A Modeling Diagnosis of the Development of Mesoscale Convective Systems over the South China Sea during the Summer Monsoon Onset in 1998

In this study, PSU/N CAR MMS is used to simulate the evolution of the 1998 East Asian summer monsoon onset over the South China Sea (SCS). Simulation outputs are further utilized to study the development of mesos­ cale convective systems (MCSs) and their interactions with the synoptic/ large-scale environment in the SCS region. During the monsoon onset, over southern China and the SCS, high equivalent potential temperature, con­ vective instability, and the southwesterly confluence in the lower tropo­ sphere provided favorable large-scale thermodynamic and dynamic condi­ tions for the MCS development. Model simulated results reveal that, in a warm-moist and unstable atmosphere, the downdraft outflows as a result of the MCS development could become a possible mesoscale triggering mechanism for the formation of a new MCS. Sensitivity tests depict the effect of latent heat release play­ ing a significant role in the establishment of monsoon circulation. The ter­ rain effect is significant for local cloud formation over land. For well orga­ nized MCSs occurring over the SCS, the influence of the mesoscale wind field is more important. The latent heat release in MCSs heats the atmosphere, reduces the surface pressure, and therefore enhances the low­ level wind convergence and cyclonic circulation. The mesoscale adjustment of wind field and mass field, on one hand, enhances the MCS development or initiates the formation of a new MCS, and the organization of MCSs, on the other hand, integrates the mesoscale disturbances into a synoptic struc­ ture and forms the local monsoon trough. Without the latent heat release, monsoon circulation cannot establish itself even if the model provides fa­ vorable large-scale boundary forcing. 1 Department of Atmospheric Sciences, National Central University, Chung-Li, Taiwan 2 Central Weather Bureau, Taiwan * Corresponding author address: Prof. Song-Chin Lin, Department of Atmospheric Sciences, National Central University, Chung-Li, Taiwan; E-mail: sclin@cc.ncu.edu.tw 369 370 TAO, Vol. 14, No. 4, December 2003 (

atmosphere, the downdraft outflows as a result of the MCS development could become a possible mesoscale triggering mechanism for the formation of a new MCS.Sensitivity tests depict the effect of latent heat release play ing a significant role in the establishment of monsoon circulation.The ter rain effect is significant for local cloud formation over land.For well orga nized MCSs occurring over the SCS, the influence of the mesoscale wind field is more important.The latent heat release in MCSs heats the atmosphere, reduces the surface pressure, and therefore enhances the low level wind convergence and cyclonic circulation.The mesoscale adjustment of wind field and mass field, on one hand, enhances the MCS development or initiates the formation of a new MCS, and the organization of MCSs, on the other hand, integrates the mesoscale disturbances into a synoptic struc ture and forms the local monsoon trough.Without the latent heat release, monsoon circulation cannot establish itself even if the model provides fa vorable large-scale boundary forcing.
1 Department of Atmospheric Sciences, National Central University, Chung-Li, Taiwan 2 Central Weather Bureau, Taiwan

INTRODUCTION
The East Asian summer monsoon is characterized by atmospheric circulations of differ ent spatial and temporal scales.The majo r components of the monsoon system comprise the western Pacific subtropical high, the northeasterly flow associated with the upper-level anticyclone, the monsoon trough over the South China Sea (SCS) and western Pacific, the low-level southwesterly flow and the Mei-Yu front.The interaction between components of different scales leads to the maintenance and change of the monsoon circulation structure during its northward migration.The SCS, with its unique geographic location and being the convergent zone of tropical southeasterly, southwesterly and cross equatorial flows, has been thought to be the origin of the East Asian summer monsoon onset (Tao and Chen 1987;Lau and Yang 1997).
The SCS monsoon onset is a multi-scale adjustment process of atmospheric circulations, which is regarded as the outset of the northward intrusion of tropical warm-moist air.It is also a significant feature of seasonal transition in the East Asia.In the Asian monsoon region, SCS is the only region in the tropics where midlatitude frontal systems can reach.Chang and Chen (1995) pointed out that, with a favorable large-scale setting, the southwesterly flow associated with the front might bring in the warm-moist air from the tropics and trigger the outburst of convection over the SCS.Lin (1996) revealed that the southwesterly flow from the Bay of Bengal provided the thermodynamic mechanism for the westward and northward develop ment of the South Asian anticyclone.The summer monsoon onset possibly could be triggered by the coincidence of the southwesterly moisture surge and the southward intrusion of the Mei-Yu front.Based on the analyses of low-level atmospheric circulations in the East Asian monsoon region, Kueh and Lin (2001) found that among 14 years during 1979-1995 the East Asian summer monsoon onset is accompanied by the frontal activities in the southern China area.Whether the southward intrusion of midlatitude fronts and the enhancement of south westerly flows are significant signals of the SCS monsoon onset is still controversial.However, the role of frontal convective disturbances produced in the northern SCS region as a mesoscale triggering mechanism of the convections over the central and southern SCS is worth further investigation.Ding and Liu (2001) pointed out that the SCS summer monsoon onset in 1998 consisted of two stages.In the first stage, sudden intensification of the southwesterly flow over the northern SCS and southward intrusion of the frontal cold air played important roles in trigger ing the monsoon onset.Johnson and Ciesielski (2002) also conducted a detailed study on the evolution of the first stage monsoon onset in the northern SCS region.However, further dy namic connection between the two stages of monsoon onset has not yet been studied.Though Li and Wu (2000) did not regard the first enhancement of the southwesterly flow over the northern SCS as the onset of summer monsoon, they also did not deny the important roles of the frontal southwesterly flow and convective activity in the monsoon onset processes.In their study, Shao and Qian (2000) depicted the SCS monsoon onset in 1998 as a process in which the southwesterly flow started first over the northern SCS, withdrew southward, and then propagated from south to north again.
Due to geographic differences, the northern SCS and the central-southern SCS exhibit different weather and climate regimes during the summer monsoon.Influences by alternate tropical southwesterly and midlatitude weather systems are found in the northern SCS region, while the tropical southwesterly is a major factor in the central-southern SCS region.However, convective activities over the central-southern SCS can still have connections with the frontal convective disturbances over the northern SCS.In this study, a numerical simulation of the 1998 summer monsoon onset is conducted by using PSU/NCAR MM5 (Pennsylvania State University/National Center for Atmospheric Research Mesoscale Model Version 5).Simula tion results were further utilized to study the development and their interactions with the syn optic/large-scale environment of the mesoscale convective systems (MCSs) occurring in the SCS area.Sensitivity tests were also performed to investigate the influences of the tropical southwesterly and the surrounding terrain on local convections over the SCS.

