The Space Conceptual Models andWater Vapor Characteristics of Typical Rainstorms during Plum Rain Season

Based on conventional observation data from the China Meteorological Administration (CMA) and reanalysis data from the American National Centers for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) between 2012 and 2021, combined with the meteorological analysis, composite synthesis, and water vapor trajectory analysis, the weather circulations of typical rainstorms during the 10 years can be divided into 4 categories: Static Front Pattern (SFP), Subtropical High Edge Pattern (SHEP), Northeast Cold Vortex Pattern (NCVP), and Low-Level Vortex and Shear Pattern (LLVSP).­e SHEP and SFP rainstorms have the characteristics of long duration and wide range, while the NCVP rainstorms are characterized by mobility and disaster weather accompaniment.­e daily precipitation of LLVSP cases has extremity feature. ­e occurrence and development of rainstorms are well coordinated with the systems on lower levels.­emain water vapor channel in lower layers of the SFP cases is from the South China Sea, while it is from Bohai for the NCVP cases and the Bay of Bengal for the SHEP and LLVSP cases. ­e main water vapor channel in middle layers is from the Bay of Bengal because of the aection of the southwest air ow. ­e south boundary of the MLYRB is the most important water vapor input boundary, followed by the west boundary, while the East and North boundaries are the outow boundaries. During the rainstorms, the low-level water vapor is exuberant with low-level water vapor convergence much stronger than the high-level divergence. Among the four types of rainstorms, the NCVP cases provide the most abundant low-level water vapor convergence, resulting in the strongest short-term precipitation among the four types. Combined with water vapor transportation and convergence, the rened spatial conceptual models of the four types of rainstorms can better judge the process intensity and falling area and provide reference for disastrous weather forecast and early warning.


Introduction
e Meiyu, which refers to the unique rainy season in the East Asian summer monsoon (EASM) region in late spring and early summer, is the second stationary period of the EASM [1][2][3][4]. e amount of precipitation and the location of the rain belt it produces are highly correlated with the ood or drought related conditions over the MLYRB [3,[5][6][7]. erefore, it is of great signi cance for social and economic activities to study the causes of Meiyu [8][9][10][11].
e Meiyu processes in the MLYRB generally have typical weather situations [35,36]. rough the classi cation of the in uencing systems of rainstorm cases, it is helpful to understand the synoptic scale mechanisms that lead to rainstorms, so as to improve the prediction accuracy of such events. Tao [37] summarized various circulation patterns of persistent heavy rain in China and pointed out that the stability of weather circulation, the transmission and convergence of water vapor, and the release and regeneration of convective unstable energy were the three basic conditions for the occurrence of persistent heavy rain in China. Maddox [38] investigated the temporal and spatial characteristics of 151 flash floods in the United States and divided these cases into four weather types according to the ground and upper air situation. Chen et al. [39] studied 117 heavy rainfall cases in Hunan from 2006 to 2014, which were divided into six types according to the experience and methods of rainstorm prediction in Hunan Province. According to the different triggering conditions of rainstorms, Deng et al. [40] established five conceptual models, including cold front type, static front type, and mesoscale trapped type. However, there are no classification studies on the impact systems of the MLYRB rainstorms during plum rain season yet, so this work has research significance.
Due to the great damage of rainstorms in the MLYRB in plum rain season, the analysis of water vapor source and transportation characteristics is also an important research aspect [41][42][43][44]. ere is a direct relationship between water vapor convergence and precipitation in rainstorm area [41]. e lifting and condensation of wet air can make the low level jet (LLJ) rise to a higher altitude [37], and water vapor condensation can enhance the vertical movement in the frontal area significantly [45]. Numerical experiments show that if the high humidity center in the Indian Ocean and the Bay of Bengal weakens in the initial field, the precipitation in the MLYRB will decrease [42,43]. Precipitation is sensitive to the initial value of water vapor on 850 hPa [46], and the release of latent heat during water vapor condensation is helpful to the triggering of new strong convection activities and increases precipitation [47]. While the water vapor in the upper troposphere has little influence on precipitation intensity and rain belt distribution, the water vapor in the middle layers plays an important role in maintaining the overall rain belt shapes, and the water vapor intensity change in the lower layers mainly affects the precipitation intensity in the center of the heavy rain. e closer the water vapor transmission position is to the shear line, the more conducive it is to the occurrence and development of the heavy rainfall [44].
Many research works on water vapor sources show that the main water vapor sources of summer rainstorms in the East China are the Bay of Bengal, the South China Sea, and the warm pool area of the Western Pacific [3,[48][49][50][51]. e studies above on water vapor sources and transportation are based on the Euler method, which cannot quantitatively distinguish the contribution of each water vapor source on precipitation. ere are also some differences between the Euler flow field and the real trajectory of the air mass [52]. In consideration of the fact that the Lagrangian method can provide details and changes of water vapor along the water vapor transport trajectory, some scholars have applied this method to study the water vapor source and budget of rainstorms in recent years [53][54][55][56][57][58][59][60][61][62]. By tracking the threedimensional position of every air block backward, this method determines the three-dimensional position and the physical quantities such as temperature, pressure, wind, and humidity of every air block at each time and defines the water vapor source contribution in the process of water vapor transportation. Currently, this method is widely used on a global scale [63] and in various subregions (the United States [64,65], Europe [66][67][68], Latin America [69], and Africa [70]).
Due to the complexity and regional differences of the impact systems in plum rain season in the MLYRB, a systematic summary of the influencing systems and the characteristics of water vapor transmission and budget in this region can not only provide a scientific basis for accurately predicting of such events but also be very necessary for disastrous weather forecast and early warning. erefore, based on the analysis of the characteristics of the rainstorm cases in the MLYRB from 2012 to 2021, this paper studies the main impact systems, summarizes the weather circulation types, and then gives water vapor transportation and budget characteristics of different types of rainstorm cases through classification, synthesis, and water vapor tracking methods. It can help us achieve a comprehensive understanding of the weather situations and water vapor characteristics during rainstorms in plum rain season in the MLYRB and deepen our understanding on the changes and differences of precipitation characteristics under different types of weather situations. It is hoped that the research results can be used to further improve the rainstorm prediction technology and provide useful reference for disaster prevention and reduction.

