Synergistic Interdecadal Evolution of Precipitation over Eastern China and the Pacific Decadal Oscillation during 1951–2015

The precipitation field over China exhibits well-defined spatiotemporal patterns associated with weather and climate systems on regional and global scales. For example, the annual precipitation in China gradually decreases from the southeast coast inland to Northwest China (Zhai et al., 2009). Less precipitation falls in the north, and more falls in the south over China, which has been previously called the south flood and north drought (SF-ND) pattern in other studies (Zhou et al., 2009; Yang et al., 2017).

The precipitation field over China exhibits multiple temporal signals, with seasonal (Lau and Li, 1984), interannual (Zhang et al., 2009, 2015), interdecadal (e.g., Zhou et al., 2006; Ren et al., 2012; Ding et al., 2020), and multi-decadal (Qian and Zhou, 2014) variabilities. For example, on the seasonal timescale, precipitation displays large seasonal differences that are significantly greater in summer than in winter. As a peculiar phenomenon associated with the East Asian monsoon rainy season (mid-June to mid-July), the mei-yu is characterized by abundant rainfall in the mid-latitude areas of China (Ding et al., 2020). On an interannual timescale, the precipitation pattern is strongly affected by El Niño-Southern Oscillation (ENSO; Chen et al., 2013; Zhou et al., 2014; Zhang and Gao, 2016; Gao et al., 2022; Hu et al. 2022a). On longer decadal-to-interdecadal scales, precipitation variability over China exhibits strong seasonal and regional dependences. For example, summer precipitation variations over eastern China exhibit several dominant modes (Ding et al., 2008), with major oscillations with cycles centered at 12 years, 30–40 years, and 80 years. Xu and Chan (2002) pointed out that there is also an interdecadal oscillation with a 10- to 32-year cycle in winter precipitation over different regions of China. Numerous processes potentially responsible for these spatial patterns and temporal variations of precipitation over China have been identified, including the local geographic configurations and mean climatic conditions. For instance, the Tibetan Plateau topography exerts direct and dynamic effects on the mei-yu (Kitoh, 2017). Accompanied by the two interdecadal abrupt changes of the East Asian summer monsoon since the 1970s, the summer precipitation pattern over eastern China has changed from the south drought and north flood (SD-NF) pattern before the late 1970s to the SF-ND pattern after the early 1990s (Ding et al., 2008). The effects of SST forcing can also modulate the interdecadal variability of precipitation patterns over eastern China. The interdecadal variations of the North Pacific SST contributed to the two interdecadal changes in the precipitation pattern over eastern China (Deng et al., 2009). Since the 1990s, the more frequent central Pacific El Niño events could cause summer precipitation patterns in China that significantly differed from traditional ENSO events by inducing different East Asian atmospheric circulation responses (Chen et al., 2014). Liu et al. (2018) pointed out that the Indian Ocean basin mode is another key driving force for the interdecadal variability of summer precipitation over Southwest China, with increased precipitation over Southwest China during the cold phase of the Indian Ocean basin mode. In terms of effects from the North Atlantic, both the North Atlantic Tripole and the Atlantic Multidecadal Oscillation can modify interdecadal variabilities of precipitation over South China via the zonal teleconnection wave train across Eurasia in the middle and high latitudes (Chen et al., 2017; Ding et al., 2020). In addition, Zhao et al. (2012) also pointed out that solar activity may regulate the decadal variations of summer precipitation over the Huaihe River Basin by amplifying or weakening the effects of East Asian summer monsoon.

The atmospheric circulation over East Asia is an important factor in controlling the changes in the spatial distribution of precipitation in China by directly altering the water vapor transport pathways (e.g., Simmonds et al., 1999; Peng and Zhou, 2017). For instance, the western Pacific subtropical high (WPSH), an important component of the atmospheric circulation system over East Asia, can alter water vapor transport from the tropical oceans to East Asia (Li and Lu, 2018). Moreover, changes in the location and intensity of the western North Pacific anomalous anticyclone can strongly influence precipitation variations over China (Fan and Fan, 2017). The anomalous anticyclone can increase the precipitation over the Yangtze River Basin and decrease the precipitation over South China and North China by modulating water vapor transport in East Asia (Chang et al., 2000; Zhang et al., 2016a, 2017b). As such, the East Asian summer monsoon can control the meridional shifts of the major seasonal rainband (Ding et al., 2008) and can further induce extensive droughts and flood disasters over East Asia (Zhang et al., 1998, 2017a). In addition, Lü et al. (2014) pointed out that the Pacific-Japan/East Asia-Pacific type of atmospheric teleconnection in terms of wave trains over East Asia can influence the precipitation distribution over most regions of China.

Furthermore, the variability of atmospheric circulation over East Asia can be affected by remote forcing from sea surface temperature (SST) in the Pacific basin (Li et al., 2019; Wu and Wang, 2019). Correspondingly, precipitation variability over China is also strongly affected by SST anomalies in the Pacific basin on different timescales. On the interdecadal scale, for example, the Pacific Decadal Oscillation (PDO), a major climate mode dominated in the North Pacific region, can strongly affect precipitation variations over China (Mantua et al., 1997). Previous studies showed that the PDO phase-related SST anomalies in the Pacific are negatively correlated with the rainband position in China due to north-south movements of the East Asian monsoon (Ding et al., 2008, 2020). Lü et al. (2014) pointed out that the PDO and the precipitation over North China in summer show a significant negative correlation on the interdecadal timescale. The PDO is also responsible for the in-phase variation in summer precipitation anomalies over the Yangtze River Basin and the Huanghe-Huaihe River Basin, which is the first leading mode (PC1) of interdecadal variability in precipitation over China (Si and Ding, 2016). Some studies further identified different patterns in precipitation variability over China during different PDO phases. For example, during the PDO warm phase, the precipitation anomaly is characterized by an SF-ND pattern over eastern China (Qian and Zhou, 2014). During the phase transition of the PDO from warm to cold in the 1940s and the late 1990s, the precipitation increased over the Huanghe-Huaihe River and decreased over the Yangtze River (Zhu et al., 2011; Dong and Xue, 2016). An opposite change was found when the PDO reversed its phase in the late 1970s (Dong and Xue, 2016).

