Cause–Effect Relationship between Meso-γ-Scale Rotation and Extreme Short-Term Precipitation: Observational Analyses at Minute and Sub-Kilometer Scales

Qiuyang ZHANG, Yali LUO, Ying TANG, Xin XU, Shuting YU, and Chong WU

1 School of Atmospheric Sciences, Chengdu University of Information Technology, Chengdu 610225 2 State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, China Meteorological Administration, Beijing 100081 3 Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters,Nanjing University of Information Science & Technology, Nanjing 210044 4 Nanjing Marine Radar Institute, Nanjing 211153 5 School of Atmospheric Sciences, Nanjing University, Nanjing 210023

ABSTRACT The cause-effect relationship between meso-γ-scale rotation and extreme short-term precipitation events remains elusive in mesoscale meteorological research.We aimed to elucidate this relationship by analyzing a rainstorm over the Pearl River Delta during the nocturnal hours of 15 May 2017 based on 6-min radar observations and 1-min rain gauge data.This rainstorm had a maximum hourly rainfall of 100.1 mm, with 26 stations recording hourly rainfall >60 mm h−1 in 5 h.Extreme heavy precipitation was produced in association with a convergence zone along the southern side of a synoptic low-level shear line, where southwesterly warm, humid airflows with precipitable water of > 60 mm, little convection inhibition (< 10 J kg−1), and a low lifting condensation level (about 300 m) dominated.A mesoγ-scale vortex was quantitatively identified during the hour with the largest number of gauges observing extreme hourly rainfall.The vortex had a mean diameter of 6.1 km and a peak intensity of 3.1 × 10−3 s−1 during its lifetime of 54 min.The vortex initialized and remained inside the region of extreme rain rates (radar-retrieved rain rates > 100 mm h−1), reached its peak intensity after the peak of the collocated 6-min rainfall accumulation, and then weakened rapidly after the extreme rainfall region moved away.The radar-retrieved liquid water path was about five to seven times the ice water path and the specific differential phase (Kdp) below 0°C increased sharply downward during the lifetime of the vortex, suggesting the presence of active warm rain microphysical processes.These results indicate that the release of the latent heat of condensation induced by extreme rainfall could have contributed to the formation of the vortex in an environment with a weak 0-1-km vertical wind shear (about 4-5 m s−1) through enhanced lowlevel convergence, although the strengthening of low-level updrafts by rotational dynamic effects and short-term rainfall cannot be ruled out.

Key words: extreme short-term precipitation, meso-γ-scale vortex, observational analysis, minute- and kilometerscale resolution

1.Introduction

Extreme short-term precipitation events have led to disasters such as floods, debris flows, and urban waterlogging worldwide.Extreme short-term precipitation events are characterized by their sudden occurrence and inhomogeneous distribution with localized centers (e.g.,Wu and Luo, 2016), which poses challenges in operational prediction, urban construction, and emergency management.South China is one of the rainiest regions in the world, with the climatological maximum annual precipitation and highest frequency of heavy rainfall events in China (Bao, 1986; Huang et al., 1986).

The rainy season in South China runs from April to early October (Ramage, 1952).The Pearl River Delta(PRD) urban agglomeration, located along the coastline of Guangdong Province, is the main economic and population center of South China.Heavy precipitation frequently occurs in this region, influenced by the East Asian summer monsoon, typhoons, sea-land breezes, and the local topography (Ding, 1994; Wang et al., 2014; Luo et al., 2017, 2020; Chen and Luo, 2018; Du and Chen,2019; Bao et al., 2021; Gao et al., 2022).The occurrence frequency of extreme hourly precipitation (exceeding the 95th percentile) has increased significantly since the beginning of rapid urbanization in this region in the early 1990s (Wu et al., 2019).This increase is probably related to the synergetic interactions between the urban-induced circulation (Changnon, 1968; Shepherd, 2005) and external forcing, including synoptic- and mesoscale disturbances and orographic forcing (Yin et al., 2020;Zhang, 2020; Sun et al., 2021).

The production of extreme short-term precipitation is closely related to convection activity and small-scale dynamic processes.The development of a mesoscale vortex has been observed during some extreme precipitation events in the USA (e.g., Morales et al., 2015; Nielsen and Schumacher, 2020a).About 50% of hourly rainfall accumulations > 75 mm h−1in the USA were reported to be associated with low-level meso-γ-scale vortices (2-20 km) embedded in supercells or organized mesoscale convective systems (MCSs) [Nielsen and Schumacher,2020b (NS20 hereafter)].