CHINA SEA SUMMER MONSOON ONSET IN 1998
Based on the criteria of intensification of 850 hPa westerly zonal wind and sudden in crease of precipitation, Ding and Liu (2001) found that a burst of southwesterly occurred first in the fourth pentad of May over the northern SCS and then over the entire SCS in the fifth pentad.Retreat of the western Pacific subtropical high was also observed in both onset stages.They emphasized that the first onset stage was closely related to the rapid development of a cyclone over the Bay of Bengal and the southward propagation of the midlatitude front.Li and Wu (2000) defined the 21 '1 of May as the 1998 summer monsoon onset date when the south westerly prevailed over the entire SCS and the subtropical high withdrew completely from the SCS.Determined by the intensifying signals oflow-level southwesterly and upper-level easterly, Shao and Qian (2000) regarded the 251h of May as the onset date.Chan et al. (2000) also treated the 2Yh of May as the onset date, which is defined by the reversal of 850 hPa zonal wind and a sudden increase of rainfall over the SCS with 5-day running mean data.Instead of a specific onset date, this study regards 18-25 May as the 1998 summer monsoon onset period based on the discussions in Kueh and Lin (2001).Major factors considered in Kueh and Lin (2001) are the 850 hPa flow pattern, the eastward retreat of the subtropical high and convec tive activities over the SCS.The onset period determined in this way also covers the specific onset dates defined by the aforementioned studies.In order to investigate the convective ac tivities occurring over the SCS in detail, this study further divides the whole evolution of the onset process into three sub-periods based on the cloud features revealed by the enhanced GMS5 satellite images.
The first sub-period, spanning 15 May to 19 May, represents the days of pre-onset and early onset.During this sub-period, cold air intruded southeastward persistently from the Asian In the following, synoptic weather conditions and convective activities for the three sub periods will be further illustrated.could possibly be the residuals of the dissipating frontal clouds.Spanning the third sub-period, from 23 May to 25 May, the synoptic setting changed remarkably over the East Asian continent and the SCS.On 23 May, while high pressure H3 moved eastward to the ocean, surface low pressure center L and the front to its northeast moved southward (Fig. lf).Following this, a strong cold-core high pressure center H4 propa gated rapidly southward from 45°N to 30°N, and then weakened (Fig. lh).During the south ward intrusion of strong cold air, the front moved eastward to the ocean and the low pressure center L moved southeastward into the Bay of Tongkin.The low pressure center L then, passing the Hainan island, moved eastward continually.

