Influence of Internal Decadal Variability on the Summer Rainfall in Eastern China as Simulated by CCSM4

Recent studies on the global warming hiatus since the beginning of the 21st century note the important contribution of internal decadal variability on global climate changes (Kosaka and Xie, 2013; Meehl et al., 2013; Trenberth and Fasullo, 2013; Dai et al., 2015). The Pacific Decadal Oscillation (PDO) (e.g., Trenberth and Hurrell, 1994; Mantua et al., 1997) and Atlantic Multidecadal Oscillation (AMO) (e.g., Enfield et al., 2001) are the two dominant modes at the decadal timescale. These oscillations can exert significant influences on regional climate in Europe, Asia, and North America (e.g., Barlow et al., 2001; Yeh and Kirtman, 2004; Knight et al., 2006; Zhang and Delworth, 2006; Ma, 2007; Sun et al., 2008; Chylek et al., 2014; Qian and Zhou, 2014).

Interdecadal climate change is an important feature of the East Asian climate. Two prominent interdecadal changes occurred in the eastern China summer rainfall in the late 1970s (Wang, 2001; Ding et al., 2008; Wang et al., 2013) and the late 1990s (Zhu et al., 2011, 2015; Si and Ding, 2013). After the late 1970s, summer rainfall over the lower reaches of the Yangtze River valley (YR) increased significantly, while it decreased over North China. After the late 1990s, summer rainfall decreased in the lower reaches of the YR, while it increased in the Huanghe River-Huaihe River valleys. Previous studies proposed either the PDO or AMO as the major modulator for these two interdecadal changes in the eastern China summer climate (e.g., Zhu and Yang, 2003; Ma, 2007; Xu and Zhu, 2010; Zhu et al., 2011; Li and Luo, 2013; Qian and Zhou, 2014; Hao et al., 2015; Yu et al., 2015; Zhu et al., 2015).

However, the PDO and AMO are not isolated; rather, they interact with each other (Wu et al., 2011). Our decadal-scale climate is not only modulated by either the PDO or AMO, but also by a combination of them, with distinct phase combinations in different periods. In the present study, we used a control multi-century simulation from CCSM4 to investigate the modulation impacts of the PDO and AMO on the summer rainfall and its associated circulation in eastern China from intrinsic variability only. The complex external forcing agents were not considered. In section 2, the model and numerical experiment are introduced. Section 3 summarizes the results, and section 4 concludes the findings and presents further discussion.

The output data of the control simulation by CCSM4 were analyzed in this study. CCSM4 is a global coupled climate model with a 1^°, 26-level atmosphere coupled to a 1^° (down to 1/48 in the equatorial tropics), 60-level ocean and state-of-the-art sea-ice and land surface schemes (Gent et al., 2011). The 501-year control simulation was conducted with no interannual variations in the external forcing agents and with greenhouse gases and the tropospheric sulfate aerosol concentration fixed at pre-industrial (1850) levels. Therefore, only the internal variability of the climate system was manifested and no long-term trend existed in this simulation.

The atmospheric variables used included precipitation, geopotential height, and wind field. The SST was used to calculate the PDO and AMO indices. The PDO index was defined as the leading principal component of annual SST anomalies in the North Pacific (20^°-60^°N, 110^°E-110^°W). The AMO index was defined as the average annual SST in the North Atlantic (0^°-60^°N, 75^°-7.5^°W). The 11-point running mean was used on the indices to highlight their decadal signals. The observed AMO and PDO indices were obtained from http://www.esrl.noaa.gov/psd/data/timeseries/ AMO/ and http://research.jisao.washington.edu/pdo/, respectively.

Figure 1.  The PDO and AMO indices in the 501-year control experiment.

Figure 2.  The composite summer precipitation rate (units: kg m$^-2$ s$^-1$) between different sets of the PDO and AMO combinations and the entire period. Dotted areas show the significant difference at the 90% confidence level.

Figure 3.  The standardized observed AMO (red) and PDO (blue) indices during 1948-2014.

Figure 4.  The composite SLP (units: Pa) between different sets of PDO and AMO combinations and the entire period. Dotted areas show the significant difference at the 90% confidence level.

Figure 5.  As in Fig. 4 but for the 850 hPa wind field (vectors) and geopotential height (color shading). The maximum zonal wind in (g) is 0.65 m s$^-1$. Dotted areas show the significant geopotential height difference at the 90% confidence level. The missing values are shown in white.