The spin term of the nonlinear dynamic termwhere ω′is the three-dimensional (3D) vorticity of the perturbation wind] in the diagnostic equation of vertical pressure disturbance implies that the cyclostrophic flow,regardless of its rotational direction, is always associated with a negative pressure perturbation (Markowski and Richardson, 2010).Such negative pressure perturbations tend to produce upward vertical acceleration (i.e., nonlinear dynamic acceleration) because the airflow tends to move toward the center of negative pressure perturbations.This, in turn, enhances the rise of warm, moist air at low levels and favors the production or maintenance of heavy precipitation.Based on idealized numerical simulations, Nielsen and Schumacher (2018) showed that,under the same environmental thermodynamic conditions, stronger meso-γ-scale rotation simulated with a higher low-level vertical wind shear (VWS) (i.e., 0-1-km VWS of 15.2 m s−1) enhanced the strength of the lowto mid-level updraft, resulting in substantially heavier precipitation.

Previous studies investigating the relationship between extreme short-term precipitation and meso-γ-scale vortices in the USA emphasized the importance of a strong low-level VWS and/or a strong 0-3-km environmental storm relative helicity (SRH) in the formation of vortices,which, in turn, enhanced rainfall (Trapp and Weisman,2003; Weisman and Trapp, 2003; NS20).The initial generation of vertical vorticity (ζ) often requires interactions between the horizontal vorticity caused by VWS and the horizontal gradient of the vertical velocity (i.e., the tilting term in the vertical vorticity equation) to function—that is, the horizontal vorticity along the storm inflow tilts and is then converted into vertical vorticity.The SRH is mainly determined by the horizontal vorticity along the direction of the relative storm streamline,which depends on the intensity of the low-level VWS and the movement of the storm.In environments with a large SRH, the horizontal streamwise vorticity in the lower troposphere is more dominant than the horizontal crosswise vorticity, meaning that the horizontal vortex tube is easier to tilt and stretch with the updraft to form a vortex (Markowski and Richardson, 2010).

A close association between the extreme short-term rainfall accumulation and a low-level meso-γ-scale vortex in the East Asian monsoon region was first documented by Li et al.(2021), who investigated the recordbreaking rainfall event affecting the megacity of Guangzhou in the PRD on 7 May 2017 using multisource observations.They found that the maximum 60-min rainfall accumulation of 219 mm was accompanied by a lowlevel (below about 3.5 km) meso-γ-scale vortex.In contrast with most of the events in the USA, this extreme rainfall event took place in an environment with a small 0-1-km VWS (< 5 m s−1) and a deep moist layer (relative humidity about 90% from the ground up to about 6 km).The initialization of the low-level rotation was closely related to the increase in the 0-1-km VWS to about 10 m s−1due to the formation of the convectively generated surface cold outflows, despite their weakness,and the slightly enhanced southerly flow in the planetary boundary layer.The vortex intensified mainly due to the stretching effect associated with the release of the strong latent heat of condensation and the weak surface cold outflows, which jointly increased the horizontal wind convergence at low levels.Yin et al.(2020) and Zeng and Wang (2022) described the low-level vortex in the Guangzhou “7 May” event in detail.Wang et al.(2021)studied a nocturnal extreme rainfall event in Northeast China in relation to the formation of a mesoscale vortex 20-30 km in diameter.

However, our understanding of the relationship between extreme short-term precipitation and meso-γscale rotation in the East Asian monsoon region is still limited.We therefore analyzed an extreme short-term precipitation event in the PRD during the night of 15 May 2017.The maximum hourly precipitation recorded by gauges was 100.1 mm h−1, whereas 26 automatic weather stations (AWSs) recorded hourly precipitation >60 mm.The relationship between the extreme hourly precipitation and meso-γ-scale rotation in this event is discussed from the perspective of observational analysis.For this purpose, we used observations at minute and sub-kilometer scales, including 1-min rain gauge data and 6-min dual-polarization Doppler radar observations.

The data and methods are described in Section 2, followed by an overview of the extreme rainfall event in Section 3.Section 4 analyzes the spatiotemporal evolution of the extreme rain rates and low-level vortex, the microphysical features, and the relevant environmental conditions.Our summary and conclusions are given in Section 5.