The SST anomalies in the Pacific Ocean play an important role in the interannual-to-interdecadal precipitation variations over China through modulating atmospheric circulation variability over East Asia (Zhou et al., 2014; Zhang et al., 2016a). While their relationships and associated processes on interannual scales have been examined extensively and intensively, those on decadal-to-interdecadal scales have been lacking. Due to the complexity of precipitation variabilities affected by different climate modes, signals with interannual-to-interdecadal SST forcing and precipitation responses over China are superimposed, making it difficult to effectively isolate their cause-effect relationships. Thus, the relative roles played by SST forcing in precipitation variations over China have not been adequately quantified on interannual and interdecadal scales. There is a clear need to isolate different bands of interannual-to-interdecadal signals for precipitation and SST variabilities and then depict their separate covarying evolutionary characteristics at different scales in a clear way.

To this end, the multi-taper method–singular value decomposition (MTM-SVD) method is applied in this study to isolate signals with different frequency bands of precipitation and SST variabilities. This allows for the focused analysis on specific decadal-interdecadal cycles of precipitation responses over China and SST forcing in the Pacific. In addition, this method not only separates the significant cyclic signals of variabilities but also obtains their coherent evolutionary patterns through spatiotemporal reconstruction at a specific cycle (e.g., a decadal timescale). We aim to understand the co-variations among the precipitation over China, North Pacific SST, and related atmospheric fields on decadal-interdecadal scales. Additionally, we analyze processes and mechanisms by which the PDO influences the East Asia atmospheric circulation, which affects the precipitation distribution over eastern China at specific interdecadal frequencies. The relative contributions of different interdecadal signals to the precipitation variations over the different regions of China can then be assessed.

The remainder of this paper is structured as follows. Section 2 describes the observational datasets and analysis methods. In section 3, we isolate the dominant cycles in the precipitation field over China and the North Pacific SST field on the interdecadal timescale. Section 4 describes the synergistic evolutionary characteristics of the precipitation, SST, and related atmospheric fields for the 11-year cycle signal and then analyzes the mechanisms involved by which the PDO can affect the precipitation distribution over eastern China. Section 5 is similar to section 4 but focuses on the 23-year cycle signal. Section 6 highlights a 1998 case for extreme summer precipitation in China by analyzing the relative contribution of the 11-year and 23-year cycle signals. The conclusions and discussion are given in section 7.

The monthly Global Precipitation Climatology Centre reanalysis data (Adler et al., 2003) with a horizontal resolution of 1° × 1° are used in this study. The monthly SST data are obtained from the Met Office Hadley Centre Sea Ice and Sea Surface Temperature version 1.1 dataset with a horizontal resolution of 1° × 1° (Rayner et al., 2003), which are interpolated onto a 2.5-degree grid to save calculation time for the significance confidence levels by the bootstrap method in the analysis. To analyze the variations of the atmospheric circulations, monthly wind, and geopotential height (GHT) fields are obtained from the National Centers for Environmental Prediction-National Centre for Atmospheric Research (NCEP-NCAR), with a horizontal resolution of 2.5° × 2.5° and 17 vertical levels (from 1000 to 10 hPa) (Kalnay et al., 1996). The monthly sea level pressure (SLP) data are from the 20th century reanalysis version 3 dataset, with a horizontal resolution of 1° × 1° (Compo et al., 2011). The SLP data from NECP-NCAR are also applied to verify our results, which showed similar features. The NCEI PDO index (Mantua, 1999) is used to quantify the similarity between the reconstructed North Pacific SST mode and the canonical PDO pattern. All datasets used in this study have the same time record length from 1951 to 2015 (65 years or 780 months) because the GPCC precipitation data are only available up to 2016. The climatological mean is defined as the mean value over the period 1971–2010.

The MTM-SVD method is adopted to detect and reconstruct climate signals on various timescales (Mann and Park, 1994). This method skillfully combines the multi-taper method (MTM) of spectral analysis (Thomson, 1982) and the singular value decomposition (SVD) of variables. Unlike typical decompositions in the time domain (e.g., standard principal component analysis), the MTM-SVD method first converts a single field or combined multiple fields into the spectral domain through the MTM process and then performs the SVD to find significant patterns of spatiotemporal variabilities. It can identify climate signals and track the evolutionary characteristics of two or more coupled climate fields (such as SST, pressure, and precipitation fields) at different frequencies (Mann and Park, 1994; Wei et al., 2013). This approach has been widely used to isolate the characteristics of global climate oscillations on different timescales (Mann et al., 2021).

First, the MTM is used to transform the time domain into the spectral domain for each time series of grid data with M points (a single spatial field or joint analysis of multiple spatial fields). The time series, x, at the m-th grid point is first standardized, and then the Fourier transform is performed at each frequency f. Thus, the eigenspectra Y(f) takes the form:

where $ n\Delta t $ is the sample interval (month, season, year, etc.), and $ {a}_{k}^{m}\left(t\right) $ is the given k-th orthogonal (Slepian) data tapers. For a given frequency f₀, the matrix Y(f) has the following form:

Then, a complex SVD on the matrix Y($ {f}_{0} $) at each frequency ($ {f}_{0} $) is performed as

where the $ {\mathrm{\gamma }}_{k}\left({f}_{0}\right) $ is the singular value. The left singular vector, $ {U}_{k}^{m}\left({f}_{0}\right) $, is the spatial EOF mode, and the right singular vector, $ {V}_{k}^{m}\left({f}_{0}\right) $, is the spectral EOF.

Next, the local fractional variance (LFV) spectrum is explained by the first mode singular value as:

The LFV spectrum provides an efficient parameter for signal detection in the frequency domain, where a peak indicates a potentially significant spatiotemporal signal at that frequency. Statistical significance confidence levels for the LFV spectrum are estimated using the bootstrap method (Mann and Park, 1999).

