Abstract

Analysis of 26 simulations from 11 general circulation models (GCMs) of the Atmospheric Model Intercomparison Project (AMIP) II reveals a basic inability to simultaneously predict the Yangtze River Valley (YRV) precipitation (PrYRV) annual cycle and summer interannual variability in response to observed global SST distributions. Only the Community Climate System Model (CCSM) and L’Institut Pierre-Simon Laplace (IPSL) models reproduce the observed annual cycle, but both fail to capture the interannual variability. Conversely, only Max Planck Institute (MPI) simulates interannual variability reasonably well, but its annual cycle leads observations by 2 months.

The interannual variability of PrYRV reveals two distinct signals in observations, which are identified with opposite subtropical Pacific SST anomalies in the east (SSTe) and west (SSTw). First, negative SSTe anomalies are associated with equatorward displacement of the upper-level East Asian jet (ULJ) over China. The resulting transverse circulation enhances low-level southerly flow over the South China Sea and south China while convergent flow and upward motion increase over the YRV. Second, positive SSTw anomalies are linked with westward movement of the subtropical high over the west-central Pacific. This strengthens the low-level jet (LLJ) to the south of the YRV. These two signals act together to enhance PrYRV. The AMIP II suite, however, generally fails to reproduce these features. Only the MPI.3 realization is able to simulate both signals and, consequently, realistic PrYRV interannual variations. It appears that PrYRV is governed primarily by coherent ULJ and LLJ variations that act as the atmospheric bridges to remote SSTe and SSTw forcings, respectively. The PrYRV response to global SST anomalies may then be realistically depicted only when both bridges are correctly simulated. The above hypothesis does not exclude other signals that may play important roles linking PrYRV with remote SST forcings through certain atmospheric bridges, which deserve further investigation.

1. Introduction

Summer rainfall over the Yangtze River Valley (YRV), as bounded by 25°–35°N and 110°–120°E, is extremely important because this region is vital to commerce, agricultural activity, transportation, and energy production in China. Excessive rainfall during the summer monsoon causes severe flooding and the dislocation of millions of people (e.g., 1998 and 2008) while a failure of the monsoon can have catastrophic economic consequences (e.g., 1985 and 2006).

The capability of numerical models to reproduce observed relationships between summer monsoon precipitation over east China (including the YRV) and the physical mechanisms that explain rainfall variability varies widely. Kang et al. (2002) found that general circulation models (GCMs) have difficulty reproducing the observed location and variations in the rainbands associated with the East Asian monsoon, where the coarse resolution of the models was attributed to substantial precipitation biases (Yu et al. 2000; Zhou and Li 2002). Although the occurrence of local precipitation variations over short time scales may not be resolvable on coarse grids (e.g., Lau and Ploshay 2009), GCMs are better suited to simulate interannual variability of regional precipitation and its teleconnections with larger-scale circulations (e.g., Liang et al. 2001, 2002, 2008). These resolvable features will be the focus of this study.

The identification of realistic model teleconnections between east China monsoon precipitation and the large-scale circulation is complicated by the fact that many GCMs have substantial regional precipitation biases. Liang et al. (2001) analyzed the East Asian monsoon simulated by the GCMs of the Atmospheric Model Intercomparison Project I (AMIP I; Gates et al. 1999). A comparison with observations led to the identification of model biases in both regional precipitation and the larger-scale circulation that are physically linked during the summer months. In particular, the poleward (equatorward) displacement of the East Asian jet was associated with negative (positive) rainfall biases over the YRV. The results indicate that the dominant atmospheric processes governing YRV rainfall variations are essentially captured by current GCMs.

The relationship between regional precipitation and the large-scale circulation is, however, time-scale dependent. Liang et al. (2002) analyzed 30 AMIP I simulations for the period 1979–88 and found that no correspondence exists between model ability to predict the annual precipitation cycle and the interannual variability of observed summer rainfall over east China. Thus, the large-scale circulation mechanisms that explain summer rainfall interannual variability may differ from those that are linked with the annual rainfall cycle. We will therefore examine both the annual cycle and summer interannual variability of YRV precipitation in this study. Data from 26 AMIP II simulations for the period 1979–2000 will be analyzed. In addition to a doubling of the integration period length, the AMIP II contains advanced models with updated physics representations and resolution refinements. The results will demonstrate that the new model suite is still unable to simultaneously simulate both features.

This inability may arise from 1) local effects that determine YRV precipitation and cannot be resolved by the GCMs and/or 2) model deficiencies in simulating observed YRV rainfall teleconnections with the large-scale circulation. However, we will identify some AMIP II simulations that are able to reproduce observed YRV precipitation teleconnections that are forced by SST anomalies through the atmospheric bridge. This suggests that the GCMs have the potential to capture the circulation features that govern observed YRV precipitation variability. As such, YRV precipitation may be predicted using indirect measures such as circulation indices established from observations (Webster and Yang 1992; Liang and Wang 1998; Wang et al. 2001; Zhou et al. 2009a).

The goal of this study is to determine the ability of the AMIP II GCMs to reproduce observed relationships between YRV rainfall, global SST, and the large-scale circulation. Section 2 describes the observed data and simulated suite that are used in this study. Section 3 compares the observed annual precipitation cycle over the YRV with those generated by the GCMs and illustrates the general failure of the models. Section 4 discusses the interannual variability of observed YRV summer rainfall and focuses on its teleconnections with Eurasian circulation and Pacific SST anomalies. Section 5 compares observed and AMIP-simulated YRV summer rainfall interannual variations, with an emphasis on their teleconnection patterns. Section 6 summarizes the results with discussion of the underlying physical mechanisms for YRV summer rainfall prediction.

2. Model simulations and observations

This study analyzes the climates simulated by 11 separate GCMs under the AMIP II protocol (Gates et al. 1999). Each model was run for the period 1979–2000, where identical “perfect” ocean surface conditions, as specified by the observed global distributions of monthly mean SST and sea ice variations, were incorporated to drive the atmosphere. Multiple realizations with the same oceanic forcing, but different initial conditions, were available for five GCMs. This resulted in a total of 26 simulations. The spread between multiple realizations of a single model defines the uncertainty due to internal variability in isolating externally forced signals. The basic model information can be found online at http://www-pcmdi.llnl.gov/projects/modeldoc/amip2/.

Observational data for the period 1979–2000 are derived from several sources. Precipitation data over the YRV are provided by daily rain gauge measurements from 98 stations in the area of (25°–35°N, 110°–120°E), including all standard monitoring sites from the China Meteorological Administration. The precipitation anomaly time series for each month is constructed by calculating the area-averaged value for that month and then computing the difference between the yearly values and the 1979–2000 climatological mean. The YRV summer rainfall anomaly time series is constructed in a similar manner but represents the average for the months of June, July, and August (JJA). Over the global oceans, SST data come from the 1° analysis of Hurrell et al. (2008) for the observational analysis and from each GCM archive at the corresponding model resolution for the AMIP suite comparison. This is to ensure a consistent comparison of the model responses to the actual SST forcings that were incorporated into individual GCMs. For the circulation fields (wind, humidity), we use the 2.5° National Centers for Environmental Prediction (NCEP)–Department of Energy (DOE) AMIP II reanalysis (Kanamitsu et al. 2002).

In this study, 1979–2000 is chosen as the reference period to construct the annual cycle (i.e., average over the AMIP II period) and interannual anomalies (i.e., departures from the average). The same period applies for all comparisons between model simulations and observations. Time series correlations will be determined through the use of the Student’s t test with 20 degrees of freedom (i.e., assuming independence between years), where significance at the 0.05 level occurs when coefficient magnitude is greater than or equal to 0.42.