2.Data and methods

2.1 Data

We use hourly and 1-min rainfall AWS data provided by the China Meteorological Administration and the quantitative precipitation estimation (QPE) product based on the Guangzhou dual-polarization radar system to analyze the spatiotemporal distribution of rainfall.The AWSs in Guangdong Province are usually 5-10 km apart, but are distributed more densely in urban agglomerations (Fig.1b).The QPE is conducted for the lowest elevation angle (0.5°) using a method based on specific attenuation and the specific differential phase (Kdp)(Wang et al., 2019).This has been shown to provide a better estimate of heavy rainfall than conventional algorithms based on radar reflectivity and differential reflectivity (Cocks et al., 2019; Seo et al., 2020).The gridded QPE data are available every 6 min at a horizontal resolution of 250 m.We define extreme precipitation as rainfall of 100 mm h−1measured by the radar-based QPE.We use rainfall of 60 mm h−1as the threshold to define the extreme hourly precipitation observed by rain gauges,which is roughly the 99.9th percentile of the cumulative density function of the gauge-based hourly precipitation over the PRD in June over a 33-yr period (1981-2013)(59.9 mm h−1; Luo et al., 2016).

The radial velocities from the Guangzhou radar system are used to quantitatively identify and track meso-γscale vortices and the dual-polarization variables are used to estimate the liquid water path (LWP) and ice water path (IWP).Sounding observations at Hong Kong are used to examine the VWS and parameters relevant to the initialization and development of moist convection—for example, the convective available potential energy(CAPE).The fifth generation ECMWF atmospheric reanalysis (ERA5) dataset (0.25° × 0.25°; hourly) is used to analyze the synoptic background of the precipitation event and the environmental dynamic conditions closely related to vortex initiation.To this end, the temperature and humidity below 900 hPa are revised by using the observations from the AWSs.

Fig.1.Distributions of (a) land use types in 2015 and (b) AWSs in Guangdong Province.The black triangle denotes the location of the Guangzhou Doppler radar system (GZRD).The blue dots represent the location of the AWSs.The red contours roughly represent the city boundaries extracted from the Defense Meteorological Satellite Program Operational Linescan System global night-light data in 2013.The gray shading in (b) denotes the terrain height.

2.2 Identification and tracking of the vortex

The two-dimensional (2D) local, linear least-squares derivatives (LLSD) technique developed by Smith and Elmore (2004) is adopted to estimate the intensity of the vortices.Traditional azimuthal shear calculations (the peak-to-peak shear) only use velocity data from two points, taking the difference in radial velocity between the two velocity peaks and dividing this by the distance between them (the vortex diameter).By contrast, the LLSD method comprehensively considers the contribution of all radial velocities within a given range to the azimuthal shear of the center point, which alleviates the noise error of the radial velocity (Smith and Elmore,2004; Newman et al., 2013).When the ambient wind speed is very strong or the vortex is far from the radar system, the vortex can also be identified (Tang et al.,2020).

The raw radar data are processed by using the 88D2ARPS program (Brewster et al., 2005) of the Advanced Regional Prediction System (Xue et al., 2000) developed at the Center for Analysis and Prediction of Storms, University of Oklahoma, including the removal of non-meteorological echoes and de-aliasing of the radial velocity.The radar data preprocessing grid adopts a horizontal grid spacing of 1 km, with a total of 403 × 403 grid points.The vertical resolution is 500 m with a total of 33 layers.The Doppler velocity data are passed through a 3 × 3 median filter to reduce speckle noise (following Newman et al., 2013; Tang et al., 2020) and the azimuthal shear at 0.5° elevation is calculated by using the LLSD method.A vortex is recognized when the shear intensity of three consecutive volume scans (about 18 min) is > 10−3s−1.This shear intensity threshold follows that used by Tang et al.(2020) to analyze the statistical characteristics of meso-γ-scale vortices (regardless of precipitation) in the Yangtze-Huaihe River region of central East China.Vortices are tracked both backward and forward to obtain their lifetime using an objective tracking method similar to the WSR-88D Storm Cell Identification and Tracking algorithm (Johnson et al.,1998).

