Received: 19 August 2019 Revised: 17 February 2020 Accepted: 27 February 2020

DOI: 10.1002/met.1904

RESEARCH ARTICLE

Analysis of a waterspout at Zhuhai, China, on June 12, 2019

Pak Wai Chan1 | S.M. Tse1 | Jeffrey Chi Wai Lee1 | Q.S. Li2

1
Hong Kong Observatory, Hong Kong,
China Abstract
2
Department of Architecture and Civil This paper reports on a waterspout presiding over the Pearl River Estuary near
Engineering, City University of Hong Zhuhai, China. It is observed mainly using the terminal Doppler weather radar
Kong, Hong Kong, China
in Hong Kong. A slanting radar reflectivity core and airflow circulating around
Correspondence was seen in the vertical cross-section across the mesoscale/microscale cyclone
Pak Wai Chan, Hong Kong Observatory, associated with the waterspout. The vertical wind profile associated with the
134A Nathan Road, Kowloon, Hong
Kong, China.
cyclone was analysed and found to exhibit characteristics similar to those of a
Email: pwchan@hko.gov.hk tornado reported in the region in a previous study. The primary characteristic of
the waterspout was a circulating flow stronger at higher levels (approximately a
few kilometres above ground) than at lower levels. In terms of nowcasting, the
performance of a convection-permitting numerical weather prediction model
(2 km horizontal resolution) was analysed. It was found to demonstrate reason-
able simulation skill for the mesoscale/microscale cyclone, although the slan-
ting feature was not well predicted. The results of this study can serve as a
useful reference for similar studies of waterspouts/tornadoes worldwide.

KEYWORDS
numerical simulation, waterspout, weather radar

1 | INTRODUCTION tornado occurred in the warmer region to the south of

the surface cold front. Flows along a trough at 850 hPa
Waterspouts are only observed occasionally in the waters (in the form of a shear line) and at 500 hPa (mid-level)
around Hong Kong, China. Since recording began in were identified. A literature search did not reveal any
September 1959 until 2018, only 49 cases of waterspouts published studies concerning the climatology of water-
were sighted within 250 nautical miles of Hong Kong. A spouts in southern China (in particular Guangdong prov-
recent waterspout/tornado event over Hong Kong occurred ince). However, there has been a report of tornado
on August 29, 2018, as reported in the literature (Hon climatology in various parts of China (Fan and Yu, 2015).
et al., 2019). It was a rather weak system with a maximum From 2004 to 2013, the number of tornadoes over Guang-
velocity of approximately 10 m
s–1 (~20 kn) near the surface. dong province reaching a magnitude of 2 on the
A waterspout (later developing into a tornado) in May enhanced Fujita scale was 14, amounting to approxi-
2015 was reported by Chen et al. (2016). It was a rela- mately 1.5 tornadoes per year during the period. How-
tively weak system with a maximum radial velocity asso- ever, no systematic discussion was found regarding the
ciated with a meso-cyclone, reaching approximately meteorological conditions favourable for the occurrence
18 m
s–1 based on weather radar observations. The of tornadoes in that paper.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any
medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2020 The Authors. Meteorological Applications published by John Wiley & Sons Ltd on behalf of the Royal Meteorological Society.

Meteorol Appl. 2020;27:e1904. wileyonlinelibrary.com/journal/met 1 of 14

https://doi.org/10.1002/met.1904
2 of 14 CHAN ET AL.

