: A method to estimate the statistical significance of a correlation when the data are serially correlated. J. Climate, 10 , 21472153.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., , and J. S. Allen, 1980: On the structure and dynamics of monthly mean sea level anomalies along the Pacific coast of North and South America. J. Phys. Oceanogr., 10 , 557578.

    • Search Google Scholar
    • Export Citation
  • Grimshaw, R., , and J. S. Allen, 1988: Low-frequency baroclinic waves off coastal boundaries. J. Phys. Oceanogr., 18 , 11241143.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kessler, W. S., 1990: Observations of long Rossby waves in the northern tropical Pacific. J. Geophys. Res., 95 , 51835217.

  • Killworth, P. D., , and J. R. Blundell, 2003: Long extratropical planetary wave propagation in the presence of slowly varying mean flow and bottom topography. Part I: The local problem. J. Phys. Oceanogr., 33 , 784801.

    • Search Google Scholar
    • Export Citation
  • Killworth, P. D., , D. B. Chelton, , and R. A. de Szoeke, 1997: The speed of observed and theoretical long extratropical planetary waves. J. Phys. Oceanogr., 27 , 19461966.

    • Search Google Scholar
    • Export Citation
  • Schneider, N., , E. Di Lorenzo, , and P. P. Niiler, 2005: Salinity variations in the southern California Current. J. Phys. Oceanogr., 35 , 14211436.

    • Search Google Scholar
    • Export Citation
  • Schopf, P. S., , D. L. T. Anderson, , and R. Smith, 1981: Beta-dispersion of low-frequency Rossby waves. Dyn. Atmos. Oceans, 5 , 187214.

  • Trenberth, K. E., 1984: Signal versus noise in the Southern Oscillation. Mon. Wea. Rev., 112 , 326332.

  • Wang, W., , and M. J. McPhaden, 1999: The surface heat balance in the equatorial Pacific Ocean. Part I: Mean seasonal cycle. J. Phys. Oceanogr., 29 , 18121831.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) Monthly sea level anomalies at San Diego (gray line) and the monthly Niño-3.4 index (black line). Niño-3.4 is the SST anomaly averaged over the east-central equatorial Pacific region 5°S–5°N, 170°–120°W. Both time series have been detrended and filtered with the interannual filter of Trenberth (1984). This 11-point symmetric filter passes more than 80% of the amplitude at a period of 24 months and longer and less than 10% of the amplitude for periods shorter than 8 months. Both time series have also been normalized by their standard deviations (σ = 37 mm for sea level and 0.70°C for Niño-3.4). The maximum correlation between the time series occurs at zero lag and is r = 0.70, significant at the 99% level. Here and elsewhere, the critical correlation coefficient was determined using the method of Ebisuzaki (1997). (b) Same as in (a), but for the San Diego sea level (gray line) and alongshore wind stress near San Diego (black line). In this case, the highest correlation (r = 0.20) at zero lag is insignificant statistically. The standard deviation of the alongshore wind stress is 0.0041 Pa. (c) Same as in (a), but for the San Diego sea level (gray line), the 5-m San Diego temperature (black line), and the San Diego surface temperature (dashed line). The highest correlation (r = 0.72) between the sea level and 5-m temperature occurs when sea level leads temperature by 1 month, but the correlation is nearly the same (r = 0.71) at zero lag. The corresponding correlations for the surface temperature are r = 0.70 and r = 0.69, respectively. All correlations are significant at the 99% level. The standard deviation of the surface temperature is 0.66°C and that of the 5-m temperature is 0.68°C.

  • View in gallery

    The CalCOFI region off of southern California showing the TOPEX/Poseidon–Jason-1 satellite tracks and 63 frequently sampled CalCOFI stations. The stations are grouped into six regions according to their zonal distance from the coast in bins of approximately 130 km. Specifically, the solid inverted triangles (region 1) represent all stations within 130 km of the coast, shaded circles (region 2) show all stations between 130 and 260 km from the coast, solid hexagons (region 3) are for stations 260–390 km from the coast, etc. (from Clarke and Dottori 2008).

  • View in gallery

    Estimates of westward sea level propagation off California using dynamic height anomalies relative to 500 m from CalCOFI hydrographic data (solid regression line, solid diamond) and from TOPEX/Poseidon–Jason-1 satellite-estimated sea level anomalies (dashed regression line, solid circles). The plotted lags for the dynamic height were found by lag correlating the gappy dynamic height data for each of the six regions in Fig. 2 with the monthly sea level anomaly record at San Diego and recording the lags at maximum correlation. Before the correlation calculations, the San Diego sea level anomalies were detrended and low-pass filtered using the Trenberth (1984) 11-point symmetric filter to obtain interannual and lower-frequency sea levels. This filter passes more than 80% of the amplitude at a period of 24 months and longer and less than 10% of the amplitude for periods shorter than 8 months. The lag correlation calculations were also carried out on the satellite data in a similar way, first binning the along-track data into the six regions to form the six monthly time series. [Redrawn from Clarke and Dottori (2008)]

  • View in gallery

    Normal and tangential coordinates (n, s) to a coastline making an angle θ with due north (from Clarke and Dottori 2008).

  • View in gallery

    One hundred thousand times the isopycnal tilt angle (rad) as a function of depth for the top 500 m for each of the five regions closest to the coast in Fig. 2. Below 500 m, the slope is negligible. Positive ϕ corresponds to isopycnals tilting upward toward the north. Region 6 was not included because there were too few grid points alongshore to estimate ϕ. The symbols represent the regions in Fig. 2.

  • View in gallery

    Buoyancy frequency squared (s−2) multiplied by 10 000 for region 3 (see Fig. 2). The other regions have a very similar profile. For regions 1 and 2 the water depths are 2 and 3 km, respectively, so the profiles for those regions end at those depths.

  • View in gallery

    (a) First mode F for nonzero ϕ (thick gray line) and first mode F for ϕ ≡ 0 (black dotted line) for region 1 closest to the coast (see Fig. 2). Also shown are the corresponding westward propagation speeds: γ = 1.34 cm s−1 for the ϕ nonzero case and γ0 = 1.48 cm s−1 for the ϕ ≡ 0 case. The solid inverted triangle symbol shown corresponds to region 1 in Fig. 2. (b)–(e) As in (a), but now for regions 2–5 in Fig. 2, respectively. The symbols corresponding to Fig. 2 are shown.

  • View in gallery

    The vertical advection coefficient (−gWT/∂z, black solid line) and alongshore advection coefficient [−gF(∂T/∂s)/(γf cosθ), gray line], for the temperature anomalies [see (3.3) and (3.4)] for (a)–(e) regions 1–5, respectively.

  • View in gallery

    Correlation between T ′ and η′ for each of the six regions (see Fig. 2) at each of the standard depths. Each region is coded by its Fig. 2 symbol. The black square in each panel corresponds to the depth of maximum correlation.

  • View in gallery

    The theoretical (black) and observed (gray) regression coefficients ν (°C m−1) [see (3.3)] for the temperature anomalies plotted as a function of depth for regions 1–5 in Fig. 2. The observed plot has solid circles when the corresponding correlation ≥0.5 and open circles otherwise. The theoretical coefficient υ was calculated using γ = 4 cm s−1 (Clarke and Dottori 2008), θ = 46°, f corresponding to 34°N, and F(z) taken from Fig. 7. The observed T ′ data had many gaps and, for the purpose of estimating the 95% confidence interval, we assumed that one degree of freedom occurred every four data points. On average, four data points covered a time interval of about 2 yr.

  • View in gallery

    (a) Monthly sea level anomalies (black curve) and monthly SST anomalies (gray curve) at San Diego. Both time series have been detrended and filtered with an 11-point symmetric Trenberth (1984) filter. The correlation coefficient between the two time series is 0.66, with rcrit(95%) = 0.34. (b) The monthly temperature time series (gray curve) and its interdecadal signal (black curve), the interdecadal signal being found by filtering with an 85-month running mean followed by a 57-month running mean.

