Submitted to Geophysical Research Letters (January 2002)
Dynamics of Seasonal and Interannual Variability of the
Peru-Chile Undercurrent
Oscar Pizarro
Departamento de Física del Océano y de la Atmósfera and Programa Regional de Oceanografía
Física y Clima, Universidad de Concepción, Chile.
Gary Shaffer
Danish Center for Earth System Science, Niels Bohr Institute for Astronomy, Physics and
Geophysics University of Copenhagen, Denmark.
Boris Dewitte
Laboratoire d'Etudes en Géophysique et Océanographie Spatiale, LEGOS/IRD Toulouse, France.
Marcel Ramos
Programa Regional de Oceanografía Física y Clima, Universidad de Concepción, Chile.
Abstract. Coastal sea level, current, satellite altimeter data and an equatorial Kelvin wave
model are used to study the dynamics of seasonal and interannual fluctuations of the Peru-Chile
Undercurrent. Satellite altimetry shows that low frequency, sea level anomalies near 30°S
propagate offshore as expected from Rossby wave theory. Those anomalies are related to
disturbances of equatorial origin observed near the coast of South America. With a wind-forced
equatorial Kelvin wave model and simple Rossby wave dynamics we show that much of the sea
level anomalies, and more than 50% of the observed undercurrent variability near 30ºS, at
seasonal and interannual periods can be explained by long Rossby waves forced by the
equatorial Kelvin waves. The amplitude of the equatorial Kelvin wave can be used to estimate
the amplitude of the Rossby wave near the South American coast, and thus, to monitor low
frequency changes in the Peru-Chile Undercurrent.
1. Introduction
The Peru-Chile Undercurrent (PCU) is a well-defined, poleward, subsurface flow, which
extends over the continental shelf and slope off the west coast of South America. The PCU
transports warm, salty Equatorial Subsurface Water from the eastern tropical Pacific to at least as
far south as 48°S [Silva and Neshiba, 1979]. This water mass in turn is the main source for the
coastal upwelling off Peru and northern Chile [e.g. Huyer et al., 1987] and is associated with a
well-developed oxygen minimum [Wyrtki, 1963]. A mean poleward flow in the PCU of 5-10 cm
s-1 was observed between 5°S and 12°S off Peru [Huyer et al. 1991]. At 30°S off Chile, a
southward mean flow of 12.8 cm s-1 was observed for a six year period near the PCU core over
the slope [Shaffer et al, 1999]. Like other poleward undercurrents of Eastern Boundary Current
systems, the PCU may be forced by upwelling-favorable, alongshore winds [McCreary, 1981;
McCreary, 1987]. However, seasonal and interannual changes in the PCU off Chile and Peru are
not clearly related to these winds [Huyer, 1991; Pizarro et al., 2001]. Near 30ºS, the PCU
exhibits large semiannual variability with poleward maxima during April and November, while
the alongshore windstress has a marked annual oscillation with only one equatorward maximum
during November [Shaffer et al. 1999]. At low frequency, sea level and current variability along
the east coast of South America is strongly modulated by disturbances originated at the equator
[e.g. Enfield and Allen 1981, Shaffer et al 1997]. At interannual periods, fluctuations of
alongshore flow off Central Chile reflect seaward-propagating, Rossby waves forced by
equatorial disturbances rather than alongshore windstress fluctuations [Pizarro et al. 2001]. In the
energetic annual and semiannual frequency bands the effects of the equatorial forcing on the
PCU is unknown, while most of the sea level and hydrographic variability has traditionally be
ascribed to the seasonal cycle of the alongshore windstress [e.g Blanco et al 2001]. Here we
consider sea level data along the western coast of South America, current observations near
30°S, a wind-forced equatorial Kelvin wave model and mid-latitude Rossby wave theory, and
find that much seasonal variability of the PCU near 30°S is also related to equatorial variability.
2. Data and Methods
Sea level heights (SLH) from October 1992 to December 2000 were derived from a
combination of TOPEX/ POSEIDON and ERS altimeters (but from January 1994 to February
1995 only T/P data was available).