Materials.
e analysis data used for classifying and summarizing the space conceptual models of rainstorms in the MLYRB in this paper include conventional ground data, radiosonde data, and FNL (Final) Operational Global Analysis reanalysis data (1°× 1°) from the American National Centers for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) from 2012 to 2021. e data used in Hybrid Single Particle Lagrangian Integrated Trajectory model (HYSPLIT 4.0) is the reanalysis data from NCEP/NCAR with temporal resolution of 6 hours and horizontal resolution of 2.5°× 2.5°, which has been packaged into an available data format (ARL) for HYSPLIT 4.0 by ARL (Air Resources Laboratory) of the National Oceanic and Atmospheric Administration (NOAA).

Study Period.
e Meiyu in China is a unique rainy season during the northward advancement of the EASM in late spring and early summer. e average Meiyu period is from mid-June to mid-July, which is also the period of the strongest precipitation. e precipitation in the Meiyu period has great interannual variation. It can be continuous from several days to several weeks or intermittent, which accounts for almost half of the precipitation in the whole summer [17]. We select June to July as the main analysis period in this paper.

Study Region.
e study region is located in the MLYRB (27°-34°N, 109°-123°E) (red rectangle area in Figure 1). is region is a key industrial and agricultural production base in China with continuous economic development and a high level of science, technology, and culture which encloses approximately 700,000 km 2 . With high temperature and rainy season in summer and mild and little rain in winter, the climate in the MLYRB is conducive to agricultural production. At the same time, this area is one of the waterlogged areas in China. During spring and summer, especially the plum rain season from June to July, the rainfall is concentrated and heavy, which is a direct cause of water logging. With a hilly topography, sudden rainstorms produce a large amount of runoff which then pour into lakes and rivers. If there are floods upstream at the same time, this is more likely to cause flood in rivers and lakes, destroy embankments, inundate houses and farmland, and even threaten the safety of people's lives and property.

