Journal of Hydrology 556 (2018) 1153–1170

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Multi-decadal 40- to 60-year cycles of precipitation variability in Chile

(South America) and their relationship to the AMO and PDO signals
Rodrigo Valdés-Pineda a,⇑, Julio Cañón b, Juan B. Valdés a
a
University of Arizona, Department of Hydrology and Atmospheric Sciences, 1133 E James E. Rogers Way, Tucson, AZ, USA
b
Grupo de Investigación en Gestión y Modelación Ambiental (GAIA), Facultad de Ingeniería, Universidad de Antioquia, SIU/UdeA, Calle 70 No. 52-21, Medellín, Colombia

a r t i c l e i n f o a b s t r a c t

Article history: High-frequency components of precipitation variability have been an important focus of study during the
Available online 1 February 2017 last few decades in Chile. Low-frequency variations, on the other hand, have received less attention, espe-
cially in association with multi-decadal cycles that can affect variability trends of precipitation in the long
Keywords: term. This study analyzes these low-frequency patterns of precipitation in Chile (>30 years), and their
Precipitation variability relationship to global Sea Surface Temperatures (SSTs), with special focus on associations with the
SSA Pacific Decadal Oscillation (PDO) and the Atlantic Multi-decadal Oscillation (AMO) indices. Singular
MSSA
Spectrum Analysis (SSA) and its Multi-Channel version (MSSA) techniques were applied to a dataset con-
Climate oscillations
PDO
taining long instrumental records of monthly precipitation aggregated yearly and seasonally. The rela-
AMO tionships between the low-frequency variability of precipitation and the PDO are significant to the
north of the country, whereas connections with the AMO are significant to the north and also to the
south. This is also evident from the global spatial correlation analysis of low-frequency precipitation
modes and SSTs, where the southernmost station shows a strong relationship with the Atlantic Ocean.
We conclude that a significant multi-decadal precipitation cycle between 40 and 60 years is evident at
the rain gauges located in the subtropical and extratropical regions of Chile. This low-frequency variabil-
ity seems to be largely linked to PDO and AMO modulation.
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction
In Chile, spectral analysis techniques have been used mainly to
observe and model paleoclimate records. Most of these studies
have shown that a substantial part of climate variability seems to
be controlled by inter-annual to decadal-scale oscillations. These
oscillations include, the 2.5-year Quasi-Biennial Oscillation (QBO)
(see Graystone, 1959; Ebdon, 1975; Baldwin et al., 2001; Dean
and Kemp, 2004), the 3- to 7-year El Niño Southern Oscillation
(ENSO) (Trenberth, 1984; Trenberth and Hoar, 1997;; Dawson
and O’Hare, 2000), and the 10- to 30-year Pacific Decadal Oscilla-
tion (PDO) (Mantua et al., 1997; Biondi et al., 2001; Deser et al.,
2010). For instance, Villalba et al. (1998) applied Singular Spectrum
Analysis (SSA) to develop long-term reconstructions (400 years) of
seasonal and annual precipitation variations for northern Patago-
nia (east of the Andes). The study used 22 precipitation records
(1912–1989) available between 39 and 43°S, and 27 chronologies
of Austrocedrus chilensis collected along the eastern slopes of the
Andean Patagonia. Oscillations observed in the power spectra of
⇑ Corresponding author.
E-mail address: rvaldes@email.arizona.edu (R. Valdés-Pineda).
tree ring chronologies, the precipitation records, and the recon-
structions were mainly associated with QBO. The precipitation
reconstructions also contained significant peaks in variance within
the frequency range of ENSO; however, their results did not reveal
a clear ENSO modulation of precipitation variability over the east
of the southern Andes.
In the Chilean Lake District (38–42°S), Fagel et al. (2008) applied
SSA to identify potential periodicities and reconstruct oscillatory
components of varve thickness over the last centuries. The proce-
dure analyzed several short and long annual laminated sediments
of Lago Peyuhue (40°S and 72°W), which is located in a region
where precipitation is often driven by the southern westerlies.
The authors found that frequency bands of varve thickness are
mainly related to the periodicities of the QBO, ENSO, and PDO. In
addition, Le Quesne et al. (2009) developed a multi-proxy approach
to analyze historical glacier variations in the Central Andes of Chile
and Argentina (33–36°S). They developed a tree-ring based multi-
variate model to reconstruct total annual precipitation of Santiago
de Chile (1866–1999). The periodic behavior of the reconstructed
precipitation time series was evaluated by extracting the dominant
oscillatory components using SSA. They found inter-annual oscilla-
tory modes (2.1–3.4 yr.) that explained 18.4% of the variance, and

http://dx.doi.org/10.1016/j.jhydrol.2017.01.031
0022-1694/Ó 2017 Elsevier B.V. All rights reserved.
1154 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
decadal/inter-decadal modes (13–23 yr.) that explained 18.5% of
the variance. Interestingly, both the inter-annual and decadal/
inter-decadal signals of the El Niño 3.4 Index (extracted by SSA)
showed better correlations with precipitation records than those
observed for the extracted PDO signals.
Aravena and Luckman (2009) analyzed the dominant spatial
and temporal patterns of a network of 23 homogenous instrumen-
tal precipitation records of southern South America (41–53°S),
finding mostly two- to seven-year oscillations for the Patagonia
Plains-Atlantic (43–50°S) and southern Patagonia (51–53°S), with
a strong biannual oscillation mode. Decadal oscillations
(
12 years) were observed for the first leading mode of northwest-
ern (41–44°S) and central Patagonia (45–47°S). The regional
annual precipitation was proposed to be strongly influenced by
Antarctic circulation modes (i.e., the Antarctic Oscillation Index
or AAO); whereas ENSO influence on precipitation variability was
less evident.
Rubio-Álvarez and McPhee (2010) studied the time series of
annual and seasonal streamflow of 44 rivers in southern Chile,
spanning the ecoregion between 34°S and 45°S (1952–2003). SSA
was applied to the original time series of composite streamflow
anomalies to obtain reconstructed time series and their peri-
odograms. Their results showed cycles of annual, seasonal, and
spatially averaged streamflows, covering a range between 2 and
11 years. The quasi-biennial and quasi-quadrennial periods were
associated with ENSO, while the six-year period was related to
the North Pacific Sea Surface Temperatures (SSTs). The authors also
indicated that longer, near-decadal periods (of approximately 7, 8,
9, and 11 years), could be related to the PDO and AAO oscillations.
Urrutia et al. (2011) developed a 410-year annual streamflow
reconstruction for the Maule River watershed (35°S–36°300 S, with
a climate like that of the Mediterranean) using Austrocedrus chilen-
sis tree ring chronologies. Their reconstruction revealed interan-
nual, interdecadal, and multi-decadal oscillation modes that were
well correlated with climatic indices. Positive correlations were
observed with the El Niño 3.4 Index (correspondence between high
tropical SSTs and high streamflows); and negative correlations
were observed with the AAO, indicating that above-average
streamflows are significantly related to the negative phase of the
AAO. This is the result of weakened sea level pressure (SLP) in
the mid-latitudes and intensified SLP in Antarctica. In this regard,
one of the most important detected oscillatory modes was an
approximately 18-year cycle (30% of explained variance), possibly
related to the 18.6-year cycle of the lunar tides. The other oscilla-
tory mode is a 50- to 60-year cycle, particularly significant after
1850, which could possibly be associated with the Half-
Gleissberg (70–100 years) solar magnetic cycle (see details in
Peristykh and Damon, 2003), and the PDO.
Although several studies have applied SSA to determine
oscillatory components in different regions of Chile, most of them
have been focused on the analysis of high- to mid-frequency
modes of precipitation, ranging particularly from quasi-biennial
to inter-decadal scales (2–30 years). Our study explores multi-
decadal patterns of precipitation (defined here as those fluctuating
over a period longer than 30 years) in Chile, by applying SSA and its
Multi-Channel version (MSSA). These analyses were applied to a
dataset of monthly series obtained from the longest instrumental
records of precipitation in Chile, a span between 1866 and 2014.
The degree of the relationship between low-frequency modes of
precipitation variability and the ENSO, AAO, and PDO signals is also
assessed. This study contributes to an increased knowledge of how
possible unknown multi-decadal modes of SST or SLP in the Pacific
Basin and other atmospheric teleconnections could explain low-
frequency variability of precipitation at an annual, seasonal, and
monthly scale.
2. Methods
2.1. Study area
Continental Chile is located in the Pacific coast of South America
between latitudes 17°300 S and 56°300 S (about 4300 km), and is
divided into 15 administrative regions. The geography of the coun-
try is mainly dominated by steep mountainous terrains, with only
around 20% of the continental territory being flat. The national ter-
ritory is composed of 101 main hydrologic basins, with more than
300 rivers flowing mainly from east to west. The Chilean territory
is also characterized by a wide variety of landscapes. From a geo-
morphological point of view, it is possible to distinguish four major
geographical units in the country: Andes Mountains (east), Inter-
mediate Depression, Coastal Mountains, and Coastal plains (west)
(see Valdés-Pineda et al., 2014).
Northern Chile (17°300 –30°S) is a hyper-arid region under the
persistent influence of the South East Pacific Anticyclone (SEPA),
which blocks the westerlies at low levels of the atmosphere as
far south as 27°S. The South American Monsoon System (SAMS)
brings easterly moisture from Amazonia (austral summer) that
generates rain shadow effect over the western Andes and Atacama
Desert – i.e., annual accumulation increases from about
20 mm year1 at 2700 m.a.s.l. to 300 mm year1 at 5000 m.a.s.l.
(Miller, 1976; Aceituno et al., 1993; Garreaud et al., 2003;
Houston, 2006; Minvielle and Garreaud, 2011). In central Chile
(30–40°S) the strong seasonality of precipitation results from the
wintertime retreat of SEPA, which allows stronger low-level west-
erlies to carry frontal systems to the continent (Montecinos and
Aceituno, 2003; Falvey and Garreaud, 2007). The annual accumula-
tion peaks during austral winter and increases southward from
around 100 mm at 30°S to nearly 2000 mm at 40°S (Quintana
and Aceituno, 2012). Due to the orographic effect of the Andes
Mountains, precipitation in central Chile also increases about twice
in the cross mountain direction (Falvey and Garreaud, 2007; Viale
and Garreaud, 2015). In southern and austral Chile (40°–56°300 S)
frontal cyclones drifting eastward along the oceanic storm tracks
are responsible for most of the precipitation and day-to-day
weather variations (Quintana and Aceituno, 2012; Garreaud
et al., 2014; Valdés-Pineda et al., 2015). This translates into abun-
dant precipitation accumulation in all seasons, reaching more than
4000 mm year1 along the western Andes (Carrasco et al., 2002;
Villarroel, 2013). In general, precipitation accumulation in Chile
tends to increase with latitude and altitude (Pizarro et al., 2012;
Quintana and Aceituno, 2012). The annual progression of this lati-
tudinal gradient shows perfect links with the interaction between
westerlies and SEPA (Fig. 1a). For instance, the band of westerlies
displaces southward during austral summer (strengthened SEPA)
(Fig. 1b and e), and moves northward during austral winter (weak-
ened SEPA) (Fig. 1c and d).
These patterns of precipitation variability are originated by the
activity of SST in the equatorial Pacific, which indeed determines
modes of variation such as ENSO and the PDO. For instance,
research from the last few decades has established that inter-
annual precipitation variability in Chile is related to ENSO
(Rutllant and Fuenzalida, 1991; Aceituno and Garreaud, 1995;
Garreaud and Batisti, 1999; Montecinos and Aceituno, 2003;
Garreaud et al., 2009), playing a major role during the austral win-
ter (Valdés-Pineda et al., 2016). The decadal to inter-decadal vari-
ability of precipitation in the country has been mainly associated
with the PDO (Mantua et al., 1997; Montecinos and Aceituno,
2003; Garreaud et al., 2009; Quintana and Aceituno, 2012). In
Chilean Patagonia another source of precipitation variability is
the AAO, which is characterized by pressure anomalies of one sign
centered in the Antarctic and anomalies of the opposite sign on a
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
Fig. 1. Spatial distribution of Averaged Annual Precipitation accumulation (mm) for 238 ground stations (1979–2012) distributed in the country and 925-hPa zonal and
meridional wind (m/s) components (U925; V925). Source: Adopted from Valdés-Pineda et al. (2016).
circumglobal band, at about 40°–50°S (Garreaud et al., 2009;
Quintana and Aceituno, 2012; Villalba et al., 2012).
2.2. Dataset: Precipitation Records, climate Indices, and gridded fields
Since our interest was essentially focused in detecting low-
frequency temporal variability (>30 years), only instrumental
records of more than 100 years were used for SSA and MSSA anal-
ysis. The meteorological yearbooks of the Meteorological Direc-
torate of Chile (DMC) were reviewed to obtain the longest
records of the country. The DMC’s database is digitally available
only from 1920 to the present (http://164.77.222.61/climatolo-
gia/). Earlier digital records were acquired from the Global Histor-
ical Climatology Network database (GHCN-Monthly) of the U.S.
National Oceanic and Atmospheric Administration (NOAA). This
database contains historical temperature, precipitation, and pres-
sure data for thousands of land stations worldwide (ftp://ftp.
ncdc.noaa.gov/pub/data/ghcn/v2/). After checking all available
data, only 12 rain gauges, evenly distributed between 26° and
53°S, were selected (see Fig. 2). These records, spanning between
1866 and 2014, represent an important collection for southern
South America (Chile) since the current availability of digital cli-
mate data over many parts of the world is only restricted to the
second half or the last third of the 20th century (Brunet and
Jones, 2011). Other long instrumental records have already been
1155
established in the United States (see Kunkel et al., 1999) and Eur-
ope (see Zolina et al., 2009).
The rain gauges located in northern regions were not included
in the analysis because they have a significant number of zeros
in the monthly and annual accumulation records that disturb the
implementation of the methods used in this study. Only one rain
gauge containing less than 100 years’ of records (Puerto Aysen at
about 45°S) was included in the analysis (see Table 1) because it
presents important information for decision-making about ongo-
ing large-scale hydroelectricity projects. The records of Valdivia,
which date back to 1853 (see Sanz Donaire, 2012), were not
included in the analysis because from 1970 onwards another rain
gauge located 15 km away was used to continue the suspended
records of Valdivia (National Directorate of Water Resources of
Chile, DGA, personal communication, 2015).
Four climate indices and gridded data from NOAA Earth System
Research Laboratory (ESRL) Physical Sciences Division (PSD) were
included in our analysis to verify their degree of relationship to
low-frequency precipitation variability: (1) the monthly Southern
Oscillation Index (SOI1) (1893–2014) obtained from NCAR-NOAA;
(2) the monthly Antarctic Oscillation Index (AAO2) (1979–2014)
1
SOI available at: http://www.cgd.ucar.edu/cas/catalog/climind/SOI.signal.ascii.
2
AAO available at: http://www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_
ao_index/aao/monthly.aao.index.b79.current.ascii.table.
1156 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170

