Meteorol. Atmos. Phys. 78, 227±244 (2001)
National Center for Medium Range Weather Forecasting (NCMRWF), Mausam Bhawan Complex, New Delhi, India
Comprehensive test of different cumulus parameterization schemes for the simulation of the Indian summer monsoon S. Das, A. K. Mitra, G. R. Iyengar, and S. Mohandas With 15 Figures Received October 3, 2000 Revised December 5, 2000 Summary The global spectral model of NCMRWF at T80 horizontal resolution and 18 vertical levels has been integrated for the summer season (July) using different cumulus parameterization schemes namely, the Simpli®ed Arakawa-Schubert scheme (SAS), the Relaxed Arakawa-Schubert Scheme (RAS), and the Kuo-type cumulus parameterization scheme (KUO). The results have been compared with mean analysis of the operational NCMRWF model (ANA) and other available observations. Results indicate that, while the global distributions of basic ®elds such as the wind, temperature and moisture are fairly well simulated by all the three schemes, there are many differences seen in the simulation of the typical features of the Indian summer monsoon. The strength of the Low Level Westerly Jet (LLWJ), the Cross Equatorial Flow (CEF), and the Tropical Easterly Jet (TEJ) are better simulated by RAS and SAS as compared to ANA than the KUO scheme. RAS and SAS produce strong rising motion owing to strong intensity of convection produced by these two schemes. This in turn produces stronger Hadley cell by RAS and SAS than compared to the KUO scheme. Simulation of the 200 mb velocity potential and divergent wind by RAS and SAS produced two prominent centers, one in the Bay of Bengal and another in the Western Paci®c, which correspond to the intense latent heating by cumulus convection during the active monsoon phase. The velocity potential and divergent winds were weaker in KUO, than compared to RAS and SAS. The simulation of OLR is improved by RAS as compared to observations. The cold bias produced by KUO at 200 mb is reduced by RAS and is substantially improved by SAS. Study of observed and simulated rainfall indicated that RAS and SAS produced better distribution of precipitation over the Western Ghat Mountains and the Arakan coast, where
deep cumulus convection is produced due to orographic forcing of the warm moist air. The KUO scheme underestimated the rainfall over these two regions, but produced slightly better distribution of rainfall over the northwest and central India, where the intensity of convection is relatively weaker. Evaluation of overall dynamics, thermal structure and rainfall indicates that in general, SAS is able to provide relatively better results compared to other two schemes.
1. Introduction Convection plays a dominant role in the tropics, in particular for the development and maintenance of the Asian summer monsoon. It affects the tropical circulation through the release of latent heat, vertical transport of heat, moisture and momentum and, through the interaction of clouds with radiation. Several methods have been proposed to parameterize the effects of cumulus convection in the large-scale numerical weather prediction models. The two most widely used schemes among them are the Kuo's scheme (Kuo, 1974) and the Arakawa-Schubert scheme (Arakawa and Schubert, 1974, hereafter referred to as A-S). The latter scheme is known for its sound physical basis, but it is also computationally very expensive. Many techniques have been proposed in the recent past to simplify this scheme, viz., the Simpli®ed Arakawa-Schubert scheme (SAS) developed by Grell (1993), Pan and Wu (1995),
228
S. Das et al.
and the Relaxed Arakawa-Schubert scheme (RAS) proposed by Moorthi and Suarez (1992). The present operational model at the National Center for Medium Range Weather Forecasting (NCMRWF) uses a Kuo-type cumulus parameterization scheme based on Anthes (1977). Intensity of the monsoon circulation is affected by many factors such as, the amount and location of diabatic heating rate, sea surface temperature and Eurasian snow cover (Hahn and Shukla, 1976; Rao and Goswami, 1986; Vernekar et al., 1995). Among them the convective latent heat release provides a vast amount of energy for the onset and maintenance of the monsoon system (Pearce and Mohanty, 1984). Krishnamurti and Ramanathan (1982) found that the onset of monsoon is sensitive to the large-scale differential heating. The importance of convection in maintaining the tropical circulation and, associated global ¯ow patterns was shown by Krishnamurti et al. (1996, 1998). The role of clouds is equally important during the monsoon period. Sharma et al. (1998) have shown that the radiative forcing associated with clouds is an important diabatic process that is responsible for interannual variation of the monsoon. The sensitivity of large-scale monsoon circulation to different cumulus parameterizations has been studied by several authors (e.g., Donner et al., 1982; Tiedtke, 1984; Slingo et al., 1988, 1992, 1994; Sud et al., 1992). The intensity of Hadley circulation is also affected