Meteorological measurements around waterspouts International Airport and provide timely alerts. It is a C-
were conducted in the United States more than 40 years band single-polarization weather radar in dual configura-
ago (Leverson et al., 1977; Golden and Bluestein, 1994). tion, which means that it has dual transmitters, dual
Aircraft measurements were made to reveal the internal receivers, dual signal processors and dual radar product
structure of the waterspouts. Such aircraft measurements generators for redundancy. These components are
have not been made in southern China; however, water- grouped into two systems and connected to an antenna
spouts were occasionally observed from weather radar servo controller, which controls the antenna. The
images. Waterspouts may be observed as areas of high antenna is 7.9 m in diameter with a gain of 49 dB. The
radar reflectivity, and detailed Doppler velocity observa- transmission frequency is 5,625 MHz. The peak and aver-
tions of the internal structure of the mesoscale/micro- age power are 250 kW and 500 W, respectively. The
scale cyclone associated with the waterspouts are rather TDWR emits a narrow pencil beam with a half-power
rare. This paper documents such observations of a water- beam width of 0.55 . The transmitted pulse width is
spout near Zhuhai over southern China. 1.0 μs, and the range size is 150 m. The pulse repetition
Waterspouts over southern China are generally rather frequencies are 1,380 Hz and 1,104 Hz, resulting in a
weak. In fact, supercells and mesoscale convective sys- maximum unambiguous velocity of 73.6 m
s–1. The
tems, in general, are rather rare in the region. The meteo- TDWR operates in two modes, monitoring and hazard-
rological reasons for weak waterspouts are not clear; they ous, depending on the weather situation. The monitoring
may be related to the relatively low vertical wind shear in mode scans 14 elevations from 0.6 to 23.0 , successively,
the region in the summer, when generally south to south- in a 5 min volume scan when the weather is not wind-
westerly flow prevails over the lower and middle tropo- shear conducive with very light or even no precipitation.
sphere. Although the waterspouts over southern China It enters hazardous mode immediately when there is
are weak and relatively short-lived, they may cause some severe weather. In this scenario, the number of scans in
damage (e.g. Kosiba et al., 2014), and their detailed study the 5 min volume scan increases to 15, and the scanning
would still be useful for the issuance of weather warnings elevation ranges from 0.6 to 17.0 . The frequency of the
upon their appearance. 0.6 scan increases to approximately once every minute,
In the afternoon of June 12, 2019, a stronger water- as it is used for wind shear and microburst detection. The
spout event occurred at Zhuhai, a city approximately TDWR is approximately 40 km east of the waterspout.
50 km to the west of Hong Kong on the western side of Boundary-layer-type radar wind profilers are operated
the Pearl River Estuary. Based on the Doppler weather in Hong Kong to measure the vertical profile of the wind
radar observations, the near-surface velocity of the meso- every 10 min. This study considers the wind profiler
cyclone associated with the waterspout reached approxi- observations measured at the city centre of Hong Kong,
mately 50 kn (approximately 25 m
s–1). The occurrence of at a location called Sham Shui Po. It uses radar beams
this waterspout has been well captured by weather radar, with a frequency of 1,299 MHz to measure the three com-
and an analysis of the background meteorological condi- ponents of the wind by Doppler swinging, namely the
tions enabling its occurrence could be performed using vertical beam and two oblique beams with an azimuth
the data from a suite of meteorological instruments, such angle of 15 from the vertical. The first range gate is
as surface weather stations, radiosonde ascent and radar approximately 300 m above ground (to avoid measure-
wind profiler. The predictability of the mesoscale/micro- ment of the low-level shear), and the gate spacing is
scale cyclone feature could also be discussed based on a approximately 200 m, measuring up to a height of nearly
real-time mesoscale meteorological model called NHM 10 km above ground. Measurements are available in clear
(non-hydrostatic model), originally developed by the air by tracing the reflectivity discontinuity of the air and
Japan Meteorological Agency (Saito et al., 2006). This in rain by tracking the movement of rain drops blown by
paper serves to document the observations and prediction the wind. The wind profiler is approximately 50 km east
of the event, which may be useful as a reference for the of the waterspout.
study of waterspouts in southern China. There is also a dense network (with separation of a
few to a few tens of kilometres) of surface weather obser-
vations over southern China, with automatic weather sta-
2 | DATA AND METHODOLOGY tions measuring the horizontal components of the wind,
pressure, air temperature and dew point. Measurements
The terminal Doppler weather radar (TDWR) system was are normally available every 1 min. Radiosonde measure-
installed in Hong Kong in 2014 and began operation in ments can be performed at King’s Park, Hong Kong
2015 (Tse et al., 2019). It is used to detect real-time wind (approximately 1 km to the south of Sham Shui Po),
shear and microbursts in the vicinity of the Hong Kong using weather balloons, which are normally launched at
CHAN ET AL. 3 of 14