Fig. 2. Time series of annual accumulation of precipitation (mm) in Chile for records longer than 100 years.

Table 1
Sampling stations analyzed in the study area.

Map number Station/location Institution Longitude Latitude Region Period Years

1 Copiapó DGA 70.33 27.38 Atacama 1897–2014 118
2 La Serena DGA 71.26 29.91 Coquimbo 1869–2014 146
3 Ovalle DGA 71.20 30.60 Coquimbo 1912–2014 103
4 Los Andes DGA 70.60 32.84 Valparaiso 1907–2014 108
5 Santiago DMCNOAA 70.66 33.46 Metropolitana 1866–2014 149
6 Rancagua DGA-NOAA 70.73 34.17 O’Higgins 1910–2014 105
7 Talca DGA-NOAA 71.66 35.42 Maule 1907–2014 108
8 Concepcion NOAA 73.04 36.82 Biobio 1912–2014 103
9 Temuco NOAA 72.59 38.75 Araucanía 1912–2014 103
10 Puerto Montt NOAA 72.94 41.46 Los Lagos 1888–2014 138
11 Puerto Aysen DGA-NOAA 72.70 45.40 Aysen 1931–2014 84
12 Punta Arenas DGA-NOAA 70.88 53.13 Magallanes 1888–2014 127