3. Annual cycle of YRV precipitation

The annual precipitation cycle over the YRV is dominated by the East Asian monsoon circulation (Ding 1994; Samel et al. 1999). Amounts during the fall, winter, and early spring are light (<2 mm day−1). Precipitation then increases rapidly during the late spring and summer as the primary monsoon rainband impacts the region. Climatologically, the monsoon rainband resides over the YRV between mid-June and mid-July. Rainband movement leads to a distinct annual precipitation cycle over the region, where the heaviest rainfall occurs during June (6.9 mm day−1) and approximately half of the mean annual total falls during summer, which is defined to be the months of JJA. Given the seasonality of YRV precipitation and the inability of coarse-domain GCMs to resolve local precipitation variations over small times scales (Yu et al. 2000; Zhou and Li 2002; Kang et al. 2002; Liang et al. 2001, 2002, 2008), JJA mean fields will be used to analyze observed and simulated YRV rainfall teleconnections.

The capability of the AMIP II models to reproduce the observed annual YRV precipitation cycle is assessed by analyzing the phase of the simulated annual cycles as well as the root-mean-square (RMS) errors of their forecasts. For a given run, the size of the GCM phase shift is determined by the number of months that the model precipitation climatology must be shifted to produce the maximum correlation with observations. Figure 1 shows that only one simulation, the single Community Climate System Model (CCSM) run, has the same phase as observations. An analysis of the six L’Institut Pierre-Simon Laplace (IPSL) realizations reveals that correlations with the observed annual cycle are large when there is no phase lag (+0.89) as well as when the runs lead observations by one month (+0.92). Because the small correlation difference makes it difficult to distinguish whether the simulations lead observations by a month or have no phase lag, we assign a −0.5 month phase lag for all IPSL runs. In every other case, model precipitation leads observations by 1–2 months with a clear correlation peak. In addition to these negative phase lags, all realizations have substantial RMS errors, between 3 and 5 mm day−1, which are comparable to the observed annual mean precipitation rate of 3.38 mm day−1.

Fig. 1.

The phase lags (month, hollow bars) and rms errors (mm day−1, solid bars) of all model simulated from observed YRV precipitation annual cycles. Each simulation is labeled as the modeling institution name, followed by an identification for a specific model (if any), and then by a dot plus a number for every run. In particular, MIROCh and MIROCm denote the high- and medium-resolution versions, respectively.

Fig. 1.

The phase lags (month, hollow bars) and rms errors (mm day−1, solid bars) of all model simulated from observed YRV precipitation annual cycles. Each simulation is labeled as the modeling institution name, followed by an identification for a specific model (if any), and then by a dot plus a number for every run. In particular, MIROCh and MIROCm denote the high- and medium-resolution versions, respectively.

When multiple realizations of a single GCM (e.g., IPSL.1–6) are compared in Fig. 1, we see that the phase lags are identical, and the RMS errors are similar. This is true for the Goddard Institute for Space Studies (GISS), Flexible Global Ocean–Atmosphere–Land System Model (FGOALS), Model for Interdisciplinary Research on Climate (MIROC), Max Planck Institute (MPI), and IPSL GCMs. Thus, the YRV precipitation annual cycles produced by these models do not appear to be sensitive to initial condition differences.

The AMIP II suite also includes two GCMs that have high- and low-resolution realizations. Figure 1 shows that the high-resolution simulation of the MIROC GCM (MIROCh) produces the same phase lag (−1 month) and a larger RMS error than the three medium-resolution (MIROCm) realizations. Thus, the increased resolution does not lead to a more realistic YRV precipitation annual cycle. On the other hand, the CCSM simulation is a higher-resolution version of the Parallel Climate Model (PCM) that also incorporates an enhanced physics package. In this case the CCSM produces no phase lag and has the smallest RMS error among all AMIP II runs. This is a substantial improvement over the PCM, which has a phase lag of −2 months and a much larger RMS error. As a result, the improved physics in the CCSM leads to a much more realistic annual precipitation cycle over the YRV. This result is supported by Chen et al. (2010), who found that the mean state and seasonal cycle of East Asian summer monsoon elements simulated by a single GCM was highly sensitive to modifications in the physics package.

Figure 2 compares climatological monthly mean precipitation variations between observations and six distinct groups of model simulations to highlight the substantial AMIP II precipitation RMS errors and annual cycle phase shifts. Clearly, the CCSM produces a more realistic prediction than the PCM (Fig. 2a) as a result of its increased resolution and improved physics. The annual cycles of these two runs also show a common bias among the AMIP II simulations, where amounts tend to be overestimated in the winter and spring but underestimated during the summer and fall. The annual cycles of the GISS and MRI lead observations by two months (Fig. 2b). These runs also underestimate peak precipitation. The four GISS realizations are very similar and illustrate the small RMS error differences indicated in Fig. 1. On the other hand, the GISS annual cycles differ substantially from that of MRI, especially between July and December. This suggests that the annual cycles simulated by different GCMs vary more than those among multiple realizations of the same model. The annual cycles of FGOALS and Centre National de Recherches Météorologiques (CNRM) lead observations by one month (Fig. 2c). The three FGOALS runs are virtually identical and very similar to the CNRM. While the phase shifts are not as large as those shown in Fig. 2b, the wet season in each of the FGOALS and CNRM simulations occurs over a longer period, and the rainfall peaks are smaller than observations. The six IPSL realizations are very similar and consistently underestimate YRV summer rainfall (Fig. 2d). The annual cycles of the MIROC differ little between its single high-resolution and three medium-resolution realizations (Fig. 2e). The remaining MPI (three realizations), Institute of Numerical Mathematics (INM), and Met Office (UKMO) runs all yield substantial overprediction between December and May and underprediction between July and October (Fig. 2f). This reflects the very large RMS errors shown in Fig. 1.

Fig. 2.

The YRV precipitation (mm day−1) annual cycle observed and modeled by six distinct groups. The legend lists OBS (thick solid) for observations and model names (thick dashed or thin curves with various patterns). Multiple realizations, if any, are depicted by a same curve pattern. The UKMO peaks at 8.7 mm day−1 in April, not shown to accommodate clearer contrast for other models.

Fig. 2.

The YRV precipitation (mm day−1) annual cycle observed and modeled by six distinct groups. The legend lists OBS (thick solid) for observations and model names (thick dashed or thin curves with various patterns). Multiple realizations, if any, are depicted by a same curve pattern. The UKMO peaks at 8.7 mm day−1 in April, not shown to accommodate clearer contrast for other models.

4. Observed YRV summer precipitation interannual teleconnections

Figures 1 and 2 clearly illustrate the failure of the AMIP II suite to simulate the observed annual precipitation cycle over the YRV. Liang et al. (2002), however, found that no linkage exists between the ability of a model to simulate the annual precipitation cycle and its skill in reproducing the interannual variability of seasonal rainfall. Thus, the purpose of the following sections will be to determine the capability of the AMIP II GCMs to simulate observed YRV summer rainfall (PrYRV) interannual variability and its linkages with the atmospheric circulation and its teleconnections. Observed relationships will be described in this section while the ability of the AMIP II suite to reproduce these associations will be determined in section 5.

Observed PrYRV is governed by the movement and intensity of both large- and regional-scale circulation features, including the upper-tropospheric East Asian polar jet over northern Eurasia (Liang and Wang 1998) and the midtropospheric subtropical high over the west-central Pacific Ocean (Chang et al. 2000). Relationships between observed PrYRV and the atmospheric circulation during JJA are identified in Fig. 3, which shows correlations of PrYRV with the 200-hPa zonal wind (U200), 500-hPa meridional wind (V500), 850-hPa meridional wind (V850), and 850-hPa relative humidity (RH850). Significant positive (negative) correlations with U200 along and north of the YRV (over extreme north China) indicate that increased PrYRV is accompanied by an equatorward shift in the location of the upper-level jet. This result is consistent with Liang and Wang (1998) and Zhou and Yu (2005), who found that displacement of the jet to the south of its mean position during the summer months causes both the ascending branch of the jet indirect transverse circulation and precipitation to intensify along the YRV. In addition, significant negative correlations over extreme southern China and adjacent areas of the South China Sea and Pacific Ocean suggest that enhanced PrYRV occurs in conjunction with a weakening of the Hadley circulation.