We can reconstruct the spatiotemporal signals for all timescales of the first mode, if a spatiotemporal signal is at a given frequency, $ {f}_{0} $:

where $ {\mathrm{\sigma }}^{m} $ is the reconstruction factor (i.e., the standard deviation of the time series), and Re means that only the real part is considered. The value $ \mathrm{\delta }\left({f}_{0}\right) $=2 is applied to the interested signals in our analyses for signals outside the secular band. The term $ {\mathrm{\alpha }}_{1}\left(n\Delta t\right) $ is taken as

The variable amplitude $ {\mathrm{\alpha }}_{1}\left(n\Delta t\right) $ represents the slowly varying temporal envelope of the oscillatory signal at frequency $ {f}_{0} $, which can be obtained by constructing a linear combination of Slepian data tapers, $ {a}_{k} $ and the k-th component of the first mode spectral EOF ($ {V}_{k}^{1} $). The term $ {\mathrm{\xi }}_{k}^{-1} $ is the retention factor of the orthogonal data tapers. Such reconstructions show the evolution of amplitude and phase information, corresponding to multi-phase $ \psi \left(t\right)=2\pi {f}_{0}n\Delta {t} $ signals over a cycle. It can describe the changes in signal over an average of time 1/$ {f}_{0} $. Once the time series and spatial fields are reconstructed, the spatial structures can be analyzed according to the explained variance (%) and phase (°) in the significant frequency bands.

In this study, the signal distributions of China precipitation and the North Pacific SST fields (120°E–60°W, 0°–80°N) in the frequency domain are analyzed through the LFV spectrums; from this, two significant interdecadal signals are identified. By reconstructing the combined fields of the precipitation, SST, and other atmospheric circulation fields (SLP, 500-hPa GHT, and 500-hPa wind) at a given interdecadal frequency, the synergistic evolution of their spatial patterns is extracted for each field. The reconstructed temporal and spatial patterns of each field contain information on the other variables. The precipitation field uses gridded data in mainland China, and the contemporaneous global spatial average is subtracted from each grid for each month before the calculations. The atmospheric circulation fields are selected for the region within 90°E–60°W, 0°–80°N.

To identify the dominant precipitation cycles over China and the North Pacific SST from 1951 to 2015, we perform MTM-SVD on the precipitation and SST fields; their LFV spectrums are shown in Fig. 1.

Figure 1.  The LFV spectrums of monthly (a) precipitation (units: mm month^–1) over China and (b) SST (units: °C) in the North Pacific from 1951 to 2015. The dashed lines represent confidence levels corresponding to 90%, 95%, and 99%.

The LFV spectrum of the precipitation in China exhibits significant peaks on the interannual and interdecadal timescales, indicating that the precipitation has multi-timescale variabilities. There is a quasi-biennial peak (significant at the 90% confidence level) at 2.4 years (f=0.41 cycle yr^–1). Within the ENSO band, three significant spectral peaks center at 3.9, 4.7, and 6.3 years (at the 99% confidence level). On the interdecadal timescale, the peaks are centered at 10.7, 13.1, 16.2, and 22.7 years (exceeding the 90% confidence level), consistent with the results of previous studies (Ding et al., 2008; Ren et al., 2012).

Correspondingly, the LFV spectrum of the SST anomaly field also shows significant peaks from the quasi-biennial to interdecadal timescales (Fig. 1b). There are multiple significant peaks in the 2- to 3-year band, indicating a robust quasi-biennial oscillation of the North Pacific SST. Three interannual signals are centered at 3.5, 4.4, and 5.1 years (exceeding the 99% confidence level). On the interdecadal timescale, there are two significant peaks near the quasi-ten-year and quasi-twenty-year (at the 95% confidence level), consistent with the study of Xian and Li (2003).

The precipitation and SST fields have two consistent cycles around the quasi-decadal and quasi-bidecadal bands, which indicates a close relationship between the North Pacific SST variations and the precipitation distribution over China on the interdecadal timescale. This finding motivates us to further analyze the synergistic interdecadal evolution of precipitation and SST. Thus, based on the significant interdecadal signals of the precipitation field, the 11-year (f = 0.094 cycles yr^–1) and 23-year (f = 0.044 cycles yr^–1) cycles are selected to reconstruct the associated combined fields (including precipitation, SST, SLP, 500-hPa GHT, and 500-hPa wind). Since the calculations are performed based on narrowband frequencies, the amplitude described below is correspondingly small. Considering that the precipitation variation over Northwest China is trivial [its standard deviation is less than 1 mm month^–1; Fig. S1 in the electronic supplementary material (ESM)], we focus our attention on the precipitation over eastern China (100°–135°E, 17°–55°N) in the following analysis. The 11-year and the 26-year cycles can be observed for the precipitation field over the domain of 100°–122°E, 20°–45°N (not shown), which means that the results are robust.

Figure 2 shows the reconstructed spatial patterns of the precipitation over China and North Pacific SST for the 11-year cycle signal. The precipitation anomalies exhibit a meridional dipole mode (the SF-ND or SD-NF), with two variability centers in the south and north of 35°N (Figs. 2a–f). The SST fields exhibit two variability centers, whose polarities are usually opposite in the subtropical North Pacific (SNP) and the equatorial Central Pacific (Figs. 2g–l). The distribution of SST variability resembles that of the PDO (Mantua et al., 1997; Zhang and Levitus, 1997), with a pattern correlation coefficient being 0.82 between the reconstructed distribution of North Pacific SST and the canonical PDO pattern regressed from the NCEI PDO index.

Figure 2.  The 10-year mean precipitation distribution [left panel: (a)–(f), units: mm month⁻¹] and SST [right panel: (g)–(l), units: °C] from (a, g) the 1950s to (f, l) the 2000s obtained from the joint reconstruction at the 11-year cycle (f = 0.094 cycle yr^–1).