acquired from NCEP-NOAA; (3) the monthly Pacific Decadal
Oscillation (PDO3) (1900–2014) obtained from the University of
Washington; and (4) the Atlantic Multidecadal Oscillation (AMO4)
(1856–2014) obtained from NCEP-NOAA. For further research and
discussion we incorporated SST, precipitable water (PW), and zonal
winds (ZW) gridded datasets as part of our study.
3
PDO available at: http://jisao.washington.edu/pdo/PDO.latest.
4
AMO available at: http://www.esrl.noaa.gov/psd/data/timeseries/AMO/.
2.3. Time series analysis of precipitation records
Total annual accumulation at each raingauge was obtained by
aggregating all monthly accumulations of each calendar year. The
seasonal series for dry (Oct.-Mar.) and wet (Apr.-Sept.) periods
were split according to the Chilean hydrologic year. These time ser-
ies were then standardized in order to give equal weight to all sta-
tions and avoid absolute-value bias (Wilks, 2011). Once all the time
series were organized, we applied time-series analysis (TSA) tech-
niques to investigate temporal patterns of precipitation. TSA is
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
commonly used to detect all essential features of the records, i.e.,
normality, stationarity, homogeneity, presence/absence of trends,
and persistence (Machiwal and Jha, 2012; Machiwal et al., 2015).
We also used the Autocorrelation Function (ACF) at different lags
(Box and Jenkins, 1970; Box et al., 1994), to determine the degree
of temporal dependency among precipitation observations at
annual, seasonal, and monthly scales. The ACF was also useful to
identify patterns in terms of trends, seasonality, and periodicity
(i.e. Huesca et al., 2015).
2.4. Low-frequency modes of precipitation variability
2.4.1. Singular Spectrum Analysis (SSA)
Taking into account the consistent length of the period of
records, we searched for significant low-frequency cycles that
could explain precipitation variability at a multi-decadal scale. In
this regard, SSA was applied to each individual annual, seasonal,
and monthly series of precipitation and the climate indices: the
SOI, AAO, PDO and AMO. The details and applications of SSA to
hydroclimatic series can be found in the last two decades of liter-
ature (i.e. Weare and Nasstrom, 1982; Broomhead and King, 1986a,
b; Fraedrich, 1986; Vautard and Ghil, 1989; Ghil and Vautard,
1991; Ghil et al., 2002; Marques et al., 2006; H. Wang et al.,
2015; R. Wang et al., 2015); therefore, they are not described in
this study.5 The approach is essentially a principal component (PC)
analysis in the time domain which allows for the study of separated
spectral contributions of the main oscillatory modes (see Vautard
and Ghil, 1989; Ghil et al., 2002; Golyandina et al., 2001; Hassani
et al., 2009, 2015; Hassani and Thomakos, 2010; Unnikrishnan and
Jothiprakash, 2015; Silva and Hassani, 2015).
There are two important parameters to take into account when
using SSA to extract each source signal, i.e., a suitable selection of
the window length (L) for constructing the trajectory matrix, and
the number of components or eigenvalues (r) to be retained (H.
Wang et al., 2015; R. Wang et al., 2015). The range of the window
length L is usually in 2 6 L 6 N/2 (H. Wang et al., 2015; R. Wang
et al., 2015). A discussion about the L parameter is provided by
Elsner and Tsonis (2013) who remarked that choosing L = N/4 is
a common practice. Others authors suggest that L should be large
enough, but not larger than N/2 (see Golyandina et al., 2001; Mah-
moudvand and ZokaeiWe, 2012; Golyandina and Zhigljavsky,
2013). In general, large values of L allow longer period oscillations
to be resolved, but choosing an L that is too large leaves too few
observations from which to estimate the covariance matrix of the
L variables.
In this study, the L parameter was fixed to one-half of the time
series length, since it has been shown to be a suitable size for the
decomposition stage of low frequency patterns (Mahmoudvand
and ZokaeiWe, 2012). For instance, the L parameter for a 100-
year series was 50 years. Additionally, in order to quantify the peri-
odicities of the cycles associated to low-frequency variability, the
first 10 leading modes (PCs) were retained and their significances
calculated according to North (1982) (Fig. 3a). Then the periodici-
ties of each PC were separately calculated using periodogram anal-
ysis as described in Ghil et al. (2002) and Chatfield (2013) (Fig. 3b).
Those series with largest periodicities (number of years or months)
as represented by the highest significant spectral peaks were
selected as low-frequency cycles for further analysis. The Multi
Taper Method (MTM) was additionally applied to verify results
from periodogram analysis (see Rao and Hamed, 2003). This
method reduces the variance of spectral estimates by using a small
set of tapers rather than the unique data taper or spectral window
used by classical methods as Fourier analysis (Ghil et al., 2002;
Golyandina and Zhigljavsky, 2013).
5
See Appendix for more details about SSA and MSSA.
1157
2.4.2. Multi-Channel Singular Spectrum Analysis (MSSA)
MSSA approach was useful for obtaining a common structure of
several series of precipitation and climatic oscillations (see i.e.
Cañón et al., 2007), and also for establishing causality between
the different series. The absence of causality between time series
(i.e. X and Y) implies that the knowledge of some predictor (Y ser-
ies) does not improve the quality of forecasts (Hassani et al., 2010),
while knowledge of some climate oscillation (X series) does not
improve the quality of precipitation predictions. MSSA is a natural
extension of SSA for analyzing multivariate time series. The algo-
rithm is almost the same as in the univariate mode; however, in
this approach the Henkel or Trajectory Matrix of the multivariate
time series consists of stacked trajectory matrices of separate time
series. MSSA also was applied to all rain gauge records in combina-
tion only with the SOI, PDO, and AMO (1912–2014) since they have
much longer records than AAO (see Section 2.2). The Grand Data
Matrix that was assembled as the Hankel Matrix (see Partington,
1988; Makowsky and Labai, 2015) was finally arranged containing
records of precipitation and climate indices. To be consistent with
SSA analysis, the L parameter was again fixed to one-half of the
time series length (L = N/2). This method has gained global popu-
larity during the last decade. For instance, Krishnamurthy and
Misra (2010) applied MSSA to a climatic series of daily outgoing
longwave radiation in the South American Monsoon System.
Recently, Zotov et al. (2015) applied MSSA to filter data from the
Gravity Recovery and Climate Experiment (GRACE) and separate
its principal components (PCs) at different periodicities over the
15 largest river basins of Russia. Despite this, no studies in Chile
have used this multi-channel version of the analysis, at least using
climate series. However, MSSA has many applications and it is
especially popular in analyzing and forecasting economic and
financial time series (Patterson et al., 2011; Hassani et al., 2013;
Hassani and Mahmoudvand, 2013), and for determining spatio-
temporal patterns using climate series (i.e. Krishnamurthy and
Stan, 2015; Krishnamurthy and Krishnamurthy, 2015).
2.5. Cross correlations between multi-decadal modes of precipitation,
climate indices and gridded fields
The large number of interactions comprising the complex
ocean-atmospheric system can be determined by analyzing the
cross-correlations between climate signals, in order to better diag-
nose and understand the whole system. Because of its simplicity,
traditional cross-correlation analysis (TCA) has become the most
widely used method for this purpose (Yuan et al., 2015). Especially
in climatology research, TCA is used in various fields, including
dynamic diagnosing and climate forecasting (Yuan et al., 2015).
In accordance with this, we analyzed the degree of the relationship
between precipitation variability and climate indices by imple-
menting traditional (and lagged) cross-correlation analysis (see
details in Bendat and Piersol, 2011). For instance, the reconstructed
low-frequency modes of precipitation at each rain gauge (RPC), cal-
culated from either SSA or MSSA, and the modes of climate indices
were used as predictors of annual, seasonal, and monthly accumu-
lation variability. The same analysis was performed to derive sig-
nificant correlation patterns between the low-frequency modes
of precipitation and SST, and also extended in the discussion sec-
tion to determine how this multi-decadal signal, mostly observed
in the ocean, can be easily detected at some atmospheric levels
of PW and ZW.
The idea of applying time-lagged cross-correlation analysis was
to identify significant lagged predictors influencing the low-
frequency precipitation variability in the country. This analysis
was conducted through the use of lagged scatterplots. One attri-
bute of the lagged scatterplots is that they can show if the cross-
correlation is representative of the majority of the data or if it is
1158 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
Fig. 3. (a) Eigenvalue spectrum for the Trajectory (Henkel) matrix of annual time series of annual precipitation in Copiapó for the period 1897–2014 (L = 59 years). The error
bars for the eigenvalues were defined according to North et al. (1982). (b) Periodogram calculated for the First Reconstructed Principal Component (RPC1) of Copiapó (
27°S
Lat.).
driven by one or more outliers (Chatfield, 2013). When the time
series is completely random, and the sample size is large, the cross
lagged-correlation coefficient is approximately normally dis-
tributed with the mean 0 and the variance 1/N. Therefore the sig-
nificance for the cross-correlation analysis was calculated
assuming that the approximate critical level of correlation at 95%
pffiffiffiffi
significance (a = 0.05) is r95 ¼ 0 þ 2= N; where N is the sample
size. According to this criterion, the required level of correlation
for significance becomes very small in a large sample size
(Chatfield, 2013).
3. Results and discussion
3.1. Leading modes of precipitation and explained variance
Most of the significant low-frequency oscillatory components of
precipitation variability were identified within the first five leading
modes of annual, seasonal, and monthly precipitation. The sum of
these five leading modes can explain annual precipitation variance
ranging from 20.3% (Santiago) to 51% (Puerto Aysen), with an aver-
age of 31.2% when considering all the rain gauges analyzed
(Table 2). These explained variances are slightly smaller than those
calculated by Aravena and Luckman (2009) for the first four lead-
ing modes of annual regional averaged indices of precipitation in
the Chilean Patagonia (41–53°S, between 1961–2000); but this is
expected, given the different periods and aggregations analyzed.
For instance, the limited length of the PC series used by Aravena
and Luckman (40 years) only allowed for the identification of fre-
quency oscillations <20 years, which limited the ability to detect
the signal of multidecadal climatic variations such as the PDO or
AMO.
3.2. Periodicities of the leading modes
The periodicities calculated for the leading modes of annual,
seasonal, and monthly precipitation ranged mainly from biennial
to multi-decadal oscillations. For instance, 79.1% of the first five
leading modes of annual precipitation (see Table 2), are dominated
by high-frequency oscillations (2–9 years) which are expected to
be associated to QBO and ENSO (see Fig. 4a) as stated in previous
studies for different regions of Chile (Villalba et al., 1998; Fagel
et al., 2008; Le Quesne et al., 2009). In addition, 11.7% of these
annual leading modes revealed oscillations between 9 and
30 years, which are expected to be connected primarily with the
decadal to inter-decadal modulation of the PDO (Fagel et al.,
2008; Masiokas et al., 2010). The remaining 9.2% of the annual
leading modes revealed multi-decadal oscillations (>30 years) that
can be linked to the multi-decadal signal of the PDO, as it can vary
irregularly from inter-annual to inter-decadal time scales (Mantua
et al., 1997; Deser, 2010; Dai, 2013; Dong and Dai, 2015). No stud-
ies have reported a possible influence of the AMO signal in Chile;
however, as revealed in the following sections of this study the
multi-decadal modulation of the North Atlantic Ocean also influ-
ences low-frequency precipitation variability in Chile. In addition,
as indicated by Urrutia et al. (2011), the Half-Gleissberg solar mag-
netic cycle could be the external forcing influencing the multi-
decadal variability of ocean-atmosphere processes. The results
found at an annual scale agree with those observed for monthly
precipitation (Fig. 4b), and they are also consistent during wet
(Fig. 4c) and dry (Fig. 4d) seasons.
3.3. Phase-space representation
Most of these frequencies can be explained by looking at the
phase diagrams that represent the trajectories of coupled PCs
and are commonly found within the first few leading modes of
variability (Fig. 5). This is designed to seek not periodicities, but
rather repeated trajectories in the phase space (Kim et al., 2015).
The PCs are related in time as they trace out oscillations in the
precipitation time series. The low- and high-frequency cycles of
precipitation are characterized by different trajectories. Multi-
decadal oscillations (>30 years), for instance, are represented by
circular trajectories. High-frequency cycles are represented by dif-
ferent geometries depending on the periodicities of the signal i.e.,
triangles (
3-yr cycle) and pentagons (5-yr cycle). Well-defined
trajectories in the phase-space are revealed at the annual scale
since the noise level is smaller relative to the precipitation signal
and vice versa. The irregularities observed in some plots can be
described mostly by inter annual variability (Hannachi, 2004).
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170 1159

Table 2
Explained variance (%) of the first five leading modes of annual precipitation obtained by SSA (L = N/2).

Station/location RPC1 RPC2 RPC3 RPC4 RPC5 Total Variance (%)

% Years % Years % Years % Years % Years
1. Copiapó 10.34(S) 60 10.18(S) 60 5.93(S) 12 5.61(S) 11 4.04 10 36.1
2. La Serena 4.68 73 4.49 2.3 4.38 2.3 4.31 2.3 3.68 2.8 21.5
3. Ovalle 6.43 3.9 5.96 3.9 5.12 2.5 4.91 2.5 4.31 5 26.7
4. Los Andes 6.76 2.5 6.66 2.5 5.57 3.9 5.45 3.9 5.08 2.3 29.5
5. Santiago 4.48 3.8 4.47 3.8 3.83 2.4 3.81 4.4 3.75 4.4 20.3
6. Rancagua 6.36 2.5 5.68 2.5 5.67 2.4 5.46 2.3 4.95 10.7 28.1
7. Talca 6.02 2.2 5.44 2.2 5.12 10.7 5.1 10.7 4.24 6.4 25.9
8. Concepcion 6.54(S) 3 6.38(S) 3 4.68 6.7 4.64 6.7 4.32 2.5 26.6
9. Temuco 8.94(S) 51 5.83(S) 6 5.72(S) 6 5.46(S) 4.5 5.42(S) 4.5 31.4
10. Puerto Montt 9.15 54 7.98 54 6.72 4.5 6.54 4.5 5.0 3.8 35.4
11. Puerto Aysen 14.69(S) 42 14.5(S) 42 8.17(S) 42 7.11(S) 14 6.51(S) 8.4 51.0
12. Punta Arenas 13.49(S) 42.7 11.85(S) 42.7 5.95(S) 7.1 5.66(S) 7.1 4.58(S) 2.56 41.5

(S) Represent significant modes calculated according to North (1982).

Fig. 4. Annual, monthly, and seasonal (wet and dry) histograms of the periodicities extracted for the first five leading modes of precipitation at each raingauge. The blue boxes
highlight high-frequency oscillations as those related to QBO and ENSO (2–9 years). The green boxes are low-frequency oscillations likely associated to the PDO (10–
30 years). The red boxes represent multi-decadal oscillations, the focus of analysis in this study. (For interpretation of the references to colour in this figure legend, the reader
is referred to the web version of this article.)