Mesoscale Convective Activities
Serial convective clouds developed and then decayed gradually over the ocean east of Hainan island after the arrival of the first front on 15 May.These convective systems reached their maximum intensity in the afternoon with distinct diurnal variations.In the subsequent 12 hours after OOOOUTC on 16 May, several fragmentary clouds organized into a couple of MCSs of relatively larger size.They are indicated by A, B, C 1 and C 2 in Fig. 2a.Six hours later at 1800UTC, C 1 and C 2 moved eastward and merged into a larger MCS C covering the whole area of the Bay of Tongkin (Fig. 2b ).The MCS B grew and moved southward a little bit, while MCS A became stationary but exhibited features of persistent rebirth and dissipation (denoted by A' in Fig. 2b).At OOOOUTC on 17 May, A', B and C merged again into an even larger MCS D (Fig. 2c).The evolution of convective systems over the ocean showed a feature that the larger MCS endured in longer life and reached their maximum intensity in the early evening.The convective clouds o_ yer Indo-China, however, maintained a distinct diurnal variation and reached their maximunfintensity in the afternoon.The MCS D then moved southward down to 15°N and evolved, with the frontal clouds, into a northeast-southwest oriented convective cloud band.In the evening of 17 May, MCS D began to dissipate and a new MCS D' appeared in the northeast near the western end of the front (Fig. 2d).Fujita (1959) pointed out that, during the decaying stage of an intense convective system, the evaporative effect of raindrops within the descending flow could lead to the development of a meso-high pressure.As seen in the satellite cloud images, the boundary of a meso-high pressure often is revealed as an arc-shaped cloud band evolving out from the decaying convec tive system (Purdom 1973).Purdom (1979) further emphasized that the arc -shaped cloud band played a critical role in the formation and development of a new intense convective system.Numerical simulation results of Zhang and Bao (1966a, b) and their related earlier studies (Zhang andFritsch 1986, 1987) also demonstrated the important role of outflow boundaries in the initiation of a new MCS.In the present study, the triggering of new convections in the vicinity of a decaying convective system could occur in a chain and is very likely the mecha nism accounting for the evolution of convective systems over the SCS.
During 18-19 May, at the western end of the frontal zone, the development of MCSs exhibited the formation, eastward moving, merging and decaying processes.Two types of convective cloud patterns, roughly in line-shape perpendicular to the frontal surface, were observed near the western end of the frontal zone.One type of clouds formed over land and then grew up over ocean, such as MCS E1 in Fig. 2e.The other type of clouds began to decay once moving over ocean, but new cells developed, such as MCS E2 in Fig. 2f.On 18 May the line convection was oriented in the northwest-southeast direction (Fig. 2e), but changed into northeast-southwest direction on 19 May (Fig. 2f) .The 850 hPa prevailing wind over the northern SCS also changed from southeasterly to southwesterly during the orientation shift of line convections, indicating a close relationship between the low-level wind field and the de velopment of convection.At l 800UTC on 19 May one of the arc-shaped cloud bands evolved and organized into a new MCS F with several intense convective cells inside (Fig. 2g).After formation, MCS F stayed temporarily over the northern SCS and then reorganized into an other new MCS F' (Fig. 2h) after cell separation and dissipation.
The MCS F' moved southward first, then mixed and ceased gradually with the frontal  2i).Twelve hours later at OOOOUTC on 22 May (Fig. 2j), another new cell H started to develop at the northern tip of the line convection.In the evening on the same day, this new cell further moved southeastward, leaving the northern SCS a clear sky temporarily (Fig. 2k).The MCS H didn't last too long.When H was decaying, an arc cloud to its west developed into a new intense MCS I (Fig. 21).In contrast to the MCSs which occurred before, MCS I moved in a west-southwest direction and ceased gradually (Fig .2m).
During the period from 1800UTC 23 May to OOOOUTC 24 May, the remnants of MCS I merged with small convective clouds near the Philippines into a new MCS K (Fig. 2n).The MCS K shifted a little bit southward but grew rapidly into a MCS of considerable size (Fig. 2o ).Accompanied with the growth of MCS K, several MCSs also developed over the northern SCS near Hainan island (indicated by J and L in Fig. 2p ).Up to this moment, regarded as the onset of summer monsoon by many previous studies, the entire SCS were nearly covered with deep convections.After the onset, dissipation and redevelopment of MCSs .sustained for sev eral cycles until early June (see one of the examples in Figs.2q, r).

Surface Observations at Dongsha and Nansha Islands
Surface observations taken at Dongsha (20°42'N, 116°43'E) and Nansha (10°23'N, 114°22'E) islands during the 1998 SCSMEX (South China Sea Monsoon Experiment) were further shown to illustrate the transitional characteristics of monsoon onset in the northern SCS and the central-southern SCS regions.Figures 3a, b show the time series of surface obser vations at the Dongsha island.Before 15 May, the sky was almost clear, except for short durations of very weak precipitation occurring on 2 and 4 May (Fig. 3a).On 15 May and during 18-20 May, intense precipitations occurred, which were related to two frontal passages in the first sub-period.Between 20 May and 30 May, perceivable precipitation was measured only on 24 May.Observed relative humidity shown in Fig. 3a behaves consistently with the measured precipitation.After full monsoon onset, the relative humidity maintained a high value, above 85%.The observed surface winds in Fig. 3b show variable wind directions be fore the onset, and in prevailing southwesterlies after the onset.
The time series of surface observations at the Nansha island show distinct seasonal varia- : .µ=:,cLCN c tions and depict clearly the monsoon onset process.Sudden increase of intense precipitation was observed coincidently with the monsoon onset (Fig. 3c).The very intense rainfall mea sured on 23 May was produced by the MCS I in Fig. 21.The relative humidity was also raised and maintained at a high value after the onset (Fig. 3c ).Transitional variations of wind direc tion and wind speed (Fig. 3d) are more evident than those revealed by the observations at the Dongsha island.In early May, Nansha island was located to the south of the subtropical high ridge.The surface wind was mainly from an east-northeast direction.In the middle of May when the subtropical high weakened, the surface wind became southeasterly with wind speed somewhat reduced.However, the wind speed increased remarkably and the wind direction switched from southeasterly to southwesterly near the stage of full monsoon onset.The change of wind direction and the sudden increase of the westerly wind component could be partly due to the eastward retreat of the subtropical high and the moving of low pressure center L into the SCS area.As revealed by Zhang and Bao (1966a, b) in a case study of a long-lived midlatitude MCS development, the convective activities could also play an important role in the mainte nance of the low pressure center and the enhancement of the southwesterly wind.This will be further discussed in Section 4.

