1. Introduction
The Madden–Julian oscillation (MJO) has been an active research subject in the last three decades since its discovery by Madden and Julian (1971, 1972). There are numerous observational studies on MJO using in situ field observations, satellite, and reanalysis data documenting its propagation, spatial, and temporal structures and its interaction with convection and surface processes (Madden and Julian 1994; Hendon and Salby 1994; Shinoda et al. 1998; Zhang 1996; Yanai et al. 2000; Sperber 2003; and others). In contrast, progress in numerical simulations of MJOs is much slower, particularly in global climate models (GCMs). Slingo et al. (1996) compared MJO simulations from 15 GCMs, and found that most of the GCMs participating in the study are unable to simulate the tropical intraseasonal variability realistically. Even those GCMs that do simulate reasonable tropical intraseasonal variability have difficulties getting the correct eastward propagation of MJOs from the Indian Ocean to the western Pacific Ocean (Sperber et al. 1997; Inness et al. 2001). One of the GCMs that produced reasonable intraseasonal variability was the second version of the National Center for Atmospheric Research (NCAR) Community Climate Model (CCM2). Unfortunately, when CCM3 replaced CCM2, the MJO simulation was degraded significantly, even though the overall performance of the model’s climate simulation is greatly improved (Kiehl et al. 1998; Zhang et al. 1998). The MJOs in CCM3 are very weak (Maloney and Hartmann 2001), and the likely factor responsible for this is the use of the Zhang–McFarlane scheme (Zhang and McFarlane 1995) in place of the Hack (1994) scheme for deep convection. Efforts to restore the MJO variability in CCM3 have not succeeded considering the general metric of climate simulation improvements. Maloney and Hartmann (2001) tested the sensitivity of the MJO simulation in CCM3 to convective parameterization schemes in a perpetual March simulation. Using a version of the Relaxed Arakawa–Schubert convection scheme (Arakawa and Schubert 1974; Moorthi and Suarez 1992) that includes cloud microphysics (Sud and Walker 1999), they showed that the amplitude of the simulated MJOs is enhanced. However, an (Atmospheric Model Intercomparison Project) AMIP-style test simulation using the Relaxed Arakawa–Schubert scheme shows that the simulated mean climate degraded too much compared to the standard CCM3 simulations (Wehner et al. 2000) to allow for meaningful coupling with ocean models (J. J. Hack 2001, personal communication). Thus, to date, weak tropical intraseasonal variability is still a serious modeling issue to be addressed in CCM3 and its successors: versions 2 and 3 of the Community Atmosphere Model (CAM2) and (CAM3).
One possible reason why the Zhang–McFarlane convection scheme fails to simulate the tropical intraseasonal variability and MJO in CCM3 is the use of convective available potential energy (CAPE) as closure to determine the amount of convection in the atmosphere. In the scheme, it is assumed that, when the atmosphere is convectively unstable, convection is activated to remove this instability within a relaxation time scale of 2 h. Observations (Xu and Emanuel 1989) indicate that the tropical atmosphere is nearly neutral to slightly unstable most of the time. Thus, convection so parameterized occurs almost at all times in the Tropics, leading to weak temporal variability.
Recently, Zhang (2002, 2003a) investigated the relationships between convection and the large-scale thermodynamic fields using field observational data from the Atmospheric Radiation Measurement (ARM) program at the U.S. Southern Great Plains (SGP) and the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) Intensive Observation Period in the tropical western Pacific. He found that under both tropical maritime and midlatitude continental conditions, convection is well correlated with the large-scale advective generation of CAPE in the free troposphere instead of with CAPE itself. Based on this result, Zhang (2002) proposed a new closure for the Zhang–McFarlane scheme. Initial results from the test of the new closure in CCM3 indicate that it has a positive impact on the simulation of the diurnal cycle of convection in North America (Zhang 2003b).
In this study, we will show that modifications to the Zhang–McFarlane scheme based on the work of Zhang (2002) significantly improve the simulation of the tropical intraseasonal variability in association with MJOs in CCM3. In section 2, we will briefly describe the model and the data used for comparison with the model results. Section 3 will present the simulation of the MJOs. In section 4, sensitivity tests are performed to examine the effects of the individual changes in the modified convection scheme. Section 5 will summarize the paper.
2. Numerical experiments
The model used in this study is the NCAR CCM3 (Kiehl et al. 1998).1 It is a global model at T42 horizontal resolution and 18-level vertical resolution. Deep convection is parameterized using the Zhang–McFarlane scheme (Zhang and McFarlane 1995), and shallow convection is parameterized using the Hack scheme (Hack 1994). For details of the model description and its climate simulation see Kiehl et al. (1998).
Two numerical simulations are conducted using the observed sea surface temperatures as boundary conditions. The first one (CTRL) uses the standard CCM3 configuration. The time integration starts on 1 September 1985 and continues for 10 years. The second one (EXP) starts on 1 September 1979 and continues for 15 years. The data from the last 10 years are used to compare with the control run. In this simulation, the modified Zhang–McFarlane scheme is used. Otherwise, the setup of the two simulations is identical. The model outputs are saved daily.
The observational data used to evaluate the model simulation are from the (National Centers for Environmental Prediction) NCEP–NCAR reanalysis (Kalnay et al. 1996). The precipitation data is from Xie and Arkin (1996) with pentad resolution. To match the temporal resolution of the observations, the model output is averaged to yield pentad means.