Fig. 3.

Geographic distributions of the summer interannual correlations (×10) of the YRV precipitation with pointwise (a) 200-hPa zonal wind, (b) 500- and (c) 850-hPa meridional wind, and (d) 850-hPa relative humidity observed during 1979–2000. Outlined in each plot is the YRV region that defines the PrYRV index with which the correlations are calculated. Contour intervals are 2 units. Shading areas denote where correlations are statistically significant at the 95% confidence level.

Fig. 3.

Geographic distributions of the summer interannual correlations (×10) of the YRV precipitation with pointwise (a) 200-hPa zonal wind, (b) 500- and (c) 850-hPa meridional wind, and (d) 850-hPa relative humidity observed during 1979–2000. Outlined in each plot is the YRV region that defines the PrYRV index with which the correlations are calculated. Contour intervals are 2 units. Shading areas denote where correlations are statistically significant at the 95% confidence level.

Significant positive correlations of PrYRV with both V500 and V850 (Figs. 3b,c) are located over the South China Sea and south China. Thus, increased PrYRV is accompanied by stronger southerly flow to the south of the YRV. Given the broad area of negative correlations located over the west-central Pacific at both levels, this increased southerly flow as well as enhanced PrYRV are likely explained by a westward shift in the position of the subtropical high. The link between PrYRV interannual variability and subtropical high movement has been established in numerous studies (e.g., Chang et al. 2000; Lu 2001; Samel and Liang 2003; Huang et al. 2004; Zhou and Yu 2005).

A second area of negative correlations at both 500 and 850 hPa (Figs. 3b,c) is located immediately to the north of the YRV. Thus, in addition to enhanced southerly flow south of the YRV, an increase in PrYRV is accompanied by stronger lower- and middle-tropospheric northerly flow to the north of the YRV. This suggests that greater PrYRV coincides with intensified lower- and middle-tropospheric convergence and vertical ascent along the YRV. A similar result was found by Liang and Wang (1998). The increase in upward motion along the YRV inferred by Figs. 3b and 3c explains the occurrence of significant positive correlations between PrYRV and local RH850 (Fig. 3d). This occurs in conjunction with large negative correlations with RH850 over the South China Sea and subtropical west Pacific and suggests that the Hadley circulation is suppressed. This is confirmed by the existence of strong positive correlations (0.45–0.65) of PrYRV with 500-hPa geopotential height (not shown) over a broad region (10°–28°N, 90°–150°E), indicating persistent local anticyclonic circulation anomalies due to the westward extension or intensification of the North Pacific subtropical high. A GCM sensitivity study by Shen et al. (2001) suggested that the anticyclonic anomalies in the subtropical western Pacific were responsible for the 1998 YRV record flood.

Figure 4 shows the observed correlation pattern between JJA mean SST in the North Pacific and PrYRV. A broad region of significant positive correlations is located over the South China Sea and subtropical west Pacific while a region of significant negative values is found over the east Pacific. Isolated regions of marginally significant positive correlations appear in the Bay of Bengal and tropical Indian Ocean (TIO). Numerous studies have highlighted the important effects of TIO SST on climate anomalies over East Asia and the northwest Pacific during the summer following an El Niño event (Shen et al. 2001; Yang et al. 2007; Chowdary et al. 2009; Xie et al. 2009, 2010). The SST signals identified with summer PrYRV interannual variations, however, are much smaller in the TIO than those in the North Pacific, and hence the latter will be the main focus in the present study.

Fig. 4.

Geographic distributions of the summer interannual correlations (×10) of the YRV precipitation with pointwise SST observed during 1979–2000. Outlined are the two key subtropical Pacific centers with high correlations, where the SSTw and SSTe indices are calculated. Contour intervals are 2 units. Shading areas denote where correlations are statistically significant at the 95% confidence level.

Fig. 4.

Geographic distributions of the summer interannual correlations (×10) of the YRV precipitation with pointwise SST observed during 1979–2000. Outlined are the two key subtropical Pacific centers with high correlations, where the SSTw and SSTe indices are calculated. Contour intervals are 2 units. Shading areas denote where correlations are statistically significant at the 95% confidence level.

To examine the relationship between Pacific SST and PrYRV more closely, area-averaged JJA SST time series were constructed for the domains with the largest positive and negative correlations. The region in the west Pacific with the largest positive correlations (11°–22°N, 112°–128°E) will be referred to as SSTw, and the area in the east Pacific with the largest negative correlations (25°–38°N, 127°–144°W) will be called SSTe. The observed PrYRV, SSTw, and SSTe time series are shown in Fig. 5. The PrYRV relationships with the SST anomalies are highly significant, where the correlations are negative for SSTee (−0.66) and positive for SSTw (+0.58).

Fig. 5.

Interannual variations during 1979–2000 of summer anomalies for the YRV precipitation (mm day−1), and subtropical west and east Pacific SST (°C).

Fig. 5.

Interannual variations during 1979–2000 of summer anomalies for the YRV precipitation (mm day−1), and subtropical west and east Pacific SST (°C).

Note that two major YRV floods (1983 and 1998) occurred in the summer following the mature phase of strong El Niño events (1982/83 and 1997/98). Numerous studies (e.g., Wu et al. 2003; Wang et al. 2009; Wu et al. 2009a,b) have found that precipitation anomalies are more predictable during the summer of a decaying El Niño. This seems to support the concept of TIO SST warming that persists through the summer that follows El Niño spring dissipation and acts as a capacitor that anchors atmospheric anomalies over the Indo–western Pacific Oceans (Xie et al. 2009; Chowdary et al. 2009). As a result, summer rainfall decreases over the subtropical Northwest Pacific but increases over the East Asian monsoon (Mei-yu or Baiu) region (Wang et al. 2000, 2003; Yang et al. 2007; Xie et al. 2010; Chowdary et al. 2010). There are, however, a larger number of cases that lack this correspondence. YRV rainfall was normal in the summer of 1987, 1992, and 1995, prior to which occurred modest, strong, and modest El Niño events (1986/87, 1991/92, and 1994/95), respectively. In addition, the record flood in 1980 and heavy precipitation during 1996 were led by El Niño signatures that were quite weak. Figure 4 clearly shows that YRV summer rainfall anomalies exhibit a much closer correspondence with SSTe and SSTw. This does not exclude the TIO forcing mechanism, which may be linked with the anomalies in the west Pacific. As demonstrated by Wu et al. (2010), from June to August, the SST forcing gradually weakens in the west Pacific but is enhanced in the TIO. This linkage can be realized, for example, as an integral part of the uniform tropospheric warming that dominates the tropics in both observations and GCM simulations in response to the El Niño forcing (Liang et al. 1997; Xie et al. 2009).

Figure 6 shows geographic correlation distributions between the SSTe anomaly time series and Eurasian circulation variables. A broad band of negative correlations with U200 (Fig. 6a) occurs along an east–west axis that is positioned just to the north of the YRV. Within this band, significant values are located over east-central and west-central China. On the other hand, bands of positive correlations are found both to the north and south, where the values within the broad southern band are significant. This pattern reveals that the East Asian jet advances toward the YRV when SSTe anomalies are negative. Meanwhile, tropical easterlies migrate toward the equator, the Hadley circulation weakens and, consequently, convection over the warm pool is suppressed, and the Walker circulation diminishes. This causes increased easterly flow over the Indian and Pacific Oceans, which corresponds to the area of significant positive U200 correlations.

Fig. 6.

As in Fig. 3, but correlations are calculated with SST anomalies over the subtropical east Pacific. (a) Outlined are the two dipole centers with high correlations, where the ULJ index is calculated.

Fig. 6.

As in Fig. 3, but correlations are calculated with SST anomalies over the subtropical east Pacific. (a) Outlined are the two dipole centers with high correlations, where the ULJ index is calculated.