MMS MODEL DESCRIPTION AND SIMULATION DESIGN
The spatial and temporal resolution of conventional sounding data is not sufficient to discern the mesoscale circulation characteristics of convective systems occurring over south ern China and the SCS during the summer monsoon.Sounding observation alone is also inca pable of examining the forcing and feedback mechanisms between the mesoscale circulation and its environment.Model simulation is therefore the most effective tool.PSU/NCAR MM5 has been used successfully in simulating MCS development and low-level jet formation asso ciated with a Mei-Yu front (Chen et al. 1997(Chen et al. , 1998a, b), b).This study thus conducted a model simulation of the 1998 summer monsoon onset by using MM5.Simulation results were further utilized to investigate the development and their interaction with the synoptic/large-scale en vironment of the MCSs occurring in southern China and the SCS regions.
The MM5 is a 3-dimensional non-hydrostatic, primitive equation model with terrain fol lowing vertical coordinates.The model is capable of multiple nesting to account for scale interactions, and has rather complete physics.The physical parameterizations include precipitation, planetary boundary layer, radiation, and surface fluxes (Grell et al. 1995).The initial inputs adopt NCEP GDAS (National Center for Environmental Predictions Global Data Assimilation System) 2.5°x2.5°latitude/longitude grid data with a 12-hour interval.Model simulations were performed on two nested interactive grids as shown in Fig. 4. The coarse grid (DOl) is a 90 km grid with a mesh size of 52x42, and the fine grid (D02) is a 30 km grid with a mesh size of 64x64.Both grids have 23 CY-levels in the vertical.The coarse grid domain covers the East Asian continental region, the Inda-China peninsula, the SCS and the western Pacific.The fine grid domain covers only the SCS region, which is intended for studying the triggering mechanism of the MCS development.
In order to examine the role of convective latent heating and the effect of surrounding terrain in the MCS development over the SCS, three model runs were designed, including a control case (CNTL), a case without latent heat release (NLHR), and a case using low resolu tion terrain data (LTER).For the control case, the coarse grid and the fine grid use 30 minute (-S6 km) and 10 minute (-19 km) terrain data, respectively.The simple ice scheme is speci fied to account for the resolvable precipitation, while the Grell cumulus parameterization scheme is chosen to treat the subgrid-scale convective clouds (Grell et al. 199S).A high-resolution Blackadar' s planetary boundary layer parameterization scheme is used to forecast the surface vertical eddy fluxes (Zhang and Anthes 1982).The NLHR case has essentially the same speci fications as those of the CNTL case, except that grid-and subgrid-scale latent heat releases of condensation are ignored.The design of the NLHR case is to test whether the MCS develop ment over the SCS has significant influences on the enhancement of the southwesterly wind or not.Both grid domains of the LTER case use one.degree (-111 km) low-resolution terrain data with the same initial inputs and physical parameterizations as specified by the CNTL case.
The LTER case is performed by reducing the terrain slopes to understand the barrier effect of the western terrain boundary on the eastward marching of the southwesterly flow.

SIMULATION RESULTS AND DIAGNOSES
To validate the model simulations, we first compare the 8SO hPa height and wind fields produced by the CNTL runs with the observed fields.Figure S shows the daily averaged 8SO hPa observed height and wind analyses for the period 18-2S May.The observed features of the first sub-period include the eastward retreat of the subtropical high, the 8SO hPa height falling over the SCS, the gradual increase of the southwesterly over the northern SCS, and the devel opment of a weak pressure trough near Hainan island (Figs.Sa, b).During the second sub period (Figs.Sc, d, e ), the low pressure system L and the pressure trough associated with the surface front moved eastward, the intense north-south pressure gradient built up and the west-erly wind component increased over the northern SCS, the 850 hPa height fell and the wind weakened over the southern SCS.At the monsoon onset (Figs.5f, g, h), southwesterly flow from the Tropics enhanced noticeably, intense north-south pressure gradient and strong south westerly extended to the central-southern SCS, a cyclonic low pressure center developed over the northern SCS.The corresponding simulated 850 hPa height and wind fields of the coarse domain are shown in Fig. 6.Comparisons show reasonable agreement between observed and simulated fields.The model simulations basically can capture the major features of each sub period.In the following discussions, the simulated precipitation is a 12-hour accumulated value.When discussed in association with the accumulated precipitation, the sea level pres sure and surface wind mentioned will be those of the last time period of the 12-hour interval, and the divergence field referred to will be the 12-hour time-averaged value.For the purpose of discussing MCS development, the locations of perceivable accumulated precipitation will be treated as the model simulated MCSs.