  • View in gallery

    The vertical advection coefficient (−gWS/∂z, black line) and alongshore advection coefficient [−gF(∂S/∂s)/(γf cosθ), gray line], for the salinity anomalies [see (4.4) and (4.5)] for (a)–(e) regions 1–5, respectively.

  • View in gallery

    Correlation of the salinity anomaly S′ with the negative surface dynamic height anomaly −η′ at standard CalCOFI depths for the six hydrographic groupings according to their zonal distances from the coast (see Fig. 2). The depth of maximum correlation, marked by the solid square, increases with increasing distance from the coast (regions 1–6).

  • View in gallery

    The theoretical (black) and observed (gray) regression coefficients μ (psu m−1) [see (4.4) and (4.5)] plotted as a function of depth for regions 1–5 (see Fig. 2). The observed plot has solid circles when the corresponding correlation ≥0.5 and open circles otherwise. As for ν, the theoretical coefficient μ was calculated using γ = 4 cm s−1 (Clarke and Dottori 2008), θ = 46°, f corresponding to 34°N, and F(z) taken from Fig. 7. The error bars correspond to a 95% confidence interval. This interval was estimated in the same way as that described in Fig. 10.

  • View in gallery

    Salinity anomaly (solid dots) averaged over the top 100 m of region 1 of the CalCOFI region plotted with the coastal La Jolla surface salinity anomaly (solid curve) and the coastal La Jolla salinity anomaly at 5-m depth (dashed curve). The CalCOFI anomaly has been low-pass filtered as described in the text. The monthly coastal surface and 5-m depth anomalies have been filtered with a Trenberth (1984) 11-point symmetric filter (see caption to Fig. 3). The correlation between the CalCOFI and coastal surface anomalous salinity is r = 0.86 [rcrit(95%) = 0.60] and that between the surface salinity and the 5-m depth salinity is r = 0.86 [rcrit(95%) = 0.60].

  • View in gallery

    (a) Anomalous alongshore wind stress (heavy solid line), and surface (light line) and 5-m depth (dashed line) salinities at San Diego. The monthly anomalies have been filtered with the Trenberth (1984) filter. (b) Lagged correlation between the alongshore wind stress time series and the surface (solid line) and 5-m depth salinity time series (dashed line). The maximum correlation between wind stress and surface salinity is 0.51 when the wind stress leads by 4 months and that between the wind stress and 5-m salinity is 0.52 when the wind stress leads by 5 months. Both correlations are significantly different from zero at the 95% confidence level but not at the 99% level.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 24 24 3
PDF Downloads 16 16 3

Rossby Waves and the Interannual and Interdecadal Variability of Temperature and Salinity off California

View More View Less
  • 1 Department of Oceanography, The Florida State University, Tallahassee, Florida
© Get Permissions
Full access

Abstract

Previous work has shown that large-scale interannual Rossby waves, largely remotely generated by equatorial winds, propagate westward from the coast off southern California. These waves have a large-scale anomalous alongshore velocity field that is proportional to the time derivative of the interannual sea level anomaly. Using these results, a theory is developed for interannual perturbations to a mean density field that varies both vertically and alongshore, like that for the California Current region off southern California. Because both the anomalous vertical and alongshore currents are proportional to the time derivative of the interannual sea level, the theory suggests that the anomalous currents associated with the Rossby waves, acting on the mean temperature field, should induce temperature fluctuations proportional to the anomalous dynamic height. The alongshore and vertical advections contribute to the temperature fluctuations in the same sense, a higher-than-normal sea level, for example, resulting in downward and poleward displacement of warmer water and a local higher-than-normal temperature. Near the surface, alongshore advection dominates vertical advection but both contribute comparably near the thermocline and below. The correlation of observed temperature and dynamic height anomalies from the California Cooperative Oceanic Fisheries Investigation (CalCOFI) data is positive, which is consistent with the theory. The correlation is highest (r ≈ 0.8) near 100-m depth in the thermocline. Although the correlation falls toward the surface, it is still between 0.5 and 0.6, suggesting that the advection mechanism is a major contributor to the temperature anomalies there.

The anomalous Rossby wave currents, acting on the mean background salinity gradient, also induce salinity anomalies. At halocline depths of 100–200 m, consistent with the theory, the correlation of observed CalCOFI salinity and dynamic height anomalies is negative and large in magnitude (r ≈ −0.8). However, the surface salinity anomaly is not due to Rossby wave dynamics; instead, much of it is driven by the alongshore wind stress, which it lags by 4 months.

* Current affiliation: Marine Optics and Remote Sensing Laboratorie, Laboratoire d’Océanographie de Villefranche (LOV), Villefranche-sur-Mer, France.

Corresponding author address: Allan J. Clarke, Dept. of Oceanography, The Florida State University, 105 N. Woodward, P.O. Box 3064320, Tallahassee, FL 32306-4320. Email: clarke@ocean.fsu.edu

Abstract

Previous work has shown that large-scale interannual Rossby waves, largely remotely generated by equatorial winds, propagate westward from the coast off southern California. These waves have a large-scale anomalous alongshore velocity field that is proportional to the time derivative of the interannual sea level anomaly. Using these results, a theory is developed for interannual perturbations to a mean density field that varies both vertically and alongshore, like that for the California Current region off southern California. Because both the anomalous vertical and alongshore currents are proportional to the time derivative of the interannual sea level, the theory suggests that the anomalous currents associated with the Rossby waves, acting on the mean temperature field, should induce temperature fluctuations proportional to the anomalous dynamic height. The alongshore and vertical advections contribute to the temperature fluctuations in the same sense, a higher-than-normal sea level, for example, resulting in downward and poleward displacement of warmer water and a local higher-than-normal temperature. Near the surface, alongshore advection dominates vertical advection but both contribute comparably near the thermocline and below. The correlation of observed temperature and dynamic height anomalies from the California Cooperative Oceanic Fisheries Investigation (CalCOFI) data is positive, which is consistent with the theory. The correlation is highest (r ≈ 0.8) near 100-m depth in the thermocline. Although the correlation falls toward the surface, it is still between 0.5 and 0.6, suggesting that the advection mechanism is a major contributor to the temperature anomalies there.

The anomalous Rossby wave currents, acting on the mean background salinity gradient, also induce salinity anomalies. At halocline depths of 100–200 m, consistent with the theory, the correlation of observed CalCOFI salinity and dynamic height anomalies is negative and large in magnitude (r ≈ −0.8). However, the surface salinity anomaly is not due to Rossby wave dynamics; instead, much of it is driven by the alongshore wind stress, which it lags by 4 months.

* Current affiliation: Marine Optics and Remote Sensing Laboratorie, Laboratoire d’Océanographie de Villefranche (LOV), Villefranche-sur-Mer, France.

Corresponding author address: Allan J. Clarke, Dept. of Oceanography, The Florida State University, 105 N. Woodward, P.O. Box 3064320, Tallahassee, FL 32306-4320. Email: clarke@ocean.fsu.edu

1. Introduction

Much of the interannual and interdecadal sea level and sea surface temperature (SST) variability along the California coast is not due to the local alongshore wind stress forcing, but instead is due to an El Niño–La Niña signal of equatorial origin (Enfield and Allen 1980; Chelton and Davis 1982; Chelton et al. 1982; Kessler 1990; Clarke and Lebedev 1999; Clarke and Dottori 2008). This is illustrated in Fig. 1, where the low-frequency sea level and SST are correlated with each other and the El Niño index (Niño-3.4) but are uncorrelated with the alongshore component of the local California wind stress. Typically during El Niño events, the sea level and SST are higher than normal and the thermocline deepens, while during La Niña events, the sea level and SST are usually lower and the thermocline shoals. Why does the SST rise when the thermocline falls and fall when the thermocline rises? Is this solely due to vertical advection or are horizontal currents also important?