2.3 Estimation of the IWP and LWP

To discuss, at least qualitatively, the release of latent heat relevant to the stretching term in the vertical vorticity equation, we retrieve the LWP and IWP from the quality-controlled, gridded dual-polarization radar measurements.These measurements are interpolated onto a Cartesian grid at a horizontal resolution of 250 m and a vertical resolution of 500 m over 0.5-15 km above mean sea level (m.s.l.) using the nearest neighbor method and vertical linear interpolation.The LWP and IWP are the vertical integrated liquid water content (LWC) and ice water content (IWC) inside clouds, respectively.We use the difference reflectivity (Zdp) method to estimate the LWC and IWC by separating the contribution of liquidand ice-phase particles toZH(Carey and Rutledge, 2000);Zdpis expressed as:

whereZHandZVare the horizontal and vertical polarized reflectivity on a linear scale (mm6m−3), respectively.This method uses therelationship to separate the fraction ofZHproduced by liquid particles.The relationship was derived based on measurements from six 2D video distrometers at the Longmen Cloud Physics Field Experiment Base of the China Meteorological Administration during the warm seasons of 2017-2018 (Li et al., 2020):

If theZdpmethod indicates the presence of pure rain(the difference betweenZHandis less than one standard error orZH< 35 dBZ below the melting layer),and.Similarly, if theZdpmethod indicates the presence of pure ice (the difference betweenZHandis less than one standard error above the melting layer orZH< 35 dBZ above the −10°C layer),= 0 and.Otherwise, the presence of mixed-phase precipitation is considered.After the determination ofand, the following equations are used to estimate the LWC (g m−3) and IWC (g m−3) (Carey and Rutledge, 2000; Cifelli et al., 2002):

where ρiceis the density of ice at 0°C (ρice= 917 kg m−3)andN0is the intercept parameter of the ice particle spectrum assumed for simplicity (N0= 4 × 106m−4).

3.Event overview

The heavy rainfall event of interest occurred over Guangdong Province from 1700 local solar time (LST;LST = UTC + 8 h) 15 to 0700 LST 16 May 2017.The highest accumulated rainfall was concentrated over the urban agglomeration and adjacent areas in the PRD (Fig.2a; rectangular box denotes the key region of this study),with a maximum 13-h accumulated rainfall of 193.6 mm.The key region-averaged hourly rainfall increased persistently after 1700 LST and reached its peak at 0000-0100 LST 16 May, and then decreased rapidly to about a half within 2 h (Fig.2b).A total of 26 AWSs recorded extreme hourly rainfall (> 60 mm h−1) from 2100 to 0200 LST, with the largest number of AWSs (13) recording extreme hourly rainfall from 0000 to 0100 LST.The following analyses focus on the time period with extreme hourly rainfall—that is, the 5 h from 2100 to 0200 LST, especially 0000-0100 LST when a low-level vortex was objectively identified.

During the extreme rainfall event, the key region was controlled by relatively weak (about 5-7 m s−1) westerly flows in the midtroposphere without prominent troughs/ridges influencing Guangdong (Fig.3a).At 850 hPa, a northeast-southwest-oriented shear line was located over Guangdong.This was originally formed in northern Guangxi and then moved southeastward to the key region.The heavy rainfall center was located on the warm and moist tongue (θe> 345 K) near the northeastern part of the shear line (Fig.3b).The heavy rainfall was produced when the wind speed on both sides of the shear line increased gradually.There was a jet of northeasterly winds to the northwest of the shear line, with the maximum wind speed exceeding 12 m s−1.

The high-speed region of southwesterly winds to the southeast was located near the central to eastern coastline of Guangdong with a clear center over the Pearl River estuary.Air parcels under the southwesterly winds to the south of the shear line had to climb along isentropic surfaces prior to the occurrence of condensation, referred to as isentropic lifting (Raymond and Jiang, 1990;Zhang and Zhang, 2012).A northeast-southwest-oriented shear line was also visible at 925 hPa, positioned slightly to the south of its counterpart at 850 hPa (Figs.3c, d).The low-level shear line system implies the presence of a relatively lower pressure along the line,mass/moisture convergence, and upward motion, as discussed by, for example, Yao et al.(2020).The enhanced uplift over the region of interest can be explained by convergence induced by the lower pressure at low levels combined with isentropic lifting.The specific humidity at 925 hPa was > 16 g kg−1(Fig.3c), reflecting the high water vapor content of the boundary layer in the key region and upstream (i.e., the lower latitudes to the southwest).The prevailing southwesterly flows, the warm, moist air, and convergence at low levels resulted in environmental dynamic and thermodynamic conditions over the key region that were favorable for the formation and development of rainstorms.