The synergistic evolution of precipitation and SST reveals an in-phase relationship between the PDO phase and the rainband position. It means that precipitation anomalies tend to increase (decrease) over North (South) China during the PDO warm (cold) phase. For example, in the 1950s, 1960s, and 2000s, precipitation anomalies exhibited the SD-NF pattern accompanied by the PDO warm phase. While during the PDO cold phase in the 1970s (Figs. 2c, i) and 1980s (Figs. 2d, j), the SF-ND pattern dominated eastern China. More precipitation appears to the south and north of 35°N over China, with only a weak warm PDO in the 1990s (Figs. 2e, k). Therefore, the changes in the PDO phase associated with the 11-year cycle correspond to the decadal change in the precipitation distribution over China.

When the SNP is dominated by a warm SST anomaly center on the 11-year timescale, the North Pacific SST distribution corresponds to the PDO cold phase. What are the differences in the precipitation distribution over China during the different stages of the PDO cold phase? To address this question, we describe the synergistic evolution of the precipitation and SST during the first (from phase 0° to phase 60°) and the post 1/4 cycle (from phase 90° to phase 150°) associated with the 11-year variability (Figs. 3 and 4a–f).

Figure 3.  The half-cycle spatiotemporal evolution [from phases (a) 0° to (f) 150°] of the precipitation (units: mm month^–1) obtained from the joint reconstruction at the 11-year cycle (f = 0.094 cycle yr^–1). (g) The phase-latitude section of precipitation (units: mm month^–1) zonally-averaged (110°–122°E) for the full-cycle (phase 0°– 360°) reconstruction.

Figure 4.  The half-cycle spatiotemporal evolution [from phases 0° to 150°] of the SST [left panel: (a)–(f), units: °C], SLP [middle panel: (g)–(l), units: Pa], and 500-hPa GHT [right panel: (m)–(r), units: gpm] obtained from the joint reconstruction at the 11-year cycle (f = 0.094 cycle yr^–1). There is a 0.9-year time difference between each frame. The black solid (dotted) lines in (a–r) mean that the anomalies are greater (smaller) than [0.06°C, 10 Pa, and 2 gpm], which represents the SST, SLP, and GHT anomaly centers.

During the first 1/4 cycle, positive precipitation anomalies occur over the middle and lower reaches of the Yangtze River (MLYR), and negative anomalies mainly appear over South China and North China (Figs. 3a–c). At the same time, warm SST anomalies can be observed from the western Pacific (the maximum exceeding 0.04°C) to the SNP (the maximum exceeding 0.06°C ) at phase 0°. As the SNP warming intensifies from phase 30° to phase 60°, the western Pacific warming weakens (Figs. 4a–c). Meanwhile, cold SST anomalies are first identified from the equatorial central-eastern Pacific to the west coast of North America at phase 0° (Fig. 4a) and then weaken from phase 30° to phase 60° (Figs. 4a–c). Such an SST evolution is closely related to the development stage of the PDO cold phase (Mantua et al., 1997). During the post 1/4 cycle, the North Pacific SST patterns related to the PDO cold phase have transformed into the decaying stage (Figs. 4d–e). As observed, two warming centers in the SNP and the western Pacific decay from phase 90° to phase 120° and turn cold by phase 150°. In the equatorial central-eastern Pacific, cold SST anomalies, which exist before phase 90°, have turned to warm anomalies at phase 120°. Corresponding to the decay stage of the PDO cold phase, positive precipitation anomalies move northwestward to North China from the MLYR, and negative precipitation anomalies increase over South China and expand northward (Figs. 5d–f).

Figure 5.  The half-cycle spatiotemporal evolution [from phases (a) 0° to (f) 150°] of the 500-hPa wind (vector, units: m s^–1) and 500-hPa GHT (shading, units: gpm) in East Asia obtained from the joint reconstruction at the 11-year cycle (f = 0.094 cycle yr^–1). The green lines represent the climatological position of the 500-hPa western Pacific subtropical high (indicated by the 5860-gpm contour) during 1951–2015.

The synergistic evolution associated with the 11-year variability exhibits a reversal in the polarity of the precipitation dipole patterns over China during the cold PDO phase. During the development stage of the PDO cold phase, the precipitation distribution over China is dominated by the SF-ND pattern, with precipitation increasing over the MLYR and decreasing over North China. When stages of the PDO cold phase change from developing to decaying, the dipole pattern over China simultaneously transitions from the SF-ND to SD-NF. The precipitation increases over North China and decreases over South China during the decay stage of the PDO cold phase.

The observed precipitation patterns of covariance with the PDO as represented by the 11-year cycle may result from the atmospheric response associated with the North Pacific SST, including the variations in the Aleutian Low (AL) over the North Pacific and the WPSH and Mongolia high (MH) over East Asia. In this subsection, we describe the evolutionary characteristics of the atmospheric pressure and circulation fields over the North Pacific and East Asia during the development and decay stages of the PDO cold phase for the 11-year cycle to explore possible processes and underlying mechanisms through which the PDO affects the precipitation over China.

Previous studies have pointed out that the AL, which can affect the East Asia climate, is strongly related to the PDO (Latif and Barnett, 1996; Deser et al., 2004). To clarify the changes in the AL associated with the PDO cold phase for the 11-year cycle, we describe the coupled evolutionary characteristics of the SLP and 500-hPa GHT fields over the North Pacific (Figs. 4g–r). During the first 1/4 cycle, positive SLP anomalies occupy the entire North Pacific at phase 0° and then decay from phase 30° to phase 60° (Figs. 4g–i). During the post 1/4 cycle (Figs. 4j–l), the positive SLP anomaly center (>10 Pa) mentioned above decays and moves northeastward; negative SLP anomalies near the Aleutian Islands then occur at phase 90° and before enhancing and expanding southward to the tropical Pacific from phase 90° to phase 150°. Meanwhile, anomalous descending motion over the SNP is enhanced at 500-hPa during the first 1/4 cycle (Figs. 4m–o) before decaying and moving southward during the post 1/4 cycle (Figs. 4p–r). The variations of SLP and GHT mean the AL is weakening (intensifying) during the development (decay) stage of the PDO cold phase.

