remote sensing
Article
Seasonal to Interannual Variability of Satellite-Based
Precipitation Estimates in the Pacific Ocean
Associated with ENSO from 1998 to 2014
Xueyan Hou 1 , Di Long 1, *, Yang Hong 1,2, * and Hongjie Xie 3
1
2
3
*
State Key Laboratory of Hydroscience and Engineering, Department of Hydraulic Engineering, Tsinghua
University, Beijing 100084, China; [email protected]
School of Civil Engineering and Environmental Science, University of Oklahoma, Norman, OK 73070, USA
Department of Geological Sciences, University of Texas at San Antonio, San Antonio, TX 78204, USA;
[email protected]
Correspondence: [email protected] (D.L.); [email protected] (Y.H.);
Tel.: +86-10-6278-7394 (D.L. & Y.H.)
Academic Editors: Qiusheng Wu, Richard Gloaguen and Prasad S. Thenkabail
Received: 27 July 2016; Accepted: 28 September 2016; Published: 11 October 2016
Abstract: Based on a widely used satellite precipitation product (TRMM Multi-satellite Precipitation
Analysis 3B43), we analyzed the spatiotemporal variability of precipitation over the Pacific Ocean
for 1998–2014 at seasonal and interannual timescales, separately, using the conventional empirical
orthogonal function (EOF) and investigated the seasonal patterns associated with El Niño–Southern
Oscillation (ENSO) cycles using season-reliant empirical orthogonal function (SEOF) analysis.
Lagged correlation analysis was also applied to derive the lead/lag correlations of the first two SEOF
modes for precipitation with Pacific Decadal Oscillation (PDO) and two types of El Niño, i.e., central
Pacific (CP) El Niño and eastern Pacific (EP) El Niño. We found that: (1) The first two seasonal EOF
modes for precipitation represent the annual cycle of precipitation variations for the Pacific Ocean and
the first interannual EOF mode shows the spatiotemporal variability associated with ENSO; (2) The
first SEOF mode for precipitation is simultaneously associated with the development of El Niño and
most likely coincides with CP El Niño. The second SEOF mode lagged behind ENSO by one year
and is associated with post-El Niño years. PDO modulates precipitation variability significantly only
when ENSO occurs by strengthening and prolonging the impacts of ENSO; (3) Seasonally evolving
patterns of the first two SEOF modes represent the consecutive precipitation patterns associated with
the entire development of EP El Niño and the following recovery year. The most significant variation
occurs over the tropical Pacific, especially in the Intertropical Convergence Zone (ITCZ) and South
Pacific Convergence Zone (SPCZ); (4) Dry conditions in the western basin of the warm pool and
wet conditions along the ITCZ and SPCZ bands during the mature phase of El Niño are associated
with warm sea surface temperatures in the central tropical Pacific, and a subtropical anticyclone
dominating over the northwestern Pacific. These findings may be useful for prediction of seasonal
precipitation anomalies over the Pacific Ocean during El Niño years and recovery years.
Keywords: precipitation; seasonal evolution; interannual variability; EOF and SEOF; the Pacific Ocean
1. Introduction
Precipitation is one of the basic climate factors that drives large-scale circulation and describes
the water cycle on Earth [1,2]. It brings fresh water into the ocean, which dilutes the concentration of
salt and has considerable impacts on ocean ecosystems [1,3]. El Niño–Southern Oscillation (ENSO),
which is the dominant mode of interannual variability over the Pacific Ocean, influences convection
and precipitation with anomalously warm sea surface temperatures (SSTs) in the eastern and central
Remote Sens. 2016, 8, 833; doi:10.3390/rs8100833
www.mdpi.com/journal/remotesensing
Remote Sens. 2016, 8, 833
2 of 18
equatorial Pacific Ocean, and thus affects the sources of fresh water upon which millions of people
rely [4,5]. Therefore, understanding the variability of precipitation and its relationship with ENSO are
of great importance.
Before the advent of satellite remote sensing, studies of ocean precipitation were rare because
long-term direct measurements of precipitation over the whole oceans were unavailable. The landmark
publications of Ropelewski and Halpert [6,7] presented the global ENSO–precipitation relationship
using data from several land- and island-based stations. Advances in remote sensing have
provided us with long-term and globally complete satellite–gauge-merged precipitation products and
satellite-based estimates, such as the Global Precipitation Climatology Project (GPCP) [8], the Climate
Prediction Center Merged Analysis of Precipitation (CMAP) [9], and the Tropical Rainfall Measuring
Mission (TRMM) Multi-satellite Precipitation Analysis (TMPA) [10]. Many studies have investigated
the impacts of ENSO on precipitation using these satellite-based precipitation datasets [11–16].
For example, Dai and Wigley [11] and Haddad et al. [14] derived the global climatology and variability
of ENSO-induced precipitation using the CMAP and TRMM datasets, respectively, and found that
ENSO is the major driver of the interannual variability of global rainfall.
In recent years, many efforts have been made to examine the climatology and variability of ocean
precipitation using the TRMM dataset [4,14,17–19]. The TRMM satellite, which was launched in late
November 1997 and ended on 15 April 2015, provided a unique 17-year dataset of global tropical
rainfall. It was followed by the Global Precipitation Measurement (GPM), which aims to improve the
accuracy of extratropical precipitation estimates. The TMPA datasets are currently among the most
recognized satellite-based precipitation products available [4] and provide us with the opportunity to
understand large-scale characteristics of precipitation over the Pacific Ocean more comprehensively.
The most prominent precipitation structures in the Pacific Ocean are the Intertropical Convergence
Zone (ITCZ) and the South Pacific Convergence Zone (SPCZ), the low-pressure zones where trade
winds converge and considerable precipitation occurs because of excess solar heating and strong
convection. Dorman and Bourke [20] reported that most rainfall north of 28◦ N occurs in winter,
whereas the maximum amount of rainfall in the tropics occurs in July and August [20]. It was found
that on interannual timescales, ENSO may drive changes in the magnitudes and positions of both
the ITCZ and SPCZ. Climatologically, the ITCZ extends across the Pacific Ocean north of the equator,
while the SPCZ extends southeast from the western central Pacific to approximately 30◦ S, 120◦ W in
boreal winter [21]. During El Niño years, the ITCZ and SPCZ tend to merge in the central Pacific,
accompanied by increasing rainfall in the ITCZ and the northeastern SPCZ [21,22].
Changes in precipitation over the ocean are significantly correlated with ocean dynamics through
air–sea interactions. For example, SST anomalies, which are the expression of ENSO events and the
response to evaporation and downward solar radiation, are generally observed positively correlated
with precipitation in the tropics [15,23,24]. Precipitation variability is also generally proportional to
wind speed variability associated with ENSO; high wind speeds are associated with higher evaporation
and latent heat flux from the ocean to the atmosphere [15,25]. Recent studies have also reported that
El Niño events can be divided into two types based on the area where the largest September–February
SST anomaly occurs: central Pacific (CP) El Niño (i.e., El Niño Modoki), and eastern Pacific (EP) El Niño
(i.e., traditional El Niño) [26–28]. In contrast to the traditionally recognized El Niño, the El Niño
Modoki phenomenon is characterized by anomalously warm central equatorial Pacific flanked by
anomalously cool regions to both the west and east. Note that different types of El Niño phenomena
may produce a variety of effects in precipitation across the globe [26].
In most published studies, the variability of precipitation over the Pacific Ocean has been
investigated individually on seasonal or interannual timescales [14,23,24,29–31]. However, a more
comprehensive understanding of the seasonal cycles of ENSO–precipitation relationships is important
considering that El Niño typically peaks in boreal winter but can be strong in any season [4,13].
The objective of this study is therefore to provide a more integrated perspective of the variability
of precipitation over the Pacific on both seasonal and interannual timescales, and of the seasonal
Remote Sens. 2016, 8, 833
3 of 18
evolution of precipitation associated with the two types of El Niño events in the Pacific, based on
TRMM precipitation estimates from 1998 to 2014. The seasonal evolution patterns associated with
ocean dynamic anomalies, including anomalies in SST, sea level, and sea surface wind (SSW), are also
examined in this manuscript.
The manuscript is organized as follows. Section 2 describes the datasets and methodologies
used in this study. Results are presented in Section 3 in four parts: (1) seasonal and interannual
variability of precipitation based on empirical orthogonal function (EOF) analysis; (2) dominant
seasonal precipitation evolution associated with interannual variability derived from season-reliant
EOF (SEOF) analysis; (3) relationships between the dominant SEOF modes of precipitation and
climate indices (the ENSO index and the Pacific Decadal Oscillation index) identified through lagged
correlation analysis; and (4) seasonal evolution patterns of ocean dynamic anomalies associated with
SEOF modes of precipitation. A discussion of determining which type of El Niño dominates the
seasonal evolution of ENSO–precipitation relationships is given in Section 4. Finally, conclusions are
provided in Section 5.
2. Datasets and Methods
Various satellite products were used to analyze the variability of precipitation and the associated
ocean dynamic factors in the Pacific Ocean, including precipitation, SST, sea level anomalies (SLA), and
SSW. Table 1 summarizes the sources, timespans, and the temporal and spatial resolutions of remote
sensing datasets. These datasets include: the Tropical Rainfall Measuring Mission (TRMM) 3B43
precipitation product, Version 7, provided by the NASA Goddard Earth Sciences Data and Information
Services Center; Optimum Interpolation SST, Version 2, provided by the NOAA/OAR/ESRL Physical
Sciences Division [32]; SLA products derived from multiple satellite sensors (Jason-1, TOPEX/Poseidon,
Geosat Follow-On, European Remote Sensing 2 and Environmental Satellite) and distributed by
AVISO [33]; and the Cross-Calibrated Multi-Platform (CCMP) V1.1 ocean surface winds dataset derived
from SSM/I, TMI, AMSR-E, SeaWinds on QuikSCAT, and SeaWinds on ADEOS-II [34] and provided
by the NASA Physical Oceanography Distributed Active Archive Center (podaac.jpl.nasa.gov).
Table 1. The sources, timespans, and spatial and temporal resolutions of the datasets.
Variable
Data Source
Timespan
Resolution
Precipitation
SST
SLA
Wind
TRMM_3B43 V7
OI SST V2
AVISO
CCMP V1.1
January 1998–present
December 1981–present
December 1992–present
July 1987–December 2011
0.25◦ , monthly
1◦ , monthly
0.25◦ , monthly
0.25◦ , monthly
In addition, four indices of Pacific climate variability were used in this study, the multivariate
ENSO index (MEI) [35], the Pacific Decadal Oscillation (PDO) index [36], the El Niño Modoki index
(EMI) [37], and the Niño 3 index [38].
