Radar Analysis of Cyclic Tornadic Mesocyclones Within the 23 June 2016 Yancheng Supercell Storm in China

Highlights

What are the main findings?

• Cyclic tornadic mesocyclones were observed using S-band stationary radar in a supercell storm over Yancheng, China, on 23 June 2016.
• The storm exhibited three discrete mesocyclones, providing a rare observational example of a severe cyclic supercell occurring in China.

What are the implications of the main findings?

• The storm reflectivity intensity appeared to be closely coupled with the occlusion and formation of mesocyclones and tornadoes.
• The timing of mesocyclogenesis may be associated with the weakening of the reflectivity core and its subsequent lowering toward the surface.

Abstract

On 23 June 2016, a supercell storm in Yancheng City, Jiangsu Province, China, produced two tornadoes: the first reached EF4 intensity and lasted approximately 50 min, while the second was rated EF2–EF3 and persisted for about 20 min. Evidence of cyclic tornadic mesocyclones within the supercell was captured by an S-band stationary Doppler radar, providing a rare case of a severe cyclic tornadic storm in China. Three discrete mesocyclones were identified during the storm’s evolution. The first occlusion persisted for approximately 42 min and was relatively prolonged; the submesocyclone-strength vortex associated with the old mesocyclone remained detectable for 18 min after the hook echo completed wrapping around itself. In contrast, the second occlusion evolved more rapidly and was completed within 18 min, during which the old and new mesocyclones rotated around each other. A temporal association was observed between the evolution of the reflectivity core and the occlusion and formation of the mesocyclones. The second and third mesocyclones formed approximately 24 and 18 min, respectively, after the reflectivity core weakened and lowered toward the surface.

1. Introduction

One supercell storm can produce a series of discrete vortices, updrafts, mesocyclones or tornadoes during its lifespan [1,2]. Burgess [3] proposed a conceptual model of cyclic mesocyclones based on radar data, in which the occlusion of mesocyclones resembled the extratropical cyclones in structure. Adlerman et al. [4] reproduced similar conceptual model using numerical simulations. Dowell and Bluestein [5,6] suggested that the cyclic tornadogenesis process is associated with an imbalance between inflow and outflow of the storm and a mismatch between the movement directions of the tornado and its parent storm. Adlerman and Droegemeier [7] presented two types of cyclic tornadogenesis in a numerical study, occluding and non-occluding, which depend on the vertical wind shear. With high-resolution mobile radars, Beck et al. [8] proposed a detailed conceptual model for the decay and reformation of hook echoes, and French et al. [9] studied the specific features of circulations at multiple levels of cyclic mesocyclones. Observations using polarimetric radar revealed more sophisticated and reliable signatures of supercell storms [10,11,12]. Kumjian et al. [13] analyzed the repetitive evolution of polarimetric features associated with the cyclic occlusion of the low-level mesocyclone using a rapid-scanning Weather Surveillance Radar-1988 Doppler (WSR-88D). Some atypical cyclic tornadoes were observed by mobile rapid scan polarimetric radars, such as the ‘failed occlusion’ during a loop-tracked tornado [14], and a hybrid case exhibiting both occlusion and non-occlusion characteristics [15]. Cyclic storms are typically among the most severe storms and may occur in tornado-prone regions worldwide. However, observations of these storms remain far less common outside North America [16,17,18].

In eastern China, tornadoes occur frequently, although the region is not among the most heavily affected. Fan and Yu [19] studied the spatial–temporal distribution of significant tornadoes in China from 1961 to 2010 and found that the average annual number of tornadoes rated higher than 1 on the Enhanced Fujita (EF) scale in China was 21, much less than in the United States and Europe. Zhang et al. [20] studied the climatology of 129 tornadoes with intensities exceeding EF2 from 1980 to 2016 based on the 5th generation ECMWF reanalysis (ERA5) dataset and found that the kinematic conditions for tornadic storms in China are much less favorable than those in the United States, which explains the lower tornado frequency and intensity in China.