A detailed description of this data –ten-day values
interpolated in 0.25º latitude x 0.25º longitude grid– can be found in Ducet et al. [2000]. These
time series were low-passed filtered using a cosine-Lanczos filter to remove intraseasonal and
higher frequencies. The filter passes 5, 50 and 95% of the original power at periods of 85, 120
and 200 days respectively. A slight smoothing in space was carried out using a 3 point triangular
filter. Weekly-averaged sea surface wind stress on 1°x1° was calculated using ERS-1 and 2
scatterometer observations (Département d'Océanographie Spatiale, IFREMER, France).
Currents at 220 m depth were obtained from a slope mooring near 30°S off Chile. The original
time series from February 1993 to February 1999 was first low passed filtered with a cosineLanczos filter (50% power at 60 hrs), and then filtered as for SLH. Hourly coastal sea level data
from tide gages located at La Libertad (2º 12’S), Lobos de Afuera (6º 56’S), Callao (12º 02’S),
Arica (18º 28’S), Caldera (27º 04’S) and Valparaiso (33º 02’S) were obtained from the Sea Level
Center, University of Hawaii. These time series were low-passed filtered to remove tides as
above and daily averaged. Gaps of up to eight days in the resulting time series were linearly
interpolated; segments with larger gaps were excluded from the analysis. Daily mean
atmospheric pressure from NCEP/NCAR reanalysis was used to correct the inverse barometer
effect in coastal SLH from Arica, Caldera and Valparaiso (but not from La Libertad, Lobos de
Afuera and Callao, weakly affected by the small air pressure fluctuations). Finally all the coastal
SLH were further filtered as for satellite SLH.
3. Sea Level Variability
We use a simple equatorial Kelvin wave model –with similar phase speed and Kelvin wave
projection coefficients as in Kessler and McPhaden [1995]– forced by satellite zonal windstress
observations along the equator (Fig.1a) to study low-frequency (lower than intraseasonal) sea
level fluctuations near the eastern boundary and evaluate the amplitude of the reflected Rossby
wave. In general, the model reproduces several of the main features observed in the eastern
equatorial Pacific (Figures 1b and 1c). According to the linear theory, for given latitude, there
exists a critical period above which the reflected Rossby wave can exist [e.g. Clarke and Shi
1991]. For 30°S, off the near-meridional Chilean coast, the critical period is about 230 days.
Thus, fluctuations with periods longer than these can propagate seaward as long Rossby waves,
while oscillations with shorter period will remain trapped against the coast. Figure 1d shows that
coastal disturbances in sea level near 30°S tend to propagate westward resembling a first mode
baroclinic Rossby-wave. Note that the observed SLH propagation speed is slightly larger than
that predicted by linear theory, and fluctuations with period somewhat smaller than the critical
one also propagate offshore. Both features have been observed at mid latitude in the Pacific
Ocean and have been associated to the effect of sheared mean currents (e.g. Fu and Chelton
2001). Figures 1b, c and d suggest a direct relation between the amplitude of SLH disturbances
originated by equatorial Kelvin waves at the eastern equatorial Pacific and the amplitude of the
midlatitude Rossby wave at 30 °S off the South American coast.
Low-passed, coastal SLH along the eastern South Pacific boundary (Fig. 2) increases greatly
during the 1997-98 El Niño period, and varies with smaller amplitude otherwise. All time series
are significantly correlated (95% confidence level) and fluctuation amplitudes tend to decrease
southward. Rms coastal sea level varies from 9.3 cm at La Libertad to 4.7 cm at Valparaiso with
a maximum of 9.9 cm at Callao. We apply EOF analysis to the sea-level time series in Fig. 2a to
extract the signal that covaries in phase along the boundary (coastal trapped disturbances
propagate from the equator to 33°S in about 20 days, essentially in phase for seasonal and lower
frequencies). The first EOF mode explains 87% of the total variance, and the spatial structure is
similar to the rms structure of the observed coastal series (Fig. 2c). EOF mode 1 time series and
model sea level at the eastern equatorial Pacific (Fig. 2b) are well correlated (r = 0.75) and
oscillate nearly in phase indicating that a large fraction of the low frequency fluctuations at the
eastern boundary can be explained by wind-forced, equatorial Kelvin waves. We also run a
linear model (results not shown) of the tropical Pacific with both zonal and meridional wind
forcing and a western boundary to estimate the relative importance of the equatorial Kelvin wave
originating from the reflection of Rossby waves at that boundary. For realistic dissipation and
full reflection efficiency there, results indicate only a minor contribution of the reflected wave in
the far eastern Pacific.
