MARCH 1999
LENTERS AND COOK
409
Summertime Precipitation Variability over South America:
Role of the Large-Scale Circulation
J. D. LENTERS*
AND
K. H. COOK
Atmospheric Science Program, Cornell University, Ithaca, New York
(Manuscript received 11 August 1997, in final form 27 January 1998)
ABSTRACT
The observed large-scale circulation mechanisms associated with summertime precipitation variability over
South America are investigated. Particular attention is paid to the Altiplano where a close relationship has been
observed between rainfall and the position and intensity of the Bolivian high. Empirical orthogonal function
(EOF), correlation, and composite analyses suggest that on intraseasonal timescales (typically 5–20 days), rainy
periods on the Altiplano are associated with at least three types of circulation anomalies, involving either
extratropical cyclones, cold-core lows, or the westward enhancement of the South Atlantic high. In each instance,
the primary support for high rainfall rates is a moist, poleward flow at low levels along the eastern flank of the
central Andes in association with the South Atlantic convergence zone (SACZ). The warm, low-level flow along
the SACZ also inflates the overlying atmospheric column, resulting in an intensification and southward shift of
the Bolivian high. Thus, the position of the SACZ (and associated frontal activity) plays a crucial role in the
variability of both the Bolivian high and Altiplano rainfall.
On longer timescales, the Bolivian high also shifts southward and intensifies during wet periods on the
Altiplano. The seasonal transition from December to the wetter month of January is accompanied by a westward
enhancement of the SACZ and northwesterly flow along the central Andes. The transition to the drier month
of February is accompanied by weakening northwesterly flow and cooler, drier conditions along the Altiplano.
Interannual precipitation variability on the Altiplano is strongly correlated with the amplitude of the fifth EOF
of the 200-mb height field, the ‘‘dry phase’’, which is much like the anomalous conditions during February. A
case study of the dry conditions during the 1987 El Niño associates reduced convection on the Altiplano with
the presence of a strong cold front over eastern South America and cold, dry air to the west. The characteristic
eastward shift of the South Pacific convergence zone during El Niño may be responsible for this enhanced frontal
activity in the SACZ (through teleconnections) and, therefore, the cool, dry, convectively unfavorable conditions
in the central Andes.
1. Introduction
Summertime precipitation in the central Andes exhibits pronounced spatial and temporal variability. High
precipitation rates, in excess of 20 mm day21 in some
areas (Fig. 1), are orographically induced along the eastern flank of the mountain belt (Lenters and Cook 1995).
Proceeding westward up the eastern cordillera of the
central Andes onto the plateau known as the Altiplano
(3.8-km elevation, spanning roughly 148–228S), climatological precipitation rates fall well below 6 mm
day21 . In the northern Altiplano, however, summer rainfall amounts are sufficient to sustain a productive, but
* Current affiliation: Center for Limnology, University of Wisconsin—Madison, Madison, Wisconsin.
Corresponding author address: Dr. Kerry H. Cook, Atmospheric
Science Program, 1110 Bradfield Hall, Cornell University, Ithaca,
NY 14853.
q 1999 American Meteorological Society
climatically sensitive, agricultural community as well
as Lake Titicaca, the largest lake of its elevation in the
world.
Precipitation on the Altiplano during austral summer
is convective and typically occurs during afternoon
thunderstorms (Aceituno and Montecinos 1993). Radiative heating of the land surface during the early daytime hours warms the overlying air, inducing low-level
inflow (including moisture-laden air from the Amazon
Basin to the east), convergence, precipitation, and the
release of latent heat, which adds to the intensity of the
convection (Gutman and Schwerdtfeger 1965; Rao and
Erdogan 1989). This localized convection can be highly
variable. For example, during the 1988/89 austral summer La Paz, Bolivia, received an average of 3.1 mm of
precipitation per day, but the rainfall was very sporadic,
with 36 out of the 90 days being rain free and 16 days
receiving more than 5 mm of rain (36 mm on one particular day).
According to Aceituno and Montecinos (1993) this
episodic nature is typical of the Altiplano, with wet
410
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 1. Climatological precipitation rate over central South America
for DJF [solid contours; from Legates and Willmott (1990)]. Contour
interval is 2 mm day21 . Shading indicates topography at 1-km intervals, starting at 0.5 km. Symbols show locations of Altiplano rainfall stations.
periods lasting a few days being ‘‘preceded and followed
by dry episodes.’’ The variability is not restricted to
daily timescales, as rainfall rates vary significantly on
5–20-day, seasonal, and interannual timescales (Kessler
1981; Horel et al. 1989; Lenters 1997a). In general, the
Altiplano is dry (wet) during negative (positive) phases
of the Southern Oscillation (Kessler 1981; Aceituno
1988; Martin et al. 1993). This is reflected in the present
study by the anomalously dry months of February and
December during the 1987 El Niño.
One possible explanation for this precipitation variability is the influence of the large-scale circulation.
Aceituno and Montecinos (1993) investigated this possibility by forming upper-air composites of geopotential
height for wet and dry days during austral summer for
the period 1980–87 (excluding the 1982/83 El Niño
summer). That study suggests that summertime precipitation variability on the Altiplano is related to the position and intensity of the Bolivian high, a prominent
anticyclone that develops over the central Andes during
austral summer (Lenters and Cook 1997). During wet
days the Bolivian high is stronger and shifted farther
south of its climatological position (see Fig. 2a) than
during dry days, and the 200-mb zonal wind speed over
the Altiplano is 5–10 m s21 stronger. Kessler (1981)
found a similar relationship for wet and dry summer
months, suggesting that the connection between precipitation and the Bolivian high applies to interannual timescales as well.
The observed relationship between the precipitation
field and the Bolivian high may indicate that latent heating associated with high precipitation in the central Andes causes the Bolivian high to strengthen, as has been
found for the Amazon Basin and surrounding regions
FIG. 2. (a) Average 200-mb geopotential height over South America
for DJF 1985–93 (from the NASA/DAO reanalyses). Contour interval
is 20 m (50 m) above (below) 12 350 m. ‘‘H’’ denotes the location
of the Bolivian high, and the dashed line indicates the approximate
position of the SACZ (see Fig. 3). (b) As in (a) but for the 850-mb
wind velocity. Vector beneath figure denotes a wind velocity of 10
m s21 .
(DeMaria 1985; Gandu and Geisler 1991; Lenters and
Cook 1997). Alternatively, wet episodes may be the
result of a strengthened and southward shifted Bolivian
high.
The purpose of this study is to identify the mechanisms responsible for the observed relationship between
the Bolivian high and the variability of summertime
precipitation on the Altiplano. The focus is broadened
to the continental scale when we find connections with
MARCH 1999
LENTERS AND COOK
rainfall and circulation features over the Amazon and
southeastern South America. Several features of the circulation are shown to be relevant in this context. One
is the South Atlantic convergence zone (SACZ), a region of high precipitation that stretches diagonally off
the southeast coast of South America (Figs. 2–3). At
low levels, the SACZ is characterized by the convergence of warm, moist air blowing off the coast along
the southeastern flank of the South Atlantic high (Fig.
2b; Kodama 1992; Lenters and Cook 1995). To the west
is the strong northwesterly flow along the Andes between 108 and 208S (Fig. 2b). To the north, the westward
extension of the South Atlantic high, combined with the
presence of tropical easterlies, brings flow onto the continent north of about 158S (Lenters and Cook 1995).
Section 2 describes the observational data used in this
study, and section 3 identifies the 200-mb height anomalies associated with high precipitation on the Altiplano
using empirical orthogonal function (EOF) analysis.
Physical mechanisms are investigated in section 4,
which explores connections with the low-level flow.
Variability on several timescales is investigated in section 5, including a composite analysis of intraseasonal
rainy episodes. Results are discussed and summarized
in section 6.
2. Data selection and analysis
Daily analyses of geopotential height, wind velocity,
temperature, specific humidity, and precipitation rate
were obtained from the Data Assimilation Office (DAO)
at the National Aeronautic and Space Administration
(NASA) Goddard Space Flight Center for eight austral
summer seasons [December–February (DJF) 1985/86–
1992/93]. The NASA/DAO analysis (Schubert et al.
1993) is a fixed 4D assimilation system that reduces the
introduction of artificial atmospheric variability by eliminating changes in the assimilation model’s parameterizations and numerics (among other things). The data
are available on a 2.58 long 3 28 lat grid at 18 pressure
levels. The 200-mb and 850-mb levels are examined,
and the geographic domain is restricted to South America (168N–588S, 858–308W).
Given the inherent uncertainties in assimilated precipitation fields, two additional independent measures
of precipitation rate are also used. The first is the Global
Precipitation Index (GPI), a satellite-derived (IR) estimate of large-scale tropical and subtropical rainfall that
is provided as an intermediate product of the Global
Precipitation Climatology Project (Janowiak and Arkin
1991). This data is available in pentads (5-day means)
for 1986–94 on a 2.58 square grid covering the Tropics
and subtropics (408N–408S). Thus, except for December
1985, the GPI data overlaps the time period of the
NASA/DAO analysis.
The second set of independent precipitation data is
the daily, ground-based records from meteorological
stations at Juliaca, Peru, and La Paz, Bolivia (located
411
in Fig. 1). These are available from the Global Daily
Summary (GDS) dataset of the National Climatic Data
Center (NCDC). These stations were chosen because of
their proximity to the wetter, northern Altiplano; drier
stations to the south capture only a fraction of the summer rain events on the Altiplano. Four other GDS stations are also located on or near the northern Altiplano,
but they have short records (less than 20 days of data)
and/or low correlation with the six-station average precipitation rate. The GDS records for Juliaca and La Paz
end in 1991, so six complete rainy seasons overlap the
NASA/DAO analysis.
