Simulating the Intraseasonal Variation of the East Asian Summer Monsoon by IAP AGCM4.0

A distinct monsoonal climate is observed in East Asia with a northerly prevailing in winter and a southerly prevailing in summer. Climatically, the East Asian summer monsoon (EASM) migrates northward in a stepwise fashion, characterized by two northward jumps of the western Pacific subtropical high (WPSH) and the related rain belt, with the first jump in mid-June and the second in late July (Tao and Chen, 1987; Ding, 2007; Lau and Ploshay, 2009). Compared with the first jump, the second is much more remarkable, resulting in a totally different circulation in East Asia after mid-July (Su and Xue, 2011). The EASM shifts to its most northern position by mid-August. Afterwards, it retreats quickly and the winter monsoon circulation begins to establish. Summer rainfall in China, therefore, can be divided into three distinct stages: the pre-flood period in southern China, the Meiyu period over the Yangtze-Huihe River basin, and the rainy season in northern China (Tao and Chen, 1987; Ding, 2007).

Much attention has been paid to the northward jumps of the WPSH due to their dominant roles in regulating summer monsoon rainfall in East Asia. It has been revealed that the anomalous convection over the western Pacific warm pool can excite a Rossby wave-train propagating northeastward, thus influencing the longitudinal displacement of the WPSH and summer rainfall anomalies in East Asia (Nitta, 1987; Huang and Wu, 1989; Lu, 2001). In addition, the northward jumps of the WPSH can also be influenced by Rossby waves in high latitudes through downstream energy dispersion (Enomoto et al., 2003). More recently, it has further been found that the northward jumps are jointly affected by both factors through their phase-locking (Bueh et al., 2008; Suzuki and Hoskins, 2009; Su and Xue, 2010).

The complex intraseasonal variations of the EASM provide a more rigorous standard for the evaluation of general circulation models (GCMs). Although the climatic features of the Asian monsoon, including the EASM, are basically simulated by most of the models participating the Atmospheric Model Intercomparison Project (AMIP) and the Monsoon GCM Intercomparison Project (Gates, 1992; Sperber and Palmer, 1996; Gadgil and Sajani, 1998; Kang et al., 2002; Zhou et al., 2009), most of them fail to capture the intraseasonal features of the EASM. In general, the northward advance of the EASM and the associated rain belt are not well reproduced in GCMs with resolutions lower than 200 km (Lau and Ploshay, 2009). A similar problem also exits in the early GCMs developed at the Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences (Wang and Bi, 1996; Lin and Zeng, 1997; Xue et al., 2001).

In recent years, horizontal resolutions of GCMs have increased greatly with the development of computer technology and numerical model designs. It is anticipated that the increased resolution will be helpful for better depicting the East Asian rain belt related to the northward advance of the EASM (Kitoh and Kusunoki, 2008; Lau and Ploshay, 2009; Oouchi et al., 2009). Recently, the fourth generation of the IAP atmospheric general circulation model (IAP AGCM4.0, hereafter IAP4) has been developed by (Zhang, 2009), with a much higher resolution compared with the former versions. The new model shows a much better performance in simulating the basic features of global atmospheric circulation (Zhang, 2009).

Based on model outputs for the period 1979-2008, we evaluate the model's performance in simulating the intraseasonal variation of the EASM, with a focus on the northward jumps of the WPSH. The remainder of the paper is organized as follows. Data and methods are introduced in section 2. In section 3, we analyze the intraseasonal variation of the EASM, with a focus on the northward jumps of the WPSH in section 4. Section 5 discusses the related biases, and finally a summary is given in section 6.