by the cumulus convection through warming and drying of the upper and lower troposphere. NCMRWF is dedicated towards providing accurate medium range weather forecasts for agricultural planning in India and, improving the skills of tropical weather forecasts in general. It uses a global spectral model originally adapted from NCEP. The horizontal resolution of the model is T80 and has 18 layers in the vertical. A description of the NCMRWF model can be obtained from Mohanty et al. (1995), and Das (1996). The motivation behind this work is to evaluate the simulation of typical features of the Indian summer monsoon using SAS vis-aÁ-vis RAS in the operational model (hereafter referred to as KUO) of NCMRWF for the simulation of the Indian summer monsoon. The need for this study has been felt for two reasons. The ®rst one is related to the general underestimation of rainfall over the Indian region by the operational model (Ramesh
and Iyengar, 1999) and, the second one is due to the limitation of the Kuo's scheme in providing basic frame work for treatment of the cloud processes by a prognostic method. The prognostic treatment of clouds are necessary for providing accurate inputs in terms of cloud liquid water content and cloud cover for radiation parameterization and realistic representation of the hydrological cycle by the model. In order to accomplish these objectives, perpetual integrations have been carried out for the month of July, a period when the monsoon is generally very active over the Indian subcontinent using SAS, RAS and KUO. There are several schemes available at present for the parameterization of cumulus convection. The basic feature that differentiates them is the closure assumption upon which the schemes are based. Majority of the schemes use a mass ¯ux type approach to parameterize the effects of cumulus convection and, they are based on the Arakawa-Schubert (A-S) scheme in some form (i.e., Tiedtke, 1989, SAS, RAS, etc.). The Kuo-type scheme uses convective instability and moisture convergence as a measure to parameterize cumulus convection, while Betts and Miller (1986) uses a mixing line approach for driving actual lapse rate toward moist adiabat. Some of the schemes use available buoyant energy or Convective Available Potential Energy (CAPE) for the closure (i.e., Fristch and Chappell, 1980), while others are based on a presumed equilibrium between surface enthalpy ¯uxes and input of low entropy air into the subcloud layer by convective updrafts (i.e., Emanuel, 1995). The real test of a physical parameterization scheme can be seen only in an operational NWP environment and, not in some isolated test cases. The fact is that almost all the operational NWP centers today use some form of the A-S scheme i.e., SAS at NCEP, RAS at NASA/DAO, mass ¯ux scheme of Tiedtke at ECMWF and economic ArakawaSchubert at JMA. Therefore, we have examined the most important schemes namely KUO, RAS and SAS for the simulation of the Indian summer monsoon. In Sect. 2, the convective parameterization schemes have been described brie¯y. Section 3 presents the simulated global wind ®elds. Sections 4 to 8 present the typical features of the monsoon circulation as simulated by different cumulus parameterization schemes. Finally, a summary of the work is presented in Sect. 9.
Comprehensive test of different cumulus parameterization schemes
2. Description of convection schemes The Kuo-type cumulus parameterization scheme is well known (Anthes, 1977) and therefore, is not described here for brevity. The SAS and RAS schemes are brie¯y described below. 2.1 The simpli®ed Arakawa-Schubert scheme (SAS) [Grell, 1993; Pan and Wu, 1995] In this scheme, the cloud spectrum is reduced to one simple cloud (updraft and downdraft couplet). The cloud size is prescribed to be the largest in the spectrum. Thus the mass ¯ux distribution equation has an exact solution. This makes the scheme very ef®cient. All terms dependent on entrainment and detrainment are set to zero. Thus, the cloud work functions for updraft and downdraft simply become the available buoyant energy for up- and downdrafts, respectively.
229
In the present scheme, the parcels start below 700 mb. The cloud base is considered to be the level of free convection (LFC). From LFC, the parcel is assumed to be non-entraining up to cloud top. Ice phase has been neglected. All convective mass ¯ux detrains at cloud top. Downdraft is included, but its starting level is set near 400 mb just above the level of minimum moist static energy (LMSE), unlike in Grell who set it exactly at the LMSE. 2.2 The relaxed Arakawa-Schubert scheme (RAS) [Moorthi and Suarez (1992)] It modi®es the entrainment relation and assumes that the normalized mass ¯ux is a linear function of height (rather than exponential as in the original A-S). This avoids the costly calculation that is necessary to ®nd the entrainment parameter of clouds. Some of the other simpli®cations made
Fig. 1. Vertical cross section of mean zonal wind obtained from a ANA, b KUO, c RAS, and d SAS