In summary, the AL co-varies with the North Pacific SST anomalies during the different PDO stages. During the development stage of the PDO cold phase, warm SST anomalies in the SNP are accompanied by a weaker-than-normal AL (Sun et al., 2017). During the decay stage, the AL strengthens since the SNP warming decays. Thus, both the variations of SLP and 500-hPa GHT contribute to the weakening (intensification) of AL during the development (decay) stage of the PDO cold phase.

To clarify the changes in the WPSH and MH during different PDO stages, the evolution of 500-hPa GHT and wind fields over East Asia for the 11-year cycle is described in Fig. 5. To the south of 25°N in the western Pacific, negative GHT anomalies increase during the first 1/4 cycle, decrease from phase 90° to phase 120° (Figs. 4p–q and 5d–e), and turn to positive at phase 150° (Figs. 4r, 5f). As seen from the wind field, the intensity of the accompanied anomalous cyclone over and to the south of 25°N in the western Pacific experienced similar changes. During the development (decay) stage of the PDO cold phase, the changes in anomalous ascending and cyclone indicate that the WPSH is weaker (stronger) than the average. In northern East Asia, negative GHT anomalies over Mongolia weaken and disappear during the first 1/4 cycle, while positive GHT anomalies on its northeastern side move southwestward (Figs. 5a–d). During the post 1/4 cycle, positive GHT anomalies move southwestward to Mongolia and cause the MH to intensify (Figs. 4p–r and 5d–f). Corresponding to the evolution of the height field, an anomalous cyclone over Mongolia weakens from phase 0° to phase 30° and is replaced by an anomalous anticyclone from phase 60° to phase 150°. This pattern demonstrates that the MH weakens (intensifies) during the development (decay) stage of the PDO cold phase. To summarize, the aforementioned evolutionary characteristics of the GHT and wind fields over East Asia demonstrate that the WPSH and MH are weaker (stronger) than the average on the 11-year timescale during the development (decay) stage of the PDO cold phase.

One possible reason for the changes in WPSH and MH is that the East Asian atmospheric circulation is affected by the AL and SST anomalies related to the PDO. The WPSH weakens and retreats eastward during the development stage of the PDO cold phase, which is mainly attributed to the combined effects of the weaker-than-average AL and a warming of the western Pacific SST. On the one hand, a weak AL can induce anomalous ascending motion over the western Pacific south of 25°N (Sun et al., 2017). On the other hand, the western Pacific warming can enhance local convection and weaken the WPSH (Zhang et al., 2016b). Both processes are conducive to the weakening and the eastward retreat of the WPSH during the development stage of the PDO cold phase. When these combined effects weaken during the decay stage, the WPSH begins to intensify and move southward. In contrast to the influence factors of WPSH, the MH is mainly affected by the AL during the development and decay stages of the PDO cold phase. During the development stage, the high-pressure anomalies related to the weak AL expand to northern East Asia and are favorable for inducing ascending motion anomalies on its south side (North China), causing the MH to weaken and move northward (Zhao et al., 2016). During the decay stage, when the high-pressure anomalies over the North Pacific weaken, the MH intensifies and moves southward.

How does an anomalous AL, related to the PDO, modulate the WPSH and MH and further modify the distribution of precipitation over China? Figures 6a and 6b show that the weakening (intensification) of the AL during the development (decay) stage of the PDO cold phase can result in a weakened (intensified) WPSH and MH, which modulates the precipitation over China. During the development stage (Fig. 6a), the weaker-than-normal WPSH and MH can induce two anomalous cyclone centers over the western Pacific and Mongolia. These circulation configurations lead to strong easterly wind anomalies over Southwest China, weak westerly wind anomalies along the MLYR, and weak northeasterly wind anomalies over the north of the MLYR, respectively. The water vapor originating from the tropical Pacific converges over the Yangtze River Basin and diverges over the Shandong Peninsula and South China (Figs. S2a–c in the ESM), contributing to increased precipitation over the MLYR with corresponding decreased precipitation over North China. During the decay stage (Fig. 6b), owing to the effects of the stronger-than-normal WPSH and MH, anomalous anticyclones dominate over the western Pacific and Mongolia, with easterly wind anomalies over North China and northwesterly wind anomalies over South China. The anomalous water vapor divergence covers the central and eastern regions of China (Figs. S2d–f). The precipitation increases over North China when the water vapor convergence center moves northeastward.

Figure 6.  Schematic diagrams showing the PDO effects on precipitation over eastern China for the (a, b) 11-year and (c, d) 23-year variabilities. Red (blue) shadings represent warm (cold) SST anomaly centers, and green shadings represent positive precipitation anomalies; yellow arrows represent the mid-level wind field distribution in key regions of China. Blue (red) lines represent the cyclonic (anticyclonic) circulation caused by the ascending (sinking) motion. The thick black arrows indicate the movement of the circulation centers. The dashed lines represent weakening effects.

Figure 7 shows the reconstructed spatial patterns of the precipitation over China and the North Pacific SST for the 23-year cycle signal. There is a precipitation variability center in the Shandong Peninsula, with anomalies of opposite polarity simultaneously present in the other areas (Figs. 7a–f). In the North Pacific, two SST variability centers with opposite polarities mainly appear in the SNP and along the west coast of North America (Figs. 7g–h), akin to the PDO cold phase (Mantua et al., 1997). The pattern correlation coefficient is 0.83 between the reconstructed North Pacific SST distribution and the canonical PDO pattern regressed from the NCEI PDO index. Overall, an anti-phase relationship between rainband position and PDO phase dominates during most periods for the 23-year cycle. This is evident by the presence of positive precipitation anomalies in southern China during the PDO warm phase and vice versa. For example, when the cold SST anomalies dominated the SNP in the 1950s and 1980s, the precipitation decreased in the Shandong Peninsula but increased in South China and Northeast China. A similar case with an opposite polarity occurs in the 1960s.

Figure 7.  The same as in Fig. 2, but for the 23-year cycle (centered at f = 0.044 cycle yr^–1).

Similar to the analysis of the 11-year cycle, we discuss the differences in the precipitation distribution over China during the different stages of the PDO cold phase for the 23-year cycle. The synergistic evolution of the precipitation and North Pacific SST during the first and post 1/4 cycles is shown in Figs. 8 and 9a–f.