Classification and Synthesis Methods of Rainstorm
Processes. e rainstorm cases in the MLYRB during plum rain season are analyzed from the perspective of weather types by the method of comprehensive analysis, and then the processes of the same type are synthesized. e synthetic field is the nearest reanalysis data at 00:00 UTC or 12:00 UTC before the rainstorm occurs: if the rainstorm occurs in the daytime, the reanalysis data at 00:00 UTC of the same day will be used; otherwise, the reanalysis data at 12:00 UTC of the same day will be used.

Introduction of Hybrid Single Particle Lagrangian
Integrated Trajectory Model (HYSPLIT v4.9). Developed by the Air Resources Laboratory of NOAA, HYSPLIT v4.9 is used in this paper to provide details and changes of water vapor along the water vapor transport trajectory [71]. e initial position of tracking is selected in the MLYRB (red rectangle area in Figure 1) and the initial tracking heights chosen are 100 m, 500 m, 1500 m, 3000 m, and 5000 m, among which the levels below 1500 m are taken as the lower troposphere and the levels between 3000 m and 5000 m are taken as the middle troposphere. e model can give hourly trajectory points and 6 hourly physical fields (temperature, height, specific humidity, pseudoequivalent potential temperature, and water vapor flux) backward to 10 days by retracking every 6 hours. Although the initial heights are fixed, the height of each trajectory varies with space and time when simulated in Lagrangian space. Due to the large number of simulated trajectories, in order to see the distribution of trajectories more directly, all trajectories are clustered by merging the trajectories close to each other [71].
Based on this, we introduce the improved regional source-sink attribution method to determine the relative contribution of different sources at different levels. By calculating the change of specific humidity in the atmosphere, the contribution of each source to the precipitation in the target area can be determined quantitatively. We introduce the following formula to calculate the contribution of every channel: (1) In the above formula, Qs represents contribution of the specific humidity or water vapor flux, qlast represents specific humidity or water vapor at the end of the channel, m represents the number of tracks included in the channel, and n represents the total number of tracks. We can fully understand the vertical structure characteristics of water vapor contribution through this method [72][73][74]. Advances in Meteorology

Calculation Method of Water Vapor Budget.
Consider that water vapor source is due to external evaporation and water vapor sink is due to internal condensation [75].
e above formula is the water vapor continuity equation: q is the specific humidity, V is the horizontal wind, ω is the vertical velocity, the first term on the left of the equation is the local variation of water vapor, the second term is the divergence of water vapor flux, the third term is the vertical transportation of water vapor, and M and E on the right of the equation are the condensation rate and evaporation rate, respectively. By integrating the above formula from the bottom to the top of the atmosphere and making average in the selected region, the average water vapor budget of the region can be diagnosed. If the influence of terrain and ground friction are not considered and the vertical velocity on the ground and the top of the atmosphere is considered to be 0, the third term on the left of the equation can be omitted. In the rainstorm area, the local variation of water vapor is much smaller than precipitation, so the first term on the left can also be omitted. It can be seen that the horizontal convergence of water vapor in the whole layer is approximately equal to the budget of water vapor in the region [76].

Weather Types and Conceptual Models of Rainstorm
Processes in the MLYRB. Using synoptic diagnostic methods, 64 cases of rainstorm processes in the MLYRB during plum rain season from 2012 to 2021 are classified into 4 categories: Static Front Pattern (14 cases, SFP), Subtropical High Edge Pattern (8 cases, SHEP), Northeast Cold Vortex Pattern (15 cases, NCVP), and Low-Level Vortex and Shear Pattern (27 cases, LLVSP). Synthetic analysis is carried out in order to further explore the main weather systems, as well as dynamic, thermal, and water vapor characteristics of each type of rainstorm cases in plum rain season.