4. Alongshore Flow
The large mean poleward flow (12.8 cm s-1 for the period from January 1993 to February
1999) observed at 220 m depth over the slope is related to a well-defined PCU present at this
depth in the study region (figure 3a). Intraseasonal oscillations with typical amplitude of 10 to 15
cm s-1 are evident in this current record. At longer periods –semiannual, annual and interannual–
the alongshore flow exhibits also significant variability (Fig. 3b and c, see also Fig. 4a). Caldera
sea level and alongshore flow spectra have similar shape at frequencies lower than 0.05 cycles
per day (Fig. 4a). Both time series are significantly coherent (at 80% confidence level) for period
longer than 30 days (Fig. 4b) with a clear maximum at the intraseasonal band centered about 70
days. In this band, phase values are about –160° consistent with a near inverse relation between
the sea level and the alongshore flow expected from geostrophic balance in a Kelvin-wave like
disturbance. In contrast, for period longer than 120 days, the phase changes to values between
about –90° and –60°.
As it is suggested in figure 1d, low frequency fluctuations (periods longer than 120 days)
observed in sea level anomalies at 30°S propagate seaward as long baroclinic Rossby waves. For
a given frequency, the local sea level disturbance associated with such a wave can be represented
as
η = A exp(ikx − iwt )
(1)
where η is the sea level disturbance, ω, k, x and t are angular frequency, wave number, offshore
distance and time respectively. For a Rossby wave the flow will be mainly in geostrophic
balance, so at the coast
v(t ) x = 0 = − gη x / f = − gc / f ηt
(2)
where f is the Coriolis parameter and c = ω /k is the long baroclinic Rossby wave phase speed
( c = β c02 / f 2 where β is the latitudinal gradient of f and c0 the phase speed of a long gravity
internal wave, about 2.4 m s-1 for the study region). We compared observed alongshore current
at 220 m depth (Fig. 3) with the alongshore flow calculated with equation (2). Two different sealevels were used in these calculations: the time-series of the Kelvin wave at the equator, and the
sea-level EOF-1 time-series at Caldera. A scaling was applied to the amplitude of the equatorial
Kelvin wave based on the ratio of the EOF-1 near 30°S to the EOF-1 at La Libertad (Fig. 2c).
This scaling accounts for the poleward reduction observed in sea-level variability due to seaward
leakage of energy and dissipation by friction close to the coast. The estimated alongshore
velocity fluctuations using the equatorial Kelvin wave model are better correlated (0.71) with the
observed current oscillations than those estimated using EOF time series (0.56). We also
consider results using the observed sea level at Caldera. There is a reasonably good agreement
with observed current (correlation of 0.57), but not as good as the result obtained with the
equatorial Kelvin wave model. Other processes, like seasonal thermal expansion, may influence
coastal sea level and decrease its correlation with the alongshore current. At interannual
frequencies, the predicted and observed current are also well correlated (Fig. 3c). A similar result
was obtained by Pizarro et al. [2001] using a more complex model for the eastern South Pacific
boundary.