Figure 8.  The same as in Fig. 3 but for the 23-year cycle (centered at f = 0.044 cycle yr^–1). There is a 1.9-year time difference between each frame.

Figure 9.  Same as in Fig. 4 but for the 23-year cycle. The black solid (dotted) line in (a–r) means that the anomalies are greater (smaller) than [0.6°C, 80 Pa, and 15 gpm], which represents the SST, SLP, GHT anomaly centers. There is a 1.9-year time difference between each frame.

During the first 1/4 cycle, more precipitation occurs within 32°–42N° over China with less precipitation on its north and south sides (Fig. 8g). Positive precipitation anomalies mainly concentrate in the Shandong Peninsula with their coverage retreating eastward, and negative precipitation anomalies dominate the MLYR and Northeast China (Figs. 8a–d). Meanwhile, large-scale warm SST anomalies occupy almost the entirety of the North Pacific at phase 0°, with the center (>0.6°C) at around 160°W, 40°N (Fig. 9a). This warm center intensifies and expands to 30°–50°N from phase 30° to phase 60°, while the enhanced cold SST anomalies appear along the west coast of North America (Figs. 9b, c). This SST distribution resembles the situation in the development stage of the PDO cold phase (Mantua et al., 1997). During the post 1/4 cycle, the cold PDO pattern enters the decay stage (Figs. 9d–f). In the North Pacific, warm SST anomalies decay from phase 90° to phase 120°, and their patterns collapse at phase 150°, while the SST cooling along the west coast of North America decays. As the cold PDO pattern changes from developing to decaying, more precipitation occurs over South China, while less precipitation occurs on its north side (Fig. 8). The positive anomalies over the Shandong Peninsula, which exist in the first 1/4 cycle, decrease and turn negative at phase 150°. Over South China, positive precipitation anomalies are first identified at phase 90° and then increase and move northward from phase 120° to phase 150°. Over Northeast China, the negative precipitation anomalies dominate almost the entire half-cycle.

In summary, accompanied by the cold PDO pattern changes, from developing to decaying from phase 0° to phase 150°, the positive precipitation anomalies first appear in the Shandong Peninsula and then move to South China. During the first 1/4 cycle, the precipitation increases over the Shandong Peninsula are accompanied by the SNP warming. When the cold PDO pattern changes from developing to decaying during the post 1/4 cycle, the rainband position has also shifted from the Shandong Peninsula to South China. Thus, more precipitation occurs over North China (South China) during the development (decay) stage of the PDO cold phase.

Similar to the analysis of the 11-year cycle, the processes and mechanisms by which the PDO forcing affects the precipitation over China for the 23-year cycle are demonstrated in this subsection.

To investigate the atmospheric responses to the North Pacific SST forcing, we focus on the evolution of SLP and 500-hPa GHT fields over the North Pacific (Figs. 9g–l and m–r). During the first 1/4 cycle, a positive SLP anomaly center (>80 Pa) is first identified near the Aleutian Islands at phase 0° and then expands southward (Figs. 9g–i). During the post 1/4 cycle (Figs. 9j–l), the anomalous high-pressure center decays and moves southward from phase 90° to phase 120°. Negative SLP anomalies only appear near the Aleutian Islands at phase 150°. Corresponding to the high-pressure center, an anomalous descending motion center (>15 gpm), with decreased intensity, can be observed over the SNP in the 500-hPa GHT field throughout the half-cycle (Figs. 9m–r). The evolutionary characteristics of SLP and GHT fields both demonstrate that the AL is weaker than normal throughout the entire half-cycle. This weaker-than-normal AL is observed during the PDO cold phase, which can induce the anomalous descending motion over North Pacific (Sun et al., 2017). When the SNP warming intensifies (weakens) during the development (decay) stage of the PDO cold phase, the anomalous descending motion associated with the weaker-than-normal AL strengthens (weakens) simultaneously.

Next, we compared the characteristic differences in the WPSH and MH associated with the development and decay stages of the PDO cold phase by describing the evolutionary features of the 500-hPa GHT and wind anomalies over East Asia (Fig. 10). To the south of 35°N in the western Pacific, negative GHT anomalies increase from phase 0° to phase 60° (Figs. 10a–c), then decrease from phase 90° to phase 120° (Figs. 10d–e) and turn positive at phase 150° (Fig. 10f). Accompanied with ascending motion, anomalous cyclones can be observed between 15°–35°N from phase 0° to phase 60°, with such patterns collapsing from phase 90° to phase 150°. This means that the WPSH is weaker than normal during the first 1/4 cycle and tends to intensify during the post-1/4 cycle. In northern East Asia, negative GHT anomalies dominate from phase 0° to phase 150°, with their center moving eastward to Japan from east of Lake Baikal (Figs. 9p–r). At the same time, positive GHT anomalies are first identified over Northeast China at phase 0° and then retreat to the east of Japan from phase 30° to phase 120°. The circulation field exhibits anomalous cyclones over and to the east of Lake Baikal that move eastward to Japan, while anomalous anticyclones weaken and retreat to the east of Japan. Thus, the MH is weaker than normal for the entire half-cycle. These results indicate that the MH and WPSH are weaker than normal during the development stage of the cold PDO phase, while the WPSH tends to intensify with weakened MH during the decay stage.

Figure 10.  The same as in Fig. 5 but for the 23-year cycle (centered at f = 0.044 cycle yr^–1). There is a 1.9-year time difference between each frame.