Correlations with both V500 and V850 (Figs. 6b,c), while generally small, reveal two distinct atmospheric responses. First, negative correlations are located to the south of the YRV, while positive values are found to the north. Although the V850 correlations are larger, the overall pattern indicates that negative SSTe anomalies occur in conjunction with increased southerly flow to the south, stronger northerly flow to the north, and greater lower-tropospheric convergence and increased relative humidity (Fig. 6d) along the YRV. These results are consistent with those of Liang and Wang (1998), who found that the indirect transverse circulation caused by the equatorward displacement of the East Asian jet strengthens lower-tropospheric convergence and vertical ascent along the YRV. The second is located at 500 hPa over the subtropical west Pacific, where a small region of significant positive correlations is centered at (20°N, 135°E) while an area of negative values is centered at (13°N, 127°E). A similar response occurs at 850 hPa, although the positive correlation center is not significant. The overall pattern reveals that, when SSTe is negative, anticyclonic flow increases over the west Pacific. The enhanced subtropical high over the west Pacific favors increased clear conditions and greater solar radiation reaching the surface. This, in turn, warms the local SST.

Figure 6 indicates that SSTe teleconnects strongly with the East Asian jet over China. Negative SSTe anomalies are identified with an equatorward shift of the jet toward the YRV, where the jet transverse circulation causes southerly (northerly) winds to strengthen south (north) of the YRV. This leads to increased lower-tropospheric convergence, vertical ascent, and relative humidity along the YRV. These teleconnections between SSTe and the Eurasian circulation are very similar to the relationships described between PrYRV and the Eurasian circulation in Fig. 3, where the equatorward movement of the East Asian jet is linked with greater PrYRV.

Spatial correlations between SSTw anomalies and the Eurasian circulation are shown in Fig. 7. The U200 pattern (Fig. 7a), while almost opposite in phase with that shown in Fig. 6a, has significant correlations which are more localized. This suggests that SSTw anomalies have a smaller impact on East Asian jet location and intensity than do SSTe variations. In contrast, a large region of significant positive correlations with V500 (Fig. 7b) extends from the South China Sea to south China. The V850 pattern (Fig. 7c) has a similar phase, but significant correlations are larger and extend further north, to the YRV. Both the V500 and V850 patterns include smaller negative correlations over the west-central Pacific. This dipole structure indicates that the subtropical high moves toward the west when SSTw anomalies are positive. This signal corresponds with the findings of Zhou et al. (2009c). Westward movement of the subtropical high strengthens the southerly flow in summer over southeastern China (Lu 2001; Zhao et al. 2007). In spite of the positive relationship with the meridional wind south of the YRV, there exists no significant correlation with RH850 over this region (Fig. 7d). On the other hand, an extensive band of significant negative RH850 correlations is located south of 20°N. This is associated with increased descent that results from westward movement of the subtropical high.

Fig. 7.

As in Fig. 3, but correlations are calculated with SST anomalies over the subtropical west Pacific. (c) Outlined is the center with high correlations, where the LLJ index is calculated.

Fig. 7.

As in Fig. 3, but correlations are calculated with SST anomalies over the subtropical west Pacific. (c) Outlined is the center with high correlations, where the LLJ index is calculated.

Figure 7 shows that SSTw anomalies teleconnect more strongly with the regional circulation than they do with the large-scale East Asian jet. Positive SSTw anomalies are associated with westward movement of the subtropical high and intensification of the LLJ to the south of the YRV. The Eurasian circulation teleconnections with SSTw are very similar to those with PrYRV identified in Fig. 3.

The patterns shown in Figs. 3, 6, and 7 reveal that PrYRV anomalies are most likely to be positive when both SSTe is negative and SSTw is positive. The resulting teleconnections with the Eurasian circulation cause the East Asian jet to shift toward the YRV while the subtropical high moves to the west. The YRV is located downstream of the jet core, where the composite analysis shows that the upper-level jet exit region both intensifies and moves south toward the YRV during summers with heavy PrYRV. The resulting indirect transverse circulation causes southerly flow to increase south of the YRV (Liang and Wang 1998). The composite analysis also shows that 200-hPa easterlies increase along the south China coast and adjacent areas of the Pacific Ocean. This occurs in response to the westward movement of the subtropical high, which further contributes to enhanced southerly flow. Both changes cause convergence, vertical ascent, and precipitation increase along the YRV. Previous studies (Wu et al. 2003; Wu et al. 2009a) have found this atmospheric response to SST to occur during the summer that follows the decay of an El Niño event. While this is true for a majority of El Niño events, there are still many cases, as discussed earlier (see Fig. 5), where the correspondence was not observed.

5. AMIP II model SST-forced interannual teleconnections

The purpose of this section is to ascertain the ability of the AMIP II models to reproduce observed teleconnections between Pacific SST anomalies and the Eurasian circulation features that explain PrYRV interannual variability.

Figure 8 is a bar plot that, for each AMIP II simulation, shows the correlation between observed and model PrYRV (black) anomalies. Observed and model correlations of PrYRV with SSTe (hatched) and SSTw (white) are also shown. Correlations between PrYRV anomaly time series reveal that only the MPI.3 realization produces a significant positive relationship (+0.44). The correlation is also large for IPSL.1 (+0.36), but not significant. The remaining models have much smaller values. This indicates that the AMIP II models generally cannot be used for direct comparisons with observed PrYRV and agrees with Zhou et al. (2008), who found that AMIP-type models have very little skill in reproducing observed precipitation over East Asia, including China. However, the possibility exists that the GCMs are able to simulate observed teleconnections between Pacific Ocean SST anomalies and the Eurasian circulation features that explain PrYRV interannual variability.

Fig. 8.

Interannual correlations between observed and model PrYRV (black) anomalies, as well as observed (OBS) and AMIP simulated correlations of PrYRV with SSTe (hatched) and SSTw (white). The simulation labels follow the convention of Fig. 1.

Fig. 8.

Interannual correlations between observed and model PrYRV (black) anomalies, as well as observed (OBS) and AMIP simulated correlations of PrYRV with SSTe (hatched) and SSTw (white). The simulation labels follow the convention of Fig. 1.

To determine this possibility, observed and GCM correlations between PrYRV and Pacific SST anomalies are compared. Observations show significant relationships with both SSTe (−0.62) and SSTw (+0.58). However, few of the AMIP simulations produce large correlations. Only the MPI.3 realization generates both significant relationships, negative with SSTe (−0.42) and positive with SSTw (+0.67). Although significant positive correlations with SSTw are also produced in the MRI and UKMO runs, neither realization simulates observed PrYRV variations and the relationship with SSTe. On the other hand, while the IPSL.1 realization has a significant negative relationship with SSTe (−0.61), the correlations with SSTw (+0.36) and observed PrYRV (+0.36) are not significant. Thus, MPI.3 is the only realization that reproduces observed PrYRV interannual variability and is able to simulate the significant observed correlations with both SSTe and SSTw. Note that both the CCSM and IPSL, which are the only two GCMs to forecast the observed PrYRV annual cycle, are unable to replicate observed PrYRV interannual variability and the relationships with both SSTe and SSTw. These results reinforce the finding of Liang et al. (2002) that no correspondence exists between model ability to predict the observed annual precipitation cycle and interannual variability of summer rainfall over east China.

The general failure of the AMIP II models to simulate PrYRV interannual variability may result from the inability of model SST variations to adequately force the atmospheric features that teleconnect with PrYRV. Li et al. (2010) found that GCMs forced with historical SST fields are able to simulate observed variations in the East Asian summer monsoon circulation but essentially fail to reproduce smaller-scale precipitation variations over the monsoon region. This indicates the possibility that the models are better able to capture relationships between PrYRV and specific circulation features. Thus, AMIP II teleconnections between Pacific Ocean SST and PrYRV will be assessed using a pair of circulation indices that are established from observations.