230
S. Das et al.
in RAS are for example, rather than requiring that ``Quasi-equilibrium'' of the cloud ensemble be achieved each time the scheme is invoked, it only relaxes the state towards equilibrium. RAS considers one cloud type at a time. It computes mass ¯ux that would require maintaining the invariance of work function as if no other clouds were present. A fraction of this mass ¯ux is used to change the surrounding and goes on to do same for another cloud type. Thus, each step is in a process of single cloud equilibrium but, in course of time, all cloud types affect one another by modifying the environment. In the following sections, we describe the results of simulations using KUO, RAS and SAS. 3. Global wind ®elds In order to examine the simulation of global winds by different cumulus parameterization schemes,
Fig. 2. As in Fig. 1, but for the meridional wind
the vertical cross sections of the zonal and meridional wind ®elds averaged for the month of July are shown in Figs. 1 and 2. Figure 1a illustrates the mean July analysis (1994±1998) of the zonal wind obtained from the NCMRWF global data assimilation system (hereafter referred to as ANA) for veri®cation purpose. The shaded areas indicate easterlies. A comparison between the NCMRWF analysis and reanalysis ®elds of the NCEP and ECMWF (not shown here) indicated that the NCMRWF analysis ®elds are very close to the two reanalysis ®elds. Results indicate that the middle latitude upper level westerly jet streams of the two hemispheres are fairly well simulated by KUO, SAS and RAS. A close comparison between the three schemes indicate that the strength of the jet stream in the southern hemisphere is better simulated by SAS than compared to the other two schemes. In the northern hemisphere, all the three convection schemes indicate slight
Comprehensive test of different cumulus parameterization schemes
underestimation of the westerlies as compared to ANA. The tropical easterlies are slightly underestimated by RAS, while it is overestimated by KUO and SAS as compared to ANA. The July mean simulation of the vertical cross section of the meridional wind ®elds obtained from ANA, KUO, RAS and SAS are shown in Fig. 2. Results indicate that the strength of the upper level northerlies and lower level southerly winds simulated by RAS is close to ANA. The strength of the tropical upper level meridional circulation is overestimated by SAS and underestimated by KUO, while RAS was comparable to ANA. 4. Low-level jet and cross equatorial ¯ow Low-Level Westerly Jet (LLWJ) is the most important feature of the summer monsoon ¯ow pattern. It interacts with the southern hemisphere's
231
Mascarene High through the Cross Equatorial Flow (CEF). The ®nal strength of the ¯ow is manifested in the north-south pressure gradient along the west coast of India. The Arabian Sea and the Bay of Bengal are among the few regions of earth, where the LLWJ is seen with some regularity in the annual cycle. These jets are related to the synoptic scale forcing and have narrow zones of high-speed ¯ow that extend over hundreds of kilometers. These LLWJ have appreciable horizontal and vertical shear. They are also important for the horizontal and vertical ¯uxes of moisture and, are generally associated with the development and evolution of deep convection over the Indian subcontinent. Figure 3a±d illustrates the strength of the LLWJ at 850 mb as obtained from ANA, KUO, RAS and SAS for the month of July. Wind speed greater than 15 m/s are shaded in the diagrams. The maximum strength of the LLWJ is observed to be
Fig. 3. Mean isotach at 850 mb obtained from a ANA, b KUO, c RAS, and d SAS
232
S. Das et al.
20 m/s in the ANA over the Arabian Sea. This has been overestimated by both RAS and SAS. The LLWJ is con®ned to a narrow belt and is seen to extend over the Bay of Bengal in the simulations by RAS and SAS. The strength of the LLWJ is very weak in KUO as compared to ANA. The LLWJ is related to intense pressure gradient produced by synoptic scale forcing, which in turn is associated with the development and evolution of deep cumulus convection over the region. The strong LLWJ produced by RAS and SAS is due to enhanced convection produced by these two schemes. The vertical cross section of the mean meridional wind at equator obtained from ANA, KUO, RAS and SAS are displayed in Fig. 4. The analyzed ®eld shows prominent maxima of 12 m/s at around 40 E and 875 mb representing the core of the Arabian Sea branch of the cross-equatorial ¯ow. The depth of the southerly wind decreases as
it approaches the longitudes of the Indian land mass. The southerly component of the cross equatorial ¯ow penetrates up to deeper level in the Arabian Sea than compared to the Bay of Bengal. These ¯ow patterns are similar to the features observed during the monsoon season by Mitra et al. (1999). The CEF in the Arabian Sea was intensi®ed by KUO, whereas the simulations by RAS and SAS was closer to analysis. The strong CEF produced by KUO over the Arabian Sea is unrealistic and, may be related to the strong pressure gradient created by unrealistic accumulation of isolated heating at lower levels by this scheme. The strength of the Bay of Bengal branch of the cross equatorial ¯ow around 85 E was more realistic in RAS and SAS compared to KUO. The southerly ¯ow is carried to higher level (up to about 800 mb) over the Indian peninsular region between 70 E±80 E by SAS than compared to the ANA and RAS, where the depth of