The differences in the MH and WPSH between the development and decay stages of the cold PDO phase result from the effects of the AL variation. During the development stage of the PDO cold phase, anomalous high pressure over the North Pacific related to a weak AL is not only conducive to the weakening and eastward retreat of the WPSH but also results in a weak MH. Two main atmospheric processes by which the AL affects the MH and WPSH may explain the above phenomenon. On the one hand, there is a “see-saw” synchronous change between the MH and the AL (i.e., the AL strengthens while the MH also strengthens, and vice versa) on the interdecadal timescale (Yang et al., 2004). On the other hand, the anomalous high pressure over the North Pacific is conducive to weakening the WPSH by inducing enhanced ascending motions over and to the south of 35°N in the western North Pacific (Sun et al., 2017). During the decay stage, the anomalous high pressure over the North Pacific persists and moves southward into the western Pacific. Thus, the effect of “see-saw” synchronous change can persist into the decay stage of the PDO cold phase, which explains why the MH is still weak during this period. In addition, western Pacific cooling can suppress local convection and contribute to weakening the WPSH (Zhang et al., 2016b). The combined effects of the anomalous high pressure and western Pacific cooling may result in the intensification and southward migration of WPSH during the decay stage of the PDO cold phase.

Figures 6c and 6d further show the processes by which the AL affects the MH and WPSH, which can result in anomalous precipitation by altering the 500-hPa wind field and moisture transport (Fig. S3 in the ESM) during the development and decay stages of the PDO cold phase for the 23-year cycle. The comparison of moisture flux divergence and precipitation demonstrates that their overall evolutionary features (intensity and main domain) are relatively consistent in most regions (including North China, Northeast China, South China, and Southwest China), except in a few cases. It indicates that moisture conditions are a vital factor in modifying the local precipitation during the PDO cold phase. During the development stage of the PDO cold phase (Fig. 6c), the MH and WPSH are weaker than normal, with two accompanied anomalous cyclones over northern East Asia and in the western North Pacific to the south of 35°N. The corresponding circulations result in southeasterly wind anomalies over Northeast China, southwesterly wind anomalies over Northwest China, and northeasterly wind anomalies over and to the south of 35°N, respectively. Accompanied by these wind anomalies, the water vapor converges over Southwest China and North China but diverges over South China (Figs. S3a–c). As a result, precipitation increases over North China and decreases over and to the south of 35°N in China. During the decay stage of the PDO cold phase (Fig. 6d), the WPSH tends to intensify with a weakened MH. The anomalous cyclone over Mongolia can cause an enhancement in westerly wind anomalies, moving southward to the MLYR, while weak southerly wind anomalies exist over South China, associated with an anomalous anticyclone. Northwesterly wind anomalies over North China are enhanced and move eastward, preventing northward water vapor transport. Regions with water vapor convergence over South China expand, while those with water vapor convergence over North China shrink (Figs. S3d–f), causing precipitation to increase along with the low-level convergence along and to the south of the MLYR.

Figure 11 shows the signals reconstructed for 1) the regionally averaged precipitation over North China, the MLYR, and South China and 2) for the SST at 180°, 37°N in the North Pacific. For the precipitation over different regions, the phase of each specific cycle signal is rearranged, resulting in contemporaneous precipitation variations. Due to the difference in the signal cycle, the 11-year cycle signal can undergo multiple phase transitions during one cycle length of the 23-year cycle signal. For example, from the 1990s to the 2000s in South China (Fig. 11c), the 23-year cycle signal was in a negative phase, while the 11-year cycle signal underwent about four phase shifts. Thus, the composite signals of precipitation are quite distinctive from the two separate single signals. Similar features of reconstituted time signals also exist in the North Pacific SST (Fig. 11d). These features indicate that the relative contributions of the 11-year and 23-year variabilities to the contemporaneous precipitation are different.

Figure 11.  Reconstructed signals of the representative grid boxes for the normalized precipitation (unit: mm month^–1) over (a) North China (110°E, 37°N), (b) the middle and lower reaches of the Yangtze River (MLYR, 115°E, 30°N), and (c) South China (113°E, 22°N), and for the normalized SST (units: °C) in (d) the subtropical North Pacific (SNP; 180°, 37°N). Shading indicates the composite signals of the 11-year (black line) and 23-year (green line) cycles.

The relative contributions of the 11-year and 23-year variabilities to precipitation are evaluated through the 1998 precipitation case, which led to major flooding in the MLYR, along the Songhua River in Northeast China, and in other basins due to the excessive rainfall in the rainy season. On the interannual timescale, previous studies have emphasized the impacts of preceding winter ENSO persisting to the following summer, which resulted in the 1998 MLYR flooding event (Li et al., 2001; Samel and Liang, 2003). Furthermore, on the interdecadal timescale, precipitation increases over the MLYR during the summer months of 1998 are related to the PDO (Fig. 11b). A peak of the composite MLYR precipitation signal is seen in the 1998 summer. Accompanied by the SNP warming, the summer of 1998 was characterized by positive precipitation anomalies over eastern China, negative anomalies in northern China, and weak positive precipitation anomalies over eastern Northeast China.

The 11-year and 23-year cycles may provide different interdecadal backgrounds upon which the summer precipitation in 1998 is rather unique. For the 11-year cycle, positive anomalies dominate eastern China, and negative anomalies are centered in northern China (Fig. 12a). For the 23-year cycle, positive anomalies are centered over and to the north and northeast of the MLYR, whereas negative anomalies dominate South China (Fig. 12b). Simultaneously, SST anomaly patterns vary with the two interdecadal PDO modes as seen in the 1998 summer, with SNP warming and tropical cooling attributed to the 11-year cycle and large-scale warming attributed to the 23-year cycle, respectively. By modifying the wind field and water vapor transport, the PDO and its associated atmospheric field contributed differently to the 1998 summer precipitation between two interdecadal modes. For the 11-year cycle, southeastern and northeastern wind anomalies over eastern China with water vapor converge over eastern China but diverge over North China (Fig. S4a in the ESM). For the 23-year cycle, an anomalous anticyclone over North China with water vapor convergence over and to the north of the MLYR and strong water vapor divergence over Poyang Lake and along its southern side (Fig. S4b). The water vapor conditions of the 11-year cycle may be more favorable for the precipitation increases over the Yangtze River Basin than the 23-year cycle. In Northeast China, water vapor convergence is observed for both the 11-year and 23-year cycles. The convergence center of the latter is further eastward than that of the former and is more consistent with the precipitation center. As a result, the combined effects of the 11-year and 23-year cycles provide a favorable interdecadal background for the 1998 summer precipitation. Their relative contributions to contemporaneous precipitation vary with the dominant interdecadal modes. The 11-year variability has a relatively large contribution to the excess precipitation over the MLYR, whereas the 23-year variability may contribute more to the contemporaneous precipitation increase over Northeast China.