The dynamical framework of IAP4 is inherited from the former IAP AGCMs (Liang, 1996; Zhang, 1990), with some newly introduced techniques or schemes, such as permissible substitution, a flexible leaping-point scheme in high latitudes, a time-splitting method, and a semi-Lagrangian vapor transport scheme (Zhang, 2009). The model has a horizontal resolution of 1.4^#cod#x000b0;#cod#215; 1.4^#cod#x000b0; and 26 vertical layers, with the top level at 2.2 hPa. The suite of physical processes incorporated in this model is taken from version 3.5 of the Community Atmosphere Model of the National Center for Atmospheric Research (NCAR CAM3.5), with a modified Zhang-McFarlane parameterization scheme (Zhang and McFarlane, 1995). The model is driven by observed sea surface temperature (SST) and sea ice from 1 January 1978 to 31 December 2008. Considering the spin-up process in the model, we adopt the last 30 years of data for evaluation.

The primary dataset used in this study is the second version of the daily reanalysis product from the National Centers for Environmental Prediction and the National Center for Atmospheric Research (NCEP/NCAR; Kanamitsu et al., 2002), including surface air temperature, geopotential height and wind fields. Outgoing long-wave radiation (OLR), which is used to infer the convection intensity, is derived from satellite observations of the National Oceanic and Atmospheric Administration (NOAA) (Liebmann and Smith, 1996). Precipitation data are taken from the Climate Prediction Center merged analysis of precipitation (CMAP; Xie and Arkin, 1997). The above datasets cover the period from 1979 to 2008 with a resolution of 2.5^#cod#x000b0;#cod#215; 2.5^#cod#x000b0;. Besides, all daily data are pretreated to the pentad mean for convenience.

As proposed by (Zhao, 1999), the 500-hPa WPSH can be well described by the following indices based on the pentad mean geopotential height: (1) the ridgeline index: the latitude of the ridgeline averaged over 110^#cod#x000b0;-150^#cod#x000b0;E; (2) the westward extension index: the westernmost longitude along the 5880 gpm contour over 90^#cod#x000b0;E-180^#cod#x000b0;; (3) the area index: the area within the 5880 gpm contour over the region 5^#cod#x000b0;-45^#cod#x000b0;N and 90^#cod#x000b0;E-180^#cod#x000b0;; (4) the northern borderline index: the northernmost latitude of the 5880 gpm contour averaged over 110^#cod#x000b0;-150^#cod#x000b0;E.

It is important to note that, instead of the 5880 gpm contour in the observation, the reference contour in the model is deliberately set to 5900 gpm due to the overestimation of geopotential height by about 20 gpm (Zhang, 2009).

To depict the abruptness of the wind field, we employ the method of normalized finite temporal variation (NFTV) proposed by (Zeng et al., 2005). A maximum of NFTV indicates an abruptness at a particular time. The method is described in detail as follows.

(1) Inner product and norm: Considering a scalar or vector variable as a function of space and time F(#cod#952;,#cod#923;,p,t), given an isobaric surface p and a region on a sphere marked with S, namely (#cod#952;,#cod#923;)#cod#x02208;S, where #cod#952; is the co-latitude and #cod#923; is the longitude, then the variables at time t₁ and t₂ are symbolized with F₁=F(#cod#952;,#cod#923;,p,t₁) and F₂=F(#cod#952;,#cod#923;,p,t₂). Therefore, the inner product (F₁, F₂) and the norm ||F|| can be defined as follows:

where S is usually assigned with an area of 10^#cod#x000b0;#cod#215; 10^#cod#x000b0; in the calculation.

(2) Normalized finite temporal variation:

where the time t=(t₁+t₂)/2 and #cod#964;_d=t₂-t₁; the value of #cod#964;_d is generally assigned as five days. It is easily inferred that values of d² range from 0 to 2. If F has no change, then d²=0; if F is opposite in sign between the consecutive time-interval, then d²=2. Before calculating, the different variables are processed using the five-day running mean.