TABLE 1 The configuration of the RAPIDS-NHM model and its outer model Meso-NHM

Meso-NHM RAPIDS-NHM
Horizontal resolution 10 km 2 km
Horizontal grid Arakawa-C
Map projection Lambert conformal conic Mercator
No. of grid points 841 × 515 305 × 305
Vertical coordinates Terrain following height coordinates using Lorenz grid
No. of vertical levels 50 60
Model top 22,244 m 20,010 m
Time step 30 s 8s
Initial time 00, 03, 06, …, 21 UTC (for using EC Tiedtke Every hour
based [HKO version] convection scheme)
Forecast range 72 hr 15 hr
Initial condition 3DVAR using background from ECMWF IFS 3DVAR using background from Meso-NHM
forecast forecast
Boundary condition ECMWF IFS forecast data Meso-NHM forecast on 50 model levels
(0000 and 1200 UTC)
At 0.125 resolution in latitude/longitude

Boundary data including:

Pressure, horizontal wind components (U
and V), temperature (T) and relative
humidity (RH) on the surface; and U, V, T,
RH and geopotential height at 21 pressure
levels (1,000, 975, 950, 925, 900, 850, 800,
700, 600, 500, 400, 300, 250, 200, 150, 100,
70, 50, 30, 20, 10 hPa)
Nesting configuration One-way nesting
Topography USGS GTOPO30 (30 s data smoothed to 1.5 times horizontal resolution) with modifications
over Hong Kong areas based on USGS-SRTM (Shuttle Radar Topography Mission) dataset
Land-use characteristics USGS GLCC (global land cover characterization) 30 s data and 24 land-use types with
modifications over Hong Kong
Dynamics Fully compressible non-hydrostatic governing equations solved by time-splitting
horizontal-explicit vertical-implicit (HEVI) scheme using fourth-order centred finite
differencing in flux form
Cloud microphysics 3-ice (ice, snow and graupel) bulk microphysics scheme
Convective parameterization EC Tiedtke based scheme (HKO version) –
Surface process Flux and bulk coefficients
Over land: Beljaars and Holtslag (1991), Donelan et al. (2004)
Over sea: Beljaars and Holtslag (1991), Donelan et al. (2004) (RAPIDS-NHM)
Wong et al. (2011) (Meso-NHM)
Roughness length: Beljaars (1995) and Fairall et al. (2003); stomatal resistance and temporal
change of wetness included; four-layer soil model to predict ground temperature and surface
heat flux
Turbulence closure model and Mellor–Yamada–Nakanishi–Niino Level 3 (MYNN-3) (RAPIDS-NHM)/
planetary boundary layer process Mellor–Yamada–Nakanishi–Niino Level 2.5 (MYNN-2.5) (Meso-NHM) (Nakanishi and
Niino, 2004) with partial condensation scheme (PCS) and implicit vertical turbulent solver
Height of planetary boundary layer calculated from virtual potential temperature profile
Atmospheric radiation Long wave radiation process following Kitagawa (2000)
Short wave radiation process using Yabu et al. (2005)
Prognostic surface temperature included; cloud fraction determined from PCS
4 of 14 CHAN ET AL.

F I G U R E 1 The waterspout
observed from Zhuhai as reported in the
news. The source of the pictures is not
known. The pictures were taken at
around 1700 HKT, June 12, 2019
(source: in Chinese, http://www.gd.
xinhuanet.com/newscenter/2019-06/13/
c_1124618280.htm)

0000 and 1200 UTC of the day (i.e. 0800 and 2000 in numerical weather prediction (NWP) forecast. The
Hong Kong time, with HKT = UTC + 8 hr). authors had access to real-time sea surface temperature
A convection-resolving model, RAPIDS-NHM, was analysis from polar-orbiting satellites; however, the per-
used in Hong Kong. RAPIDS-NHM is a 2 km resolution formance in past model runs is similar to that based on
model that runs on an hourly basis, i.e. initialized each the ECMWF analysis. Details of RAPIDS-NHM and its
hour. The initial condition was based on 3D variational outer model are available in Table 1.
data assimilation analysis, which combined conventional
(e.g. surface stations, Aircraft Meteorological Data
Reports) and remote sensing observations (e.g. radar 3 | WEATHER O BSERVATIONS
reflectivity and atmospheric motion vector from
Himawari-8) over the Pearl River Estuary. The boundary According to a news report, the waterspout was observed
conditions were prescribed, in one-way nesting, by an over the offshore waters of Zhuhai between approxi-
outer 10 km resolution model of the NHM. The sea sur- mately 1630 and 1700 HKT. The observations of the
face temperature was based on the European Centre for waterspout from the news are shown in Figure 1. The
Medium-Range Weather Forecasts (ECMWF) analysis suspected location of the waterspout is shown in
and was fixed throughout the 15 hr forecast. Notably, the Figure 2b. Synoptically, southern China was under the
coarse resolution of the EMCWF analysis could not cap- influence of a low-pressure trough (Figure 2a), which
ture the small-scale variation, owing to the complicated brought unsettled weather to the region. Thunderstorms
land–sea boundary in the area. However, there were no occurred over time over the Pearl River Estuary in the
better observation data available to the authors for the afternoon. The surface trough was associated with upper-
CHAN ET AL. 5 of 14