Figure 12.  Distributions of the reconstructed signals in the summer (June–July–August) of 1998: (a–c) precipitation over China (shading, units: mm month⁻¹) and the 500-hPa wind (vector, units: m s⁻¹) over East Asia, and (d–f) the SST (shading, units:°C) and SLP (contour, units: Pa). (a, d) and (b, e) represent the fields reconstructed at the 11-year and 23-year cycles, respectively; panels (c, f) are the composite fields for the 11-year and 23-year cycles. The contour interval is 15 Pa for the SLP in (d), (e), and (f).

We employ MTM-SVD analysis to investigate the synergistic interdecadal evolution of the precipitation distribution over China and the SST in the North Pacific from 1951 to 2015. The main conclusions are as follows.

Both the precipitation and SST fields show two significant interdecadal signals with 11-year and 23-year cycles. These two dominant interdecadal modes are related to the PDO. The synergistic evolution shows that the PDO can modulate the interdecadal distribution of precipitation over China during different periods. For the 11-year cycle, precipitation anomalies tend to increase (decrease) over North (South) China during the PDO warm (cold) phase. For the 23-year cycle, there are positive precipitation anomalies in southern China during the PDO warm phase and vice versa. For the two cycles, the variations of the AL are essential to the processes by which the PDO affects the WPSH and MH and further influences the precipitation distribution over China.

For the 11-year cycle, a weakened AL can weaken the WPSH and MH during the development stage of the PDO cold phase. Accompanied by this phase, two anomalous mid-level cyclones induce the westerly and easterly wind anomalies along 30°N over China, resulting in wind convergence and increased precipitation over the MLYR. During the decay stage of the PDO cold phase, the intensified WPSH and MH working in association with the enhanced AL can weaken the westerly wind anomalies and promote the northward movement of southeasterly wind anomalies, resulting in increased precipitation, which tends to more northward to North China. For the 23-year cycle, during the development stage of the PDO cold phase, the weakened WPSH and MH, in association with decreased AL, can cause southwesterly wind anomalies over Northwest China, southeasterly wind anomalies over Northeast China, and northeasterly wind anomalies over South China. During the decay stage of the PDO cold phase, the combined effects of the intensified WPSH and weakened MH cause northwesterly wind anomalies to intensify and move eastward. The eastward expansion of northwesterly wind anomalies prevents northward water vapor transport and causes low-level convergence and precipitation south of the MLYR. The peaks of the precipitation and SST phasings are not consistent within the quasi-decadal and quasi-bidecadal bands. For the 23-year cycle, there is a 3-year difference in their peaks. More exploration of this 3-year difference is required, but its analysis and explanation are beyond the scope of this study since the lead-lag relationship between the precipitation and SST can not be determined from their dominant cycle differences and evolution with the same cycles.

Additionally, the 11-year and 23-year cycles jointly regulate the interdecadal distribution of precipitation over China. For example, in the 1998 flooding case, the 11-year cycle contributes significantly to the excess precipitation over the MLYR, whereas the 23-year cycle may contribute more to the contemporaneous precipitation increase over Northeast China.

Decadal-to-interdecadal variabilities are much more complicated because they can involve processes in other ocean basins with longer timescales, including those from the Atlantic and Indian Oceans (e. g., Feng et al. 2021). For instance, the anomalous SST tripole pattern over the North Atlantic can also modulate the precipitation patterns over China on an interdecadal timescale (Wu and Wu, 2019). The North Atlantic Oscillation-related SST tripole pattern could excite a downstream propagating Rossby wave train in the atmosphere to modulate East Asian summer monsoon variations (Li and Ruan, 2018; Li et al., 2019) and further influence the precipitation. The possible links between the North Atlantic Ocean and East Asia need to be explored. In addition, the responses of precipitation to the effects of tropical SST and the related processes cannot be ignored. Particularly, the PDO is closely related to the tropical SST on decadal and interdecadal timescales. The amplitudes of SST anomalies in the tropic and subtropic Pacific reconstructed at the 11-year cycle are similar. We further examined possible variations in tropical SST-induced convective activity on decadal and interdecadal timescales. For the 11-year cycle, the cold SST anomalies in the central tropic Pacific can induce descending motions over the dateline, while warm SST anomalies induce ascending motions over the western Pacific. For the 23-year cycle, similar features can be seen between the SST and vertical velocity. Tropical SST associated with the PDO can induce convective activity, modulating the WPSH and AL via a large-scale divergent circulation (Jo et al., 2015; Hu et al., 2022b). Moreover, solar activity, which has dominantly an 11-year cycle variation, can be closely linked to the Northeast Pacific and Tropical Pacific SSTs (e.g., Lin et al., 2021). This implies that there may be a response of precipitation to solar activity on the 11-year timescale. In this study, we focused on the response of precipitation to SST related to the PDO mainly because the North Pacific SST reconstructed at the 11-year cycle generally exhibits a PDO-like mode. We only qualitatively analyze the processes and mechanisms of the North Pacific SST forcing that affect precipitation over China. In the future, model sensitivity experiments should be conducted to further clarify the relevant dynamic processes and physical mechanisms (Hu et al., 2019; Zhang et al., 2020, 2022). We are hopeful that the relationships between the PDO and precipitation examined by this study can be used for decadal predictions of the PDO and corresponding precipitation over China.

Acknowledgements. The authors thank the reviewers for their precious and insightful comments, which greatly helped improve our manuscript. This work is supported by the National Natural Science Foundation of China (Grant No. 42030410), Laoshan Laboratory (No. LSKJ202202403-2), the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB40000000), and the Startup Foundation for Introducing Talent of NUIST.

Electronic supplementary material: Supplementary material is available in the online version of this article at https://doi.org/10.1007/s00376-023-3011-z.