The first index is a measure of East Asian westerly jet intensity and will be called the upper-level jet (ULJ). Figure 3a clearly shows that PrYRV increases when the East Asian jet is displaced equatorward toward the YRV. This jet movement occurs in concert with a weakening of the Walker circulation, which causes increased upper-level easterly flow over the South China Sea and subtropical west Pacific. Figure 6a reveals that the teleconnections over both regions are forced by negative anomalies in the SSTe domain. Thus, the ULJ index is constructed such that positive (negative) values occur in conjunction with positive (negative) anomalies over the northern region and negative (positive) anomalies in the southern region. Based on these criteria, the ULJ is defined to be the time series of area averaged U200 anomalies in the region bounded by (32°–36°N, 110°–130°E) minus those in the region bounded by (18°–22°N, 110°–130°E).

The second index is an indicator of lower-tropospheric southerly flow along and south of the YRV and will be called the low-level jet (LLJ). Figure 3c shows that PrYRV increases when the subtropical high moves to the west of its mean position and causes V850 to strengthen along and south of the YRV. In addition, Fig. 7c indicates that this teleconnection is forced by positive anomalies in the SSTw domain. Thus, the LLJ index is constructed such that positive (negative) values occur in conjunction with positive (negative) V850 anomalies along and south of the YRV. Given this, the LLJ is defined to be the time series of area averaged V850 anomalies in the region bounded by (15°–25°N, 105°–115°E).

Figure 9a is a bar plot that shows observed and model ULJ correlations with PrYRV (black), SSTe (hatched), and SSTw (white) anomalies. The observed significant positive correlation with PrYRV (+0.58) indicates that summer rainfall increases when both the East Asian jet migrates toward the YRV and the Hadley circulation weakens (Fig. 3a). The significant negative correlation with SSTe (−0.52) reveals that these circulation features occur in response to cold SST anomalies over the subtropical east Pacific (Fig. 6a). The positive correlation between ULJ and SSTw (+0.33), however, is not significant. Thus, SST anomalies in the subtropical west Pacific have a less meaningful impact on the East Asian jet.

Fig. 9.

Observed (OBS) and AMIP simulated interannual correlations of (a) ULJ and (b) LLJ indices with PrYRV (black), SSTe (hatched), and SSTw (white) anomalies. The simulation labels follow the convention of Fig. 1.

Fig. 9.

Observed (OBS) and AMIP simulated interannual correlations of (a) ULJ and (b) LLJ indices with PrYRV (black), SSTe (hatched), and SSTw (white) anomalies. The simulation labels follow the convention of Fig. 1.

The model correlations in Fig. 9a show that several of the simulations are able to capture one or more of the observed relationships between PrYRV, ULJ, and SSTe. In particular, the GISS.1, FGOALS.1, FGOALS.2, MPI.2, and MPI.3 realizations produce significant positive correlations between PrYRV and ULJ. However, among these runs, only the FGOALS.1 and MPI.3 generate the observed negative correspondence between ULJ and SSTe. Thus, while several simulations indicate that PrYRV increases when the East Asian jet migrates toward the YRV and the Hadley circulation weakens, only two realizations link these circulation changes with negative SSTe anomalies.

Figure 9b shows observed and model LLJ correlations with PrYRV, SSTe, and SSTw anomalies. The observed significant positive correlation with PrYRV (+0.69) indicates that summer rainfall increases when southerly flow along and south of the YRV intensifies (Fig. 3c). In addition, the very large positive correlation with SSTw (+0.86) means that this circulation feature occurs in response to warm SST anomalies over the subtropical west Pacific (Fig. 7c). On the other hand, the negative correlation between LLJ and SSTe (−0.41) is substantial but not significant. Thus, SSTe anomalies do not have nearly as important an impact on low-level jet intensity as SSTw does.

The model correlations shown in Fig. 9b reveal that several simulations capture one or more of the observed relationships between LLJ, PrYRV, and SSTw. In particular, the CNRM, FGOALS.1, IPSL.2, MIROCh, MPI.1, MPI.3, MRI, and UKMO realizations reproduce the observed positive relationship between LLJ and PrYRV. Among these runs, the MIROCh, MPI.1, MPI.3, and UKMO show the observed positive correspondence between LLJ and SST. This indicates that the AMIP II models possess some skill in simulating both the observed relationship between increased PrYRV and enhanced V850 along and south of the YRV as well as the SSTw anomalies that force this circulation response.

The above comparisons clearly demonstrate that, among the 26 AMIP II realizations, only MPI.3 is able to reproduce observed PrYRV interannual variations that occur in response to specified global SST forcings. This success results from the ability of the model to simulate observed relationships with prominent Eurasian circulation features, including the ULJ and LLJ, as well as their teleconnections with Pacific Ocean SST anomalies. It appears that PrYRV is governed primarily by coherent ULJ and LLJ variations, which act as the atmospheric bridges to remote SSTe and SSTw forcings, respectively. The PrYRV response to global SST anomalies may then be realistically depicted only when both bridges are correctly simulated. Our finding is supported by Sampe and Xie (2010), who identified both the upper westerly jet and the low-level southerlies as the essential environmental forcing mechanisms for the Mei-yu-Baiu rainband. This result also concurs with several studies (e.g., Wu et al. 2003; Wu et al. 2009a) that document the existence of a signal between SST forcing and PrYRV during the summer that follows the decay of an El Niño event.

Figure 10 shows that MPI.3 has skill in predicting PrYRV responses to specified global SST forcings. The PrYRV teleconnection pattern with SST anomalies (Fig. 10a) closely resembles observations (Fig. 4), with significant positive (negative) correlations over broad regions of the subtropical west (east) Pacific Ocean. Meanwhile, MPI.3 realistically depicts the overall temporal evolution of PrYRV during 1979–2000 (Fig. 10b). In particular, the model accurately predicts the major YRV summer floods in 1983 and 1998 as well as the severe drought in 1985. However, MPI.3 overestimates precipitation in 1993 and reverses the anomaly signs in 1994 and during the 1980 major flooding event. We speculate that the MPI.3 failure in 1980 may result from model initialization errors or the weak SSTw forcing in observations (Fig. 5).

Fig. 10.

As in Fig. 4, but for the (a) MPI.3 realization; (b) interannual variations during 1979–2000 of summer YRV precipitation anomalies (mm day−1) as observed (OBS) and simulated by the MPI.3.

Fig. 10.

As in Fig. 4, but for the (a) MPI.3 realization; (b) interannual variations during 1979–2000 of summer YRV precipitation anomalies (mm day−1) as observed (OBS) and simulated by the MPI.3.

6. Discussion and conclusions

Our analysis shows that the AMIP II models generally fail to simultaneously predict PrYRV annual cycle and summer interannual variability in response to observed global SST forcings. Only two models (CCSM and IPSL) reproduce the observed annual cycle, but both are unable to capture the interannual variability. On the other hand, among the 26 AMIP II realizations, only MPI.3 correctly simulates the interannual variability. Yet, its annual cycle leads observations by 2 months. This result reinforces the finding of Liang et al. (2002) that no correspondence exists between model ability to predict the observed annual precipitation cycle and interannual variability of summer rainfall over east China.

Results also indicate that AMIP II model spread is substantial for both the PrYRV annual cycle and interannual variability, while initial condition differences critically impact only the latter. The sensitivity of simulated PrYRV interannual variations and teleconnection patterns to initial conditions makes it unlikely that a direct comparison of a single-model realization or ensemble mean with observations can determine GCM predictive skill. Thus, the subsequent focus is to determine the underlying physical mechanisms that explain why the MPI.3 is the only realization to successfully predict PrYRV interannual variability. This is facilitated by correlation analyses to first identify observed teleconnection patterns with regional circulation features and global SST anomalies. Observations reveal two distinct signals: 1) the exit region of the ULJ advances toward the YRV and intensifies when SSTe anomalies are negative, where the associated indirect jet transverse circulation causes convergent flow along the YRV; and 2) the subtropical high moves toward the west when SSTw anomalies are positive, which leads to LLJ intensification south of the YRV. Therefore, PrYRV is most likely to increase when subtropical Pacific SST anomalies are both negative in the east and positive in the west. The resulting movements of the ULJ and subtropical high (associated with LLJ intensification) enhance mass convergence and vertical ascent along the YRV.