3.4. Low-frequency modes of precipitation
3.4.1. Annual and monthly scales
The low-frequency modes of annual and monthly precipitation
(>30 years) explain the largest variance at those rain gauges
located in the northern and southern regions of the country
(Fig. 6). For instance, the RPC1 of annual precipitation variability
in Copiapó, located in northern Chile at around 27°S, reveals a sig-
nificant semi-centennial cycle fluctuating between 54 and 64 years
that explains 10.3% of the total variance. This rain gauge is placed
in the southern part of the Atacama Desert, where precipitation is
largely controlled by the sub-tropical high-pressure belt
(15°–30°S) (Houston, 2006). The upwelling cold Peruvian Current
inhibits the moisture capacity of onshore winds (below 1000 m.a.
s.l), and the proximity of the Andean Cordillera restricts moisture
advection from the east (Houston, 2006). Therefore, winter precip-
itation originates largely from northerly and easterly moving fron-
tal systems originating from the Pacific (Vuille and Ammann,
1997). Surprisingly, this multi-decadal pattern was also significant
at those rain gauges located along the westerlies wind belt located
between 40° and 55°S. As a result, the RPC1 of annual accumula-
tion in Puerto Aysen revealed a marked 42-year cycle that
explained 14.3% of the variance. Punta Arenas showed two signif-
icant leading modes (RPC1 and RPC2) that explain 20.5% of the
total variance and evolve in a cycle fluctuating between 42 and
64 years. In these regions, high-frequency precipitation is largely
influenced by the Southern Annular Mode (SAM) or AAO, in
combination with ENSO (Aravena and Luckman, 2009; Elbert
et al., 2011). However, the regime is quite variable, both in space
and time, because of the topography of the Andes and changes in
the strength and latitudinal position of the southern Westerlies
(Garreaud et al., 2009). For the same region of southern
South-America, Aravena and Luckman (2009) detected frequencies
of fewer than 15 years, observing decadal variations (
12 years) in
northwestern (41–44°S) and central Patagonia (45–47°S); and
2–7 year cycles with a strong biennial oscillation mode in the
Patagonia Plains-Atlantic (43–50°S) and southern Patagonia
(51–53°S).
1160 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170

Fig. 5. Phase-space representation of different eigencouples representing trajectories for low and high-frequency oscillations in long rain gauge records of Chile. The
explained variance of each eigencouple is represented as percentage and the periodicity of the signal is stated in parentheses.

Fig. 6. Reconstructed leading modes of annual and monthly precipitation.

3.4.2. Wet and dry seasons
The rain gauges shown in Fig. 6 contain a notable multi-decadal
cycle, which is also apparent at seasonal scale. The low-frequency
RPC1 of those precipitation records is statistically significant in both
wet (Apr.-Sept.) and dry (Oct.-Mar.) seasons, being slightly stronger
during the wet periods (Fig. 7). The RPC1 observed for those rain
gauges located in central Chile revealed mostly high-frequency
oscillations (2–9 years), which cannot be attributed to the PDO or
AMO modulation. For instance, a significant biennial cycle was
revealed for the RPC1 of Santiago during dry seasons, which suggests
high ENSO-modulation (Rutllant and Fuenzalida, 1991; Aceituno
and Garreaud, 1995; Garreaud and Batisti, 1999; Montecinos and
Aceituno, 2003; Valdés-Pineda et al., 2016). The tropical and extra-
tropical factors affecting precipitation in central Chile are not inde-
pendent. Thus, the positive (negative) phase of the Southern Oscilla-
tion and a strengthened (weak) subtropical high tend to be associated
with an increased (reduced) baroclinicity at mid-latitudes in the
southeast Pacific (Quintana and Aceituno, 2012). Our results agree
with those found by Núñez et al. (2013) who showed that changes
in precipitation in central Chile (30–36°S) are not strongly related
to shifts between the warm and cold phases of the PDO.
3.4.3. Climate indices
The low-frequency signals of the SOI and AAO did not share
similarities with the low-frequency cycles of precipitation
observed in the extreme arid (north) and cold (south) regions of
Chile. This pattern is consistent at annual, seasonal, and monthly
scales. The periodicities calculated for the low-frequency signal
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
Fig. 7. Reconstructed low-frequency oscillations for the leading modes of precipitation during wet (Apr.-Sept.) and dry (Oct.-Mar.) seasons.
(RPC1 + RPC2) of mean annual SOI (Fig. 8a) ranged between 5.8 and
6.8 years, with a total explained variance of 5.9% (Fig. 8b and c);
and the low-frequency component of the mean annual AAO ranged
between 5.1 and 6.0 years with an explained variance of 14.7%.
These results are consistent with those described in Section 1 of
this paper i.e., studies in which ENSO and AAO activity are strongly
modulated by high-frequency modes of variability, which are not
the focus of this study.
The first two reconstructed leading modes (RPC1 and RPC2) of
the PDO (Fig. 8j) reveal a significant multi-decadal 51- to 58-year
cycle with a total explained variance of 33.2% (Fig. 8k and l) which
is consistent at annual, seasonal, and monthly scales. This lower
frequency component of the PDO, with its characteristic multi-
decadal timescale of about 60 years (d’Orgeville and Peltier,
2007), has indicated regime shifts associated to SSTs that can range
between 50 and 80 years (see i.e. Minobe, 1997; Lee et al., 2013;
Steinman et al., 2015). A 20-year quasi-periodic oscillation
observed in RPC5 and RPC6 is also suggested in previous studies
(Minobe, 1997; d’Orgeville and Peltier, 2007), as well as a quasi-
30- to 35-year cycle observed in the RPC9, which is not significant
under this SSA (L = N/2). This cycle, however, could be significantly
related to inter-decadal low-frequency variability in north-central
Chile as indicated by Núñez et al. (2013).
The existence of multi-decadal cycles was first evinced by
Schlesinger and Ramankutty (1994), who detected a 65- to 70-
year cycle for global mean temperature records. In addition, SST
records of 11 globally distributed geographical regions showed that
this multi-decadal 50- to 90-year cycle is related to the first Empir-
ical Orthogonal Function(EOF)of SST anomalies in the North Atlantic
region, known as the Atlantic Multi-decadal Oscillation (AMO)
(Knight et al., 2005; Justino and Peltier, 2005; Trenberth and Shea,
2006). Specifically, positive (negative) phases of the AMO coincide
with warmer (colder) North Atlantic SSTs (Enfield et al., 2001;
McCarthy et al., 2015). The leading low-frequency modes of AMO
were calculated to determine possible unrevealed linkages between
the multi-decadal oscillations of precipitation observed over the
extreme arid and cold regions of Chile. Results reveal that RPC1
+ RPC2 of AMO (Fig. 8m) evolve in a significant multi-decadal 50-
to 80-year cycle, which explains more than 50% of the total variance
(Fig. 8n and o). The determination of common oscillatory signals
1161
between the multi decadal modes of precipitation and climate
indices using MSSA is detailed in the following section.
3.5. Multi-channel analysis of precipitation variability
Compared to the results from using SSA, the determination of
common oscillatory signals between precipitation and climate
indices using MSSA was associated with lower variances of the
leading modes at all temporal scales. This is due to the fact that
precipitation is easily influenced by local factors, which translates
into more complex spatio-temporal variability within the matrix
containing the time series (Antunes et al., 2006; Gomes, 2011).
The first three low-frequency modes (L = 52 years) associated with
the matrix – which includes all rain gauges, the SOI, PDO, and AMO
(1912–2014) – revealed significant explained variances of 4.7%
(RPC1), 3.6% (RPC2), and 3.5% (RPC3). The low values observed in
these multi-decadal modes are in central Chile, which is mostly
dominated by high-frequency ENSO modulation. Weighted cross-
correlations between normalized multi-decadal components of
precipitation and climate indices confirm positive relationships
between pairs of multi-decadal modes extracted from those rain
gauges located in central Chile. These normalized modes are posi-
tively correlated with the PDO and negatively correlated with the
SOI and AMO (Fig. 9a), suggesting possible large-scale multi-
decadal patterns that affect precipitation variability in central
Chile. The multi-decadal cycles, however, are much stronger and
more significant in the extreme regions of the country. Therefore,
they constitute a primary basis upon which to explain the past,
present, and future periodicities of extreme hydrologic events in
those regions of Chile.
When only Copiapó and Punta Arenas are considered in the
analysis with the SOI, PDO, and AMO (L = 52 years), the explained
variances doubled compared to the previous results, and the
multi-decadal modes are easily found within the first two signifi-
cant components (Fig. 9b). In this regard, the periodicity of the
PDO signal indicates a clear multi-decadal cycle that evolves in an
almost identical phase to the multi-decadal cycle of precipitation
extracted for the Atacama Region’s rain gauge (Fig. 9c). The
amplitude of the multi-decadal PDO signal is greater than the ampli-
tude of the AMO and SOI. Therefore, these three large-scale patterns
1162 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170

Fig. 8. First two reconstructed components (first column); eigenvalues with phase diagram (second column); and power spectra of first two reconstructed components (third
column) of annually averaged time series of Copiapó (1897–2014), Punta Arenas (1888–2014), the SOI (1866–2014), the PDO (1900–2014), and the AMO (1860–2014).
Eigenvalue error bars are defined according to North et al. (1982). The black line represents the original time series of Copiapó, Punta Arenas, the SOI, PDO, and AMO, and the
red line represents the low-frequency RPC1 + RPC2 obtained from SSA with L = N/2. (For interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)

should have different levels of influence over long-term precipita-
tion variability in arid (i.e., Copiapó) and cold (i.e., Punta Arenas)
regions of Chile. For instance, if both ENSO and the PDO are in the
same phase, it is suggested that El Niño/La Niña impacts can be
intensified. Conversely, if ENSO and the PDO are out of phase, they
may offset one another, preventing the impacts of a strong ENSO.
3.6. Phase-space of multi-decadal modes
The evolution between each pair of multi-decadal cycles of pre-
cipitation and climate indices can be defined as a dynamic system.
This system is described by trajectories such as those represented
by phase-space diagrams (see i.e. Sivakumar et al., 2002; Nolte,
2010; Kim et al., 2015). These trajectories reveal significant offsets
i.e., the trajectory representing the relationship between each pair
of multi-decadal modes showed that the PDO (RPC1 + RPC2) is sig-
nificantly but slightly out of phase (
4-year-lagged) with the series
of Copiapó (Fig. 10a). Surprisingly, the multi-decadal AMO signal
(RPC1 + RPC2) shows a strong negative (
20-year-lagged) offset
from the same multi-decadal component (Fig. 10b). For Punta Are-
nas (RPC1 + RPC2) the positive-lagged PDO relationship is less evi-
dent (
10-year-lagged), since it shows a weaker linkage when
outlining the trajectory of both components (Fig. 10c). The AMO
signal indicates phase shifts of about 25 years, which are weaker
than the multi-decadal mode of Punta Arenas (Fig. 10d). Moreover,
the phase-space representation of both rain gauges did reveal a
relationship to a small degree, and the most notable pattern is
essentially that of Copiapó offsetting Punta Arenas by about
12 years (Fig. 10e and f).
Keeping in mind these previous relationships, it is important to
add that on the timescale of the PDO series (1900–2014), AMO is
shown to lag the PDO by
17 years (Fig. 10g) and to lead the
PDO by approximately 13 years (Fig. 10h), and. These latter results
are in agreement with those found by d’Orgeville and Peltier
(2007), who stated that this relationship between the AMO and
the 60-year component of the PDO are signatures of the same oscil-
lation cycle.
3.7. Spatial significances of multi-decadal modes
The projection of the multi-decadal signals of the PDO and AMO
(MSSA, L = 52 years) onto grids of global SSTs revealed unprece-
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170 1163

Fig. 9. (a) Matrix of w-correlations for the 15 reconstructed components of annual precipitation, the SOI, PDO, and AMO. (b) Eigenvalue spectrum for the Trajectory (Henkel)
matrix of annually averaged time series of Copiapó, Punta Arenas, the SOI, PDO, and AMO, obtained from MSSA analyses. The error bars for the eigenvalues were defined
according to North et al. (1982). (c) Multi-decadal modes (RPC1 + RPC2) of Copiapó, Punta Arenas, the SOI, PDO, and AMO.

