2. University of Chinese Academy of Sciences, College of Resources and Environment, Beijing 100049
Adaptation and mitigation strategies for climate change in the near future require decadal forecasts of both the natural variability of climate and the climate system response to additional greenhouse gases and aerosol forcing (WCRP, 2009). Currently, there is little scientific understanding as to whether the climate can be predicted by up to a few decades (Meehl et al., 2009). For example, most uninitialized free-running simulations for CMIP5 (Coupled Model Intercomparison Project Phase 5) do not reproduce the hiatus of the early 2000s well, when forced by specified historical forcing up to and including 2005 and RCP4.5 (+4.5 W m–2 Representative Concentration Pathway) in 2005–12 (Meehl et al., 2014). Such challenges have led to the study of the decadal variability of the climate system becoming a research hotspot in recent years.
Professor Shaowu Wang was one of the pioneers in studies of the decadal variability of climate in China. In 1962, he investigated the interannual to decadal variation of atmospheric centers of action (ACAs) within global circulation and found that there were evident decadal oscillations for most ACAs, with cycles of quasi-22 and 35 yr (Wang, 1962a). Specifically, the quasi-35-yr oscillation was more evident for ACAs over eastern Asia, including the Siberian high, Aleutian low, and Equatorial low in wintertime, and the Pacific high and Indian low in summertime. The quasi-35-yr oscillations of the Siberian high, Aleutian low, and Equatorial low were highly correlated with the decadal variation of winter temperature, and the oscillations of the Pacific high and Indian low were highly correlated with the decadal variation of summer rainfall over most of China (Wang, 1962b,c). Following that initial work, he and his colleagues published a series of papers (Wang, 1963,1964a,b,1965;Wang et al., 1963) on the decadal variation of atmospheric circulation and the anomalous climate in China, including studies on the possible linkage between decadal oscillations of atmospheric circulation with cooling around the 1960s and the variation of summer rainfall with a spatial pattern of flooding in the south and drought in the north over eastern China starting from the early 1970s (Wang, 1973). Based on a 500-yr dataset of the grade of dryness/wetness derived from Chinese historical documents, he found that the quasi-35-yr oscillation was more pronounced for the dryness/wetness change over eastern China (Wang and Zhao, 1979a). He classified the spatial patterns of drought and flood into six categories [(i) floods over all of eastern China, especially in the Yangtze River valley; (ii) floods in the Yangtze River valley along with droughts in the north and the south; (iii) floods in the south along with droughts in the north; (iv) droughts in the Yangtze River valley along with floods in the north and the south; (v) floods in the north along with droughts in the south; and (vi) droughts over all of eastern China] and investigated the change in frequency of each category of drought and flood distribution (Wang and Zhao, 1979b). These studies laid the foundations for many subsequent studies focusing on the decadal variability of precipitation in China.
Many recent studies have confirmed that precipitation over eastern China experienced decadal variation during the last 50 years (Ding et al., 2008;The Council of the Second National Assessment Report on Climate Change, 2011;Zhang, 2015). In the 1960s and 1970s, relatively more precipitation occurred over the North China Plain (NCP) and southern China (SC) respectively, while less precipitation occurred over the middle to lower reaches of the Yangtze River (YZR). Then, a decadal shift in the precipitation pattern occurred in 1977–79, and precipitation increased over the YZR and decreased over the NCP and SC afterwards (Hu, 1997;Huang et al., 1999,2011). Around 1992/93, another decadal change in precipitation was detected, with a dipole pattern involving an increase over the south of the YZR and a decrease over the north (Zhu et al., 2011;Wang et al., 2016). Finally, a further decadal shift occurred in 1999/2000, in which precipitation increased over the area between the Yellow River and the Huaihe River (HHR, southern part of the NCP) and SC, but decreased over the YZR and the area north of the Yellow River (i.e., the northern part of the NCP) (Xu et al., 2015;Zhu et al., 2015). Additionally, there might also be one more decadal shift in the precipitation pattern over eastern China, since the summer of 2012 (Gong et al., 2013). It has also been found that these decadal shifts in precipitation are significantly associated with changes in global- or large-scale circulation (e.g.,Hu, 1997;Zhang et al., 2007;Ding et al., 2009;Zhu et al., 2011;Huang et al., 2011,2013), especially the shift of the Pacific Decadal Oscillation (PDO) (e.g.,Zhu and Yang, 2003;Chan and Zhou, 2005;Zhou et al., 2007;Gong et al., 2013;Qian and Zhou, 2014;Xu et al., 2015;Zhu et al., 2015). During 1961–78, a cool phase of the PDO, a stronger East Asian summer monsoon, a weaker and more eastward western Pacific subtropical high (WPSH), and weaker cold-air activities in the north, were all beneficial to the northward transport of water vapor from the south, thus resulting in more summer precipitation over the NCP. From 1979 to the mid-2000s, the PDO was largely in its warm phase; however, during 1979–92, the sea surface temperature (SST) over the western Pacific was below normal and the WPSH was stronger and westward-extended, which led to more summer precipitation over the YZR and less summer precipitation over the NCP and SC. After 1993, the SST over the western Pacific was above normal, which resulted in a spatial pattern of more precipitation over SC and the HHR but less precipitation over the YZR and NCP, with a belt of increased rainfall dominant over the HHR. In the mid-2000s, the PDO changed back to its cool phase (the SST over the western Pacific in particular has been below normal again since 2010), which might have led to the increased precipitation over the NCP and SC but decreased precipitation over the YZR (Zhou and Xia, 2012;Gong et al., 2013;Huang et al., 2013;Xu et al., 2015;Zhu et al., 2015). However, all of these studies were mainly based on observational data limited to within the last 60 years. Over longer timescales, however, the stability of these observation-based findings is unclear. Accordingly, in this paper, we investigate the decadal variability of summer precipitation over eastern China using long-term reconstructed and simulated precipitation data.2 Data and methods 2.1 Data
Three kinds of datasets are used in this study: monthly precipitation observations; reconstructed and proxy precipitation derived from Chinese historical documents; and the control simulation of the Community Earth System Model (CESM) with fixed pre-industrial external forcing.