Fig. 4. Vertical cross section of the mean meridional wind at equator obtained from a ANA, b KUO, c RAS, and d SAS
Comprehensive test of different cumulus parameterization schemes
233
Fig. 5. Vertical cross section of the mean zonal wind at 75 E obtained from a ANA, b KUO, c RAS, and d SAS
the southerly wind is much shallower between the two longitudes. RAS produced very strong northerly ¯ow between 70 E±80 E above 850 mb. As the southwesterly ¯ow approaches the Western Ghat mountains, the warm moisture laden current is forced to rise up and produce deep convection, which results in intense precipitation over the region. Most of the numerical models fail to simulate the intensity of rainfall observed over the Western Ghat. Figure 5 depicts the vertical cross section of the mean zonal wind at 75 E obtained from ANA, KUO, RAS and SAS. The analyzed wind shows that the core of the LLWJ is located between 800±850 mb around 10 N. Easterly wind appears above 700 mb north of about 25 N, which is the Bay of Bengal branch of the monsoon ¯ow. The simulated ¯ow indicates that the LLWJ is overestimated by 3± 4 m/s by RAS and SAS, but the position of the
LLWJ was close to the analysis. The simulations of RAS produced double cores of the jet, the second one was at 3 N between 600±650 mb. The easterly wind above 700 mb north of about 25 N was also overestimated by 2±3 times by RAS and SAS as compared to ANA. The position of the LLWJ was seen at 17 N in KUO simulation, which was very much north of the analyzed position (9±12 N). 5. Eastery jet and the monsoon trough The Tropical Easterly Jet (TEJ) stream is an inherent feature of the Indian summer monsoon. It is a belt of strong easterly winds, which is a part of the southern periphery of an upper tropospheric anticyclone. The easterly jet is observed between 200 and 100 mb and has speeds often exceeding 100 knots. The TEJ can be traced in the upper
234
S. Das et al.
Fig. 6. Mean isotach at 200 mb obtained from a ANA, b KUO, c RAS, and d SAS
troposphere right up to the west coast of Africa. The location of the easterly jet moves north and south in phase with the northward and southward movement of the monsoon trough (Rao, 1976). Figure 6a±d presents the strength of the TEJ at 200 mb as obtained from ANA, KUO, RAS and SAS averaged for the month of July. Wind speed greater than 20 m/s are shaded in the diagrams. The maximum strength of the TEJ is observed to be 25 m/s in the ANA over the Arabian Sea. This has been considerably overestimated by SAS. The core of the TEJ exceeding 35 m/s is located southeast of Srilanka in SAS and, it extends from western Arabian Sea to the Indian Ocean and Paci®c. The location of the core of TEJ simulated by RAS is similar to that of SAS and is weaker in strength. The TEJ is a part of the southern periphery of an upper tropospheric anticyclone. It may be noted that the intense deep cumulus
convection results in strong upper level divergence. This will be evident in the simulations of the velocity potential by the three schemes as we shall see latter. The strong TEJ compliments well with the strong low level monsoon ¯ow, which is a result of enhanced deep convection produced by SAS. Figure 6a also shows the subtropical westerly jet stream having speeds exceeding 30 m/s north of the Himalayas. None of the simulations have been able to produce the westerly jet stream as seen in the ANA. In order to examine the characteristics of the mean circulation in the vertical plane, Fig. 7a±d illustrates the sectoral average (45 E±100 E) vertical cross section of the zonal wind as obtained from ANA, KUO, RAS and SAS. The diagrams indicate that the core of the TEJ is located at 100 mb and, is strongest in SAS. This is similar to what was found earlier. The strength of
Comprehensive test of different cumulus parameterization schemes
235
Fig. 7. Vertical cross section of the mean zonal wind averaged between 45 E±100 E as obtained from a ANA, b KUO, c RAS, and d SAS
the low-level westerly ¯ow is also overestimated by both SAS and RAS, but the position of the maxima was more close to analysis. The monsoon trough is the most important quasi-permanent feature of the monsoon circulation. Its location and intensity directly affects the distribution of precipitation over the Indian region. During the established phase of the monsoon (July), the trough runs from Northwest India to the Head Bay of Bengal. A northward shift of the trough near the foothills of the Himalayas is usually associated with weaker monsoon activity over major parts of India. During the monsoon season, several transient depressions form over the Bay of Bengal and move towards northwest across the land. Fig. 8 presents the mean streamlines at 850 mb as obtained from ANA, KUO, RAS and SAS. The
diagrams indicate that the axis of the monsoon trough is relatively north of the analyzed position in the simulation by KUO. The monsoon trough is stronger in RAS and SAS as compared to ANA. The RAS and SAS have also produced a closed circulation at the eastern end of the monsoon trough over the Bay of Bengal, but the associated easterlies to the north of the monsoon trough seem to be more realistic. The easterlies are completely missing over the Indo-Gangetic plain in the simulation by KUO scheme. 6. Vertical motion and divergent circulation The sectoral average (45 E±100 E) vertical cross section of the vertical velocity as obtained from ANA, KUO, RAS and SAS are illustrated in Fig. 9. The analyzed vertical motion shows a