Fig. 10. Phase-space diagrams representing the trajectory between the multi-decadal (low-frequency) modes of annual precipitation in Copiapó (
27.4°S) and Punta Arenas
(
53.1°S), and the multi-decadal signals the PDO and AMO.

dented ocean-climate linkages for Chile. We observed that the
multi-decadal PDO signal projected onto global SSTs resulted in
two significant bands concentrated over low (strong pattern) and
high (weak pattern) latitudes of South America; this suggests that
large-scale precipitation variability at annual and seasonal scales
(wet and dry conditions) is driven by a significant multi-decadal
dipole warming over the Pacific Ocean (Fig. 11a). As expected from
previous results, this spatial dipole was also found (at all temporal
scales) when we projected the multi-decadal mode of Copiapó onto
global SSTs (Fig. 11b). The spatial distribution of this signal showed
to be significantly stronger during dry conditions (see Appendix).
The multi-decadal warming over the North Atlantic Ocean denotes
global AMO-like significant patterns. This low-frequency signal,
which is significantly strong over the North Atlantic, contrasts with
1164 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
a weakened and not significant pattern over the eastern Equatorial
and subtropical Pacific Ocean (Fig. 11c). These weaker conditions
over the eastern Pacific basin are not evident over the western
Pacific, which reveals significant patterns concentrated over low
latitudes that can be associated with enhanced precipitation condi-
tions over northeast Australia (Sun et al., 2015) and Southeast Asia
(i.e., Indonesia, the Philippines, China, Japan, etc.) (see Lu et al.,
2006; Zhang and Delworth, 2006; Li et al., 2015). This pattern also
reveals two elongated significant bands of AMO that extend
towards mid- and high latitudes of the eastern Pacific at both
hemispheres. This ocean-connection (warm AMO) means more
precipitation along the northwestern United States (see Enfield
et al., 2001; McCarthy et al., 2015), but it also shows clear influ-
ences over the multi-decadal signal of annual precipitation found
for Punta Arenas (Fig. 11d). This annual signal also reveals a con-
siderable connection with the South Atlantic Ocean, which is sig-
nificant but weakened during wet and dry conditions (see
Figs. B1 and B2 in the Appendix).
4. Discussion
4.1. Multi-decadal cycles of precipitation variability
The emergence of an approximately multi-decadal oscillation
has been investigated in different areas of the earth sciences
(Klyashtorin et al., 2009). For instance, long tide gauge records
show a significant oscillation with a period of around 60 years
for the majority of the tide gauges examined during the 20th cen-
tury in every ocean basin of the world (Chambers et al., 2012). In
general these multi-decadal oscillations are characterized by 30-
year and 50- to 70-year periodic cycles (Klyashtorin and
Lyubushin, 2007; Klyashtorin et al., 2009). In this study, we
extracted significant multi-decadal components of precipitation
that explain the largest variances at those rain gauges located in
the extreme northern (arid) and southern (cold) regions of Chile.
The periodicities calculated for these multi-decadal components
indicate cycles ranging between 40 and 64 years. These cannot
be associated with the modes of the SOI (AAO), since their compo-
nents (at least SOI) show a weak multi-decadal signal, with a
source of variability mostly dominated by strong high-frequency
cycles (<10 years). The strong high-frequency cycles introduce
more noise into the precipitation records because of intensification
or weakening of precipitation contributions. This led to a much
more difficult identification of multi-decadal cycles of precipita-
tion in central Chile and, as a result, most of the rain gauges pro-
duced no significant outcomes. This relationship also remained
consistent and significant during wet and dry seasons.
On the other hand, this large-scale temporal variability of pre-
cipitation seems to be highly associated with the multi-decadal
features observed over the Pacific (PDO) and Atlantic (AMO)
Oceans (i.e. Enfield et al., 2001; McCabe et al., 2004; McCarthy
et al., 2015). In fact, these multi-decadal signals are significantly
correlated with those records of rain gauges located in the subtrop-
ical and extratropical regions of Chile. The PDO signal is stronger to
the north of the country (Copiapó), whereas the AMO signal is
strong in both the north and the south of the country (Punta Are-
nas). Positive SST anomalies are observed during the positive phase
of both oceanic oscillations. This pattern is evident from the global
spatial correlation of low-frequency precipitation modes and SSTs.
4.2. Multi-decadal oscillations and extreme hydrometeorological
events
These unprecedented spatial linkages between the multi-
decadal precipitation variability over Chile and the PDO and AMO
signals constitute an additional and important element to be con-
sidered when predicting future impacts associated with extreme
hydrological events such as floods and droughts. For example, the
extreme flood events that occurred during March and July of 2015
in the Atacama Desert (Wilcox et al., 2016), which resulted in sev-
eral deaths and economic loses, were generated by two different
ocean-atmosphere processes that occurred during the high phase
of AMO. The first event that occurred at the end of the dry season
(end of March) was recently defined by Barrett et al. (2016) as a cut-
off low with anomalous heights that moved east and was directed
ashore near 25°S (see Fig. C1 in the Appendix) by circulation around
a very strong ridge centered near 60°S. This phenomenon coincided
with warmer temperatures in central Chile and anomalously high
PW content (20 mm above climatological norms) over the Peruvian
Bight. North-westerlies picked up higher PW over an area of the
subtropical eastern Pacific Ocean (similar to that observed in
Fig. 11) that had warmed rapidly during the prior week, and was
transported over the northern regions of Chile. The second event
occurred during the rainiest month of the wet season (July) where
moisture from westerlies reached the region as a frontal system due
to a weakened SEPA (northward displacement). This finding coin-
cides with the significant positive correlation of AMO with PW at
500 mb over the north and south regions of Chile during the wet
season (Apr.-Sep.). In the North of Chile the correlation concen-
trates over the area spanned by the SEPA, but in the South it appears
as an elongated band extended from New Zealand towards south-
ern Chile. This correlation pattern is also significantly positive along
the North-East Pacific coastline and reaches countries all along
South America, Central America, Southern Mexico and the United
States (Fig. 12a). It is important to mention that, during the high-
phase of AMO, the PW reduces in the Pacific Basin and produces
neutral to negative anomalies of PW in Central Chile; however,
the subtropical and extratropical regions of Chile reveal positive
anomalies of PW, which is part of a pattern extended all over the
American Continent. On the other hand, during the dry season
(Oct.-Mar.) this pattern of positive correlations is still observed in
the coastline of northern Chile but no longer in southern Chile
(Fig. 12b). According to these findings, both March and July events
could be potentially associated to the high phase of AMO (probably
intensified by 2015–2016 ENSO), which is strongly associated to a
warmer ocean (Bozkurt et al., 2016), higher precipitable water
amounts and positive ZW anomalies at 500 mb (Fig. 13a and b) in
northern and southern regions of Chile. Such events must be also
related to the interaction between different phases of ENSO, the
PDO, and the AMO which can intensify or diminish the impacts of
each other according to their phases (see i.e. Wang et al., 2013).
When both ENSO and the PDO are in the same phase, it is expected
that El Niño (La Niña) impacts may be magnified generating posi-
tive (negative) precipitation anomalies over Chile. Accordingly,
we hypothesize that this reinforcing was also evident during the
recent positive phases of both ENSO and AMO, producing enhanced
positive anomalies of precipitation in Chile and other regions of the
World. The application of this knowledge will potentially benefit
the predictive capacity of future extreme hydrometeorological
events in semiarid basins of northern Chile (ungauged) where times
of concentration are low and lead times are essentially required to
minimize impacts on the society.
4.3. Ongoing and future research
The recent opposition between the warm (positive) phase of the
AMO and a substantially negative-trending (cold phase) of the
multi-decadal PDO signal (or Pacific Multi-decadal Oscillation,
PMO) was apparently producing the slowdown or false pause of
warming temperatures of the past decade (Steinman et al., 2015).
This global warming hiatus has been also described as the result
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170 1165

Fig. 11. Multi-decadal modes of annual signals of (a) the PDO; (b) Copiapó (
27.4°S); (c) the AMO; and (d) Punta Arenas (
53.1°S), projected onto SSTs (see Hirahara et al.,
2014). The green marker is the Copiapó rain gauge, and the yellow marker is the Punta Arenas rain gauge. Colored patterns represent significant correlation coefficients
calculated by the large sample approximation method. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