(1) Monthly precipitation observations. This dataset, strictly quality-controlled, is from the China Meteorological Administration ( http://data.cma.cn), and includes data from 403 weather stations over eastern China (east of 105°E) for the period 1951–2014.
(2) Reconstructed and proxy precipitation. The first reconstruction used is the Meiyu (rainy season dominated by the Asian summer monsoon) precipitation over the middle and lower reaches of the YZR (30°–33°N and east of 110°N, approximately) (Ge et al., 2008) and the JJA (June–August) precipitation over the middle and lower reaches of the Yellow River (34°–39°N and east of 108°N, approximately) (Zheng et al., 2005) for the period 1736–2000. Both of these datasets were produced by using Chinese historical archives of quantified rainfall and snowfall since 1736. The correlation coefficient between Meiyu precipitation and observed precipitation since 1951 is 0.876, and the variance explained can reach up to 76.7% (Ge et al., 2008). The dataset of JJA precipitation over the middle and lower reaches of the Yellow River has been verified by field experiments, which showed a correlation coefficient of 0.9325 and explained variance of 87% (Zheng et al., 2005). The second reconstruction used is the regional dry–wet index (DWI) series over the NCP (34°–40°N and east of 105°N, approximately), the Jiang–Huai (JH) area (31°–34°N and east of 110°N, approximately), and the Jiang–Nan (JN) area (25°–31°N and east of 110°N, approximately) during A.D. 500–2000. These data were reconstructed from the grade of drought/flooding derived from descriptions of drought and flood disasters (with direct impacts on agriculture and society) recorded in Chinese historical documents. This dataset has also been verified by comparison with observations. The correlation coefficients between the DWI series and observations are 0.66, 0.84, and 0.78 for the NCP, JH area, and JN area, respectively (Zheng et al., 2006).
(3) Simulation data. The model used for these data is CESM version 1.02, which was developed and released by the National Center for Atmospheric Research (NCAR) (Hurrell et al., 2013). In the experiment, the atmosphere, ocean, land, and sea-ice components are activated by using the Community Atmosphere Model version 4 (Neale et al., 2013), the Community Land Model version 4 (Lawrence et al., 2011), the Parallel Ocean Program version 2, and version 4 of the Los Alamos National Laboratory Community Ice Code. The resolution is set as medium, with grid sizes of 2.5° longitude × 1.9° latitude for atmosphere and land, and 1.0° longitude × 1.0° latitude, approximately, for ocean and sea-ice. The simulation is started with a fixed pre-industrial external forcing, in which the initial conditions are total solar irradiation of 1360.89 W m–2, CO2 concentration of 284.7 ppmv, CH4 concentration of 791.6 ppbv, and N2O concentration of 275.68 ppbv, as well as the aerosol and land use forcing in 1850 provided by NCAR. The simulation runs for 1200 yr in total, in which the first 200 yr is used for spin-up and the last 1000 yr (from 201 to 1200) for analysis. It has been reported that CESM with fixed pre-industrial external forcing is able to reproduce the spatial patterns and temporal reversal of decadal variation of summer precipitation over eastern China (Zheng et al., 2016).