Teleconnections between the AMIP II simulations and PrYRV are then assessed using the ULJ and LLJ regional circulation indices that are established from observations. Many simulations capture one or more of the observed relationships between PrYRV, ULJ or LLJ, and SSTe or SSTw. However, only MPI.3 is consistently able to reproduce the observed relationships between PrYRV, ULJ and SSTe, as well as LLJ and SSTw. The MPI.3 realization also realistically simulates overall PrYRV temporal evolution during 1979–2000, including the 1983 and 1998 floods and the 1985 drought. It appears that PrYRV is governed primarily by coherent ULJ and LLJ variations, which act as atmospheric bridges to remote SSTe and SSTw forcings, respectively. The PrYRV response to global SST anomalies may then be replicated only when both bridges are correctly simulated, as in the single MPI.3 realization. The general failure of the remaining AMIP II suite to simulate PrYRV interannual variability in response to global SST forcings may result from model inability to adequately represent the atmospheric bridges that teleconnect with PrYRV.

Note that PrYRV is more closely linked with SSTe and SSTw than with TIO SST anomalies. Although several studies have suggested possible TIO forcing of climate anomalies over East Asia and the northwest Pacific during the summer following an El Niño event (Shen et al. 2001; Yang et al. 2007; Chowdary et al. 2009; Xie et al. 2009, 2010), we found only small areas of marginally significant positive PrYRV correlations with TIO SST variations. Figure 11 depicts the lagged relationships between PrYRV, SSTe, SSTw, and the Niño-3.4 index (5°N–5°S, 170°–120°W). Clearly, PrYRV does not have a direct link with Niño-3.4, where all correlation magnitudes are less than 0.25 regardless of the lag period. Thus, PrYRV predictability from Niño-3.4 is low. The correlations with SSTe and SSTw are also small for the preceding seasons, while they are significant during subsequent seasons; SSTe correlations are −0.64 (JAS) and −0.49 (ASO) and SSTw values are 0.49 (JAS), 0.49 (ASO), and 0.50 (SON). Regarding persistence, the lagged signal is stronger for SSTe than SSTw back to the preceding February–April (FMA) while the relative strength is reversed in subsequent seasons through September–November (SON). The Niño-3.4 persistence is highly skewed toward the seasons that follow JJA. In addition, SSTw correlations with Niño-3.4 from the preceding seasons are significant, ranging from 0.58 [November–January (NDJ)] to 0.45 [March–May (MAM)], while the SSTe values are small for all lag periods. These results indicate that both PrYRV and Niño-3.4 lead SSTw. Liang and Wang (1998) demonstrated that the ULJ fluctuations governing PrYRV are strongly coupled with Southern Oscillation variations and that their interactions tend to precede (follow) El Niño phenomena during October through May (summer). This teleconnection may work with the TIO forcing mechanism to bridge the delayed influence of the PrYRV anomalies and El Niño on SSTw. The actual physical links between these components (monsoon, SSTe, SSTw, TIO and El Niño) are complex and deserve further investigation.

Fig. 11.

Observed lag correlations of interannual variations during 1979–2000 with 3-month running means during the seasons preceding and following the summer (JJA) of the central variable: (a) JJA PrYRV with SSTw, SSTe, and Niño-3.4 at various lags; (b) autocorrelations of SSTw, SSTe, and Niño-3.4; and (c) JJA SSTw and SSTe with Niño-3.4 at various lags.

Fig. 11.

Observed lag correlations of interannual variations during 1979–2000 with 3-month running means during the seasons preceding and following the summer (JJA) of the central variable: (a) JJA PrYRV with SSTw, SSTe, and Niño-3.4 at various lags; (b) autocorrelations of SSTw, SSTe, and Niño-3.4; and (c) JJA SSTw and SSTe with Niño-3.4 at various lags.

The physical mechanisms that explain the occurrence of the two distinct SST signals identified in the Fig. 4 correlation analysis are difficult to discern without conducting comprehensive model sensitivity experiments. However, to derive a plausible interpretation, we performed a composite analysis of 200 and 850 hPa winds and SST during years when PrYRV anomalies were significantly positive (negative) and corresponded with positive (negative) SSTw and negative (positive) SSTe. Based on these criteria (i.e., Fig. 5), the years used for the positive (negative) PrYRV composite were 1983, 1993, and 1998 (1985, 1994, and 1997). Figure 12 illustrates the geographic distributions of the differences between the positive and negative composites as observed and simulated by MPI.3. The SST composite difference pattern closely resembles the correlation map with PrYRV shown in Fig. 4, especially over the SSTw and SSTe regions. This indicates that the composite analysis essentially captures the two signals.

Fig. 12.

Composite differences in summer wind circulations at (a),(c) 200- and (b),(d) 850-hPa between the PrYRV positive (1983, 1993, 1998) and negative (1985, 1994, 1997) extremes as (a),(b) observed and (c),(d) simulated by MPI.3. The wind anomalies are drawn by vectors that are scaled to 4 m s−1. Overlaid are the corresponding SST composite differences, normalized by grid-point standard deviations during 1979–2000, and shaded with the grayscale as shown. Small discrepancies between the observed and simulated SST patterns are due to resolution differences.

Fig. 12.

Composite differences in summer wind circulations at (a),(c) 200- and (b),(d) 850-hPa between the PrYRV positive (1983, 1993, 1998) and negative (1985, 1994, 1997) extremes as (a),(b) observed and (c),(d) simulated by MPI.3. The wind anomalies are drawn by vectors that are scaled to 4 m s−1. Overlaid are the corresponding SST composite differences, normalized by grid-point standard deviations during 1979–2000, and shaded with the grayscale as shown. Small discrepancies between the observed and simulated SST patterns are due to resolution differences.

As the ULJ shifts toward the YRV, observations reveal a distinct dipole circulation pattern over an extensive area of the East Asia-west Pacific sector, with a cyclonic anomaly centered in Northeast China and an anticyclonic anomaly center over Southeast China and the subtropical west Pacific (Figs. 12a,b). The anticyclonic center produces increased clear conditions and greater incident solar radiation which, in turn, warms surface waters over the SSTw region. Meanwhile, as the ULJ migrates toward the YRV, the observed midlatitude long wave pattern is altered, causing an anomalous anticyclonic (cyclonic) circulation to develop over the northeast Pacific (western North America). This indicates that the North American upper-level jet stream shifts toward the west. The enhanced anticyclonic circulation over the northeast Pacific produces anomalous north to northeast flow along the southeast flank of the circulation, which strengthens the California Current. As a result, cold water is advected along the south and southeast flanks of the circulation to produce the observed negative anomaly over the SSTe region.

We speculate that the SSTe signal is an oceanic response to atmospheric forcing that is bridged by the ULJ, which explains anomalous PrYRV and accompanies changes in the midlatitude long wave circulation pattern. This interpretation is consistent with the lagged SSTe correlation with PrYRV that is shown in Fig. 11a. On the other hand, as indicated above, the large positive SSTw anomaly appears to be a local response to clear conditions beneath the intensified subtropical high. Positive SSTw may, in turn, cause the subtropical high to intensify. Thus, the lagged correlation between PrYRV and SSTw shown in Fig. 11a is consistent with Liang and Wang (1998), who demonstrated that the upper level jet fluctuations governing PrYRV are strongly coupled with Southern Oscillation variations and that their interactions tend to precede (follow) El Niño phenomena during October–May (summer).

There are three additional SST anomaly centers in the composite difference map that do not appear in the Fig. 4 correlations maps. The first is a negative region located in the central Pacific along 15°N. This center is associated with 200-hPa southwesterly and 850-hPa northeasterly anomalies, typical of tropical convection responses to local SST forcing. The SST anomaly is clearly shown in the positive composite while very weak in the negative composite. A second negative SST anomaly center is located in the vicinity of Korea and Japan. This center occurs only in the negative composite map. The third SST anomaly is positive and located in the TIO region, where the aerial coverage of the anomaly pattern is much smaller in the negative composite. The lack of opposite SST anomalies with comparable magnitudes in the positive and negative composites may explain why the teleconnection with PrYRV in each of these three regions is absent in the Fig. 4 correlation map.