236
S. Das et al.
Fig. 8. Mean streamlines at 850 mb obtained from a ANA, b KUO, c RAS, and d SAS
major zone of ascent associated with the northern hemispheric equatorial trough (monsoon trough) around 20 N. Another weak zone of rising motion is located near 10 S, which is associated with the southern hemispheric equatorial trough. In between, there exists a very narrow region of descent over the equatorial region. A major zone of descent is seen around 25 S. The rising and sinking motion around 20 N and 25 S are associated with the ascending and descending branch of the Hadley cell. Results indicate that the rising and sinking motion of the northern and southern hemisphere are 2 to 3 times stronger in RAS and SAS than compared to the ANA. This indicates that the Hadley cell is strengthened by the two convection schemes. Donner et al. (1982), and Zhang (1994) found that the Hadley cell is weakened in the simulation with cumulus parameterization than without using the cumulus parameterization. Tiedtke (1984) studied the Arakawa
Schubert (1974) scheme and the Kuo scheme and his results were consistent with results of Donner et al. (1982). On the other hand, Sud et al. (1992) showed that the Arakawa-Schubert scheme makes the Hadley circulation stronger in the GLA-GCM. In the present simulation, RAS and SAS show strong rising motion than compared to KUO and ANA, which indicates a strong intensity of convection produced by the mass ¯ux type of convection schemes. The level of maxima is seen at a higher level in RAS than compared to other schemes. The present results are consistent with Sud et al. (1992). The 200 mb velocity potential is widely used to study the relationship between convection and the monsoon circulation (Krishnamurti and Ramanathan, 1982; Zhang, 1994; Sperber et al., 1994). The velocity potential is a gross measure of the global-scale divergent and convergent centers and has a wave number 1 structure. Figure 10
Comprehensive test of different cumulus parameterization schemes
237
Fig. 9. Vertical cross section of the mean vertical velocity averaged between 45 E±100 E as obtained from a ANA, b KUO, c RAS, and d SAS
presents the July mean velocity potential and divergent winds at 200 mb as obtained from ANA, KUO, RAS and SAS. The latent heat release in cumulus convection plays a major role in the location of the centers of divergent circulation, which in turn is related to the location of warmest SSTs. Since the monsoon circulation is quite sensitive to the location and intensity of heating rates, use of different cumulus parameterization schemes can cause major differences in the simulated monsoon circulation. The analyzed ®eld shows a strong upper level divergent center located north of the equator near 130 E. Associated with it are the strong meridional and zonal mass ¯uxes that constitute the upper branches of Hadley and Walker circulations. The gradient of the isopleths of the velocity potential determine the strength of overturning of the Hadley and Walker circulations.
The isopleths are nearly parallel to longitude meridians over India in ANA suggesting an east-west overturning there. An active monsoon is characterized by a strong east-west overturning (Krishnamurti, 1971). In the KUO simulation, the intensity of this divergent circulation is weaker and its center is not well de®ned. The simulations by RAS and SAS produced two prominent centers, one in the Bay of Bengal and another in the Western Paci®c. The divergent circulation in the Bay of Bengal corresponds to the intense heating by latent heat release in the cumulus convection during the active monsoon phase of July. The strong upper-level divergent ¯ow over the Bay of Bengal is consistent with the strong TEJ and, formation of organized cyclonic circulation at lower level seen in Fig. 8c,d. The strength of the 200 mb velocity potential and the divergent
