Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320463142

Interannual Variations and Prediction of Spring Precipitation over China

Article in Journal of Climate · October 2017


DOI: 10.1175/JCLI-D-17-0233.1

CITATIONS READS

33 355

2 authors:

Yujia You Xiaojing Jia


Columbia University Zhejiang University
8 PUBLICATIONS 161 CITATIONS 46 PUBLICATIONS 558 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Xiaojing Jia on 28 January 2019.

The user has requested enhancement of the downloaded file.


15 JANUARY 2018 YOU AND JIA 655

Interannual Variations and Prediction of Spring Precipitation over China

YUJIA YOU AND XIAOJING JIA


School of Earth Sciences, Zhejiang University, Hangzhou, Zhejiang, China

(Manuscript received 11 April 2017, in final form 11 August 2017)

ABSTRACT

The interannual variations and the prediction of the leading two empirical orthogonal function (EOF)
modes of spring (April–May) precipitation over China for the period from 1951 to 2014 are investigated using
both observational data and the seasonal forecast made by six coupled climate models. The leading EOF
mode of spring precipitation over China (EOF1-prec) features a monosign pattern, with the maximum
loading located over southern China. The ENSO-related tropical Pacific SST anomalies in the previous winter
can serve as a precursor for EOF1-prec. The second EOF mode of spring precipitation (EOF2-prec) over
China is characterized by a dipole structure, with one pole near the Yangtze River and the other one with
opposite sign over the Pearl River delta. A North Atlantic sea surface temperature (SST) anomaly dipole in
the preceding March is found contribute to the prec-EOF2 and can serve as its predictor. A physics-based
empirical (P-E) model is then formulated using the two precursors revealed by the observational analysis to
forecast the variations of EOF1-prec and EOF2-prec. Compared to coupled climate models, which have little
skill in forecasting the time variations of the two EOF modes, this P-E model can significantly improve
the forecast skill of their time variations. A linear regression model is further established using the time series
forecast by the P-E model to forecast the spring precipitation over China. Results suggest that the seasonal
forecast skill of the spring precipitation over southeastern China, especially over the Yangtze River area, can
be significantly improved by the regression model.

1. Introduction standard deviation of AM precipitation (Fig. 1b) is sim-


ilar to that of the climatology, indicating strong year-to-
Precipitation over China exhibits prominent variability
year variability of the AM precipitation over southern
on both interannual and interdecadal time scales. Be-
China. The ratio of the climatological AM precipitation
cause of its considerable influence on agricultural and
to summer [June–August (JJA)] precipitation over China
economic productivity, extensive research has been de-
is depicted in Fig. 1c. It shows that, over southern China,
voted to exploring its mechanisms and predictability (e.g.,
the amount of the AM precipitation is the equivalent of
Lau and Li 1984; Huang et al. 2003; Wang et al. 2003; Gao
approximately 50%–70% of the summer precipitation.
et al. 2008; Li et al. 2015). Most previous studies focused
Because the early planting season in southern China is
on investigating summer precipitation anomalies (e.g.,
heavily dependent on the amount of spring precipitation,
Chang et al. 2000; Xie et al. 2009), whereas relatively less
improving the accuracy of the seasonal forecast for spring
attention has been paid to spring precipitation. Spring
precipitation will greatly benefit society and the economy
precipitation over southern China actually accounts for
(Qiu et al. 2009; Wu et al. 2014; Chen et al. 2014; Feng
a considerable fraction of the total annual rainfall.
et al. 2014; Wu and Mao 2016). Moreover, since spring is a
Figure 1a presents the climatological mean state of pre-
transitional season that links the preceding winter to the
cipitation for the average of April and May (AM) over
succeeding summer, a better understanding of the causes
China. It shows that the AM precipitation has the largest
of the variations in spring precipitation will provide a
loading over southern China and achieves peak values at
fundamental basis for cross-season climate prediction by
two centers, with one located near the Pearl River delta
numerical models.
and the other centered over coastal southeastern China,
The mechanisms that account for the variation of
at approximately 288N. The spatial distribution of the
precipitation over China have been examined in many
previous studies (e.g., Yang and Lau. 2004; Xin et al.
Corresponding author: XiaoJing Jia, jiaxiaojing@zju.edu.cn 2006; Qiu et al. 2009; Feng and Li. 2011; Wu et al. 2014;

DOI: 10.1175/JCLI-D-17-0233.1
Ó 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright
Policy (www.ametsoc.org/PUBSReuseLicenses).
656 JOURNAL OF CLIMATE VOLUME 31

Yang 1992; Zhang and Sumi 2002; Wang et al. 2003; Xie
et al. 2009). Wang et al. (2000) showed that the anom-
alous sea surface temperature (SST) associated with El
Niño in the central-eastern equatorial Pacific induces
anomalous convergence in the upper-level eastern
tropical Pacific and descent over the Philippine Sea,
favoring the development of an anomalous lower-level
Philippine Sea anticyclone (PSAC). The anomalous
PSAC can act as a medium bridging remote El Niño
forcing and variations in the East Asian climate as it
enhances southerly winds to its northwest flank and
generates above-average precipitation and temperatures
over China. The PSAC can persist through local air–sea
interactions over the western North Pacific, causing
rainfall anomalies over southern China in the ensuing
spring and summer. In the work of Feng and Li (2011),
it was further shown that canonical El Niño events and
central Pacific warming have different effects on spring
precipitation over China.
The relationship between spring precipitation over
China and the Indian Ocean has also been explored. The
remote effect of the tropical Indian Ocean in main-
taining the PSAC via capacitor effects has been ex-
plored by Xie et al. (2009). Chen et al. (2014) proposed
an interdecadal change in the relationship between
south Indian Ocean SST anomalies (SSTAs) and in-
terannual variations in winter–spring persistent pre-
cipitation over southern China. The effects of Indian
Ocean SST played a more central role in modulating the
variability in winter–spring persistent precipitation for
1974–94 than in modulating that for 1953–73, which was
most likely caused by enhanced SST variation in the
south Indian Ocean. The south Indian Ocean subtropical
dipole (IOSD) was also found to influence spring pre-
cipitation over regions near the Yangtze and Yellow
Rivers through the modulation of meridional circulation
(e.g., Feng et al. 2014). Some studies have investigated
the impact of the Eurasian–Tibetan Plateau snow cover
FIG. 1. (a) Climatology and (b) standard deviation of AM pre- on the rainfall over East Asian. For example, Wu and
cipitation over China for 1951–2014. (c) The ratio of AM pre- Kirtman (2007) suggested that anomalous excessive
cipitation to summer (JJA) precipitation over China represented spring precipitation in southern China can be associated
by percentage. The contour interval is 20 mm month21 for (a), with spring snow cover in western Siberia. Zuo et al.
10 mm month21 for (b), and 10% for (c).
(2012) examined the relationship between the snow water
equivalent (SWE) and the spring rainfall and found that
Chen et al. 2014; Feng et al. 2014). Previous studies in- decreased spring SWE in Eurasia is related to negative
dicate that the precipitation over China can be affected spring rainfall anomalies over southeastern and north-
by many anomalous states of lower boundary conditions eastern China, and positive rainfall anomalies over
(e.g., Tao and Chen 1987; Wang et al. 2000, Huang et al. southwestern and northwestern China could be found.
2004; Wu and Kirtman. 2007; Wu et al. 2009; Zuo et al. Some other studies investigated the characteristics of
2012; Wu et al. 2014; Feng et al. 2014). Among them, El the long-term trends of spring precipitation over China.
Niño–Southern Oscillation (ENSO) is one of the most For example, pronounced positive and negative trends
important factors that impacts the precipitation over of spring precipitation were observed by Yang and Lau
China on the interannual time scale (e.g., Webster and (2004) in southeastern China and central-eastern China,
15 JANUARY 2018 YOU AND JIA 657