At present, Doppler weather surveillance radar is an effective tool for tornado detection and warning [21]. Since the 1990s, a network of surveillance radars has been established in China [22]. These radars, known as China New Generation Weather Radar (CINRAD) or WSR-98D, have been used extensively. With the help of CINRADs, many studies have been conducted on tornadic storms in the Yangtze–Huaihe region in China, which has the highest frequency of severe tornadoes in the country. For example, based on analyses of multiple cases in this region, Wu et al. [23] and Wang et al. [24] reported that mid-level mesocyclones were more frequently associated with hail production, whereas low-level mesocyclones were more often linked to tornado occurrence.

On 23 June 2016, a supercell storm formed and swept across the Yangtze–Huaihe region, producing two tornadoes in Funing County (FN) and Sheyang County (SY) of Yancheng City, Jiangsu Province, China, and causing the most severe tornadic damage and loss since 1977 in the country. Damage surveys show that the tornado in FN reached EF4 intensity. Detailed records of the supercell storm, including meteorological background, full development of tornadoes, and damage surveys, can be found in Xue et al. [25], Zhang et al. [26] and Meng et al. [27]. Several numerical studies were performed on this tornado case. Sun et al. [28] investigated the resolution dependence of simulated tornado structures, while Huang and Xue [29,30,31] performed a series of dynamical diagnostics that revealed the roles of unstable vortex Rossby waves, vortex rivers, and related dynamical processes.

Previous work by Meng et al. [27] briefly discussed a possible occlusion process between two tornadoes within the Yancheng supercell, primarily inferred from a backward-moving tornadic vortex signature (TVS) relative to the storm’s main inflow during the dissipation stage of the first tornado. However, given the rarity of cyclic supercells in China and their association with the most severe convective events, the cyclic behavior of the Yancheng storm is worth further investigation.

Compared with earlier studies that mainly focus on the first EF4 tornado, this work examines the full life cycle of the supercell, including the cyclic evolution of two tornadic mesocyclones followed by a subsequent non-tornadic mesocyclone. The study documents more detailed and comprehensive radar-based observational evidence of cyclic mesocyclones, including the evolution of hook-echo geometry and associated inferred flow patterns, and the temporal association between mesocyclogenesis and reflectivity core evolution. However, due to the limitations of single-Doppler radar observations, the interpretations presented in this study are partly based on existing conceptual models of cyclic supercells. The findings of this study may improve understanding of the cyclic behavior of the Yancheng storm and could provide useful reference information for operational warning practices in the region.

The remainder of the paper is organized as follows: Section 2 provides an overview of the supercell storm; Section 3 discusses the cyclic tornadic mesocyclone occlusion-genesis processes; Section 4 compiles statistical data on characteristics of the mesocyclones and tornadoes; and Section 5 presents a summary of the study.

2. Event and Data

2.1. Synoptic Background

At 1400 local standard time (LST; UTC+8) on 23 June 2016, about 11 min before the first tornado struck FN in Yancheng City, a cold-core low was situated in northeastern China at 500 hPa, while the subtropical high was located over the northwest Pacific (Figure 1a). The cold air from the northwest of the cold-core low converged with moist warm air along the western edge of the subtropical high over the Huai River Basin in eastern China, creating favorable conditions for deep convection. At the surface, a strengthened surface low was located to the west of Yancheng City, and the city was in the warm sector behind the warm front (Figure 1c). Notably, a west–east alley of strong low-level wind shear, indicated by the weak surface winds and relatively strong southwesterly flow at 850 hPa (yellow boxes in Figure 1b,c), was located just south of FN. The maximum convective available potential energy estimated from the radiosondes near Yancheng City at 1400 LST ranged from 2991 J kg⁻¹ [27] to 3290 J kg⁻¹ [26]. The lifting condensation level was as low as 984.6 hPa, about 240 m above sea level [26]. The storm-relative helicity was 139 m² s⁻² for 0–1 km layer and 170 m² s⁻² for 0–3 km layer [26]. Overall, the combination of strong instability, adequate low-level moisture, and enhanced vertical wind shear provided a synoptic environment highly conducive to tornadic storm development. More detailed synoptic and mesoscale analyses can be found in references [26,27] and are, therefore, not repeated here.