F I G U R E 2 (a) The surface analysis chart at 0800 HKT, June 12, 2019. The blue curves are the surface troughs of low pressure. For each
observation, wind barbs show the speed and direction of the surface wind (full barb 10 kn [5 m
s–1] and half barb 5 kn [2.5 m
s–1]). The
number associated with each surface station is the temperature. (b) The surface observations near the waterspout at about 1644 HKT, June
12, 2019. The filled circle in (b) shows the suspected location of the waterspout. The number at the upper left corner of the wind barb is the
temperature, the number in red is the dew point and the number at the upper right corner of the wind barb is surface pressure (e.g. 033
means 1,003.3 hPa)
6 of 14 CHAN ET AL.

F I G U R E 3 (a) The observations from the radar wind profiler. The wind barbs show the speed and direction of the wind (full barb
10 kn [5 m
s–1] and half barb 5 kn [2.5 m
s–1]). The height is above the mean sea level. The horizontal axis is the time in Hong Kong time
(HKT = UTC + 8 hr). The vertical shear is calculated from the wind profiler measurements and displayed in (b). The horizontal axis is the
time in Hong Kong time. WGL refers to the wind at Waglan Island (location in Figure 2b). The vertical wind shear at the time of the
waterspout (1644 HKT) can be read from the vertical line in (b)

level troughs at the 850 and 700 hPa levels. This could be southern China in the early summer and has been exten-
explained by the phenomenon of quasi-geostrophic front- sively studied as the Meiyu–Baiu front (e.g. Sampe and
ogenesis (Chen, 1993), which normally occurs over Xie, 2010).
CHAN ET AL. 7 of 14

F I G U R E 4 The radiosonde
measurements of King's Park, Hong
Kong station (45004) at 0800 HKT,
June 12, 2019. The wind barbs show
the speed and direction of the wind
(full barb 10 kn [5 m
s–1] and half
barb 5 kn [2.5 m
s–1]). ET85 means
equivalent potential temperature at
850 hPa. Similar ET values apply to
the 700 and 500 hPa level

Based on the surface observations, the flow appeared trough. The southwesterly flow brought relatively warm
to be cyclonic at the suspected location of the waterspout and abundant moisture to southern China, which would
(Figure 2b). Unfortunately, there were no surface obser- favour the occurrence of unstable weather, such as heavy
vations sufficiently close to the waterspout to measure its rain and thunderstorms. This increased the instability
strength directly. According to a news report, the water- (conditional instability) of the atmosphere. In the middle
spout remained quasi-stationary over there and dis- troposphere, there was a jet of strong to gale force wind
appeared afterwards. speed (22–34 kn, i.e. approximately 11–17 m
s–1). This
Synoptically, at 850 and 700 hPa, an east–west ori- type of jet is rather common over southern China during
ented trough occurred over central China, and the south the summer monsoon season. From the synoptic analysis
China coast was under the influence of a southwesterly of the 500 hPa level (not shown), the west to southwest-
flow. At 500 hPa, a westerly jet appeared along this erly jet stream extended along the coastal area of south
coastal area. At 200 hPa, an east–west oriented ridge axis China. The waterspout event occurred right over the jet
associated with a broad anticyclone occurred along the core but not in the entrance/exit region of the jet.
coast. This was related to the significant divergence near According to Sampe and Xie (2010), the westerly jet cau-
the Pearl River Delta region, favouring the occurrence of sed horizontal warm advection, triggering the upward
significant convection. motion along the Meiyu–Baiu rainband. The advection of
The wind profiler observations for that day are shown warm air led to convection over southeastern China by
in Figure 3a. The lower troposphere was mainly domi- inducing an adiabatic ascent along the westerly jet. This
nated by a south to southwesterly flow under the south- favours the occurrence of intense convection along the
west monsoon associated with the low-pressure surface jet. The vertical wind shear based on the wind profiler
8 of 14 CHAN ET AL.