A number of statistical methods were used to differentiate the seasonal and interannual
variabilities of precipitation, including EOF [39] and SEOF [40] analysis. The EOF is a statistical
method used to decompose the variability of a field into sums of modes, each of which is
described as a spatial pattern and a time series [41]. To derive the variability of precipitation on
an interannual timescale, 13-month running means were used to filter out annual variation and
seasonal variability. EOF analysis was able to reveal the spatial patterns and temporal evolution
of the dominant modes on seasonal or interannual timescales individually. However, the EOF
was unable to determine the combined seasonal and interannual evolution of a single dominant
mode. Therefore, we used the SEOF, which was discussed in detail by Wang and An [40], to identify
the dominant season-dependent interannual variability modes and to depict seasonally evolving
anomalies throughout a full year. El Niño is a strongly phase-locked phenomenon with a seasonal
cycle, which initiates during the boreal spring or summer, peaks during autumn or winter, and
Remote Sens. 2016, 8, 833
4 of 18
terminates during spring [42]. Therefore, the precipitation anomalies applied in the SEOF analysis
were in the seasonal sequence of a “monsoon year” [43], beginning in the boreal summer of year 0,
denoted as June–July–August (JJA(0)), to the boreal spring of the following year (year 1), referred to as
March–April–May (MAM(1)). The precipitation anomalies for JJA(0), September–October–November
(SON(0)), December–January–February (DJF(0) or D(0)JF(1)) and MAM(1) were together treated as
a “yearly block” to construct a covariance matrix. After the EOF decomposition was performed, the
Remote Sens. 2016, 8, 833
4 of 18
yearly block was then divided into four consecutive seasonal anomalies to identify seasonally evolving
modes.(MAM(1)).
The statistical
significanceanomalies
test applied
the EOF
and SEOF modes was based
on the rule of
The precipitation
for toJJA(0),
September–October–November
(SON(0)),
or D(0)JF(1))
MAM(1) of
were
treated of
as aeach
“yearly
thumbDecember–January–February
proposed by North et al.(DJF(0)
[44]. The
relativeand
separation
thetogether
eigenvalues
mode were
1/2 , where was
block”
construct a the
covariance
matrix.
After
EOF
decomposition
the yearly and N is
examined
bytocomputing
sampling
error,
i.e.,the
λ×
(2/N)
λ is performed,
a given eigenvalue
block was then divided into four consecutive seasonal anomalies to identify seasonally evolving
the length
of time series, and checking that the spacing between the two eigenvalues is larger than the
modes. The statistical significance test applied to the EOF and SEOF modes was based on the rule of
sampling
error
of them [44,45].
thumb proposed by North et al. [44]. The relative separation of the eigenvalues of each mode were
We
also
used
the laggedthe
correlation
and i.e.,
regression
identify the
1/2, where methods
examined by computing
sampling error,
λ × (2/N)analysis
λ is a giventoeigenvalue
andrelationships
N is
between
seasonal
modes
forspacing
precipitation
climate-related
indices
and ocean
the principal
length of time
series, evolving
and checking
that the
between and
the two
eigenvalues is larger
than
the variables,
sampling error
them
[44,45].
dynamic
i.e.,ofSST,
SLA,
and SSW.
We also used the lagged correlation and regression analysis methods to identify the
relationships between principal seasonal evolving modes for precipitation and climate-related
3. Results
indices and ocean dynamic variables, i.e., SST, SLA, and SSW.
3.1. Seasonal and Interannual Variability of Precipitation Based on EOF Analysis
3. Results
The EOF was applied to the monthly precipitation anomalies to identify the leading modes
3.1. Seasonal and
Interannual
Variability
of Precipitation
Based to
on EOF
of precipitation
over
the Pacific
Ocean.
According
the Analysis
rule of North et al. [44], only the
first two seasonal
modes
were
statistically
significant
and
distinguishable.
leading
The EOF was applied to the monthly precipitation anomalies to identify theThe
leading
modestwo
of modes,
precipitation
over
the
Pacific
Ocean.
According
to
the
rule
of
North
et
al.
[44],
only
the
first
two
which account for 20.8% and 7.4% of the total variance respectively, presented dominant seasonal
seasonalinmodes
were statistically
significant
and distinguishable.
two modes
modes, which
variabilities
precipitation.
The spatial
patterns
of the firstThe
twoleading
leading
are shown in
account for 20.8% and 7.4% of the total variance respectively, presented dominant seasonal
Figure 1a,b. The corresponding temporal coefficients of the two leading modes are presented in
variabilities in precipitation. The spatial patterns of the first two leading modes are shown in Figure 1a,b.
FigureThe
1c,d.
The positive
(negative)
spatial
positive
(negative)
temporal
coefficients
corresponding
temporal
coefficients
of distributions
the two leadingand
modes
are presented
in Figure
1c,d. The
yield positive
precipitation
anomalies,
whereas
the
positive
(negative)
spatial
distributions
and
positive (negative) spatial distributions and positive (negative) temporal coefficients yield positive
precipitation
the positive
distributions
and negative (positive)
negative
(positive)anomalies,
temporalwhereas
coefficients
yield(negative)
negativespatial
precipitation
anomalies.
yield variation,
negative precipitation
Intemporal
term coefficients
of temporal
the PCanomalies.
of the first mode shown in Figure 1c indicates
In term of temporal variation, the PC of the first mode shown in Figure 1c indicates seasonal
seasonal variation
of precipitation with peaks in June–August (boreal summer) and troughs in
variation of precipitation with peaks in June–August (boreal summer) and troughs in December–
December–February
(boreal winter), whereas the PC of the second mode shown in Figure 1d exhibits
February (boreal winter), whereas the PC of the second mode shown in Figure 1d exhibits variation
variation
with
peaks
in September–November
(boreal
autumn)
troughs in
March–May
with peaks in September–November
(boreal autumn)
and
troughs and
in March–May
(boreal
spring). (boreal
spring).
The
combination
the
for modes
the two
modes
thus
describes annual
a consecutive
annual cycle of
The
combination
of the of
PCs
forPCs
the two
thus
describes
a consecutive
cycle of variation
in precipitation
over the
Pacific
variation
in precipitation
over
theOcean.
Pacific Ocean.
(a) Mode1
(b) Mode2
40oN
0.3
40oN
0.2
20oN
0.2
20oN
0.1
0.1
0o
0o
0
20oS
-0.1
20oS
40oS
-0.2
40oS
120oE
160oE
160oW
120oW
80oW
-0.1
120oE
160oE
(c)
3
Precipiation
2
1
160oW
120oW
80oW
(d)
3
Precipiation
2
1
0
0
-1
-1
-2
-2
-3
1998
0
2000
2002
2004
2006
2008
2010
2012
2014
-3
1998
2000
2002
2004
2006
2008
2010
2012
2014
Figure 1. Seasonal variability of precipitation in the Pacific Ocean characterized by the first two
Figure 1. Seasonal variability of precipitation in the Pacific Ocean characterized by the first two
empirical orthogonal function (EOF) modes: (a,b) are the spatial patterns of EOF modes (Unit: mm/h);
empirical orthogonal function (EOF) modes: (a,b) are the spatial patterns of EOF modes (Unit: mm/hr);
(c,d) are
the time series of normalized principal components (PC).
(c,d) are the time series of normalized principal components (PC).
Remote
RemoteSens.
Sens. 2016,
2016,8,
8,833
833
55 of
of18
18
Figure 1a shows the spatial variation pattern of precipitation in boreal summer and winter. The
Figure 1a shows the spatial variation pattern of precipitation in boreal summer and winter.
most notable spatial feature is the approximately symmetrical but opposite variation structure north
The most notable spatial feature is the approximately symmetrical but opposite variation structure
and south of the latitude 5°N. Specifically, there is a zonal positive anomaly band along the ITCZ
north and south of the latitude 5◦ N. Specifically, there is a zonal positive anomaly band along the
within the latitudes of 5–20°N and negative anomaly bands in the SPCZ and eastern equatorial Pacific
ITCZ within the latitudes of 5–20◦ N and negative anomaly bands in the SPCZ and eastern equatorial
between latitudes of 5°S–5°N. These patterns mean that in boreal summer, rainfall increases in the
Pacific between latitudes of 5◦ S–5◦ N. These patterns mean that in boreal summer, rainfall increases
ITCZ and tropical offshore regions, including the South China Sea, Philippine Sea, and the seas west
in the ITCZ and tropical offshore regions, including the South China Sea, Philippine Sea, and the
of Mexico and Colombia, whereas rainfall decreases within 20°S–5°N, especially in the SPCZ region
seas west of Mexico and Colombia, whereas rainfall decreases within 20◦ S–5◦ N, especially in the
and the eastern equatorial Pacific between 5°S and 5°N. For the midlatitudes of 20–50°N, precipitation
SPCZ region and the eastern equatorial Pacific between 5◦ S and 5◦ N. For the midlatitudes of 20–50◦ N,
increases along the western offshore region but decreases in the northeastern Pacific in boreal
precipitation increases along the western offshore region but decreases in the northeastern Pacific in
summer, and the reverse situation occurs in boreal winter. In the midlatitudes of the South Pacific
boreal summer, and the reverse situation occurs in boreal winter. In the midlatitudes of the South
Ocean, in contrast, precipitation increases less significantly during austral winter but decreases in
Pacific Ocean, in contrast, precipitation increases less significantly during austral winter but decreases
austral summer.
in austral summer.
Figure 1b presents the spatial variation pattern of precipitation in boreal spring and autumn.
Figure 1b presents the spatial variation pattern of precipitation in boreal spring and autumn.
One remarkable feature is the shrinking of the ITCZ in boreal autumn and a widespread decrease in
One remarkable feature is the shrinking of the ITCZ in boreal autumn and a widespread decrease
precipitation in the equatorial Pacific, whereas it increases in the SPCZ and over the Maritime
in precipitation in the equatorial Pacific, whereas it increases in the SPCZ and over the Maritime
Continent. During boreal spring, however, a double ITCZ forms: one located north and another south
Continent. During boreal spring, however, a double ITCZ forms: one located north and another south
of the equator, along the latitudes of 5°N and 5°S, respectively. Meanwhile, a narrow ridge of high
of the equator, along the latitudes of 5◦ N and 5◦ S, respectively. Meanwhile, a narrow ridge of high
pressure and reduced precipitation forms between these two convergence zones. In the midlatitudes
pressure and reduced precipitation forms between these two convergence zones. In the midlatitudes
of the Northern Hemisphere, conditions become dry on the eastern coast of China and wet across
of the Northern Hemisphere, conditions become dry on the eastern coast of China and wet across most
most of the North Pacific Ocean during boreal autumn. For the midlatitudes of the South Pacific
of the North Pacific Ocean during boreal autumn. For the midlatitudes of the South Pacific Ocean,
Ocean, precipitation increases in austral autumn but decreases in austral spring.
precipitation increases in austral autumn but decreases in austral spring.
The interannual variability of precipitation was determined using EOF analysis on 13-month
The interannual variability of precipitation was determined using EOF analysis on 13-month
running mean datasets. The first mode of precipitation over the Pacific accounts for 36.5% of the total
running mean datasets. The first mode of precipitation over the Pacific accounts for 36.5% of the total
variance. The spatial distribution and principal components of interannual variability in precipitation
variance. The spatial distribution and principal components of interannual variability in precipitation
from July 1998 to June 2014 are shown in Figure 2.
from July 1998 to June 2014 are shown in Figure 2.
(a) Mode1
o
0.1
40 N
o
20 N
0
0.05
o
0
o
20 S
-0.05
o
40 S
o
120 E
o
160 E
o
160 W
o
o
120 W
80 W
(b) r=0.8101
3
Precipiation
2
MEI
1
0
-1
-2
-3
1999
2001
2003
2005
2007
2009
2011
2013
Figure 2. The EOF mode of interannual variability of precipitation in the Pacific Ocean. (a) Spatial
Figure 2. The EOF mode of interannual variability of precipitation in the Pacific Ocean. (a) Spatial
distribution (Unit: mm/h); (b) green line represents the normalized principal component (PC) of the
distribution (Unit: mm/hr); (b) green line represents the normalized principal component (PC) of the
first EOF mode; red line represents the Multivariate ENSO index (MEI).
first EOF mode; red line represents the Multivariate ENSO index (MEI).