Although much of the California coast’s interannual sea level and SST seem to be remotely driven, the remote signal does not explain the low-frequency surface salinity since the latter is not well correlated with El Niño indices (Schneider et al. 2005). Chelton et al. (1982) suggested that both anomalous alongshore and vertical advection of salt may contribute to the low-frequency variability of salinity. Applying the alongshore advection hypothesis, Schneider et al. (2005) hypothesized that near-surface salinity anomalies at a hydrographic cross section off Long Beach, California, are due to an accumulation of salt caused by the advection of the main meridional salinity gradient by anomalies in the alongshore velocity. Schneider et al. (2005) calculated a time series of salinity using a model based on the alongshore advection of salt and an ad hoc dilution factor. However, the correlation coefficient between the model and the observed time series was a modest 0.41.

The main goal of this work is to understand the influence of the low-frequency (interannual and interdecadal) remote signal on the observed California low-frequency variability in temperature and, to some extent, in salinity. To this end, in section 2 we develop a theory for anomalous low-frequency perturbations to a mean density field that varies vertically and alongshore. This theory is related to the open-ocean theory of Killworth et al. (1997) but here we are concerned with the California Cooperative Oceanic Fisheries Investigation (CalCOFI) region and some near-boundary constraints are applied. After discussing the low-frequency currents and their connection to long, westward-propagating waves in section 2, we examine how the anomalous currents cause temperature anomalies in a sloping mean temperature field in section 3. In section 4 we analyze a model and its physics for the salinity anomalies, and in section 5 we compare the model predictions with observations. A concluding section 6 summarizes the main results and discusses the controversy over whether anomalous variability in the California Current system is locally or remotely forced.

2. Westward-propagating low-frequency waves and their currents

a. Waves and surface flows

Theory suggests that at interannual and lower frequencies, sea level at the California coast should propagate westward as long Rossby waves (Schopf et al. 1981; Cane and Moore 1981; Clarke 1983; Grimshaw and Allen 1988; Clarke and Shi 1991). Clarke and Dottori (2008) showed observationally that westward propagation from the coast does indeed occur. In their analysis they first grouped the 63 frequently sampled CalCOFI hydrographic stations off southern California into six bins according to their zonal distance from the coast (Fig. 2) to obtain six long (1949–2001) gappy records of dynamic height relative to 500 m. At each hydrographic station the temporal trend in dynamic height was removed. Then the 12-point annual cycle (average January, February, March, … , December dynamic height) was calculated and subtracted from the gappy, detrended record to obtain a record of the anomalous dynamic height. The lags at maximum correlation of these time series with the complete interannual sea level record at San Diego, plotted against the zonal distance from the coast, then gave the 4.10 cm s−1 westward propagation speed seen in Fig. 3. A similar calculation using the Ocean Topography Experiment (TOPEX)/Poseidon–Jason-1 interannual sea level data gave a similar westward propagation speed equal to 4.34 cm s−1. Clarke and Dottori found that this speed was about double that predicted by theory, even when the mean flow and bottom topographic variability were taken into account using the results of Killworth and Blundell (2003).

Following Clarke and Dottori (2008), the large-scale sea level anomaly westward propagation at speed γ ≈ 4.1 cm s−1 enables a simple expression for the large-scale anomalous alongshore velocity. For such a westward-propagating signal, the sea level anomaly η′ must be of the form
i1520-0485-39-10-2543-e21
where t is the time, x is the distance eastward, and G is some general function. In this section, a prime (′) associated with a variable means it is an anomaly. Differentiation of (2.1) with respect to x and t shows that
i1520-0485-39-10-2543-e22
Note that
i1520-0485-39-10-2543-e23
where i is the unit vector in the eastward direction, is the horizontal gradient operator, en is the unit vector in the direction of the n axis perpendicular to the coast, and es is the unit vector in the direction of the s axis along the coast (see Fig. 4). The interannual anomalous flow is approximately geostrophic and so the condition of no normal flow into the coast implies that ∂η′/∂s ≈ 0. Thus, in (2.3),
i1520-0485-39-10-2543-e24
where θ is the angle between i and en and also the angle that the coastline makes with due north (see Fig. 4). Therefore, from (2.4), (2.2) may be written
i1520-0485-39-10-2543-e25
The alongshore surface velocity anomaly is, by geostrophy,
i1520-0485-39-10-2543-e26
where g is the acceleration due to gravity, f is the Coriolis parameter, ρ* is the mean water density, and the anomalous pressure at the surface is
i1520-0485-39-10-2543-e27
Substituting for ∂η′/∂n in (2.6) by using (2.5) gives (Clarke and Dottori 2008)
i1520-0485-39-10-2543-e28

b. Low-frequency currents beneath the surface

Equation (2.8) enables us to estimate the anomalous alongshore velocity at the surface. But we will need anomalous alongshore and vertical velocities throughout the water depth. We will assume that the anomalous pressure is
i1520-0485-39-10-2543-e29
where the function F is unity at the surface since p′ is ρ*′ there. The function F is the first vertical mode of an appropriate eigenvalue problem to be discussed in the next subsection. By geostrophy, the anomalous alongshore velocity is, from (2.9),
i1520-0485-39-10-2543-e210
or, by (2.8),
i1520-0485-39-10-2543-e211
To obtain an expression for the anomalous vertical velocity w′, we begin with the incompressibility condition that following particles the density does not change. The anomaly version of this condition in terms of the anomalous density ρ′ linearized about the mean conditions is
i1520-0485-39-10-2543-e212
where the overbar refers to a time mean and u is the current component in the n direction perpendicular to the coast.
Equation (2.12) may be considerably simplified. From (2.9) and the hydrostatic balance,
i1520-0485-39-10-2543-e213
where
i1520-0485-39-10-2543-e214
and hence, by (2.5),
i1520-0485-39-10-2543-e215
From (2.15), we see that the second term on the left-hand side of (2.12) is negligible provided that |u| ≪ γ cosθ ≈ 3 cm s−1. Since the California Current has a speed of about 5 cm s−1 and is nearly parallel to the coast, |u| ≪ 3 cm s−1 and uρ′/∂n is negligible in (2.12). Also, observe that ∂ρ/∂n ∼ ∂ρ/∂s but, from low-frequency dynamics near the eastern ocean boundaries, |u′| ≪ |υ′| (see, e.g., Clarke and Shi 1991) so the third term on the left-hand side of (2.12) can be neglected when compared with the fifth term. Since also w′ ∼ w (see Plate 4 in Bograd et al. 2001) and |∂ρ′/∂z| ≪ |∂ρ/∂z|, wρ′/∂z is negligible compared with w′∂ρ/∂z. Finally, although ∼ 5 cm s−1 and υ′ ∼ 1 cm s−1, |∂ρ′/∂s| is of order 1% of ∂ρ/∂s and so ρ′/∂s is negligible compared with υ′∂ρ/∂s. Under these approximations, (2.12) reduces to
i1520-0485-39-10-2543-e216
Equation (2.16) may also be written as
i1520-0485-39-10-2543-e217
where
i1520-0485-39-10-2543-e218
is the slope that a constant density surface makes with the horizontal in a vertical/alongshore plane. Since this slope is very small, ϕ is an excellent approximation of the angle between the horizontal and the constant density surfaces in the vertical/alongshore plane. Equation (2.17) shows that the anomalous vertical velocity is partly due to particles sliding parallel to isopycnals (υϕ) and partly due to particles crossing isopycnals (−ρt/ρz). Substitution of (2.11) and (2.14) into (2.17) then gives
i1520-0485-39-10-2543-e219
where N is the buoyancy frequency defined by
i1520-0485-39-10-2543-e220

To summarize, (2.11) and (2.19) enable us to calculate the alongshore and vertical currents provided we know F(z). We will determine this function theoretically in the next subsection.