A 5-day running mean is applied to the NASA/DAO
analysis and the Altiplano station data to smooth highfrequency variability (Horel et al. 1989; Kiladis and
Weickmann 1992) and short-lived events, allowing a
focus on timescales of 3–5 days or longer. Such longerterm variability is more likely to be directly related to
the large-scale circulation since short-lived events recorded at individual stations are often associated with
localized convection. The smoothing is also valuable
for filling in gaps from missing data, minimizing the
effects of potential errors in the precipitation records,
and establishing a timescale that is consistent with the
GPI pentads.
Anomalous 5-day running means (for all three datasets) are created by subtracting the seasonal mean of
each of the eight (or six) summers. In this way the results
for each summer are normalized to each other. Much of
the interannual variability is still retained in the anomalies, since individual summer months are often more
different from each other than are entire summer seasons. An example of this is the 1987 El Niño, which
we examine more closely in section 5.
Empirical orthogonal function analysis is used to
identify the dominant circulation patterns. This method
was chosen over composite techniques based on the
results of Lenters (1997a), in which 200-mb geopotential height composites from the NASA/DAO data were
formed for wet and dry episodes on the Altiplano. Their
analysis confirmed the results of Aceituno and Montecinos (1993), namely, that the Bolivian high is stronger and shifted southward during wet episodes. However, there was considerable variability between individual years, suggesting that Altiplano wet episodes are
associated with more than one type of circulation pattern. Here we use an EOF analysis to select recurring
circulation patterns first, and then correlate these patterns with the precipitation field. [For a comprehensive
description of EOFs, including applications to the atmospheric sciences, the reader is referred to Preisendorfer (1988). Useful introductions to EOF analysis are
also provided by Kutzbach (1967) and Wilks (1995).]
The EOF analysis is performed on the standardized
5-day running mean 200-mb geopotential height anomalies over South America for the eight summer seasons
DJF 1985/86–1992/93. The standardization indicates division by the (climatological) standard deviation at each
412
MONTHLY WEATHER REVIEW
grid point. Since the geopotential height field is typically
much more variable in the extratropics than the Tropics,
this standardization ensures that extratropical height
anomalies do not have a heavily weighted influence on
the EOF analysis. The analysis is restricted to 40 degrees
of latitude (08–408S) and 55 degrees of longitude (858–
308W), resulting in a correlation matrix of size 483 3
483.
The EOFs, or eigenvectors, are scaled such that the
magnitude of the mth EOF is equal to the square root
of the mth eigenvalue of the correlation matrix. Together
with the fact that the 200-mb height anomalies are standardized, this allows the EOF loadings to be interpreted
as coefficients of correlation between the time-dependent amplitude (or strength) of the EOFs and the original
height field. Furthermore, the scaling results in each of
the EOF amplitudes having unit variance.
3. Circulation anomalies associated with South
American rainfall
a. Correlations with Altiplano precipitation
The 483 eigenvectors and eigenvalues of the correlation matrix were calculated and sorted using standard
techniques. The first 10 EOFs together explain over 93%
of the variance. The log-eigenvalue diagram [not shown;
Wilks (1995)] indicates that approximately the first 30
EOFs are ‘‘signal’’ and the remaining 453 are ‘‘noise.’’
However, none of the EOFs after the fifth are significantly correlated with Altiplano precipitation and these
first five resemble recognizable circulation patterns. For
these reasons, we focus on EOFs 1–5, which account
for over 77% of the variance in the 200-mb height field.
Figure 4 shows the first five EOFs; their correlations
with the 5-day running mean precipitation average of
the La Paz and Juliaca station data are shown in Table
1. Use of the Spearman rank correlation coefficient instead of the ‘‘standard’’ Pearson product-moment coefficient yields similar results. Statistical tests of significance were based on a two-tailed t test of the correlation coefficient after it has been subjected to a Fisher
transformation (Wilks 1995, 154). To account for the
reduced degrees of freedom resulting from serial correlation in the time series, the method of Quenouille
(1952, 168) was adopted.
The first EOF (Fig. 4a) is a relatively zonal pattern
characterized by anomalously high pressure south of
258S, low pressure to the north, and a weak trough along
558W. When the amplitude of this anomaly is positive
the Bolivian high is shifted south of its climatological
position (Fig. 2a), inducing anomalous easterly flow between 108 and 208S. The amplitude of EOF1 is significantly correlated with Altiplano precipitation (Table 1).
This agrees with the results of Kessler (1981) and Aceituno and Montecinos (1993), who also found that a
southward shift of the Bolivian high is associated with
anomalous precipitation on the Altiplano.
VOLUME 127
The second EOF (Fig. 4b) has high pressure to the
southwest of the Altiplano and a trough to the east.
Although this pattern is not strongly correlated with the
Altiplano precipitation time series on a 0-day lag (Table
1), we find a significant correlation (r ø 0.23, significant
at 95%) when lagging the precipitation by 5–10 days.
That is, EOF2 tends to be strong (weak) 5–10 days after
high (low) precipitation on the Altiplano. This indicates
that EOF2 may represent an atmospheric response to
rainfall in the central Andes region. This suggestion is
reinforced by previous modeling remits. Using a linear,
primitive equation model, Lenters and Cook (1997) simulate a strong Bolivian high to the southwest of the
Altiplano and a trough to the east in response to condensational heating. Previous linear modeling studies
(Silva Dias et al. 1983; DeMaria 1985) show similar
response patterns to Amazonian condensational heating,
and indicate that the time-dependent response to tropical
heating over South America includes a westward-propagating Bolivian high. EOF2 shows similar time-dependent behavior, in the sense that lagged correlations
of its amplitude with the 200-mb height field indicate
a westward propagating high pressure cell (Lenters
1997a).
The third (Fig. 4c), fourth (Fig. 4d), and fifth (Fig.
4e) EOFs are associated with anomalous strengthening
and/or shifting of the Bolivian high to the southeast,
east, and south of the Altiplano, respectively. Enhancement of the climatological trough (Fig. 2a) east of the
Bolivian high is also evident in Figs. 4c and 4e. Of these
three EOFs, EOF5 most closely resembles the ‘‘wet
phase’’ identified by Kessler (1981) and Aceituno and
Montecinos (1993) and is the only one significantly correlated with Altiplano precipitation rates (Table 1).
EOFs 3 and 4 are suggestive of patterns associated with
the northeastward intrusion of frontal systems that typically occur along the SACZ (Kousky and Gan 1981;
Kodama 1993; Kousky and Kayano 1994; Paegle and
Mo 1997).
EOF4 shows no significant correlation with Altiplano
rainfall (Table 1), and the amount of variance it explains
(10.8%) and, hence, its eigenvalues are very close to
the third (11.1%) and fifth (10.4%) EOFs. According
to the test proposed by North et al. (1982), EOF4 is
statistically indistinguishable from EOF3 and EOF5; the
sample size used in generating the EOFs is not large
enough to allow one to treat EOF4 as a distinct mode
of variability. In fact, EOF4 may simply be an evolution
of the same general pattern associated with EOF3 (given
the similar, but shifted features of Figs. 4c and 4d). For
these reasons, the fourth EOF will not be a focus of the
remaining analysis.
The eigenvalues of the third and fifth EOFs are also
close enough to raise some doubts about their distinguishability. In addition, the relationship between EOF3
and Altiplano precipitation (Table 1) is statistically insignificant. However, as discussed below, we find that
high precipitation on the Altiplano is often associated
MARCH 1999
LENTERS AND COOK
413
with both strong positive and negative phases of EOF3,
suggesting an important underlying relationship that is
not captured by the simple correlations presented in Table 1. A pattern resembling EOF3 has also been previously shown to be related to regional precipitation
anomalies in the vicinity of the central Andes (Kiladis
and Weickmann 1992; Kousky and Kayano 1994; Paegle and Mo 1997). Thus, the third EOF is retained in
the analysis.
b. Correlations with regional South American
precipitation
Besides being correlated or otherwise associated with
precipitation on the Altiplano, EOFs 1, 3, and 5 are
strongly linked to precipitation anomalies throughout
South America. Figures 5a,b show maps of the coefficient of correlation between EOF1’s amplitude and the
NASA/DAO (5-day running mean) and GPI precipitation fields, respectively, with unshaded regions indicating statistically significant correlation at the 95% level.
(For the GPI correlation, the EOF amplitude is sampled
every five days to match the corresponding GPI pentad.)
In both precipitation datasets EOF1 is associated with
at least three regional precipitation anomalies (Figs.
5a,b). The central Andes shows a positive correlation,
as does a larger region to the southeast, around 258S,
508W. A band of negative correlation is present in eastern Brazil, with highest values attained around 158S,
458W. This band is somewhat diagonally oriented and
positioned east-northeast of the regions of above-normal
precipitation. This suggests a possible connection with
the SACZ (Fig. 3), namely that high (low) precipitation
in the central Andes is associated with a southwestward
(northeastward) positioning of the SACZ, with portions
of the eastern Amazon Basin showing a similar, but outof-phase relationship.
Regional precipitation anomalies associated with
EOF3 are shown in Figs. 5c,d. Neither precipitation
dataset exhibits significant correlations in the central
Andes, in agreement with the station data (Table 1).