FIGURE 5 (a) The TDWR radar reflectivity measurements at an elevation angle of 4 at about 1644 HKT, June 12, 2019. The radius of
each circle around the airport is 25 nautical miles. (b) The corresponding Doppler velocity measurements. Positive radial velocity means
wind component away from the radar and negative radial velocity means wind component towards the radar. The radius of each circle
around the radar’s location is 50 km. The circled areas are the areas with higher radar reflectivity and Doppler velocity couplet
CHAN ET AL. 9 of 14

et al., 2018). The sea surface temperature was 27.5 C,

whereas the 850 hPa temperature was 20.7 C, producing
a difference of 6.8 C. The lifting condensation level was
973 hPa, and the equilibrium level was 141 hPa. The
heights were 303 and 14,769 m above mean sea level, giv-
ing a height difference of 14,466 m or 47,460 ft. The
waterspout was marginally within the category of
thunderstorm-related waterspouts in the Szilagyi Water-
spout Nomogram, and the SWI was approximately
10, suggesting the possibility of waterspout occurrence.
More cases are being collected to determine if the SWI is
applicable to southern China.

4 | TDWR OBSERVATIONS

TDWR observations taken at a 4 elevation angle at

approximately the time when the mesoscale/microscale
F I G U R E 6 Vertical sections along the white line shown in cyclone associated with the waterspout was the strongest
Figure 5a. The viewing direction is shown in red. (a), (b) The cross- are shown in Figure 5. They are plan position indicator
section of reflectivity and Doppler velocity respectively. The scans. The approximate location of the waterspout is
horizontal distance between adjacent vertical lines is 1 nautical
shown in Figure 5a. This is an area of relatively high
mile. The vortex is confined to the lowest kilometre or so. The
reflectivity. In southern China, severe convection is
shear signature tilts with height. It is roughly in the direction of the
expected when the radar reflectivity is 33 dBZ or greater,
low-level wind shear. Time is 1644 HKT, June 12, 2019
although this criterion has not been documented and is
only used empirically. The waterspout appeared to be an
measurements is shown in Figure 3b, which represents ordinary multi-cellular storm. The Doppler velocity is
the magnitude of the vector differences between the shown in Figure 5b. An opposing velocity couplet was
winds at 1, 3 and 6 km above the mean sea level, as observed, and there was a mesoscale/microscale cyclone
recorded by the wind profiler, and the wind at Waglan at the location of the waterspout.
Island (Figure 2c). This is taken to be representative of A vertical cross-section was made at the line in
the surface wind in Hong Kong, divided by the depth of Figure 5a. The reflectivity and Doppler velocity along
the layer. Since approximately noon of that day, the verti- the plane are shown in Figure 6a and 6b, respectively.
cal wind shear rose above the wind shear mean value of From Figure 6a, there is a tilted area of the higher
the day. One possible reason for the increase in vertical radar reflectivity core. A similar slanting area is shown
shear may have been the mesoscale vertical circulation in the positive velocity in Figure 6b. This is a slanting
solenoidally induced by differential heating between the reflectivity core surrounded by circulating air. The
land and water. However, this needs to be confirmed by tilting may be associated with a gust front; however,
further studies (e.g. numerical simulation). The relatively further studies of the evolution of the thunderstorms
large value of vertical wind shear (of the order of are required.
12 × 10−3 s−1 for the 0–1 km vertical wind shear) is Slanting tornadoes in the intensification phase have
favourable for the occurrence of a waterspout been reported in the literature (e.g. Bluestein et al., 2019).
(Li et al., 2017). It has been observed in many cases in the The slanting tornado is considered to be a misalignment
United States and China that such a high vertical wind between the mid-level and low-level meso-cyclones. The
shear value is favourable for waterspout/tornado forma- separation is expected to enhance the stretching (increas-
tion; however, the physical mechanism is not certain. ing the length of the vortex and tilting it onto the hori-
The radiosonde measurement at 0800 HKT on the zontal) and intensification of the pre-existing vortex,
day is shown in Figure 4. The troposphere was rather acting to amplify tornadogenesis. Owing to the limited
humid up to the tropopause, and the atmosphere was amount of weather radar observations of waterspouts/
unstable with a K-index of 38. The convective available tornadoes over southern China, it is not clear whether
potential energy was approximately 1,173 J
kg–1. such slanting vortices are a common feature in the
The Szilagyi Waterspout Index (SWI) based on the region, representing an intensifying vortex. More obser-
radiosonde measurement is considered (Renko vational studies are required.
10 of 14 CHAN ET AL.