Static Front Pattern Rainstorms.
ere occurred altogether 14 SFP rainstorm cases during the 10 years. By compositing and analyzing the synoptic situations in the SFP rainstorms, it is found that the main impact systems are stably maintained Western Pacific Subtropical High (WPSH), mid-level trough, low-level shear lines, and ground stationary front. As shown in Figure 2, the circulation characteristics of SFP rainstorms are as follows.
ere is a deep mid-level trough in the mid-high latitudes with strong positive vorticity advection in front of it, in spite of the small meridionality. Large-scale dynamic forcing is conducive to maintaining strong upward movement in a large area. WPSH is strong enough to control Southern China, resulting in blocking the movement of the mid-low latitude systems, which is conducive to the maintenance of heavy rainfall in the MLYRB. ere are cold shear lines in lower levels which are accompanied by LLJ with moderate intensity. e heavy rainfall area of these processes is located between the cold shear line on 700 hPa and the static front, with a wide range. e distribution of the heavy rain area is similar to that of the shear lines in a northeast to southwest belt shape. e highest precipitation intensity is mainly around 50 mm/24 h. When the shear lines and strong LLJ occur at the same time, precipitation above 100 mm/24 h may also occur. Because the rain belt can last for several days or even more in the same area, rainstorms of this kind have great influences.

Subtropical High Edge Pattern Rainstorms.
ere occurred altogether 8 SHEP rainstorm cases during the 10 years.
e main impact systems are stably maintained WPSH, mid-level short wave troughs, low-level shear lines, and LLJ. As shown in Figure 3, the circulation characteristics of SHEP rainstorms are as follows.
e WPSH is stably sustained in the Southeast area of China with its ridge line located near 20°N, north boundary at 25°E, and west ridge point at 100°E. Warm and humid air is transported to the MLYRB along the southwest air flow on the northwest side of the WPSH. In mid-high latitudes, there are short wave troughs developing and moving eastward continuously. erefore, the northerly air flow behind the troughs and the southwest air flow on the north side of the WPSH converge in the MLYRB. In low levels, there occurs warm shear lines between southwest airflow and southeast airflow. Under the combined action of divergent suction of upper-level distributary area and convergence in the low levels, heavy rain forms in the MLYRB. e heavy rain locates in the south of the warm shear on 850 hPa with a quasi-east-west shape.

Northeast Cold Vortex Pattern Rainstorms.
ere occurred altogether 15 NCVP rainstorm cases during the 10 years.
e main impact systems are stably maintained Northeast Cold Vortex, mid-level trough, low-level shear lines, LLJ, SLLJ, and ground cold front. As shown in Figure 4, the circulation characteristics of NCVP rainstorms are as follows.
e Northeast Cold Vortex is established and remains in the Northeastern China with its central intensity reaching 5360 gpm. While the intensity of the Northeast Cold Vortex continues to strengthen because of continuous supplement of cold air from the west side of Baikal Lake, the ridge in the north of Okhotsk Sea also develops, which prevents the Northeast Cold Vortex from moving eastward. At the same time, the WPSH continues to extend westward and northward to South China, which brings rich water vapor and unstable energy to the MLYRB along the LLJ and SLLJ in the northeast of WPSH. e strong convergence of warm air from the edge of the WPSH and cold air from the Northeast Cold Vortex occurs in the MLYRB, resulting in heavy rain processes with the characteristics of strong intensity and coexistence of multiple disasters. During this type of rainstorms, the high-and low-level jets remain on 200 hPa, 850 hPa, 700 hPa, and 925 hPa, the Northeast Cold Vortex is strong and stable, the WPSH is in a northeast to southwest belt, and the trough is in a large meridional direction. Such kind of precipitation has mobility characteristics, which is well coordinated with the shape of the Northeast Cold Vortex and low-level shear lines.