Figure 1 shows the time-latitude diagrams of the observed and simulated pentad-mean precipitation averaged over 110^#cod#x000b0;-130^#cod#x000b0;E. While the tropical rain belt exhibits a clear intraseasonal oscillation, the subtropical rain belt in East Asia migrates northward in a stepwise way, characterized by slow movements and two northward jumps. Based on the criterion that the rain belt moves northward at least three degrees in latitude within one pentad, we identify that the jump occurs at Pentad 34 (P34 for short, 15-19 June) and P41 (20-24 July), respectively (Fig. 1a). The model generally simulates the observed intraseasonal variation in the tropics and subtropics (Fig. 1b), except that rainfall is systemically underestimated, as also seen in the monthly and seasonal means (Zhang, 2009). Furthermore, the simulated second jump is not as evident as the first. This is probably due to the rain belt being located at relatively high latitudes after the first jump.

Figure 1.  Latitude-time cross section of pentad mean precipitation averaged over 110^#cod#x000b0;-130^#cod#x000b0;E (units: mm d^-1): (a) observation; (b) simulation. The bold dashed lines indicate the jump stages in the observation and simulation.

In addition, there also exist some biases in the intraseasonal evolution. The observation shows that two maxima of precipitation appear over the tropical western Pacific in P34 (15-19 June) and P41 (20-24 July), respectively (Fig. 1a). Although the two maxima are captured by the model, the first maximum occurs in P36 (25-29 June), two pentads later than CMAP, and the second occurs in P40 (15-19 July), one pentad earlier than CMAP (Fig. 1b). In other words, the interval between the two maxima is shorter than the observation. Similarly, the interval between the two jumps of the East Asian rain belt is shorter as well.

It can also be found in Fig. 1b that the rain belt in the tropics and subtropics is located more northward during May through September, especially in May and June. This period is often called the pre-flood period, when heavy precipitation is observed in southern China. In contrast, the observed center is almost absent in the model in southern China. Instead, the model simulates a center in the Yangtze River basin. Furthermore, the observed rain belt in East Asia starts to retreat by mid-August, while the simulated 4 mm d^-1 contour does not disappear until mid-September. Therefore, the simulated EASM tends to maintain for one month longer than the observation.

Figure 2.  Time series of various indices of the WPSH for the observation (solid circles with solid line) and simulation (open circles with dashed line): (a) the ridgeline index (units: ^#cod#x000b0;N); (b) the westward extension index (units: ^#cod#x000b0;E); (c) the northern borderline index (units: ^#cod#x000b0;N); (d) the area index (10⁶ km²). The dashed circles in (a) indicate the jump stages in the observation and simulation.

Figure 3.  Height-time cross section of the normalized finite temporal variation of wind for the observation (left panels) and simulation (right panels) averaged over the different regions (units: dimensionless): (a, b) East Asia (20^#cod#x000b0;-45^#cod#x000b0;N, 110^#cod#x000b0;-140^#cod#x000b0;E); (c, d) South China Sea (5^#cod#x000b0;-20^#cod#x000b0;N, 110^#cod#x000b0;-120^#cod#x000b0;E); (e, f) western Pacific warm pool (5^#cod#x000b0;-20^#cod#x000b0;N, 120^#cod#x000b0;-150^#cod#x000b0;E).

Figure 2 shows the temporal evolution of various indices of the WPSH. The simulated ridgeline index and northern borderline index are in agreement with the observation, except that both indices are located somewhat northward (Figs. 1a and c). The ridgeline index, in particular, is shifted more northward by 5^#cod#x000b0; in latitude. Similar to the rain belt shown in Fig. 1, the simulated WPSH exhibits two jumps with a shorter interval than the observation.

The simulated western extension index and area index are relatively poor (Figs. 2b and d). Although the variations before mid-July are generally simulated, the model fails to simulate a rapid eastward retreat and a sharp reduction in intensity after mid-July. Compared with June and July, the observed WPSH in August retreats eastward with a weak intensity. The simulated WPSH, however, tends to be located more northwestward with a stronger intensity. As a result, there still exists a rain belt in August in East Asia (Fig. 1b).

To investigate the intraseasonal variation of the EASM in three dimensions, Fig. 3 shows the temporal evolution of normalized finite temporal variation (NFTV). The model captures the main features of the intraseasonal evolution in East Asia, such as the maximum in the lower stratosphere in June, the maximum in the whole troposphere during mid-July, and the rapid adjustment of the tropospheric circulation after mid-July (Fig. 3b). Overall, the model simulates the whole trend varying from a small value to a large one.