c. Estimation of the structure function F(z)

The curl of the horizontal momentum equations leads to the vorticity equation:
i1520-0485-39-10-2543-e221
where is the horizontal gradient operator, u the horizontal velocity, and ζ the relative vorticity. On the large spatial scales relevant here, ζ is negligible dynamically and so, with the help of the continuity and incompressibility condition
i1520-0485-39-10-2543-e222
(2.21) can be written
i1520-0485-39-10-2543-e223
The anomaly version of this equation is
i1520-0485-39-10-2543-e224
where j is a unit vector in the northward direction and β is the northward gradient of the Coriolis parameter. Since θ is the angle that the coastline makes with due north, (2.24) can be written
i1520-0485-39-10-2543-e225
or, since u′ ≪ υ′,
i1520-0485-39-10-2543-e226
Substitution into this equation using (2.11) and (2.19) leads to the following field equation for F(z):
i1520-0485-39-10-2543-e227
Since we are interested in the first baroclinic mode, the rigid-lid surface boundary condition w′ = 0 on z = 0 is valid and, from (2.19), this condition is
i1520-0485-39-10-2543-e228
In most of the CalCOFI region the bottom topography has negligible slope but near the shore in region 1 the bottom topography slopes steeply enough to affect the dynamics. The condition of nonnormal flow into the slowly varying bottom is
i1520-0485-39-10-2543-e229
Substitution of (2.11) and (2.19) into (2.29) gives
i1520-0485-39-10-2543-e230
Equations (2.27), (2.28), and (2.30) form an eigenvalue problem that can be solved numerically to find F(z) and the eigenvalue γ. Because the derivatives of N−2 and ϕ are not smooth, it is better to solve an equivalent eigenvalue problem in which N2 and ϕ are not differentiated. We write
i1520-0485-39-10-2543-e231
Then, (2.27) is
i1520-0485-39-10-2543-e232
Differentiating (2.32) gives
i1520-0485-39-10-2543-e233
and from (2.31) and (2.32)
i1520-0485-39-10-2543-e234
Substitution of (2.34) into (2.33) then yields the field equation
i1520-0485-39-10-2543-e235
From (2.28) and (2.31), the surface boundary condition is
i1520-0485-39-10-2543-e236
while from (2.30)(2.32) the bottom boundary condition is
i1520-0485-39-10-2543-e237

We solved (2.35)(2.37) numerically for the regions 1–5 shown in Fig. 2 using the isopycnal tilt angles shown in Fig. 5. Calculations were not done for region 6 because there were insufficient data to estimate ϕ. The square of the buoyancy frequency was calculated for each region and, as an example, values for region 3 are plotted in Fig. 6. The alongshore topography slope dH/ds only affected the results in region 1, so in regions 2–5 we set dH/ds = 0 while in region 1 we estimated dH/ds as 1.3 × 10−3. Once we had obtained W and γ for the first vertical mode in each region, we calculated F using (2.32). The eigenfunctions are determined up to an arbitrary multiplicative constant and we normalized F by setting F(0) = 1.

Figure 7 shows the first mode F and γ as well as the first vertical mode and eigenvalues when ϕ ≡ 0. Not surprisingly, the eigenfunctions and eigenvalues deviate most from the ϕ ≡ 0 case in regions 1 and 2 where ϕ is biggest. Farther from the coast, the isopycnal tilt hardly affects the solution. The Rossby wave speeds in the first two regions are lower because the water is shallower. In region 1, where dH/ds is nonzero, γ is increased from 1.05 to 1.34 cm s−1 by the alongshore bottom slope. Nowhere does the isopycnal tilt cause the Rossby wave speeds to increase to the 4 cm s−1 speed seen in the observations.

Based on the above results and the previous analysis of Clarke and Dottori (2008), neither the mean flow, variations in bottom topography, nor the mean alongshore isopycnal tilt can explain the quantitative discrepancy between the observed and theoretical Rossby wave propagation speed. One of the authors has derived a theory suggesting that the vigorous eddy field in the California Current may have a diffusive effect on the large-scale, low-frequency dynamics, leading to an increase in the theoretical propagation speed. But testing this possible explanation requires extensive calculations that are beyond the scope of the present analysis.

d. Particle displacements

When an anomalous flow advects warm southern water northward along the coast, the water in the north will be warmer than normal. In our analysis of temperature and salinity variability, it is therefore useful to consider anomalous particle displacement. The interannual and lower-frequency flow anomalies have a major influence on particle displacement because they flow consistently in one direction over long periods.

If we let Y′ be the anomalous alongshore particle displacement, then
i1520-0485-39-10-2543-e238
Substitution of (2.11) into the above equation and integrating once with respect to time gives (Clarke and Dottori 2008)
i1520-0485-39-10-2543-e239
that is, the anomalous alongshore particle displacement is northward for η′ > 0 and southward for η′ < 0. Note that the relationship (2.39) is crucially dependent on υ′ being proportional to ∂η′/∂t [see (2.11)], a relationship that holds because of westward Rossby wave propagation. At higher frequencies, the motion does not propagate westward but rather decays away from the coast (see, e.g., Clarke and Shi 1991) and υ′ is proportional to η′ rather than ∂η′/∂t.
Similarly to υ′ and Y′, we can define Z′ to be the upward vertical displacement so that
i1520-0485-39-10-2543-e240
and using (2.19) and an integration with respect to time gives
i1520-0485-39-10-2543-e241
From (2.31) the above relationship can also be written
i1520-0485-39-10-2543-e242
Since F(0) = 1, by (2.32) Wz > 0 near the surface. But W = 0 at the surface, so near the surface W < 0. This result and (2.42) imply that near enough to the surface, but beneath it, the vertical particle displacement is opposite to the sea level displacement. Note also from (2.42) that the parameter gW(z) is dimensionless and is a measure of how many times the vertical particle displacement in the interior of the water column is greater than the sea level displacement. Typical values of gW in the region of the thermocline where W is maximum are of the order of 100.

3. The temperature anomalies

Now that we have estimated the anomalous low-frequency velocities and corresponding particle displacements, we can examine how such anomalies act upon the mean sloping temperature field to induce low-frequency temperature fluctuations.

a. Theory

Beneath the surface layer, the temperature is approximately conserved. This leads to a linearized equation for anomalous temperature that has the same form as (2.12) for anomalous density. Similarly to our analysis for density, this equation may be further simplified to [cf. (2.16)]
i1520-0485-39-10-2543-e31
Substituting for υ′ and w′ using (2.38) and (2.40) and integrating once with respect to time gives
i1520-0485-39-10-2543-e32
or, from (2.39) and (2.42),
i1520-0485-39-10-2543-e33
with
i1520-0485-39-10-2543-e34
Since the temperature decreases poleward, (∂T/∂s) < 0 and the first term on the left-hand side of (3.4) is positive in the upper part of the water column where F > 0. Because the mean ocean temperature increases upward toward the surface and W < 0 in the upper ocean, the second term on the right-hand side of (3.4) is also positive. In other words, both the alongshore and vertical particle displacements contribute to positive ν in (3.3). Physically, when the sea level is (say) anomalously high, Y′ > 0 and Z′ < 0 [see (2.39) and (2.42)]; that is, particles are transported northward alongshore and downward. Thus, higher-temperature water in the south nearer the surface is displaced northward and downward, making the local temperature higher than normal. Opposite effects occur when η′ < 0 and lower-temperature northern water is transported southward and upward.

Equations (3.2)(3.4) show that the size of the alongshore contribution to the temperature anomaly is proportional to the first term on the right-hand side of (3.4) and the size of the vertical contribution to the second. Figure 8 shows that, for the five examined CalCOFI regions in Fig. 2, the alongshore advection contribution to the temperature anomaly dominates the vertical advection contribution in the top 50 m. At greater depths, the alongshore advection contribution is typically greater than, but comparable to, the vertical advection contribution.