The sounding observation in Hong Kong and the grid point in the ERA5 dataset closest to the rainstorm (see Fig.2a) at 2000 LST 15 May 2017 are used to analyze the upstream environmental conditions for extreme shortterm precipitation.The ERA5 dataset at 1000 hPa is corrected using the 2-m temperature and dewpoint from four nearby surface meteorological stations.The data between 1000 and 900 hPa are obtained by linear interpolation.The revised vertical profiles of the temperature and dewpoint (Fig.4a) suggest a large amount of precipitable water (63 mm), a small to moderate CAPE (718 J kg−1), a small convective inhibition (CIN) (8 J kg−1), and a low lifting condensation level (LCL) and level of free convection (LFC) (387 and 833 m, respectively).The dewpoint deficit up to 400 hPa is mostly < 3°C.The sounding observation in Hong Kong provides similar results(Fig.4b).

Fig.2.(a) Spatial distribution of accumulated precipitation from 1700 LST 15 to 0700 LST 16 May 2017.The black rectangle outlines the key region in this study (22.5°-23.75°N, 112.5°-114.5°E).The symbol “X” indicates the location of the maximum accumulated rainfall.The small black box denotes the Hong Kong sounding station (HK)and the star denotes the grid point in the ERA5 dataset used in Fig.4a.(b) The gray bars denote the average precipitation in the key region,whereas each line represents the hourly precipitation at the AWS with the largest hourly precipitation in any hour from 2100 to 0200 LST.

These results collectively indicate that convection could be initiated with weak lifting over the key region and the deep near-saturated moist layer and thick warm cloud layer (estimated to be about 4.3 km with 0°C at about 4.7 km) are favorable for active warm rain processes and the production of extreme short-term rainfall.The moisture is notably more abundant than during extreme short-term precipitation events in the USA (i.e., a mean precipitable water of about 50 and 47 mm with and without the presence of low-level rotation, respectively;NS20).The moderate convective intensity of the rainstorm is expected given the small to moderate CAPE.

Fig.3.Meteorological fields at 0000 LST 16 May 2017.(a) 500-hPa geopotential height (blue contours; dagpm) and wind barbs (half-barb 2.5 m s−1, full barb 5 m s−1, and pennant 25 m s−1); (b) 850-hPa wind speed (shading; m s−1), wind vectors (arrows), and the 345-K equivalent potential temperature (θe) (red contours); (c) 925-hPa specific humidity (shading; g kg−1), wind vectors (arrows), and 345-K θe (red contours); and (d)925-hPa wind speed (shading; m s−1), wind vectors (arrows), and strong convergence (white solid contours; divergence of −10−4 and −2 × 10−4 s−1).Black box indicates the region of heavy rainfall (i.e., the key region in this study).Gray areas denote terrain above 850 and 925 hPa, respectively.The thick red solid line indicates the shear line.

Fig.4.Skew T-logp diagrams at 2000 LST 15 May 2017 based on (a) the ERA5 dataset and (b) the sounding observation in Hong Kong.The red and green solid lines are the atmospheric temperature and dewpoint temperature curves, respectively.The area of shading represents the CAPE.Values are given for the CAPE, CIN, LCL, LFC, and precipitable water (PW).

4.Evolution of the storm producing extreme rainfall and the meso-γ-scale vortex

Fig.5.Hourly radar basic reflectivity factor (shading; dBZ) at 0.5° elevation over the key region from 2000 LST 15 to 0300 LST 16 May 2017.Each “X” symbol indicates one AWS with the hourly accumulated precipitation in the last hour > 60 mm.The black triangle denotes the location of the Guangzhou radar system and the four circles are centered on the radar system with radii of 30, 60, 90, and 120 km, respectively.

Scattered convective cells initiated over the northwest of the key region at about 1700 LST 15 May and then moved southeastward under the influence of the propagating synoptic shear line.The cells intensified with the enhancement of the low-level southwesterly flows (not shown) and were organized to form a northeast-southwest-oriented, quasi-linear MCS at 2100 LST(Figs.5a-c).The strongest portion of the MCS exhibited a bow shape at about 0000 LST, which became more obvious during the following hour, with 13 AWSs observing extreme hourly rainfall (Figs.5d-f).The MCS later became less organized, with weakened convective intensity (Figs.5g, h) and reduced rainfall rates, before moving out of the key region at about 0600 LST.