238
S. Das et al.
Fig. 10. Mean divergent wind and velocity potential at 200 mb obtained from a ANA, b KUO, c RAS, and d SAS
circulations is slightly stronger in RAS and, located southward than compared to SAS. The orientation and gradients of the isopleths of the velocity potential have also changed over Indian region in the simulations by RAS and SAS as compared to ANA. Sharma et al. (1998) have shown that the radiative forcings due to clouds can change the orientation and gradients of the isopleths of the velocity potential and it could also broaden the longitudinal extent of the divergent circulation to other areas of the monsoon region. 7. Precipitation and the outgoing long wave radiation Figure 11 displays the July mean precipitation (mm/day) obtained from observations, KUO, RAS and SAS. The observed precipitation rates are obtained by combining precipitation estimates from INSAT IR data and the rain-gauge observations
(Mitra et al., 1997). Observed precipitation rates show the two typical maxima of precipitation over the Western Ghat and the Bay of Bengal off the Arakan coast. Intense rainfall is also observed over the Himalayan region. The semi arid regions of northwest India and the eastern coast of the southern peninsular India have much less precipitation owing to the rain shadow effect. Precipitation rate simulated by KUO is reasonably well as compared to the observations. The west coast maxima has been simulated fairly well though not so intense as in observations. Rainfall over the central, west and, northwest India are fairly well represented. The rainfall simulation by RAS indicates that the west coast maxima over the Western Ghat are very well depicted, while the Bay of Bengal rainfall maxima has been shifted much to the south and is close to the east coast of India. Best simulations of the rainfall are obtained by SAS, which shows fairly well distributions of the
Comprehensive test of different cumulus parameterization schemes
239
Fig. 11. Mean rainfall in mm/day obtained from a observations, b KUO, c RAS, and d SAS
two rainfall maxima. The rainfall intensity has also been simulated reasonably well by the SAS. The realistic rainfall simulation by SAS can be attributed to the overall realistic simulation of LLWJ, TEJ and the associated divergent ¯ow pattern at 200 mb by this scheme. KUO is able to simulate the rainfall fairly well where the intensity of convection is weaker, but underestimates the rainfall when the convection is intense. Krishnamurti (1995) has shown that the realistic rainfall patterns can be simulated appropriately only by a high resolution model with proper description of moisture and divergence ®eld. The present de®ciency at T80 resolution may be overcome with a higher resolution model. The Outgoing Longwave Radiation (OLR) from the top of the atmosphere is a good measure of the intensity of convection and, depends upon the surface temperature, cloud amount and height.
Figure 12 presents the July mean OLR obtained from observations (Earth Radiation Budget Experiment, ERBE), KUO, RAS and SAS. The values less than 220 w m 2 are shaded in the diagram. Lower values of OLR indicate deep convection with high-level cirrus. The observations (Fig. 12a) show that a major portion of the Indian subcontinent, the Bay of Bengal and the Southeast Asia are covered by deep convective clouds during July. The simulations by KUO indicate that the lower values of OLR are con®ned to the southern parts of the Indian peninsular and extend over the Arabian sea. The OLR simulated by RAS shows relatively larger areas of low OLR indicating larger areas of deep convection produced by this scheme. The OLR simulated by SAS shows two prominent minima corresponding to the heavy zones of precipitation associated with deep convection over the Western Ghat and the
240
S. Das et al.
Fig. 12. Mean outgoing longwave radiation obtained from a observations (ERBE), b KUO, c RAS, and d SAS
Bay of Bengal off the Arakan coast. In general, RAS produced relatively better OLR than compared to other schemes. 8. Temperature and moisture anomalies In order to get an idea of errors in the simulated temperature and moisture ®elds, the July mean temperature at 850 mb obtained from ANA, and its difference from the simulated temperatures by KUO, RAS and SAS are shown in Fig. 13. The shaded regions indicate cooling. Results show that KUO produces warming of about 1 K over most parts of the Northwest India. It produces cooling of about 1±2 K over rest of the India and adjoining sea. The simulations by RAS indicates that the cooling is increased to 2 K over the southeast coast of India but, it reduced the warming over the
northwest India, which was produced by KUO simulation. The simulated temperature by SAS shows better values than compared to KUO and RAS. At upper level 200 mb (Fig. 14), KUO produced cooling by 3 K almost everywhere over the Indian region. The cooling is reduced by RAS. The cold bias is further reduced by SAS. The reduction of cold bias by RAS and SAS is because of the proper redistribution of heat by deep convective clouds over the region. Convection can redistribute the heating through mixing in the vertical and, warm the atmospheric column by latent heat release. Figure 15 presents the errors in the simulated relative humidity ®eld at 850 for the month of July as obtained from KUO, RAS and SAS. The shaded regions indicate drying. Results show that
Comprehensive test of different cumulus parameterization schemes
241
Fig. 13. Mean temperature at 850 mb obtained from a ANA, and its difference from the simulations by b KUO, c RAS, and d SAS
KUO and RAS produced drying by 10% over the northern India at 850 mb and moistening by 10% over the Arabian Sea. RAS produced excessive moistening by 20±30% over the oceanic regions. SAS produced mixed improvements. It is dif®cult to explain the drying produced by SAS over the eastern equatorial ocean region because, the quality of moisture analysis over the tropical oceanic region is generally not very reliable due to the lack of observations. 9. Summary In order to investigate the impact of different cumulus parameterization schemes in the simulation of the Asian summer monsoon, the operational T80 model of NCMRWF has been integrated for an active monsoon phase (July) using KUO, RAS and SAS convection schemes. Results have
been compared with the mean analysis of wind ®elds obtained from the NCMRWF global data assimilation system (ANA) and other available observations. Examination of results indicate several advantages of using SAS and RAS for the simulation of the Indian summer monsoon as compared to the KUO scheme. This is true in particular, with regard to the simulation of the dynamics of the monsoon ¯ow, thermal structure and the rainfall. For example, results indicate better simulations of the Low Level Westerly Jet (LLWJ), the Cross Equatorial Flow (CEF), and the Tropical Easterly Jet (TEJ) by RAS and SAS than compared to the KUO scheme. The results are also physically consistent, viz., the LLWJ is related to intense pressure gradient produced by synoptic scale forcing, which in turn is associated with the development and evolution of deep cumulus