We also note that there are some discrepancies between the observation and simulation. While the observed maximum after mid-July is located below 400 hPa, the simulated one is confined to the middle and upper troposphere. In contrast, the NFTV in the lower troposphere is much smaller in the model. This is attributed to an unusually stronger WPSH (Fig. 2d) and a later retreat of the EASM in August (Fig. 1b).

The circulation adjustments in the tropics, such as the South China Sea and the warm pool region, are confined to the lower level, especially after mid-July (Figs. 3c-f). Compared with East Asia, the simulated intensity and temporal evolution agree well with the observation, indicating that an AGCM driven by the observed SST has a better performance in the tropics than that in the mid and high latitudes (Sperber and Palmer, 1996).

As discussed above, three biases related to the EASM intraseasonal variations are identified in the model: (1) an absence of the pre-flood period in South China and an earlier beginning of the Meiyu period in the Yangtze-Huaihe River basin; (2) a later retreat of the EASM due to an unusually stronger WPSH in August; (3) a shorter interval between the two jumps of the WPSH. In the following, we discuss the first and second of these three biases; the third is discussed in the next section.

Figure 4.  (a) Observed and (b) simulated precipitation averaged over the period from 16 May to 14 June (units: mm d^-1). (c) The WPSH represented by 5880, 5890 gpm for the observation (solid) and 5890, 5900 gpm for the simulation (dashed) over the same period, respectively (units: gpm).

Figure 5.  Latitude-time cross section of the (a) observed and (b) simulated land-sea thermal contrast as represented by the difference of surface air temperature between 150^#cod#x000b0;E and 110^#cod#x000b0;E. (c) The corresponding difference between simulation and observation.

Figure 6.  The (a) observed and (b) simulated 850-hPa wind fields; and (c) the corresponding difference between simulation and observation (units: m s^-1).

Figure 4 shows the observed and simulated precipitation and the WPSH averaged from 16 May to 14 June, a typical pre-flood period in southern China. A precipitation center is observed in southern China and the South China Sea, as shown by the 6 mm d^-1 contour (Fig. 4a). In the model, however, rainfall in southern China during this period is underestimated significantly. The discrepancy in southern China is related to the WPSH. As shown in Fig. 4c, the observed WPSH has retreated to the east of the South China Sea during this period, but the simulated WPSH still occupies the South China Sea.

The East Asian monsoon is primarily driven by land-sea thermal contrast (Zhou and Zou, 2010). Figure 5 shows the difference of surface air temperature between 150^#cod#x000b0;E and 110^#cod#x000b0;E. In general, the land-sea temperture contrast tends to change from a positive value in winter to a negtive one in summer, leading to a northerly in winter and a southerly in summer over East Asia (Fig. 5a). Although this sesonal variation is reasonably simulated by IAP4 (Fig. 5b), a negative bias is still evident in summer, particularly in August, as indicated by a minimum from low to high latitudes (Fig. 5c). Accordingly, the western Pacific anticyclone extends much farther westward with a notable southerly bias in East Asia (Figs. 6b and c), thereby resulting in a later retreat of the EASM than the observation.

During the northward jump of the WPSH, there exhibits a remarkable change in the circulation pattern, which can be well represented by the inter-pentad difference between the jumping pentad and the previous pentad (Su and Xue, 2010). Since the jump time in the model is different from the observation, the inter-pentad difference is taken at a different time. During the first jump stage, for example, P34 (15-19 June) minus P33 (10-14 June) is selected for the observation, while P36 (25-29 June) minus P35 (20-24 June) is selected for the simulation.