Low-Level Vortex and Shear Pattern Rainstorms.
ere occurred altogether 27 LLVSP rainstorm cases during the 10 years. e main impact systems are mid-level trough, low-level vortexes, shear lines, LLJ, SLLJ, and ground   Advances in Meteorology inverted trough. As shown in Figure 5, the circulation characteristics of LLVSP rainstorms are as follows. ere is a mid-level trough occurring in mid-high latitudes with large meridionality. Under the action of positive vorticity advection in front of the trough, a low-level vortex generates and moves eastward with strong LLJ and SLLJ in front of it. e WPSH in this type is relatively weak and located on sea. Located between the trough and the WPSH, the MLYRB is under the influence of strong warm and humid southwest airflow and warm inverted trough extending from the southwest to the northeast, providing favorable thermal and dynamic environment conditions for the generation of small-and medium-scale disturbances. Heavy rainfall occurs in the southeast of the vortex, the south of the warm shear lines, and the east of the cold shear lines. Frequently, it also occurs in the outlet of the LLJ and SLLJ in the warm air region [77,78]. e distribution of rain belt is consistent with the shape of the vortex with the maximum intensity exceeding 100 mm/24 h, which is prone to extremely heavy rain magnitude precipitation (250 mm/24 h). erefore, when faced with the LLVSP rainstorms, forecasters should pay attention to the configuration of the shear lines and the LLJ in addition to the generation and disappearance of the low vortex and its moving path.