The description of the modifications to the Zhang–McFarlane scheme and their effect on the simulation of the tropical precipitation climatology are presented elsewhere (Zhang and Mu 2005). Here we will outline the main changes. Three changes are made to the Zhang–McFarlane scheme in the modification. First, a new closure based on the work of Zhang (2002) is used to replace the CAPE-based closure in the original Zhang–McFarlane scheme. Second, a relative humidity threshold (RHc) of 80% is included in the scheme for convection trigger. The third change to the Zhang–McFarlane scheme is the removal of the restriction in the code that convection only originates from below the PBL top. This allows midlevel convection with a cloud base above the PBL top to be included in the Zhang–McFarlane scheme.
The new closure is based on the notion that CAPE change with time can be separated into contributions from the boundary layer and contributions from the free troposphere above it. Since the majority of the air parcels that form the convective updrafts originate from the boundary layer, changes in boundary layer temperature and moisture fields affect the buoyancy of the air parcels when they are lifted, thereby affecting CAPE. For the convenience of description, we refer to this part of the CAPE change as the boundary layer component. The buoyancy of an air parcel depends not only on its own temperature but also on the temperature of its surrounding environment when it is lifted above the boundary layer into the free troposphere. For a given parcel temperature, the colder the ambient air is, the more buoyant the air parcel. Therefore, large-scale (environmental) temperature and moisture changes in the free troposphere also affect CAPE. We call this part of the CAPE change the free tropospheric component. The new closure assumes that there is a quasi equilibrium between the free tropospheric component of CAPE change due to the large-scale processes and that due to convection during convective activities. This is independent of the boundary layer component of the CAPE change, which may result from large-scale and/or convective processes during convection. Thus, under this quasi-equilibrium assumption, the boundary layer temperature and moisture are allowed to change freely to modify the total CAPE values at any time. In physical terms, it means that, given the large-scale forcing in the free troposphere, just enough convection is allowed to occur so that the resultant convective heating in the free troposphere balances the large-scale cooling, giving a free-tropospheric mean temperature change much smaller than what would have resulted from either large-scale forcing or convective heating alone.
A relative humidity threshold has been widely used in convective parameterization schemes to suppress convection in situations when the boundary layer air is deemed too dry (Slingo et al. 1996; Moorthi and Suarez 1992). It has had varying effect on the simulation of MJOs. Using three different types of convection parameterization schemes and relative humidity thresholds, varying from 0% to 90%, Wang and Schlesinger (1999) showed that the intensity of MJOs is sensitive to the relative humidity threshold in all three schemes they tested: the higher the threshold, the stronger the MJO intensity. In contrast, using a modified version of the Relaxed Arakawa–Schubert scheme, (McRAS: Sud and Walker 1999), in CCM3, Maloney and Hartmann (2001) found that imposing a relative humidity threshold almost has the opposite effect to those found by Wang and Schlesinger (1999) on the simulated MJO intensity. They showed that the simulation with no relative humidity threshold has the strongest MJO intensity whereas the simulation using a relative humidity threshold of 91% has the weakest MJO intensity. The default RHc for convection trigger is 81% in their study. As we will show later in sensitivity tests, inclusion of a relative humidity threshold enhances the MJO intensity in our simulation, in agreement with Wang and Schlesinger (1999).
3. MJO simulation
a. General features
Before delving into the MJO simulation, it is important to demonstrate that the modified convection scheme gives a reasonable simulation of the mean climate. For this purpose, Fig. 1 shows the annual mean of the 850-mb wind in the latitude range of 60°S–60°N from the NCEP–NCAR reanalysis, EXP, and CTRL, respectively. The contours are for the zonal component of the wind. From the NCEP–NCAR reanalysis, the Tropics is dominated by easterly winds except over the northern Indian Ocean. The higher latitudes in both hemispheres are occupied by westerly winds. These features are simulated well in both CTRL and EXP. From the tropical western Pacific east of the Philippines to Thailand, the easterly wind is too strong in CTRL, whereas the EXP simulation is in better agreement with the reanalysis. Sperber (2004) also found overestimated easterlies in CAM2 and CCSM2 simulations and suggested that it is an important factor that may have contributed to the weak MJO signature in these models. More on the climatology of the simulations, including tropical precipitation, can be found in Zhang and Mu (2005). It suffices to say that overall there is considerable improvement in the mean climate from CTRL to EXP.
To examine the intraseasonal variability and MJO simulation, we use a 20–80-day Lanczos filter to bandpass filter the saved daily output from the model simulations. Slingo et al. (1999) found that an index of MJO activity, defined as the 20–100-day variance of the zonal mean 200-mb zonal wind averaged between 10°S and 10°N and smoothed with a 101-day running mean, is a good measure of the intraseasonal variability of the upper-tropospheric wind in the Tropics. Figure 2 shows the time series of this index for the NCEP–NCAR reanalysis, CTRL, and EXP, respectively. Here a 20–80-day window instead of 20–100-day filter window is used to be consistent with the rest of the results. This narrower window for variance calculation should not change the characteristics qualitatively, though. The time series from the NCEP–NCAR reanalysis clearly shows large interannual variation of the MJO activity during the 16 years from 1980 to 1995. The CTRL run has quiet MJO activities throughout the simulation period except in 1988. In comparison, EXP in general is able to simulate the magnitude of the interannual variation of the MJO activity, although in some years the simulated MJO activity is high while the observed MJO activity is quiescent, and vice versa.