Fig. 12. Correlation map for the relationship between the raw signal of AMO and Fig. 13. Correlation map for the relationship between the raw AMO signal and
the amount of PW at 500 mb. Blue colors represent positive correlations and red zonal wind speeds at 500 mb. Blue colors represent positive correlations and red
colors negative correlations. (For interpretation of the references to colour in this colors negative correlations. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.) figure legend, the reader is referred to the web version of this article.)
1166 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
of interactions between a secular warming tendency due in a large
part to the buildup of anthropogenic CO2 concentrations, and an
internally generated cooling due to a cool phase of a quasi-60-
year oscillatory variability that is closely associated with the
AMO and PDO (Yao et al., 2015). Despite the hiatus periods that
have appeared in surface air temperature records, other compo-
nents of the climate system associated with warming have contin-
ued at multi-decadal scales. These include displacements in the
mean latitudinal position of the Intertropical Convergence Zone
(ITCZ) that vary on a similar cycle to the AMO (Knudsen et al.,
2011), and likely linkages with external forces related to astronom-
ical dynamics (see i.e. Scafetta, 2010; Knudsen et al., 2014). These
interrelations must be clarified in future studies in order to
improve forecasting schemes and climate scenarios for near- and
long-term impact assessment of extreme hydrological events.
These studies will need to include the effect of the multi-decadal
climate variability on the hydrological regime of Chilean water-
sheds as a way to improve water governance schemes.
5. Conclusions
Multi-decadal cycles of precipitation in Chile are significant at
annual, seasonal, and monthly scales. The high-frequency compo-
nent associated with precipitation records of central Chile is mostly
related to the QBO, ENSO, and AAO. The ENSO (represented by SOI)
and the AAO did not explain the multi-decadal precipitation vari-
ability, since their frequencies are lower than those observed for
the multi-decadal leading modes of precipitation variability. The
strong influence of the PDO and AMO signals is significant over
the multi-decadal precipitation variability of the extreme regions
of Chile i.e., those located in low and high latitudes. Specifically,
the PDO correlations are evident in the north of the country,
whereas correlations with the AMO are strong both in the north
and in the south of the country. For instance, the raw signal of
AMO revealed to be significantly correlated with SST, PW and zonal
winds (both measured at 500 mb), suggesting that this multi-
decadal variability is not only observed in the ocean but also in tro-
pospheric processes, since the positive phase of AMO is also related
to higher PW and positive anomalies of zonal winds at 500 mb.
Based on our results, we conclude that the PDO and AMO are two
of the most important factors influencing multi-decadal precipita-
tion variability in these northern and southern regions of Chile.
The result of the relationships between warm and cold phases
of the ENSO, AAO, PDO, and AMO must be researched further to
determine the physical factors or patterns that explain this
multi-decadal mode of variability in Chile and other regions of
the World. The possible connection between this multi-decadal
pattern and external forcing also must be examined in future
studies. Gridded data and assembled models can be used as part
of a future integrated approach to draw new conclusions about
the physical origin and interrelations of these signals and these
conclusions can be used to improve water governance schemes
at catchment scale.
Acknowledgments
The authors acknowledge the Dirección General de Aguas de
Chile and the Dirección Meteorológica de Chile for providing the
data used in this study. The first author received support from
the Comisión Nacional de Investigación Científica y Tecnológica
de Chile (CONICYT) to carry out this research. The second author
received partial support from the NSF-USAID program PEER cycle
1 (project 31) and Universidad de Antioquia project PI12-1-03 to
participate in this research. These contributions are greatly
acknowledged.
Appendix A. Singular Spectrum Analysis (SSA)
Traditionally the EOF method finds a combination of variables
which explains most of the variability, usually observing high cor-
relation in space. However, the significant correlation in time is
not taken into account. Auto- and cross-correlation in time are
very useful for prediction purposes and also for building stochas-
tic time series models. An extension of the traditional EOF
method is the Extended Empirical Orthogonal Functions (EEOFs),
which is a technique to deal not only with spatial but also with
temporal correlations observed in climate data. The method was
first introduced by Weare and Nasstrom (1982), who applied it
to the 300-mb Relative Vorticity. A similar approach was devel-
oped to deal with the dynamical reconstruction of low order
chaotic systems by Broomhead and King (1986a,b) who called it
Singular System Analysis (SSA) in the one-dimensional case.
Fraedrich (1986) proposed the name of the method as Singular
Spectrum Analysis (SSA), and a modified version named
Multichannel-SSA (M-SSA). Recently, Patterson et al. (2011)
developed the non-parametric method of Multivariate Singular
Spectrum Analysis (MSSA) for multi-vintage data. This approach
has proven much more flexible than the standard methods of
modelling, which involve at least one of the restrictive assump-
tions of linearity, normality, and stationarity.
Over the last few decades, SSA has been considered a powerful
technique in time series analysis, and it has been developed and
applied to many practical problems. SSA has been described as a
very useful tool that can be used for solving the following prob-
lems: finding trends of different resolution; smoothing; extraction
of seasonality components; simultaneous extraction of cycles with
small and large periods; extraction of periodicities with varying
amplitudes; simultaneous extraction of complex trends and peri-
odicities; finding structure in short time series; and change-point
detection. (Vautard and Ghil, 1989; Ghil et al., 2002; Golyandina
et al., 2001; Hassani et al., 2009; Hassani and Thomakos, 2010).
The SSA decomposes the original time series into a sum of small
numbers of interpretable components, such as a slowly varying
trend, an oscillatory component, and noise. The basic SSA method
consists of two complementary stages, Decomposition and Recon-
struction. Each stage includes two separate steps. At the first stage,
the series are decomposed to obtain the Trajectory Matrix; at the
second stage the noise-free series are reconstructed for forecasting
new data points. A detailed description of the SSA technique is
given in several publications. For more explanations and compar-
ison with other time series analysis techniques, refer to the refer-
ences provided in this article.
In general terms, the whole procedure of the SSA technique
depends upon two parameters:
I. The lagged window length (L).
II. The number of needed singular values for reconstruction (r).
An improper choice of parameters L and r may yield incomplete
reconstruction and misleading results in forecasting. Recent results
obtained by Mahmoudvand and Zokaei (2012), show that the
choice of L close to one-half (N/2) of the time series length is a suit-
able choice for the decomposition stage in most cases of SSA. How-
ever, in the case of climatic indices the oscillations reported in
previous studies may be used as an appropriate lagged window
length for analysis purposes.
Appendix B. Spatial significances of multi-decadal modes of wet
and dry seasons
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170 1167