242
S. Das et al.
Fig. 14. As in Fig. 13, but for 200 mb
convection over the region. The strong LLWJ produced by RAS and SAS is due to enhanced convection produced by these two schemes. The CEF simulated by RAS and SAS are more realistic than compared to KUO. The strong CEF produced by KUO over the Arabian Sea is unrealistic and, may be related to accumulation of isolated heating at lower levels by this scheme. The strength of the Bay of Bengal branch of the cross equatorial ¯ow around 85 E was also more realistic in RAS and SAS than compared to KUO. Our results also show that the TEJ is strongest in SAS. The TEJ is a part of the southern periphery of an upper tropospheric anticyclone. It may be noted that the intense deep cumulus convection results in strong upper level divergence. This is evident in the simulations of the velocity potential by the three schemes. The strong TEJ compliments well with the strong low level monsoon ¯ow, which is a result of enhanced deep
convection produced by SAS. The cold bias at 200 mb seen in the simulation of KUO is reduced by RAS and, is further improved by SAS. The reduction of cold bias by RAS and SAS is because of the proper redistribution of heat by deep convective clouds over the region. Convection can redistribute the heating through mixing in the vertical and, warm the atmospheric column by latent heat release. Proper representation of convective clouds lead to better simulations of the Outgoing Longwave Radiation (OLR) as seen from the results of RAS and SAS. Study of rainfall indicates that the RAS and SAS are able to produce better distribution of rainfall over the Western Ghat Mountains and the Arakan coast, where deep cumulus convection occurs due to orographic forcing of the warm and moist air. However, the rainfall is more close to observations in SAS than compared to the other two schemes. KUO is able to produce good rainfall over the
Comprehensive test of different cumulus parameterization schemes
243
Fig. 15. As in Fig. 13, but for the relative humidity
regions where the intensity of convection is relatively weaker. The realistic rainfall simulation by SAS can be attributed to the overall realistic simulation of LLWJ, TEJ and the associated divergent ¯ow pattern at 200 mb. Thus, evaluation of all these features of the Indian summer monsoon simulated by the three schemes indicate that in general, SAS is able to produce relatively better results than compared to others. Acknowledgements The original model of NCMRWF was adopted from NCEP (USA). The authors are grateful to Drs. Srinivas Moorthi and H.-L. Pan of NCEP for providing the RAS and SAS schemes. Thanks are due to the Head, NCMRWF and Head, Research Division, NCMRWF for their constant encouragement. We would also like to thank the two anonymous referees for their valuable comments, which has led to substantial improvements in the quality of the paper.
References Anthes RA (1977) A cumulus parameterization scheme using a one-dimensional cloud model. Mon Wea Rev 105: 270±286 Arakawa A, Schubert WH (1974) Interaction of cumulus cloud ensemble with the largescale environment, Part I. J Atmos Sci 31: 674±701 Betts AK, Miller MJ (1986) A new convective adjustment scheme, Part I: Observational and theoretical basis. QJRMS 112: 677±691 Das Someshwar (1996) Study of convective heating rates generated at different resolutions with special reference to a depression over the Bay of Bengal. In: Advances in Tropical Meteorology (Datta RK, ed.) New Delhi: Concept Publishing, pp 92±102 Donner LJ, Kuo H-L, Pitcher EJ (1982) The signi®cance of thermodynamic forcing by cumulus convection in a general circulation model. J Atmos Sci 39: 2159± 2181 Emanuel KA (1995) The behavior of a simple hurricane model using a convective scheme based on subcloud-layer entropy equilibrium. Vol 52, pp 3959±3968
244
S. Das et al.: Comprehensive test of different cumulus parameterization schemes