A pair of positive and negative radial velocities accompanying extreme rain rates near the bow echo can be seen in the radial velocity at 0.5° scanning angle from 0012 to 0054 LST (Figs.6e-l and 7e-l).This cyclonic rotation indicates the likely existence of a meso-γ-scale vortex.With the LLSD technique, the vortex is objectively identified (see Table 1 for time, center location,diameter, and intensity).This vortex was present for 54 min and moved little during its lifetime.Its maximum diameter was 8.9 km and its mean diameter was 6.1 km;the maximum azimuth shear was 3.09 × 10−3s−1with a mean of 1.90 × 10−3s−1.

Tang et al.(2020) identified 3790 meso-γ-scale vortices (regardless of the rainfall) over the lower reaches of the Yangtze River valley in central East China using the LLSD method with a mean lifetime, diameter, and intensity of 26.3 min, 8.1 km, and 2.3 × 10−3s−1, respectively.Analysis of the radar climatology of tornadic and non-tornadic vortices in the mid-Atlantic and southeastern USA used 6 × 10−3s−1as the minimum value of the azimuthal shear (Davis and Parker, 2014).Based on these comparisons, we conclude that the vortex in this study was weak to moderate compared with those in Tang et al.(2020) and much weaker than those in Davis and Parker (2014).Note that NS20 did not quantify the size and intensity of meso-γ-scale vortices associated with extreme short-term precipitation over the USA because they identified the vortices subjectively.

To better understand the relationship between the vortex and the extreme hourly rainfall, we carefully examine the spatiotemporal evolution of the rain rate and the azimuthal shear at 6-min intervals using the gridded 6-min QPE product.We find that the vortex initiated inside the region with rain rates > 100 mm h−1at 0012 LST(Fig.7e) and remained within this region until 0030 LST(Figs.7e-h).During this time period, the rotation intensified and reached its peak intensity (Table 1).The region with an extreme rain rate > 100 mm h−1later shrank (Fig.7i) and propagated more rapidly southeastward, while the vortex only moved a small distance (Figs.7j-l).When the vortex separated from the region with the extreme rain rate at 0042 LST (Fig.7j), it weakened significantly from 2.64 × 10−3to 1.22 × 10−3s−1(Table 1).

Table 1.Time, longitude and latitude of the center, diameter, and shear intensity of the vortex identified by using the LLSD method

Each 6-min rainfall accumulation at the three AWSs located closest to the center of the vortex (at distances of 1.07, 1.10, and 3.58 km, respectively) is also compared with the intensity of the vortex (Fig.8).The 6-min rainfall accumulation at the AWSs reached a peak at 0006 LST (G1956) and 0018 LST (G1915, G1982) before the emergence of the peak shear intensity at 0030 LST.These results suggest that the stretching effect enhanced by the latent heat of condensation may have contributed to the formation and intensification of the vortex, similar to the “7 May” event in Guangzhou (Li et al., 2021).However, the potential enhancement of rainfall by the nonlinear dynamic effects of rotation cannot be ruled out because the vortex was collocated with the region of extreme rainfall from 0012 to 0036 LST (Figs.7e-i).However, it is difficult to quantify the feedbacks between the release of latent heat, the strength of the vortex, and the intensity of rainfall using only the observations.

To gain more evidence of the strong release of the latent heat of condensation accompanying the low-level vortex, we analyzed the extreme precipitation feature(EPF) using observations from the dual-polarimetric radar system following Yu et al.(2022).An EPF is defined as the contiguous area of strong composite reflectivity (≥ 40 dBZ) containing extreme rain based on the radar QPE.The maximum height of the 40-dBZ echo-top (maxHt_40dBZ) is used as a proxy of strong updrafts in deep convection (Zipser et al., 2006).EPFs with maxHt_40dBZ reaching 9 km are defined as intense convection (Liu and Zipser, 2015), indicating that large precipitation particles (e.g., graupel and hail) are lifted to about −22°C over the PRD during the warm seasons.Following Yu et al.(2022), EPFs with maxHt_40dBZ below 6 km are defined as weak convection because their updrafts are not strong enough to support graupel/hail beyond about −4°C, whereas the other EPFs are defined as moderate convection with graupel/hail between 6- and 9-km altitude.From 0000 to 0100 LST,the maxHt_40dBZ was about 7.5 km (about −16°C)—that is, the EPF is of moderate convective intensity, consistent with the small to moderate CAPE of the inflow air(Fig.4).