Remote Sens. 2016, 8, 833
6 of 18
The time
series
Remote
Sens. 2016,
8, 833of the PC has a close relationship with ENSO, with a correlation coefficient
6 of 18
of 0.81 (passing the 95% Student’s t-test). This correlation indicates that this EOF mode reveals
The timeprecipitation
series of the PC
has a close
relationship
withENSO.
ENSO,The
withpositive
a correlation
coefficient
of
the interannual
variability
associated
with
PC values
plotted
in
0.81 (passing the 95% Student’s t-test). This correlation indicates that this EOF mode reveals the
Figure 2b coincide with the El Niño years 2002–2003, 2004–2005, 2006–2007, 2009–2010, 2012–2013, and
interannual precipitation variability associated with ENSO. The positive PC values plotted in Figure 2b
2014–2015, and the negative values coincide with the La Niña years 1998–1999, 1999–2001, 2007–2008,
coincide with the El Niño years 2002–2003, 2004–2005, 2006–2007, 2009–2010, 2012–2013, and 2014–2015,
and 2010–2011
[46,47].
The spatial
in Figure
2a is characterized
the significantly
and the negative
values
coincidepattern
with theshown
La Niña
years 1998–1999,
1999–2001, by
2007–2008,
and
positive,
wedge-shaped,
westward-pointing
feature
centered
along
the
equator
in
the
central
tropical
2010–2011 [46,47]. The spatial pattern shown in Figure 2a is characterized by the significantly
positive,
Pacific
that extendswestward-pointing
eastward along the
IPCZcentered
during along
El Niño
negative
precipitation
wedge-shaped,
feature
the years.
equatorMeanwhile,
in the centralatropical
Pacific
that
anomaly
exists
in thealong
western
region
of the
warm
pool,
i.e., overathe
Maritime
Continent
and seas
extends
eastward
the IPCZ
during
El Niño
years.
Meanwhile,
negative
precipitation
anomaly
exists in
western region
of the warm
pool,precipitation
i.e., over the Maritime
Continent
seas northeast
northeast
of the
Australia.
The reverse
of these
anomalies
occursand
during
La Niñaofyears.
Australia.
The reverse
of these precipitation
anomalies
occurs during
La Niña
years.
Overall,
thisyears.
EOF
Overall,
this EOF
mode describes
the precipitation
variability
during
El Niño
and
La Niña
mode describes the precipitation variability during El Niño and La Niña years.
3.2. Dominant Seasonally Evolving Precipitation Patterns from SEOF Analysis
3.2. Dominant Seasonally Evolving Precipitation Patterns from SEOF Analysis
The season-dependent interannual variability of precipitation in the Pacific Ocean was evaluated
The season-dependent interannual variability of precipitation in the Pacific Ocean was evaluated
through SEOF analysis. The first two leading modes account for 25.2% and 7.8% of the total variance,
through SEOF analysis. The first two leading modes account for 25.2% and 7.8% of the total variance,
respectively.
According to the rule of North et al. (1982), these two modes were well distinguishable
respectively. According to the rule of North et al. (1982), these two modes were well distinguishable
fromfrom
eacheach
other
andand
from
thethe
remaining
the 95%
95%confidence
confidence
level.
other
from
remainingmodes
modes at
at the
level.
Figure
3 shows
thethe
time
of the
thefirst
firsttwo
twoSEOF
SEOF
modes
of precipitation
Figure
3 shows
timeseries
seriesofofthe
the PCs
PCs of
modes
of precipitation
in thein the
Pacific.
BothBoth
of these
modes
show
variabilities.
PC1
of the
SEOF
(SEOF-PC1)
Pacific.
of these
modes
showprominent
prominent interannual
interannual variabilities.
PC1
of the
SEOF
(SEOF-PC1)
reflects
the ENSO
cycle,
withpositive
positiveand
and negative
negative values
to El
and and
La Niña
reflects
the ENSO
cycle,
with
valuescorresponding
corresponding
toNiño
El Niño
La Niña
years,
respectively.
This
relationship
indicates
that
the
first
SEOF
mode
(SEOF1)
describes
the
years, respectively. This relationship indicates that the first SEOF mode (SEOF1) describes the seasonal
seasonal
evolution of precipitation
associated
with
of the
SEOF (SEOF-PC2)
correlates
evolution
of precipitation
associated with
ENSO.
PC2ENSO.
of thePC2
SEOF
(SEOF-PC2)
correlates
closely with
closely with PC1 with a lag time of approximately one year, with a lagged correlation coefficient of
PC1 with a lag time of approximately one year, with a lagged correlation coefficient of 0.6 (which
0.6 (which passes the 95% Student’s t-test); therefore, the second SEOF mode (SEOF2) lags behind
passes the 95% Student’s t-test); therefore, the second SEOF mode (SEOF2) lags behind SEOF1 by
SEOF1 by about one year. Figure 4 shows the seasonally evolving spatial patterns of the first two
aboutSEOF
one modes
year. Figure
4 shows the
seasonally
of the
firstpattern
two SEOF
modes of
of precipitation.
There
is notableevolving
similarityspatial
betweenpatterns
the general
spatial
of SEOF1
precipitation.
There
is
notable
similarity
between
the
general
spatial
pattern
of
SEOF1
in
Figure
4a and
in Figure 4a and the precipitation pattern of the first EOF mode in Figure 2a. The main difference is
the precipitation
pattern
of
the
first
EOF
mode
in
Figure
2a.
The
main
difference
is
that
the
spatial
that the spatial pattern of the SEOF mode depicts the seasonal evolution of precipitation with
pattern
of the SEOF
mode Because
depictsthe
thepositive
seasonal
evolution
of precipitation
withtointerannual
variability.
interannual
variability.
values
of the SEOF-PC1
correspond
El Niño years,
the
spatial
Figureof
4athe
reveal
the seasonal
evolution
El
Because
thepatterns
positiveinvalues
SEOF-PC1
correspond
to of
El precipitation
Niño years, associated
the spatialwith
patterns
in
Niño
Figure
4a events.
reveal the seasonal evolution of precipitation associated with El Niño events.
2
(a)
1
0
-1
-2
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
2
(b)
1
0
-1
-2
-3
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Year
Figure
3. (a)
season-reliant
empirical
orthogonal
function
Figure
3. Normalized
(a) NormalizedPCs
PCsof
ofthe
the first
first season-reliant
empirical
orthogonal
function
(SEOF) (SEOF)
mode formode
precipitation inin
thethe
Pacific
Ocean
from summer
1998 to spring
(b) as in2014;
(a), but(b)
for as
thein
second
mode.
for precipitation
Pacific
Ocean
from summer
19982014;
to spring
(a), but
for the
second mode.
Remote Sens. 2016, 8, 833
Remote Sens. 2016, 8, 833
7 of 18
7 of 18
Figure
4. (a)
Spatial
patternsofofthe
thefirst
firstSEOF
SEOF mode
mode (SEOF1)
precipitation
anomalies
fromfrom
Figure
4. (a)
Spatial
patterns
(SEOF1)ofofseasonal
seasonal
precipitation
anomalies
June–July–August
of
year
0
(JJA(0))
to
March–April–May
of
the
following
year
(MAM(1));
(b)
June–July–August of year 0 (JJA(0)) to March–April–May of the following year (MAM(1)); (b) as
asin
in (a),
(a), but for the second mode (SEOF2). Units: mm/hr.
but for the second mode (SEOF2). Units: mm/h.
The most notable feature in Figure 4a is that the positive precipitation anomaly dominates the
The most
notable
feature
in equatorial
Figure 4a Pacific,
is that the
positive
precipitation
dominates
ITCZ
and SPCZ
regions
in the
while
the negative
anomalyanomaly
mainly occupies
the the
ITCZ
and SPCZ
regions
in the
Pacific,
whileIslands.
the negative
anomaly mainly
occupies
Maritime
Continent
region
andequatorial
southeast of
the Solomon
More specifically,
during the
boreal the
summer
of year zero,
i.e., and
JJA(0),
above-average
occurs over
regions eastduring
of the the
Maritime
Continent
region
southeast
of the precipitation
Solomon Islands.
Morethespecifically,
Philippines,
as
well
as
along
the
ITCZ
and
SPCZ
bands.
During
this
period,
a
below-average
boreal summer of year zero, i.e., JJA(0), above-average precipitation occurs over the regions east of
precipitation pattern
onlythe
around
islands
Indonesia
and the
western
coast
of Mexico.
the Philippines,
as welloccurs
as along
ITCZthe
and
SPCZofbands.
During
this
period,
a below-average
For the boreal
autumn,
i.e.,only
SON(0),
however,
anomalously
dry conditions
to coast
a largeofarea,
precipitation
pattern
occurs
around
the islands
of Indonesia
and theexpand
western
Mexico.
especially in the western Pacific from the South China Sea to the Maritime Continent. Meanwhile, the
For the boreal autumn, i.e., SON(0), however, anomalously dry conditions expand to a large area,
ITCZ band of high precipitation narrows and moves slightly equatorward. During the following
especially in the western Pacific from the South China Sea to the Maritime Continent. Meanwhile,
boreal winter, i.e., DJF(1), both the positive and negative precipitation variation patterns of the tropics
the ITCZ band of high precipitation narrows and moves slightly equatorward. During the following
are strengthened, and the development of El Niño reaches its mature phase. Precipitation increased
boreal
winter,
DJF(1),
both the
negative
variation
of the
tropics
in both
thei.e.,
ITCZ
and SPCZ,
but positive
the bandand
regions
northprecipitation
of the ITCZ and
south patterns
of the SPCZ
in the
are strengthened,
and
the
development
of
El
Niño
reaches
its
mature
phase.
Precipitation
increased
central Pacific experience declines in precipitation. During the subsequent boreal spring, i.e., in
bothMAM(1),
the ITCZthe
and
SPCZ, butdry
the region
band regions
of the
ITCZ and
south of the
SPCZ
the central
anomalously
extends north
from the
Philippines
to Australia
across
the in
Maritime
Pacific
experience
precipitation.
During
subsequent
borealthe
spring,
i.e.,part
MAM(1),
Continent
in thedeclines
western in
Pacific.
Meanwhile,
in the the
central
tropical Pacific,
western
of the the
SPCZ
starts
to
shrink,
and
the
diagonally-oriented
eastern
portion
of
the
SPCZ
that
extends
into
anomalously dry region extends from the Philippines to Australia across the Maritime Continentthe
in the
sub-tropics
to in
experience
high
precipitation.
Note
that the
equatorial
western
Pacific. continues
Meanwhile,
the central
tropical
Pacific, the
western
parteastern
of the SPCZ
startsPacific
to shrink,
3°S and 10°S andeastern
east of portion
150°W receives
lowerthat
than
averageinto
precipitation
duringcontinues
austral to
and between
the diagonally-oriented
of the SPCZ
extends
the sub-tropics
spring
in
El
Niño
years.
A
negative
precipitation
anomaly
also
occurs
in
the
midlatitude
Pacific
◦
◦
experience high precipitation. Note that the eastern equatorial Pacific between 3 S and 10 SOcean
and east
in ◦both the Northern and Southern Hemispheres. The negative anomalies are found mainly over the
of 150
W receives lower than average precipitation during austral spring in El Niño years. A negative
Maritime Continent and the South Pacific during the onset of El Niño events, but these anomalies
precipitation anomaly also occurs in the midlatitude Pacific Ocean in both the Northern and Southern
shift to the North Pacific during the decay stage [13]. The opposite of the seasonal evolution of
Hemispheres. The negative anomalies are found mainly over the Maritime Continent and the South
precipitation associated with El Niño development described above occurs with La Niña.