Figure 3 presents the global distribution of the 20–80-day variance for precipitation from the Xie–Arkin observations, the EXP run, and the CTRL run, respectively. There is large precipitation variance in the Indian Ocean, the western Pacific, and along the intertropical convergence zone (ITCZ) and the South Pacific convergence zone (SPCZ) in the Pacific in the observations. All of these major features are simulated in EXP with comparable magnitude. In contrast, the 20–80-day precipitation variance in CTRL is much weaker, about half of that in EXP and the observations.
Since most of the variance is concentrated in the equatorial region, Fig. 4 shows the 20–80-day variances of 850-mb and 200-mb zonal winds, precipitation, and outgoing longwave radiation (OLR) as functions of longitude averaged over 10°S–10°N. The variance of the 850-mb zonal wind from EXP is in good agreement with the observations in both magnitude and its longitudinal variation except over Southeast Asia (near 120°E), where EXP fails to show a local minimum in variance compared to the observations. The CTRL simulation shows considerably weaker variance in the Indian Ocean and the western to central Pacific. The 200-mb zonal wind variance shows similar degree of improvement from CTRL to EXP. The precipitation variance in the Indian Ocean and the western Pacific is well simulated in EXP, with the peak in the western Pacific from EXP even stronger than the observations. On the other hand, the variance from CTRL is about half or less of the observed variance in these regions. The simulated OLR variance from EXP is slightly weaker in the western Pacific and much weaker in the Indian Ocean. Even so, the enhancement of the variance from CTRL to EXP is significant in these regions.
Figure 5 shows the frequency–wavenumber spectra of precipitation and 200-mb zonal wind filtered with the 20–80-day bandpass window and averaged over 10°S–10°N from the Xie–Arkin observations/NCEP–NCAR reanalysis, EXP, and CTRL, respectively. For precipitation, the observations are dominated by eastward propagating waves at periods near 41 days and wavenumbers 1 to 4. There is also a weak power of westward propagating waves of similar wavenumbers and frequencies. The EXP simulation shows similar power spectrum structure to the observations, with dominant power at periods 33–41 days and wavenumbers 1 to 4. However, the amplitude of the power is weaker than that of the observations. This will be seen more clearly in the following subsection when a composite MJO is presented. The power spectrum for CTRL shows that the dominant power in the wavenumber–frequency space is concentrated in westward propagating waves at periods near 40 days and wavenumbers 1 to 5, while the secondary peak of the power is in the eastward propagating waves. Furthermore, the power is only about half that in EXP. For 200-mb zonal wind, the NCEP–NCAR reanalysis shows a power peak near 41 days for eastward propagation at wavenumber 1. The EXP simulation shows a similar power peak, but with weaker magnitude. In addition, it has a secondary power peak at 28-day period. The CTRL simulation has a power peak at periods 33–41 days at wavenumber 1, with weaker magnitude than in EXP. The westward propagating component of the power spectrum in EXP and CTRL is similar to that in the NCEP–NCAR reanalysis.
b. EOF composites
To further examine the simulation of the intraseasonal variability and MJOs, the EOF analysis is conducted using the 200-mb velocity potential field. Figure 6 shows the first two modes of the EOFs from the NCEP–NCAR reanalysis, EXP, and CTRL, respectively. Since the Laplacian of velocity potential gives divergence, the center of low velocity potential corresponds to maximum upper-level divergence, or maximum convective activity, and the center of high velocity potential corresponds to suppressed convection. The first EOF from the NCEP–NCAR reanalysis accounts for 40% of the total 20–80-day variance, with the center of the low located in the western Pacific. The second EOF accounts for 31% of the variance, with the center of the low located in the central to eastern Pacific. In comparison, the centers of EOF1 in EXP are located in the same positions as those in the NCEP–NCAR reanalysis, with slightly weaker amplitude. The centers of EOF2 are located somewhat west of those in the NCEP reanalysis. EOF1 accounts for 29% and EOF2 accounts for 23% of the total variance. The locations of the EOF centers in CTRL are farther to the west, with the first EOF low centered over the eastern part of the Indian Ocean, accounting for 28% of the variance. The second mode has its low center slightly east of the date line, accounting for 21% of the variance. Consistent with Figs. 2 –5, the amplitude of the EOF modes in CTRL is considerably smaller than in the observations.