However, both datasets show strong correlations over
southeastern South America in the vicinity of the SACZ.
These anomalies (Figs. 4c, 5c,d) are similar to those
found by Kiladis and Weickmann (1992) in their observational study of austral summer circulation anomalies associated with tropical convection. Figure 12b
(12c) in Kiladis and Weickmann (1992) shows an upperlevel anomaly, much like the negative (positive) phase
of EOF3, which is significantly correlated with strong
convection in northeastern (western) Brazil on 14–30day timescales. In both regions the correlation is strongest one day after the peak in convection, suggesting
that the convection forces the circulation anomaly. Similar wet/dry anomalies in the vicinity of the SACZ have
also been noted previously by Kousky and Casarin
(1986), Kousky and Kayano (1994), and Paegle and Mo
(1997).
FIG. 3. (a) Average DJF precipitation rate over South America from
the NASA/DAO reanalyses (1985–93). Contour interval is 1 mm
day21 , and the dashed line indicates the approximate position of the
SACZ (as determined by the band of maximum precipitation). (b)
As in (a), but for the Global Precipitation Index data. DJF average
approximated from pentads by averaging 2 December–24 February
for 1985–93 (excluding Dec of 1985).
Of the three EOFs, EOF5 shows the highest correlation
with precipitation in the central Andes, especially for the
GPI (Fig. 5f) and station data (Table 1). The correlation
with NASA/DAO precipitation in this region is weak but
positive (Fig. 5e). Both precipitation datasets show generally negative correlations everywhere north of 108–
158S, with significant values in the eastern equatorial
Pacific (just off the coasts of Ecuador and Peru) and over
central Brazil. In the GPI dataset the negative correlation
in central Brazil extends along a south-southeastward-
414
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 4. (a) First EOF of the 5-day running mean
200-mb geopotential height field for the eight summer seasons DJF 1985–93. Units are equivalent to
the coefficient of correlation between the height field
and the amplitude of EOF1, and negative contours
are dashed. Percent of the total variance explained
is noted in parentheses. (b)–(e) As in (a) but for
EOFs 2–5, respectively.
oriented band down to 308–358S, where the NASA/DAO
data also shows a negative anomaly.
According to Aceituno (1988) and Rao and Hada
(1990), the spatial pattern of precipitation associated with
EOF5 fits characteristics of the positive phase of the Southern Oscillation index (SOI). During La Niña (high SOI),
positive precipitation anomalies occur over the Altiplano
with negative anomalies along the coasts of Ecuador and
Peru. The opposite is typical during El Niño (low SOI)
events. This suggests a close relationship between ENSO
and the 200-mb height anomalies associated with EOF5,
which we explore in greater detail in section 5.
4. Physical mechanisms
In this section we further investigate the observed
correlations to identify causal relationships between the
large-scale circulation and the precipitation field, and
MARCH 1999
LENTERS AND COOK
TABLE 1. Coefficient of correlation between Altiplano precipitation
and EOF amplitudes. Asterisk indicates correlation is significant at
the 95% level.
EOF
r value
1
2
3
4
5
0.283*
0.104
20.048
0.082
0.321*
the mechanisms that lead to the causality. Since periods
of high precipitation during the summer often indicate
low-level atmospheric conditions that favor convective
development (e.g., Kodama 1992; Kodama 1993), we
examine the low-level circulation fields.
Figure 6 shows correlations between the amplitude
of the first EOF and the 5-day running mean 850-mb
wind velocity (Fig. 6a), specific humidity (Fig. 6b), temperature (Fig. 6c), and moist static stability (Fig. 6d).
All of the fields are from the NASA/DAO dataset, and
unshaded areas again represent regions with statistically
significant correlations. The moist static stability in the
lower troposphere is calculated using
S52
(u *|
]u *e
e p1 2 u *|
e p2 )
ø2
,
]p
p1 2 p2
(1)
where u*e | p is the saturation equivalent potential temperature at pressure level p, p1 5 850 mb, and p 2 5
700 mb. Here S , 0 indicates conditional instability.
Calculation of S using p 2 5 500 mb yields qualitatively
similar, but weaker, results.
The southward shift of the Bolivian high represented
by EOF1 (with signs as in Fig. 4a) is accompanied by
anomalous anticyclonic flow at low levels over southeastern South America (Fig. 6a), with generally dry
(Fig. 6b), cool (Fig. 6c) conditions to the north. The
temperature pattern resembles the somewhat zonal 200mb height pattern associated with EOF1 (Fig. 4a), as
expected from hydrostatic balance. The anomalous flow
onto the east coast (108–208S) is associated with relatively dry, stable air; the flow off the continent (238–
288S) coincides with moist, warm, unstable low-level
conditions. The precipitation perturbation that is associated with EOF1 is, therefore, consistent with the perturbation of the low-level circulation; both the positive
and negative precipitation anomalies over southeastern
South America in Figs. 5a and 5b correspond to lowlevel moist static stability anomalies of the opposite sign
(Fig. 6d).
The low-level circulation associated with the third EOF
(Fig. 7) has some features similar to those of EOF1, but
with less zonal orientation and more diagonally oriented
structure. Warm (Fig. 7c), moist (Fig. 7b) air and unstable
conditions (Fig. 7d) are associated with northwesterly
flow (Fig. 7a) from the central Andes into the South
Atlantic, coinciding with regions of high precipitation
415
(Figs. 5c and 5d). The band of low precipitation to the
east generally shows more stable low-level conditions.
The diagonal orientation of bands of low-level warm
(cold), moist (dry), unstable (stable) air, and high (low)
precipitation associated with EOF3 and (to some extent)
EOF1 suggest an association with the strength and/or
position of the SACZ and associated frontal activity.
During positive phases of EOF1 and EOF3 (i.e., south
or southeastward shifts of the Bolivian high), the SACZ
is shifted west (Figs. 5a,d) of its average position (Fig.
3). Conditions to the east and north are then cooler, drier,
and less favorable for convection, resulting in belownormal precipitation for that region. During negative
phases, the opposite conditions prevail, and high precipitation shifts to the northeast. This ‘‘seesaw’’ behavior in the SACZ was first noted by Kousky and Casarin
(1986) and discussed more recently by Paegle and Mo
(1997); both studies suggest a relationship between the
SACZ and the 30–60-day (Madden–Julian) oscillation.
The present EOF analysis reveals a further connection
with the SACZ seesaw, namely, the position and intensity of the Bolivian high.
The low-level wind regime (Figs. 2b, 6a, 7a) plays
an important role in the position and strength of the
SACZ by advecting warm, moist air from the northwest
and cool, dry air from the southeast. This is in agreement
with Kodama (1993), who found that poleward flow in
the SACZ is a critical ingredient for maintaining intense
convection. Paegle and Mo (1997) also noted strong
northerly flow in northern Argentina during westward
shifts of the SACZ (their ‘‘positive’’ phase).
The close correspondence between the low-level temperature field (Figs. 6c and 7c) and the 200-mb height
field (Figs. 4a and 4c) suggests that the southward shift
of the Bolivian high and the corresponding precipitation
anomalies (Figs. 5a–d) are both reflections of a warm,
unstable lower troposphere. It is also possible that the
temperature and height anomalies are, at least in part,
responding to diabatic heating within the SACZ; there
is a strong resemblance between the linear response to
condensational heating in the SACZ [Fig. 7g, Lenters
and Cook (1997)] and the high pressure cells of EOFs
3 and 4 (Figs. 4c,d). This interpretation is also supported
by the results of Kiladis and Weickmann (1992).
The low-level conditions associated with EOF5 (Fig.
8) are rather distinct from those of the first and third
EOFs, in that the circulation features show less diagonal
structure. A positive moisture anomaly (Fig. 8b) stretches eastward from the central Andes near 208S. Anomalously high temperatures (Fig. 8c) occur nearly directly
below the 200-mb high that characterizes EOF5 (Fig.
4e). Anomalous instability is also present (Fig. 8d) in
this region, and these features coincide with the associated central Andean precipitation maxima (Figs. 5e,f).
Considering that the Altiplano receives most of its moisture from the east, the high humidity along the flank of
the central Andes (Fig. 8b) may play an especially im-
416
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 5. Coefficient of correlation between the amplitude of EOF1 and the precipitation field of the (a)
NASA/DAO (5-day running mean) and (b) GPI datasets. Regions of statistically insignificant correlation (at
the 95% level) are shaded. Contour interval is 0.1, with negative values dashed. (c) and (d), (e) and (f ) As
in (a) and (b), but for EOF3 and EOF5, respectively.
portant role in the association between EOF5 and precipitation on the Altiplano.
The fifth EOF, like the first and third, is associated
with anomalously strong northwesterly flow along the
eastern flank of the Andes (Fig. 8a). In this case, however, the anomaly is strongest over Peru and Bolivia,
well north of the wind regimes associated with EOF1
and EOF3 in Paraguay and northern Argentina. This
MARCH 1999
LENTERS AND COOK
417
FIG. 6. Coefficient of correlation between the amplitude of EOF1 and the 5-day running mean 850-mb
(a) wind velocity, (b) specific humidity, (c) temperature, and (d) moist static stability from the NASA/DAO
analyses. Unshaded regions indicate a statistically significant correlation at the 95% level; for (a) either the
zonal or meridional wind speed must be significantly correlated. The vector beneath (a) indicates a correlation
magnitude, (ru2 1 ry2)1/2 , of 0.6, and the contour interval in (b)–(d) is 0.1, with negative contours dashed.
suggests that the moist air just east of the central Andes
may be more tropical in origin than the SACZ-like bands
associated with EOF1 and EOF3.