respectively. The forcing mechanism accounting for (NCEP)–National Center for Atmospheric Research
these trends in association with the global SST has been (NCAR) Global Reanalysis 1 (Kalnay et al. 1996) (these
examined. Qiu et al. (2009) demonstrated that the rainfall data are available at a horizontal resolution of 2.58 3 2.58);
in May over southeastern China has significantly de- 2) the observed monthly mean precipitation of 160 Chi-
creased in recent decades along with a phase transition to nese meteorological stations, obtained from the National
La Niña over the Indo-Pacific Ocean. In addition to Climate Center of the China Meteorological Adminis-
SSTAs, Wu et al. (2014) reported that mid-to-high-latitude tration; 3) the NOAA Extended Reconstructed SST,
atmospheric circulation anomalies induced by anomalous version 3b (ERSST.v3b; Smith et al. 2008), which has a
states of lower boundary conditions, such as Eurasian resolution of 2.08 3 2.08; and 4) the Global Precipitation
snow cover and depth, could influence Eurasian climate Climatology Project (GPCP) monthly mean precipitation
anomalies in spring and summer. data with a horizontal resolution of 2.58 3 2.58 (Adler
Although efforts have been devoted to investigating et al. 2003). All the datasets used in the study cover the
the mechanisms of the variability of spring precipitation period 1951–2014 except for the GPCP, which covers the
over China, questions regarding which of the proposed period 1979–2014.
factors is most essential to understanding its variability An EOF analysis was applied to the spring precipi-
still remain. Furthermore, the question of which factors tation over China to subtract the leading EOF modes.
can be used in the seasonal forecasting of the spring To determine the climate anomalies associated with
precipitation over China has not been well investi- these leading EOF modes, Pearson correlation and re-
gated. In the current work, the variations of the spring gression analyses were carried out between a physical
precipitation over China are investigated through an field and the time series associated with these EOF
empirical orthogonal function (EOF) analysis. We modes. The ENSO variability is represented by averaging
concentrate on examining the leading two EOF modes the SST over the Niño-3.4 region (58S–58N, 1708–1208W).
of spring precipitation. Their associated climate anom- To explore the mechanism behind the connection
alies are investigated with a focus on identifying their between large-scale circulation anomalies and the spring
respective related SST precursors. In addition, empiri- precipitation over China, a phase-independent wave
cal models are established according to the results of activity flux proposed by Takaya and Nakamura (2001)
the observational analysis. We demonstrate that the was used. The flux is based on the conservation of wave
seasonal forecast skill of the spring precipitation in activity pseudomomentum and is a useful diagnostic tool
southern China can be significantly improved by these for identifying sources or sinks of wave activity. In the
empirical models. current work, the horizontal component of this flux is
The organization of the rest of the text is as follows. calculated which is can be given by
The datasets, methodology, and model used in this study
1
are introduced in section 2. Section 3 presents the W5 [U(c02 0 0 0 0 0 0
x 2 c cxx ) 1 V(cx cy 2 c cxy )
characteristics of the climatology of spring precipitation 2jUj
over China. Section 4 examines the features of the first 3 U(c0x c0y 2 c0 c0xy ) 1 V(c02 0 0
y 2 c cyy )] ,
two EOF modes of spring precipitation, as well as their
associated atmospheric circulation and SST anomalies. where c is the streamfunction and subscripts are partial
In section 5, the forecast skill of the multimodel en- derivatives in x and y directions; also, U 5 (U, V) de-
semble (MME) of six coupled climate models in pre- notes the 500-hPa two-dimensional geostrophic zonal
dicting the variation of spring precipitation is examined. and meridional velocity components for spring mean
Empirical models are then established trying to improve flow and primes denote deviations from the climatology.
the seasonal forecast skill of spring precipitation. A
b. Models
comprehensive summary of the results and a discussion
are presented in section 6. The MME of the seasonal forecast of six coupled cli-
mate models was used to examine the performance of the
numerical models in predicting the variation of spring
2. Data, methodology, and models precipitation over China. The climate numerical models
used include the CAWCR from the Asia–Pacific Eco-
a. Observational data and methodology
nomic Cooperation (APEC) Climate Center/Climate
The datasets used in this study include the following: Prediction and Its Application to Society (APCC/
1) monthly mean reanalyses of sea level pressure (SLP), CliPAS) project and the CMCC–INGV, ECMWF,
geopotential heights, specific humidity, and wind from Leibniz Institute of Marine Sciences at Kiel University
the National Centers for Environmental Prediction (IFM-GEOMAR), Météo-France (MF), and Met Office
658 JOURNAL OF CLIMATE VOLUME 31

FIG. 2. The climatological mean of (a) March and (b) AM precipitation (shading, mm day21) and the vertically
integrated water vapor transport [vector, kg (m s)21]. (c) The difference between (a) and (b) obtained by sub-
tracting (a) from (b).

(UKMO) models from the Ensemble-Based Predictions characteristics and variations of the precipitation over
of Climate Changes and Their Impacts (EMSEMBLES) southern China in March are more closely related to those
project (Weisheimer et al. 2009; Wang et al. 2009a,b; Jia in the preceding January–February (JF). Therefore, AM
et al. 2014a,b). None of the coupled models has flux ad- rather than March–May (MAM) is used to represent the
justments. The seasonal forecasts from March to May, spring season in the current study.
initialized from the first day of February for the period The climatological mean precipitation (shading) and
from 1960 to 2005, are used in the current study. For the the vertically integrated moisture fluxes (vector) for
1-month lead seasonal forecasts, the first month of the March and AM are presented in Figs. 2a and 2b, re-
integration was not used, as we are mainly interested in spectively. The differences between Figs. 2a and 2b are
the forecast signal coming from air–sea coupling rather depicted in Fig. 2c, which was obtained by subtracting
than the atmospheric initial conditions. The MME pre- Fig. 2a from Fig. 2b. As shown in Fig. 2a, the climato-
dictions were made based on the equally weighted aver- logical mean precipitation in March is primarily con-
age of the six coupled climate model ensemble mean centrated over southeastern China. The moisture
anomalies after removing their own climatology. Further transportation to China in March mainly includes two
information about the climate models can be found in branches. One branch is with a southwesterly along the
Jia et al. (2014a,b). northern flank of the western Pacific subtropical high
(WPSH), and the other branch is along a subtropical
westerly at 308N. From March to AM, the magnitude of
3. The climatology of spring precipitation in China
the time-averaged precipitation obviously increased in
In this section, we first present evidence to show that magnitude, with the maximum center of the precipitation
there are remarkable differences in the mean state and moving southwestward to the Pearl River delta (Fig. 2b).
the temporal variations of the precipitation over southern Differences in the water vapor transportation to southern
China between March and the following AM. The China are also observed between AM and the previous
15 JANUARY 2018 YOU AND JIA 659

TABLE 1. The fraction of the variance explained by EOF1 of the TABLE 2. TCCs of PC1hf and PC2hf between JF and March,
monthly mean precipitation over China from January to May between March and AM, and between April and May. An asterisk
during 1951–2014 represented using percentages. indicates correlation significant at the 95% confidence level.

