Synoptic Characteristics Related to Warm-Sector Torrential Rainfall Events in South China During the Annually First Rainy Season

The precipitation during the first rainy season (April-June) accounts for approximately half of the annual precipitation in south China (Luo et al. ^[1]). During this period, there are usually two types of heavy rainfall events, i. e., squall lines (SLs) and warm-sector torrential rainfall (WSTR). The former occurs along with low pressure and front, while the latter develops in the warm sector at least 200 km ahead of cold fronts (Wang et al. ^[2]; Wu et al. ^[3]; Luo et al. ^[1]).

Squall line and WSTR are the two most common heavy rainfall events that occur in the annually first rainy season in south China every year. Squall line is typically related to strong-intensity convection, while WSTR is often related to medium-intensity convection. Previous studies found that SL convection can extend to several kilometers high above the freezing level, where ice particles and lightning are active (Wang et al. ^[4]; Yuan and Qie ^[5]; Qie et al. ^[6]; Qian et al. ^[7]; Wu et al. ^[8]; Pan et al. ^[9]; Meng et al. ^[10]). Their occurrence and development are closely associated with fronts and strong dynamic disturbances, so the related synoptic-scale environments are clear (Weisman and Klemp ^[11]; LeMone et al. ^[12]; Tetzlaff and Peters ^[13]; Sow et al. ^[14]; Wu et al. ^[15]; Meng et al. ^[10]). Compared with SL events, WSTR events are related to less intense convection but heavier rainfall intensity (Du and Chen ^[16]; Huang and Luo ^[17]; Wu et al. ^[18]). They initiate in the warm sectors far from cold fronts without prominent influences from large-scale forcing. The land-sea breeze, the warm shear line, the low-level jets, and the onset of summer monsoon are important factors providing favorable conditions for the initiation of WSTR convection (Du and Chen ^[16]; Huang et al. ^[19]; Lin et al. ^[20]; Chen et al. ^[21]; Chen et al. ^[22]; Chen et al. ^[23]; Chen et al. ^[24]). High Moisture content in the upstream oceanic airflow, especially that in the boundary layer, is found to be critical to WSTR rainfall intensity (Hamada et al. ^[25]; Chen et al. ^[24]; Huang et al. ^[26]).

WSTR events usually occur suddenly and can last for ten to twenty hours, during which they move less, keep quasi-stationary and produce heavy rainfall to the same place. This feature distinguishes WSTR from pre-frontal squall lines and warm season afternoon thunderstorm. First, SLs are also likely to occur in the pre-frontal warm sectors, but they initiate near the front and then quickly move toward warm sectors (Karan et al. ^[27]; Meng et al. ^[10]; Wu et al. ^[15]; Chen et al. ^[28]). The large discontinuities in meteorological conditions near the fronts are key conditions for the formation of pre-frontal squall lines, while WSTRs occur under weak synoptic-scale forcing environments at least 200 km ahead of the fronts. Second, warm season afternoon thunderstorms caused by diurnal heating also occur suddenly under weak synoptic-scale forcing. However, they often occur on small temporal scales which only last for a few hours (Weckwerth ^[29]; Lin et al. ^[30]) and they can produce lightning, destructive hail, and tornadoes.

Despite numerous detailed individual case studies, the general understanding of the synoptic characteristics of WSTR remains unclear. To derive general conclusions, multiple SL and WSTR events that occurred during 2013-2017 are selected for analysis in this study. Preliminary investigation on these events showed weak similarities in the horizontal distribution of WSTR events, and their differences from SL events were also unclear. Therefore, this study examines the vertical structures of the atmospheric conditions in WSTR events with those in SL events as references, aiming at a better understanding of WSTR events and thus providing basis for the improvement of the predictability of WSTR events.

Hourly precipitation data from approximately 8, 000 automatic weather stations and hourly three-dimensional radar reflectivity mosaics with a 3-km horizontal grid spacing provided by the Guangdong Meteorological Observatory were used in this study. From these data, 32 episodes of heavy rainfall were selected. They consist of 16 SLs and 16 WSTR events that occurred in the annually first rainy seasons of 2013-2017 in south China (Table 1). The locations of these events are shown in Fig. 1. The study areas vary with each rainfall event, covering 8 degrees apart from the rainfall centers in the north-south direction and east-west direction ([±8°N, ±8° E]).