The First Sub-period -Frontal Activities
In the first sub-period, a frontal cloud band appeared stationary along the coast of south ern China, several MCSs formed in the warm sector and gradually organized into a convective cloud band perpendicular to the front.Figure 7 shows the accumulated precipitation with 850  Figure 8 shows the accumulated precipitation with 1000 hPa wind analysis (left panels) and sea level pressure (right panels) of the fine domain CNTL runs for the same time periods as shown in Fig. 7.Over the northern SCS, low-level wind divergence or convergence was found in local areas within and around the MCS.This kinematic feature plays an important role in the MCS development.At OOOOUTC on 18 May, the MCS located to the southeast of Hainan island was accompanied with low-level convergence (Fig. 8a).Thus, it grew vigor-'�\ • ' ' • • � -"">,,,., fr.,_ )�,., .' 'i'" '-.'.\ tr,.--.-/"i,,I " ' \, .":/•, , .. ' r "t' Fig. 6.Same as Fig. 5 except for coarse domain CNTL runs.
ously in the following 12 hours (Fig. 8c).Another example is the line-shaped MCS located to the west of the Philippines.At OOOOUTC on 18 May, low-level wind divergence and conver gence were seen in the northern and the southern sections of the MCS, respectively (Fig. 8a).
In the next 12 hours, the southern section of the MCS continued growing, but the northern section dissipated due to unfavorable conditions (Fig. 8c).As seen in 3-hour simulation out puts (not shown), the initiation of a new MCS often happens in a time scale shorter than 12 hours and most frequently in the western and the southern peripheries of a decaying MCS since the environmental flows are mainly from the southwest.Phenomena observed in model simulations suggest that the low-level outflow of a decaying MCS may play a triggering role in the formation of a nearby new MCS if the outflow and its environment constitute a favor able condition for convection.This result supports the observational analyses of Purdom (1979) and is consistent with the numerical simulations of Zhang and Bao (l 996a, b ).During 1200UTC 17 May-1200UTC 18 May over the northern SCS, though the large-scale environment had high equivalent potential temperature and was convectively unstable in the lower troposphere, the horizontal temperature gradient across the frontal zone was weak.Thus, mesoscale forcing plays the primary role in the development of MCS, but not the frontal lifting.In the sea level pressure analyses shown in Figs.8b, d, mesoscale low pressure and high pressure are found in the developing and decaying stages of a MCS, respectively.Taking the MCS which located to the southeast of Hainan island as an example, the low-level wind convergence occurred in a local area of low pressure at its growing stage (Fig. 8b ).This low pressure zone became a high pressure center as the MCS began to decay (Fig. 8d).
Coarse domain CNTL and NLHR model simulations of sea level pressure, 8SO hPa equiva lent potential temperature, 8SO hPa wind and 200 hPa wind analyses along 20°N and 1S0N latitude lines are displayed in time sequence as shown in Fig. 9.The left panels of CNTL runs (Fig. 9a, b) show that high 8SO hPa equivalent potential temperature, low sea level pressure and increasing of 8SO hPa southerly or southwesterly wind were found during the MCS development.On the other hand, increasing of sea level pressure, decreasing of 8SO hPa equiva lent potential temperature, weakening of 8SO hPa southwesterly wind or diffluent flow were found during the MCS decaying stage.The north-south differences in the upper-level 200 hPa winds are more evident.Short durations of 200 hPa wind direction variations were associated with the MCS development over the northern SCS (20°N).To its south over the central SCS (1S0N), the upper-level winds are relatively weak, and there appear increasing northerly wind components during the MCS development.The NLHR simulations shown in Figs.9c, d depict that, without latent heat release, the 8SO hPa southwesterly wind over the entire SCS became weaker apparently, and the 850 hPa easterly wind component coming from the outflow of the subtropical high became stronger over the central SCS.The upper-level northwesterly wind strengthened persistently, particularly over the central SCS.Sea level pressures along both latitudes were also raised appreciably.These contrasts between CNTL and NLHR runs reveal the significant influence of the MCS development on the environmental flow.
-: : .,,,. ,,,,. .ence of mesoscale circulation forcing is more important.west-southeast direction (Figs. 2f,g,h).The coarse domain accumulated precipitation analy ses (Fig. 10) show, over the northeastern SCS and the Bashi Channel, an organized convective cloud band detaching from the frontal cloud band.Within the 12 hours from 1200UTC 19 May to OOOOUTC 20 May (Fig. lOa), along 20°N latitude line, the 850 hPa westerly wind compo nent entering the MCS area from the west increased, and the low-level south inflows of MCSs occurring over the northwestern and the northeastern SCS were coming from the west and the east, respectively.However, during the period from OOOOUTC 20 May to 1200UTC 20 May (Fig. lOc), the south inflow of the MCSs occurring over the northern SCS was mainly southwesterly, coming from the west.A weak low-level cyclonic circulation was found in the neighborhood of the MCSs, and the 850 hPa winds over the central-southern SCS were getting weaker.The variation of wind field depicts the eastward retreat of the subtropical high and the propagation of the southwesterly flow, which exerts direct influences on the MCS develop ment over the SCS.In addition, strong southwesterly winds appeared in the southern China region.This is the southwesterly flow ahead of the low pressure system L, which will be further discussed later.In Figs.lOb , d, the 925 hPa convergent zones coincided with the loca tions of MCSs.In the upper troposphere, the 200 hPa winds maintained a westerly-northwest erly diffluent flow pattern and the outflows from the MCSs were almost toward the east.In  lb, d).An other feature of the MCS development during this sub-period is worth mentioning.On the 850 hPa isobaric surface, weak northeasterly winds (Fig. 11) with low pressure and low equivalent potential temperature (not shown) were often found to the northeast of a newly formed MCS.The northeasterly wind brings in cold air from the north, which is anticipated to increase the low-level instability and favor the MCS development.
In contrast to the sporadic feature of MCS development during the first sub-period, the simulated results of the CNTL runs reveal that MCSs occurring in the second sub-period are rather organized, the low pressure centers are more evident with larger spatial extent, and the 850 hPa southwesterly winds are stronger and more widely spread.However, without latent heat release, the NLHR runs are not able to produce similar features (not shown).For the LTER runs, the simulated results over the SCS are similar to those of the CNTL runs (not shown).As mentioned before, the terrain effect is only significant for local convections over land.
Time-sequence results show that organized MCSs were mainly occurring over the north ern SCS during the period from 1200UTC 19 May to OOOOUTC 21 May and exhibited mesos cale characteristics similar to those depicted in the first sub-period (not shown).After 21 May, significant decrease of sea level pressure was noted extensively over the SCS, and the increase of westerly wind components was also very pronounced over the southern SCS (Figs. 13a, b ).
In the upper-level, the flow over the northern SCS changed gradually from westerly to northerly, while the outflow associated with the MCSs occurring over the southern SCS showed difflu ent pattern obviously.The simulated results of the NLHR runs presented in Figs.13c, d, however, didn't show extensive decreasing of sea level pressure, intensifying of westerly wind compo nent in the southern SCS region, and changing of upper-level flow.This indicates that the convective latent heat release plays an important role in the fall of sea level pressure and the enhancement of westerly wind over the SCS.Again, this is consistent with the findings of Over the northern SCS along latitude 20°N, the simulated results of the NLHR runs (Fig. 13c) didn't show low 850 hPa equivalent potential temperature to the east of MCSs similar to that of the CNTL runs (Fig. 13a).During the time period shown in Fig. 13, the sky was virtu ally clear in the vicinity of Taiwan and the Bashi Channel.The differences between CNTL and NLHR runs suggest that the northeasterly wind and cold dry air to the north of MCS H (Figs. 2j,k) could be due to local intensifying of northerly wind components caused by low-level convergence of MCSs and the channel effect of the Taiwan Strait.