The MPI.3 composites (Figs. 12c,d) capture the major circulation anomalies over the East Asia–west Pacific sector. This includes the cyclonic anomaly in northeast China and the anticyclonic response in the subtropical west Pacific. It, however, simulates a cyclonic anomaly, opposite to observations, over the northeast Pacific, while producing a realistic response over western North America. For the 6 extreme event years used in the composite analysis, MPI.3 captured PrYRV anomalies during all but 1993 and 1994, when the model substantially overpredicted precipitation (see Fig. 10b). These failures may indicate some model deficiency in capturing air–sea interactions over the midlatitude region. As discussed earlier, the SSTe anomaly may likely be the mixed layer ocean response to the midlatitude circulation pattern forcing induced by the ULJ movement that governs PrYRV variations. The lack of two-way interaction in the AMIP type experiment (see below) may explain why the MPI.3 fails to simulate the response in the North Pacific Ocean. The realistic simulation over western North America suggests that the interactive land surface generates a correct response to the atmospheric forcing.

Our conclusion may be affected by the AMIP experimental design, where observed SST variations are prescribed globally to force the atmospheric responses. This prescription excludes feedback mechanisms that contribute to SST regional variability (i.e., atmosphere forces oceans) and, consequently, Asian monsoon evolution (Meehl and Arblaster 1998; Kitoh and Arakawa 1999; Zhou et al. 2009b). Lau et al. (1996) showed that most AMIP I GCMs are able to predict observed tropical rainfall responses to ENSO-related SST forcing but have very limited skill in the extratropics. Liang et al. (2001, 2002) found that the prescribed SST field limits model ability to simulate realistic teleconnections of east China monsoon precipitation with the large-scale circulation. Wang et al. (2004) also attributed this prescription to the common AMIP failure in reproducing the observed inverse relationship between summer local rainfall and SST anomalies over the Philippine Sea, the South China Sea, and the Bay of Bengal. Fu et al. (2002) and Wu et al. (2006) demonstrated the need to incorporate air–sea interactions for realistic simulation of summer monsoon and rainfall variations in tropical Indo–western Pacific Ocean regions and the midlatitudes. Over these areas, where the atmospheric effect (primarily from negative convection–SST feedback) is significant, the AMIP-type simulations produce excessive SST forcing. Thus, the impact that the prescribed AMIP SST pattern has on the general circulation plays a major role in determining the extent to which the models are able to simulate observed teleconnections with summer PrYRV. We plan to use available fully coupled atmosphere–ocean GCM simulations, following Liang et al. (2008), to revisit the issue and focus on how air–sea interactions actually affect our findings.

Acknowledgments

We thank Jinhong Zhu and Tiejun Ling for help on data processing. We acknowledge LLNL/PCMDI and the modeling groups for making available the AMIP II simulations, and NCSA/UIUC for the computing support. The research was partially supported by the National Natural Science foundation of China Award No. 40875050 to Wang and the National Aeronautics and Space Administration Award NNX08AL94G to Liang. The views expressed are those of the authors and do not necessarily reflect those of the sponsoring agencies or the Illinois State Water Survey.