January February March April May JF/March March/AM April/May


EOF1 49% 48% 40% 26% 22% PC1 0.27* 0.09 0.24*
PC2 0.25* 0.12 0.26*

March. The center of the WPSH moves eastward from


The TCC between the PC2 for March and AM is only
approximately 1208E in March to 1358E in AM, causing
0.12 and does not pass the significance test, whereas it is
an eastward shift of the associated moisture transport on
0.24 between March and JF, which is again significant at
its northwestern flank. Moisture transportation along the
the 95% confidence level.
subtropical westerly at 308N still remains in AM. A third
In summary, the above results indicate that the char-
branch of water vapor transport can be observed origi-
acteristics of water vapor transportation to China in
nating from southern India, traveling across the whole of
March are quite different from those in AM. The tem-
Indo-China and finally reaching southern China. Differ-
poral variation of the precipitation in March bears more
ences in the transportation of water vapor between
similarities to those of the previous JF than to those of
March and AM can be even more clearly seen from
the succeeding AM. Therefore, in the following, AM is
Fig. 2c. Notable water vapor transportation from the
used to represent the spring season for precipitation
Indian Ocean to southern China in AM is observed,
over China. The TCC between the PC1 obtained using
denoting that the Indian Ocean plays a more important
AM and that obtained using MAM is 0.90 while that for
role in the variation in precipitation over southern
PC2 is 0.91; both are significant at the 99% significant
China in AM compared to that in the previous March.
level according to the Student’s t test examination.
To further understand the variations of the charac-
Therefore we believe that the results are not very sen-
teristics of spring precipitation over China, an EOF
sitive to the choice of the selected months in the current
analysis was performed on the monthly mean pre-
work although differences also exist.
cipitation of 160 Chinese meteorological stations from
January to May to identify the dominant precipitation
patterns. The leading two EOF modes can distinct from 4. The leading two modes of spring precipitation
one another and from the remaining eigenvectors and associated climate anomalies
according to the criterion defined by North et al. (1982).
a. Characteristics of the leading two EOF modes of
The spatial distributions of the leading EOF mode
the spring precipitation
(EOF1) of the monthly mean precipitation and their
corresponding principal component (PC1) from January In this section, we continue concentrate on exploring
to May were examined. It was found that the spatial the two leading modes of AM precipitation (EOF1-prec
patterns of EOF1 share many similarities from January and EOF2-prec) to investigate the characteristics of
to May (not shown). Table 1 presents the fractions of spring precipitation over China and their associated
variance explained by the EOF1, represented using climate anomalies. EOF1-prec and EOF2-prec account
percentages. It shows that the variance explained by for 26% and 17% of the total variance in spring pre-
EOF1 exceed 40% for January, February, and March, cipitation, respectively, and are distinct from one an-
while it decreases drastically to 26% and 22% in April other and from the remaining eigenvectors according to
and May, respectively. The drastic decline from March the criterion defined by North et al. (1982). The nor-
to AM in the percentage of the variance explained by malized principal components of the leading two EOF
EOF1 might suggest a greater similarity of the vari- modes (PC1 and PC2, respectively) are presented as
ability in precipitation in March to the previous January bars in Figs. 3b and 3d, respectively. The 9-yr running
and February than to the following April and May. averages of PC1 and PC2 are also overlaid in Figs. 3b
The temporal correlation coefficients (TCCs) of the and 3d, represented by solid black lines. Both PC1 and
principal components (PCs) associated with the leading PC2 show clear interannual and interdecadal variability
two EOF modes among various months for the period of for the period under examination. The spatial distribu-
1951–2014 are presented in Table 2. The TCC for the tions of the leading two EOF modes, represented by
PC1 between March and JF is 0.27, statistically signifi- linear regression of the spring precipitation over China
cant at the 95% confidence level. The TCC is, however, onto PC1 and PC2, are presented in Figs. 3a and 3c,
not significant and presents a value of only 0.09 between respectively. The spatial distribution of EOF1-prec
March and AM, which cannot pass the significance test. features a monosign pattern with the maximum
660 JOURNAL OF CLIMATE VOLUME 31

FIG. 3. The linear regression of AM precipitation over China onto its corresponding time series of (a) EOF1 and
(c) EOF2 for 1951–2014. The normalized time series (bars) are presented for (b) PC1 and (d) PC2. The black lines
in (b) and (d) denote the 9-yr running mean of the time series.

loading located over southern China and with amplitudes b. Large-scale circulation anomalies associated with
decreasing from southeast to northwest (Fig. 3a). A posi- EOF1-prec
tive (negative) EOF1-prec is represented by enhanced
We first examine the large-scale circulation anomalies
(reduced) rainfall anomalies in southern China. To un-
associated with EOF1-prec. The regression maps of the
derstand the representativeness of EOF1-prec for spring
winds at 850 hPa onto PC1hf are depicted in Fig. 5a. The
precipitation over China, an index is constructed using the
EOF1-prec associated water vapor transportation and
normalized area-averaged AM precipitation over 208–
the divergence are calculated and presented in Fig. 5b.
308N, 1108–1208E [precipitation index (PI); Fig. 4]. The
In Fig. 5a, associated with a positive EOF1-prec an
TCC between PC1 and PI is 0.98 for the period 1951–2014,
anomalous lower-level anticyclone is clearly seen cen-
far exceeding the 99% significance level, indicating that
tered at 208N, 1408E. Anomalous southwesterlies are
PC1 can serve as a reasonable indicator of the variation in
observed along the northwest of the anomalous anticy-
the spring precipitation over southern China. The main
clonic system transporting moisture from the western
feature of EOF2-prec is a north–south dipole structure
tropical Pacific Ocean to southern China (Fig. 5b).
with an out-of-phase relationship between the Pearl River
Water vapor converges in southeastern China consistent
delta and the middle and lower reaches of the Yangtze
with the enhanced precipitation over there during a
River (Fig. 3c). A positive (negative) phase of EOF2-prec
positive EOF1-pr.
is featured by enhanced (suppressed) precipitation over
As we mentioned in the introduction, many previous
the Yangtze River and suppressed (enhanced) pre-
studies have revealed that ENSO is one of the most
cipitation over the Pearl River delta.
important factors that influences precipitation anoma-
Some previous studies have noted that mechanisms
lies in southern China (e.g., Li 1988; Wang et al. 2000;
accounting for the climate anomalies over East Asia are
Xie et al. 2009; Jia et al. 2014a,b). The PSAC is the key
different for interannual and interdecadal time scales.
In the current study, we focus on examining the in-
terannual variations of spring precipitation over China
and therefore a Fourier harmonic bandpass filter is used
to remove the variations with periods longer than 10 yr.
Only the time scales between 2 and 9 yr have been kept
in the data and are denoted by subscript ‘‘hf’’ in the
following sections. This will also reduce the influence of
plausible unrealistic interdecadal data variation on the FIG. 4. The solid line depicts the normalized area-mean AM
precipitation anomalies for 208–308N, 1108–1308E (PI), and PC1 is
interannual relationship (e.g., Inoue and Matsumoto denoted by the dashed line. The TCC between PC1 and PI is 0.98
2004; Wu et al. 2005). The results obtained are not for the period 1951–2014, which is shown in the upper-right corner
sensitive to the filter method used in the current study. of the figure.
15 JANUARY 2018 YOU AND JIA 661