Pacific during
thethe
onset
of El
Niño events,
butSEOF-PC1
these anomalies
shift toitthe
North Pacific
Based on
lagged
correlation
between
and SEOF-PC2,
is inferred
that theduring
spatial the
decay
stageof[13].
The(see
opposite
of the
seasonal
evolution
precipitation
associated
witha El
Niño
pattern
SEOF2
Figure 4b)
is one
year out
of phaseofwith
SEOF1. Figure
4b features
less
development described above occurs with La Niña.
Based on the lagged correlation between SEOF-PC1 and SEOF-PC2, it is inferred that the spatial
pattern of SEOF2 (see Figure 4b) is one year out of phase with SEOF1. Figure 4b features a less
Remote Sens. 2016, 8, 833
8 of 18
significant spatial variation pattern that is approximately opposite to that of SEOF1 during the period
from JJA(0) to DJF(1). In the boreal summer, negative precipitation anomalies occur over the seas east
of the Philippines and north of Papua New Guinea and extend eastward along latitudes of 3–5◦ N to
◦
◦
the Columbian
Remote Sens.coast
2016, 8,and
833 southeastward to the southeast Pacific (~30 S, 120 W). This pattern
8 of 18 implies
that the ITCZ in the northern hemisphere moves northward to about 10◦ N and that the SPCZ starts
significant spatial variation pattern that is approximately opposite to that of SEOF1 during the period
to shrink. The northwestern Pacific also receives above average precipitation during the summer.
from JJA(0) to DJF(1). In the boreal summer, negative precipitation anomalies occur over the seas east
During of
SON(0),
wet conditions occur on the coasts of Asia, especially over the South China Sea and
the Philippines and north of Papua New Guinea and extend eastward along latitudes of 3–5°N to
the Maritime
Continent
Meanwhile,
thesoutheast
band that
extends
northeastward
fromimplies
New Guinea
the Columbian
coastregion.
and southeastward
to the
Pacific
(~30°S,
120°W). This pattern
to the central
Pacific
continues
to experience
dry conditions.
Byand
thethat
boreal
winter,
these dry
that thetropical
ITCZ in the
northern
hemisphere
moves northward
to about 10°N
the SPCZ
starts
to shrink.
northwestern
Pacific
also receives
average
precipitation
during
summer.Islands,
conditions
occurThe
over
the Maritime
Continent
andabove
north
of New
Guinea and
thethe
Solomon
During
SON(0), wetwet
conditions
occur on
the coaststhe
of Asia,
especially
overPacific.
the SouthThe
China
Sea and
while the
anomalously
conditions
dominate
central
tropical
ITCZ
band east
the Maritime Continent region. Meanwhile, the band that extends northeastward from New Guinea
◦
of 160 W moves closer to the equator during this season. A circle of above-average precipitation
to the central tropical Pacific continues◦ to experience
dry◦conditions. By the boreal winter, these dry
occurs in
the North
around
140Continent
E–170◦and
W, 8–30
andGuinea
extends
toward the
conditions
occurPacific
over the
Maritime
north N,
of New
andnortheastward
the Solomon Islands,
California
coast.
During the wet
following
spring,
the ITCZ
in thetropical
Northern
Hemisphere
moves
while
the anomalously
conditions
dominate
the central
Pacific.
The ITCZ band
eastnorthward
of
160°Wfrom
movesthe
closer
to the equator
during this
A circle
above-average
precipitation
and extends
Philippines
eastward
to season.
Mexico,
whileofmost
of the SPCZ
regionoccurs
experiences
in the North
Pacific around
140°E–170°W, 8–30°N, and extends northeastward toward the California
below-average
precipitation
conditions.
coast. During the following spring, the ITCZ in the Northern Hemisphere moves northward and
extends from
the Philippines
eastward
to Mexico,
most of the
SPCZ region experiences below3.3. Relationships
between
the Dominant
SEOF
Modeswhile
and Climate
Indices
average precipitation conditions.
To further demonstrate the relationships between the dominant SEOF modes of precipitation
3.3. Relationships
the Dominant
SEOF Modes
and Climate
Indices
and climate
variability,between
the lagged
correlations
between
the first
two precipitation SEOF-PCs and the
MEI and PDO
index are
shown in
5. Here,
Year(the
−1)dominant
and Year(1),
the abbreviations
To further
demonstrate
theFigure
relationships
between
SEOFwhich
modes are
of precipitation
variability,
the lagged
correlations
between the
two precipitation
and the by one
of Year(and
−1)climate
and Year(1),
indicate
that
the SEOF-PCs
lagsfirst
behind
or leads SEOF-PCs
its counterpart
MEI
and
PDO
index
are
shown
in
Figure
5.
Here,
Year(−1)
and
Year(1),
which
are
the
abbreviations
year, respectively. Year(0) , which is the abbreviation of Year(0), denotes simultaneous correlation.
of Year(−1) and Year(1), indicate that the SEOF-PCs lags behind or leads its counterpart by one year,
The lagged correlations between the first SEOF mode and the MEI and PDO index are presented
respectively. Year(0) , which is the abbreviation of Year(0), denotes simultaneous correlation. The
in Figure
5a as
red and between
blue lines,
respectively.
Note
the PDO
correlation
between the
lagged
correlations
the first
SEOF mode and
thethat
MEI and
index arecoefficients
presented in Figure
SEOF-PC1
and
in FigureNote
5a) are
and exceed
the 5%
significance
level from JJA
5a as
redthe
andMEI
blue(red
lines,line
respectively.
thatpositive
the correlation
coefficients
between
the SEOF-PC1
MEI (red
line in indicates
Figure 5a) are
positive
exceed the 5% significance
from JJA
to precipitation
MAM
to MAMand
forthe
Year(0),
which
that
there and
is simultaneous
correlationlevel
between
the
for
Year(0),
which
indicates
that
there
is
simultaneous
correlation
between
the
precipitation
SEOF1
SEOF1 and ENSO. Note that the SEOF1 and the MEI reach their maximum correlation coefficient of
and ENSO. Note that the SEOF1 and the MEI reach their maximum correlation coefficient of 0.97 in
0.97 in DJF(0).
ENSO exhibits significant phase locking to the seasonal cycle; it starts in boreal spring,
DJF(0). ENSO exhibits significant phase locking to the seasonal cycle; it starts in boreal spring,
maturesmatures
in boreal
winter and decays in following spring or summer [48]. Because ENSO events tend
in boreal winter and decays in following spring or summer [48]. Because ENSO events tend
to reachtotheir
at the
of the
year,
thatthe
the
SEOF1
is primarily
associated
reachpeaks
their peaks
at end
the end
of the
year,we
wespeculate
speculate that
SEOF1
is primarily
associated
with with
episodes
of ENSO
development.
episodes
of ENSO
development.
(a) corr (SEOF1, MEI & PDO)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
JJA
Y(-1)
SON
DJF
Y(0)
MAM
JJA
SON
DJF
Y(1)
MAM
JJA
SON
JJA
SON
DJF
MAM
(b) corr (SEOF2, MEI & PDO)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
JJA
Y(-1)
SON
DJF
Y(0)
MAM
JJA
SON
DJF
Y(1)
MAM
DJF
MAM
Figure
5. (a) Lead–lag
correlations
betweenthe
the first
first SEOF
and
thethe
MEIMEI
(red (red
line) and
Figure 5.
(a) Lead–lag
correlations
between
SEOFPC
PC
and
line)PDO
andindex
PDO index
(blue line); (b) As in (a), but for the second mode. The dashed lines indicate at the 5% significance
(blue line); (b) As in (a), but for the second mode. The dashed lines indicate at the 5% significance level
level of the Student’s t-test. Year(−1) (Year(1)) denotes that the PC lags behind (leads) its counterpart
of the Student’s
t-test. Year(−1) (Year(1)) denotes that the PC lags behind (leads) its counterpart by one
by one year; Year(0) denotes simultaneous correlation. JJA is boreal summer, SON is boreal autumn,
year; Year(0)
denotes
simultaneous
DJF is boreal
winter,
and MAM iscorrelation.
boreal spring.JJA is boreal summer, SON is boreal autumn, DJF is
boreal winter, and MAM is boreal spring.
Remote Sens. 2016, 8, 833
9 of 18
Lagged correlations between the SEOF-PC1 and the PDO index (blue line in Figure 5a) are
significant from boreal autumn, i.e., SON(0) to the summer of the next year, i.e., JJA(1). This lagged
relationship indicates that PDO plays a crucial role in modulating the variability of precipitation on
an interannual timescale. The correlations between the SEOF1 and the PDO index are weaker than
those between the SEOF1 and the MEI and lag by one season, which signifies that PDO impacts the
variability of precipitation from boreal autumn to the following spring and prolongs the impact of
ENSO on precipitation by one season, specifically during the ENSO-decaying summer.
Lagged correlations of the SEOF-PC2 with the MEI and PDO index are shown in Figure 5b as
red and blue lines, respectively. The simultaneous correlations between them are not statistically
significant at the 5% level, but the lagged correlation is significant and positive from JJA to MAM in
Year(−1), i.e., the SEOF2 lags behind ENSO by one year. This trend is consistent with the previous
conclusions that the SEOF2 lags behind the SEOF1 by approximately one year and that the SEOF1
correlates simultaneously with ENSO. Therefore, the SEOF2 corresponds to the years immediately
after ENSO events. The lagged correlation between the SEOF2 and the PDO index are not statistically
significant at the 5% level, which indicates that PDO has no significant effect on the variability of
precipitation during post-ENSO years.
3.4. Seasonally Evolving Patterns of Ocean Dynamic Anomalies Associated with SEOF Modes of Precipitation
The seasonally evolving patterns of ocean dynamic factors (SST, SLA and SSW) in the Pacific
associated with the first two SEOF modes of precipitation are shown in Figures 6–8. The anomalies of
these factors were regressed upon the SEOF-PC time series. To include the lagged correlations between
SEOF modes and ENSO, we exhibit the simultaneous ocean dynamic patterns associated with SEOF1
in Figure 6, the one-year-leading ocean dynamic patterns associated with SEOF2 in Figure 7, and the
simultaneous ocean dynamic patterns associated with SEOF2 in Figure 8.
3.4.1. SEOF1
The seasonally evolving patterns of SST anomalies from boreal summer, i.e., JJA(0), to the
following spring, i.e., MAM(1), are shown in Figure 6a. These anomalies demonstrate the development
of the El Niño event, most likely El Niño Modoki, with strong anomalous warming starting in the
central tropical Pacific [26]. Specifically, this figure shows the seasonal evolution of El Niño Modoki
as it initiates during JJA(0), develops during SON(0), peaks in D(0)JF(1) and decays in MAM(1).