5. Temporal variability of the circulation patterns
There are a number of similarities among the EOFs,
such as enhanced northwesterly flow over western South
America and its association with a band of warm, moist,
unstable air and high precipitation, as well as low precipitation over eastern portions of the continent. This is
somewhat unsatisfying, as the distinctions among the
EOFs are blurred, and it may be inappropriate to treat
the EOFs as truly independent modes of variability.
However, further distinction among the EOFs emerges
by comparing the temporal variability of their respective
amplitudes.
Three timescales are considered. The first concerns
the seasonal cycle, specifically that portion which occurs
within DJF. Figure 9a shows the six-summer mean, nor-
malized precipitation for DJF from the stations at La
Paz and Juliaca. The seasonal cycle for these three
months consists of a progression from a relatively dry
December, through January (the wettest month of the
year on average), and back to a drier February. According to a standard t test (assuming independence),
precipitation differences between January and the other
two months are statistically significant well beyond the
95% level. We examine the seasonal cycle in each of
the three EOFs to understand how each contributes to
establishing the observed cycle in precipitation. Interannual variability and the intraseasonal ‘‘event’’ timescale (roughly 5–20-day episodes) are also considered.
a. Climatological seasonal cycle
To decompose the time-mean seasonal cycle for DJF,
the amplitudes of EOFs 1, 3, and 5 are averaged over
the eight summer seasons to form three time series; these
are indicated by the solid lines in Figs. 9b–d. Since an
418
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 7. As in Fig. 6 but for EOF3.
eight-sample average is not large enough to identify a
robust seasonal cycle, a monthly average was also
formed by sampling the time series every five days
(starting with 3 December) and averaging the first middle, and last six samples. These three 240-day averages
are indicated by the dashed lines in Figs. 9b,d. (Similar
methodology was followed to obtain the 6-yr mean precipitation shown in Fig. 9a.) Although the time period
used in creating the seasonal cycle includes two moderate El Niño episodes (1987 and 1992), excluding these
years from the average has a minimal effect. To maintain
as large a sample size as possible, the full datasets are
used.
The first EOF shows a statistically significant increase
in amplitude from December to January, but a much
smaller (and insignificant) decrease from January to
February (Fig. 9b). This suggests that EOF1 is associated with the Altiplano’s summer transition from dry
(December) to wet (January), but not from wet to dry
(February). EOF3 exhibits no significant seasonal variability (Fig. 9c). The fifth EOF shows only a weak
transition from December to January but a significant
high-to-low transition from January to February (Fig.
9d). The seasonal cycle of EOF2 (not shown) is similar
to that of EOF1 and lags the precipitation time series
by about 7 days, consistent with the earlier results.
The seasonal variability in EOF1 and EOF5 together
account for the transitions that lead to the observed
seasonal cycle in the Altiplano precipitation record. Are
the circulation anomalies associated with these EOFs
forcing the transitions, or are they reflecting changes in
the diabatic heating field when the precipitation regimes
change for some other reason? The transition of EOF1
from December to January in the 5-day-mean time series
(solid line in Fig. 9b) clearly leads the transition in the
precipitation field (Fig. 9a) by 2–3 days. Similarly, the
transition of EOF5 from January to February leads the
rainfall transition by at least 3–4 days. This suggests
that the tendency of the Altiplano to be wetter during
January than during December or February is, at least
in part, dynamical in origin and is not simply a local,
convective response to the change in solar zenith angle.
This is similar to the idea offered by Horel et al. (1989)
regarding the Amazon Basin’s rapid transition to the
west season.
The significant correlation between the amplitude of
MARCH 1999
LENTERS AND COOK
419
FIG. 8. As in Fig. 6 but for EOF5.
EOF1 and the South American precipitation field (Table
1) suggests that the transition from a relatively dry December (negative phase of EOF1) to a wetter January
(positive phase of EOF1) on the Altiplano is accompanied by a westward (eastward) displacement of the
SACZ. Composite precipitation anomaly maps (not
shown) for December and January support this interpretation. According to Fig. 6 (with change of sign),
northwesterly flow over the eastern Amazon Basin and
western South Atlantic during December supports a
band of warm, moist, unstable air at 850 mb, whereas
conditions to the southwest, near the central Andes, are
nearly opposite. These low-level anomalies reverse in
January, when EOF1 enters its positive phase.
The transition to relatively dry conditions in February, in association with a sign change in EOF5, is by a
different mechanism. For the wet-to-dry transition, convection is suppressed by a reversal in the anomalous
850-mb northwesterly wind flow along the northern central Andes and an accompanying drop in temperature
and humidity east of the Altiplano (Fig. 8).
This analysis suggests that positive phases of EOF1
and EOF5 are each necessary, but insufficient conditions
for high precipitation on the Altiplano during January.
During December and February cool, dry, southeasterly
flow at 850 mb is present along at least some portion
of the central Andes, since the amplitude of either EOF1
or EOF5 is negative. During January, warm, moist,
northwesterly flow spans the full length of the central
Andes, providing abundant moisture and low-level instability. This northwesterly flow often manifests itself
as a low-level jet, and is capable of transporting significant amounts of moisture poleward (Berri and Inzunza 1993; Paegle and Mo 1997).
b. Interannual variability
Time series for studying interannual variability were
generated by first subtracting the climatological seasonal cycle from the 8-yr EOF time series and the 6-yr
Altiplano station data precipitation record. Monthly averages were formed by sampling the resulting time series
every five days and averaging the first, middle, and last
six samples. The monthly mean anomalies are shown
for precipitation (solid line) and EOF5 (dashed line) in
Fig. 10. The precipitation anomalies show a moderate
420
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 9. (a) Seasonal cycle of normalized summertime precipitation on the Altiplano. Solid line
indicates the six-summer average (DJF 1985–91) of the standardized 5-day running mean Altiplano
precipitation rate. Dashed line represents 30-day averages. Numbers on x axis indicate the day of
the season. (b), (c), and (d) As in (a) but for the amplitudes of the first, third, and fifth EOFs,
respectively.
amount of interannual variability, with February and
December of 1986 and January of 1987 and 1988 being
particularly wet months. The three driest months of the
period are January of 1986 and February and December
of 1987, the two months that appear to be most strongly
affected by the 1987 El Niño.
The interannual variability of EOF5 (dashed line in
Fig. 10) is similar to that of the Altiplano precipitation.
The coefficient of correlation is 0.82, which is statistically significant at the 99% level. This is notably higher than corresponding correlations for EOF1 and EOF3
(r ø 0.55 and 20.14, respectively), indicating that in-
FIG. 10. Interannual variability of the Altiplano precipitation rate (solid), amplitude of EOF5
(dashed), and SOI (dotted) for DJF 1985–93 (SON 1985–92 in the case of the SOI, which leads
the other two time series by 3 months). Anomalies represent monthly deviations from the 8-yr
(or 6-yr) monthly mean and have been normalized by the std. dev. for the period shown.
MARCH 1999
LENTERS AND COOK
terannual fluctuations in summertime precipitation on
the Altiplano are primarily associated with circulation
anomalies captured by EOF5.
When the amplitude of EOF5 is low, such as during
the 1987 El Niño, anomalously cool, dry, southeasterly
flow prevails along the central Andes, shifting the Bolivian high northward and reducing precipitation rates
on the Altiplano (Figs. 4e; 5e,f; and 8, with change of
sign). During above-normal phases of EOF5, such as
February 1986, warm, moist, air is advected from the
northwest, the Bolivian high is displaced southward, and
high precipitation rates occur on the Altiplano. Thus,
the connection between Altiplano precipitation and the
north–south position of the Bolivian high also applies
to interannual timescales, as was first suggested by Kessler (1981).
To further explore possible connections between
ENSO and Altiplano precipitation, standardized monthly anomalies of the SOI for SON 1985–92 are plotted
in Fig. 10 (dotted line; seasonal means removed, as in
the precipitation and EOF5 time series). The SOI is
moderately correlated with both EOF5 and the Altiplano
precipitation, with r ø 0.51 and 0.63, respectively (both
significant at the 90% level). The three time series correspond particularly well during the 1987 El Niño.
These results corroborate the previously noted connections between ENSO and precipitation variability on the
Altiplano (Aceituno 1988; Martin et al. 1993). It should
be noted that some extremes in the SOI (e.g., the 1988/
89 La Niña) do not appear in Fig. 10 since the seasonal
means have been removed.
A case study of the 1987 El Niño event is presented
here to explore the physical mechanisms that connect
the circulation anomalies of EOF5 with the interannual
variability of Altiplano rainfall. This period is chosen
because of its anomalous nature (the large negative
anomalies in Fig. 10), as well as the significant influence
that El Niño events, in general, have on Altiplano precipitation rates.
The focus of the case study is the abrupt transition
from wet conditions in January 1987 to dry conditions
in February 1987, which coincided with the development of classical El Niño conditions along the west coast
of South America (Kousky and Leetmaa 1989). The
transition is reflected in EOF5 as a significant drop in
amplitude (Fig. 10), which occurred approximately six
days before the precipitation decrease.
Figure 11 shows six Hovmöller diagrams, which depict the evolution of precipitation and large-scale circulation anomalies from the NASA/DAO reanalysis
from 22 January to 26 February 1987. The anomalies
represent 5-day running mean deviations from the 8-yr
February mean, averaged between 128S and 188S (four
grid points; less than four in regions of topography,
where missing data is ignored). The longitude domain
includes the coastal areas of Peru, west of about 708W;
the northern Altiplano and eastern flank of the central
421
Andes, between approximately 708W and 608W; and the
Amazon Basin, east of about 608W.