Additionally, the PDO index series derived both from observations during 1900–2015 (released by the Joint Institute for the Study of the Atmosphere and Ocean; http://research.jisao.washington.edu/data_sets/) and tree-ring-based reconstruction for the period 993–1996 (MacDonald and Case, 2005) are also used for comparison. The PDO index in the simulation data is defined as the leading EOF (empirical orthogonal function) of SST anomalies for the Pacific Ocean to the north of 20°N (Zhang et al., 1997).2.2 Methods
In this study, the sub-regions for decadal variation of summer precipitation over eastern China are identified by EOF analysis of the observation and simulation data of summer (May–September) precipitation. Based on the leading three EOFs (Fig. 1) derived from observational data, which account for 70.3% of the total variance, the dominant spatial patterns for the decadal variation of summer precipitation over eastern China are: (1) the roughly dipole pattern (Fig. 1a) demarcated by 35°N, approximately; (2) the tripole belts (Fig. 1b) with demarcation around 30°N and 40°N, respectively; (3) the four zonal belts (Fig. 1c) with alternating anomalous centers over the south of 30°N, the JH area (30°–35°N, approximately), the NCP (35°–42°N, approximately), and northeastern China (north of 42°N), respectively. Since the variance explained by the leading three EOFs is large enough and the spatial patterns show common dividing lines, we identify three sub-regions according to the spatial patterns of the three leading EOFs, as follows: the NCP (35°–40°N, approximately); the JH area (30°–35°N, approximately); and the JN area (20°–30°N, approximately). Such divisions are also roughly consistent with the sub-regions for the precipitation and DWI reconstructions within historical times. Whereas, the three sub-regions for the decadal variation of summer precipitation over eastern China (except for northeastern China), shown from the simulation data, should be demarcated by 32°N and 26°N from north to south, because the EOF1 (Fig. 1d) shows a dipole pattern with a demarcation around 32°N, and the EOF2 (Fig. 1e) shows a tripole pattern with demarcations around 26°N and 38°N, respectively; the demarcations in both EOF1 and EOF2 appear to shift southwards compared with those from the observational data. Meanwhile, the EOF3 (Fig. 1f) derived from the simulation data shows a monopole pattern (i.e., consistently dry or wet) throughout eastern China. The sub-regional precipitation is then calculated by the arithmetic mean among all stations (for the observationaldata) and the weighted average of all grids (for the simulation data) located in each sub-region. As we can see, the leading EOFs from the observation and simulation do not exactly agree, and there may be many reasons for this. For instance, firstly, the observations span 54 yr, while the simulation spans 1000 yr. The observational data have an irregular distribution, based on site location, while the simulation has a grid size of 1.9° (latitude) by 2.5° (longitude). The different temporal spanning and spatial distribution lead to discrepancy in the temporal and spatial samples in the calculation of EOFs, and therefore, may contribute to discrepancy in the EOFs. Secondly, simulations exclude any influences of external forcing variation, while observations are essentially influenced by actual external forcing, such as human-induced CO2, aerosol, land use, and volcanic eruption. In addition, model uncertainty may also lead to the discrepancy in the EOFs.
Power spectrum analysis is then employed to detect the cycles of decadal variations of summer precipitation for each sub-region, and low-pass or band-pass fast Fourier transform (FFT) filter smoothing is adopted to investigate their temporal evolutions. Correlation analysis is then performed to investigate the consistency of the decadal variations of summer precipitation among sub-regions. Additionally, comparison between the consistency of the decadal variations of summer precipitation among sub-regions and the variation of the PDO index is also carried out.3 Results and discussion 3.1 Decadal variation of summer precipitation from observation and historical reconstruction
Figure 2 shows the percentage of the summer precipitation anomaly (relative to the mean of 1961–90) from observations during 1951–2014 for the three belts (35°–40°N, 30°–35°N, and 20°–30°N), and the PDO index. The results suggest that the summer precipitation exhibits evident decadal variability with various phases in different sub-regions. Over the NCP, the summer precipitation experiences a decadal decrease twice, at around 1965 and 1979, with the driest period during 1991–2002, as well as a decadal increase from 2003 to 2013 (Fig. 2a). In the JH area, the summer precipitation decreases from 1959, increases at around 1979, decreases from 1992, and increases again from 2003 (Fig. 2b). Meanwhile, in the JN area, the summer precipitation decreases from 1978 and turns to increase from 1993 (Fig. 2c). These decadal variations lead to opposite phases between the NCP and JH area within 1966–95, and between the JH and JN areas within 1955–63 and 1987–2013. The correlation coefficients between the low-pass smoothed series by a 10-yr FFT of the PDO index (Fig. 2d) in 1951–2014 and sub-regional summer precipitation are –0.516 (significant at the 0.1 level with effective degrees of freedom) for the NCP, 0.035 for the JH area, and –0.059 for the JN area. This finding suggests that the decadal increase (decrease) of summer precipitation over the NCP generally corresponds to a PDO phase shift from cold to warm (from warm to cold).