The Third Sub-period -Mesoscale Convective Activities
The third sub-period is a period with MCSs developing over the entire SCS.During 1200UTC 23 May-OOOOUTC 24 May, MCSs near Hainan island developed under the favor able condition of low-level confluence of strong northerly and westerly winds (Fig. 14a).The MCSs near the Philippines were accompanied by westerly winds to their west, but the strong southwesterly winds were limited to the southern flanks of MCSs only.The local feature of southwesterly winds could be an evidence of the accelerating effect as a result of the MCS 30N-.--�-.,..-:-:-:-,:T"'.:-=,...,. .. .-. .. .. .,. ..,. ..,. .. .. ..,.,  Figure 15 shows the fine domain CNTL simulations; the low-level outflows as a result of the MCS development near the Philippines during the time period 1200UTC 23 May-OOOOUTC 24 May (Fig. 15a) led to vigorous convections near the southern flanks of the preceding MCSs in the next 12 hours (Fig. 15c).Similar MCS developments also occurred near Hainan island.This chain development of MCS persisted for a few days over the SCS after the monsoon onset (not shown).As shown in Figs.15b, d, high sea level pressure is found in association with the descending outflows.Over the northern SCS, on the average, the sea level pressure decreases appreciably during the 12-h period.-90 -70 Displayed in Fig. 16 are time-sequence coarse-domain simulations for the time period from 0600UTC 23 May to OOOOUTC 25 May.The 850 hPa equivalent potential temperature of CNTL runs (Figs.16a, b) exhibits temporarily increasing and then decreasing corresponding to the MCS developing and decaying stage, respectively.The lowering of sea level pressure and intensifying of 850 hPa southwesterly wind are also very evident over the SCS.These important features, however, were not seen in the NLHR runs (Figs. 16c,d).After 1800UTC 23 May, along latitude 20°N in the longitude range 105°-115°£, the 850 hPa northerly winds of the NLHR runs were much weaker than those of the CNTL runs.The strengthening of northerly winds is partly attributed to the southward propagation of high pressure, but the differences between CNTL and NLHR runs also indicate the importance of the latent heat effect.
Summarized from the simulated results of the three sub-periods, the monsoon trough and regional low pressure center seem to be constructed progressively from the local pressure and wind disturbances associated with the MCS development.The circulation pattern over the SCS also changes gradually from local small scale disturbances into a flow regime of synop tic-scale feature.This reveals the multi-scale characteristics of monsoon evolution over the SCS.The NLHR simulations of both coarse and fine domains for the three sub-periods all show similar conclusions that the southwesterly flow and monsoon trough would not develop successfully over the SCS without the effect of latent heat release.The surface low pressure center as indicated by Lin Fig. 1 exhibited different character istics during the monsoon evolution.In the early stage, its formation and southward propaga tion were occurring mainly in the midlatitude baroclinic zone over land.In the later stage, the low pressure center moved southeastward over the ocean where observations are virtually void.Benefited by the model simulation, we are able to examine the evolution of the low pressure center over the northern SCS and its relationship to the development of monsoon trough and convective activities over the SCS.
The coarse-domain simulations of sea level pressure and 6-hour accumulated precipita tion for time period OOOOUTC 22 May-0600UTC 24 May are shown in Fig. 17 with an inter val of 6 hours.While L was moving southward over land, the low pressure center coincided with the distribution of MCS pretty well.During 1200UTC-1800UTC on 22 May, the local low pressure center near Hainan island weakened considerably when the MCS over the north ern SCS dissipated, which points to an evidence of the mesoscale mass adjustment in associa tion with the MCS development.After 1200UTC on 23 May, L moved southward over the Bay of Tongkin and was further maintained as a local low pressure center in company with the MCS development over the SCS.While L was over land, the NLHR runs (not shown) can still simulate the southward movement of the low pressure center except with weaker strength.However, the low pressure center disappeared immediately when it moved over ocean, sug gesting the critical role of the latent heat effect in the maintenance of the low pressure center over the SCS.The convective impact on the evolution of the low pressure system is quite  similar to a case study conducted by Zhang and Bao (1996a, b) on the oceanic cyclogenesis as induced by a MCS moving offshore.
Model simulated results reveal that, in a warm-moist and unstable atmospheric environment, the mesoscale descending outflow resulting from the MCS development becomes a possible mesoscale triggering mechanism for the formation of a new MCS.The latent heat release in MCS heats the atmosphere, reduces the surface pressure, and therefore enhances the low-level wind convergence and cyclonic circulation.The mesoscale adjustment of wind field and mass field, on the one hand, feeds back and enhances the MCS development or initiates the forma tion of a new MCS, and the organization of MCSs, on the other hand, integrates the mesoscale disturbances into a synoptic structure and establishs the local monsoon trough.Sensitivity tests depict the effect of latent heat release as it plays a significant role during the monsoon onset process.The terrain effect is significant for local convective activities over land.For  well organized MCSs occurring over the SCS, influence of mesoscale circulation forcing is more important.Without the latent heat release, the monsoon trough cannot form even if the model provides favorable large-scale boundary forcing.Thus, it can be concluded that the establishing of the monsoon trough, lowering of sea level pressure, and increasing of relative vorticity are all closely related to the MCS development.Over the SCS, the MCS development causes mesoscale wind disturbances fi rst, then adjusts the mass field, and finally changes the synoptic environment and establishes the monsoon circulation.
Finally, it should be emphasized that although local low-level wind convergence is an important factor in the formation of an MCS, the fine domain simulation results also indicate that it is still not sufficient to initiate the formation of a new MCS.Other dynamic or thermo dynamic triggering mechanisms, such as surface fluxes need to be further studied.In addition, there were frontal passages along the coast of southern China and over the northern SCS before and during the monsoon onset.What are the differences in the role, circulation characteristics, and structure of the fronts occurring before and during the monsoon onset?It seems worthy of further investigation.
continent with cold front extending westward to the coast of southern China and the northern SCS.Convective clouds developed continually in the warm sector of frontal zone along 20°N latitude, while it was still clear in the central-southern SCS south of 15°N latitude.The second sub-period (20-22 May) represents a short span of transition from local convective activities in the northern SCS to a full development of convection in the central-southern SCS.The convective clouds ceased gradually in the northern SCS and began to develop in the central southern SCS.In the meantime, a new cloud cluster associated with a surface low pressure center began to grow and move southward.From 23 May to 25 May is the third sub-period that denotes the full onset over the entire SCS.Comparably larger MCSs were observed in the central-southern SCS with fragmentary clouds evolving in the northern SCS.Until the 251h of May, the SCS was covered with convective clouds.Strong convections were found in a zone along 10°N latitude.