So far we have discussed the temperature anomaly beneath the surface layer. In the surface layer we must take into account heat fluxes through its top and bottom. Consequently, temperature is no longer conserved and for a surface layer of depth h the temperature satisfies (see, e.g., Wang and McPhaden 1999)
i1520-0485-39-10-2543-e35
where ρ* is the (constant) water density, cp the specific heat at constant pressure, Q0 the net surface heat flux across the air–sea interface, and Qw the heat flux through the base of the surface layer. The net surface heat flux Q0 takes into account the incoming shortwave radiation, the outgoing longwave radiation, the latent heat loss due to evaporation, and the sensible heat flux. The heat through the base of the surface layer Qw is mainly due to the turbulent transfer of heat because at 100 m the surface layer is deep enough that no shortwave radiation penetrates through its bottom.
By similar arguments that led to (3.1), the simplified anomaly version of (3.5) is
i1520-0485-39-10-2543-e36
Substituting for υ′ from (2.11) and integrating with respect to time gives
i1520-0485-39-10-2543-e37

b. Testing the theory

At each standard depth we calculated, from the CalCOFI data, six time series of anomalous T ′ in a similar way to that done for dynamic height (see section 2a). We first tested (3.3) by correlating T ′ against η′. The correlation (see Fig. 9) is of the right sign and is high (about 0.8) for all six regions near 100-m depth. The correlation gradually falls with increasing depth, possibly because the observed temperature fluctuations are smaller there. The correlation also falls toward the surface but even in the top 50 m it still averages about 0.6 in regions 1–4. This suggests that the advection term, which is proportional to η′ in (3.7), is an important contributor to the temperature anomaly in the upper 100 m. The fall in correlation may indicate that the net heat flux anomaly [second term on the right-hand side of (3.7)] is contributing to the interannual surface layer anomalous temperature. However, in contrast with some previous work (e.g., Di Lorenzo et al. 2005), we do not think that it is a dominant contributor.

If indeed anomalous advection is a major contributor to low-frequency temperature anomalies, then not only should η′ and T ′ be highly correlated, but the temperature fluctuations predicted should also be about the right size. In other words, the theoretical regression coefficient ν [see (3.3)] should be about the same size as that found observationally. Figure 10 shows that while the theoretical regression coefficients ν often do not fall inside the error bars of the corresponding observationally determined regression coefficients, nevertheless, for the most part the theoretical coefficients have similar spatially varying structures and sizes to those estimated from the data. This is further evidence that alongshore and vertical advection by anomalous flow are major contributors to the upper-ocean large-scale, low-frequency temperature fluctuations in the CalCOFI region.

Figure 11b shows that while interannual variability dominates T ′, there is also some interdecadal variability. The analysis of Clarke and Lebedev (1999) suggests that the interdecadal variability in T ′ is not due to local net heat flux anomalies because the California coastal interdecadal SST and sea level signals are highly correlated (r = 0.88) and the interdecadal sea level signal is largely generated along the equator.

Since the sea surface temperature and dynamic height anomalies are positively correlated across the CalCOFI region and, in particular, in region 1 nearest the coast, we would expect that monthly coastal sea level and coastal sea surface temperature anomalies to also be positively correlated and they are (see Fig. 1 and also the longer record in Fig. 11a). The anomalous coastal temperature records also show that the low-frequency variability is predominantly interannual rather than interdecadal (Fig. 11b). In the next section we will see that this is not the case for anomalous surface salinity.

4. Salinity anomalies

a. Low-frequency salinity beneath a surface layer

Analysis of observed S′ by Schneider et al. (2005) suggests (see their Fig. 10) that salinity anomalies beneath a 50–100-m-thick surface layer are mainly due to movements of the halocline whereas salinity anomalies in the surface layer occur independent of density anomalies. This indicates that the salinity anomalies in the surface layer originate from mechanisms different from those beneath the surface layer and we therefore will consider these regions separately.

Beneath the surface layer we expect the salinity S to be conserved following the flow, so there
i1520-0485-39-10-2543-e41
Linearizing this equation for low-frequency perturbations about the mean and simplifying gives, similar to Eqs. (2.16) and (3.1),
i1520-0485-39-10-2543-e42
Similarly to the temperature anomaly case, substituting for υ′ and w′ using (2.38) and (2.40) and integrating once with respect to time gives
i1520-0485-39-10-2543-e43
or, from (2.39) and (2.42),
i1520-0485-39-10-2543-e44
with
i1520-0485-39-10-2543-e45
Thus, beneath about 100-m depth, where (4.2) is valid, the local sea level and salinity anomalies should be highly linearly correlated with a z-dependent regression coefficient μ given by (4.5). We will check this prediction with our salinity observations in section 5.

Equation (4.3) shows that the salinity anomaly S′ can be regarded as the sum of a salinity anomaly due to an alongshore displacement (−Y′∂S/∂s) and a salinity anomaly due to a vertical displacement (−Z′∂S/∂z). Figure 12 shows that alongshore displacement dominates near the surface and that alongshore and vertical contributions tend to be small and of opposite sign in water deeper than 150 m.

b. Low-frequency salinity in the surface layer

In the surface layer salinity can be changed, not only by alongshore and vertical advection, but also by the addition or removal of freshwater at the surface and subsequent mixing. However, Schneider et al. (2005) estimated that freshwater fluxes due to evaporation, precipitation, and river runoff are too small to account for the observed changes in surface salinity. Salinity anomalies could also in theory arise due to anomalous Ekman pumping via the wind stress curl, but Chhak et al. (2009) recently showed that this contribution is negligible. However, they suggest that anomalous coastal upwelling, driven by the anomalous Ekman transport perpendicular to the coast, has an influence on the surface salinity anomalies. The upwelling-favorable offshore Ekman transport, proportional to τs, the equatorward component of the alongshore wind stress, will tend to increase the salinity of the surface layer since mean salinity increases with depth. This effect can be included in the model of the previous section by including a term proportional to τs on the right-hand side of (4.2):
i1520-0485-39-10-2543-e46
where a is a positive constant.
In our analysis of salinity anomalies in the next section we will find that the salinity anomaly is dissipated in the surface layer over a time scale Δt. We will take this into account by adding a dissipation term S′/Δt to (4.6) so that it becomes
i1520-0485-39-10-2543-e47
or, using (2.38) and (2.40),
i1520-0485-39-10-2543-e48

We will test our theoretical predictions of S′ beneath the surface layer and close to the surface in the next section.

5. Observed and model salinity anomalies

a. Testing the Rossby wave advection theory

At each standard depth we calculated, from CalCOFI data, six time series of anomalous salinity S′ in a similar way to that done for η′ except that, since S′ did not have a significant trend in time, no trend was removed. According to (4.4), correlation of each S′ time series with the corresponding η′ time series should be high with the regression coefficient μ given by (4.5). Plots of the correlation of S′ and −η′ with depth for each of the six zonal distances from the coast (Fig. 13) show two regions of low correlation separated by a region of high correlation. The maximum correlation is about 0.8 and deepens with increasing distance from the coast; it is at 100 m deep for regions 1–3, at 150 m deep for regions 4 and 5, and at 200 m deep for region 6. The high correlations coincide approximately with the depth of the halocline, which increases in depth with distance from the coast. Figure 14 shows that for the most part when the correlation is high, the theoretical regression coefficient μ [see (4.4) and (4.5)] has the same sign and is of similar size to that observed.

b. Testing the theory for low-frequency salinity near the surface

The correlation in Fig. 13 falls near the surface, suggesting that (4.4) and advection by anomalous Rossby wave currents do not describe the dominant process there. This being so, we will drop the two advection terms on the right-hand side of (4.8) so that, after multiplying by Δt, it reduces to
i1520-0485-39-10-2543-e51
By Taylor series expansion,
i1520-0485-39-10-2543-e52
so, within an error of order 0.5(Δt)2(∂2S′/∂t2)/S′(t), (5.1) may be written
i1520-0485-39-10-2543-e53
At frequency ω,
i1520-0485-39-10-2543-e54
so we expect (5.3) to hold for frequencies ω low enough that 0.5ω2t)2 ≪ 1.