At the beginning of the analysis period, the Altiplano
was experiencing the last wet episode of the 1986/87
summer season (Fig. 11a). Associated with this wet period was a strengthening of the Bolivian high (Fig. 11b),
a strong northerly low-level wind anomaly (Fig. 11c),
and above-normal 850-mb specific humidity (Fig. 11d).
The Bolivian high reached a maximum over the Altiplano 2 days after the rainfall maximum and propagated
westward (Fig. 11b). This was associated with a drop
in amplitude of EOF2 (not shown).
As the Altiplano precipitation rates decreased after
26 January, the northerly wind anomaly propagated to
the east (Fig. 11c) and the specific humidity decreased
with the intrusion of dry air from the east (Fig. 11d).
It is difficult to ascertain why the dry air spreads westward since the moisture advection and divergence fields
are noisy (not shown). However, the westward propagation of the trough to the east of the Bolivian high
(Fig. 11b) may be associated with horizontal wind convergence and compensatory subsidence at lower levels
as in the upper-tropospheric cyclonic vortices studied
by Virji (1981) and Kousky and Gan (1981).
The dry period from 2 to 7 February is not particularly
unusual. However, on 8 February low-level winds east
of the Altiplano became more southerly (Fig. 11c) and
advected cool, dry, stable air northward (Figs. 11d,f).
This occurred in association with a strong cold front,
which extended well to the southeast and triggered convection in the SACZ and Amazon basin (Fig. 11a). At
about the same time, convection and low-level humidity
began to increase along the coast of Peru (Figs. 11a and
11d), where atmospheric conditions had been relatively
warm and unstable since the end of January (Figs.
11e,f).
By 11–12 February, 5-day mean precipitation rates
on the Altiplano and nearby low-level temperature and
humidity levels were at a minimum. A deep upper-level
trough developed in association with the cold front, and
the southerly wind anomaly at 850 mb was greater than
7 m s21 (Fig. 11c). The full 850-mb wind was greater
than 5 m s21 , the largest, positive, meridional wind
speed of the entire eight-summer time period. The 850mb temperature anomaly at 638W was also the coldest
reached along the eastern flank of the central Andes.
Temperature, humidity, and 200-mb heights returned
to normal by about 20 February, but Altiplano precipitation rates remained below normal through the remainder of the month (Fig. 11a). Rainfall along the coast
of Peru subsided by 22 February, but conditions there
remained relatively favorable for convective development (Figs. 11d and 11f). The only circulation anomaly
that persisted was the low-level southerly wind (Fig.
11c). It was weak by 26 February, but may be connected
with the continued dry conditions on the Altiplano by
preventing the influx of tropical moisture from the north.
This case study suggests that, on interannual time-
422
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 11. Hovmöller diagram of anomalous 5-day running mean (a) precipitation rate (mm day 21 ), (b) 200mb geopotential height (m), and 850-mb (c) meridional wind speed (m s 21 ), (d) specific humidity (g kg21 ),
(e) temperature (8C), and (f ) moist static stability [8C (100 mb)21 ] for 22 Jan–26 Feb 1987. Anomalies
represent deviations from the 8-yr ‘‘Feb’’ average (30 Jan–28 Feb) and are averaged over 128–188S. Negative
values are shaded. Data obtained from the NASA/DAO analyses.
scales, dry conditions on the Altiplano may be the
result of anomalously strong or persistent cold fronts
in association with the SACZ. The corresponding
southerly advection of dry, stable air (one of the defining characteristics of the negative phase of EOF5)
would then be responsible for suppressing convection
in the region and shifting the Bolivian high northward.
Such frontal activity would also be associated with
enhanced convection over the southern Amazon Basin,
as occurred during the case study and is evident in the
precipitation field associated with the negative phase
of EOF5 (Figs. 5e and 5f). This is consistent with the
results of a number of authors (Nobre and Oliveira
1986; Aceituno 1988; Rao and Hada 1990; Gan and
MARCH 1999
LENTERS AND COOK
Rao 1991), who show that southern Brazil and northern
Argentina experience increased cyclogenesis, frontal
activity, and precipitation during negative phases of
the Southern Oscillation.
If dry conditions on the Altiplano during El Niño are
associated with extreme or persistent frontal activity to
the east (in the SACZ), then what is the causal connection? One possibility is that the connection is made
through the South Pacific convergence zone (SPCZ).
During negative phases of EOF5, dry conditions on the
Altiplano and wet conditions in the SACZ are associated
with an eastward shift of the SPCZ of about 208–308
relative to the positive phase (Lenters 1997b). An eastward shift of the SPCZ is also a general characteristic of
El Niño (Kousky et al. 1984) as a result of higher sea
surface temperatures in the central and eastern equatorial
Pacific. Teleconnections between the SPCZ and SACZ
have been suggested (Lau and Chan 1983; Kalnay et al.
1986; Kousky and Casarin 1986; Grimm and Silva Dias
1995; Paegle and Mo 1997), and Kalnay et al. (1986)
note that ‘‘the SACZ tends to occur when the SPCZ is
strong and displaced to the east.’’ This is in agreement
with the results of Grimm and Silva Dias (1995), who
simulate an anomalous trough over the SACZ in response
to a southeastward-displaced SPCZ. Eastward displacement of the SPCZ during El Niño may cause such enhancement of frontal activity in the SACZ through blocking activity or changes in the Walker circulation (Kousky
et al. 1984). This increased frontal activity in eastern
South America is then likely to embed the Altiplano in
cool, dry, convectively unfavorable conditions.
c. Intraseasonal events
The first, third, and fifth EOFs are not clearly distinguishable on intraseasonal timescales. For example, the
1986/87 rainy season on the Altiplano was characterized
by three distinct episodes of high precipitation. EOF5
has maxima during each of the three wet episodes (not
shown); but EOF1 also has amplitude maxima during
the latter two wet events, and EOF3 is large and positive
(negative) during the second (third) event.
In general, no one particular EOF appears to be associated with intraseasonal wet or dry events on the
Altiplano. When one EOF is strong, the remaining
EOFs tend to have large amplitudes as well. This is a
limitation of the EOF analysis and reflects a lack of
‘‘simple structure’’ in the EOF patterns. One can often
obtain simple structure and avoid the above difficulties
by rotating the EOFs, a procedure that is described in
detail by Richman (1986). Rotated EOFs would likely
provide additional understanding of the circulation features associated with intraseasonal wet and dry episodes on the Altiplano.
Instead of rotating the EOFs, however, we use compositing to focus on the synoptic meteorology associated
with high precipitation events in the central Andes. Although more subjective than EOF rotation, this approach
423
emphasizes the meteorological conditions associated
with Altiplano rainfall. Also, continued use of EOFs
(rotated or otherwise) does not show promise for resolving the difficulties encountered with EOF3, namely,
associating both its positive and negative phases with
Altiplano precipitation.
We first define wet 5-day periods, using all three precipitation datasets. A time series consistent with the sampling of the station data was constructed from the NASA/
DAO analyses by averaging the 5-day running mean precipitation rates at three grid points in the Altiplano region
(centered at 168S, 708W; 188S, 67.58W; 208S, 67.58W).
The same was done for the GPI data, except only two
grid points were used (centered at 16.258S, 68.758W and
18.758S, 68.758W) and the temporal resolution was restricted to pentads. ‘‘Wet’’ pentads were defined as those
in which the 5-day-averaged precipitation rate was above
the summer median in at least two out of the three datasets
(two out of two for 1991/92 and 1992/93). The remaining
pentads are designated as ‘‘dry.’’ Of the 136 pentads, 61
are wet, and 75 are dry.
The daily circulation and precipitation anomalies (deviations from the seasonal means shown in Figs. 2 and
3; NASA/DAO data) within each wet pentad were visually examined to subjectively identify distinct circulation modes that characterize wet episodes. Guided by
the EOF analysis, we paid close mention to the presence
of warm, moist, unstable air at 850 mb, the positions
and intensities of the Bolivian high and SACZ, and the
low-level flow east of the Andes. Three distinct circulation types were identified and assigned to each of the
305 days of the 61 wet pentads. Not all of the precipitation events fit a particular category equally well; in
a number of cases the resemblance is borderline. However, to maintain a large sample size for compositing,
no wet days were left uncategorized. Although this introduces more variability within each category, the larger sample size allows the averaging process to ‘‘filter
out’’ the less robust circulation features of the wet episode, thereby increasing one’s confidence in the anomalies that remain. An implicit assumption in our categorization process is that each of the five days within
a wet pentad is characterized by high rainfall. This assumption, which is not likely to be true in all cases, is
difficult to avoid given the restriction of the GPI data
to pentads.
Despite these inherent problems with the subjective
categorization process, the results that follow show statistically significant anomalies associated with each of
the three types of synoptic conditions. Furthermore, in
many instances the results are closely corroborated by
the preceding EOF analysis.
1) EXTRATROPICAL CYCLONE ACTIVITY—
WESTWARD SHIFT OF SACZ
The first type of precipitation event, which we associate with extratropical cyclone activity and a west-
424
MONTHLY WEATHER REVIEW
ward shift of the SACZ, is characterized by wet conditions in the central Andes and dry conditions in the
Amazon Basin. This type accounts for 104 of the 305
rainy days in the time series. Figures 12a,b show precipitation anomalies (difference from the dry pentads)
for this type of wet period for the NASA/DAO and GPI
precipitation records. The suggestion of links between
Altiplano rainfall and the position of the SACZ in these
figures is consistent with the EOF analysis.