The EPF-averaged IWP and LWP are about 2.0-3.5 and 15-18 kg m−2(Fig.9), respectively, which are within the 44-57 and 60-73 percentiles of the 9292 EPFs over the PRD during April-September 2016-2017 (Yu et al.,2022).It is important that the retrieved LWPs are about five to seven times the corresponding IWPs.TheKdpvalue increases rapidly below the 0°C level—that is, the vertical profile of 95th percentile (VP95) increases from about 1° km−1at 4 km to 2.3° km−1at 1 km above m.s.l.(Fig.10), which is even faster than the average obtained by Yu et al.(2022).AsKdpreflects the mass of the liquid hydrometeors, such a rapid increase inKdpindicates the efficient collection of cloud droplets by falling raindrops in the EPF.

Fig.6.Every 6-min radar basic reflectivity factor (shading; dBZ) at 0.5° elevation from 2348 LST 15 to 0054 LST 16 May 2017.Each “X”symbol indicates one AWS with precipitation > 60 mm (a, b) from 2300 to 0000 LST or (c-l) from 0000 to 0100 LST.The small blue circles in parts (e-l) denote the location and horizontal size of a vortex identified from Figs.7e-l, respectively.The dotted curves represent circles 30 and 60 km from the Guangzhou radar system.

Fig.7.Radial velocity (shading; m s−1) at 0.5° elevation every 6 min from 2348 LST 15 to 0054 LST 16 May 2017.Each “X” symbol indicates one AWS with precipitation > 60 mm (a, b) from 2300 to 0000 LST and (c-l) from 0000 to 0100 LST.The small red circles in parts (e-l) denote the location and horizontal size of the vortex.The blue and pink lines indicate contours with QPE rain rates of 60 and 100 mm h−1, respectively.The black dotted curves represent circles 30 and 60 km from the Guangzhou radar system, respectively.

Fig.8.Time series of the azimuthal shear intensity (black line) and gauge-based 6-min accumulated precipitation (blue, green, and red lines representing G1915, G1982, and G1956, respectively) from 0006 to 0100 LST 16 May 2017.

Fig.9.The dots indicate the (a) IWP and (b) LWP of the EPF every 6 min from 0006 to 0100 LST 16 May 2017.Different colors represent different times.The solid line represents the probability density function of the (a) IWP and (b) LWP of the 9292 EPFs in the PRD from April to September 2016-2017 identified by Yu et al.(2022).

Fig.10.The VP95 of Kdp for the EPF from 0000 to 0100 LST 16 May 2017.The gray line represents the average VP95 and the bars denote the range from the 25th to 75th percentiles of all the EPFs in Yu et al.(2022).

As the strength of the release of latent heat is proportional to the mass of the cloud droplets, these results further support the significant contribution of not only warm rain microphysical processes in producing the extreme short-term rainfall, but also the release of the latent heat of condensation in helping to form and enhance the lowlevel convergence and thus vertical vorticity.The latent heat of condensation induced by the extreme short-term rainfall through the stretching effect could therefore contribute to the formation and intensification of the meso-γscale vortex.However, it is hard to quantify the latent heat and its contribution to the formation of the vortex based only on observations.The pre-existing environmental convergence in the planetary boundary layer associated with the synoptic shear line could also contribute to the formation of the vortex through the stretching effect, although the maximum convergence was located to the south of the key region (Fig.3d).

The stretching effect cannot play a part in the evolution of the vortex unless the vertical vorticity is initialized by other mechanisms, which are often related to the environmental VWS and SRH (Trapp and Weisman,2003; Weisman and Trapp, 2003).About 12 min prior to the initialization of the vortex, the 0-1-km VWS over the key region was about 4-5 m s−1(Fig.11a), which is comparable with the average of the 299 extreme hourly rainfall events (> 75 mm h−1) with meso-γ-scale rotation over the USA (NS20).To calculate the 0-3-km SRH, the storm movement is estimated (3.88 m s−1) by tracking the strong echo (> 40 dBZ).The 0-3-km SRH was about 75-100 m2s−2(Fig.11b), which is about half the average value obtained by NS20 (about 150-200 m2s−2).These results suggest that the dynamic environmental conditions in the lowest kilometer for the initialization of the vortex in this event were roughly comparable with the average of those in the USA (NS20), whereas the 0-3-km SRH was significantly weaker.