Equation (5.3) suggests that S′ should be well correlated with τs and lag it by some time Δt. We tested this with representative California Current salinity anomaly and alongshore wind stress anomaly time series (to be described below).

Two long records of surface and 5-m depth salinities at the coast at La Jolla, California, were provided by the Stephen Birch Aquarium–Museum at the Scripps Institute of Oceanography. Analysis suggests that this coastal salinity anomaly is representative of the salinity anomaly offshore. Specifically, Schneider et al. (2005) analyzed the salinity anomaly along line 90, an 800-km hydrographic line perpendicular to the coast that begins about 100 km north along the coast from La Jolla. The first mode of their empirical orthogonal function analysis, describing 38% of the variance, had a single-signed structure that extended over the whole 800 km.

We confirmed that the coastal salinity anomaly is representative of the salinity anomaly offshore using salinity data from our region 1 (see Fig. 2). We chose to analyze region 1 because data are sparse and region 1 easily has the most data. Even though region 1 has the most data, the salinity anomalies were still very noisy so we filtered in time using an 11-month running mean. Usually, many points in the 11-month window were missing, but we still recorded a value for a given month by averaging over the data present provided at least 3 months had data. The high correlation of the surface coastal salinity and the coastal 5-m depth salinities with the gappy region 1 CalCOFI upper-ocean salinity (Fig. 15) suggests that the gappy region 1 CalCOFI signal is approximately valid and that the coastal salinity measurements can be used to monitor variations of average salinity anomalies seaward of the coast.

Figure 16 compares the low-frequency La Jolla surface and 5-m depth coastal salinity anomalies with the low-frequency alongshore wind stress anomaly estimated from daily data available from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR; Kalnay et al. 1996). The two time series are maximally correlated with the anomalous wind stress when the wind stress leads the surface salinity by 4 months (r = 0.51) and the 5-m salinity by 5 months (r = 0.52). Both correlations are significantly different from zero at the 95% level but not at the 99% level. These results are consistent with (5.3) with Δt = 4 or 5 months. Note that the dominant periodicity in the time series is decadal, so with ω = 2π/10 yrs and Δt = 5 months, the criterion 0.5ω2t)2 = 0.03 ≪ 1 is satisfied. Although the maximum correlations occur at 4 or 5 months’ lead, Fig. 16b shows that the maximum is not sharply peaked but occurs in a broad plateau of about 3–13 months. In fact, if we (say) remove the first 20 yr of each record and repeat the correlation calculations, then for both the surface and 5-m salinities, the maximum correlation increases to 0.64 with the wind stress now leading by 12 months. As before, this maximum occurs in a similar broad region of high lag correlation and is significantly different from zero at the 95% but not 99% confidence levels. With Δt = 12 months, 0.5ω2Δt2 = 0.20 is still small.

We conclude that much of the surface anomalous salinity variability in the California Current region is driven by local alongshore wind stress anomalies. It is therefore dynamically very different from the salinity anomalies beneath the surface layer, which the evidence in section 5a suggests are largely due to anomalous geostrophic remotely driven Rossby wave currents advecting mean salinity gradients.

6. Conclusions

Theory and analysis of the CalCOFI and coastal data lead us to the following main conclusions.

  • (i) Low-frequency anomalous upper-ocean temperature is mainly due to alongshore and vertical advection of the mean temperature gradient by the anomalous flow.
  • (ii) The large-scale anomalous flow is associated with westward-propagating, large-scale Rossby waves. These waves are associated with coastal sea level anomalies, which are largely driven remotely at the equator.
  • (iii) The westward propagation is key to understanding why anomalous alongshore advection contributes temperature anomalies that are proportional to the dynamic height anomalies. Vertical advection also predicts that the temperature anomalies should be proportional to the dynamic height anomalies. Both alongshore and vertical advections are such that dynamic height and temperature anomalies should be positively correlated; a positive dynamic height anomaly, for example, results in downward and poleward displacement of warmer water and a local higher-than-normal temperature.
  • (iv) Temperature anomalies are mainly interannual and even near the surface seem to be due more to alongshore advection than anomalous heat flux.
  • (v) Salinity anomalies are also generated by the anomalous Rossby wave flow, especially in the halocline region between about 100- and 200-m depths.
  • (vi) Surface salinity anomalies are not due to the anomalous Rossby wave flow. Rather, much of the anomalous variability is due to anomalous local alongshore wind stress, which drives a surface Ekman transport perpendicular to the coast.
  • (vii) The easily measured coastal surface salinity can be used to monitor the large-scale salinity anomalies offshore in the California Current.

Some of the above conclusions disagree with related previous California Current work, particularly that of Di Lorenzo et al. (2008), and reviewers of our manuscript asked that we discuss this. For some time there has been a debate about the relative importance of remotely and locally generated low-frequency variabilities in the California Current. As mentioned in the introduction (see also Fig. 1), early work by Enfield and Allen (1980) and Chelton and Davis (1982) showed that the coastal interannual sea level is largely of equatorial origin all along the eastern boundary of the Pacific. Clarke and Lebedev (1999) obtained a similar result for decadal variability. This remote signal is not limited to within a few tens of kilometers of the California coast as independent hydrographic and satellite altimeter observations show that the signal can be seen propagating right across the CalCOFI region (Fig. 3) and that the amplitude of the propagating signal decreases only slightly with zonal distance over almost all of the CalCOFI region (see Fig. 5 in Clarke and Dottori 2008). We argue that this large-scale low-frequency propagating signal is fundamental to the low-frequency California Current dynamics and should not be excluded as in some recent numerical models (e.g., Di Lorenzo et al. 2008) of the region.

Regarding the contribution of local wind forcing to the geostrophic flow, the fact that the large-scale dynamic height–sea level signal has a highly correlated lag with zonal distance from the coast (see Fig. 3) and that this signal changes only slightly in amplitude over 500 km (see Fig. 5 in Clarke and Dottori 2008) suggests that this signal is not significantly influenced by the wind stress or the wind stress curl. We checked this by calculating empirical orthogonal functions (EOFs) of these quantities using wind stress data from the NCEP–NCAR reanalysis dataset (Kalnay et al. 1996) and lag correlating the principal components with both San Diego sea level and the Niño-3.4 index. Tables 1 and 2 report our analysis of the wind stress curl anomalies over the CalCOFI region and the coastal alongshore wind stress anomalies within 350 km of the coast from about 32.5° to 37.5°N. The correlations are low everywhere except perhaps for the EOF mode 3 of the alongshore wind stress with Niño-3.4 (r = 0.46 at 1-month lag), but this mode describes only 1.44% of the variance of the alongshore wind stress anomalies. The conclusion that local wind stress forcing does not contribute substantially to the low-frequency dynamics near the California coast south of San Francisco was originally given by Enfield and Allen (1980).

While local wind forcing does not apparently contribute significantly to the large-scale low-frequency sea level, the dynamic height, temperature, and geostrophic flow variability, nor the salinity beneath the surface layer, it does contribute to the salinity variability in the surface layer. The latter is apparent from Fig. 16 and previous work by Di Lorenzo et al. (2008; see the fifth plot in their Fig. 1). Local forcing appears to not only affect the surface salinity anomalies, but also the anomalies in nitrate and chlorophyll. This follows from the model results of Di Lorenzo et al. (2008). Their model excludes the remote ocean signal from the equator and yet they obtain significant correlations between model- and observationally estimated CalCOFI nitrate (r = 0.55) and chlorophyll a (r = 0.5). Di Lorenzo et al. (2008) point out that fluctuations in these variables are related to fluctuations in a climate mode they term the North Pacific Gyre Oscillation (NPGO). This oscillation is defined as the second empirical orthogonal function of the SST or sea surface height over the northeast Pacific region 25°–62°N, 180°–110°W. The CalCOFI region is a small southeastern part of this region, and dynamically we expect, as do Di Lorenzo et al., that the connection between the NPGO and local CalCOFI nutrients and chlorophyll is due to the fact that the local alongshore wind stress forcing is correlated with the large-scale forcing that drives the NPGO.