Fritsch JM, Chappell C (1980) Numerical simulation of convectively driven pressure systems. Part I: Convective parameterization. J Atmos Sci 37: 1722±1733 Grell GA (1993) Prognostic evaluation of assumptions used by cumulus parameterization. Mon Wea Rev 121: 764±787 Hahn D, Shukla J (1976) An apparent relationship between Eurasian snow cover and Indian monsoon rainfall. J Atmos Sci 33: 2461±2463 Krishnamurti TN, Sinha MC, Mishra V, Sharma OP (1998) Tropical/Middle latitude interaction viewed via wave energy ¯ux in frequency Domain. Dyn Atmos Ocean (Special issue) 27(1±4): 383±412 Krishnamurti TN, Bedi HS, Rohaly GD, Oosterhof D (1996) Partitioning of the seasonal simulation of a Monsoon Climate. MWR 124(7): 1499±1519 Krishnamurti TN (1995) Numerical weather prediction. Annu Rev Fluid Mech 27: 195±224 Krishnamurti TN, Ramanathan Y (1982) Sensitivity of the monsoon onset to differential heating. J Atmos Sci 39: 1290±1306 Krishnamurti TN (1971) Observational study of the upper tropospheric motion ®eld during the northern hemispheric summer. J Appl Meteorol 10: 1066±1096 Kuo HL (1974) Further studies of the parameterization of the in¯uence of cumulus convection on large scale ¯ow. J Atmos Sci 31: 1232±1240 Mitra AK, Dasgupta M, Rajan D (1999) Mascarene high, cross equatorial ¯ow, low-level westerly jet and northsouth pressure gradient. In: Evaluation and maintenance of the summer monsoon 1998 (Global Data Assimilation ± Forecast System of India, ed) NCMRWF, DST, New Delhi, pp 45±55 Mitra AK, Bohra AK, Rajan D (1997) Daily rainfall analysis for Indian summer monsoon region. Int J Climatol 17: 1083±1092 Mohanty UC, Das Someshwar, Saseendran SA, et al. (1995) Medium range prediction of atmospheric systems over Indian region by NCMRWF forecasting system. Proc 5th regional workshop on Asian/African monsoon emphasizing training aspects (WMO Tropical Meteorology Research Program Series, Report no: 52, WMO/TDNO: 698), pp 51±62 Moorthi S, Suarez MJ (1992) Relaxed Arakawa-Schubert: A parameterization of moist convection for general circulation models. Mon Wea Rev 120: 978±1002 Pearce RP, Mohanty UC (1984) Onsets of the Asian monsoon 1979±1982. J Atmos Sci 41: 1620±1639 Pan Hua-Lu, Wu W-S (1995) Implementing a mass ¯ux convection parameterization package for the NMC medium-range forecast model. Of®ce Note 409, available from EMC/NCEP/NOAA
Ramesh KJ, Iyengar GR (1999) Characteristics of the medium range rainfall forecasts of the Asian summer monsoon. Int J Climatol 19: 627±637 Rao KG, Goswami BN (1988) Interannual variation of sea surface temperature over the Arabian Sea and the Indian monsoon: A new perspective. Mon Wea Rev 116: 558± 568 Rao YP (1976) Southwest monsoon. Meteorological monograph. Synoptic meteorology No 1/1976. India Meteorological Department, New Delhi, 367 pp Sperber KR, Hameed S, Potter GL, Boyle JS (1994) Simulation of the northern summer monsoon in the ECMWF Model: Sensitivity to horizontal resolution. Mon Wea Rev 122: 2461±2481 Sharma OP, Le-Treut H, Seze G, Fairhead L, Sadourny R (1998) Interannual variations of the summer monsoons: sensitivity to cloud radiative forcing. J Climate 11: 1883± 1905 Slingo JM, Blackburn M, Betts A, Brugge R, Hoskins B, Miller M, Steenman-lark L, Thuburn J (1994) Mean climate and transience in the tropics of the UGAMP GCM: Sensitivity to convective parameterization. Q J Roy Meteorol Soc 120: 881±922 Slingo JM, Sperber KR, Morcrette J-J, Potter GL (1992) Analysis of the temporal behavior of convection in the tropics of the ECMWF model. J Geophys Res 97D: 18119±18135 Slingo JM, Mohanty UC, Tiedtke M, Pearce RP (1988) Prediction of the 1979 summer monsoon onset with modi®ed parameterization schemes. Mon Wea Rev 116: 328±346 Sud YC, Chao WC, Walker GK (1992) Role of cumulus parameterization scheme in simulating atmospheric circulation and rainfall in the nine-layer Goddard Laboratory for Atmospheres General Circulation Model. Mon Wea Rev 120: 594±611 Tiedtke M (1989) A comprehensive mass ¯ux scheme for cumulus parameterization in large-scale models. Mon Wea Rev 117: 1799±1800 Tiedtke M (1984) The effect of penetrative cumulus convection on the large scale ¯ow in a general circulation model. Beitr Phys Atmos 57: 216±239 Vernekar A, Zhou J, Shukla J (1995) The effect of Eurasian snow cover on the Indian monsoon. J Climate 8: 248±266 Zhang GJ (1994) Effects of cumulus convection on the simulated monsoon circulation in a general circulation model. Mon Wea Rev 122: 2022±2038 Author's address: Dr. Someshwar Das, NCMRWF, Satmet building, Mausam Bhawan, Lodhi Road, New Delhi110003, India (E-mail: somesh@ncmrwf:gov:in; somesh@ scientist:com)