In the boreal summer, i.e., JJA(0), the positive SST anomaly pattern is centered mainly on the central
equatorial Pacific and the western American coast, while the negative SST anomaly is less apparent
and only appears east of Japan and in the midlatitude South Pacific. In boreal autumn, i.e., SON(0),
El Niño develops with positive SST anomalies extending eastward along the equator toward the South
American coast and negative anomalies expanding to the Kuroshio extension, the Maritime Continent,
and southeastward in a band to the subtropical South Pacific. In the boreal winter, i.e., D(0)JF(1),
El Niño reaches its mature phase, which is characterized by the most significant expansion of positive
SST anomalies in the central and eastern equatorial Pacific. Note that the SST pattern of the PDO warm
phase is also depicted, i.e., the anomalously cold SSTs in the central North Pacific and anomalously
warm SSTs along the west coast of the Americas [49]. During the spring of the following year, i.e.,
MAM(1), positive SST anomalies in the tropics start to decay, but the warm phase of PDO continues to
dominate the North Pacific. The significant PDO pattern in the North Pacific could only be seen from
SON(0) to MAM(1), which is consistent with the conclusion revealed by Figure 5a that the SEOF1 is
significantly correlated with the PDO index only from SON(0) to JJA(1). In fact, PDO experiences its
highest energetic fluctuation during boreal winter and spring [49,50]. Therefore, the first SEOF mode
of precipitation was associated with the development of El Niño and the warm phase of PDO.
Remote Sens. 2016, 8, 833
Remote Sens. 2016, 8, 833
10 of 18
10 of 18
(a)
JJA(0)
(b)
JJA(0)
40o N
40oN
20o N
20o N
20oN
0o
0o
0o
20oS
20o S
20o S
40oS
40o S
120o E
160o E
160oW
120oW
80o W
40o S
120oE
SON(0)
160oE
160o W
120o W
80o W
120o E
SON(0)
40o N
40o N
40oN
20o N
20oN
0o
0o
0o
20oS
20o S
20o S
40o S
120o E
160o E
160oW
120oW
80o W
160oE
160o W
120o W
80o W
120o E
D(0)JF(1)
40o N
40o N
40oN
20o N
20oN
0o
0o
0o
20oS
20o S
20o S
40o S
120o E
160o E
160oW
120oW
80o W
160oE
160o W
120o W
80o W
120o E
MAM(1)
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20oS
20o S
20o S
40o S
120o E
-1.5
-1
160o E
-0.5
160oW
0
120oW
0.5
1
80o W
(℃)
1.5
160o E
160o W
120o W
80o W
160o W
120o W
80o W
160o W
120o W
80oW
160o E
MAM(1)
40o N
40oS
80o W
40o S
120oE
MAM(1)
120o W
D(0)JF(1)
20o N
40oS
160o W
40o S
120oE
D(0)JF(1)
160o E
SON(0)
20o N
40oS
(c)
JJA(0)
40o N
40o S
120oE
-0.2
160oE
-0.1
160o W
0
120o W
0.1
80o W
0.2
(m)
120oE
-1
160o E
-0.5
0
0.5
(m/s)
1
Figure 6. Seasonally
Seasonally evolving
evolving patterns
patterns of
of (a)
(a) SST;
SST; (b) sea level; and (c) SSW anomalies from JJA(0) to
MAM(1) regressed with the PC of the first SEOF mode of
of precipitation.
precipitation.
The
The simultaneous
simultaneous seasonally
seasonally evolving
evolving patterns
patterns of
of SLA
SLA regressed
regressed with
with the
the SEOF-PC1
SEOF-PC1 are
are shown
shown
in
Figure
6b.
This
figure
exhibits
the
typical
SLA
structure
of
El
Niño,
i.e.,
the
sea
level
was
in Figure 6b. This figure exhibits the typical SLA structure of El Niño, i.e., the sea level was lower
lower
than
average
in
the
western
tropical
Pacific
basin
and
higher
than
average
in
the
central
and
eastern
than average in the western tropical Pacific basin and higher than average in the central and eastern
tropical
Pacific basin
basin[51,52].
[51,52].More
Morespecifically,
specifically,
negative
SLA
expands
eastward
toward
tropical Pacific
thethe
negative
SLA
expands
eastward
toward
160◦160°E,
E, and
and
the
center
of
the
positive
anomaly
also
moves
eastward
from
the
boreal
summer
to
winter.
the center of the positive anomaly also moves eastward from the boreal summer to winter. During the
During
the
decaying
ofthe
El Niño
in thespring,
following
sea level
startsto
toits
recover
toconditions,
its normal
decaying
phase
of El phase
Niño in
following
sea spring,
level starts
to recover
normal
conditions,
as
shown
by
declines
in
the
sizes
and
magnitudes
of
both
the
positive
and
negative
SLA
as shown by declines in the sizes and magnitudes of both the positive and negative SLA regions.
regions.
SLAresembles
pattern resembles
the composite
SLAcentral
for thePacific
centralElPacific
i.e., Modoki,
El Niño
This SLAThis
pattern
the composite
SLA for the
Niño, El
i.e.,Niño,
El Niño
Modoki,
described
by
Chang
et
al.
[53]
(see
their
Figures
5b
and
7),
from
which
we
can
further
described by Chang et al. [53] (see their Figures 5b and 7), from which we can further infer thatinfer
the
that
first
SEOF
of precipitation
is more
likely associated
with Modoki
El Niño rather
Modoki
rather
first the
SEOF
mode
ofmode
precipitation
is more likely
associated
with El Niño
than
withthan
the
with
the
traditional
El
Niño.
traditional El Niño.
Figure
6c depicts
depicts the
the seasonally
seasonallyevolving
evolvingpatterns
patternsofofwind
windanomalies
anomaliesfrom
from
the
boreal
summer
Figure 6c
the
boreal
summer
to
to
the
following
spring
in
relation
to
the
precipitation
SEOF1.
In
general,
this
figure
reveals
the
ENSO
the following spring in relation to the precipitation SEOF1. In general, this figure reveals the ENSO
patterns
circulation cells
cells over
over the
the tropical
tropical Pacific
Pacific and
and wind
wind variability
patterns related
related to
to the
the Walker
Walker circulation
variability over
over the
the
midlatitudes
[54,55].
During
the
boreal
summer,
i.e.,
JJA(0),
westerly
winds
strengthen
in
the
midlatitudes [54,55]. During the boreal summer, i.e., JJA(0), westerly winds strengthen in the western
western
basin
South China
China Sea
Sea and
and the
the Maritime
Maritime Continent,
Continent, while
wind
basin of
of the
the warm
warm pool,
pool, mainly
mainly over
over the
the South
while wind
speeds
weaken
and
reverse
direction
within
160°E–160°W
along
the
equator.
From
JJA(0)
to
D(0)JF(1),
◦
◦
speeds weaken and reverse direction within 160 E–160 W along the equator. From JJA(0) to D(0)JF(1),
the
in the
the central
central and
and eastern
eastern equatorial
equatorial
the ongoing
ongoing wind
wind anomalies
anomalies help
help to
to reinforce
reinforce the
the warm
warm SSTs
SSTs in
Pacific
lead to
to the
the maturation
maturation of
of El
El Niño.
Niño. Overall,
Pacific and
and therefore
therefore lead
Overall, the
the weakened
weakened trade
trade winds
winds in
in the
the
central
tropical
Pacific
coincide
with
the
increasing
precipitation.
The
negative
wind
anomaly
central tropical Pacific coincide with the increasing precipitation. The negative wind anomaly that
that
extends
to the
the North
extends northeastward
northeastward and
and southeastward
southeastward to
North and
and South
South American
American coasts
coasts is
is also
also consistent
consistent
with
positive precipitation
precipitation anomaly
anomaly in
the ITCZ
ITCZ and
SPCZ shown
shown in
Figure 4a.
4a. Note
with the
the positive
in the
and SPCZ
in Figure
Note that
that aa band
band
of
Pacific
along
thethe
equator.
This
band
appears
during
the
of strong
strong easterly
easterlywinds
windsoccurs
occursininthe
theeastern
eastern
Pacific
along
equator.
This
band
appears
during
boreal
summer,
develops
during
autumn
and extends
to the to
Peruvian
coast during
the boreal
the boreal
summer,
develops
during
autumn
and extends
the Peruvian
coast during
thewinter.
boreal
However,
these
strong
winds
vanish
during
the
decaying
period
of
El
Niño
in
the
following
spring
winter. However, these strong winds vanish during the decaying period of El Niño in the following
when a negative precipitation anomaly emerges in the band region. In the boreal autumn and winter,
the subtropical anticylone develops and dominates the northwestern Pacific. During winter, a
Remote Sens. 2016, 8, 833
11 of 18
spring when a negative precipitation anomaly emerges in the band region. In the boreal autumn and
Remote Sens. 2016, 8, 833
11 of 18
winter,
the subtropical anticylone develops and dominates the northwestern Pacific. During
winter,
a southwesterly wind anomaly over the East China Sea and the South China Sea weakens the prevailing
southwesterly wind anomaly over the East China Sea and the South China Sea weakens the
winds, which is associated with an increase in precipitation over the coastal waters of southeastern
prevailing winds, which is associated with an increase in precipitation over the coastal waters of
China.
An anomalous cyclone occupies the midlatitude North Pacific during the boreal winter and
southeastern China. An anomalous cyclone occupies the midlatitude North Pacific during the boreal
following
spring
whenspring
the precipitation
is below average.
The reverse
of theseofanomalies
appears
winter and
following
when the precipitation
is below average.
The reverse
these anomalies
during
La
Niña
events.
appears during La Niña events.
3.4.2.
SEOF2
3.4.2.
SEOF2
The
seasonally
variablesassociated
associatedwith
withthe
the
second
SEOF
mode
The
seasonallyevolving
evolving oceanic
oceanic dynamic
dynamic variables
second
SEOF
mode
areare
shown
in
Figures
7
and
8.
Figure
7
shows
the
one-year-ahead
evolution
patterns
of
SST,
sea
level
shown in Figures 7 and 8. Figure 7 shows the one-year-ahead evolution patterns of SST, sea level and
and
SSW
anomalies,
which
significantly
correlated
with
ENSO.
Theone-year-leading
one-year-leading
seasonally
SSW
anomalies,
which
areare
significantly
correlated
with
ENSO.
The
seasonally
evolving
SST
from JJA
JJAin
inYear
Year(−1)
(−1)
MAM
Year(0)
shown
evolving
SSTanomalies
anomaliesassociated
associatedwith
with the
the SEOF2 from
toto
MAM
in in
Year(0)
areare
shown
in in
Figure
7a.
This
SST
anomaly
pattern
represents
the
development
of
a
conventional
El
Niño
event,
Figure 7a. This SST anomaly pattern represents the development of a conventional El Niño event,
i.e.,
the
eastern
warmerin
inthe
theeastern
easternequatorial
equatorial
Pacific
during
i.e.,
the
easternPacific
Pacific(EP)
(EP)El
ElNiño.
Niño. SSTs
SSTs become
become warmer
Pacific
during
thethe
boreal
summer,continue
continuetotodevelop
develop in
in the
the eastern
eastern and
inin
thethe
boreal
autumn
boreal
summer,
and central
centralequatorial
equatorialPacific
Pacific
boreal
autumn
until
ENSOreaches
reachesthe
themature
mature phase
phase in the boreal
normal
in in
thethe
boreal
until
ENSO
boreal winter,
winter,and
andfinally
finallyreturn
returntoto
normal
boreal
spring.
Because
the
SEOF2
corresponds
to
the
post-ENSO
years
in
Figure
5b,
we
further
infer
that
the
spring. Because the SEOF2 corresponds to the post-ENSO years in Figure 5b, we further infer
that
SEOF
mode
of precipitation
is associated
with post-EP
El Niño
a warm aPDO
thesecond
second
SEOF
mode
of precipitation
is associated
with post-EP
Elyears.
Niño Moreover,
years. Moreover,
warm
pattern
withwith
cold cold
SSTs SSTs
in theinNorth
PacificPacific
and warm
along
western
AmericanAmerican
coast is shown
PDO
pattern
the North
and SSTs
warm
SSTsthe
along
the western
coast is
in
Figure
7a
from
SON(−1)
and
MAM(0).