F I G U R E 7 (a) The time series of the maximum Doppler positive and negative velocities at 0.6 and 1 elevation angles, in association
with the mesoscale/microscale cyclone. The horizontal axis is the local time in hour–minute–second (HHMMSS). (b) The difference between
the maximum outbound and maximum inbound Doppler velocities as a function of time and height

The maximum Doppler velocity associated with the 1,630 H to 1,645 H in the first 4 km above ground, indi-
mesoscale/microscale cyclone at the lowest two elevation cating the strengthening of the cyclonic vortex. The
angles, namely 0.6 and 1 , were selected and used to corresponding profiles of the maximum positive velocity
produce the time series in Figure 7a. It is noted that there and maximum negative velocity with height are shown
is a positive value in the maximum negative velocity at in Figure 8a and 8b, respectively. The maximum positive
1637 HKT as no negative value is identified, and this velocity falls from approximately 40–50 kn (20–25 m
s–1)
small positive value appears to be at the location of the at the lowest 1 km to approximately 30 kn (15 m
s–1) at
maximum negative value by time continuity. The maxi- altitudes above the boundary layer. The fall of the maxi-
mum value of the velocity (approximately 52 kn, approxi- mum positive velocity with height is rather gentle and
mately 26 m
s–1) reaches the largest value at 1646 HKT. not very significant. Meanwhile, the maximum negative
The difference between the maximum outbound and velocity, which is the background southwest monsoon
maximum inbound Doppler velocities as a function of superimposed on the mesoscale/microscale cyclonic flow,
time and height is shown in Figure 7b. It increases from remains at a rather low value of −20 kn (−10 m
s–1)
CHAN ET AL. 11 of 14

F I G U R E 8 The vertical profile of

maximum positive (a) and negative
(b) Doppler velocities with height at the
time of maximum values of lower
elevation Doppler velocities in
association with the mesoscale/
microscale cyclone. The y axis is the
height in metres and the x axis is the
velocity in knots

within the atmospheric boundary layer and increases to be analysed at the location of heavy rain and there
approximately −40 kn (−20 m
s–1) in the middle tropo- appears to be a “notch” of simulated rainfall (sum of rain
sphere. This is consistent with the observation of the and snow) associated with the cyclone. The notch is a
mid-tropospheric jet stream from the wind profiler in weaker weather radar echo region that may be associated
Figure 3a. A similar increase in Doppler velocity was also with a tornado. A similar cyclonic flow is analysed in the
reported in a previous study of tornadoes by Hon upper part of the boundary layer. For instance, at
et al. (2019). More cases of waterspouts/tornadoes for 850 hPa in Figure 9b there is an area of strong positive
southern China would be needed to determine whether vorticity. At 200 hPa in Figure 9c, there is an area of sig-
this Doppler velocity profile shape is generic or not; how- nificant divergence in the northwesterly flow. The low-
ever, they serve as useful references for wind engineering level cyclonic flow and upper-level divergence favour the
applications for tornadic winds in southern China. occurrence of strong convection.
From Figure 9b and 9c, apart from the location of the
waterspout, there are other areas of significant vorticity
5 | N W P RE S U L T S at 850 hPa that exhibit significant divergence at 200 hPa
(e.g. over inland areas of the Pearl River Delta). It is
The real-time NHM run initiated at 1400 HKT, June noted that these are also areas with significant updraft
12, 2019, was considered. The 3 hr forecast results are helicity.
shown in Figures 9 and 10. The horizontal sections are From Figure 9d, an upper air divergence was present
presented in Figure 9. An area of heavy rainfall was fore- even before the triggering of the waterspout-related con-
cast in the Pearl River Estuary, as indicated in Figure 9a. vective activity. Thus, it is not only an outcome of con-
Cyclonic flow and the associated low-pressure area could vection but also a support to convection.
12 of 14 CHAN ET AL.