Figure 3 illustrates the percentage of the anomaly (relative to the mean of 1961–90) for JJA precipitation over the NCP (Fig. 3a), and the Meiyu precipitation over the JH area (Fig. 3b), reconstructed from historical archives during 1736–2000. Note that the Meiyu season only covers 27 days (16 June to 12 July on average, approximately) within May–September, but the rainfall in the Meiyu season accounts for more than 30% of the May–September precipitation, and more than 50% of the total variance of precipitation variation. The results show that the dominant cycle for the decadal variation of summer precipitation is 22–24 yr (Fig. 3c) over the NCP, and quasi-36 yr (Fig. 3d) over the JH area, both of which agree with the results reported by Wang (1962a,b,c). These different dominant decadal cycles thus result in different phases of decadal variation of precipitation between the NCP and JH. The correlation coefficient for the 30-yr running interval between the decadal variation of JJA precipitation over the NCP and the Meiyu precipitation over the JH area shows that an opposite pattern of change occurs during 1736–80 and 1966–91, but a consistent pattern occurs in 1785–1834, 1886–1915, and 1942–60. Comparison of the decadal variations of precipitation with PDO index in 1900–2000 shows that, on the whole, less/more precipitation over the NCP corresponds to a warm/cool PDO phase. Although the reverse pattern is found during 1916–78 between Meiyu precipitation and PDO index, their correlation during 1900–2000 is not statistically significant.3.2 Decadal variation of summer precipitation from proxy data
The proxy data of summer precipitation used here are the sub-regional DWI series during A.D. 500–2000. Note that a 10-yr moving average was involved in developing the sub-regional DWI (Zheng et al., 2006), which means that the high frequency [> 1 (10 yr)–1] variability in the DWI series has been removed. Thus, these DWI series depict the decadal variability of summer precipitation directly, and only the decadal resolution series (i.e., the value in 501–510, 511–520, and so on) are selected to investigate the dominant cycle via power spectrum analysis, so as to avoid the autocorrelation induced by the 10-yr moving average.Figure 4 shows the power spectrum of the DWI series for the NCP, JH area, and JN area. The results show that the decadal-scale variation is dominated by 22–24 yr over the NCP and 32 yr over the JH and JN areas, which agree with the results shown in Figs. 3c and 3d. Moreover, the dominant cycle of multi-decadal oscillation is also found: 70–74 yr over the NCP (Fig. 4a); 45–48 and 82 yr over the JH area (Fig. 4b); and 44 yr over the JN area (Fig. 4c).
However, the evolution of band-pass FFT filter smoothing (Fig. 5) shows that each dominant periodic pattern of variability is not consistently present over the whole of A.D. 500–2000. For example, the period at the scale of 20–35 yr dominates the decadal variation, accounting for 29.3%, 22.6%, and 22.6% of the total variance of summer precipitation change over the NCP, JH area, and JN area, respectively (Fig. 5a), but the signal is notably weak in most decades during A.D. 700–800, 840–900, 980–1500, and the 20th century, approximately, for the NCP; A.D. 500–680, 750–850, 1130–70, 1370–1440, 1650–1700, and after 1850, approximately, for the JH area; and 500–50, 800–1050, 1350–1440, and 1500–1650, approximately, for the JN area. The reverse variational pattern between the NCP and JN area is found in 550–640, 850–80, 950–90, 1520–80, and 1730–85. The period at the scale of 35–50 yr accounts for 12.3%, 17.2%, and 18.8% of the total variance of summer precipitation change for the NCP, JH area, and JN area (Fig. 5b), with a very weak signal during 900–1000, 1130–1350, and after 1750 over the NCP, during 500–50, 650–700, 800–50, 1050–1350, and after 1800 over the JH area, and during 500–750, 1250–1300, 1440–90, and 1680–1740 over the JN area, approximately, as well as a reverse variational pattern between the NCP and JN area within 500–740, 1060–1220, 1310–70, and 1720–1800.