Figure 1.  Terrain height (shaded, m) overlaid by the locations of the 32 heavy rainfall events selected for analysis. Red (black) dots indicate the rainfall centers of the 16 WSTR (SL) events. Red curves indicate trumpet-shaped topography to the north of the WSTR centers.

The synoptic characteristics associated with the WSTR and SL events were analyzed using a six-hourly ERA-Interim reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts. The dataset is presented at a 0.75° × 0.75° horizontal grid that has 37 vertical layers at standard pressure levels from 1, 000 to 1 hPa (Dee et al. ^[31]). Many previous studies on heavy precipitation events indicated that the synoptic conditions which we focus on are represented well enough by the ERA-Interim reanalysis dataset (Zhong et al. ^[32]; Bohlinger et al. ^[33]; Maranan et al. ^[34]; Yokoyama and Takayabu ^[35]; Raveh and Wernli ^[36]).

We defined rainfall centers (shown in Fig. 1) as grid cells where 1-h rainfall maxima occurred. They were also used as domain centers in composite analyses. Torrential rainfall was defined as a process that persisted for more than 6 h, during which the rainfall amount exceeded 20 mm in each hour, or the total rainfall amount throughout the event exceeded 100 mm. Convection initiation time was defined as the time when the composite radar echo first exceeded 35 dBZ, and dissipating time was defined as the time when the echo fell and remained continuously, below 35 dBZ (Roberts and Rutledge ^[37]).

The radar echo centroid was defined as the average height of reflectivity exceeding a certain threshold within the [±0.5°N, ±0.5°E] domain around each rainfall center. For each event, three centroids were calculated for the thresholds of 40, 45, and 50 dBZ at the time of the 1-h rainfall maxima.

Composite analyses were implemented as the spatial average covering the same distance apart from each rainfall center. Convective instability was defined as the vertical gradient of pseudo equivalent potential temperature with air pressure. The p-coordinate fluxes $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\rightharpoonup$}} \over F} $ is defined in Eq. (1) (Wang and Paegle ^[38]),

where M represents a meteorological field, e. g., the temperature or specific humidity, ${{{\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\rightharpoonup$}} \over V} }_h}}$ represents the horizontal wind vector, and g is the gravitational acceleration constant.

The reflectivities of the WSTR (Fig. 2a) and SL (Fig. 2b) events were similar in magnitude but different in vertical structure. The echo centroids of the SLs were much higher than those of the WSTR events. For example, for the 16 SL events examined, almost all the 50 dBZ centroids were above the 0℃ level, with some extending to an altitude of 12 km; the 45 dBZ centroids were between the 50 and 40 dBZ centroids, resulting in an upward distribution. In contrast, the centroids of the WSTR events formed a downward distribution, with almost all the 50 dBZ centroids below the 0℃ level and at lower altitudes than the 45 and 40 dBZ centroids.

Figure 2.  Radar echo centroids of (a) 16 WSTR events and (b) 16 SL events, in which the event number is on the x-axis and the altitude (km) is on the y-axis. Green, blue, and red segments represent reflectivities above the 40, 45, and 50 dBZ thresholds, respectively, and the centroids are indicated by dots. The solid line indicates the height of the 0℃ level. The dashed line indicates the height of the -10℃ level. The histograms indicate observed 24-h accumulated rainfall amount (mm) of each event. The horizontal distribution of the composite 1-h rainfall (mm) of (c) 16 WSTR events and (d) 16 SL events is illustrated by the shaded area.

Acquired at the same time as the radar echo centroids shown in Fig. 2a and Fig. 2b, the composite 1-h accumulated precipitation amounts of the WSTR events (Fig. 2c) were generally higher than those of the SLs (Fig. 2d). In almost all WSTR events, the 24-h rainfall amounts exceeded 150 mm, and some even exceeded 300 mm (Fig. 2a). The higher rainfall intensity but lower echo centroids in the WSTR events are considered to be associated with warm-rain processes (Hamada et al. ^[39]; Hamada et al. ^[25]; Luo et al. ^[1]). One possible reason is that raindrops mainly grow in warm layers below the 0 ℃ level. When they reach a size that is too large to be held up, they fall through the cloud and grow continuously until they reach cloud base, which results in a downward increase in radar reflectivity (Xu and Zipser ^[40]; Xu and Zipser ^[41]).