Water Vapor Trajectory Analysis.
Since water vapor locates mainly below 500 hPa, the temporal and spatial evolution of water vapor in lower and middle levels are analyzed. e water vapor transport characteristics in each type of rainstorms are analyzed using clustered backward trajectories.
ere are four main water vapor transport channels in the lower layers of the STP rainstorms according to Figure 6(a). e most important channel is from the South China Sea (accounts for 51.47% of the total number of lowlevel trajectories, and the contribution rates of specific humidity and water vapor flux are about 53%). e second one is from the Bay of Bengal (accounts for 23.13% and the contribution rates of specific humidity and water vapor flux are about 24% (Table 1)) and it enters into the MLYRB through the Indochina Peninsula. e third one is from the Yellow Sea, followed by the high latitude inland channel, which indicates the cold air from West Siberia. ere are also four main water vapor transport channels in the lower layers of the SHEP rainstorms ( Figure 6(c)). e main channel is from the Bay of Bengal (accounts for 40.73% and the contribution of specific humidity and water vapor flux is about 38% (Table 1)), followed by the channel from the Bay of Bengal-South China Sea, the Yungui Plateau, and the west Pacific Ocean. e channel from the Yungui Plateau is the shortest, which indicates the slow motion of water vapor in this channel. In the lower layers of the NCVP rainstorms ( Figure 6(e)), the main channel is from Bohai (accounts for 38.17% and the contribution of specific humidity and water vapor flux is about 34% (Table 1)) and it enters into the MLYRB through the Shandong Peninsula. It is followed by the Bay of Bengal channel, the Bay of Bengal-South China Sea channel, and the South China Sea channel. In the lower layers of the LLVSP rainstorms ( Figure 6(g)), the main channel is from the Bay of Bengal (accounts for 28.9% and the contribution of specific humidity and water vapor flux is about 28% (  It can be seen from the height evolution of the lowerlevel water vapor channels that, in the SFP rainstorms ( Figure 6(b)), channels 1 and 2 coming from the ocean surface in low latitude with height between 940 hPa and 900 hPa are slightly elevated due to the topographic force of Nanling Mountain after landing. Channel 6 is the shortest, indicating that the air mass in this channel moves slowly and is slightly elevated through the hilly area of Shandong Province. All the 3 channels are transported to the height around 900 hPa in MLYRB. Channel 7 coming from the high latitude inland with the initial height around 650 hPa sinks into the MLYRB in form of cold air, which strengthens the atmospheric baroclinity. e long track of channel 7 indicates the fast movement of dry and cold air.
In the SHEP rainstorms ( Figure 6(d)), channels 1, 3, and 4 all come from ocean surface. Due to the obvious elevation of Nanling Mountain and Luoxiao Mountain, the water vapor of channel 1 is finally transported to the MLYRB around 850 hPa, while water vapor of channel 3 and that of channel 4 are also elevated slightly by the Nanling Mountains and finally transported to the MLYRB near 920 hPa and 940 hPa, respectively. e water vapor of channel 8 moves slowly across the Qinling Mountain, so it rises earlier and sinks later to around 880 hPa. Channels 1, 2, and 4 of the NCVP rainstorms ( Figure 6(f )) coming from ocean surface in low latitude move quickly through the Yungui Plateau and the Nanling Mountain with height elevated obviously and finally are transported to the vicinity of 900 hPa (for channels 2 and 4) and 860 hPa (for channel 1) in the MLYRB. e water vapor of channel 5 from Bohai is also elevated to a higher altitude during passing through the hilly area of Shandong and transported to the MLYRB near 900 hPa. e water vapor channels 1, 2, and 4 of the LLVSP rainstorms ( Figure 6(h)) which are similar to those of the NCVP are finally transported to the height between 920 hPa and 850 hPa in the MLYRB. Channel 6 coming from the Yellow Sea is lifted obviously to around 900 hPa after passing through the Yangtze River Delta and the Dabie Mountain.
According to the previous analysis, it can be seen that the water vapor channels in lower levels and lower latitude regions account for the majority, while the air mass from the north in higher latitude has little influence. So we further explore the distribution of water vapor channels in middle layers.
ere are four main water vapor transport channels in the middle layers of the STP rainstorms according to Figure 7(a). e most important channel is from the Bay of Bengal, which accounts for 43.65% of the total number of the middle-level trajectories, and the contribution rates of specific humidity and water vapor flux are about 50% ( Table 2). It is followed by the South China Sea channel, the high latitude inland channel, and the Hetao channel. e high latitude inland channel and the Hetao channel indicate the cold air from the north. In the middle layers of the SHEP rainstorms (Figure 7(c)), the main channel remains that from the Bay of Bengal with trajectories accounting for 62.95% of the total number of the middle-level trajectories and the contribution of specific humidity and water vapor flux being about 70% (Table 2), which indicates the transportation of large amount of water vapor through the southwest air flow on the edge of the WPSH. It is followed by the Indo-China Peninsula channel, the Bay of Bengal-South China Sea channel, and the high latitude inland channel. In the middle layers of the NCVP rainstorms (Figure 7(e) Table 2)). It is followed by the Southwest China channel with a slow moving speed, the Somali jet channel, and the high latitude inland channel. In the middle layers of the LLVSP rainstorms (Figure 7(g)), the main channel remains that from the Bay of Bengal (accounts for 49.62% and the contribution of specific humidity and water vapor flux is about 52% ( Table 2)). It is followed by the Somali jet channel, the high latitude inland channel, and the Yangtze River Delta channel. It can be seen that the main water vapor channel in middle layers is from the Bay of Bengal with the contribution rates of specific humidity and water vapor flux more than half (Table 2), which indicates the continuous transportation of water vapor from the Bay of Bengal to the MLYRB through the southwest jet stream and sufficient water vapor conditions for the occurrence of heavy rain in the MLYRB.
It can be seen from the height evolution of middle-level water vapor channels that, in the SFP rainstorms (Figure 7(b)), channels 1 and 2 coming from ocean surface in low latitude with track height around 950 hPa are obviously elevated due to the quick movement through topographic area. Channels 6 and 7 coming from the high latitude inland with a high initial height sink to the vicinity of 700 hPa in the MLYRB finally.
In the SHEP rainstorms (Figure 7(d)), channels 1, 3, and 5 coming from ocean surface with initial height below 800 hPa are finally transported to the MLYRB around 700 hPa, while water vapor of channel 7 from the high latitude inland with initial height around 550 hPa is finally transported to the MLYRB in the form of cold air. e water vapor in channels 1 and 4 of the NCVP rainstorms (Figure 7(f )) coming from ocean surface in the low latitude moves quickly through the mountain areas with height elevated obviously and is transported to the vicinity of 700 hPa in the MLYRB. e water vapor of channel 7 from the high latitude inland with initial height around 600 hPa is finally transported to the MLYRB in the form of cold air. e water vapor of channel 8 from Southwest China is slightly elevated to a higher altitude due to the slow motion when passing through hilly areas. Both channels 7 and 8 are transported to the vicinity of 700 hPa in the MLYRB. e water vapor in channels 1 and 4 of the LLVSP rainstorms (Figure 7(h)) which is similar to that of the NCVP rainstorms is finally transported to the vicinity of 700 hPa in the MLYRB. Channel 7 coming from the high  latitude inland with initial height near 600 hPa sinks and invades the MLYRB in the form of cold air. Channel 9 coming from the Yangtze River Delta is lifted slowly to around 700 hPa in the MLYRB under the influence of Dabie Mountain.