Figure 7 shows the observed and simulated inter-pentad differences of 850-hPa wind and OLR during the jump stages. Due to enhanced convection over the tropical western Pacific at P34 (15-19 June), a Rossby wave-train appears, as represented by a cyclonic or anticyclonic anomaly alternating from the Philippines to the Bering Strait (Fig. 7a). This feature is generally simulated by the model, except that the propagation path is located more westward because the observed OLR anomaly near the Philippines shifts to the South China Sea in the model (Fig. 7b). Different from the first jump, the second one is not only related to the propagation of the Rossby wave-train from the warm pool region, but also to the downstream propagation of Rossby waves in high latitudes of Eurasia (Fig. 7c). The model simulates the second jump well due to a good representation of the phase-locking of the two factors mentioned by (Su and Xue, 2010) (Fig. 7d). It is also noted that the second jump is not as evident as that in the observation.

Figure 7.  The inter-pentad differences of 850-hPa wind (units: m s^-1, vector) and outgoing longwave radiation (OLR) (units: W m^-2, shaded) during the jump stages for observation (left panels) and simulation (right panels): (a) the difference between P34 (15-19 June) and P33 (10-14 June; the observed first jump); (b) the difference between P36 (25-29 June) and P35 (20-24 June; the simulated first jump); (c) the difference between P41 (20-24 July) and P40 (15-19 July; the observed second jump); (d) the difference between P40 (15-19 July) and P39 (10-14 July; the simulated second jump).

The discrepancies of northward jumps of the WPSH between simulation and observation are possibly related to the climatological intraseasonal oscillation (CISO) over the warm pool region in the model. (Wang and Xu, 1997) showed that transient intraseasonal oscillation (TISO) in the tropical western Pacific tends to be phase-locked to the annual cycle during some specific stages, thereby forming the CISO. Therefore, in a certain year, ISO can be viewed as the sum of CISO and TISO, of which CISO accounts for the most variance of ISO. In other words, CISO can reflect the dominant periodicity of ISO in different years.

Using the same method, Kang et al. (1999, 2002) studied the intraseasonal variation of the Asian monsoon and conducted the related model evaluation. The CISO component (20-90 days) of OLR over the warm pool region is extracted from the annual cycle (greater than 90 days) by using the fast Fourier transform. The original series of OLR is preprocessed with 5-day running mean.

Figure 8 shows the observed and simulated OLR in the warm pool region from 1 April to 1 October, together with a power spectrum analysis. The observed OLR over the warm pool decreases from 260 W m^-2 in April to 220 W m^-2 in summer, with the lowest value in August. Although the model simulates this variation well, it overestimates OLR during the whole period (Fig. 8b), showing that weaker convective activities in the tropical western Pacific not only appear in summer, but also appear in other seasons. In particular, there appears a minimum of OLR during the jump stage of the WPSH (mid-June and late July), indicating that the jumps of the WPSH are associated with the CISO in the warm pool (Fig. 8c). A similar phenomenon is also found in the model, but with a shorter interval between the two minimums (Fig. 8d). It can be concluded that the phase disagreement of CISO in the warm pool region leads to the jump bias in the model.

Figure 8.  (a, b) Time series of OLR over the warm pool region (5^#cod#x000b0;-20^#cod#x000b0;N, 120^#cod#x000b0;-150^#cod#x000b0;E) (units: W m^-2) (the solid and dashed lines represent the original curve with 5-day running mean and the base annual cycle, respectively); (c, d) 20-90-day filtered time series of OLR over the region; (e, f) the corresponding power spectrum (solid line) and the standard spectrum of red noise at the 95% confidence level. For the sake of clarity, the observed and simulated power spectra including the significance line are all scaled by 100. The observation and simulation are shown in the left and right panels, respectively. Triangles in (c) and (d) indicate the two jump stages in observation and simulation.

The phase disagreement of CISO between simulation and observation is directly ascribed to a shorter periodicity of CISO in the model. As shown in Fig. 8e, strong powers at 20 and 30 days are exhibited, and the strongest one is at 45 days. In contrast, the model simulates two periodicities (Fig. 8f), one being 20-40 days with a peak of 30 days, and the other being greater than 50 days with a peak of 60 days (not as significant as 20-40 days). Therefore, the simulated periodicity of CISO is about 15 days shorter than the observation.