F I G U R E 9 The mean sea level pressure, surface (10 m) wind and 1 hr accumulated rainfall amount (a), 850 hPa vorticity (b), 200 hPa
divergence (c), updraft helicity (e) at the time of 3 hr forecast as initialized at the RAPIDS-NHM run at 1400 HKT, June 12, 2019, and
200 hPa divergence (d) at the initial time of the model run. The wind barbs show the speed and direction of the wind (full barb 10 kn [5 m
s–
1
] and half barb 5 kn [2.5 m
s–1]). The brown circle in (a) is the rough location of the actual occurrence of the waterspout

The development of convection is shown in the east of the intense convection. Thus, cyclonic flow
Figure 10. It is a vertical cross-section at the location appears to be associated with intense convection.
given in Figure 9a. The wind barbs show the speed and Cyclonic flow with strong reflectivity is forecast in the
direction of the wind (full barb 10 kn [5 m
s–1] and half- model, although it does not appear to be tilting in height,
barb 5 kn [2.5 m
s–1]). Near the mid-troposphere as suggested in the TDWR observations. The lack of
(e.g. 500 hPa), a westerly flow appears to the west of the tilting in the modelled convection is due to the limited
intense convection, and a southwesterly flow appears to resolution of the NHM model. The horizontal scale of the
CHAN ET AL. 13 of 14

F I G U R E 1 0 Vertical cross-section
across the high reflectivity core at the
line in Figure 9a. The time is 1700 HKT,
June 12, 2019

tilting is 3–4 nautical miles, which is too small for the weather radar is not correctly simulated. The NWP model
2 km NHM model to resolve. The NHM model is capable with convection-resolving resolution or even higher (sub-
of forecasting the occurrence of mesoscale/microscale kilometre) resolution appears to demonstrate skill in pro-
cyclones 3 hr ahead. The occurrence of mesoscale/micro- viding a reasonable forecast of mesoscale/microscale
scale cyclones is also forecast in subsequent model runs cyclones, namely the cyclonic flow embedded in the con-
(07Z and 08Z) with slight variation in time (±30 min) vection extending to the upper troposphere and the rela-
and location up to 40 km (not shown). tively high vorticity.
Additional cases of waterspouts/tornadoes in the
southern China region will be studied in the future to
6 | C ON C L U S I ON S evaluate the performance of the NWP, in general, and
the vertical structure of the airflow associated with that
The analysis of a waterspout occurring over the Pearl phenomenon.
River Estuary is presented in detail in this paper. It is
well captured by weather radar, permitting examination ORCID
of the vertical structure of the waterspout through con- Pak Wai Chan https://orcid.org/0000-0003-2289-0609
sideration of the reflectivity core and Doppler velocities. S.M. Tse https://orcid.org/0000-0001-6498-6372
The southwesterly flow (negative Doppler velocity)
appears to have a structure similar to the results pres- RE FER EN CES
ented in a previous study of a tornado in the region (Hon Beljaars, A.C.M. and Holtslag, A.A.M. (1991) Flux parameterization
et al., 2019), namely a lower radial velocity in the first and land surfaces in atmospheric models. Journal of Applied
few kilometres above ground, which then increases sig- Meteorology, 30, 327–341.
nificantly at higher levels (by a factor of 1.5–2). More Beljaars, A. (1995) The parametrization of surface fluxes in large-
scale models under free convection. Quarterly Journal of the
analysis must be performed to determine if this is a
Royal Meteorological Society, 121, 255–270.
generic structure of the vertical wind profiles for water- Bluestein, H.B., Thiem, K.J., Snyder, J.C. and Houser, J.B. (2019)
spouts/tornados in the region. Tornadogenesis and early tornado evolution in the El Reno,
As in the previous case, the high-resolution numerical Oklahoma, supercell on May 31, 2013. Monthly Weather Review,
weather prediction (NWP) model is able to capture the 147, 2045–2066.
occurrence of strong convection. The vertical structure of Chen, Y.-L. (1993) Some synoptic-scale aspects of the surface fronts
the mesoscale/microscale cyclone appears to be well over southern China during TAMEX. Monthly Weather Review,
predicted, although the tilting structure observed by the 121, 50–64.
14 of 14 CHAN ET AL.