The 200-mb circulation field associated with this type
of precipitation event has a zonally elongated high pressure anomaly centered on the coast to the southeast of
the precipitation maximum (Fig. 12c), with anomalous
low pressure over the Amazon Basin. In the lower troposphere, this structure is reflected by same-sign temperature anomalies (Fig. 12d) and northwesterly flow east
of the central Andes (Fig. 12e). Low-level specific humidity (Fig. 12f) and moist static stability anomalies (not
shown) also mimic the structure in the precipitation field.
Note the evidence of low-level cyclonic circulation
southeast of the central Andes (Fig. 12e). This is a reflection of transient cyclonic vortices. Our examination
of the daily synoptic conditions associated with this type
of precipitation event shows these cyclones forming or
intensifying in the lee of the southern central Andes
(generally between 208S and 408S), and propagating to
the east-southeast. Many of these cyclones are ‘‘Chaco
lows’’ (Satyamurty et al. 1990), but there are also mature, eastward-propagating midlatitude systems that
cross the southern Andes in association with this type
of precipitation. [See Lenters (1997a) for case studies
of the composite events presented in this section.]
Strong pressure gradients to the northeast of the extratropical cyclones lead to warm, moist, low-level flow
from the northwest and an intense, active SACZ, which
is displaced westward of its climatological position (Fig.
3). The associated convection behind the warm front is
often observed to wrap around the cyclone in a classic
comma-shaped pattern [as in Fig. 1 of Virji (1981)],
with a strong cold front following the convection. These
extratropical characteristics of the SACZ have been noted before (e.g., Kodama 1993), but have not been previously associated with Altiplano precipitation to our
knowledge.
Structures identified in several of the EOFs are represented, including wet (dry) SACZ-like structure in the
central Andes (Amazon Basin), the enhanced Bolivian
high southeast of the Altiplano, and the northwesterly
wind flow and associated warm, moist air at low levels.
The average amplitudes of EOF1, EOF2, and EOF5
during this type of precipitation event are significantly
higher than during dry periods. (EOF3 also shows a
positive difference, but it is not statistically significant.)
This reiterates the difficulties in interpreting the unrotated EOFs as distinct modes of intraseasonal variability.
On the other hand, one can conclude that at least three
of the EOFs are (in part) an upper-level manifestation
of cyclone activity.
VOLUME 127
The composite fields (Fig. 12) also bear close resemblance to the ‘‘weakened’’ or ‘‘positive’’ phase of the
SACZ seesaw identified by Paegle and Mo (1997). This
suggests that the mode that they identified is associated
with cyclone activity and the early stages of a developing SACZ. This would account for the observed westward displacement and/or weakening of the convergence zone.
2) COLD-CORE
SACZ
SUBTROPICAL LOWS—INTENSIFIED
The composite wet-minus-dry precipitation pattern
associated with the second type of event is shown in
Fig. 13. This type accounts for 94 of the 305 rainy days
in the time series. Both datasets show anomalously high
precipitation in the central Andes and reduced precipitation over southern Brazil and northern Argentina
(Figs. 13a,b). Increased rainfall is also present over the
average position of the SACZ (Fig. 3), suggesting that
this type of precipitation event is associated with an
intensification of the convergence zone.
Aside from increased central Andean rainfall, this
second type of precipitation event has conditions nearly
opposite to those identified in the previous type of wet
episode. The most prominent circulation anomaly is a
deep, cold-core low centered at about 308S, 458W and
between at least 200 and 850 mb (Figs. 13c and 13e).
At low levels the circulation about the low brings cool,
dry air (Figs. 13d and 13f) onto the coast near 558W,
coincident with the negative precipitation anomaly
(Figs. 13a,b).
The anomalous low-level northwesterly flow along
the eastern flanks of the northern central Andes (Fig.
13e) and the associated region of relatively high specific
humidity (Fig. 13f) suggest that the winds along the
Andes may be advecting tropical moisture from the
northwest, thereby fueling convection on the Altiplano.
Results from a case study (Lenters 1997a) suggest that
low-level wind convergence to the northwest of the
cold-core low (i.e., the confluence of northwesterly and
southeasterly winds in Fig. 13e) may also play a role
in supporting the high rainfall rates.
Aside from the central Andes region, the precipitation
anomalies associated with an intensification of the
SACZ (Figs. 13a,b) are similar to the negative phase of
EOF3 (Figs. 5c,d) and the ‘‘intensified SACZ’’ phase
identified by Paegle and Mo (1997). Furthermore, the
composite amplitude of EOF3 for these events is strongly negative, and the low-level circulation patterns in
Figs. 13d,f are very similar to those associated with the
negative phase of EOF3 (Figs. 7a–c, with change of
sign). Thus, the low correlation between EOF3 and Altiplano precipitation (Table 1) does not indicate a lack
of significance for this type of circulation pattern. Rather, it simply does not capture the tendency for Altiplano
rainfall to be above normal during times when the Bolivian high is stronger to the southeast (as in Fig. 12c)
MARCH 1999
LENTERS AND COOK
FIG. 12. Anomalous precipitation field associated with wet conditions on the Altiplano during composite episodes of extratropical cyclone-induced westward shifts of the SACZ based on (a) NASA/
DAO analyses and (b) GPI data. Contour interval is 1 mm day21 (negative values dashed) and shading
indicates statistically insignificant differences from dry conditions at the 95% level (based on twosided Student’s t-test, assuming independence). (c)–(f ) As in (a), but for the (c) 200-mb height and
wind fields (10-m contour interval, 10-m s21 wind vector) and 850-mb (d) temperature (0.38C contour
interval), (e) height and wind (5-m contour interval, 5-m s21 wind vector), and (f ) specific humidity
(0.3 g kg21 contour interval).
425
426
MONTHLY WEATHER REVIEW
VOLUME 127
FIG. 13. As in Fig. 12 but for wet episodes associated with cold-core subtropical lows and
intensification of the SACZ.
as well as during times when the high is weaker to the
southeast (as in Fig. 13c).
3) WESTWARD ENHANCEMENT OF SOUTH
ATLANTIC HIGH AND SACZ
The third type of atmospheric conditions that we relate
to above-normal precipitation events in the central Andes
is associated with a westward enhancement of the South
Atlantic high. The precipitation anomalies (Figs. 14a,b)
are similar to those associated with extratropical cyclone
activity (Figs. 12a,b), with high precipitation over the central Andes accompanied by anomalously dry conditions
over the eastern Amazon Basin, suggesting a westward
shift of the SACZ. The region of enhanced precipitation
(Figs. 14a,b) does not have the diagonal orientation typ-
MARCH 1999
LENTERS AND COOK
427
FIG. 14. As in Fig. 12 but for wet episodes associated with westward enhancement of the South
Atlantic high and SACZ.
ically associated with the SACZ, but the low-level circulation features are similar to those of the SACZ, as
discussed below.
The composite 200-mb circulation anomalies include
easterly flow over the Altiplano, with a well-defined
high pressure cell roughly 158 to the south (Fig. 14c).
This anticyclonic anomaly is similar to EOF5 (Fig. 4e)
and represents a southward shift of the Bolivian high.
(The latitudinal position of the two cells in Figs. 4e and
14c is more similar if Fig. 14c is normalized by its
standard deviation, as was done for the EOFs.) At 850
mb a high temperature anomaly west of 508W and south
of about 208S (Fig. 14d) is positioned beneath the 200mb anticyclone. This warm air appears to have been
advected from the north along the western flank of an
anomalous low-level high (Fig. 14e), which is centered
428
MONTHLY WEATHER REVIEW
over Uruguay and southern Brazil. This low-level feature represents the most distinct difference from the first
type of wet event, which is otherwise very similar, but
shows cyclonic circulation southeast of the central Andes (Fig. 12e). Both composites, however, are characterized by moist, poleward flow at low levels (Figs. 12e,f
and 14e,f) along a region of enhanced convection (Figs.
12a,b and 14a,b) and decreased static stability (not
shown). Such features are characteristic of the SACZ
(Kodama 1992; Kodama 1993), but in this case are located near the Andes, west of the SACZ’s climatological
position.
Positive (negative) humidity anomalies at low levels
(Fig. 14f) along the western (eastern) flank of the 850mb high pressure anomaly suggest the southward (northward) advection of moist (dry) air in connection with
the anticyclonic circulation. This high humidity along
the eastern flank of the southern central Andes, together
with some conditional instability associated with the
band of warm air, appear to be the primary mechanisms
responsible for high rainfall on the Altiplano during
these episodes. The full-field 850-mb high pressure cell
manifests itself as a westward enhancement of the South
Atlantic high, which typically does not extend far onto
the continent during the summer. The SACZ is usually
located along the western flank of the South Atlantic
high (e.g., Kodama 1992), and so a westward enhancement of the high is consistent with the aforementioned
westward displacement of the SACZ. The associated
low-level moisture flux along the western flank of the
high and, therefore, the moisture source for the Altiplano are from the northeast (the South Atlantic and
eastern Amazon), as opposed to the northwesterly flux
that is characteristic of the other two types of wet episodes.