It can also be seen that CISO in late June reaches a minimum, but it is not in a negative phase. Comparison between Fig. 7c and Fig. 7d shows that the observed OLR minimum during the first jump is shifted westward in the model. Since the selected region in Fig. 8d is the same as the observation (5^#cod#x000b0;-20^#cod#x000b0;N, 120^#cod#x000b0;-150^#cod#x000b0;E), the simulated CISO is accordingly in a positive phase. If the key region in the model is taken as the simulated OLR minimum [Fig. 7d, (5^#cod#x000b0;-20^#cod#x000b0;N, 110^#cod#x000b0;-130^#cod#x000b0;E)], the simulated CISO is obviously in a negative phase during the first jump (Fig. 9b). Besides, the interval between the two OLR minima (marked with triangles in Fig. 9b) is still shorter than the observation (Fig. 8c), and the corresponding power spectrum is similar to that in the warm pool region with a significant power at 30 days (Fig. 9c). Although a bias exists in the simulated OLR minimum during the first jump, the model generally simulates the relationship between CISO in the tropical northwestern Pacific and northward jumps of WPSH and East Asian rain belt.

(Madden and Julian, 1994) pointed out that remarkable intraseasonal oscillations exist in the equatorial convections, known as Madden-Julian oscillation (MJO). Although it is most prominent in boreal winter, the MJO also displays a notable signal with an obvious eastward propagation in summer (e.g., Lau and Chan, 1986; Zhang et al., 2009). In addition, a distinct northward propagation is observed as well, especially in the Indian Ocean and the tropical western Pacific (e.g., Annamalai and Slingo, 2001; Chou and Hsueh, 2010).

As shown in Fig. 10a, there exists a clear eastward propagation of the MJO, taking about 20 days to propagate from the eastern equatorial Indian Ocean to the tropical western Pacific. When it arrives in the Maritime Continent, a remarkable northward propagation is observed as far as 30^#cod#x000b0;N, with an almost unchanged oscillating periodicity during the whole process (Fig. 10b). Compared with the observation, the model simulates the phase and amplitude of MJO well in the equatorial Indian Ocean, but the eastward propagation signal is poorly simulated, behaving like standing waves separated by the Maritime Continent (Fig. 10c). Despite the northward propagation in the tropical western Pacific being represented by IAP4, the simulated oscillating periodicity is obviously shorter, with a peak of about 30 days (Figs. 8f and 10d), thereby leading to the shorter interval between the jumps of the WPSH.

Figure 9.  The same as the right panels in Fig. 8, but for the region (5^#cod#x000b0;-20^#cod#x000b0;N, 110^#cod#x000b0;-130^#cod#x000b0;E).

Figure 10.  Time-longitude cross section of lagged correlation coefficients between 20-90-day band-pass filtered OLR (shaded), zonal wind at 850 hPa (contours) averaged over 10^#cod#x000b0;S-10^#cod#x000b0;N and 20-90-day filtered OLR over the reference region (10^#cod#x000b0;S-5^#cod#x000b0;N, 75^#cod#x000b0;-100^#cod#x000b0;E) in summer (units: dimensionless): (a) observation; (c) simulation. (b, d) The same as (a, c), but for time-latitude cross section averaged over 110^#cod#x000b0;-120^#cod#x000b0;E. Contours and shading indicate significance at the 90% confidence level based on the t-test.

Due to the specified SST and sea ice in the model, the simulated sea surface air temperature is almost identical to the observation. Therefore, the negative bias in land-sea temperature contrast must be attributed to the overestimated land surface temperature in the model, which is closely associated with the Community Land Model 3.5 (CLM3.5) used in IAP4. In addition, it is also possible that a stronger monsoon results from an overestimated WPSH and warmer air in the corresponding region. Whether there exists a relationship between the stronger WPSH and the overestimated land-sea temperature difference needs further study.