FIG. 5. The regression maps of (a) 850-hPa wind (m s21) and (b) water vapor transport [vector, kg (m s)21] and
divergence (shading, mm day21) onto PC1hf. Red (blue) shading in (a) denotes positive (negative) correlations sig-
nificant at the 95% confidence level. Stippled regions in (b) denote correlations significant at the 95% confidence level.

system that bride remote ENSO forcing to East Asian persist from wintertime to the following summer. As
climate variation (e.g., Wang et al. 2000; Xie et al. 2009), suggested by previous studies (e.g., Wang et al. 2000,
which can also be seen from Fig. 5. To see the time evo- 2003), during positive ENSO events, in the ensuing
lution impact of ENSO on EOF1-prec, the EOF1-prec- spring and summer, as the eastern tropical Pacific posi-
related tropical Pacific SSTAs from the preceding tive SSTAs decay rapidly, over the western tropical
autumn fOctober–December, for year 21 [OND(21)]g Pacific the western tropical Pacific anticyclone can be
to the succeeding summer [JJA(0)] are presented in Fig. 6. maintained through local air–sea interactions. More
Corresponding to a positive EOF1-prec, in the preceding specifically, anomalous northeastern winds on the east
autumn (Fig. 6a), significant El Niño–like SSTAs domi- side of the anticyclonic system strengthen climatological
nate the eastern tropical Pacific. Meanwhile, pronounced trade winds, enhancing the latent heat flux from the
negative SSTAs are noticed over the western tropical ocean to the atmosphere and cooling its underlying SST.
Pacific. In the following winter (Fig. 6b), to the west flank Negative SSTAs excite anomalous westward propagat-
of the western tropical Pacific SSTAs, a horseshoe-like ing and descending Rossby waves, reinforcing the anti-
positive SSTA appears, forming a tropical Pacific tripolar cyclonic system in return. On the west side of the
SSTA pattern. In the following spring (Fig. 6c), the eastern negative SSTAs, southerlies prevail along the western
tropical Pacific SSTAs decrease sharply and almost dis- flank of the anticyclonic system, transporting warm air
appear. In contrast, the western tropical Pacific negative from low-latitude oceans and contributing to the posi-
SSTAs and the positive SSTAs around southern China tive SSTAs around the South China Sea and surround-
remain throughout this season, with the positive SSTAs ing regions.
strengthening in amplitude and reaching their peak. The To better understand the time relationship between
eastern tropical Pacific SSTAs transfer to a negative phase EOF1-prec and ENSO, the lead–lag correlation be-
in the following summer and combine with the negative tween PC1hf and the Niño-3.4 index is depicted in Fig. 7.
SSTAs over the western tropical Pacific, forming La Niña– It shows that the simultaneous TCC between the PC1hf
like SSTAs (Fig. 6d). The positive SSTAs around the and the Niño-3.4 index is not significant and, with a value
South China Sea still remain in the following summer and of only 0.12, is consistent with Fig. 6c. The TCC between
weaken in amplitude. the PC1hf and the Niño-3.4 index in the previous sea-
Compared to the SSTAs over the eastern tropical sons, however, is significant at the 95% confidence level
Pacific, which adapt from an El Niño–like SSTA pattern and can be traced back to the time when the Niño-3.4
in the preceding autumn to a La Niña–like SSTA pattern SST leads PC1hf by 12 months. The TCC between PC1hf
in the following summer, the western tropical Pacific and Niño-3.4 index remains significant until the follow-
negative SSTAs and the positive SSTAs on its northwest ing year, but with a reversed sign. The above analysis
662 JOURNAL OF CLIMATE VOLUME 31

FIG. 6. The regression maps of SST in (a) OND(21), (b) January–March [JFM(0)], (c) AM(0),
and (d) JJA(0) onto PC1hf. Solid (dashed) contours indicate positive (negative) values. Light and
dark red (blue) shading denotes positive (negative) correlation coefficients significant at the 95%
and 99% confidence levels, respectively. The contour interval is 0.058C, and zero contour is omitted.

indicates that the Niño-3.4 SST in the previous season contribute to the EOF2-prec are not well explored. The
can serve as a good predictor for the EOF1-prec for regression maps of SLP and 500-hPa geopotential height
seasonal forecasting. (Z500) onto PC2hf are further examined and displayed
in Figs. 9a and 9b, respectively. At the surface, pro-
c. Large-scale circulation anomalies associated with nounced positive SLP anomalies are clearly seen dom-
EOF2-prec inating the midlatitude western North Pacific, inducing
Similar to Fig. 5, the EOF2-prec related climate cir- cyclonic anomalous southerly wind over the coastal
culation anomalies are presented in Fig. 8. The EOF2-
prec associated winds at 850 hPa (Fig. 8a) show that
southwesterlies prevail along the coastal area of eastern
China from 208 to 308N while southeasterlies dominate
from 308 to 408N. The winds along the coastal area turn
to southwesterlies over the regions north of the Yangtze
River. The spatial distribution of the EOF2-prec-related
water vapor transport (Fig. 8b) suggests a low-level cy-
clonic southerly wind inducing anomalous moisture
convergence over coastal China north of 308N and
moisture divergence over southeastern China, consis-
tent with the north–south dipole structure of EOF2-
prec, which shows an out-of-phase relationship between
the Pearl River delta area and the middle and lower
FIG. 7. The lead–lag TCCs between PC1hf and the Niño-3.4 index
reaches of the Yangtze River region. from the previous December [Dec(22)] to the following Septem-
Compared to the ENSO-related PSAC that can im- ber [Sep(11)]. The short-dashed lines represent the 95% and 99%
pact the EOF1-prec, the circulation anomalies that confidence levels according to the Student’s t test.
15 JANUARY 2018 YOU AND JIA 663

FIG. 8. As in Fig. 5, but for the regression maps onto PC2hf.

regions of China along its western flank (Fig. 9a). In the onto PC2hf is also overlaid on this map (contours). It
midtroposphere, the atmospheric circulation anomalies shows that the wave activity flux originates from the
over the mid-to-high latitudes of the Northern Hemi- midlatitude North Atlantic Ocean, flows eastward
sphere are dominated by a wave train–like structure, across the Eurasian continent, and reaches the Japan
originating from the North Atlantic Ocean, across the Sea. This distribution of the wave activity flux indicates
Ural Mountains, propagating eastward across the whole that the North Atlantic Ocean may serve as a possible
Eurasian continent (Fig. 9b). As we mentioned before, energy source region for the EOF2-prec-related wave
the wave activity flux proposed by Takaya and train–like circulation anomalies.
Nakamura (2001) is a good diagnostic tool for exploring Inspired by Fig. 10, the EOF2-prec-related North
the source region of atmospheric motion. To trace the Atlantic SSTAs from the previous February to the fol-
source of the EOF2-prec related wave train–like circu- lowing June are examined and presented in Fig. 11.
lation anomalies in Fig. 9b, the wave activity flux at Z500 Statistically significant negative SSTAs over the sub-
associated with PC2hf is calculated and presented in polar North Atlantic and positive SSTAs to the south
Fig. 10 represented by vectors. The regression of Z500 are observed in spring (Fig. 11c). This north–south