5. Summary and Conclusions
Sea level from the eastern South Pacific and satellite altimetry show that much of the sea level
variability along the coasts of Peru and Chile at seasonal, annual and interannual frequencies is
related to equatorial Kelvin waves reaching the South American coast. This equatorial forcing
has also been widely documented for intraseasonal and higher frequencies. At seasonal and
lower frequencies, longitude-time plots of sea-level height anomalies show that coastal
disturbances propagate offshore as long baroclinic Rossby waves. The analysis of 7-year long
current record over the slope and near the core of the Peru-Chile Undercurrent at 30ºS indicate
that this flow is strongly modulated at seasonal and interannual periods by Rossby waves forced
by equatorial Kelvin waves arriving at the South American coast. There, about 50% of the low
frequency (lower than intraseasonal) variability of the Undercurrent can be explained using the
amplitude of the equatorial Kelvin waves at the eastern equatorial Pacific to estimate the
amplitude of the Rossby wave near the coast in the eastern South Pacific. The seasonal cycle of
coastal upwelling supporting high biological production off the west coast of South America
follows the seasonal cycle of local winds. However, it is now clear that remotely-forced,
seasonal variability in the Peru-Chile Undercurrent may also influence the local, yearly cycle of
biological production as the Undercurrent provides upwelling source water along much of this
coast.
Acknowledgments. The long-term current measurement program off Chile has been supported
by grants from the Swedish agency SAREC/SIDA the Danish and Chilean National Research
council (the latter in part from the FONDAP program), and for the Danish National Research
Foundation. Sea level data was obtained from the Sea-Level Center, University of Hawaii and
SHOA-Chile. OP and MR are grateful for support from Fundación Andes-Chile (Grants C13600/4 and D-13615).
References
Blanco, J. L., A. C. Thomas, M.-E. Carr, and P. T. Strub, Seasonal climatology of hydrographic
conditions in the upwelling region off northern Chile. J. Geophys. Res., 106, 11451-11467, 2001.
Clarke, A. J., and C. Shi, Critical frequencies at ocean boundaries. J. Geophys. Res., 96, 10 73110 738, 1991.
Ducet, N., P. Y. Le Traon and G. Reverdin, Global high resolution mapping of ocean circulation
from TOPEX/POSEIDON and ERS-1/2. J. Geophys. Res., 105, 19477-19498, 2000.
Enfield, D. B., and J. S. Allen, On the structure and dynamics of monthly mean sea level
anomalies along the Pacific coast of North and South America. J. Phys. Oceanogr., 10, 557-578,
1980.
Fu, L-L, and D.B. Chelton, Large-Scale ocean circulation. In Satellite altimetry and earth
sciences, edited by L-L. Fu and A. Cazenave. Academic Press, 2001.
Kesler, and M. J. McPhaden, Oceanic equatorial waves and the 1991-1993 El Niño. J. of
Climate, 8, 1 757-1 774, 1995.
Hormazabal, S., G. Shaffer and O. Pizarro. Tropical Pacific control of intraseasonal oscillations
off Chile by way of oceanic and atmospheric pathways. Geophys Res. Letter. (in press)
Huyer, A., R. L. Smith and T. Paluszkiewicz, Coastal upwelling off Peru during normal and El
Niño times. J. Geophys. Res., 92, 14 297-14 307, 1987.
Huyer, A., M. Knoll, T. Paluszkiewicz, and R. L. Smith, The Peru Undercurrent: a study in
variability. Deep-Sea Res., 38 Suppl.1, 247-279, 1991.
McCreary, J. P., A linear stratified ocean model of the coastal undercurrent. Phil. Trans. Roy.
Soc., London, A302, 385-413, 1981.
McCreary, J. P., P. K. Kundu and S-Y. Chao, On the Dynamics of the California Current
System. J. Mar. Res., 45, 1-32, 1987.
Pizarro, O. A. J. Clarke and S. Van Gorder, El Niño sea level and currents along the South
American coast: comparison of observations with theory. J. Phys. Oceanogr. 31, 1891-1903,
2001.
Shaffer, G., O. Pizarro, L. Djurfeldt, S. Salinas and J. Rutllant, Circulation and low frequency
variability near the Chile coast: Remotely-forced fluctuations during the 1991-1992 El Nino. J.
Phys. Oceanogr., 27, 217-235, 1997.
Shaffer, G., S. Hormazabal, O. Pizarro and S. Salinas, Seasonal and interannual variability of
winds, current and temperature near the coast of Chile. J. Geophys. Res., 104, 29951-29961,
1999.