Fig.11.(a) The 0-1-km VWS (shading and contours; m s−1) and 950-hPa wind vectors (arrows) and (b) the 0-3-km SRH (shading and contours; m2 s−2) and 700-hPa wind vectors (arrows) over the key region at 0000 LST 16 May 2017.The black circle denotes the location of the identified vortex.

5.Summary and conclusions

Meso-γ-scale rotation has been found to accompany extreme short-term precipitation over the USA (NS20)and the Guangzhou “7 May” record-breaking rainfall event (Li et al., 2021; Zeng and Wang, 2022); however,the relationship between meso-γ-scale rotation and extreme short-term precipitation in the East Asian monsoon region remains unclear.We analyzed an extreme short-term rainfall event over the PRD associated with a low-level vortex during the night of 15 May 2017 using integrated observations on minute and sub-kilometer scales from the densely distributed AWSs and a dual-polarization Doppler radar system at Guangzhou.Our main conclusions are as follows.

(1) This event took place with a synoptic shear line dominating the initiation and organization of convection.Strong southwesterly flows transported warm, moist air into the PRD and generated boundary layer convergence with northeasterly flows to the north of the shear line.The near-surface atmosphere over the key region had a low CIN, LCL, and LFC.The abundant moisture (precipitable water about 63 mm), the near-saturated lower to midlevels, and the deep warm cloud layer (about 4.3 km)favored the production of heavy rainfall and warm rain microphysical processes.

(2) A meso-γ-scale vortex was objectively identified during the hour with the largest number of AWSs recording extreme hourly rainfall (> 60 mm h−1).The vortex was present for 54 min with a mean diameter of 6.1 km and a mean azimuthal shear intensity of 1.90 × 10−3s−1—that is, it had a longer duration and comparable size and intensity to the average of 3790 meso-γ-scale vortices in East China regardless of the amount of precipitation (Tang et al., 2020).The maximum azimuthal shear was 3.06 × 10−3s−1, below the minimum threshold used in studies of tornadic and non-tornadic vortices in the USA (6 × 10−3s−1; Davis and Parker, 2014).

(3) Close to the time of the initialization of the vortex,the 0-1-km VWS was about 4-5 m s−1and the 0-3-km SRH was about half the average value of extreme shortterm precipitation events in the USA (NS20).The vortex initiated and remained inside the region of extreme rainfall (radar-based QPE > 100 mm h−1) until it reached its peak intensity and therefore the potential enhancement of rainfall by the nonlinear dynamic effects of rotation cannot be ruled out.The peak intensity of rotation lagged behind the peak of the collocated AWS-observed 6-min rainfall accumulation.The vortex weakened rapidly when the region of extreme rainfall propagated away from it and disappeared within 18 min.

(4) The EPF associated with the vortex was of moderate convective intensity and had a high LWC.The LWPs were about five to seven times the corresponding IWPs during the lifetime of the vortex andKdpbelow 0°C increased rapidly downward, indicating active warm rain microphysical processes.The release of latent heat in the lower levels of the rainstorm might therefore contribute to the formation and intensification of the vortex by enhancing low-level convergence.The pre-existing environmental convergence associated with the synoptic shear line could also contribute to the vortex.

This study has documented the features and evolution of a low-level meso-γ-scale vortex associated with an extreme short-term precipitation event in the PRD on the monsoon coast of South China using observations on minute and sub-kilometer scales.Our fine-scale analysis provides a more detailed and precise illustration of the temporospatial distribution of extreme rain rates and lowlevel rotation than many previous studies.These results add to our knowledge of the production of extreme precipitation in the East Asian monsoon region.The use of advanced techniques and tools, such as the four-dimensional (4D) variational Doppler radar analysis system(Sun and Crook, 1997), will help to reveal the 3D structure of the vortex and quantify the cause-effect relationship between the low-level rotation and extreme rain rates—that is, the likely feedbacks among the convergence enhanced by the rear inflow, the increase in the release of latent heat, the enhanced strength of the vortex,and the intensity of the rear inflow (Zhang, 1992; Davis et al., 2004).More general and robust conclusions call for future observational analysis of a large number of events with a coexisting vortex and extreme rainfall.Numerical modeling studies are required to deepen our understanding of their relationships and associated physical processes.

Acknowledgments.We thank three anonymous reviewers for providing constructive comments.This study is a part of the Southern China Monsoon Rainfall Experiment, which is a Research and Development Project of the World Weather Research Programme of the World Meteorological Organization.