FIG. 9. The regression maps of (a) SLP (hPa) and (b) Z500 (m) onto PC2hf. Solid (dashed)
contours represent positive (negative) values. Light and dark red (blue) shading denotes
positive (negative) correlation coefficients significant at the 95% and 99% confidence levels,
respectively. The contour interval is 0.25 hPa in (a) and 4 m in (b), and zero contour is omitted.
664 JOURNAL OF CLIMATE VOLUME 31

FIG. 10. The regression map of the Z500 (contour, m) and the wave activity flux (vector, m2 s22)
onto PC2hf.

SSTAs dipole can be traced back to the preceding when the NASDI leads PC2hf by two months. The TCC
February and peaks in magnitude in the preceding reaches its maximum in March and then decreases with
March and almost disappears in the following June. time, consistent with Fig. 11. The above analysis sug-
These results indicates that the SSTA over subtropical– gests that the SSTA dipole over the North Atlantic
midlatitude North Atlantic is a possible source of the Ocean of the previous March can excite a stationary
wave train–like circulation anomalies. A North Atlantic Rossby wave train–like atmospheric pattern that
SST dipole index (NASDI) is then defined by differ- propagates eastward to the downstream region. The
ences in normalized SSTAs between 358–458N, 658– positive anomalies of the wave train–like atmospheric
458W and 508–608N, 508–108W. This index is not very pattern over the western tropical Pacific can influence
sensitive to the region selected above. Figure 12 presents the moisture transport from the low-latitude ocean to
the lead–lag temporal correlation coefficients between southern China in the following AM, implying that the
PC2hf and the NASDI. It shows that the TCC between North Atlantic SSTAs dipole in the previous March could
PC2hf and the NASDI exceeds the 95% confidence level serve as a predictor for the EOF2-prec. The TCC

FIG. 11. The regression maps of SST in (a) February, (b) March, (c) AM, and (d) June onto PC2hf. Solid (dashed)
contours represent positive (negative) values. Light and dark red (blue) shading denotes positive (negative) cor-
relation coefficients significant at the 95% and 99% confidence levels, respectively. The contour interval is 0.058C,
and zero contour is omitted.
15 JANUARY 2018 YOU AND JIA 665

between the NASDI and the NAO index in the previous


winter has been examined. It shows that the TCC be-
tween NASDI in March is significantly correlated to the
NAO index in the previous February and January with
TCCs of 0.44 and 0.41, respectively, significant at the 99%
significant level, indicating that the North Atlantic SST
dipole may be the result of the NAO from the previous
winter. However, more studies are needed to better un-
derstand their mechanisms.
Wave train–like patterns originating from the North
Atlantic and propagating downstream to East Asia have
been observed in several previous studies (e.g.,
Watanabe. 2004; Wu et al. 2009; Wu et al. 2011; Li et al.
2015). For example, a significant wet trend in winter was FIG. 12. The lead–lag correlation coefficients between PC2hf and
NASDI from the previous October [Oct(21)] to the following No-
observed after the 1970s over the Yangtze River by Li vember [Nov(0)]. The short-dashed lines represent the 95% and
et al. (2015). They related the wet trend to a wave train– 99% confidence levels, respectively, according to the Student’s t test.
like pattern originating in western Europe and extend-
ing downstream to northeastern China (Li et al. 2015;
their Fig. 6). Wu et al. (2009) suggested that a tripolar An EOF analysis similar to that used for the obser-
SST pattern over the North Atlantic in the spring could vational data was applied to the MME forecast of the
persist into the following summer and excite a down- AM precipitation over China for the period of 1960–
stream teleconnection pattern prevailing along the pole- 2005 to obtain the forecast leading two EOF modes. The
ward flank of the westerly jet and that this pattern could regression maps of the MME forecast AM precipitation
affect the Asian subtropical front of the following summer. onto the associated normalized time series have been
Watanabe (2004) showed that in February, when the presented in Fig. 13. It shows that the forecast leading
NAO accompanies the Mediterranean convergence two EOF modes of the spring precipitation share many
anomaly, the NAO signals can extend downstream similarities to EOF1-prec and EOF2-prec in the obser-
through a wave train–like pattern along the Asian jet vations, indicating that the MME of the six coupled
stream. In the current study, we further demonstrate that, numerical models can reasonably capture the spatial
on the interannual time scale, the SSTA dipole over the structures of the observed leading two EOF modes of
North Atlantic Ocean of the previous winter can excite a the spring precipitation over China. However, the TCC
stationary Rossby wave train–like pattern that propa- between the forecast PC1hf and the observed PC1hf is
gates eastward and contributes to precipitation over only 0.08 while that between the forecast PC2hf and the
China in the following spring. More specifically, it can observed PC2hf is 0.07, indicating that the numerical
contribute to the second EOF mode of spring pre- models present almost no capacity to predict variations
cipitation over China by modulating the circulation of the leading two EOF modes in the observations.
anomalies over the western North Pacific regions. In the following, to reveal the ability of the numerical
models’ forecasting the leading two EOF modes of the
observational precipitation over China, the MME fore-
5. The seasonal prediction of spring precipitation
cast of AM precipitation is projected onto the observa-
over China
tional EOF1-prec and EOF2-prec as shown in Figs. 3a
In this section, the seasonal prediction of spring pre- and 3c, respectively, to obtain the MME forecast time
cipitation over China is examined. The performance of series of the observational EOF1-prec and EOF2-prec.
six coupled climate models in predicting the variations The normalized forecast time series, namely MME-
in spring precipitation over China is first examined. PC1hf and MME-PC2hf, are presented as blue lines in
We focus on examining the capacity of the numerical Figs. 14a and 14b, respectively. The corresponding time
models’ to forecast the leading two EOF modes of series of observational PC1hf and PC2hf are also overlaid
spring precipitation. A physics-based empirical (P-E; as black lines in Fig. 14 for the purpose of a better
Wu et al. 2009; Yim et al. 2014; Wang et al. 2015) model comparison. The TCC between MME-PC1hf and PC1hf
is then established according to the observational anal- is only 0.03, whereas that between MME-PC2hf and
ysis results presented in the last section with the aim to PC2hf is only 0.07 for the period 1960 to 2005; both
improve the seasonal forecast skill of spring pre- cannot pass the significance test, indicating that the MME
cipitation over China. forecast AM precipitation presents almost no capacity to
666 JOURNAL OF CLIMATE VOLUME 31

FIG. 13. The linear regression maps of MME forecast of AM precipitation over China onto (a) PC1hf and (b) PC2hf.
The contour interval is 3 mm month21, and zero contour is omitted.