Meanwhile, at the 50–85-yr period, which accounts for 13.5%, 12.1%, and 7.5% of the total variance of summer precipitation change for the NCP, JH area, and JN area (Fig. 5c), the strong signal is exhibited in the 11th to 12th centuries, 14th to 15th centuries, and after 1750 for NCP; the 6th century, the 9th century, the 14th century, and the 18th to 20th centuries for the JH area; and the 6th century, the 9th to 11th centuries, and the 16th to 19th centuries for the JN area. Moreover, a reverse variational pattern between the NCP and JN area is found during 760–870, 1180–1390, 1490–1650, and after 1730. This period looks likely to be related with the multi-decadal cycle of the PDO from 993 to 1996 (MacDonald and Case, 2005). As shown by Fig. 5c, at the scale of 50–85 yr, the warm PDO phase corresponds generally to dry conditions in both the NCP and JH area but wet conditions in the JN area, and the cool PDO phase corresponds generally to wet conditions in both the NCP and JH area but dry conditions in the JN area, since AD 1800, during which the PDO turns to strengthen. However, such a correspondence is not consistently present before AD 1800. Possible reasons for this unstable relationship might be the variation of the dominant periods of precipitation and PDO along with time. As shown in Fig. 5, the power of precipitation on different timescales varies greatly with time. The power of NCP and JH precipitation is much stronger on the 50–80-yr scale and weaker on the 35–50-yr scale after 1800, while the power of JN precipitation shows the opposite characteristic. The power of the PDO on the scale of 50–70 yr is also unstable and obviously strengthens after 1800, which has been verified in previous studies (e.g.,Minobe, 1997). Due to such variations of the dominant periods, the correspondence between the PDO and the precipitation anomaly will also vary. In addition, the 100-yr low-pass FFT filter smoothing of the DWI series shows that the low-frequency variability accounts for 25.0%, 24.1%, and 29.1% of the total variance of summer precipitation change in the NCP, JH area, and JN area, respectively (Fig. 5d).3.3 Decadal variation of summer precipitation from simulation
To isolate the decadal variation of precipitation, we calculate the 10-yr running average of the CESM simulations to remove the interannual variability. Then, the 10-yr mean series is used to carry out power spectrum analysis. The power spectrum of CESM-simulated precipitation (Fig. 6) shows that the significant decadal cycle periods exist across all of the sub-regions. For the NCP, there are significant decadal cycle periods around 20 and 25 yr, and a centennial cycle of 127–140 yr; plus, the peaks of the power spectrum at 30–40 and 60–80 yr do not pass the 90% confidence level. For the JH area, there are significant cycle periods of around 22, 38, 64–67, and 140–155 yr. For the JN area, there are only significant cycle periods of 22–23 and 108 yr, as well as strong power at the period of 30–40 yr, albeit not passing the 90% confidence level. Moreover, the dominant cycles of 29 and 42 yr are also found in the PDO variation (Fig. 6d). These decadal cycles from simulations partly match well with those from the historical DWI. For instance, there are common peaks of the power spectrum at 32–38 and around 22 yr for the JH area, and at 22–25 yr and a quasi-centennial cycle for the NCP. It is notable that the peaks of the power spectrum from the simulations and historical DWI do not match exactly. For instance, the periods of 32 and 44 yr from the historical DWI are not shown by the simulation.
Figure 7 shows the variability of individual periodic bands for all simulations, which indicates that they vary with model year, though there are no changes of external forcing in the simulation. The period of 20–35 yr accounts for 19.5%, 12.8%, and 18.2% of the total variance over the NCP, JH, and JN area, respectively. The signal looks stronger from 550 to 800, and it is weak from 450 to 550 and after 1050. Meanwhile, it accounts for 21.8% of the total variance in the PDO variability, with a very weak signal during 800–1000 and around 250, 380, 550, and 1050. At the period of 35–50 yr, it accounts for 5.4%, 8.3%, and 11.2% of the total variance over the NCP, JH, and JN area, respectively. The weak signal exists from 450 to 700 for the NCP, and from 450 to 550 for the JH and JN areas. Meanwhile, the variability of the PDO at the period of 35–50 yr accounts for 11.5% of the total variance, with a more powerful signal during 250–400, 500–80, 700–850, and 1070–1150, approximately. For the period of 50–85 yr, it accounts for 9.4%, 8.6%, and 6.2% of the total variance of the NCP, JH, and JN area, respectively. There are weak signals from 400 to 700 for the NCP and JH area, while it is weak from 600 to 700 for the JN area. A weak signal can also be found from 300 to 400 for the JN area, from 850 to 1000 for the NCP, and from 1000 to 1200 for the JH area. For the variability of the PDO at the period of 50–85 yr, it accounts for only 8.2% of the total variance, with a weak signal during 280–680 and after 1000, approximately. Moreover, the low-frequency [< 1 (100 yr)–1] variability is also notable, which accounts for 14.2%, 10.9%, and 9.4% of the total variance of the NCP, JH, and JN area, respectively.