Fig. B1. Multi-decadal modes of wet signal of (a) Copiapó (
27.4°S); (b) PDO; (c) Punta Arenas (
53.1°S), and (d) AMO; projected onto gridded SSTs (see Hirahara et al., 2014) for
1912–2014. Green marker is the Copiapó rain gauge and the yellow marker is the Punta Arenas raingauge. Colored patterns represent significant correlation coefficients
calculated by the large sample approximation method. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. B2. Multi-decadal modes of dry signals of (a) Copiapó (
27.4°S); (b) PDO; (c) Punta Arenas (
53.1°S), and (d) AMO; projected onto gridded Sea Surface Temperatures (see
Hirahara et al., 2014) for 1912–2014. Green-marker is Copiapó raingauge and yellow-marker is Punta Arenas raingauge. Colored patterns represent significant correlation coefficients
calculated by the large sample approximation method. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1168 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
Appendix C. Extreme event of March 2015 in the Atacama
Desert Chile
Fig. C1. Scheme of a trapped cutoff-low in the Pacific in front of the Atacama Region
and centered about 25°S. The phenomenon triggered the transport of moisture from
the Peruvian Bight and Northern Chile towards the Atacama Desert Region Source:
Own source, adapted from Di Liberto (2015).
References
Aceituno, P., Fuenzalida, H., Rosenblüth, B., 1993. Climate along the extratropical
west coast of South America. In: Mooney, H.A., Fuentes, E.R., Kronberg, B.I.
(Eds.), Earth System Responses to Global Change: Contrast between North and
South America. Academic Press, New York, pp. 61–69.
Aceituno, P., Garreaud, R., 1995. Impacto de los fenómenos el Niño y la Niña en el
régimen pluviométrico Andino. Revista Chilena de Ingeniería Hidráulica 9, 12–
20.
Antunes, S., Pires, O., Rocha, A., 2006. Detecting spatio-temporal precipitation
variability in Portugal using multichannel singular spectral analysis. Int. J.
Climatol. 26 (15), 2199–2212.
Aravena, J.C., Luckman, B.H., 2009. Spatio-temporal rainfall patterns in Southern
South America. Int. J. Climatol. 29 (14), 2106–2120. http://dx.doi.org/10.1002/
joc.1761.
Baldwin, M.P., Gray, L.J., Dunkerton, T.J., Hamilton, K., Haynes, P.H., Randel Holton, J.
R., Alexander, M.J., Hirota, I., Horinouchi, T., Jones, D.B.A., Kinnersley, J.S.,
Marquardt, C., Sato, K., Takahashi, M., 2001. The quasi-biennial oscillation. Rev.
Geophys. 39 (2), 179–229.
Barrett, B.S., Campos, D.A., Veloso, J.V., Rondanelli, R., 2016. Extreme temperature
and precipitation events in March 2015 in central and northern Chile. J.
Geophys. Res.: Atmos. 121 (9), 4563–4580.
Bendat, J.S., Piersol, A.G., 2011. Random Data: Analysis and Measurement
Procedures, vol. 729. John Wiley & Sons.
Biondi, F., Gershunov, A., Cayan, D.R., 2001. North Pacific decadal climate variability
since 1661. J. Clim. 14 (1), 5–10.
Box, G., Jenkins, G., 1970. Time Series Analysis: Forecasting and Control. Holden-
Day, San Francisco.
Box, G.E.P., Jenkins, G.M., Reinsel, G.C., 1994. Time Series Analysis: Forecasting and
Control. Prentice Hall, Englewood Cliff, N.J..
Bozkurt, D., Rondanelli, R., Garreaud, R., Arriagada, A., 2016. Impact of Warmer
Eastern Tropical Pacific SST on the March 2015 Atacama Floods. Mon. Weather
Rev. 144 (11), 4441–4460.
Broomhead, D.S., King, G.P., 1986a. Extracting qualitative dynamics from
experimental data. Physica D 20 (2), 217–236.
Broomhead, D.S., King, G.P., 1986b. On the qualitative analysis of experimental
dynamical systems. Nonlinear Phenomena Chaos 113, 114.
Brunet, M., Jones, P., 2011. Data rescue initiatives: bringing historical climate data
into the 21st century. Clim. Res. 47 (1), 29.
Cañón, J., González, J., Valdés, J.B., 2007. Precipitation in the Colorado River Basin
and its low frequency associations with PDO and ENSO signals. J. Hydrol. 333
(2), 252–264.
Carrasco, J.F., Casassa, G., Rivera, A., 2002. Meteorological and climatological aspects
of the Southern Patagonia Icefield. In: The Patagonian Icefields. Springer, US, pp.
29–41.
Chambers, D.P., Merrifield, M.A., Nerem, R.S., 2012. Is there a 60-year oscillation in
global mean sea level? Geophys. Res. Lett. 39 (18).
Chatfield, C., 2013. The Analysis of Time Series: An Introduction. CRC Press.
Dai, A., 2013. The influence of the inter-decadal Pacific oscillation on US
precipitation during 1923–2010. Clim. Dyn. 41, 633–646.
Dean, J.M., Kemp, A.E.S., 2004. A 2100 year BP record of the Pacific Decadal
Oscillation, El Niño Southern Oscillation and Quasi-Biennial Oscillation in
marine production and fluvial input from Saanich Inlet, British Columbia.
Palaeogeogr., Palaeoclimatol., Palaeoecol. 213, 207–229.
Deser, C., Alexander, M.A., Xie, S., Phillips, A.S., 2010. Sea surface temperature
variability: patterns and mechanisms. Ann. Rev. Mar. Sci. 2 (1), 115–143. http://
dx.doi.org/10.1146/annurev-marine-120408-151453. PMID 21141660.
Dong, B., Dai, A., 2015. The influence of the Interdecadal Pacific Oscillation on
Temperature and Precipitation over the Globe. Clim. Dyn., 1–15
d’Orgeville, M., Peltier, W.R., 2007. On the Pacific decadal oscillation and the
Atlantic multidecadal oscillation: might they be related? Geophys. Res. Lett. 34
(23).
Dawson, A.G., O’Hare, G., 2000. Ocean-atmosphere circulation and global climate:
the El Niño-Southern Oscillation. Geography 85, 193–208.
Ebdon, R.A., 1975. The quasi-biennial oscillation and its association with
tropospheric circulation patterns. Met. Mag. 104, 282–297.
Elbert, J., Grosjean, M., von Gunten, L., Urrutia, R., Fischer, D., Wartenburger, R.,
Ariztegui, D., Fujak, M., Hamann, Y., 2011. Quantitative high-resolution winter
(JJA) precipitation reconstruction from varved sediments of Lago Plomo 47 S,
Patagonian Andes, AD 1530–2002. The Holocene, 0959683611425547.
Elsner, J.B., Tsonis, A.A., 2013. Singular Spectrum Analysis: A New Tool in Time
Series Analysis. Springer Science & Business Media.
Enfield, D.B., Mestas- Núñez, A.M., Trimble, P.J., 2001. The Atlantic Multidecadal
Oscillation and its relationship to rainfall and river flows in the continental U.S.
Geophys. Res. Lett. 28, 2077–2080.
Fagel, N., Boës, X., Loutre, M.F., 2008. Climate oscillations evidenced by spectral
analysis of Southern Chilean lacustrine sediments: the assessment of ENSO over
the last 600 years. J. Paleolimnol. 39 (2), 253–266.
Falvey, M., Garreaud, R., 2007. Wintertime precipitation episodes in Central Chile:
associated meteorological conditions and orographic influences. J.
Hydrometeorol. 8 (2), 171–193.
Fraedrich, K., 1986. Estimating the dimensions of weather and climate attractors. J.
Atmos. Sci. 43 (5), 419–432.
Garreaud, R., Batisti, D., 1999. Inter-annual (ENSO) and inter-decadal (ENSO-like)
variability of the Southern Hemisphere tropospheric circulation. J. Clim. 12,
2113–2123 (PDF).
Garreaud, R., Vuille, M., Clement, A.C., 2003. The climate of the Altiplano: observed
current conditions and mechanisms of past changes. Palaeogeogr.,
Palaeoclimatol., Palaeoecol 194 (1), 5–22. Chicago.
Garreaud, R., Vuille, M., Compagnucci, R., Marengo, J., 2009. Present-day South
American climate. Palaeogeogr. Palaeoclimatol. Palaeoecol. 281 (3–4), 180–195.
http://dx.doi.org/10.1016/j.palaeo.2007.10.032.
Garreaud, R.D., Gabriela Nicora, M., Bürgesser, R.E., Ávila, E.E., 2014. Lightning in
western Patagonia. J. Geophys. Res.: Atmos. 119 (8), 4471–4485.
Graystone, P., 1959. Meteorological office discussion on tropical meteorology. Met.
Mag. 88, 117.
Ghil, M., Vautard, R., 1991. Interdecadal oscillations and the warming trend in
global temperature time series. Nature 350 (6316), 324–327.
Ghil, M., Allen, M.R., Dettinger, M.D., Ide, K., Kondrashov, D., Mann, M.E., Robertson,
A.W., Saunders, A., Tian, Y., Varadi, F., Yiou, P., 2002. Advanced spectral methods
for climatic time series. Rev. Geophys. 40 (1). 3-1.
Golyandina, N., Nekrutkin, V., Zhigljavsky, A.A., 2001. Analysis of Time Series
Structure: SSA and Related Techniques. CRC Press.
Golyandina, N., Zhigljavsky, A., 2013. Singular Spectrum Analysis for Time Series.
Springer Briefs in Statistics. Springer.
Gomes, P.T., 2011. Interannual oscillations in winter rainfall over Europe: Iberia
study case. Finisterra: Revista portuguesa de geografia 46 (91), 27–45.
Hannachi, A., 2004. A Primer for EOF Analysis of Climate Data. Department of
Meteorology, University of Reading. 1–33.
Hassani, H., Heravi, S., Zhigljavsky, A., 2009. Forecasting European industrial
production with singular spectrum analysis. Int. J. Forecast. 25 (1), 103–118.
Hassani, H., Thomakos, D., 2010. A review on singular spectrum analysis for
economic and financial time series. Stat. Its Interface 3 (3), 377–397.
Hassani, H., Zhigljavsky, A., Patterson, K., Soofi, A., 2010. A Comprehensive Causality
Test based on the Singular Spectrum Analysis. Causality in Science, Oxford
University Press.
Hassani, H., Heravi, S., Zhigljavsky, A., 2013. Forecasting UK industrial production
with multivariate singular spectrum analysis. J. Forecasting 32 (5), 395–408.
Hassani, H., Mahmoudvand, R., 2013. Multivariate singular spectrum analysis: a
general view and new vector forecasting approach. Int. J. Energy Stat. 1 (01),
55–83.
Hassani, H., Webster, A., Silva, E.S., Heravi, S., 2015. Forecasting US tourist arrivals
using optimal singular spectrum analysis. Tourism Manage. 46, 322–335.
Hirahara, S., Ishii, M., Fukuda, Y., 2014. Centennial-scale sea surface temperature
analysis and its uncertainty. J. Clim. 27, 57–75. http://dx.doi.org/10.1175/JCLI-
D-12-00837.1.
Houston, J., 2006. Variability of precipitation in the Atacama Desert: its causes and
hydrological impact. Int. J. Climatol. 26, 2181–2198. http://dx.doi.org/10.1002/
joc.1359.
Huesca, M., Merino-de-Miguel, S., Eklundh, L., Litago, J., Cicuéndez, V., Rodríguez-
Rastrero, M., Ustin, S., Palacios-Orueta, A., 2015. Ecosystem functional
 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
assessment based on the ‘‘optical type” concept and self-similarity patterns: An
application using MODIS-NDVI time series autocorrelation. Int. J. Appl. Earth
Observ. Geoinformation.
Justino, F., Peltier, W.R., 2005. The glacial North Atlantic Oscillation. Geophys. Res.
Lett. 32 (21).
Kim, S., Kim, Y., Lee, J., Kim, H.S., 2015. Identifying and evaluating chaotic behavior
in hydro-meteorological processes. Adv. Meteorol.
Klyashtorin, L.B., Lyubushin, A.A., 2007. Cyclic Climate Changes and Fish
Productivity. VNIRO (All Russian Institute Fisheries and Oceanography)
Publishing, Moscow. 224.
Klyashtorin, L.B., Borisov, V., Lyubushin, A., 2009. Cyclic changes of climate and
major commercial stocks of the Barents Sea. Mar. Biol. Res. 5 (1), 4–17.
Knight, J.R., Allan, R.J., Folland, C.K., Vellinga, M., Mann, M.E., 2005. A signature of
persistent natural thermohaline circulation cycles in observed climate.
Geophys. Res. Lett. 32 (20).
Knudsen, M.F., Seidenkrantz, M.S., Jacobsen, B.H., Kuijpers, A., 2011. Tracking the
Atlantic Multidecadal Oscillation through the last 8,000 years. Nat. Commun. 2,
178.
Knudsen, M.F., Jacobsen, B.H., Seidenkrantz, M.S., Olsen, J., 2014. Evidence for
external forcing of the Atlantic Multidecadal Oscillation since termination of the
Little Ice Age. Nat. Commun. 5.
Krishnamurthy, V., Misra, V., 2010. Observed ENSO teleconnections with the South
American monsoon system. Atmos. Sci. Lett. 11 (1), 7–12.
Krishnamurthy, V., Stan, C., 2015. Simulation of the South American climate by a
coupled model with super-parameterized convection. Clim. Dyn. 44 (9–10),
2369–2382.
Krishnamurthy, L., Krishnamurthy, V., 2015. Teleconnections of Indian monsoon