However, we should point out that the local alongshore wind stress variability and the related NPGO are not the primary drivers of variations in the CalCOFI ecosystem. If they were, then the zooplankton population would be strongly related to variations in the alongshore wind stress rather than, as pointed out by Chelton et al. (1982) and more recently by Clarke and Dottori (2008), large-scale variations in alongshore geostrophic flow. Specifically, the correlation of the anomaly of the logarithm of the zooplankton concentration averaged over the CalCOFI region with the upwelling-favorable alongshore wind stress time series τs is maximum when the wind stress leads by 1 month; it is, however, only 0.28, which is not significantly different from zero at the 90% level. On the other hand, as shown by Clarke and Dottori (2008), the corresponding correlation with dynamic height averaged over the CalCOFI region is r = −0.67, which is significantly different from zero at the 99% level. Note that correlating dynamic height with zooplankton abundance is consistent with advection by anomalous alongshore Rossby wave currents (Clarke and Dottori 2008).

To summarize, local alongshore wind stress, which is linked to the NPGO, appears to be mainly responsible for the low-frequency variation of the CalCOFI surface salinity, nutrients, and chlorophyll a, while low-frequency variations of CalCOFI sea level/dynamic height, temperature, subsurface salinity, and zooplankton concentration result mainly from a remotely driven equatorial signal and westward Rossby wave propagation from the California coast.

Acknowledgments

We gratefully acknowledge the support of the National Science Foundation (OCE-0220563 and ATM-06233402) and the Brazilian National Council for the Development of Science and Technology (CNPq-200132/2001-6). We also thank the reviewers for their extensive comments on our manuscript.

REFERENCES

  • Bograd, S. J., , T. K. Chereskin, , and D. Roemmich, 2001: Transport of mass, heat, salt and nutrients in the southern California Current System: Annual cycle and interannual variability. J. Geophys. Res., 106 , (C5). 92559275.

    • Search Google Scholar
    • Export Citation
  • Cane, M. A., , and D. W. Moore, 1981: A note on low-frequency equatorial basin modes. J. Phys. Oceanogr., 11 , 15781584.

  • Chelton, D. B., , and R. E. Davis, 1982: Monthly mean sea level variability along the west coast of North America. J. Phys. Oceanogr., 12 , 757784.

    • Search Google Scholar
    • Export Citation
  • Chelton, D. B., , P. A. Bernal, , and J. A. McGowan, 1982: Large-scale interannual physical and biological interaction in the California Current. J. Mar. Res., 40 , 10951125.

    • Search Google Scholar
    • Export Citation
  • Chhak, K. C., , E. Di Lorenzo, , N. Schneider, , and P. F. Cummins, 2009: Forcing of low-frequency ocean variability in the northeast Pacific. J. Climate, 22 , 12551276.

    • Search Google Scholar
    • Export Citation
  • Clarke, A. J., 1983: The reflection of equatorial waves from oceanic boundaries. J. Phys. Oceanogr., 13 , 11931207.

  • Clarke, A. J., , and C. Shi, 1991: Critical frequencies at ocean boundaries. J. Geophys. Res., 96 , 1073110738.

  • Clarke, A. J., , and A. Lebedev, 1999: Remotely driven decadal and longer changes in the coastal Pacific waters of the Americas. J. Phys. Oceanogr., 29 , 828835.

    • Search Google Scholar
    • Export Citation
  • Clarke, A. J., , and M. Dottori, 2008: Planetary wave propagation off California and its effect on zooplankton. J. Phys. Oceanogr., 38 , 702714.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., , A. J. Miller, , N. Schneider, , and J. C. McWilliams, 2005: The warming of the California Current: Dynamics and ecosystem implications. J. Phys. Oceanogr., 35 , 336362.

    • Search Google Scholar
    • Export Citation
  • Di Lorenzo, E., and Coauthors, 2008: North Pacific Gyre Oscillation links ocean climate and ecosystem change. Geophys. Res. Lett., 35 , L08607. doi:10.1029/2007GL032838.

    • Search Google Scholar
    • Export Citation
  • Ebisuzaki, W., 1997: A method to estimate the statistical significance of a correlation when the data are serially correlated. J. Climate, 10 , 21472153.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., , and J. S. Allen, 1980: On the structure and dynamics of monthly mean sea level anomalies along the Pacific coast of North and South America. J. Phys. Oceanogr., 10 , 557578.

    • Search Google Scholar
    • Export Citation
  • Grimshaw, R., , and J. S. Allen, 1988: Low-frequency baroclinic waves off coastal boundaries. J. Phys. Oceanogr., 18 , 11241143.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kessler, W. S., 1990: Observations of long Rossby waves in the northern tropical Pacific. J. Geophys. Res., 95 , 51835217.

  • Killworth, P. D., , and J. R. Blundell, 2003: Long extratropical planetary wave propagation in the presence of slowly varying mean flow and bottom topography. Part I: The local problem. J. Phys. Oceanogr., 33 , 784801.

    • Search Google Scholar
    • Export Citation
  • Killworth, P. D., , D. B. Chelton, , and R. A. de Szoeke, 1997: The speed of observed and theoretical long extratropical planetary waves. J. Phys. Oceanogr., 27 , 19461966.

    • Search Google Scholar
    • Export Citation
  • Schneider, N., , E. Di Lorenzo, , and P. P. Niiler, 2005: Salinity variations in the southern California Current. J. Phys. Oceanogr., 35 , 14211436.

    • Search Google Scholar
    • Export Citation
  • Schopf, P. S., , D. L. T. Anderson, , and R. Smith, 1981: Beta-dispersion of low-frequency Rossby waves. Dyn. Atmos. Oceans, 5 , 187214.

  • Trenberth, K. E., 1984: Signal versus noise in the Southern Oscillation. Mon. Wea. Rev., 112 , 326332.

  • Wang, W., , and M. J. McPhaden, 1999: The surface heat balance in the equatorial Pacific Ocean. Part I: Mean seasonal cycle. J. Phys. Oceanogr., 29 , 18121831.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

(a) Monthly sea level anomalies at San Diego (gray line) and the monthly Niño-3.4 index (black line). Niño-3.4 is the SST anomaly averaged over the east-central equatorial Pacific region 5°S–5°N, 170°–120°W. Both time series have been detrended and filtered with the interannual filter of Trenberth (1984). This 11-point symmetric filter passes more than 80% of the amplitude at a period of 24 months and longer and less than 10% of the amplitude for periods shorter than 8 months. Both time series have also been normalized by their standard deviations (σ = 37 mm for sea level and 0.70°C for Niño-3.4). The maximum correlation between the time series occurs at zero lag and is r = 0.70, significant at the 99% level. Here and elsewhere, the critical correlation coefficient was determined using the method of Ebisuzaki (1997). (b) Same as in (a), but for the San Diego sea level (gray line) and alongshore wind stress near San Diego (black line). In this case, the highest correlation (r = 0.20) at zero lag is insignificant statistically. The standard deviation of the alongshore wind stress is 0.0041 Pa. (c) Same as in (a), but for the San Diego sea level (gray line), the 5-m San Diego temperature (black line), and the San Diego surface temperature (dashed line). The highest correlation (r = 0.72) between the sea level and 5-m temperature occurs when sea level leads temperature by 1 month, but the correlation is nearly the same (r = 0.71) at zero lag. The corresponding correlations for the surface temperature are r = 0.70 and r = 0.69, respectively. All correlations are significant at the 99% level. The standard deviation of the surface temperature is 0.66°C and that of the 5-m temperature is 0.68°C.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 2.
Fig. 2.