Therefore,
although
the
lagged
correlation
between
the
shown in Figure 7a from SON(−1) and MAM(0). Therefore, although the lagged correlation
between
SEOF2-PC
and
the
PDO
index
is
not
statistically
significant
in
Figure
5b,
PDO
may
be
an
important
the SEOF2-PC and the PDO index is not statistically significant in Figure 5b, PDO may be an important
factor, nonetheless, and may dominate ocean precipitation in the year after an ENSO event.
factor, nonetheless, and may dominate ocean precipitation in the year after an ENSO event.
40o N
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
40o S
120o E
160o E
160o W
120o W
80o W
40o S
120o E
160o E
160o W
120o W
80o W
40o N
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
40o S
120o E
160o E
160o W
120o W
80o W
160o E
160o W
120o W
80o W
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
120o E
160o E
160o W
120o W
80o W
160o E
160o W
120o W
80o W
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
120o E
-1.5
-1
160o E
-0.5
160o W
0
120o W
0.5
1
160o W
120o W
80o W
120o E
160o E
160o W
120o W
80o W
120o E
160o E
160o W
120o W
80o W
120o E
160o E
160o W
120o W
80o W
40o S
120o E
40o N
40o S
160o E
40o S
120o E
40o N
40o S
120o E
40o S
80o W
120o E
160o E
1.5
-0.2
-0.1
160o W
0
120o W
0.1
80o W
0.2
-1
-0.5
0
0.5
1
Figure
of (a)
(a) SST;
SST; (b)
(b) sea
sealevel;
level;and
and(c)(c)SSW
SSWanomalies
anomalies
from
Figure7. 7.Seasonally
Seasonally evolving
evolving patterns
patterns of
from
oneone
year
prior
totoprecipitation,
−1) to
to MAM(0),
MAM(0),regressed
regressedwith
withthe
thePC
PCofof
the
second
SEOF
mode
year
prior
precipitation,i.e.,
i.e.,JJA(
JJA(−1)
the
second
SEOF
mode
of of
precipitation.
precipitation.
RemoteSens.
Sens.2016,
2016,8,8,833
833
Remote
12of
of18
18
12
40o N
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
40o S
120o E
160o E
160o W
120o W
80o W
40o S
120o E
160o E
160o W
120o W
80o W
40o N
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
40o S
120o E
160o E
160o W
120o W
80o W
160o E
160o W
120o W
80o W
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20o S
40o S
120o E
160o E
160o W
120o W
80o W
160o E
160o W
120o W
80o W
40o N
40o N
20o N
20o N
20o N
0o
0o
0o
20o S
20o S
20 S
40o S
40o S
120o E
-1.5
-1
160o E
-0.5
160o W
0
120o W
0.5
1
80o W
1.5
160o W
120o W
80o W
120o E
160o E
160o W
120o W
80o W
120o E
160o E
160o W
120o W
80o W
120o E
160o E
160o W
120o W
80o W
40o S
120o E
40o N
o
160o E
40o S
120o E
40o N
40o S
120o E
40o S
120o E
160o E
160o W
-0.2
-0.1
0
120o W
0.1
80o W
0.2
-1
-0.5
0
0.5
1
Figure
Figure8.
8.Seasonally
Seasonallyevolving
evolvingpatterns
patternsof
of(a)
(a)SST;
SST;(b)
(b)sea
sealevel;
level;and
and(c)
(c)SSW
SSWanomalies
anomaliesfrom
fromJJA(0)
JJA(0)to
to
MAM(1)
MAM(1)regressed
regressed with
with the
the PC
PC of
of the
the second
second SEOF
SEOF mode
mode of
of precipitation.
precipitation.
Figure
seasonally
evolving
pattern
of SLA
one year
priorprior
to theto
second
SEOF mode
Figure7b
7bshows
showsthe
the
seasonally
evolving
pattern
of SLA
one year
the second
SEOF
of
precipitation.
This
pattern
depicts
similar
SLA
structure
to
that
shown
in
Figure
6b
but
mode of precipitation. This pattern depicts similar SLA structure to that shown in Figure 6b butwith
with
anomalously
anomalouslyhigh
high SLAs
SLAs in
in the
the eastern
eastern equatorial
equatorial Pacific.
Pacific. Figure
Figure 7c
7c shows
shows the
the seasonally
seasonally evolving
evolving
pattern
patternof
of wind
wind anomalies
anomaliesfrom
fromsummer
summerto
tothe
thefollowing
followingspring
springone
oneyear
yearprior
priorto
to the
the second
second SEOF
SEOF
mode.
Similar
to
Figure
6c,
Figure
7c
exhibits
a
westerly
wind
anomaly
in
the
western
tropical
Pacific
mode. Similar to Figure 6c, Figure 7c exhibits a westerly wind anomaly in the western tropical Pacific
and
central
equatorial
Pacific.
In the
midlatitude
North
Pacific,
the
andanomalously
anomalouslyweak
weakwinds
windsininthe
the
central
equatorial
Pacific.
In the
midlatitude
North
Pacific,
westerly
winds
also
strengthen
from
the
boreal
autumn
to
winter.
The
main
difference
is
that
the
the westerly winds also strengthen from the boreal autumn to winter. The main difference is that
negative
windwind
speed
anomalies
associated
with
thethe
SEOF2
areare
focused
the negative
speed
anomalies
associated
with
SEOF2
focusednot
notonly
onlyon
onthe
the central
central
equatorial
Pacific,
but
extend
east
during
JJA(−1)
and
SON(−1).
Furthermore,
the
cyclone
equatorial Pacific, but extend east during JJA(−1) and SON(−1). Furthermore, the cyclone vanishes
vanishes
during
the decaying
decayingphase
phaseofof
El Niño
in MAM(0)
in Figure
but is maintained
during
the
during the
EPEP
El Niño
in MAM(0)
in Figure
7c, but7c,
is maintained
during the
decaying
decaying
phase
of CPinElMAM(0)
Niño in MAM(0)
6c.both
Overall,
both
Figure
7b and
Figure 7c confirm
phase of CP
El Niño
in Figure in
6c.Figure
Overall,
Figure
7b,c
confirm
the interpretation
that
the
interpretation
that
the
second
SEOF
mode
of
precipitation
is
associated
with
the
after EP El
the second SEOF mode of precipitation is associated with the years after EP El Niñoyears
events.
Niño Figure
events.8a shows the seasonally evolving SST anomalies simultaneously associated with the second
Figure
the seasonally
SST
associated
with
the
SEOF mode8a
of shows
precipitation,
i.e., fromevolving
JJA in Year
(0)anomalies
to MAM insimultaneously
Year(1). In general,
the SSTs
recover
second
SEOF
mode
of
precipitation,
i.e.,
from
JJA
in
Year
(0)
to
MAM
in
Year(1).
In
general,
the
SSTs
to the normal conditions. North of New Guinea in the western Pacific, precipitation decreases and the
recover
the normal
conditions. in
North
of after
NewEP
Guinea
in the
precipitation
decreases
easterlyto
trade
winds strengthens
JJA(0)
El Niño,
butwestern
the SSTPacific,
does not
decrease significantly
and
the
easterly
trade
winds
strengthens
in
JJA(0)
after
EP
El
Niño,
but
the
SST
does
not
decrease
during JJA(0). This finding indicates that precipitation and wind respond to ENSO more quickly
than
significantly
during
JJA(0).
This
finding
indicates
that
precipitation
and
wind
respond
to
ENSO
more
does SST and that decreasing precipitation is more closely related to strengthening wind than to SST
quickly
does SST
and
decreasing
is moreprecipitation
closely related
to strengthening
wind
cooling.than
However,
over
thethat
central
tropicalprecipitation
Pacific, the positive
anomalies
in zonal bands
than
to
SST
cooling.
However,
over
the
central
tropical
Pacific,
the
positive
precipitation
anomalies
in
along the equator are more strongly correlated with warmer SSTs. Overall, the seasonal evolution
zonal
bands along
more strongly
correlated
warmer
Overall,
the seasonal
of precipitation
inthe
theequator
tropics are
is positively
correlated
with with
SST after
EP SSTs.
El Niño.
Furthermore,
the
evolution
of
precipitation
in
the
tropics
is
positively
correlated
with
SST
after
EP
El
Niño.
Furthermore,
subtropical anticyclone that occupies the northwestern Pacific is associated with higher precipitation
the
subtropical
anticyclone
thatafter
occupies
theNiño
northwestern
Pacific is associated
with
in this
region during
the summer
an EP El
event. The Northeasterly
trade winds
arehigher
strong
precipitation
in
this
region
during
the
summer
after
an
EP
El
Niño
event.
The
Northeasterly
trade
winds
◦
◦
within 0–20 N and 140–170 E in the boreal summer but decay in autumn and reach their minimum
are strong within 0–20°N and 140–170°E in the boreal summer but decay in autumn and reach their
minimum strength in the winter. In the South Pacific around 10–30°S, the Southeasterly trade winds
Remote Sens. 2016, 8, 833
13 of 18
Remote Sens. 2016, 8, 833
13 of 18
◦ the Southeasterly trade winds are strong
strength
in in
theboth
winter.
the South
Pacific
are strong
the In
boreal
summer
and around
winter. 10–30
In the S,
eastern equatorial Pacific, wind speeds are
in
bothinthe
and winter.
the eastern
Pacific,
wind speeds
weak in the
weak
theboreal
borealsummer
spring and
start to In
recover
in theequatorial
boreal winter.
Altogether,
theseare
simultaneous,
boreal
spring
and start
to recover
in dynamic
the boreal
winter.represent
Altogether,
simultaneous,
seasonally
seasonally
evolving
patterns
of ocean
variables
the these
ocean-surface
conditions
after El
evolving
patterns
of
ocean
dynamic
variables
represent
the
ocean-surface
conditions
after
El
Niño
Niño events, and contribute to the evolution of precipitation patterns.
events, and contribute to the evolution of precipitation patterns.
4. Discussion
4. Discussion
4.1. Which Type of El Niño Dominates the Seasonal Evolution of Precipitation?
4.1. Which Type of El Niño Dominates the Seasonal Evolution of Precipitation?
Our results have demonstrated that the first two SEOF modes of precipitation were associated
Our results have demonstrated that the first two SEOF modes of precipitation were associated
with El Niño development and post-El Niño years, respectively. However, types of El Niño could not
with El Niño development and post-El Niño years, respectively. However, types of El Niño could not
be separated confidently based only on the regressed spatial patterns of SST, SLA and SSW shown in
be separated confidently based only on the regressed spatial patterns of SST, SLA and SSW shown
Figures 6 and 7. To identify which type of ENSO event most affects the dominant SEOF modes of
in Figures 6 and 7. To identify which type of ENSO event most affects the dominant SEOF modes of
precipitation, the lagged correlations between the first two SEOF modes with the El Niño Modoki
precipitation, the lagged correlations between the first two SEOF modes with the El Niño Modoki
index (EMI) and Niño3 index for each season were calculated, and the results are shown in Figure 9.
index (EMI) and Niño3 index for each season were calculated, and the results are shown in Figure 9.