As for the weak eastward propagation of the MJO in the model, there are four possible reasons. First, as noted by (Inness and Slingo, 2003), the coarsely distributed islands over Indonesia possibly act as some sort of barrier to the eastward propagation of the MJO. Clearly, it is difficult for IAP4 with a horizontal resolution of 1.4^#cod#x000b0;#cod#215; 1.4^#cod#x000b0; to distinguish these small islands in Indonesia. With the horizontal resolution being increased, it is anticipated that the model will simulate the intraseasonal variation of the EASM better by improving the eastward propagation of the MJO. Actually, a study by (Lau and Ploshay, 2009) has demonstrated that, by increasing the model horizontal resolution by half a degree, a high-resolution AGCM with imposed SST can capture the timing and onset of monsoon rainfall quite well.

Second, since air-sea coupling acts as a magnitude amplifier and a periodicity selector of the MJO propagating eastward, a lack of air-sea coupling in a stand-alone AGCM tends to weaken the eastward propagation of the MJO (Woolnough et al., 2001). (Zhang et al., 2006) also showed that for AGCMs that already have some MJO signals, coupling with an ocean model can strengthen eastward propagating signals.

Besides, the mechanism for air-sea coupled intraseasonal interaction proposed by (Flatau et al., 1997) indicates that the westerly at the lower level is favorable for the eastward propagation of the MJO. As shown in Fig. 6c, the easterly bias is found from the Indochina Peninsula to the Philippines, indicating a weaker westerly in the model, which partly leads to the weaker eastward propagation of the MJO.

Finally, as an inherent phenomenon in the tropical atmosphere (Waliser et al., 1999), the MJO simulation can also be improved by tuning the convective parameterization in the model. In fact, the MJO is sensitive to the diabatic heating profiles of the atmosphere (Li et al., 2009). The enhancement of diabatic heating in the lower troposphere is favorable for water vapor convergence and the eastward propagation of the MJO. (Benedict et al., 2013) also indicated that, by modifying the convective closure and trigger assumptions to inhibit deep cumuli in the convective parameterization scheme, AGCMs can produce reasonable tropical intraseasonal oscillation, but often at the expense of a degraded mean state. Therefore, how to tune convective parameterization to improve MJO simulation remains a key question concerning model development in the future.

Based on model outputs driven by observed SST during 1979-2008, we evaluated the performance of the newly developed model, IAP AGCM4.0, in simulating the intraseasonal variation of the EASM, with a focus on the northward jumps of the WPSH and the related rain belt.

In general, the model simulates the intraseasonal oscillation of the tropical rain belt and the northward advances of the rain belt in East Asia. However, there exist some evident biases in the model. Since the simulated WPSH in early summer tends to extend too far northwestward, the pre-flood rain belt in southern China is absent in the model. During the late summer, especially in August, a remarkable negative bias in the land-sea thermal contrast leads to a stronger southerly anomaly in East Asia and a longer maintenance of the EASM. Since SST and sea ice are specified, the bias in the temperature contrast can only be related to the bias in the land surface model.

Although the northward jumps of the WPSH are reproduced reasonably, the model simulates a shorter interval between the two jumps than the observation. In the model, the first jump occurs in P36 (25-29 June), which is two pentads later, while the second one occurs in P40 (15-19 July), which is one pentad earlier. This discrepancy is directly related to a shorter periodicity of ISO in the tropical northwestern Pacific.

Since the model simulates the northward propagating signals in the tropical western Pacific, the bias in the tropical northwestern Pacific ISO could be ascribed to the weak eastward propagation of the MJO originating from the equatorial Indian Ocean. Further discussion indicated that there are four possible reasons responsible for the weak eastward propagating signals in the model: the relatively lower horizontal resolution; the lack of air-sea interaction; the weaker westerly from the Indochina Peninsula to the tropical western Pacific; and the imperfect convective parameterization. Summarizing the above discussion, we suggest that the model's resolution, along with the land surface model and cumulus parameterization scheme, should be further improved in the near future in order to better simulate the intraseasonal variation of the EASM.