Figure 6.  As in Fig. 4 but for the 200 hPa level.The maximum zonal wind in (g) is 1.56 m$^-1$

Figure 7.  As in Fig. 4 but for the latitude-geopotential height cross-section of mean vertical velocity over 110$^\circ$-120$^\circ$E. Shading shows the significant difference in vertical velocity at the 90% confidence level. The contour interval is 10$^-3$ Pa s$^-1$. Red (blue) contours represent descending (ascending) anomalies. Black contours show the contours of zero.

Figure 8.  The running correlation coefficient between the PDO and AMO in the first, second, third, fourth and fifth 100 years (a-e) and the whole period of the control simulation. Dashed lines denote the values at the 95% confidence level.

The simulated PDO and AMO indices are shown in Fig. 1. To identify the connections between the PDO/AMO and summer rainfall in eastern China, we first divided the 501 model years into eight different sets: + PDO (i.e., PDO index >0); - PDO (i.e., PDO index <0); + AMO (i.e., AMO index >0); - AMO (i.e., AMO index <0); + PDO+ AMO; + PDO- AMO; - PDO+ AMO; and - PDO- AMO. There were 297, 204, 273, 228, 169, 128, 104, and 100 model years for the eight sets, respectively. The composite analysis was conducted by subtracting the long-term mean of the whole period from the average of each set.

The + PDO (Fig. 2a) and - PDO (Fig. 2b) can induce more and less rainfall over the lower reaches of the Yangtze River-Huaihe River valleys (centered over the Huaihe River valley), respectively, though the + PDO signals are not statistically significant. In contrast, the + AMO (Fig. 2c) and - AMO (Fig. 2d) are related to decreased and increased rainfall centered over the lower reaches of the YR. It seems that the - AMO (Fig. 2d) can induce significantly increased Mei-yu rainfall over eastern Asia. The anomalies are weak and insignificant in eastern China in the + PDO+ AMO set (Fig. 2e). However, significant increased rainfall is evident over the western subtropical Pacific, which is a mutual reinforcing effect between the + PDO (Fig. 2a) and + AMO (Fig. 2c). Differently, in the + PDO- AMO set, these two decadal signals reinforce each other over eastern China (Figs. 2a, d and f), leading to positive rainfall anomalies over the lower reaches of the YR, centered near (30^°N, 120^°E). The - PDO and + AMO can strengthen each other and induce a band of significant negative rainfall anomalies covering the YR and southern Japan (Figs. 2b, c and g). The - PDO and - AMO reinforce each other in the Pacific Ocean, with a negative-positive-negative-positive pattern from the western tropical to subtropical Pacific (not shown). However, their effects counteract each other and no significant signals occur over eastern China (Figs. 2b, d and h).

The connection between the combined PDO and AMO and summer rainfall in eastern China can be observed in the two recent interdecadal changes in the late 1970s and the late 1990s. Accompanying + PDO- AMO (Fig. 3), summer rainfall increases over the YR and decreases over North China after the late 1970s (e.g., Wang, 2001; Ding et al., 2008; Wang et al., 2013), while summer rainfall decreases over the YR and increases over the Huanghe River-Huaihe River valleys when accompanying + PDO- AMO after the late 1990s (e.g., Zhu et al., 2011; Si and Ding, 2013; Zhu et al., 2015). This observed consistency between the interdecadal changes of summer precipitation and the phases of the PDO and AMO are consistent with the model results.

The + PDO corresponds to decreased SLP centered over East Asia and the adjacent ocean (Fig. 4a), while the -PDO corresponds to positive SLP anomalies there (Fig. 4b). In contrast, the + AMO is related to positive SLP anomalies over East Asia and opposite anomalies over the ocean to the south, showing a strengthened zonal land-sea thermal contrast (Fig. 4c), while the - AMO is related to negative SLP anomalies over East Asia and opposite anomalies over the ocean to the south, showing a weakened zonal land-sea thermal contrast (Fig. 4d). However, the impact of a single PDO or AMO on SLP in East Asia is weak. The - PDO+ AMO set is connected with strong positive SLP centered over East Asia, and thus the land-sea thermal contrast there is significantly strengthened (Fig. 4g). In the other three sets, the land-sea thermal contrast becomes somewhat reduced, though the signals are relatively weak.

In the lower-level wind and geopotential height field, the effect of the PDO and AMO in their opposite phases can enhance each other over eastern China (Fig. 5), as in the summer rainfall field. For example, both the - PDO and + AMO (Figs. 5b and c) are related to positive height anomalies over East Asia and negative height anomalies over the Okhotsk Sea, which resembles an East Asia-Pacific (EAP) teleconnection pattern. This combination of - PDO+ AMO exhibits stronger significant anomalies over East Asia and the Okhotsk Sea (Fig. 5g). Southerly anomalies occur over eastern China north of 30^°N, showing a strengthened East Asian summer monsoon. Compared with Fig. 5g, opposite changes occur in the + PDO- AMO composite (Fig. 5f), but the amplitude and area with significant values are smaller. The anomalies in the monsoon circulation are consistent with the SLP anomalies reflecting the land-sea thermal contrast.