Figure 8 shows the evolution of the 200-mb velocity potential in one composite MJO cycle, superposed with precipitation anomalies, for the NCEP–NCAR reanalysis, EXP, and CTRL, respectively. From the NCEP–NCAR reanalysis, at phase 1 the positive center of the 200-mb velocity potential, corresponding to the negative precipitation anomaly, is located over the Maritime Continent of Southeast Asia. The negative center is located in the tropical Atlantic and Central America. By the time when the MJO reaches phase 3, the positive 200-mb velocity potential center has moved to near the date line. The negative precipitation anomaly occupies the western and central equatorial Pacific while the positive precipitation anomaly is located in the eastern Indian Ocean. At phase 5, the mature phase of the MJO, the positive velocity potential center is located in Central America, whereas the Southeast Asia Maritime Continent is occupied by the negative velocity potential center and positive precipitation anomaly. Phase 7 is practically opposite to phase 3, with active convection moving to the central and eastern Pacific. Phase 9 is similar to phase 1. Both CTRL and EXP show similar evolution and eastward propagation of the MJO signal in the 200-mb velocity potential field except that the CTRL simulation has a much weaker amplitude, about two-thirds of that of the EXP simulation, which is 10%∼20% weaker than the NCEP–NCAR reanalysis. Compared to the observations, the composite MJO cycle of precipitation in EXP is also well simulated. For instance, at phase 1, the negative precipitation anomaly center is located in a broad region from the western equatorial Pacific to Southeast Asia; at phase 5, the spatial pattern is approximately opposite to that in phase 1. Both are in qualitative agreement with the observations. The positive precipitation anomalies in the Indian Ocean at phase 3 has propagated to Southeast Asia and the western Pacific by phase 5, and farther eastward to the central and eastern Pacific by phase 7. The precipitation signal in CTRL is very weak and its spatial organization and eastward propagation are much less coherent compared with the observations and EXP. While the improvement from CTRL to EXP is significant, there remain obvious deficiencies in the EXP run compared to the observations. First, the magnitude of the precipitation anomalies is smaller compared to that of the observed precipitation anomalies. Second, the spatial scales of the MJO signal in the model precipitation field are smaller than those in the observations. In addition, the precipitation signal in the eastern part of the Indian Ocean is weak in comparison with the observations, and the precipitation pattern in EXP is shifted more eastward into the central equatorial Pacific at the mature phase.
Figure 9 demonstrates the eastward propagation characteristics of the MJOs through longitude–phase Hovmöller plots of composite 850-mb zonal wind, precipitation, and OLR averaged over 10°S–10°N for the observations, EXP, and CTRL, respectively. The shadings denote that the values are at or above the 95% confidence level. The 850-mb zonal wind shows clearly eastward propagation over the entire equatorial belt, whereas the eastward propagation of precipitation and OLR are mostly confined to west of the date line in the observations. In EXP, the eastward propagation of the MJO signals in the 850-mb zonal wind, precipitation, and OLR is well simulated. Compared to the observations, however, the amplitudes of the signals in all these fields are weaker, particularly in precipitation and OLR. In the CTRL simulation, the MJO signal in the 850-mb zonal wind is largely stationary in the western Pacific and the central and eastern Pacific. Owing to the phase shift between the east and west Pacific, it appears as if the signal were propagating eastward discretely. The MJO signals in precipitation and OLR are very weak everywhere, and are mostly below the 95% confidence level. Both precipitation and OLR in the Indian Ocean show a slow westward propagation from phase 1 to 7. The westward propagation of MJO signals in the Indian Ocean was also noticed by Maloney and Hartmann (2001) in their perpetual March simulation using CCM3 and by Sperber (2004) in CAM2.
The structure of the MJOs can be examined through phase–height plots. From Fig. 8, maximum precipitation at the MJO mature phase is located near 120°E for the NCEP–NCAR reanalysis and CTRL and 150°E for EXP. Thus, to compare the structural similarities and differences between the simulations and the observation, Fig. 10 shows the phase–height plots of velocity potential, zonal wind, and specific humidity at these longitudes for the NCEP–NCAR reanalysis, EXP, and CTRL, respectively. The phases are plotted from right to left so that the figures may also be viewed qualitatively as equivalent to a longitudinal cross section along the equatorial belt for a steadily propagating MJO. At the initial phase of the observed MJO, the velocity potential is negative below 400 mb and positive aloft. This corresponds to low-level divergence and upper-level convergence. Phase 3 corresponds to the transition from suppressed convection to active convection; at this phase the flow is nondivergent. At phase 5, there is strong positive velocity potential anomaly below 400 mb, and even stronger negative anomaly above it, corresponding to active convection. Phase 7 is similar to phase 3, except it is a transition from active convection to suppressed convection, and phase 9 is similar to phase 1. Both EXP and CTRL simulate the phase–height structure well, except as noted before, that the magnitude in CTRL is substantially weaker, about one-third that of the observations.
The zonal wind in the observations shows maximum easterly at phase 3 and maximum westerly at phase 7 near the 700-mb height. The wind anomalies switch signs in the upper troposphere, with a maximum at the 150-mb level. Both EXP and CTRL show similar general patterns. However, there are important differences in details. For example, at the phase of maximum MJO convection, the NCEP–NCAR reanalysis, and EXP show weak westerly flow near the surface while CTRL shows weak easterlies.
The specific humidity field in the NCEP–NCAR reanalysis shows a dry anomaly at phase 1. Moistening starts to develop in the lower troposphere around phase 2. By phase 4, the entire troposphere below 300 mb is occupied by a moist anomaly. At phase 5, the moist anomaly reaches a maximum near 600 mb whereas the dry anomaly starts to develop in the lower troposphere afterward. By phase 8, the troposphere is occupied by the dry anomaly. The specific humidity field in EXP shows similar evolution. The leftward tilt of the moisture anomalies, indicating lower-level moistening and midlevel drying before convection and midlevel moistening and lower-level drying during and after convection appear in both the NCEP–NCAR reanalysis and EXP. The control run has a deep layer of moistening before convection (between phase 3 and 5) and a similar strength drying after convection (between phase 7 and 9). This indicates that shallow convection is not active in the MJO development in the CTRL run. Indeed, shallow convection in the composite MJO cycle is very weak (Fig. 12b).