(a) corr (SEOF1, EMI & Nino3)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
JJA
Y(-1)
SON
DJF
Y(0)
MAM
JJA
SON
DJF
Y(1)
MAM
JJA
SON
JJA
SON
DJF
MAM
(b) corr (SEOF2, EMI & Nino3)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
JJA
Y(-1)
SON
DJF
Y(0)
MAM
JJA
SON
DJF
Y(1)
MAM
DJF
MAM
Figure 9. (a) Lagged correlations of the first SEOF PC with the EMI (red line) and Niño3 index
Figure 9. (a) Lagged correlations of the first SEOF PC with the EMI (red line) and Niño3 index (blue
(blue line); (b) lagged correlations between the second SEOF PC and the EMI (red line) and Niño3
line); (b) lagged correlations between the second SEOF PC and the EMI (red line) and Niño3 index
index (blue line). Year(−1) (Year(1)) denotes that the PC lags behind (leads) its counterpart by one year;
(blue line). Year(−1) (Year(1)) denotes that the PC lags behind (leads) its counterpart by one year;
Year(0) denotes simultaneous correlation. JJA is boreal summer, SON is boreal autumn, DJF is boreal
Year(0) denotes simultaneous correlation. JJA is boreal summer, SON is boreal autumn, DJF is boreal
winter, and MAM is boreal spring.
winter, and MAM is boreal spring.
Figure
shows that
that the
thefirst
firstSEOF
SEOFmode
modehas
hassignificant
significant
correlation
with
both
Figure 9a shows
correlation
with
both
thethe
EMIEMI
andand
the
the
Niño3
index
for
Year(0).
The
simultaneous
correlation
coefficients
between
the
SEOF1
and
EMI
Niño3 index for Year(0). The simultaneous correlation coefficients between the SEOF1 and EMI were
were
in JJA,
SON,
DJF
and
MAM,respectively,
respectively,which
whichare
are higher
higher than those
0.77, 0.77,
0.90, 0.90,
0.97 0.97
and and
0.910.91
in JJA,
SON,
DJF
and
MAM,
those
between
the
SEOF1
and
Niño3
index
(Table
2).
Therefore,
the
first
SEOF
mode
of
precipitation
between the SEOF1 and Niño3 index (Table 2).
SEOF mode of precipitation is
is
associated
with the
thecombination
combinationofofEP
EPElEl
Niño
and
El Niño
events,
likely
coincided
associated with
Niño
and
CPCP
El Niño
events,
but but
moremore
likely
coincided
with
with
theEl
CPNiño.
El Niño.
inference
is further
supported
theSST
SSTpattern
patternshown
shownin
in Figure
Figure 6a with
the CP
ThisThis
inference
is further
supported
byby
the
with
anomalously
warm
SSTs
mainly
in
the
central
equatorial
Pacific,
and
by
the
similarity
between
anomalously
SSTs mainly in the central equatorial Pacific, and by the similarity between the
the
spatial
spatial pattern
pattern of
of precipitation
precipitation in
in the
the boreal
boreal winter
winter for
for SEOF1
SEOF1 shown
shown in
in Figure
Figure 4a
4a and
and the
the correlation
correlation
map
map of
of precipitation
precipitation with
with the
the EMI
EMI shown
shown in
inFigure
Figure11a
11aof
ofAshok
Ashoket
etal.
al.[26].
[26].
Both
Figures
2a
and
4a
reveal
the
striking
ITCZ
and
SPCZ
bands
Both Figures 2a and 4a reveal the striking ITCZ and SPCZ bands with
with positive
positive precipitation
precipitation
anomalies
anomalies that extend eastward and southeastward from the Solomon Islands, as well as the dry
conditions
Although different
different from Figure 2a, Figure 4a shows
conditions over
over the
the Maritime
Maritime Continent.
Continent. Although
shows the
the
seasonal evolution of precipitation associated with the development of El Niño. The responses of
precipitation patterns to the development of El Niño from the boreal summer to winter include the
Remote Sens. 2016, 8, 833
14 of 18
seasonal evolution of precipitation associated with the development of El Niño. The responses of
precipitation patterns to the development of El Niño from the boreal summer to winter include the
following: the ITCZ becomes progressively wetter and moves southward toward the equator, the SPCZ
also receives more precipitation and shifts northeastward, and the location of maximum precipitation
along the equator moves east. The opposite patterns of variation in precipitation occur during La
Niña, as described by Chung and Power [21]. Overall, precipitation in the western basin of the warm
pool experiences dry conditions with the development of either EP or CP El Niño events, and both the
SPCZ and IPCZ move toward the equator during El Niño years. During La Niña events, these trends
are reversed.
For the SEOF2 of precipitation, the lagged correlation coefficients between the SEOF2 and Niño3
index were significant in Year(−1), whereas the correlation between the SEOF2 and EMI did not reach
the 5% significance level. These findings strongly indicate that the second SEOF mode of precipitation
was associated with the decaying year of EP-type ENSO events. Therefore, we may conclude that
Figure 4a,b demonstrate the consecutive seasonal evolution pattern of precipitation anomalies over
the entire development of EP El Niño events, as well as the subsequent recovery and return of normal
conditions after EP El Niño.
Table 2. Significant lagged correlation coefficients of the PCs of the SEOF modes and the El Niño
Modoki Index (EMI) and Niño3 index.
Correlations
Season
EMI
Niño3 Index
corr(SEOF1, EMI, & Niño3)
SEOF1-JJA
SEOF1-SON
SEOF1-DJF
SEOF1-MAM
0.77
0.90
0.97
0.91
0.69
0.86
0.88
0.52
corr(SEOF2, EMI, & Niño3)
SEOF2-JJA
SEOF2-SON
SEOF2-DJF
SEOF2-MAM
*
*
*
*
0.57
0.59
0.57
0.57
* Correlation that did not reach the 5% significance level.
4.2. Impact of PDO on the ENSO-Induced Precipitation Seasonal Evolution
As an interdecadal climate oscillation, the impact of PDO on the variation of precipitation is
accompanied by ENSO. As was indicated by the lagged correlation shown in Figure 5, significant
correlation between the PDO and precipitation SEOFs exists only when the MEI correlates significantly
with the SEOFs. Moreover, these indices were in the same phase in influencing precipitation. Therefore,
PDO may impact precipitation over the Pacific primarily by strengthening and prolonging the effects
of ENSO. In addition, Figure 5a shows that PDO still correlated significantly with the precipitation
SEOF1 mode in the El Niño decaying summer, which is consistent with the SST anomaly pattern of the
PDO warm phase in JJA(0) shown in Figure 8a. This correlation indicates that the precipitation pattern
in JJA(0) of Figure 4b is also influenced by PDO, i.e., precipitation is still dominated by SST anomalies
in the North Pacific during the decaying summer of EP El Niño.
5. Conclusions
Compared with previous studies, we provide a more integrated perspective on the seasonal
and interannual variability of precipitation over the Pacific Ocean as well as the seasonal evolution
of precipitation associated with the ENSO during the period of 1998–2014. In addition, lagged
correlation analysis is applied to derive the lead–lag correlations between the first two SEOF modes
for precipitation and climate-related air–sea phenomena, i.e., ENSO and PDO, and further to reveal
the large-scale oceanic patterns associated with the distributions of high and low precipitation over
Remote Sens. 2016, 8, 833
15 of 18
the Pacific Ocean. These findings may be useful for seasonal precipitation anomaly predictions for the
Pacific Ocean during El Niño years and the following recovery years.
Using EOF analysis, we have examined the spatiotemporal variability of precipitation, at
seasonal and interannual timescales separately, to provide an overview of precipitation variability.
The combination of the first two seasonal EOF modes represents the consecutive annual cycle of
precipitation variations of the Pacific Ocean. Specifically, a double ITCZ exists in boreal spring,
with one north and the other south of the equator; during boreal summer, the northern zone moves
northward to 10◦ N, and the southern one vanishes. The SPCZ is drier than average during the austral
autumn and winter, but wetter during the austral spring and summer. In terms of the interannual
variability of precipitation, the first EOF mode for 13-month running mean precipitation datasets
represents the spatiotemporal variation associated with ENSO. During El Niño, global conditions
are characterized by significantly positive wedge-shaped anomalies in the central Pacific along the
equator toward the ITCZ and the northern part of the SPCZ, as well as negative anomalies over the
South China Sea, the Maritime Continent, and the southern part of the SPCZ. The reverse patterns
exist during La Niña.
Using SEOF and lagged correlation analyses, we have revealed two dominant seasonally
evolving modes of precipitation anomalies and their relationships with ENSO and PDO. The principal
component of the first SEOF mode reflects El Niño and La Niña events, with the positive (negative)
values corresponding to El Niño (La Niña) years. The lagged correlation between the SEOF PCs and
MEI indicates that the first SEOF mode is simultaneously associated with ENSO developing episodes,
while SEOF2 lagged behind ENSO by one year and is associated with post-ENSO years. In addition,
PDO modulates the precipitation variability significantly only when ENSO occurs. More specifically,
PDO could impact the precipitation over the Pacific primarily by strengthening and prolonging the
impact of ENSO on the precipitation during the ENSO decaying summer. By using lagged correlation
analysis on the first two SEOF modes with the El Niño Modoki index and Nino3 index, respectively, we
further distinguish different associations of two types of El Niño (i.e. EP El Niño and CP El Niño) with
the two dominant SEOF modes. It is revealed that the first SEOF mode of precipitation is associated
with both EP and CP El Niño years, but more likely coincides with the CP type, while the second SEOF
mode is associated with the years after EP El Niño occurs.
In fact, the spatial seasonally evolving pattern of the first two SEOF modes (Figure 4) describes the
consecutive seasonal evolution pattern of precipitation anomaly associated with the entire developing
episodes of EP El Niño as well as the recovering to normal episodes afterwards. The significant variation
pattern mainly exists on the tropical Pacific, especially ITCZ and SPCZ regions. Regressed anomalies
of ocean dynamic variables (SST, SLA, and WIND) associated with the first two SEOF modes depict
the synchronously spatial variation of ocean and climate dynamics. During the developing episodes
of either EP or CP El Niño, the western basin of warm pool experiences dry conditions while both
SPCZ and IPCZ become progressively wetter and move toward the equator. This precipitation pattern
during the mature phase of El Niño is associated with warm SSTs in the central tropical Pacific and
a subtropical anticyclone dominating conditions over the northwestern Pacific.
In this study, the asymmetric responses of precipitation to El Niño and La Niña SST anomalies
are not distinguished. Therefore, the direction for future work is to investigate the spatiotemporal
variations in precipitation during the warm and cold phase of ENSO as well as the underlying oceanic
and atmospheric mechanisms.
Acknowledgments: This work was supported by the National Natural Science Foundation of China under Grants
No. 91437214 and No. 71461010701.
Author Contributions: Xueyan Hou carried out the research and prepared the manuscript under the guidance of
Di Long and Yang Hong. Hongjie Xie contributed to the preparation of the paper through review, editing, and
comments. All authors made substantial intellectual contributions to the paper and the findings reported herein.
Conflicts of Interest: The authors declare no conflict of interest.
Remote Sens. 2016, 8, 833
16 of 18
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Curtis, S.; Adler, R. ENSO Indices Based on Patterns of Satellite-Derived Precipitation. J. Clim. 2000, 13,
2786–2793. [CrossRef]
Kodama, Y.; Tamaoki, A. A Re-examination of precipitation activity in the subtropics and the mid-latitudes
based on satellite-derived data. J. Meteorol. Soc. Jpn. Ser. II 2002, 80, 1261–1278. [CrossRef]
Huang, P.; Xie, S. Mechanisms of change in enso-induced tropical pacific rainfall variability in a warming climate.