rainfall with AMO and Atlantic tripole. Clim. Dyn., 1–17
Kunkel, K.E., Andsager, K., Easterling, D.R., 1999. Long-term trends in extreme
precipitation events over the conterminous United States and Canada. J. Clim.
12 (8), 2515–2527. http://dx.doi.org/10.1175/1520-0442(1999) 012 2515:
LTTIEP 2.0.CO;2.
Le Quesne, C., Acuña, C., Boninsegna, J.A., Rivera, A., Barichivich, J., 2009. Long-term
glacier variations in the Central Andes of Argentina and Chile, inferred from
historical records and tree-ring reconstructed precipitation. Palaeogeogr.
Palaeoclimatol. Palaeoecol. 281 (3), 334–344.
Lee, T., Ouarda, T.B.M.J., Li, J., 2013. An orchestrated climate song from the Pacific
and Atlantic Oceans and its implication on climatological processes. Int. J.
Climatol. 33 (4), 1015–1020.
Li, S., Jing, Y., Luo, F., 2015. The potential connection between China surface air
temperature and the Atlantic Multidecadal Oscillation (AMO) in the pre-
industrial period. Sci. China Earth Sci., 1–13
Lu, R., Dong, B., Ding, H., 2006. Impact of the Atlantic Multidecadal Oscillation on the
Asian summer monsoon. Geophys. Res. Lett. 33 (24).
Mahmoudvand, R., Zokaei, M., 2012. On the singular values of the Hankel matrix
with application in singular spectrum analysis. Chilean J. Stat. 3 (1), 43–56.
Machiwal, D., Jha, M.K., 2012. Hydrologic Time Series Analysis: Theory and Practice.
Springer Science & Business Media.
Machiwal, D., Kumar, S., Dayal, D., 2015. Characterizing rainfall of hot arid region by
using time-series modeling and sustainability approaches: a case study from
Gujarat, India. Theor. Appl. Climatol., 1–15
Makowsky, J.A., Labai, N., 2015. Hankel Matrices: From Words to Graphs. In
Language and Automata Theory and Applications. Springer International
Publishing, pp. 47–55.
Mantua, N.J., Hare, S., Zhang, Y., Wallace, J.M., Francis, R.C., 1997. A Pacific
interdecadal climate oscillation with impacts on salmon production. Bull. Am.
Meteorol. Soc. 78, 1069–1079.
Marques, C.A.F., Ferreira, J.A., Rocha, A., Castanheira, J.M., Melo-Gonçalves, P., Vaz,
N., Dias, J.M., 2006. Singular spectrum analysis and forecasting of hydrological
time series. Phys. Chem. Earth, Parts A/B/C 31 (18), 1172–1179.
Masiokas, M.H., Villalba, R., Luckman, B.H., Mauget, S., 2010. Intra-to multidecadal
variations of snowpack and streamflow records in the Andes of Chile and
Argentina between 30 and 37 S. J. Hydrometeorol. 11 (3), 822–831.
McCabe, G.J., Palecki, M.A., Betancourt, J.L., 2004. Pacific and Atlantic Ocean
influences on multidecadal drought frequency in the United States. Proc. Natl.
Acad. Sci. 101 (12), 4136–4141.
McCarthy, G.D., Haigh, I.D., Hirschi, J.J.M., Grist, J.P., Smeed, D.A., 2015. Ocean impact
on decadal Atlantic climate variability revealed by sea-level observations.
Nature 521 (7553), 508–510.
Miller, A., 1976. The climate of Chile. In: Schwerdtfeger, W. (Ed.), Climates of Central
and South America. Elsevier, Amsterdam, pp. 113–145.
Minobe, S., 1997. A 50–70 year climatic oscillation over the North Pacific and North
America. Geophys. Res. Lett. 24, 683–686.
Minvielle, M., Garreaud, R.D., 2011. Projecting rainfall changes over the South
American Altiplano. J. Clim. 24 (17), 4577–4583.
Montecinos, A., Aceituno, P., 2003. Seasonality of the ENSO-related rainfall
variability in central Chile and associated circulation anomalies. J. Clim. 16
(2), 281–296.
Di Liberto, T., 2015. Flooding in the Atacama Desert: How did that happen?. NOAA –
National Oceanic and Atmospheric Administration. Available at: <https://
www.climate.gov/news-features/event-tracker/flooding-atacama-desert-how-
did-happen> (accessed Sept 14th, 2015).
Nolte, D.D., 2010. The tangled tale of phase space. Phys. Today.
North, G.R. et al., 1982. Sampling errors in the estimation of empirical orthogonal
functions. Mon. Weather Rev. 110 (7), 699–706.
1169
Núñez, J., Rivera, D., Oyarzún, R., Arumí, J.L., 2013. Influence of Pacific Ocean
multidecadal variability on the distributional properties of hydrological
variables in north-central Chile. J. Hydrol. 501, 227–240.
Partington, J.R, 1988. An introduction to Hankel operators. LMS Student Texts 13.
Cambridge University Press. ISBN 0-521-36791-3.
Patterson, K., Hassani, H., Heravi, S., Zhigljavsky, A., 2011. Multivariate singular
spectrum analysis for forecasting revisions to real-time data. J. Appl. Stat. 38
(10), 2183–2211.
Peristykh, A.N., Damon, P.E., 2003. Persistence of the Gleissberg 88-year solar cycle
over the last 12,000 years: evidence from cosmogenic isotopes. J. Geophys. Res.
108 (A1), 1003.
Pizarro, R., Valdés-Pineda, R., García-Chevesich, P., Vallejos, C., Sangüesa, C.,
Morales, C., Balocchi, F., Abarza, A., Fuentes, R., 2012. Latitudinal analysis of
rainfall intensity and mean annual precipitation in Chile. Chilean J. Agric. Res.
72 (2), 252–261.
Quintana, J., Aceituno, P., 2012. Changes in the rainfall regime along the
extratropical west coast of South America (Chile): 30–43°S.Atmósfera 25 (1),
1–22.
Rao, A.R., Hamed, K., 2003. Multi-taper method of analysis of periodicities in
hydrologic data. J. Hydrol. 279 (1), 125–143.
Rubio-Álvarez, E., McPhee, J., 2010. Patterns of spatial and temporal variability in
streamflow records in south central Chile in the period 1952–2003. Water
Resour. Res. 46 (5).
Rutllant, José., Fuenzalida, Humberto., 1991. Synoptic aspects of the Central Chile
Rainfall variability associated with the Southern Oscillation’. Int. J. Climatol. 11,
63–76.
Sanz Donaire, J.J., 2012. Las series anuales de precipitación más largas de Chile:
estudio y enseñanzas. Estudios Geográficos 73 (273), 625–656.
Scafetta, N., 2010. Empirical evidence for a celestial origin of the climate oscillations
and its implications. J. Atmos. Solar Terr. Phys. 72 (13), 951–970.
Schlesinger, M.E., Ramankutty, N., 1994. An oscillation in the global climate system
of period 65–70 years. Nature 367 (6465), 723–726.
Silva, E.S., Hassani, H., 2015. On the use of singular spectrum analysis for forecasting
US trade before, during and after the 2008 recession. Int. Econ. 141, 34–49.
Sivakumar, B., Jayawardena, A.W., Fernando, T.M.K.G., 2002. River flow forecasting:
use of phase-space reconstruction and artificial neural networks approaches. J.
Hydrol. 265 (1), 225–245.
Steinman, B.A., Mann, M.E., Miller, S.K., 2015. Atlantic and Pacific multidecadal
oscillations and Northern Hemisphere temperatures. Science 347 (6225), 988–
991.
Sun, C., Li, J., Feng, J., Xie, F., 2015. A decadal-scale teleconnection between the North
Atlantic Oscillation and subtropical eastern Australian rainfall. J. Clim. 28 (3),
1074–1092.
Trenberth, K.E., 1984. Signal versus noise in the Southern Oscillation. Mon. Wea.
Rev. 112, 326–332.
Trenberth, K.E., Hoar, T.J., 1997. El Niño and climate change. Geophys. Res. Lett. 24
(23), 3057–3060.
Trenberth, K.E., Shea, D.J., 2006. Atlantic hurricanes and natural variability in 2005.
Geophys. Res. Lett. 33, 12.
Unnikrishnan, P., Jothiprakash, V., 2015. Extraction of nonlinear rainfall trends using
singular spectrum analysis. J. Hydrol. Eng. 0. 05015007.
Urrutia, R.B., Lara, A., Villalba, R., Christie, D.A., Le Quesne, C., Cuq, A., 2011.
Multicentury tree ring reconstruction of annual streamflow for the Maule River
watershed in south central Chile. Water Resour. Res. 47 (6).
Valdés-Pineda, R., Pizarro, R., García-Chevesich, P., Valdés, J.B., Olivares, C., Vera, M.,
Balocchi, F., Pérez, F., Vallejos, C., Fuentes, R., Abarza, A., Helwig, B., 2014. Water
governance in Chile: availability, management and climate change. J. Hydrol.
519, 2538–2567. http://dx.doi.org/10.1016/j.jhydrol.2014.04.016.
Valdés-Pineda, R., Pizarro, R., Valdés, J.B., Carrasco, J.F., García-Chevesich, P.,
Olivares, C., 2015. Spatio-temporal trends of precipitation, its aggressiveness
and concentration, along the Pacific coast of South America (36°–49° S). Hydrol.
Sci. J. http://dx.doi.org/10.1080/02626667.2015.1085989.
Valdés-Pineda, R., Valdés, J.B., Díaz, H.F., Pizarro-Tapia, R., 2016. Analysis of spatio-
temporal changes in annual and seasonal precipitation variability in South
America-Chile and related ocean–atmosphere circulation patterns.
International Journal of Climatology 36 (8), 2979–3001.
Vautard, R., Ghil, M., 1989. Singular spectrum analysis in nonlinear dynamics, with
applications to paleoclimatic time series. Physica D 35 (3), 395–424.
Viale, M., Garreaud, R., 2015. Orographic effects of the subtropical and extratropical
Andes on upwind precipitating clouds. J. Geophys. Res. Atmos. 120. http://dx.
doi.org/10.1002/2014JD023014.
Villalba, R., Cook, E.R., Jacoby, G.C., D’Arrigo, R.D., Veblen, T.T., Jones, P.D., 1998.
Tree-ring based reconstructions of northern Patagonia precipitation since AD
1600. The Holocene 8 (6), 659–674.
Villalba, R., Lara, A., Masiokas, M.H., Urrutia, R.B., Luckman, B.H., Marshall, G.J.,
Mundo, I.A., Christie, D.A., Cook, E.R., Neukom, R., Allen, K., Fenwick, P.,
Boninsegna, J.A., Srur, A.M., Morales, M.S., Araneo, D., Palmer, J.G., Cuq, E.,
Aravena, J.C., Holz, A., Le Quesne, C., 2012. Unusual Southern Hemisphere tree
growth patterns induced by changes in the Southern Annular Mode. Nat. Geosci.
5, 793–798.
Villarroel, C., 2013. Eventos extremos de precipitación y temperatura en Chile:
Proyecciones para fines del siglo XXI. Available at: <http://www.repositorio.
uchile.cl/handle/2250/114066>.
Vuille, M., Ammann, C., 1997. Regional snowfall patterns in the high, arid Andes. In:
Climatic Change at High Elevation Sites. Springer, Netherlands, pp. 181–191.
1170 R. Valdés-Pineda et al. / Journal of Hydrology 556 (2018) 1153–1170
Wang, B., Liu, J., Kim, H.J., Webster, P.J., Yim, S.Y., Xiang, B., 2013. Northern
Hemisphere summer monsoon intensified by mega-El Niño/southern
oscillation and Atlantic multidecadal oscillation. Proc. Natl. Acad. Sci. 110
(14), 5347–5352.
Wang, H., Sankarasubramanian, A., Ranjithan, R.S., 2015. Understanding the low-
frequency variability in hydroclimatic attributes over the southeastern US. J.
Hydrol. 521, 170–181.
Wang, R., Ma, H.G., Liu, G.Q., Zuo, D.G., 2015. Selection of window length for singular
spectrum analysis. J. Franklin Inst. 352 (4), 1541–1560.
Weare, B.C., Nasstrom, J.S., 1982. Examples of extended empirical orthogonal
function analyses. Mon. Weather Rev. 110 (6), 481–485.
Wilcox, A.C., Escauriaza, C., Agredano, R., Mignot, E., Zuazo, V., Otárola, S., Mao, L.,
2016. An integrated analysis of the March 2015 Atacama floods. Geophys. Res.
Lett. 43 (15), 8035–8043.
Wilks, D.S., 2011. Statistical Methods in the Atmospheric Sciences. Elsevier
Academic Press, California, USA.
Yao, S.L., Huang, G., Wu, R.G., Qu, X., 2015. The global warming hiatus—a natural
product of interactions of a secular warming trend and a multi-decadal
oscillation. Theoret. Appl. Climatol., 1–12
Yuan, N., Fu, Z., Zhang, H., Piao, L., Xoplaki, E., Luterbacher, J., 2015. Detrended
partial-cross-correlation analysis: a new method for analyzing correlations in
complex system. Sci. Rep. 5.
Zhang, R., Delworth, T.L., 2006. Impact of Atlantic multidecadal oscillations on
India/Sahel rainfall and Atlantic hurricanes. Geophys. Res. Lett. 33 (17).
Zolina, O., Simmer, C., Belyaev, K., Kapala, A., Gulev, S., 2009. Improving estimates of
heavy and extreme precipitation using daily records from European rain gauges.
J. Hydrometeorol. 10 (3), 701–716.
Zotov, L.V., Shum, C.K., Frolova, N.L., 2015. Gravity Changes over Russian River
Basins from GRACE. In: Planetary Exploration and Science: Recent Results and
Advances. Springer, Berlin Heidelberg, pp. 45–59.