predict the variations of the observational leading two Following Wu et al. (2009) and Lin and Wu (2011),
EOF modes of spring precipitation over China. a ‘‘leaving 16 out’’ strategy is used to determine the
In the last section, the analysis results suggest that the robustness of the hindcast results for the period of 1951–
Niño-3.4 index in the previous season can serve as a 2014. The cross-validation method systematically re-
predictor for PC1hf, whereas the NASDI in the previous moves 16 years. A forecast model is then derived for the
March is significantly correlated with PC2hf. In the fol- remaining years and can be tested on the omitted cases.
lowing, the Niño-3.4 index (NINO34) of the previous In the present study, 16 years account for about 25% of
winter and the NASDI in the preceding March are se- the total hindcast period (64 yr), which can prevent
lected to construct a physics-based empirical model using overfitting and data wasting in the process. To test
the linear regression method to predict the observational the predictive capability of the P-E model, the cross-
PC1hf and PC2hf. The P-E model is expressed as validation method is applied to perform the hindcast.
The cross-validated estimates of PC1hf and PC2hf are
PC1 5 a0 1 a1 NINO34 1 «1 , (1) shown as red lines in Fig. 14. The TCC between the
PC2 5 b0 1 b1 NASDI 1 «2 , (2) hindcast PC1hf from the P-E model and the observa-
tional PC1hf (black line) is 0.35, which exceeds the 99%
where a0, a1, b0, and b1 are regression coefficients, and «1 confidence level. The TCC between the hindcast PC2hf
and «2 are the residuals. from the P-E model and the observational PC2hf is 0.26,

FIG. 14. (a) The PC1hf of the observational data (black lines), the P-E model (red lines) and
the MME (blue lines). The TCC of PC1hf between the observational data and the P-E model as
well as that between the observational data and the MME are indicated on the bottom-left
corner. An asterisk indicates correlation significant at the 95% confidence level. Purple plus
signs represent the predicted PC1hf derived from the P-E model based on data before 2000.
(b) As in (a), but for PC2hf.
15 JANUARY 2018 YOU AND JIA 667

with a statistical significance t test P value of 0.001, time evolution of the observational EOF1-prec and
again, statistically significant at the 99% confidence EOF2-prec, in the following another regression model is
level. Therefore, the results indicate that the P-E model constructed aim to improve the seasonal forecasting of
clearly performs better in predicting the variations of spring precipitation over China. In essence, this approach
the observational EOF1-prec and EOF2-prec than the uses the preforecast time series from the P-E model [(1)
MME forecasts of the six coupled climate models for the and (2)] and the spatial structures of the observational
period under examination. EOF1-prec and EOF2-prec to construct a new set of
To further evaluate the real forecasting ability of the seasonal forecasts. The regression model can be written as
P-E model, another experiment was performed. A similar
P-E model is established; however, it is trained using the Pre(x, y, t) 5 PC1(t)EOF1(x, y) 1 PC2(t)EOF2(x, y) 1 «,
data before 2001. This P-E model is then used to do pre- (4)
dictions for 2001–14 (shown as purple plus signs in
Fig. 14). The result shows that the P-E model can still where PC1 and PC2 are the time series obtained from
reasonably predict the variation of EOF1-prec and EOF2- the P-E model [(1) and (2)]. EOF1 and EOF2 are the
prec for this period. Some extreme years of PC1, such as spatial structures of the observational EOF1-prec and
2010 and 2011, are predicted well by the P-E model, al- EOF2-prec, respectively, and « is the residual and is
though at a smaller magnitude compared to the observa- omitted in the current calculations. Again, a cross-
tions, possibly due to fewer predictors in the equations. validation method is used when calculating PC1 and
As we know, in the Northern Hemisphere various ac- PC2 to mimic an operational environment.
tive teleconnection patterns over the mid-to-high lati- The correlation skill of the above regression model in
tudes can affect climate variations (e.g., Wallace and forecasting spring precipitation over China is depicted in
Gutzler 1981). Some of these may also be related to the Fig. 15b and that for the MME forecast AM pre-
pattern shown in Fig. 9. We examined the TCCs between cipitation of the six coupled climate models is presented
the PC2hf and the indices of several well-known tele- in Fig. 15a. The results show that the TCC skill of
connection patterns that have been identified by previous the MME forecast AM precipitation barely exceeds the
studies. Results show that the TCC between PC2hf and 90% confidence level over southern China, where the
the index of the west Pacific (WP) pattern (WPI) reaches climatological mean and the variations have the largest
0.34, statistically significant beyond the 99% confidence loadings. In contrast, the TCC skill of the regression
level. As the WP pattern dominates the subtropical model is clearly improved over southern China, espe-
northwestern Pacific and Kamchatka (figures not shown), cially over the Yangtze River area. The results suggest
displaying as a dipole structure in the north–south di- that the regression model can efficiently improve the
rection, and the PCC between WP and EOF2-prec rea- seasonal forecasting of spring precipitation over south-
ches 0.38. Therefore, we speculate that the WP may ern China by improving the forecast skill of the time
interact with the North Atlantic SSTA dipole forced series of the observational EOF1-prec and EOF2-prec.
wave train–like pattern over the western North Pacific Another linear regression model is established using
and together contribute to EOF2-prec variations. Specific PC1 forecast by P-E model in (1) and PC2 forecast by
mechanisms of this interaction are, however, beyond the P-E model in (3). The result of the TCC skill (Fig. 15c)
scope of the present study and may be investigated in shows that the spring precipitation forecast skill has
future studies. The TCC between the WPI and NASDI is been further improved over the southern China area.
not significant, suggesting that they can serve as an in-
dependent precursor for EOF2-prec. The P-E model for
6. Summary and discussion
PC2 is now written in the following format:
Using both observational data and the MME of the
PC2 5 b0 1 b1 NASDI 1 b2 WPI 1 «2 . (3) 1-month-lead seasonal forecast made by six coupled
climate models, the interannual variations and the pre-
The TCC between the time series of the P-E model in diction of spring precipitation over China are examined.
(3) predicted and the observational PC2 is now in- Evidence has been first presented to show that the pre-
creased to 0.39, which is statistically significant at the cipitation over China in spring is better represented by
99% confidence level. It suggests that the WP can clearly that of April–May (AM). The characteristics and the
improve the seasonal forecast skill for the time variation variations of the precipitation in March show a closer
of EOF2-prec. relationship with those of the previous January–
Inspired by the above results that show that the P-E February (JF) compared to those of the following AM.
model can significantly improve the forecast skill of the Therefore, AM is used to represent the spring season for
668 JOURNAL OF CLIMATE VOLUME 31

FIG. 15. (a) The TCC (contours) of AM precipitation between the observational data and the MME forecasts of the six coupled climate
models for 1960–2005. (b) As in (a), but for the TCC between observational data and the regression model using time series forecast by the
P-E model in (1) and (2). (c) As in (b), but using time series forecast by the P-E model in (1) and (3). Light and dark shading denotes
correlation significant at the 90% and 95% confidence levels, respectively.