Silva N., and S. Neshyba, On the southernmost extension of the Peru-Chile Undercurrent. DeepSea Res., 26, 1387-1393, 1979.
Wyrtki, K., The horizontal and vertical field of motion in the Peru Current. Bull. Scripps Inst. of
Ocean., 8, 313-346.
DYNAMICS OF THE PERU-CHILE UNDERCURRENT
b) Model sea level
c) TP/ERS sea level
d) TP/ERS sea level, 30ºS
2000 1999 1998 1997 1996 1995 1994 1993
a) Zonal windstress
140ºE
-0.1 Pa
180º
140ºW 100ºW
0.0
0.1
140ºE
180º
-20 cm
140ºW 100ºW
0.0
20
140ºE
180º
-20 cm
140ºW 100ºW
0.0
20
75ºW
80ºW
-10 cm 0.0
85ºW
10
Figure 1. Low-passed time series of zonal windsterss averaged between 2°S and 2°N along the
equatorial Pacific Ocean (a). Sea level disturbances from a wind-forced, equatorial Kelvin wave
model (b). Low-passed sea level anomalies obtained from satellite altimetry along the equatorial
Pacific (c), and at 30°S in the eastern South Pacific (d). The straight slanted line indicates the
theoretical phase speed of the first mode long Rossby wave (c = 0.02 cm s-1). Longitude was
reversed to illustrate the relationship between equatorial and midlatitude disturbances.
Equatorial Anomalies were averaged between 2°S and 2°N. All time series were low-passed to
remove intraseasonal and higher frequencies. Note the different color scales in figures c and d.
DYNAMICS OF THE PERU-CHILE UNDERCURRENT
1992 1993 1994 1995 1996 1997 1998 1999
(a)
cm
20
0
-20
40
(b)
cm
20
0
-20
1992 1993 1994 1995 1996 1997 1998 1999
cm
15
(c)
rms observed
EOF mode 1
10
5
0
0
5ºS
10ºS
15ºS
20ºS
25ºS
30ºS
35ºS
Figure 2. Low-passed time series of adjusted sea level at La Libertad (2º 12’S), Lobos de
Afuera (6º 56’S), Callao (12º 02’S), Arica (18º 28’S), Caldera (27º 04’S) and Valparaiso (33º
02’S) along the coast of South America (a). Time series of sea level anomalies at the eastern
boundary computed from a wind-forced, equatorial Kelvin wave model (thin line), and EOF
mode 1 (thick line) estimated from the observed sea level time series and projected in la Libertad
(b). The eigenvector of the EOF mode 1 is presented in (c) together with the observed rms value
at each location.
speed (cm s-1)
DYNAMICS OF THE PERU-CHILE UNDERCURRENT
40
(a)
0
-40
speed (cm s-1)
speed (cm s-1)
-80
20
(b)
0
-20
10
(c)
5
0
-5
-10
1993
1994
1995
1996
1997
1998
Figure 3. Subinertial (thin line) and low-passed (intra- seasonal and lower frequencies removed;
thick line) alongshore velocity at 220 m depth over the slope near 30°S off Chile (a). Low-passed
time series as in top panel (thick line), estimated alongshore current using the amplitude of the
equatorial Kelvin wave (thin line) and estimated alongshore flow using the sea-level EOF mode
1 (dashed line) (b). As in b, but for the interannual time series (annual and other higher
frequencies removed) (c).
DYNAMICS OF THE PERU-CHILE UNDERCURRENT
Coherence Squared
Spectral density
103
2
10
10
1
1
(a)
0
(b)
0.8
-40
0.6
-80
95%
0.4
-120
80%
0.2
0
0.02
0.04
Frequency (cpd)
0
Phase (degrees)
104
-160
0
0.02
0.04
Frequency (cpd)
-200
Figure 4. Auto-spectra of the alongshore current at 30°S (dashed line) and sea level at Caldera
(solid line) (a), and coherence squared (line) and phase (dots) between the time series (b).
Horizontal lines indicate the 80% and 95% significance level for coherence. Only phase results
for coherence values above 80% confidence level are plotted.