Figure 1
Figure 1 shows JMA (Japan Meteorological Agency) OOOOUTC surface weather maps of selected dates during the summer monsoon onset.The first cold front which occurred in the first sub-period arrived at the coast of southern China at OOOOUTC on 15 May (Fig. la).This front with a weakening cold core high pressure center H 1 behind it moved eastward to the ocean on 16 May (not shown).Another high pressure center H 2 moved southward immedi ately from north to around 30°N and then moved northeastward again.At OOOOUTC on 18 May, the associated cold front had moved to the ocean east of Taiwan and became stationary with frontal surface passing through Taiwan to the coast of southern China (Fig. lb).This is the second cold front appearing in the first sub-period.Over the SCS, no significant weather change was observed during this sub-period.The major feature of the second sub-period is the development of a surface low pressure system (denoted by Lon the maps).At OOOOUTC on 20 May, in the northern high-latitude area, a low pressure center was formed in the warm sector to the south of an eastward moving cold front (Fig. le).This low pressure system, accompanied with cloud clusters at its south eastern flank, moved southeastward progressively and reached the Szuchuan Basin at OOOOUTC on 21 May (Fig. ld).Meanwhile, high pressure H 2 moved to the ocean northeast of Korea and finally weakened at 1800UTC 21 May (not shown).During the evolutions of Land H 2 , the southerly wind component increased appreciably in the central and southern China areas due to the increase of an east-west pressure gradient.In a subsequent time, at OOOOUTC on 22 May, while L moved southward out of the Szuchuan Basin, another high pressure system formed north of 40°N and moved rapidly eastward.To the northeast of L, a new front formed (Fig. le).The synoptic setting over the northern SCS had apparent changes during this sub period.An east-west oriented low pressure zone was found along 20°N in the northern SCS region and the coastal area of southern China.The relatively weak low pressure center embed ded in the low pressure zone near the Bashi Channel (Fig. ld) moved eastward and then north eastward following the frontal path (Fig. le).These small and weak low pressure centers

Fig. 1 .
Fig. 1.JMA OOOOUTC surface weather maps of selected dates during 15-25 May.Date is indicated in the upper-left corner of each panel.

Fig. 2 .
Fig. 2. GMS enhanced satellite IR images of selected times during 16-25 May.Date and time are indicated in the upper-left corner of each panel.