the precipitation over China in the current study. The monitored in real time; therefore, the proposed P-E
interannual variations of spring precipitation over model and the regression model can serve as an efficient
China are investigated by applying an EOF analysis to real-time forecasting tool for spring precipitation over
the observational precipitation data. Results show that southern China.
EOF1-prec is characterized by a monosign pattern with It is worth noting that in this study we only examined
dominant loading over southern China. A positive the first two EOF modes of spring precipitation over
EOF1-prec, which features enhanced rainfall in China, which together explain approximately 43% of
southern China, is closely related to El Niño–type the total variation in precipitation for 1951–2014. The
SSTAs over the eastern tropical Pacific in the pre- remaining EOF modes of spring precipitation over
ceding winter. The EOF2-prec, which features a north– China are not discussed. However, over southern China,
south dipole pattern of opposite signs centered over the the EOF1-prec and EOF2-prec together can explain
Yangtze River and Pearl River delta, is significantly more than 70% of the variability over the Yangtze River
correlated with equivalent wave train–like circulation and more than 60% of variability over southeastern
anomalies in northern mid-to-high latitudes. This wave China for spring precipitation (figure not shown). This
train–like pattern can impact the variations of spring may explain why the improved forecasting skill of the
precipitation over China by modulating the circulation time series associated with prec-EOF1 and prec-EOF2
anomalies over the western North Pacific, causing could efficiently improve the seasonal forecasting of
anomalous moisture convergence over the Yangtze spring precipitation over these regions. Furthermore, in
River alongside divergence conditions in southern the current study, the underlying assumption of the P-E
China. The prec-EOF2 is found to be related to a me- model and the linear regression model is that the EOF
ridional SST anomaly dipole over the North Atlantic, modes remain relatively stable during the period under
which can be traced back to the previous March. examination. However, some previous studies suggest
The MME forecast AM precipitation of six coupled that the precipitation over China undergoes significant
climate models show almost no skill in predicting the decadal modulations (e.g., Xin et al. 2006; Qiu et al.
variations of EOF1-prec and EOF2-prec. A physics- 2009; Wu and Mao 2016). For example, Xin et al. (2006)
based empirical (P-E) model was formulated based on suggested that late spring precipitation in southern
the results of the observational analysis to improve the China decreased significantly after the late 1970s in as-
seasonal forecast skill. It has been demonstrated that sociation with upper troposphere anomalous cooling
this P-E model performs much better in predicting the over central China. Qiu et al. (2009) showed that late
time series of EOF1-prec and EOF2-prec. A linear re- spring rainfall over southeastern China has decreasing
gression model is further constructed using the P-E by more than 30% since 1951. Wu and Mao (2016) also
model predictions to forecast the spring precipitation showed that the impact of ENSO on spring precipitation
over China. Results suggest that the forecast skill of the in southern China can be modulated by the Pacific de-
spring precipitation over southeastern China, especially cadal oscillation (PDO) through effects of variations in a
over the Yangtze River area, can be significantly im- low-level subtropical anticyclonic system over the western
proved by the regression model. The precursors used North Pacific. Additional studies on interdecadal changes
in the experiments described above can be easily of the predictive capacity of the P-E model and the
15 JANUARY 2018 YOU AND JIA 669

regression model may therefore need to be conducted in a Li, C. Y., 1988: The frequent actions of the East Asian trough and
future study. the occurrences of El Niño. Sci. China, 31B, 667–674.
Li, X.-F., J. Li, and Y. Li, 2015: Recent winter precipitation in-
crease in the middle-lower Yangtze River valley since the
Acknowledgments. This research was funded by Na- late 1970s: A response to warming in the tropical Indian
tional Natural Science Foundation of China (Grants Ocean. J. Climate, 28, 3857–3879, https://doi.org/10.1175/
41475065 and 41530425) and by the Fundamental Re- JCLI-D-14-00701.1.
search Funds for the Central Universities. The authors Lin, H., and Z. Wu, 2011: Contribution of the autumn Tibetan
Plateau snow cover to seasonal prediction of North American
would like to acknowledge the support from the training winter temperature. J. Climate, 24, 2801–2813, https://doi.org/
center of atmospheric sciences of Zhejiang University. 10.1175/2010JCLI3889.1.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982:
Sampling errors in the estimation of empirical orthogonal
REFERENCES
functions. Mon. Wea. Rev., 110, 699–706, https://doi.org/
Adler, R. F., and Coauthors, 2003: The Version-2 Global Precipitation 10.1175/1520-0493(1982)110,0699:SEITEO.2.0.CO;2.
Climatology Project (GPCP) monthly precipitation analysis Qiu, Y., W. Cai, X. Guo, and A. Pan, 2009: Dynamics of late spring
(1979–present). J. Hydrometeor., 4, 1147–1167, https://doi.org/ rainfall reduction in recent decades over southeastern China.
10.1175/1525-7541(2003)004,1147:TVGPCP.2.0.CO;2. J. Climate, 22, 2240–2247, https://doi.org/10.1175/2008JCLI2809.1.
Chang, C. P., Y. Zhang, and T. Li, 2000: Interannual and inter- Smith, T. M., R. W. Reynolds, T. C. Peterson, and J. Lawrimore,
decadal variations of the East Asian summer monsoon 2008: Improvements to NOAA’s historical merged land–
and tropical Pacific SSTs. Part I: Roles of the subtropical ocean surface temperature analysis (1880–2006). J. Climate,
ridge. J. Climate, 13, 4310–4325, https://doi.org/10.1175/ 21, 2283–2296, https://doi.org/10.1175/2007JCLI2100.1.
1520-0442(2000)013,4310:IAIVOT.2.0.CO;2. Takaya, K., and H. Nakamura, 2001: A formulation of a phase-
Chen, J.-P., Z.-P. Wen, R. Wu, Z.-S. Chen, and P. Zhao, 2014: independent wave-activity flux for stationary and migra-
Interdecadal changes in the relationship between southern tory quasigeostrophic eddies on a zonally varying basic
China winter–spring precipitation and ENSO. Climate Dyn., flow. J. Atmos. Sci., 58, 608–627, https://doi.org/10.1175/
43, 1327–1338, https://doi.org/10.1007/s00382-013-1947-x. 1520-0469(2001)058,0608:AFOAPI.2.0.CO;2.
Feng, J., and J. Li, 2011: Influence of El Niño Modoki on spring Tao, S., and L. Chen, 1987: A review of recent research on the East
rainfall over south China. J. Geophys. Res., 116, D13102, Asian summer monsoon in China. Monsoon Meteorology,
https://doi.org/10.1029/2010JD015160. C. P. Chang and T. N. Krishnamurti, Eds., Oxford University
Feng, J.-Q., L.-J. Yu, and D.-X. Hu, 2014: Influence of Indian Press, 60–92.
Ocean subtropical dipole on spring rainfall over China. Int. Wallace, J. M., and D. S. Gutzler, 1981: Teleconnections in the
J. Climatol., 34, 954–963, https://doi.org/10.1002/joc.3732. geopotential height field during the Northern Hemisphere
Gao, H., and Coauthors, 2008: Analysis of the severe cold surge, ice- winter. Mon. Wea. Rev., 109, 784–812, https://doi.org/10.1175/
snow and frozen disasters in South China during January 2008: II. 1520-0493(1981)109,0784:TITGHF.2.0.CO;2.
Possible climate causes (in Chinese). Meteor. Mon., 34, 101–106. Wang, B., R. Wu, and X. Fu, 2000: Pacific–East Asian teleconnec-
Huang, R., L. Zhou, and W. Chen, 2003: The progresses of recent tion: How does ENSO affect East Asian climate? J. Climate, 13,
studies on the variabilities of the East Asian monsoon and 1517–1536, https://doi.org/10.1175/1520-0442(2000)013,1517:
their causes. Adv. Atmos. Sci., 20, 55–69, https://doi.org/ PEATHD.2.0.CO;2.
10.1007/BF03342050. ——, ——, and T. Li, 2003: Atmosphere–warm ocean in-
——, G. Huang, and Z. Wei, 2004: Climate variations of the sum- teraction and its impacts on Asian–Australian monsoon
mer monsoon over China. East Asian Monsoon, C.-P. Chang, variation. J. Climate, 16, 1195–1211, https://doi.org/10.1175/
Ed., World Scientific Publishing, 213–270. 1520-0442(2003)16,1195:AOIAII.2.0.CO;2.
Inoue, T., and J. Matsumoto, 2004: A comparison of summer sea ——, and Coauthors, 2009a: Advance and prospectus of seasonal
level pressure over East Eurasia between NCEP-NCAR prediction: Assessment of the APCC/CliPAS 14-model en-
reanalysis and ERA-40 for the period 1960–99. J. Meteor. semble retrospective seasonal prediction (1980–2004). Climate
Soc. Japan, 82, 951–958, https://doi.org/10.2151/jmsj.2004.951. Dyn., 33, 93–117, https://doi.org/10.1007/s00382-008-0460-0.
Jia, X. J., J.-Y. Lee, H. Lin, A. Alessandri, and J.-K. Ha, 2014a: ——, J. Liu, J. Yang, T. Zhou, and Z. Wu, 2009b: Distinct
Interdecadal change in the Northern Hemisphere seasonal principal modes of early and late summer rainfall anoma-
climate prediction skill: Part I. The leading forced mode of lies in East Asia. J. Climate, 22, 3864–3875, https://doi.org/
atmospheric circulation. Climate Dyn., 43, 1595–1609, https:// 10.1175/2009JCLI2850.1.
doi.org/10.1007/s00382-013-1988-1. ——, J.-Y. Lee, and B. Xiang, 2015: Asian summer monsoon
——, ——, ——, H. Hendon, and J.-K. Ha, 2014b: Interdecadal rainfall predictability: A predictable mode analysis. Climate
change in the Northern Hemisphere seasonal climate pre- Dyn., 44, 61–74, https://doi.org/10.1007/s00382-014-2218-1.
diction skill: Part II. Predictability and prediction skill. Climate Watanabe, M., 2004: Asian jet waveguide and a downstream
Dyn., 43, 1611–1630, https://doi.org/10.1007/s00382-014-2084-x. extension of the North Atlantic Oscillation. J. Climate, 17,
Kalnay, E., and Coauthors, 1996: The NCEP–NCAR 40-Year Re- 4674–4691, https://doi.org/10.1175/JCLI-3228.1.
analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, https:// Webster, P. J., and S. Yang, 1992: Monsoon and ENSO: Selectively
doi.org/10.1175/1520-0477(1996)077,0437:TNYRP.2.0.CO;2. interaction systems. Quart. J. Roy. Meteor. Soc., 118, 877–926,
Lau, K.-M., and M.-T. Li, 1984: The monsoon of East Asia and its https://doi.org/10.1002/qj.49711850705.
global associations—A survey. Bull. Amer. Meteor. Soc., Weisheimer, A., and Coauthors, 2009: ENSEMBLES: A new
65, 114–125, https://doi.org/10.1175/1520-0477(1984)065,0114: multi-model ensemble for seasonal-to-annual predictions—
TMOEAA.2.0.CO;2. Skill and progress beyond DEMETER in forecasting tropical
670 JOURNAL OF CLIMATE VOLUME 31