The relationship between the evolution of the moisture field and MJO convection can be further examined through longitude–height cross sections. Figure 11 shows the longitude–height cross sections of the anomalies of deep convective heating from the Zhang–McFarlane scheme and shallow convective heating from the Hack scheme over a composite MJO cycle, superposed with specific humidity anomalies for EXP. At phase 1, the western Pacific and eastern Indian Ocean region is dominated by dry anomalies and negative deep convective heating anomalies (convectively suppressed region). By phase 3, the region west of 150°E is occupied by positive moisture and deep convective heating anomalies. The dry anomaly and suppressed convection region has moved to the date line and farther eastward. However, the lower troposphere near the date line has a shallow layer of positive humidity anomaly, which is clearly related to shallow convection as evident from the strong shallow convective heating from the Hack scheme in this layer (Fig. 11b). At phase 5, a deep layer of strong positive humidity and deep convective heating anomalies occupy a large region from east of the date line to 120°E. Dry anomaly and suppressed convection are located in the Indian Ocean. Again, ahead of the strong deep convection, there is a large region of strong shallow convection from the Hack scheme, corresponding to a layer of moist anomaly in the lower half of the troposphere and dry anomaly aloft. At phase 7, deep convection and moist anomalies moved to the central and eastern Pacific, and the dry anomaly and suppressed convection moved to the Southeast Asia Maritime Continent and the western Pacific. Phase 9 is similar to phase 1. From phase 1 to 7, the deep convective heating anomalies show clearly eastward propagation from the Indian Ocean to the central Pacific.
The same longitude–height cross sections for CTRL are shown in Fig. 12. Overall, deep convection is much less organized, with weaker intensity. Shallow convection virtually has no MJO signal, and there is no coherence between the lower-tropospheric humidity anomalies and the shallow convection anomalies (Fig. 12b). At phase 1, active convection is located in the western Pacific near 140°E, the eastern Indian Ocean near 80°E, and Africa west of 30°E. Suppressed convection regions are located in the central Pacific, Indonesia, and the western Indian Ocean. Positive moisture anomalies are associated with active convection and negative moisture anomalies are associated with suppressed convection. The deep convective heating anomalies (both positive and negative) in the Indian Ocean slowly propagate westward to Africa during the MJO cycle. At phase 5, convection also develops in Indonesia and near the date line. By phase 7, the convective heating and the moisture anomaly pattern is almost entirely reversed from that at phase 5. The vast region from the Indian Ocean to the western Pacific is occupied by negative heating and moisture anomalies, and the eastern Pacific is occupied by positive heating and moisture anomalies. The pattern at phase 9 is similar to that at phase 7.
Several observational studies (e.g., Bladé and Hartmann 1993; Johnson et al. 1999; Kemball-Cook and Weare 2001) have argued that the low-level moistening or preconditioning is necessary for the development of the MJO’s mature phase. In a recent study using a linear primitive equation model, Wu (2003) emphasized the role of shallow convection in the development of MJOs. He suggests that latent heating from shallow convection drives the low-level moisture convergence, which in turn feeds the convection; this eventually leads to the outbreak of deep convection. If these arguments are valid, then the moistening in the lower troposphere by shallow convection before the development of deep convection has apparently played an important role in the MJO evolution in EXP.
In the EXP simulation, there is no change in shallow convection parameterization. Yet, shallow convection is enhanced significantly. This must be a result of the interaction between deep and shallow convection. In CTRL, deep convection occurs continuously with CAPE-based closure. Thus, there is less convective energy for shallow convection. In the EXP simulation, use of both the new closure and the relative humidity threshold suppresses deep convection when the large-scale forcing does not favor convection or when the boundary layer is too dry. This allows CAPE to build up. Since the shallow convection scheme is based on local instability, part of this CAPE can be used to increase shallow convection.
4. Sensitivity to individual changes
In section 3, we showed that the MJO simulation is significantly improved by changes in the Zhang–McFarlane convection scheme. This section attempts to assess the roles of each individual changes. For this purpose, we conduct two additional experiments. In the first one, denoted as no relative humidity critical value (NRHC), the relative humidity threshold is removed. In the second experiment, denoted as no midlevel initiation of deep convection (NMID), the convection base is restricted to within the PBL, as in the control run. For both the experiments, the model is integrated for 7 years starting from 1 September 1979. Similar analysis to that for CTRL and EXP was performed for these simulations.
We find that use of the relative humidity threshold affects the amplitude of the MJO signal in precipitation significantly. Figure 13 shows the longitudinal variation of precipitation and 850-mb zonal wind anomalies at the mature phase of the composite MJO, averaged from 10°S to 10°N. When only the convection closure is changed, that is, from CTRL to NRHC, the amplitude of the composite MJO precipitation increases slightly. When the relative humidity threshold is included in addition to the closure change, from NRHC to NMID, the amplitude more than doubles. When a further change is made to allow convection to initiate from above the PBL top, that is, from NMID to EXP, the MJO amplitude decreases slightly and the maximum precipitation center at the mature phase is shifted eastward by about 10°. For the 850-mb zonal wind field, change of closure (from CTRL to NRHC) increases the amplitude of the MJO signal considerably by increasing the westerly wind west of the convection center, with a maximum at 100°E. East of 150°E there is little systematic change in the easterly wind. Including the relative humidity threshold (from NRHC to NMID) does not increase the amplitude of the MJO signal in the wind field, but it increases the easterly wind ahead of the convection center while it decreases the westerly behind the convection center. Allowing convection to initiate above the PBL top (from NMID to EXP) decreases the easterly wind east of the convection and increases the westerly wind west of convection modestly. Based on these results, the relative humidity threshold has the largest impact on the intensity of the simulated MJO signal in precipitation among the three changes in the modification to the Zhang–McFarlane convection scheme, whereas use of the new closure has the largest impact on the MJO signal in the 850-mb zonal wind.