6. Discussion and summary
The large-scale circulation mechanisms associated
with daily to interannual rainfall variability over South
America during austral summer have been investigated
using observations from eight summer seasons (DJF
1985/86–1992/93). Particular emphasis has been placed
on understanding the role of the Bolivian high in the
episodic nature of precipitation on the Altiplano of the
central Andes. Gridded atmospheric analyses and three
independent measures of precipitation have been analyzed using a variety of methods, including composite,
EOF, and correlation techniques, as well as a case study
of the 1987 El Niño. The results reveal a number of
important relationships between South American precipitation variability and features of the large-scale circulation.
a. Intraseasonal wet episodes
On the intraseasonal ‘‘event’’ timescale, summer
rainy periods on the Altiplano have been observed to
VOLUME 127
be associated with at least three types of synoptic conditions. The first is characterized by the presence of low
pressure southeast of the Altiplano, either in the form
of a developing Chaco low in northern Argentina, or as
a propagating extratropical cyclone further south. Similar cyclonic vortices have been identified by Satyamurty et al. (1990). The low pressure systems are usually associated with strong low-level northwesterly
winds, which advect warm, moist, unstable air along
the eastern flank of the central Andes, triggering convection on the Altiplano. The convection extends southeastward into the SACZ, which is westward of its climatological position during such episodes, creating drier-than-normal conditions over the Amazon Basin. Kodama (1993) also noted that poleward low-level flow is
instrumental in maintaining intense convection in the
SACZ, but the connection with precipitation on the Altiplano has not been previously revealed. Because of
the warm front and intense convection, the Bolivian high
is stronger and extends anomalously southeastward
along the band of warm air. Meanwhile, southerly flow
along the western flank of the low-level cyclone advects
cool, dry air along a cold front that lies beneath an
upper-level trough immediately west of the SACZ. Others have found similar 200-mb height patterns to be
associated with strong convection in the western Amazon Basin and reduced precipitation in the climatological SACZ (Kiladis and Weickmann 1992; Paegle and
Mo 1997).
The second type of synoptic condition is characterized by northerly flow along the northern central Andes
and along the eastern flank of a cold-core low positioned
over southeastern South America. The SACZ is intensified as a result of the diversion of moisture around the
northern periphery of the low. High precipitation on the
Altiplano is maintained by the northwesterly flux of
tropical moisture along the northern central Andes, as
well as the convergence of this flux with southerly flow
along the western flank of the low. Because of the coldcore nature of the cyclone, an upper-level trough prevails over southern Brazil, and the Bolivian high is shifted southwestward instead of southeastward. A similar
relationship to strong convection over eastern Brazil has
been noted by Kiladis and Weickmann (1992) and Paegle and Mo (1997).
The third type of large-scale circulation pattern observed to be associated with intraseasonal wet episodes
on the Altiplano is related to a westward enhancement
of the South Atlantic high and SACZ. The anomalous
high pressure over south-central South America advects
warm, moist air along the eastern flank of the southern
central Andes, providing convectively favorable conditions. The flux of moisture toward the central Andes
is directed from the South Atlantic and eastern Amazon
Basin, as opposed to the northwesterly flux typical of
the other two episodes. Anomalously dry, southerly flow
over eastern portions of the continent inhibit convection
over much of the Amazon Basin. The Bolivian high is
MARCH 1999
LENTERS AND COOK
stronger and shifted southward as a result of the warm
anomaly along the southern central Andes.
b. Climatological seasonal cycle
The Altiplano receives more precipitation during the
month of January than during December or February.
This longer-timescale variability represents a 3-month
segment of the 12-month annual cycle, and is a reflection of a greater number of wet episodes during January.
During December anomalously dry (moist), southeasterly (northwesterly) flow in eastern Bolivia (Brazil) is
associated with an eastward displacement of the SACZ
and, correspondingly, fewer rainy episodes on the Altiplano. A positive (and anomalous) meridional temperature gradient at low levels results in a northward
shift of the Bolivian high. January is characterized by
a westward displacement of the SACZ and a southward
shift of the Bolivian high. Northwesterly moisture flux
at 850 mb along the entire eastern flank of the central
Andes leads to more numerous rain events. Drier conditions on the Altiplano during February are accompanied by a reduction in the northwesterly flux of moisture along the northern central Andes, a drop in temperature and humidity east of the Altiplano, and a weakening and northward displacement of the Bolivian high.
There is some evidence that the transitions to wetter
conditions on the Altiplano in January and drier conditions in February lag related transitions in the 200mb geopotential height field by a few days. This suggests that the seasonal cycle of summer rainfall on the
Altiplano is not simply due to local changes in solar
insolation, but is more closely associated with changes
in the large-scale circulation, such as variations in the
moisture supply east of the central Andes.
c. Interannual variability
Summer rainfall on the Altiplano is characterized by
a significant amount of interannual variability. Some of
this variability, such as the dry period during February
and December of 1987, is associated with El Niño and
the Southern Oscillation. In addition, over 67% of the
interannual variance in monthly mean precipitation rate
on the Altiplano is accounted for by the amplitude of
the fifth EOF of the 200-mb geopotential height field.
The negative (dry) phase of this EOF, such as occurred
during the 1987 El Niño, is associated with a weakened
and northward displaced Bolivian high and a region of
cool, dry, southerly flow at low levels east of the Altiplano.
During the January–February onset of the 1987 El
Niño dry period, the cold, low-level air associated with
the negative phase of EOF5 was related to the northward
penetration of a strong cold front along the eastern flank
of the central Andes. The dry, stable conditions and
persistent southerly flow that followed the passage of
the front inhibited convection on the Altiplano for
429
weeks. Though the case study was restricted to the dry
period of February 1987, the statistically significant correlation between EOF5 and Altiplano precipitation rates
suggests an explanation that is more generally applicable, perhaps even to other ENSO events. The proposed
mechanism is also consistent with previous studies (e.g.,
Aceituno 1988; Rao and Hada 1990; Gan and Rao
1991), which indicate increased cyclogenesis, frontal
activity, and precipitation in southeastern South America during El Niño. The fundamental link to ENSO may
be through the SPCZ, which is shifted eastward during
El Niño and is evidenced to have teleconnections with
the SACZ.
d. The Bolivian high–Altiplano rainfall connection
The southward displacement of the Bolivian high during intraseasonal wet episodes on the Altiplano is related
(through hydrostatic balance) to warm, low-level air that
has been advected southward along the Andes. This
warm air is generally accompanied by reduced vertical
stability and high humidity, providing conditions favorable for convection on the Altiplano. Thus, the southward displacement of the Bolivian high does not necessarily ‘‘cause’’ high precipitation on the Altiplano, but
is simply an upper-level manifestation of the low-level
mechanisms responsible for triggering the convection.
Both the southward shift of the Bolivian high and the
accompanying Altiplano rainfall are intimately related
to the position and intensity of the SACZ. In this respect,
then, the SACZ and its associated frontal activity are
an important source of variability for the Bolivian high
and Altiplano precipitation.
A second reason for the observed shift of the Bolivian
high during wet episodes is simply that the upper-level
circulation responds to the release of anomalous latent
heat during rainy periods. This is suggested by the second EOF of the 200-mb geopotential height field, which
closely resembles the linear response to diabatic heating
over the central Andes (high pressure to the southwest
and easterly flow to the north) and is significantly correlated with Altiplano precipitation on a 5–10-day lag.
Modeling studies have also simulated the formation and
southwestward propagation of anomalous high pressure
in response to diabatic heating over the South American
region (e.g., DeMaria 1985).
On the longer timescale of the seasonal cycle, the
southward shift of the Bolivian high remains evident
(during January, the wettest month), but the 200-mb
anomaly is more zonally elongated (as represented by
the first EOF). This is partly because the various types
of wet episodes, each of which have distinctly different
200-mb height patterns away from the Altiplano, are
blended together in the January average. However, the
more zonally uniform structure may also reflect the influence of the annual cycle of solar insolation on the
low-level temperature field.
Finally, on interannual timescales, wet (dry) months
430
MONTHLY WEATHER REVIEW
are once again characterized by a strengthened (weakened) and southward (northward) displaced Bolivian
high. As evidenced by the 1987 El Niño, this connection
(unlike that for the seasonal cycle) appears to be less
related to a blending of wet (or dry) episodes, but is
instead a reflection of extreme conditions, such as the
anomalous northward penetration of strong cold fronts.
e. Caveats and directions for future research
A significant limitation of the present study is the
relatively short analysis period, which is restricted to
eight summer seasons (six for the station data) by the
limited temporal coverage of the atmospheric datasets.
Use of longer datasets would increase the statistical significance of the results, particularly over the seasonalcycle and interannual timescales. The full range of interannual variability should be examined as well, not
just anomalies within individual seasons. Another drawback of this study is the subjective analysis of intraseasonal wet events. With some of the important largescale circulation mechanisms identified, it would be instructive to undertake a more objective diagnosis, perhaps through rotated EOFs or an objective scheme for
categorizing events. The incorporation of a more rigorous treatment of statistical significance would also be
worthwhile.
The influence of tropical disturbances, relative to extratropical disturbances, may be underestimated in this
study. An inherent bias arises since temperature and
geopotential height anomalies at higher latitudes are typically much larger than those in the Tropics and so are
more visible in the composite and case study analyses.
(The EOF analysis was performed on normalized geopotential height fields to reduce this bias.) This problem
is exacerbated by the difficulty in obtaining accurate
estimates of important tropical fields such as vertical
velocity.