Along with the unstable strength of individual periodic bands across all simulations, the phase correspondence among the precipitation anomalies of the three sub-regions and the PDO phase shift are also unstable. For example, for the period of 35–50 yr, there is a strongly similar phase shift from 750 to 830, while it is the reverse phase shift from 1000 to 1200, between the NCP and JH area. The same phase shift is likely to be dominant between the JH and JN areas; however, there are strongly opposite phase shifts, such as from 200 to 330 and from 850 to 1000, for the period of 50–80 yr. The correspondence between the PDO shift and precipitation variation also varies with time, regardless of the periodic band or region. Generally, as shown by Fig. 7e, at a scale of longer than 10 yr, the duration with significant (p < 0.1) correlations between the PDO phase shift and precipitation variation for the 3 sub-regions accounts for only 10%–15% of total years. This duration, meanwhile, consists of both the same phase shift, i.e., positive correlation, and the reverse phase shift, i.e., negative correlation. Moreover, the variation of correlation also varies with region. For the NCP, the same phase shift occurs from 650 to 750 and from 840 to 960, and the reverse phase shift occurs from 360 to 460, around 800, and from 1090 to 1030. For the JH and JN areas, there are common phase shifts from 360 to 420 and from 750 to 800, and common reverse phase shifts from 660 to 720. In addition, there is the same phase shift from 1030 to 1060 for the JH area, while in this period it is the reverse shift for the JN area.3.4 Discussion
The above-mentioned findings demonstrate that the dominant periods of cycle from observed and reconstructed precipitation, the DWI, and simulated precipitation from the control run of CESM, mostly match with one another. There are common cycle periods, such as 20–25 yr and around 120 yr in the NCP, and around 20, 32–38, and 60–80 yr in the JH area. These common cycle periods from different data sources not only prove the ability of CESM in capturing the variation of precipitation over eastern China, but also confirm existence of these significant cycle periods.
It is noted, however, that mismatch exists too. For instance, the cycle of 70–74 yr for the NCP, the cycle of 45–48 yr for the JH, and the cycle of 32 and 44 yr for the JN area, are not detected in the CESM simulation. Such mismatches might be partly derived from the defects in CESM. However, two definite factors contributing to these mismatches should be noted. One is that historical data are derived from reality, including the effects of all external forcing, such as solar activity, volcanic eruption, and increased greenhouse gases, while precipitation variation in the CESM simulation is only derived from internal variability of the climate system. The other is that the DWI data are derived from descriptions of drought and flood disasters (with direct impacts on agriculture and society), rather than fully from the precipitation amount, as simulated by CESM.
In addition to having a similar power spectrum distribution, both the DWI and the CESM simulation exhibit varied strengths of individual period bands with time. As shown in Figs. 5 and 7, the variation of signal strength with time is dependent on both sub-region and periodic band. Due to these complex variations, the correspondence of the precipitation anomaly among sub-regions is unstable for each period band. It is also noteworthy that the correspondence between the precipitation anomaly and PDO phase shift is unstable along with time, regardless of region and period band. All of these unstable correspondences are present in both the DWI and the CESM simulation.
Against the background of the above-mentioned unstable correspondences, the decadal shift around 1977–79 looks different from that around 1992/93. Around 1977–79, the precipitation decreases over the NCP and JN area look like being derived from the phase shift of the quasi-70-yr period and the period of 20–35 yr, respectively; the precipitation increase in the JH area is derived from the phase shift of quasi-45–48 yr. Around 1992/93, the precipitation increase over the south of the YZR and the decrease in the north is principally due to the existing negative phase of the quasi-70-yr period and a negative phase in the 22–24-yr period over the NCP, but positive phases in all of the decadal cycles over the JN area.
As the CESM simulation excludes the effect of external forcing, the simulation exhibits the full internal variability of the climate system. As shown by Fig. 7, the internal variability of the climate system could lead to strong decadal to centennial precipitation variability. Moreover, the strength of the variation tends to vary with time, rather than being constant. Such a characteristic looks very similar to the variations of the reconstructed historical DWI (Fig. 5). These findings suggest that the decadal variation of precipitation over eastern China may be derived from internal variability alone. Thereby, the decadal oscillations over eastern China, including the decadal shifts observed in the late 1970s, early 1990s, and early 2000s, may possibly result from internal variability of the climate system.4 Conclusion
This study investigates the decadal variability of summer precipitation, with a focus on the dominant decadal cycle, the temporal evolution and pattern of variation at individual periods, and the relationship between variations of the PDO and precipitation, for the NCP, JH area, and JN area in eastern China, at the millennial scale, by synthesis of observations, reconstructions, proxy data, and simulation data. The conclusions can be summarized as follows.
(1) The decadal oscillations of summer precipitation over eastern China are significant at peaks of 22–24, 32–36, 44–48, and quasi-70 yr, in which the cycle of 22–24 and 32–36 yr had been pointed out by Shaowu Wang previously (Wang, 1962a,b,c;Wang and Zhao, 1979a). However, the dominant cycle for the decadal variation of summer precipitation varies across sub-regions, with 22–24 and quasi-70 yr for the NCP, 32–36, 44–48, and quasi-70 yr in the JH area, and 32–36 and 44–48 yr in the JN area, which result in different phases of decadal variation of precipitation along with time among the three sub-regions over eastern China. The variability of sub-regional summer precipitation at each decadal period, including the scale of 20–35, 35–50, and 50–80 yr, is not consistent across the entire millennium. Thus, both result in different spatial patterns of the decadal variation of precipitation along with time over eastern China, including the decadal shifts of the spatial pattern in the precipitation change observed in the late 1970s, early 1990s, and early 2000s.