Vertical profiles of the composite convective instability (Fig. 3a) and the composite p-coordinate vertical velocity (Fig. 3b) were calculated over the [± 0.5°N, ±0.5°E] domain around each rainfall center. These profiles were defined as the regional averages of the ERA-Interim reanalysis dataset that was nearest in time to the observed time of convection initiation. The convective instabilities in the WSTR events were generally stronger (weaker) below 850 (between 850-600) hPa than those in the SLs. Similarly, the vertical velocities were stronger (weaker) below (above) 975 hPa in the WSTR events than those in the SLs. The stronger vertical velocity and convective instability near the surface benefit the triggering of WSTR convection in the lower layers, while the weaker convective instability above 850 hPa means that less convective energy is available for the extension of convection to higher altitudes, which favors the formation of low echo centroids, as shown in Fig. 2a.

Figure 3.  Vertical distribution of (a) the composite domain-averaged convective instabilities (K hPa^-1), and (b) the p-coordinate vertical velocities (hPa s^-1). The solid (dashed) line indicates WSTR (SL) events.

Following the definition in Eq. (1), fluxes of the temperature and moisture at the onset of the 1-h rainfall maxima were analyzed along the north-south section of the rainfall center. From Fig. 4a-b, and Fig. 4d-e, we can see that there are clear differences between the WSTR and SL events. In the SLs, the convergence of north and south winds near the rainfall centers tilted northward with height, and gradients of temperature and moisture in the meridional direction were evident. In contrast, the rainfall centers in the WSTR events were controlled by a single southerly airflow, and the southerly warm moist fluxes were transported up to + 3° north of the rainfall centers. In the vertical direction, the WSTR warm moist fluxes were distributed in parallel and close to the surface without any obvious tilt or gradient, and the warm moist layers were deeper than observed in the SLs. The reason for the warmer, wetter and deeper environment in the WSTR events might be that they occur primarily along the coastlines or coastal inland where moisture and warm temperature is abundant under the influence of southerly flow. In contrast, the SL events occur further inland where moisture is limited compared to that in the WSTR events. The onset of the East Asian subtropical summer monsoon might also contribute to the temperature and moisture differences between WSTR and SL events (Chen et al. ^[24]; Luo et al. ^[1]).

Figure 4.  Composite fields of the (a) temperature flux (℃ s^-1), (b) moisture flux (g s kg^-1), and (c) divergence (shaded, 10⁵ s^-1) and winds (vectors, m s^-1) of the WSTR events. Composite fields of the (d) temperature flux, (e) moisture flux, and (f) divergence and winds of the SLs. Shaded areas and vectors on (a), (b), (d), and (e) indicate the magnitude and direction of the fluxes, respectively.

Although the temperature and moisture fluxes were deeper and stronger, the convergence and lifting were shallower and weaker in the WSTR events than in the SL events at the onset of the 1-h rainfall maxima (Fig. 4c and Fig. 4f). For the SLs, there was considerable north-south wind convergence at lower levels and it tilted northward with height; the maximum values were distributed near the 850-900 hPa level. In comparison with those of the SLs, the dynamic disturbances of the WSTR events were much shallower near the rainfall centers and the convergence was below 950 hPa. In addition, the convergence was dominated by speed differences rather than by directional differences. It is noted that the dynamic disturbances below 950 hPa in the WSTR events were equivalent to those in the SLs. These dynamic disturbances were distributed in parallel and close to the surface, which facilitated the triggering of convection at lower levels. The distinct vertical structure of dynamic conditions in the WSTR events might be partially explained by the trumpet-shaped terrain to the north of the WSTR centers (discussed in section 5), while the land-sea contrast and the urban landscape might also help trigger WSTR convection under shallow convergence conditions (Chen et al. ^[21]; Chen et al. ^[22]; Chen et al. ^[23]).