Water Vapor Budget
Analysis. In order to analyze the water vapor transportation and budget during the four different types of rainstorms, the boundary water vapor flux and regional water vapor revenue and expenditure in the MLYRB during the rainstorm period are calculated as follows.
It can be seen from the vertical integrated water vapor flux at the boundaries of MLYRB that there are same characteristics in the four types of rainstorm cases ( Figure 8): Firstly, the south boundary of the MLYRB is the most important water vapor input boundary, followed by the west boundary, while the east and north boundaries are the outflow boundaries. From the above trajectory analysis in Section 3.2, it can be seen that water vapor mainly flows into the rainstorm area through Southwest and Southern China. Secondly, there is obvious positive water vapor income in the MLYRB, which means that the MLYRB is the water vapor convergence area. We can see from the water vapor flux at the south and west boundaries that the water vapor input at different boundaries is determined by the difference of the moisture channels and the strength of LLJ.
In the NCVP (Figure 8(c)) and LLVSP (Figure 8(d)) rainstorm cases, the position of the WPSH is eastward, and the positive vorticity advection in front of the mid-level troughs with large meridionality is strong. e strong potential gradient between the mid-level troughs and the WPSH strengthens the LLJ and SLLJ, too. erefore, the water vapor flux at the south boundary of the above two types (especially in the LLVSP) of rainstorms is the strongest. At the north boundary, there are relatively strong outputs in the SHEP (Figure 8(b)) and the LLVSP ( Figure 8(d)) rainstorm cases, while there is a weak output in the SFP (Figure 8(a)) rainstorms and weak input in the NCVP rainstorms. Because of the obvious cold air activity in the NCVP and the SFP rainstorm cases, there is obvious northerly wind flow in either lower or middle layers, which makes the integral water vapor transport positive in the north boundary. From the perspective of total regional water vapor income, the largest net water vapor income occurs in the NCVP and SFP rainstorm cases, resulting in short-duration heavy rain and strong convective weather with the accompaniment of the strong convergence of cold and warm air. However, because of the obvious mobility character in the NCVP rainstorms, the 24-hour accumulated precipitation is not the largest.
It can be seen from the vertical distribution of the averaged water vapor flux of the four types of rainstorms ( Figure 9) before rainstorms occur that there is obvious outflow at the east boundary except in the low levels of the LLVSP rainstorms (strong inflow), while there is inflow in all levels above 850 hPa at the west boundary in all the four types. e water vapor transportation centers at the east and south boundaries are below 800 hPa, which are mainly coordinated with LLJ and SLLJ. e water vapor transportation at the south boundary is generally the same as the center located around 850 hPa, while there are obvious differences at the north boundary. At the north boundary, there is obvious outflow on lower layers and weak inflow on middle layers of the SHEP and LLVSP rainstorms, while there is obvious inflow on both middle and lower layers of the NCVP rainstorms.
ere are weak transportation centers \in the middle troposphere (around 600-500 hPa) at the north boundary, indicating that the four types of rainstorm cases are accompanied by the invasion of weak cold air on middle layers, especially in the NCVP rainstorms. From the vertical distribution of the water vapor budget, we can see that the SHEP and the NCVP rainstorm cases have the strongest water vapor flux convergence of the low levels and are more likely to produce strong short-term rainstorms.