(2) The relationship between precipitation variations and the PDO shift in eastern China shows that the warm (cold) PDO phase generally corresponds to dry (wet) conditions in the NCP but wet (dry) conditions in the JN area after AD 1800, when the PDO turns to strengthen. However, this correspondence is not consistently present before AD 1800. In the simulation, such a correspondence only exists in model years 350–400, 770–820, and 1080–1140. Both the data and the simulation suggest that the relationship between the PDO phase shift and precipitation variation is unstable. Such unstable correlations imply that the regional precipitation variations may be regulated significantly by the PDO in some periods, while it is not in other periods, such as from 1951 to 2014 for the JH and JN areas.
These findings suggest that the decadal oscillations resulting from the internal variability of the climate system are various and unstable. Therefore, caution should be exercised when using the decadal behavior of precipitation and the PDO observed in the 20th century as a basis for anticipating the decadal shift of the spatial pattern of precipitation variation in the near future.
|C. L. Chan, W. Zhou J. ,2005: PDO, ENSO and the early summer monsoon rainfall over South China. Geophys. Res. Lett. , 32 , L08810. DOI:10.1029/2004GL022015|
|Y. H. Ding, Z. Y. Wang, Y. Sun ,2008: Inter-decadal variation of the summer precipitation in East China and its association with decreasing Asian summer monsoon. Part I: Observed evidences. Int. J. Climatol. , 28 , 1139–1161.|
|Y. H. Ding, Y. Sun, Z. Y. Wang, et al ,2009: Inter-decadal variation of the summer precipitation in East China and its association with decreasing Asian summer monsoon. Part II: Possible causes. Int. J. Climatol. , 29 , 1926–1944.|
|Q. S. Ge, X. F. Guo, J. Y. Zheng, et al ,2008: Meiyu in the middle and lower reaches of the Yangtze River since 1736. Chinese Sci. Bull. , 53 , 107–114.|
|Z. Q. Gong, J. H. Zhao, G. L. Feng ,2013: Analysis of the summer precipitation of 2012 in East China and its possibility of decadal shift. Acta Phys. Sinica , 62 , 099205. DOI:10.7498/aps.62.099205|
|Z. Z. Hu ,1997: Interdecadal variability of summer climate over East Asia and its association with 500-hPa height and global sea surface temperature. J. Geophys. Res. , 102 , 19403–19412.|
|R. H. Huang, Y. H. Xu, L. T. Zhou ,1999: The interdecadal variation of summer precipitations in China and the drought trend in North China. Plateau Meteor. , 18 , 465–476.|
|R. H. Huang, J. L. Chen, Y. Liu ,2011: Interdecadal variation of the leading modes of summertime precipitation anomalies over eastern China and its association with water vapor transport over East Asia. Chinese J. Atmos. Sci. , 35 , 589–606.|
|R. H. Huang, Y. Liu, T. Feng ,2013: Interdecadal change of summer precipitation over eastern China around the late-1990s and associated circulation anomalies, internal dynamical causes. Chinese Sci. Bull. , 58 , 1339–1349.|
|J. W. Hurrell, M. M. Holl, P. R. Gent, et al ,2013: The community earth system model: A framework for collaborative research. Bull. Amer. Meteor. Soc. , 94 , 1339–1360.|
|D. M. Lawrence, K. W. Oleson, M. G. Flanner, et al ,2011: Parameterization improvements and functional and structural advances in version 4 of the Community Land Model. J. Adv. Model. Earth Syst. , 3 , M03001. DOI:10.1029/2011MS00045|
|G. M. MacDonald, R. A. Case ,2005: Variations in the Pacific Decadal Oscillation over the past millennium. Geophys. Res. Lett. , 32 , L08703. DOI:10.1029/2005GL022478|
|G. A. Meehl, L. Goddard, J. Murphy, et al ,2009: Decadal prediction: Can it be skillful?. Bull. Amer. Meteor. Soc. , 90 , 1467–1485.|
|G. A. Meehl, L. Goddard, G. Boer, et al ,2014: Decadal climate prediction: An update from the trenches. Bull. Amer. Meteor. Soc. , 95 , 243–267.|
|R. B. Neale, J. Richter, S. Park, et al ,2013: The mean climate of the Community Atmosphere Model (CAM4) in forced SST and fully coupled experiments. J. Climate , 26 , 5150–5168.|
|C. Qian, T. J. Zhou ,2014: Multidecadal variability of North China aridity and its relationship to PDO during 1900–2010. J. Climate , 27 , 1210–1222.|