As shown in Fig. 1, the selected 16 WSTR events all occurred to the south of low mountains. 4 occurred in the middle of Guangdong Province, 9 were distributed in western coastal areas, and the remaining 3 were in eastern coastal areas. The 16 WSTR cases were therefore divided into three categories: middle, western, and eastern, based on their locations. To examine the effect of terrain on the WSTR events, we performed composite analyses on the terrain and water vapor fluxes at 975, 950, and 900 hPa at the onset of the 1-h rainfall maxima (Fig. 5). The patterns and orientations of the composite terrain agreed well with those of individual events, and the trumpet-shaped topographical features were easily identifiable (black curves in Fig. 5).

Figure 5.  Composite terrain (gray contours, m) and water vapor fluxes (shaded, g s kg^-1) at the (a) 975-hPa level in west Guangdong, (b) 950-hPa level in west Guangdong, (c) 900-hPa level in west Guangdong, (d) 975-hPa level in east Guangdong, (e) 950-hPa level in east Guangdong, (f) 900-hPa level in east Guangdong, (g) 975-hPa level in central Guangdong, (h) 950-hPa level in central Guangdong, and (i) 900-hPa level in central Guangdong. Trumpet-shaped topography is represented by black contours.

At the 975-hPa level (approximately 300 m above sea level), almost all the southerly water vapor was blocked on the windward side of the mountains. At the 950-hPa level (approximately 500 m above sea level), most of the southerly airflow remained blocked by the higher mountains to the north. However, the trumpet-shaped topography allowed a narrow stream to cross the lower terrain, move northward, and converge at the channels. At the 900-hPa level (approximately 1, 000 m above sea level), the blocking effect was weakened because most of the mountains were less than 1, 000 m high. Thus, the effect of terrain on WSTR events could be classified into two aspects: blocking of the southerly airflow near the surface, and generating convergence of low-level warm and moist air in front of the trumpet-shaped terrain.

It should be noted that topography alone cannot trigger heavy rainfall event; favorable large-, meso-, and micro-scale conditions are also required (Du and Chen^[16]; Huang et al. ^[26]).

In this study, the precipitation, radar reflectivity, and associated synoptic characteristics associated with 16 WSTR events and 16 SL events, which occurred during the annually first rainy seasons of 2013-2017 in south China, were compared to examine the link between rainfall intensity and radar echo structure, as well as relevant environmental conditions supporting this association.

Observation analyses showed that precipitation intensities were higher, but radar echo centroids were lower in the WSTR events than those in the SLs. The WSTR reflectivities increased in a downward direction and the 50 dBZ centroids were below the 0℃ level. In contrast, the SL reflectivities increased in an upward direction and most 50 dBZ centroids were above the 0℃ level. This is consistent with the case studies shown in Luo et al. ^[1]. The different echo structures indicate that WSTR events are related to warm-rain processes, while SL events are related to stronger convective processes.

Composite analysis was used to examine ERA-Interim reanalysis data to determine the synoptic characteristics that governed the distinct reflectivity characteristics. In addition to the deeper and richer warm and moist environment, the shallow convergence and lifting conditions in the WSTR events favor the formation of warm rain and low echo centroids. Different from the research by Hamada et al. ^[39], which reported that the environment related to extreme rainfall events is convectively more stable than that of extreme convective events, our study found that the near-surface convective instability and lifting in the WSTR events are the same as, or stronger than, that in the SL events. This difference might be partially attributable to the trumpet-shaped terrain located in the north of rainfall centers.

The unique synoptic-scale similarities in the WSTR events presented in this paper provide references for future studies on evaluating and improving numerical weather prediction (NWP) model performance. For example, NWP models should be able to represent the shallow and close-to-surface dynamic lifting conditions correctly. It means intensifying the low-level layers in the vertical direction and improving the boundary layer parameterization scheme are possible methods to improve the predictability of WSTR. Note that this work provide only general descriptions of the synoptic characteristics of 16 warm-sector torrential rainfall events in south China in a 5-year period, since the temporal and spatial resolutions of ERA-Interim reanalysis data are limited. In the future, we will explore convection triggering mechanisms and more detailed microphysical features using analyses or forecasts from a convective-scale NWP model.