Although the new closure only has a small impact on the intensity of the simulated MJO precipitation, it has a pronounced effect on the MJO propagation characteristics. Figure 14 shows the phase–longitude Hovmöller plots of 850-mb zonal wind and precipitation anomalies averaged over 10°S–10°N for NRHC and NMID, respectively. This can be compared with Fig. 9 for the CTRL and EXP simulations. By changing the closure from CTRL to NRHC, the eastward propagation of the 850-mb zonal wind anomalies becomes distinct, with amplitude comparable to that in the EXP. Same as in Fig. 9, the precipitation anomalies show less coherence than the wind field. Nevertheless, eastward propagation from the Indian Ocean to the western Pacific is apparent in NRHC, as compared to westward propagation in the Indian Ocean in the CTRL. From NRHC to NMID, that is, when a relative humidity threshold is imposed in addition to the closure change, the eastward propagation characteristics remain about the same for both the 850-mb zonal wind and precipitation fields. However, the intensity becomes much stronger in precipitation. When the additional change is made by allowing deep convection to initiate above the PBL top, that is, from NMID to EXP (see Fig. 9), the difference in the eastward propagation is small while the amplitude becomes somewhat weaker, particularly in the precipitation field.
To summarize, we showed through sensitivity tests by adding one change at a time that use of the new convection closure is instrumental to the eastward propagation of the MJOs, while its effect on the MJO intensity is variable dependent. Addition of a relative humidity threshold for the convection trigger retains the eastward propagation; in addition, it enhances the MJO intensity significantly. The sensitivity to the relative humidity threshold agrees with the work of Wang and Schlesinger (1999). These two changes are the main factors responsible for the improved MJO simulation. Allowing deep convection to initiate above the PBL top reduces the MJO intensity slightly. But overall, it has no significant systematic impact.
5. Conclusions
This study examined the simulation of the tropical intraseasonal variability and MJO in CCM3 using a modified Zhang–McFarlane scheme. It is shown that modifications to the Zhang–McFarlane scheme lead to a much improved simulation of the intraseasonal variability and MJO compared with the standard CCM3 simulation. The intraseasonal variability of the various fields examined is in close agreement with the observations. Using a technique similar to that of Maloney and Hartmann (1998), we constructed a composite MJO cycle. The composite MJO signals for 850-mb zonal wind, precipitation, and OLR all have comparable intensity to that of the observations and the NCEP–NCAR reanalysis. The eastward propagation is also realistically simulated. These are in sharp contrast to the MJO simulation in the standard CCM3 model, which shows weak intraseasonal variability and westward propagation of the MJO signals in the Indian Ocean. The examination of the interaction between the moisture field in the lower and middle troposphere and convection suggests that shallow convection ahead of the deep convection in the experiment seems to play an important role in preconditioning the atmosphere for MJO development, as previous observations indicate.
Sensitivity tests suggest that both closure change and imposing a relative humidity threshold for the convection trigger have an important effect on the improvement of the MJO simulation. On the other hand, lifting the restriction of convection initiating only in the boundary layer does not seem to have systematic, qualitative impact on the MJO simulation.
While this study presents the improvement of the MJO simulation in the NCAR CCM3 model with modifications to the Zhang–McFarlane convection scheme, we have not addressed the question of why there is such an improvement. For instance, physically and dynamically what mechanisms lead to the improved simulation of MJO intensity and eastward propagation in the model? Maloney (2002) examined the MJO simulation in the NCAR CCM3 using McRAS (Sud and Walker 1999), and concluded that the enhanced MJOs with McRAS are strongly tied to the surface frictional convergence. Since the modified Zhang–McFarlane scheme determines convection based on the free tropospheric large-scale forcing, we speculate that the enhanced MJO in our study may be more related to the wave-CISK (conditional instability of the second kind) mechanism (Lindzen 1974; Wu 2003). This will be the subject of our future study.
Acknowledgments
This research was supported by the NSF under Grant ATM-0204798 and by the Environmental Science Division of the U.S. Department of Energy under Grant DE-FG02-03ER63532. We thank Jialin Lin for providing us the numerical code for the wavenumber–frequency spectrum analysis. The constructive comments from the reviewers greatly helped to improve the quality of this paper.
REFERENCES
Arakawa, A., and W. H. Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment. Part I. J. Atmos. Sci., 31 , 674–701.
Bladé, I., and D. L. Hartmann, 1993: Tropical intraseasonal oscillations in a simple nonlinear model. J. Atmos. Sci., 50 , 2922–2939.
Collins, W. D., and Coauthors. 2003: Description of the NCAR Community Atmosphere Model (CAM2). 190 pp. [Available online at www.ccsm.ucar.edu/models/atm-cam/docs/cam2.0/description.pdf.].