|The Council of the Second National Assessment Report on Climate Change, 2011:Second National Assessment Report on Climate Change. Science Press, Beijing, 710 pp. (in Chinese)|
|L. Wang, Y. Qian, Y. C. Zhang, et al ,2016: Observed variability of summer precipitation pattern and extreme events in East China associated with variations of the East Asian summer monsoon. Int. J. Climatol. , 36 , 2942–2957. DOI:10.1002/joc. 4530|
|S. W. Wang ,1962a: The multi-year variation of atmospheric centers of action. Acta Meteor. Sinica , 32 (S1) , 304–318.|
|S. W. Wang ,1962b: The multi-year variation of atmospheric centers of action over eastern Asia and the oscillation of climate in China. Acta Meteor. Sinica , 32 , 19–36.|
|S. W. Wang ,1962c: Analysis on the climate oscillation in Shanghai, China. Acta Meteor. Sinica , 32 , 322–336.|
|S. W. Wang ,1963: A preliminary study on the characteristics and evolution of mean monthly circulation. Part I: Zonal index. Acta Meteor. Sinica , 33 , 361–374.|
|S. W. Wang ,1964a: A preliminary study on the characteristics and evolution of mean monthly circulation. Part II: Ultra-long waves. Acta Meteor. Sinica , 34 , 316–328.|
|S. W. Wang ,1964b: Oscillation of atmospheric circulation in recent 90 years, Part I. Acta Meteor. Sinica , 34 , 486–506.|
|S. W. Wang ,1965: Oscillation of atmospheric circulation in recent 90 years, Part II. Acta Meteor. Sinica , 35 , 200–214.|
|S. W. Wang ,1973: Atmospheric circulation and climate anomalies and outlook. Meteor. Sci. Technol. (3) , 24–30.|
|S. W. Wang, Z. C. Zhao ,1979a: The 36-yr wetness oscillation in China and its mechanism. Acta Meteor. Sinica , 37 , 64–73.|
|S. W. Wang, Z. C. Zhao ,1979b: An analysis of historical data of droughts and floods in last 500 years in China. Acta Geogr. Sinica , 34 , 329–341.|
|S. W. Wang, Q. E. Gong, F. L. Niu ,1963: Study on climate oscillation in China. Acta Sci. Nat. Univ. Pek. (4) , 343–358.|
|WCRP, 2009: The World Climate Research Programme Implementation Plan 2010-2015, WMO/TD No. 1503, Geneva, WMO, 43 pp.|
|Z. Q. Xu, K. Fan, H. J. Wang ,2015: Decadal variation of summer precipitation over China and associated atmospheric circulation after the late 1990s. J. Climate , 28 , 4086–4106.|
|Y. Zhang, M. Lü Q., M. Yang J., et al ,2007: The interdecadal variation of precipitation pattern over China during summer and its relationship with the atmospheric internal dynamic processes and extra-forcing factors. Chinese J. Atmos. Sci. , 31 , 1290–1300.|
|R. H. Zhang ,2015: Natural and human-induced changes in summer climate over the East Asian monsoon region in the last half century: A review. Adv. Climate Change Res. , 6 , 131–140.|
|Y. Zhang, J. M. Wallace, D. S. Battisti ,1997: ENSO-like interdecadal variability: 1900-93. J. Climate , 10 , 1004–1020.|
|J. Y. Zheng, Z. X. Hao, Q. S. Ge ,2005: Variation of precipitation for the last 300 years over the middle and lower reaches of the Yellow River. Sci. China Ser. D Earth Sci. , 48 , 2182–2193.|
|J. Y. Zheng, W. C. Wang, Q. S. Ge, et al ,2006: Precipitation variability and extreme events in eastern China during the past 1500 years. Terr. Atmos. Oceanic Sci. , 17 , 579–592.|
|J. Y. Zheng, M. W. Wu, Z. X. Hao, et al ,2016: Spatial pattern of decadal variation of summer precipitation in eastern China: Comparison of observation and CESM control simulation. Geogr. Res. , 35 , 14–24.|
|B. T. Zhou, D. D. Xia ,2012: Interdecadal change of the connection between winter North Pacific Oscillation and summer precipitation in the Huaihe River valley. Sci. China Earth Sci. , 55 , 2049–2057.|
|W. Zhou, C. Y. Li, X. Wang ,2007: Possible connection between Pacific oceanic interdecadal pathway and East Asian winter monsoon. Geophys. Res. Lett. , 34 , L01701. DOI:10.1029/2006GL027809|
|Y. L. Zhu, H. J. Wang, W. Zhou, et al ,2011: Recent changes in the summer precipitation pattern in East China and the background circulation. Climate Dyn. , 36 , 1463–1473.|
|Y. L. Zhu, H. J. Wang, J. H. Ma, et al ,2015: Contribution of the phase transition of Pacific Decadal Oscillation to the late 1990s' shift in East China summer rainfall. J. Geophys. Res. , 120 , 8817–8827. DOI:10.1002/2015JD023545|
|Y. M. Zhu, X. Q. Yang ,2003: Relationships between Pacific Decadal Oscillation (PDO) and climate variabilities in China. Acta Meteor. Sinica , 61 , 641–654.|