Chen, Y., Yu, X., Chen, X., Wang, S. and Luo, M. (2016) A tornado Nakanishi, M. and Niino, H. (2004) Improvement of the Mellor–
in South China in May 2015. Journal of Applied Meteorological Yamada level 3 model with condensation physics: its design
Science, 27, 334–341 (in Chinese with English abstract). and verification. Boundary-Layer Meteorology, 112, 1–31.
Donelan, M.A., Haus, B.K., Reul, N., Plant, W.J., Stiassnie, M., Renko, T., Ivusic, S., Prtenjak, M.T., Solian, V. and Horvat, I. (2018)
Graber, H.C., Brown, O.B. and Saltzman, E.S. (2004) On the Waterspout forecasting method over the eastern Adriatic using
limiting aerodynamic roughness of the ocean in very strong a high-resolution numerical weather model. Pure and Applied
winds. Geophysical Research Letters, 31, L18306. Geophysics, 175, 3579–3778.
Fairall, C.W., Bradley, E.F., Hare, J.E., Grachev, A.A. and Edson, J. Saito, K., Fujita, T., Yamada, Y., Ishida, J., Kumagai, Y., Aranami, K.,
B. (2003) Bulk parameterization of air-sea fluxes: updates and Ohmori, S., Nagasawa, R., Kumagai, S., Muroi, C., Kato, T.,
verification for the COARE algorithm. Journal of Climate, 16, Eito, H. and Yamazaki, Y. (2006) The operational JMA non-
571–591. hydrostatic model. Monthly Weather Review, 134, 1266–1298.
Fan, W. and Yu, X. (2015) Characteristics of spatial–temporal distri- Sampe, T. and Xie, S.-P. (2010) Large-scale dynamics of the Meiyu–
bution of tornadoes in China. Meteorology Mon., 41, 793–805 Baiu rainband: environmental forcing by the westerly jet. Jour-
(in Chinese with English abstract). nal of Climate, 23, 113–134.
Golden, J.H. and Bluestein, H.B. (1994) The NOAA-National Geo- Tse, S.M., Hagio, M. and Maeda, Y. (2019) Windshear detection by
graphic Society waterspout expedition (1993). Bulletin of the terminal Doppler weather radar during tropical cyclone
American Meteorological Society, 75, 2281–2288. Mujigae in 2015. Meteorological Applications, accepted on
Hon, K.K., Tse, S.M., Chan, P.W. and Li, Q.S. (2019) Observation March 9, 2019. https://doi.org/10.1002/met.1789.
and real-time simulation of a tornado event in Hong Kong on Wong, W.K., S. Sumdin and S.T. Lai, 2011: Development of air–sea
August 29, 2018. Advances in Meteorology, 2019, 8571430. bulk transfer coefficients and roughness length in JMA non-
Kitagawa, H. (2000) Radiation process. Rep. Numer. Predict. Div., hydrostatic model. Tech. Rep. Meteorol. Res. Inst. No. 65: "Inter-
46, 16–31 (in Japanese). national Research for Prevention and Mitigation of Meteorologi-
Kosiba, K.A., Robinson, P., Chan, P.W. and Wurman, J. (2014) cal Disasters in Southeast Asia", 82-85.
Wind field of a nonmesocyclone anticyclonic tornado crossing Yabu, S., Murai, S. and Kitagawa, H. (2005) Clear sky radiation
the Hong Kong international airport. Advances in Meteorology, scheme. Rep. Numer. Predict. Div., 51, 53–64 (in Japanese).
2014, 597378.
Leverson, V.H., Sinclair, P.C. and Golden, J.H. (1977) Waterspout
wind, temperature and pressure structure deduced from aircraft
measurements. Monthly Weather Review, 105, 725–733. How to cite this article: Chan PW, Tse SM,
Li, Z., Wang, D., Mai, X., Wu, K., Liu, Y., Wang, H., Zhu, J., Lee JCW, Li QS. Analysis of a waterspout at
Yan, L., Li, C., Chen, H. and Long, Y. (2017) Observations of Zhuhai, China, on June 12, 2019. Meteorol Appl.
the tornado occurred at Foshan on October 4, 2015. Acta Met- 2020;27:e1904. https://doi.org/10.1002/met.1904
eorologica Sinica, 75, 288–313 (Chinese with English abstract).