The CalCOFI region off of southern California showing the TOPEX/Poseidon–Jason-1 satellite tracks and 63 frequently sampled CalCOFI stations. The stations are grouped into six regions according to their zonal distance from the coast in bins of approximately 130 km. Specifically, the solid inverted triangles (region 1) represent all stations within 130 km of the coast, shaded circles (region 2) show all stations between 130 and 260 km from the coast, solid hexagons (region 3) are for stations 260–390 km from the coast, etc. (from Clarke and Dottori 2008).

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 3.
Fig. 3.

Estimates of westward sea level propagation off California using dynamic height anomalies relative to 500 m from CalCOFI hydrographic data (solid regression line, solid diamond) and from TOPEX/Poseidon–Jason-1 satellite-estimated sea level anomalies (dashed regression line, solid circles). The plotted lags for the dynamic height were found by lag correlating the gappy dynamic height data for each of the six regions in Fig. 2 with the monthly sea level anomaly record at San Diego and recording the lags at maximum correlation. Before the correlation calculations, the San Diego sea level anomalies were detrended and low-pass filtered using the Trenberth (1984) 11-point symmetric filter to obtain interannual and lower-frequency sea levels. This filter passes more than 80% of the amplitude at a period of 24 months and longer and less than 10% of the amplitude for periods shorter than 8 months. The lag correlation calculations were also carried out on the satellite data in a similar way, first binning the along-track data into the six regions to form the six monthly time series. [Redrawn from Clarke and Dottori (2008)]

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 4.
Fig. 4.

Normal and tangential coordinates (n, s) to a coastline making an angle θ with due north (from Clarke and Dottori 2008).

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 5.
Fig. 5.

One hundred thousand times the isopycnal tilt angle (rad) as a function of depth for the top 500 m for each of the five regions closest to the coast in Fig. 2. Below 500 m, the slope is negligible. Positive ϕ corresponds to isopycnals tilting upward toward the north. Region 6 was not included because there were too few grid points alongshore to estimate ϕ. The symbols represent the regions in Fig. 2.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 6.
Fig. 6.

Buoyancy frequency squared (s−2) multiplied by 10 000 for region 3 (see Fig. 2). The other regions have a very similar profile. For regions 1 and 2 the water depths are 2 and 3 km, respectively, so the profiles for those regions end at those depths.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 7.
Fig. 7.

(a) First mode F for nonzero ϕ (thick gray line) and first mode F for ϕ ≡ 0 (black dotted line) for region 1 closest to the coast (see Fig. 2). Also shown are the corresponding westward propagation speeds: γ = 1.34 cm s−1 for the ϕ nonzero case and γ0 = 1.48 cm s−1 for the ϕ ≡ 0 case. The solid inverted triangle symbol shown corresponds to region 1 in Fig. 2. (b)–(e) As in (a), but now for regions 2–5 in Fig. 2, respectively. The symbols corresponding to Fig. 2 are shown.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 8.
Fig. 8.

The vertical advection coefficient (−gWT/∂z, black solid line) and alongshore advection coefficient [−gF(∂T/∂s)/(γf cosθ), gray line], for the temperature anomalies [see (3.3) and (3.4)] for (a)–(e) regions 1–5, respectively.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 9.
Fig. 9.

Correlation between T ′ and η′ for each of the six regions (see Fig. 2) at each of the standard depths. Each region is coded by its Fig. 2 symbol. The black square in each panel corresponds to the depth of maximum correlation.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 10.
Fig. 10.

The theoretical (black) and observed (gray) regression coefficients ν (°C m−1) [see (3.3)] for the temperature anomalies plotted as a function of depth for regions 1–5 in Fig. 2. The observed plot has solid circles when the corresponding correlation ≥0.5 and open circles otherwise. The theoretical coefficient υ was calculated using γ = 4 cm s−1 (Clarke and Dottori 2008), θ = 46°, f corresponding to 34°N, and F(z) taken from Fig. 7. The observed T ′ data had many gaps and, for the purpose of estimating the 95% confidence interval, we assumed that one degree of freedom occurred every four data points. On average, four data points covered a time interval of about 2 yr.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 11.
Fig. 11.

(a) Monthly sea level anomalies (black curve) and monthly SST anomalies (gray curve) at San Diego. Both time series have been detrended and filtered with an 11-point symmetric Trenberth (1984) filter. The correlation coefficient between the two time series is 0.66, with rcrit(95%) = 0.34. (b) The monthly temperature time series (gray curve) and its interdecadal signal (black curve), the interdecadal signal being found by filtering with an 85-month running mean followed by a 57-month running mean.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 12.
Fig. 12.

The vertical advection coefficient (−gWS/∂z, black line) and alongshore advection coefficient [−gF(∂S/∂s)/(γf cosθ), gray line], for the salinity anomalies [see (4.4) and (4.5)] for (a)–(e) regions 1–5, respectively.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 13.
Fig. 13.

Correlation of the salinity anomaly S′ with the negative surface dynamic height anomaly −η′ at standard CalCOFI depths for the six hydrographic groupings according to their zonal distances from the coast (see Fig. 2). The depth of maximum correlation, marked by the solid square, increases with increasing distance from the coast (regions 1–6).

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 14.
Fig. 14.

The theoretical (black) and observed (gray) regression coefficients μ (psu m−1) [see (4.4) and (4.5)] plotted as a function of depth for regions 1–5 (see Fig. 2). The observed plot has solid circles when the corresponding correlation ≥0.5 and open circles otherwise. As for ν, the theoretical coefficient μ was calculated using γ = 4 cm s−1 (Clarke and Dottori 2008), θ = 46°, f corresponding to 34°N, and F(z) taken from Fig. 7. The error bars correspond to a 95% confidence interval. This interval was estimated in the same way as that described in Fig. 10.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 15.
Fig. 15.

Salinity anomaly (solid dots) averaged over the top 100 m of region 1 of the CalCOFI region plotted with the coastal La Jolla surface salinity anomaly (solid curve) and the coastal La Jolla salinity anomaly at 5-m depth (dashed curve). The CalCOFI anomaly has been low-pass filtered as described in the text. The monthly coastal surface and 5-m depth anomalies have been filtered with a Trenberth (1984) 11-point symmetric filter (see caption to Fig. 3). The correlation between the CalCOFI and coastal surface anomalous salinity is r = 0.86 [rcrit(95%) = 0.60] and that between the surface salinity and the 5-m depth salinity is r = 0.86 [rcrit(95%) = 0.60].

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Fig. 16.
Fig. 16.

(a) Anomalous alongshore wind stress (heavy solid line), and surface (light line) and 5-m depth (dashed line) salinities at San Diego. The monthly anomalies have been filtered with the Trenberth (1984) filter. (b) Lagged correlation between the alongshore wind stress time series and the surface (solid line) and 5-m depth salinity time series (dashed line). The maximum correlation between wind stress and surface salinity is 0.51 when the wind stress leads by 4 months and that between the wind stress and 5-m salinity is 0.52 when the wind stress leads by 5 months. Both correlations are significantly different from zero at the 95% confidence level but not at the 99% level.

Citation: Journal of Physical Oceanography 39, 10; 10.1175/2009JPO3898.1

Table 1.

EOF modes for the anomalous CalCOFI alongshore wind stress. Column 3 lists the maximum (in magnitude) correlation with the corresponding lag (in months) for the given EOF principal component with the Niño-3.4 anomaly index. Column 4 is similar except for San Diego sea level anomalies.

Table 1.
Table 2.

As in Table 1, but for anomalous CalCOFI wind stress curl.

Table 2.
Save