Fig. 3 .
Fig. 3. Daily surface observations of the Dongsha (a, b) and the Nansha (c, d) islands for time period 5 May-15 June.In the left panels are 24-hour accumulated rainfall (vertical bar with units in mm) and daily averaged relative humidity (thick line with units in % ), while in the right panels are daily averaged wind direction (black dot) and wind speed (grey line with units in m s• 1 ).

Fig. 5 .
Fig. 5. Daily averaged 850 hPa observed height (solid line) and wind (vector) analyses for time period 18-25 May.Date is indicated at the top of each panel.The contour interval of height field is 10 m with values above 1520 m shaded.

130E
hPa wind analysis (left panels) and the 925 hPa divergence with 200 hPa wind analysis (right panels) of the 12-hour periods, 1200UTC 17 May-OOOOUTC 18 May and OOOOUTC 18 May-1200UTC 18 May, for the coarse domain CNTL runs.The accumulated precipitation analysis reveals an intense cloud band extending northeastward from the coast of southern China.To the south of the cloud band, there is an organized north-south oriented convective cloud band.The former is the frontal cloud band and the latter is a band-shaped organized MCSs developing in the warm sector of the front.Compared with Figs.2d, e, the model seems to be able to simulate the major convective activity in the first sub-period.For the period 1200UTC 17 May-OOOOUTC 18 May, over the northern SCS, the 850 hPa southwesterly flow near the southern end of the organized MCS mainly came from the outflow of the subtropical high (Fig.7a).But, during the period OOOOUTC 18 May-1200UTC 18 May, the frontal cloud band somewhat weakened and the MCS formation in the warm sector became active, the origin of southwesterly flow over the northwestern SCS gradually switched to the western side (Fig.7c).Also shown in Figs.7b, d, the 925 hPa convergent zones match well with MCS (accumulated precipitation) locations, and an upper-level diffluent flow pattern can be seen over southern China and the northern SCS in the 200 hPa wind analyses.

Fig. 7 .
Fig. 7. Coarse domain CNTL simulations.a) accumulated 12-hour precipitation (shaded) and 850 hPa wind vector and isotach (grey thick line) at the last hour of time period 5/17/12Z-5/18/00Z, b) accumulated 12-hour pre cipitation (thin line), mean 925 hPa convergence (shaded), and 200 hPa wind vector at the last hour of the same period as in a), c) same as in a) except for time period 5/18/00Z-5/18/12Z, d) same as in c) except for time period 5/18/00Z-5/18/12Z.Contour interval of isotachs is 5 m s• 1 , values less than 10 m s• 1 are not shown.Precipitation is shaded at inter vals 1, 5, 10, 15 , 20, and 25 cm as indicated by the bar chart at the right side of the panel.Convergence is shaded at an interval of 10x10-6 s• 1 as indicated by the bar chart at the right side of the panel.

Fig. 8 .
Fig. 8. Fine domain CNTL simulations.a) accumulated 12-hour precipitation (shaded) and 1000 hPa wind vector at the last hour of time period 5117 I 12Z-5 /18/00Z, b) sea level pressure at the last hour of the same period as in a), c) same as in a) except for time period 5/18/00Z-5 /18/12Z, d) same as in c) except for time period 5/18 /00Z-5 /18/12Z.Precipitation is shaded at intervals 1, 5, 10, 15, 20, and 25 cm as indicated by the bar chart at the right side of the panel.Sea level pressure is contoured at an interval of 1 hPa.

4. 2 Fig. 9 .
Fig. 9. Time-longitude cross sections along latitudes 20°N (a, c) and 15°N (b, d) of coarse domain CNTL (a, b) and NLHR (c, d) simulations for time period OOOOUTC 17 May-0600UTC 19 May.In each panel, the upper diagram shows sea level pressure (shaded) and 200 hPa wind vectors, the lower diagram shows 850 hPa equivalent potential temperature (shaded) and wind vectors.Sea level pressure and equivalent potential temperature are shaded at intervals 2 hPa and 4 K, respectively, as indi cated by the bar charts at the right side of the diagrams.The white thick lines represent 1008 hPa and 352 K, respectively.The thick black lines shown in the CNTL simulations (a, b) are contour lines of hourly accu mulated precipitation with an interval of l cm.
Fig. 1 la, MCSs developed over the northern Philippines and to the east of Hainan island during 0600UTC 21 May-1800UTC 21 May.This corresponds well with the MCSs shown by satellite image in Fig. 2i.In the following 12 hours (Fig. 1 lc), these MCSs gradually organized into a northwest-southeast oriented cloud band, which is indicated by MCS H in Fig. 2j.As compared with Figs.lOa, c, during the time periods shown in Figs. 1 la, c, the low level flows over the northern SCS were nearly from the west, while over the southern SCS flows were mainly from the south and the west.Figure 1 la depicts a cyclonic circulation accompanied with the MCS near the northern Philippines and weak northeasterly winds be tween the northern Bashi Channel and the Taiwan Strait.Twelve hours later (Fig. l lc), with the MCS development, a cyclonic circulation of considerable horizontal extent was observed in the northern SCS region, which reveals a close link between the MCS development and the low pressure center shown in Fig. ld.In response to the MCS development, the southward diffluent wind component became more evident in the upper troposphere (Figs. 1
Zhang and Bao (l 996a, b ).Another numerical study of MCS development on a Mei-Yu front by Chen et al. (l 998a) also demonstrated the convective enhancement of low-level jet in the southern periphery of a MCS.