REFERENCES

REFERENCES
Chang
,
C. P.
,
Y.
Zhang
, and
T.
Li
,
2000
:
Interannual and interdecadal variations of the East Asian summer monsoon and tropical Pacific SSTs. Part I: Roles of the subtropical ridge
.
J. Climate
,
13
,
4310
4325
.
Chen
,
H.
,
T.
Zhou
,
R. B.
Neale
,
X.
Wu
, and
G. J.
Zhang
,
2010
:
Performance of the New NCAR CAM3.5 model in East Asian summer monsoon simulations: Sensitivity to modifications of the convection scheme
.
J. Climate
,
23
,
3657
3675
.
Chowdary
,
J. S.
,
S.-P.
Xie
, ,
J.-J.
Luo
,
J.
Hafner
,
S.
Behera
,
Y.
Masumoto
, and
T.
Yamagata
,
2009
:
Predictability of Northwest Pacific climate during summer and the role of the tropical Indian Ocean
.
Climate Dyn.
,
36
,
607
621
.
Chowdary
,
J. S.
,
S.-P.
Xie
,
J.-Y.
Lee
,
Y.
Kosaka
, and
B.
Wang
,
2010
:
Predictability of summer Northwest Pacific climate in eleven coupled model hindcasts: Local and remote forcing
.
J. Geophys. Res.
,
115
,
D22121
,
doi:10.1029/2010JD014595
.
Ding
,
Y.-H.
,
1994
:
Monsoons over China
.
Kluwer Academic, 419 pp
.
Fu
,
X.
,
B.
Wang
, and
T.
Li
,
2002
:
Impacts of air–sea coupling on the simulation of mean Asian summer monsoon in the ECHAM4 model
.
Mon. Wea. Rev.
,
130
,
2889
2904
.
Gates
,
W. L.
, and
Coauthors
,
1999
:
An overview of the results of the Atmospheric Model Intercomparison Project (AMIP I)
.
Bull. Amer. Meteor. Soc.
,
80
,
29
55
.
Huang
,
R.
,
W.
Chen
,
B.
Yang
, and
R.
Zhang
,
2004
:
Recent advances in studies of the interaction between the East Asian winter and summer monsoons and ENSO cycle
.
Adv. Atmos. Sci.
,
21
,
407
424
.
Hurrell
,
J. W.
,
J. J.
Hack
,
D.
Shea
,
J. M.
Caron
, and
J.
Rosinski
,
2008
:
A new sea surface temperature and sea ice boundary dataset for the Community Atmosphere Model
.
J. Climate
,
21
,
5145
5153
.
Kanamitsu
,
M.
,
W.
Ebisuzaki
,
J.
Woollen
,
S.-K.
Yang
,
J. J.
Hnilo
,
M.
Fiorino
, and
G. L.
Potter
,
2002
:
The NCEP-DOE AMIP-II Reanalysis (R-2)
.
Bull. Amer. Meteor. Soc.
,
83
,
1631
1643
.
Kang
,
I.-S.
, and
Coauthors
,
2002
:
Intercomparison of the climatological variations of Asian summer monsoon precipitation simulated by 10 GCMs
.
Climate Dyn.
,
19
,
383
395
,
doi:10.1007/s00382-002-0245-9
.
Kitoh
,
A.
, and
O.
Arakawa
,
1999
:
On overestimation of tropical precipitation by an atmospheric GCM with prescribed SST
.
Geophys. Res. Lett.
,
26
,
2965
2968
.
Lau
,
K.-M.
,
Y.
Sud
, and
J.
Kim
,
1996
:
Intercomparison of hydrologic processes in AMIP GCMs
.
Bull. Amer. Meteor. Soc.
,
77
,
2209
2227
.
Lau
,
N.-C.
, and
J. J.
Ploshay
,
2009
:
Simulation of synoptic- and subsynoptic-scale phenomena associated with the East Asian summer monsoon using a high-resolution GCM
.
Mon. Wea. Rev.
,
137
,
137
160
.
Li
,
H.
,
A.
Dai
,
T.
Zhou
, and
J.
Lu
,
2010
:
Responses of East Asian summer monsoon to historical SST and atmospheric forcing during 1950–2000
.
Climate Dyn.
,
34
,
501
514
.
Liang
,
X.-Z.
, and
W.-C.
Wang
,
1998
:
Associations between China monsoon rainfall and tropospheric jets
.
Quart. J. Roy. Meteor. Soc.
,
124
,
2597
2623
.
Liang
,
X.-Z.
,
K. R.
Sperber
,
W.-C.
Wang
, and
A. N.
Samel
,
1997
:
Predictability of SST forced climate signals in two atmospheric general circulation models
.
Climate Dyn.
,
13
,
391
415
.
Liang
,
X.-Z.
,
W.-C.
Wang
, and
A. N.
Samel
,
2001
:
Biases in AMIP model simulations of the east China monsoon system
.
Climate Dyn.
,
17
,
291
304
.
Liang
,
X.-Z.
,
A. N.
Samel
, and
W.-C.
Wang
,
2002
:
China rainfall interannual predictability: Dependence on the annual cycle and surface anomalies
.
J. Climate
,
15
,
2555
2561
.
Liang
,
X.-Z.
,
J.
Zhu
,
K. E.
Kunkel
,
M.
Ting
, and
J. X. L.
Wang
,
2008
:
Do CGCMs simulate the North American monsoon precipitation seasonal-interannual variability?
J. Climate
,
21
,
4424
4448
.
Lu
,
R.
,
2001
:
Interannual variability of the summertime North Pacific subtropical high and its relation to atmospheric convection over the warm pool
.
J. Meteor. Soc. Japan
,
79
,
771
783
.
Meehl
,
G. A.
, and
J.
Arblaster
,
1998
:
The Asian–Australian monsoon and El Niño–Southern Oscillation in the NCAR Climate System Model
.
J. Climate
,
11
,
1356
1385
.
Samel
,
A. N.
, and
X.-Z.
Liang
,
2003
:
Understanding relationships between the 1998 Yangtze River flood and northeast Eurasian blocking
.
Climate Res.
,
23
,
149
158
.
Samel
,
A. N.
,
W.-C.
Wang
, and
X.-Z.
Liang
,
1999
:
Monsoon rainfall over China and relationships with the Eurasian circulation
.
J. Climate
,
12
,
115
131
.
Sampe
,
T.
, and
S. P.
Xie
,
2010
:
Large-scale dynamics of the Meiyu-Baiu rainband: Environmental forcing by the westerly jet
.
J. Climate
,
23
,
113
134
.
Shen
,
X.
,
M.
Kimoto
,
A.
Sumi
,
A.
Numaguti
, and
J.
Matsumoto
,
2001
:
Simulation of the 1998 East Asian summer monsoon by the CCSR/NIES AGCM
.
J. Meteor. Soc. Japan.
,
79
,
741
757
.
Wang
,
B.
,
R.
Wu
, and
X.
Fu
,
2000
:
Pacific–East Asian teleconnection: How does ENSO affect East Asian climate?
J. Climate
,
13
,
1517
1536
.
Wang
,
B.
,
R.
Wu
, and
K. M.
Lau
,
2001
:
Interannual variability of the Asian summer monsoon: Contrasts between the Indian and the western North Pacific–East Asian monsoons
.
J. Climate
,
14
,
4073
4090
.
Wang
,
B.
,
R.
Wu
, and
T.
Li
,
2003
:
Atmosphere–warm ocean interaction and its impacts on the Asian–Australian monsoon variation
.
J. Climate
,
16
,
1195
1211
.
Wang
,
B.
,
I. S.
Kang
, and
J. Y.
Lee
,
2004
:
Ensemble simulations of Asian–Australian monsoon variability by 11 AGCMs
.
J. Climate
,
17
,
803
818
.
Wang
,
B.
, and
Coauthors
,
2009
:
Advance and prospectus of seasonal prediction: assessment of the APCC/CliPAS 14-model ensemble retrospective seasonal prediction (1980–2004)
.
Climate Dyn.
,
33
,
93
117
,
doi:10.1007/s00382-008-0460-0
.
Webster
,
P. J.
, and
S.
Yang
,
1992
:
Monsoon and ENSO: Selectively interactive systems
.
Quart. J. Roy. Meteor. Soc.
,
118
,
877
926
.
Wu
,
B.
,
T.
Zhou
, and
T.
Li
,
2009a
:
Seasonally evolving dominant interannual variability modes of East Asian climate
.
J. Climate
,
22
,
2992
3005
.
Wu
,
B.
,
T.
Zhou
, and
T.
Li
,
2009b
:
Contrast of rainfall–SST relationships in the western North Pacific between the ENSO-developing and ENSO-decaying dummers
.
J. Climate
,
22
,
4398
4405
.
Wu
,
B.
,
T.
Li
, and
T.
Zhou
,
2010
:
Relative contributions of the Indian Ocean and local SST anomalies to the maintenance of the western North Pacific anomalous anticyclone during the El Niño decaying summer
.
J. Climate
,
23
,
2974
2986
.
Wu
,
R.
,
Z.-Z.
Hu
, and
B. P.
Kirtman
,
2003
:
Evolution of ENSO-related rainfall anomalies in East Asia
.
J. Climate
,
16
,
3742
3758
.
Wu
,
R.
,
B. P.
Kirtman
, and
K.
Pegion
,
2006
:
Local air–sea relationship in observations and model simulations
.
J. Climate
,
19
,
4914
4932
.
Xie
,
S.-P.
,
K.
Hu
,
J.
Hafner
,
H.
Tokinaga
,
Y.
Du
,
G.
Huang
, and
T.
Sampe
,
2009
:
Indian Ocean capacitor effect on Indo–western Pacific climate during the summer following El Niño
.
J. Climate
,
22
,
730
747
.
Xie
,
S.-P.
,
Y.
Du
,
G.
Huang
,
X.-T.
Zheng
,
H.
Tokinaga
,
K.
Hu
, and
Q.
Liu
,
2010
:
Decadal shift in El Nino influences on Indo–western Pacific and East Asian climate in the 1970s
.
J. Climate
,
23
,
3352
3368
.
Yang
,
J.
,
Q.
Liu
,
S.-P.
Xie
,
Z.
Liu
, and
L.
Wu
,
2007
:
Impact of the Indian Ocean SST basin mode on the Asian summer monsoon
.
Geophys. Res. Lett.
,
34
,
L02708
,
doi:10.1029/2006GL028571
.
Yu
,
R.
,
W.
Li
,
X.
Zhang
,
Y.
Liu
,
Y.
Yu
,
H.
Liu
, and
T.
Zhou
,
2000
:
Climatic features related to eastern China summer rainfalls in the NCAR CCM3
.
Adv. Atmos. Sci.
,
17
,
503
518
.
Zhao
,
P.
,
R.-H.
Zhang
,
J.-P.
Liu
,
X.-J.
Zhou
, and
J.-H.
He
,
2007
:
Onset of southwesterly wind over eastern China and associated atmospheric circulation and rainfall
.
Climate Dyn.
,
28
,
797
811
.
Zhou
,
T.
, and
Z.
Li
,
2002
:
Simulation of the East Asian summer monsoon by using a variable resolution atmospheric GCM
.
Climate Dyn.
,
19
,
167
180
.
Zhou
,
T.
, and
R.
Yu
,
2005
:
Atmospheric water vapor transport associated with typical anomalous summer rainfall patterns in China
.
J. Geophys. Res.
,
110
,
D08104
,
doi:10.1029/2004JD005413
.
Zhou
,
T.
,
R.
Yu
,
H.
Li
, and
B.
Wang
,
2008
:
Ocean forcing to changes in global monsoon precipitation over the recent half-century
.
J. Climate
,
21
,
3833
3852
.
Zhou
,
T.
, and
Coauthors
,
2009a
:
The CLIVAR C20C Project: Which components of the Asian–Australian Monsoon circulation variations are forced and reproducible?
Climate Dyn.
,
33
,
1051
1068
.
Zhou
,
T.
,
B.
Wu
, and
B.
Wang
,
2009b
:
How well do atmospheric general circulation models capture the leading modes of the interannual variability of the Asian–Australian monsoon?
J. Climate
,
22
,
1159
1173
.
Zhou
,
T.
, and
Coauthors
,
2009c
:
Why the western Pacific subtropical high has extended westward since the late 1970s
.
J. Climate
,
22
,
2199
2215
.