Although the Bolivian high is clearly a feature of
tropical origin (Lenters and Cook 1997), this study suggests that much of its variability reflects extratropical
influences. Further exploration of these tropical–extratropical interactions would be relevant for understanding variability on a wide range of timescales. In addition,
placing the identified mechanisms in a more global context would be useful, such as examining potential links
with the 30–60-day (Madden–Julian) oscillation.
Finally, the physical mechanisms responsible for dry
conditions on the Altiplano during ENSO should be
more rigorously and thoroughly investigated. The preliminary explanation offered here attributes the cause
to frontal activity in the SACZ and, ultimately, to convection in the SPCZ. However, other mechanisms
should be explored as well, and over numerous ENSO
events. For example, the close proximity of the Altiplano to the ENSO-sensitive coastal areas of Peru suggests a potentially more direct link to El Niño. More
rigorous examination of teleconnections between the
VOLUME 127
SPCZ, SACZ, and other subtropical convergence zones
in the Southern Hemisphere would also contribute greatly to our understanding of the ENSO phenomenon.
Acknowledgments. The authors wish to thank three
anonymous reviewers, who provided valuable contributions to the revised manuscript. Some of the data used
in this study were produced through funding from the
Earth Observing System of NASA’s Mission to Planet
Earth in cooperation with the National Oceanic and Atmospheric Administration. The data were provided by
the Earth Observing System Data and Information System Distributed Active Archive Center at Goddard
Space Flight Center, which archives, manages, and distributes this dataset. Special thanks to Arlindo da Silva
of the Data Assimilation Office at GSFC for providing
the South American subset of the NASA/DAO data.
Precipitation data were provided by NCDC and the National Center for Atmospheric Research. Figures were
produced using the Grid Analysis and Display System
(GrADS), provided by the Center for Ocean–Land–Atmosphere Interactions at the University of Maryland.
This work represents a portion of J. D. Lenters’ Ph.D.
dissertation at Cornell University and was supported by
NASA Grant NAGW-2638 and NSF Grant ATM9300311.
REFERENCES
Aceituno, P., 1988: On the functioning of the Southern Oscillation
in the South American sector. Part I: Surface climate. Mon. Wea.
Rev., 116, 505–524.
, and A. Montecinos, 1993: Circulation anomalies associated
with dry and wet periods in the South American Altiplano. Preprints, Fourth Int. Conf. on Southern Hemisphere Meteorology
and Oceanography, Hobart, Australia, Amer. Meteor. Soc., 330–
331.
Berri, G. J., and J. Inzunza, 1993: The effect of the low-level jet on
the poleward water vapour transport in the central region of
South America. Atmos. Environ., 27A, 335–341.
DeMaria, M., 1985: Linear response of a stratified tropical atmosphere
to convective forcing. J. Atmos. Sci., 42, 1944–1959.
Gan, M. A., and V. B. Rao, 1991: Surface cyclogenesis over South
America. Mon. Wea. Rev., 119, 1293–1302.
Gandu, A. W., and J. E. Geisler, 1991: A primitive equations model
study of the effect of topography on the summer circulation over
tropical South America. J. Atmos. Sci., 48, 1822–1836.
Grimm, A. M., and P. L. Silva Dias, 1995: Analysis of tropical–
extratropical interactions with influence functions of a barotropic
model. J. Atmos. Sci., 52, 3538–3555.
Gutman, G. J., and W. Schwerdtfeger, 1965: The role of latent and
sensible heat for the development of a high pressure system over
the subtropical Andes, in the summer. Meteor. Rundsch., 18, 69–
75.
Horel, J. D., A. N. Hahmann, and J. E. Geisler, 1989: An investigation
of the annual cycle of convective activity over the tropical Americas. J. Climate, 2, 1388–1403.
Janowiak, J. E., and P. A. Arkin, 1991: Rainfall variations in the
tropics during 1986–1989, as estimated from observations of
cloud-top temperature. J. Geophys. Res., 96 (Suppl.), 3359–
3373.
Kalnay, E., K. C. Mo, and J. Paegle, 1986: Large-amplitude, shortscale stationary Rossby waves in the Southern Hemisphere: Ob-
MARCH 1999
LENTERS AND COOK
servations and mechanistic experiments to determine their origin. J. Atmos. Sci., 43, 252–275.
Kessler, A., 1981: Fluctuations of the water budget on the Altiplano
and of the atmospheric circulation (in German). Aachener Geographische Arbeiten, 14, 111–122.
Kiladis, G. N., and K. M. Weickmann, 1992: Circulation anomalies
associated with tropical convection during northern winter. Mon.
Wea. Rev., 120, 1900–1923.
Kodama, Y.-M., 1992: Large-scale common features of subtropical
precipitation zones (the Baiu Frontal Zone, the SPCZ, and the
SACZ). Part I: Characteristics of subtropical frontal zones. J.
Meteor. Soc. Japan, 70, 813–835.
, 1993: Large-scale common features of sub-tropical convergence zones (the Baiu Frontal Zone, the SPCZ, and the SACZ).
Part II: Conditions of the circulations for generating the STCZs.
J. Meteor. Soc. Japan, 71, 581–610.
Kousky, V. E., and M. A. Gan, 1981: Upper tropospheric cyclonic
vortices in the tropical South Atlantic. Tellus, 33, 538–550.
, and D. P. Casarin, 1986: Rainfall anomalies in southern Brazil
and related atmospheric circulation features. Proc. Second Int.
Conf. on Southern Hemisphere Meteorology, Wellington, New
Zealand, Amer. Meteor. Soc., 435–438.
, and A. Leetmaa, 1989: The 1986–87 Pacific warm episode:
Evolution of oceanic and atmospheric anomaly fields. J. Climate,
2, 254–267.
, and M. T. Kayano, 1994: Principal modes of outgoing longwave
radiation and 250-mb circulation for the South American sector.
J. Climate, 7, 1131–1143.
,
, and I. F. A. Cavalcanti, 1984: A review of the Southern
Oscillation: Oceanic-atmospheric circulation changes and related
rainfall anomalies. Tellus, 36A, 490–504.
Kutzbach, J. E., 1967: Empirical eigenvectors of sea-level pressure,
surface temperature and precipitation complexes over North
America. J. Appl. Meteor., 6, 791–802.
Lau, K.-M., and P. H. Chan, 1983: Short-term climate variability and
atmospheric teleconnections from satellite-observed outgoing
longwave radiation. Part I: Simultaneous relationships. J. Atmos.
Sci., 40, 2735–2750.
Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial
variability in gauge-corrected, global precipitation. Int. J. Climatol., 10, 111–127.
Lenters, J. D., 1997a: Climate dynamics of South America during
summer: Connections between the large-scale circulation and
regional precipitation. Ph.D. dissertation, Cornell University,
268 pp. [Available from J. D. Lenters, Center for Climate Research, Institute for Environmental Studies, University of Wisconsin—Madison, 1225 W. Dayton St., Madison, WI 537061695.]
, 1997b: Interannual variability of summertime precipitation on
431
the South American Altiplano: Connections with the large-scale
circulation. Preprints, 22d Conf. on Hurricanes and Tropical
Meteorology, Fort Collins, CO, Amer. Meteor. Soc., 457–458.
, and K. H. Cook, 1995: Simulation and diagnosis of the regional
summertime precipitation climatology of South America. J. Climate, 8, 2988–3005.
, and
, 1997: On the origin of the Bolivian high and related
circulation features of the South American climate. J. Atmos.
Sci., 54, 656–677.
Martin, L., M. Fournier, P. Mourguiart, A. Sifeddine, B. Turcq, M.
L. Absy, and J.-M. Flexor, 1993: Southern oscillation signal in
South American palaeoclimatic data of the last 7000 years. Quat.
Res., 39, 338–346.
Nobre, C. A., and A. S. de Oliveira, 1986: Precipitation and circulation anomalies in South America and the 1982–83 El Niño/
Southern Oscillation episode. Proc. Second Int. Conf. on Southern Hemisphere Meteorology, Wellington, New Zealand, Amer.
Meteor. Soc., 442–445.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions.
Mon. Wea. Rev., 110, 699–706.
Paegle, J., and K. C. Mo, 1997: Alternating wet and dry conditions
over South America during summer. Mon. Wea. Rev., 125, 279–
291.
Preisendorfer, R. W., 1988: Principal Component Analysis in Meteorology and Oceanography. Elsevier, 425 pp.
Quenouille, M. H., 1952: Associated Measurements. Academic Press,
242 pp.
Rao, G. V., and S. Erdogan, 1989: The atmospheric heat source over
the Bolivian plateau for a mean January. Bound.-Layer Meteor.,
46, 13–33.
Rao, V. B., and K. Hada, 1990: Characteristics of rainfall over Brazil:
Annual variations and connections with the Southern Oscillation.
Theor. Appl. Climatol., 42, 81–91.
Richman, M. B., 1986: Rotation of principal components. J. Climatol., 6, 293–335.
Satyamurty, P., C. C. Ferreira, and M. A. Gan, 1990: Cyclonic vortices
over South America. Tellus, 42A, 194–201.
Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated
dataset for earth science applications. Bull. Amer. Meteor. Soc.,
74, 2331–2342.
Silva Dias, P. L., W. H. Schubert, and M. DeMaria, 1983: Large-scale
response of the tropical atmosphere to transient convection. J.
Atmos. Sci., 40, 2689–2707.
Virji, H., 1981: A preliminary study of summertime tropospheric
circulation patterns over South America estimated from cloud
winds. Mon. Wea. Rev., 109, 599–610.
Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences.
Academic Press, 467 pp.