In addition, we also studied the upper-level circulation over a larger area to examine the related large-scale features. The anomalies are stronger and better organized than those in the lower level. The PDO can induce significant anomalies over the subtropical North Pacific, close to the Kuroshio extension region (Figs. 6a and b). Significant anomalies also appear over western and eastern Eurasia, forming a wave-train pattern in the northern high latitudes. The most prominent anomalies induced by the AMO locate over the high latitudes of North America (Figs. 6c and d). Besides, significant anomalies also occur over East Asia in a small area. The effect of the + PDO and + AMO can reinforce each other in the low latitudes and northwestern Eurasia, but counteract each other over East Asia (Fig. 6e). The -PDO+AMO set is related to the wave-train pattern in the northern high latitudes (Fig. 6g), as well as significant cyclonic anomalies over East Asia and the North Pacific. The - PDO+ AMO combination also has the strongest anomalies (Fig. 6g). Significant positive geopotential height anomalies occur over Europe and North America, implying the presence there of significant climate anomalies. These significant large-scale anomalies imply that the PDO and AMO can not only influence the summer rainfall in eastern China, but also have an impact on climate in other regions, such as India and West Africa (Li et al., 2012). The strong anticyclonic anomalies over East Asia, which are associated with the EAP-like pattern, can induce significant easterly anomalies related to the southern branch of these anticyclonic anomalies. Therefore, the East Asian westerly jet stream (EAWJS) is weakened and moves northward.

Moreover, the weakened EAWJS is related to the decreased rainfall over the YR (Fig. 2g) by suppressing the jet-related secondary circulation (Ding, 2008). A weakened EAWJS is connected to a weaker jet-related secondary circulation, and the ascending branch centered over the YR is reduced, thereby decreasing the rainfall. Figure 7 shows the vertical velocity: robust signals occur in the + PDO- AMO (Fig. 7f) and - PDO+ AMO (Fig. 7g) sets. Both the - PDO and + AMO can induce somewhat descending anomalies over the YR (Figs. 7b and c). However, their combination is related to much stronger descending anomalies over the YR than their individual counterparts (Fig. 7g). Therefore, the rainfall over the YR is decreased. Conversely, for the + PDO- AMO set, significant ascending anomalies occur over the YR (Fig. 7f), and the rainfall there increases.

Using the pre-industrial control simulation of CCSM4, we analyzed the impact of different combinations of PDO and AMO phases on summer rainfall in eastern China. The results indicated that combinations of the PDO and AMO in opposite phases have the most robust impact on eastern China summer rainfall and related circulations: - PDO+ AMO (+ PDO- AMO) is related with decreased (increased) rainfall over the YR; the land-sea thermal contrast is somewhat stronger; the East Asian summer monsoon is strengthened (weakened); the EAWJS is weaker (stronger); and ascending motion related with the EAWJS over YR is inhibited (enhanced). In our recent decadal climate changes, the situation was consistent with that simulated in CCSM4: + PDO- AMO corresponded to more rainfall over the YR from the late 1970s to the late 1990s, while - PDO+ AMO corresponded to less rainfall over the YR after the late 1990s. However, the decreased rainfall in North China after the late 1970s, as well as the increased rainfall in the lower reaches of the Huanghe River-Huaihe River valleys after the late 1990s, were not presented in CCSM4. One of the possible causes may be the model's insufficiency in simulating the East Asian climate, which is a common disadvantage for state-of-the-art climate models. Another may be that the impact of anthropogenic forcing, to which the interdecadal changes in East Asian summer rainfall could be partly attributed, is not represented in the pre-industrial simulation of CCSM used in this study.

Although combinations of the PDO and AMO are good indicators for summer rainfall variation in eastern China at the decadal time scale, correlations between them were inconsistent in CCSM4. For the 501-year time series used in this study, their relationship varied (Fig. 8): sometimes positive, sometimes negative, sometimes PDO leading the AMO, and sometimes the other way round. Although (Wu et al., 2011) revealed that, based on observational data for the period 1870-2002, significant correlations show up when the PDO leads the AMO by 1 year and when the AMO leads the PDO by 11-12 years, when based on our longer model output dataset the correlations are probably unstable. In addition, to predict either the PDO or AMO is difficult for numerical and statistical models at present. Therefore, to predict the combinations of the PDO and AMO and decadal changes of summer rainfall in eastern China are more challenging issues.