Discussion.
e study of weather system classification is helpful for forecasters to identify the weather situations and their configuration relationship with heavy rain in the MLYRB and judge the possibility of heavy rain, which is essential for the prediction of heavy rain in plum rain season. However, the classification is based on the synoptic scale systems, while the direct producers of rainstorms are usually the small-and medium-scale systems [34,79,80]. ere exists obvious convective character with 6-hour precipitation more than 160 mm in the NCVP and SHEP rainstorms, while it is a little weaker in the other two types (figure omitted). We can also see Mesoscale Convective Complex (MCC) and Mesoscale Convective Systems (MCSs) [16,28,81,82] which are the main cause of convective precipitation embedded in the large-scale precipitation cloud systems (Figure omitted)  the WPSH and the South Asia High makes the rain belt remain in the MLYRB. e synoptic systems such as shear lines, LLJ, SLLJ, low vertexes, and fronts make the environment suitable for convective systems. Convergence of dry and cold airflow with warm and wet airflow from the ground to the middle troposphere and upward movement along the Meiyu front increase the instability and stimulate the release of large amount of unstable energy [29]. erefore, in the operational forecasting, forecasters should analyze the environmental conditions, structural characteristics, and evolution of the small-and medium-scale systems on the basis of the large scale weather classification, make accurate forecast of the possible falling area, intensity, and starting and ending time of rainstorms, and release the early warning signal in the nowcasting stage with the help of the numerical forecast and observation data.

Conclusions
Based on conventional observation data and reanalysis data from NCEP/NCAR between 2012 and 2021, combined with the meteorological analysis, composite synthesis, and water vapor trajectory analysis, the weather circulations of typical rainstorms during the 10 years can be divided into 4 categories: Static Front Pattern (14 cases, SFP), Subtropical High Edge Pattern (8 cases, SHEP), Northeast Cold Vortex Pattern (15 cases, NCVP), and Low-Level Vortex and Shear Pattern (27 cases, LLVSP). e SHEP and SFP rainstorm cases have the characteristics of long duration and wide range, while the NCVP rainstorms are characterized by mobility and disaster weather accompaniment. e LLVSP rainstorms are accompanied by strong low-level jet (LLJ) and super low-level jet (SLLJ), so the daily precipitation of this type has extremity feature. Most of the rainstorms are accompanied by highlevel jet and low-level jet, indicating the obvious coupling effect of high-level and low-level systems in the rainstorm cases during plum rain season. e occurrence and development of rainstorms are all well coordinated with the systems on lower levels.
By analyzing the characteristics of water vapor transport in lower layers (below 1500 m) and middle layers (3000 m-5000 m) during the four types of rainstorms, the vertical structure of water vapor transport is obtained. e main water vapor channel in the lower layers of the SFP is from the South China Sea, while it is from Bohai for the NCVP rainstorms and the Bay of Bengal for the SHEP and LLVSP rainstorms. e main water vapor channel in middle layers is from the Bay of Bengal with contribution rate accounting for 62.95% in the SHEP, followed by NCVP (56.01%), the LLVSP (49.26%), and the SFP (43.65%) because of the affection of the southwest air flow. e south boundary of the MLYRB is the most important water vapor input boundary, followed by the west boundary, while the east and north boundaries are the outflow boundaries. During the rainstorm cases, the low- level water vapor is exuberant, and the low-level water vapor convergence is stronger than the high-level divergence. Among the four types of rainstorms, the NCVP rainstorms provide the most abundant low-level water vapor convergence, resulting in the strongest short-term precipitation intensity among the four types. Combined with water vapor transportation and convergence, the refined spatial conceptual models of the four types of rainstorms can better judge the process intensity and falling area and provide reference for disastrous weather forecast and early warning.

Data Availability
e Grid data used in this study are obtained from https:// rda.ucar.edu/datasets/ds083.2/index.html and ftp://arlftp. arlhq.noaa.gov/pub/archives/reanalysis. e rainstorm cases used in this paper are from the rainstorm case datasets of the China Meteorological Administration.

Conflicts of Interest
e authors declare that they have no conflicts of interest.