Pacific SSTs. Geophys. Res. Lett., 36, L21711, https://doi.org/ Xie, S.-P., K. Hu, J. Hafner, H. Tokinaga, Y. Du, G. Huang,
10.1029/2009GL040896. and T. Sampe, 2009: Indian Ocean capacitor effect on Indo–
Wu, R., and B. P. Kirtman, 2007: Observed relationship of spring western Pacific climate during the summer following
and summer East Asian rainfall with winter and spring El Niño. J. Climate, 22, 730–747, https://doi.org/10.1175/
Eurasian snow. J. Climate, 20, 1285–1304, https://doi.org/ 2008JCLI2544.1.
10.1175/JCLI4068.1. Xin, X., R. Yu, T. Zhou, and B. Wang, 2006: Drought in late spring
——, J. L. Kinter III, and B. P. Kirtman, 2005: Discrepancy of south China in recent decades. J. Climate, 19, 3197–3206,
of interdecadal changes in the Asian region among the NCEP– https://doi.org/10.1175/JCLI3794.1.
NCAR reanalysis, objective analyses, and observations. Yang, F.-L., and K.-M. Lau, 2004: Trend and variability of China
J. Climate, 18, 3048–3067, https://doi.org/10.1175/JCLI3465.1. precipitation in spring and summer: Linkage to sea-surface
——, S. Yang, S. Liu, L. Sun, Y. Lian, and Z. Gao, 2011: Northeast temperatures. Int. J. Climatol., 24, 1625–1644, https://doi.org/
China summer temperature and North Atlantic SST. J. Geophys. 10.1002/joc.1094.
Res., 116, D16116, https://doi.org/10.1029/2011JD015779. Yim, S.-Y., B. Wang, and W. Xing, 2014: Prediction of early
——, G. Liu, and P. Zhao, 2014: Contrasting Eurasian spring summer rainfall over South China by a physical-empirical
and summer climate anomalies associated with western and model. Climate Dyn., 43, 1883–1891, https://doi.org/10.1007/
eastern Eurasian spring snow cover changes. J. Geophys. Res. s00382-013-2014-3.
Atmos., 119, 7410–7424, https://doi.org/10.1002/2014JD021764. Zhang, R., and A. Sumi, 2002: Moisture circulation over East Asia
Wu, X., and J. Mao, 2016: Interdecadal modulation of ENSO- during El Niño episode in northern winter, spring and autumn.
related spring rainfall over South China by the Pacific decadal J. Meteor. Soc. Japan, 80, 213–227, https://doi.org/10.2151/
oscillation. Climate Dyn., 47, 3203–3220, https://doi.org/ jmsj.80.213.
10.1007/s00382-016-3021-y. Zuo, Z.-Y., R.-H. Zhang, B.-Y. Wu, and X. Rong, 2012: Decadal
Wu, Z., B. Wang, J. Li, and F.-F. Jin, 2009: An empirical seasonal variability in springtime snow over Eurasia: Relation with
prediction model of the East Asian summer monsoon using circulation and possible influence on springtime rainfall over
ENSO and NAO. J. Geophys. Res., 114, D18120, https://doi.org/ China. Int. J. Climatol., 32, 1336–1345, https://doi.org/10.1002/
10.1029/2009JD011733. joc.2355.

View publication stats

You might also like