Hack, J. J., 1994: Parameterization of moist convection in the National Center for Atmospheric Research community climate model (CCM2). J. Geophys. Res., 99 , 5551–5568.
Hendon, H. H., and M. L. Salby, 1994: The life cycle of the Madden–Julian oscillation. J. Atmos. Sci., 51 , 2225–2237.
Inness, P. M., J. M. Slingo, S. J. Woolnough, R. B. Neale, and V. D. Pope, 2001: Organization of tropical convection in a GCM with varying vertical resolution: Implications for the simulation of the Madden–Julian Oscillation. Climate Dyn., 17 , 777–793.
Johnson, R. H., T. M. Rickenbach, S. A. Rutledge, P. E. Ciesielski, and W. S. Schubert, 1999: Trimodal characteristics of tropical convection. J. Climate, 12 , 2397–2418.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Kemball-Cook, S. R., and B. C. Weare, 2001: The onset of convection in the Madden–Julian oscillation. J. Climate, 14 , 780–793.
Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, D. L. Williamson, and P. J. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3. J. Climate, 11 , 1131–1149.
Lindzen, R. S., 1974: Wave-CISK in the Tropics. J. Atmos. Sci., 31 , 156–179.
Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci., 28 , 702–708.
Madden, R. A., and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40–50 day period. J. Atmos. Sci., 29 , 1109–1123.
Madden, R. A., and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122 , 814–837.
Maloney, E. D., 2002: An intraseasonal oscillation composite life cycle in the NCAR CCM3.6 with modified convection. J. Climate, 15 , 964–982.
Maloney, E. D., and D. L. Hartmann, 1998: Frictional moisture convergence in a composite life cycle of the Madden–Julian oscillation. J. Climate, 11 , 2387–2403.
Maloney, E. D., and D. L. Hartmann, 2001: The sensitivity of intraseasonal variability in the NCAR CCM3 to changes in convective parameterization. J. Climate, 14 , 2015–2034.
Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev, 120 , 978–1002.
Shinoda, T., H. H. Hendon, and J. Glick, 1998: Intraseasonal variability of surface fluxes and sea surface temperature in the tropical western Pacific and Indian Oceans. J. Climate, 11 , 1685–1702.
Slingo, J. M., and Coauthors. 1996: Intraseasonal oscillations in 15 atmospheric general circulation models: Results from an AMIP diagnostic subproject. Climate Dyn., 12 , 325–357.
Slingo, J. M., D. P. Rowell, K. R. Sperber, and F. Nortley, 1999: On the predictability of the interannual behavior of the Madden–Julian oscillation and its relationship with El Niño. Quart. J. Roy. Meteor. Soc., 125 , 583–609.
Sperber, K. R., 2003: Propagation and the vertical structure of the Madden–Julian oscillation. Mon. Wea. Rev., 131 , 3018–3037.
Sperber, K. R., 2004: Madden–Julian variability in NCAR CAM2.0 and CCSM2.0. Climate Dyn., 23 , 259–278.
Sperber, K. R., J. M. Slingo, P. M. Inness, and W. K-M. Lau, 1997: On the maintenance and initiation of the intraseasonal oscillation in the NCEP/NCAR reanalysis and in the GLA and UKMO AMIP simulations. Climate Dyn., 13 , 769–795.
Sud, Y. C., and G. K. Walker, 1999: Microphysics of clouds with the relaxed Arakawa–Schubert scheme (McRAS). Part I: Design and evaluation with GATE phase III data. J. Atmos. Sci., 56 , 3196–3220.
Wang, W., and M. E. Schlesinger, 1999: The dependence on convective parameterization of the tropical intraseasonal oscillation simulated by the UIUC 11-layer atmospheric GCM. J. Climate, 12 , 1423–1457.
Wehner, M. F., K. Taylor, C. Doutriaux, K. Achutarao, P. Gleckler, J. Hnilo, and J. Boyle, 2000: PCMDI analysis of candidate atmospheric models for CCSM. PCMDI Tech. Rep. UCRL-JC-141573, 10 pp. + figures.
Wu, Z., 2003: A shallow CISK, deep equilibrium mechanism for the interaction between large-scale convection and large-scale circulations in the Tropics. J. Atmos. Sci., 60 , 377–392.
Xie, P., and P. A. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840–858.
Xu, K-M., and K. A. Emanuel, 1989: Is the tropical atmosphere conditionally unstable? Mon. Wea. Rev., 117 , 1471–1479.
Yanai, M., B. Chen, and W-W. Tung, 2000: The Madden–Julian oscillation observed during the TOGA COARE IOP: Global view. J. Atmos. Sci., 57 , 2374–2396.
Zhang, C., 1996: Atmospheric intraseasonal variability at the surface in the tropical western Pacific Ocean. J. Atmos. Sci., 53 , 739–758.
Zhang, G. J., 2002: Convective quasi-equilibrium in midlatitude continental environment and its effect on convective parameterization. J. Geophys. Res., 107 .4220, doi:10.1029/2001JD001005.
Zhang, G. J., 2003a: Convective quasi-equilibrium in the tropical western Pacific: Comparison with midlatitude continental environment. J. Geophys. Res., 108 .4592, doi:10.1029/2003JD003520.
Zhang, G. J., 2003b: Roles of tropospheric and boundary layer forcing in the diurnal cycle of convection in the U.S. Southern Great Plains. Geophys. Res. Lett., 30 .2281, doi:10.1029/2003GL018554.
Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33 , 407–446.
Zhang, G. J., and M. Mu, 2005: Effects of modifications to the Zhang–McFarlane convection parameterization on the simulation of tropical precipitation in the National Center for Atmospheric Research Community Climate Model, version 3. J. Geophys. Res., 110 .D09109, doi:10.1029/2004JD005617.
Zhang, G. J., J. T. Kiehl, and P. J. Rasch, 1998: Response of climate simulation to a new convective parameterization in the National Center for Atmospheric Research Community Climate Model (CCM3). J. Climate, 11 , 2097–2115.
Annual average 850-mb wind (vectors) and its zonal component (contours) for (a) NCEP–NCAR reanalysis, (b) EXP, and (c) CTRL. The contour intervals are at 3 m s−1, with areas less than −3 m s−1 shaded.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Time series of the 20–80-day variance of the zonal mean 200-mb zonal wind averaged between 10°S and 10°N with a 101-day running mean for the NCEP–NCAR reanalysis (solid line), the CTRL (short-dashed line), and the EXP (long-dashed line).
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Global distributions of the 20–80-day variance of precipitation from (a) the Xie–Arkin observations, (b) EXP, and (c) CTRL, respectively (units: mm2 day−2).
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
The 20–80-day variances of (a) 850- and (b) 200-mb zonal winds (units: m2 s−2), (c) precipitation (units: mm2 day−2), and (d) outgoing longwave radiation (units: W2 m−4) as functions of longitude averaged over 10°S–10°N. The solid line is for NCEP–NCAR reanalysis, the long-dashed line is for EXP, and the short-dashed line is for CTRL.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Frequency–wavenumber power spectra for (a), (b), (c) precipitation and (d), (e), (f) 200-mb zonal wind filtered with a 20–80-day bandpass and averaged over 10°S–10°N from (a) Xie–Arkin observations; (d) NCEP–NCAR reanalysis; (b), (e) EXP; and (c), (f) CTRL, respectively. The contour intervals for precipitation are 0.25 mm2 day−1 starting at 0.5 mm2 day−1 for (a), 0.1 mm2 day−1 starting at 0.5 mm2 day−1 for (b), and 0.1 mm2 day−1 starting at 0.2 mm2 day−1 for (c). The contour intervals for zonal wind are 0.1 m2 s−2 day−1.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
First two principal modes of the EOFs for the 200-mb velocity potential from the NCEP–NCAR reanalysis, EXP, and CTRL, respectively (units: 106 m2 s−1).
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
(a) Time–lag correlation between the first two principal components of the EOF analyses of the 200-mb velocity potential and (b) the constructed MJO index for a composite cycle centered at the peak. The solid, long-dashed and short-dashed lines are for NCEP–NCAR reanalysis, EXP, and CTRL, respectively. Negative time in (a) means PC2 leads PC1.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Composite 200-mb velocity potential anomalies (units: 106 m2 s−1), superposed with precipitation anomalies (units: mm day−1), for (a) the NCEP–NCAR reanalysis, (b) EXP, and (c) CTRL at phases 1, 3, 5, 7, and 9, respectively. Contour intervals are 0.5 × 106 m2 s−1.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Longitude–phase cross sections of the composite 850-mb zonal wind (contour interval: 0.5 m s−1), precipitation (contour interval: 0.5 mm day−1), and OLR (contour interval: 3 W m−2) averaged over 10°S–10°N for (a) the observations, (b) EXP, and (c) CTRL, respectively. Negative values are in dashed lines, and the shaded areas represent the 95% significance level.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Phase–height cross sections of the 200-mb velocity potential (contour interval: 0.5 × 106 m2 s−1), zonal wind (contour interval: 0.5 m s−1), and specific humidity (contour interval: 0.05 g kg−1) for (a) the NCEP–NCAR reanalysis, (b) EXP, and (c) CTRL, respectively. Negative values are in dashed lines, and the shaded areas represent the 95% significance level.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Longitude–height cross sections of the anomalies of specific humidity (contours) and (a) deep convective heating from the Zhang–McFarlane scheme (color shades) and (b) shallow convective heating from the Hack scheme (color shades), at phases 1, 3, 5, 7, and 9, respectively, for EXP. The contour interval for specific humidity is 0.05 g kg−1, and the color shade interval for deep and shallow convective heating is 0.05 K day−1.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Zonal variation of composite MJO (a) precipitation and (b) 850-mb zonal wind at the mature phase averaged over 10°S–10°N.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
Longitude–phase cross sections of the composite 850-mb zonal wind (contour interval: 0.5 m s−1) and precipitation (contour interval: 0.5 mm day−1) averaged over 10°S–10°N for (a) the run without a relative humidity threshold for the convection trigger and (b) the run with deep convection initiating only within the boundary layer. Negative values are in dashed lines, and the shaded areas represent the 95% significance level.
Citation: Journal of Climate 18, 19; 10.1175/JCLI3508.1
The reason why we use CCM3 instead of its successor CAM2 is for ease of implementation of the code modifications. Although there are numerous changes from CCM3 to CAM2 (Collins et al. 2003), the climate simulations from CCM3 and CAM2 are, in general, similar. Furthermore, the same convection parameterization scheme is used in both models. Therefore, the results from this study are expected to apply to CAM2 as well as the newly released CAM3.