Nat. Geosci. 2015, 8, 922–926. [CrossRef]
Curtis, S. The El Niño–southern oscillation and global precipitation. Geogr. Compass 2008, 2, 600–619. [CrossRef]
Trenberth, K.E.; Hoar, T.J. El Niño and climate change. Geophys. Res. Lett. 1997, 24, 3057–3060. [CrossRef]
Ropelewski, C.F.; Halpert, M.S. Precipitation patterns associated with the high index phase of the
southern oscillation. J. Clim. 1989, 2, 268–284. [CrossRef]
Ropelewski, C.F.; Halpert, M.S. Global and regional scale precipitation patterns associated with the
El Niño/southern oscillation. Mon. Weather Rev. 1987, 115, 1606–1626. [CrossRef]
Adler, R.F.; Huffman, G.J.; Chang, A.; Ferraro, R.; Xie, P.; Janowiak, J.; Rudolf, B.; Schneider, U.; Curtis, S.;
Bolvin, D.; et al. The Version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation
analysis (1979–present). J. Hydrometeorol. 2003, 4, 1147–1167. [CrossRef]
Xie, P.; Arkin, P.A. Global Precipitation: A 17-Year monthly analysis based on gauge observations, satellite
estimates, and numerical model outputs. Bull. Am. Meteorol. Soc. 1997, 78, 2539–2558. [CrossRef]
Huffman, G.J.; Bolvin, D.T.; Nelkin, E.J.; Wolff, D.B.; Adler, R.F.; Gu, G.; Hong, Y.; Bowman, K.P.; Stocker, E.F.
The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor
precipitation estimates at fine scales. J. Hydrometeorol. 2007, 8, 38–55. [CrossRef]
Dai, A.; Wigley, T.M.L. Global patterns of ENSO-induced precipitation. Geophys. Res. Lett. 2000, 27,
1283–1286. [CrossRef]
Chou, C.; Tu, J.; Yu, J. Interannual variability of the western north pacific summer monsoon: Differences
between ENSO and Non-ENSO Years. J. Clim. 2003, 16, 2275–2287. [CrossRef]
Curtis, S.; Adler, R.F. Evolution of El Niño-precipitation relationships from satellites and gauges. J. Geophys.
Res. Atmos. 2003, 108. [CrossRef]
Haddad, Z.S.; Meagher, J.P.; Adler, R.F.; Smith, E.A.; Im, E.; Durden, S.L. Global variability of precipitation
according to the Tropical Rainfall Measuring Mission. J. Geophys. Res. Atmos. 2004, 109, D17. [CrossRef]
Chou, C.; Huang, L.; Tu, J.; Tseng, L.; Hsueh, Y. El Niño impacts on precipitation in the Western North
Pacific–East Asian Sector. J. Clim. 2009, 22, 2039–2057. [CrossRef]
Wu, Z.; Wang, B.; Li, J.; Jin, F. An empirical seasonal prediction model of the east asian summer monsoon
using ENSO and NAO. J. Geophys. Res. Atmos. 2009, 114. [CrossRef]
Curtis, S.; Salahuddin, A.; Adler, R.F.; Huffman, G.J.; Gu, G.; Hong, Y. precipitation extremes estimated by
GPCP and TRMM: ENSO relationships. J. Hydrometeorol. 2007, 8, 678–689. [CrossRef]
Dayem, K.E.; Noone, D.C.; Molnar, P. Tropical Western Pacific warm pool and maritime continent
precipitation rates and their contrasting relationships with the walker circulation. J. Geophys. Res. Atmos.
2007, 112, D06101. [CrossRef]
Yang, S.; Smith, E.A. Convective–stratiform precipitation variability at seasonal scale from 8 years of TRMM
observations: Implications for multiple modes of diurnal variability. J. Clim. 2008, 21, 4087–4114. [CrossRef]
Dorman, C.E.; Bourke, R.H. Precipitation over the Pacific Ocean, 30◦ S to 60◦ N. Mon. Weather Rev. 1979, 107,
896–910. [CrossRef]
Chung, C.Y.; Power, S. Precipitation response to La Niña and global warming in the Indo-Pacific. Clim. Dyn.
2014, 43, 3293–3307. [CrossRef]
Trenberth, K.E.; Caron, J.M. the southern oscillation revisited: Sea level pressures, surface temperatures,
and precipitation. J. Clim. 2000, 13, 4358–4365. [CrossRef]
Adler, R.F.; Gu, G.; Wang, J.; Huffman, G.J.; Curtis, S.; Bolvin, D. Relationships between global precipitation
and surface temperature on interannual and longer timescales (1979–2006). J. Geophys. Res. Atmos. 2008, 113,
D22. [CrossRef]
Liu, C.; Allan, R.P. Multisatellite observed responses of precipitation and its extremes to interannual
climate variability. J. Geophys. Res. Atmos. 2012, 117. [CrossRef]
Remote Sens. 2016, 8, 833
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
17 of 18
Back, L.E.; Bretherton, C.S. The relationship between wind speed and precipitation in the Pacific ITCZ.
J. Clim. 2005, 18, 4317–4328. [CrossRef]
Ashok, K.; Behera, S.K.; Rao, S.A.; Weng, H.; Yamagata, T. El Niño modoki and its possible teleconnection.
J. Geophys. Res. Oceans 2007, 112. [CrossRef]
Yeh, S.; Kug, J.; Dewitte, B.; Kwon, M.; Kirtman, B.P.; Jin, F. El Niño in a changing climate. Nature 2009, 461,
511–514. [CrossRef] [PubMed]
Kug, J.; Jin, F.; An, S. Two types of El Niño events: Cold tongue El Niño and warm pool El Niño. J. Clim.
2009, 22, 1499–1515. [CrossRef]
Gu, G.; Adler, R.F.; Huffman, G.J.; Curtis, S. Tropical rainfall variability on interannual-to-interdecadal and
longer time scales derived from the gpcp monthly product. J. Clim. 2007, 20, 4033–4046. [CrossRef]
Huang, P.; Xie, S.; Hu, K.; Huang, G.; Huang, R. Patterns of the seasonal response of tropical rainfall to
global warming. Nat. Geosci. 2013, 6, 357–361. [CrossRef]
Lau, K.; Wu, H.T. Principal modes of rainfall–sst variability of the Asian summer monsoon: A reassessment
of the monsoon–enso relationship. J. Clim. 2001, 14, 2880–2895. [CrossRef]
Reynolds, R.W.; Rayner, N.A.; Smith, T.M.; Stokes, D.C.; Wang, W. An improved in situ and satellite SST
analysis for climate. J. Clim. 2002, 15, 1609–1625. [CrossRef]
AVISO. Available online: http://www.aviso.altimetry.fr/en/data/products/sea-surface-height-products/
global/msla-mean-climatology.html#c10358 (accessed on 1 October 2016).
Atlas, R.; Hoffman, R.N.; Ardizzone, J.; Leidner, S.M.; Jusem, J.C.; Smith, D.K.; Gombos, D. A Cross-calibrated,
multiplatform ocean surface wind velocity product for meteorological and oceanographic applications.
Bull. Am. Meteorol. Soc. 2011, 92, 157–174. [CrossRef]
Multivariate ENSO Index (MEI). Available online: http://www.esrl.noaa.gov/psd/enso/mei/ (accessed on
1 October 2016).
The Pacific Decadal Oscillation (PDO). Available online: http://research.jisao.washington.edu/pdo
(accessed on 1 October 2016).
MODOKI ENSO: A New Phenomenon is Found in the Tropical Pacific. Available online: http://www.
jamstec.go.jp/frsgc/research/d1/iod/modoki_home.html.en (accessed on 1 October 2016).
Climate Indices: Monthly Atmospheric and Ocean Time Series. Available online: http://www.esrl.noaa.
gov/psd/data/climateindices/list/ (accessed on 1 October 2016).
Venegas, S.A.; Mysak, L.A.; Straub, D.N. Atmosphere–ocean coupled variability in the South Atlantic. J. Clim.
1997, 10, 2904–2920. [CrossRef]
Wang, B.; An, S. A method for detecting season-dependent modes of climate variability: S-EOF analysis.
Geophys. Res. Lett. 2005, 32, L15710. [CrossRef]
Messié, M.; Chavez, F.P. Physical-biological synchrony in the global ocean associated with recent variability
in the central and Western Equatorial Pacific. J. Geophys. Res. Oceans 2013, 118, 3782–3794. [CrossRef]
Spencer, H. Role of the atmosphere in seasonal phase locking of El Niño. Geophys. Res. Lett. 2004, 31.
[CrossRef]
Yasunari, T. The monsoon year—A new concept of the climatic year in the tropics. Bull. Am. Meteorol. Soc.
1991, 72, 1331–1338. [CrossRef]
North, G.R.; Bell, T.L.; Cahalan, R.F.; Moeng, F.J. Sampling errors in the estimation of empirical orthogonal
functions. Mon. Weather Rev. 1982, 110, 699–706. [CrossRef]
Messié, M.; Chavez, F.P. A global analysis of ENSO synchrony: The Oceans’ biological response to
physical forcing. J. Geophys. Res. 2012, 117. [CrossRef]
Lee, T.; McPhaden, M.J. Increasing intensity of El Niño in the Central-Equatorial Pacific. Geophys. Res. Lett.
2010, 37. [CrossRef]
Singh, A.; Delcroix, T.; Cravatte, S. Contrasting the flavors of El Niño-Southern oscillation using sea surface
salinity observations. J. Geophys. Res. Oceans 2011, 116. [CrossRef]
Neelin, J.D.; Jin, F.; Syu, H. Variations in ENSO phase locking. J. Clim. 2000, 13, 2570–2590. [CrossRef]
Mantua, N.J.; Hare, S.R.; Zhang, Y.; Wallace, J.M.; Francis, R.C. A pacific interdecadal climate oscillation with
impacts on salmon production. Bull. Am. Meteorol. Soc. 1997, 78, 1069–1079. [CrossRef]
Minobe, S. Resonance in bidecadal and pentadecadal climate oscillations over the North Pacific: Role in
climatic regime shifts. Geophys. Res. Lett. 1999, 26, 855–858. [CrossRef]
Remote Sens. 2016, 8, 833
51.
52.
53.
54.
55.
18 of 18
Delcroix, T. Observed surface oceanic and atmospheric variability in the tropical Pacific at seasonal and
ENSO timescales: A tentative overview. J. Geophys. Res. 1998, 103, 18611–18633. [CrossRef]
Hou, X.; Dong, Q.; Xue, C.; Wu, S. Seasonal and interannual variability of chlorophyll-a and associated
physical synchronous variability in the western tropical Pacific. J. Mar. Syst. 2016, 158, 59–71. [CrossRef]
Chang, Y.; Du, L.; Zhang, S.; Huang, P. Sea level variations in the tropical Pacific ocean during two types of
recent El Niño events. Glob. Planet. Chang. 2013, 108, 119–127. [CrossRef]
Pan, J.Y.; Yan, X.H.; Zheng, Q.; Liu, W.T. Vector empirical orthogonal function modes of the ocean surface
wind variability derived from satellite scatterometer data. Geophys. Res. Lett. 2001, 28, 3951–3954. [CrossRef]
Pan, J.Y.; Yan, X.H.; Zheng, Q.N.; Liu, W.T.; Klemas, V.V. Interpretation of scatterometer ocean surface wind
vector EOFs over the Northwestern Pacific. Remote Sens. Environ. 2002, 84, 53–68. [CrossRef]
© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).