2.2. The Supercell Storm

At 0855 LST, a storm developed in northern Anhui Province, about 450 km west of Yancheng City, and moved eastward. At 1003 LST, the storm entered Jiangsu Province and turned toward the east–northeast (Figure 2). After 1158 LST, a mesocyclone appeared in the storm for the first time, and the storm gradually organized and intensified as several cells merged. At 1408 LST, the storm developed a hook echo. Shortly after that, according to damage surveys, an EF4 tornado struck FN at about 1410 LST, and lasted until 1500 LST. About 10 min later, another tornado hit SY from 1510 to 1530 LST. The supercell storm moved over the sea at about 1607 LST and subsequently degraded into a non-supercell storm. During the lifetime of the storm, the hook echo formed cyclically three times at about 1408, 1511, and 1545 LST.

2.3. Radar Data

This study used CINRAD Doppler radar data from Yancheng station, which was about 40–65 km south–southeast of the supercell storm during its tornadic period. At the lowest elevation angle, the radar beam height ranged from 0.68–1.24 km. The radar used volume coverage pattern 21, which consisted of nine elevation angles ranging from 0.5° to 19.5°. At 0.5° and 1.5° elevations, pulse repetition frequencies were 322 Hz and 1014 Hz for reflectivity and radial velocity, with maximum ranges of 460 km and 230 km, respectively. The update interval for a single volume scan was approximately 6 min. Velocity dealiasing, removal of ground clutter and non-meteorological echoes, and mitigation of range folding were performed within the CINRAD system, followed by manual review and correction.

Detection of mesocyclone signatures and tornadic vortex signatures (TVSs) was based on the CINRAD system, which uses the same algorithms as those of the WSR-88D system. For TVS detection, the minimum gate-to-gate azimuthal radial velocity difference (DV) was set to 25 m s⁻¹ at the detection base and 36 m s⁻¹ anywhere within the detection, with a detection height of 8 km and a minimum depth of 1.5 km [32]. Only at 1511 LST was the base DV threshold temporarily reduced to 20 m s⁻¹, which remains a reasonable threshold given that a tornado was identified in the damage survey and that the same threshold was applied by Meng et al. [27] throughout the case for TVS detection.

The CINRAD system uses two sets of thresholds for mesocyclone detection, one of high angular momentum (540 km² h⁻¹) and low wind shear (7.2 h⁻¹), and the other with low angular momentum (180 km² h⁻¹) with high wind shear (14.4 h⁻¹) [33]. A vortex exceeding either set is identified as a mesocyclonic vortex. To track submesocyclone-strength vortices after mesocyclones weakened, additional mesocyclone detections were performed using reduced angular momentum thresholds: 300 km² h⁻¹ and 100 km² h⁻¹ at 1448 LST, 200 km² h⁻¹ and 90 km² h⁻¹ at 1454 LST, and 300 km² h⁻¹, 100 km² h⁻¹ at 1459 LST.

3. Cyclic Mesocyclogenesis

3.1. Overview

Figure 3 illustrates the evolution of mesocyclones within the storm from 1408 to 1556 LST, during which three cyclic mesocyclones were identified. The first mesocyclone (M1) was initially detected at 1156 LST by another CINRAD located approximately 100 km west of the Yancheng radar. At 1411 LST, an EF4 tornado struck FN and persisted until 1500 LST. From 1442 to 1459 LST, M1 weakened and occluded while a new mesocyclone (M2) formed to its east. M2 subsequently intensified and produced a tornado in SY from 1511 to 1528 LST. During 1516 to 1528 LST, M2 occluded as a third mesocyclone (M3) developed. M3 dissipated at 1556 LST as the storm moved offshore.

3.2. Transition Between M1 and M2

Figure 4 shows the reflectivity at 0.5° elevation, radial velocity at 0.5° and 3.4° elevation during the occlusion of M1 and formation of M2. At 1431 LST, M1’s maximum shear reached 94 × 10⁻³ s⁻¹. At 1436 LST, a new submesocyclone-strength vortex formed at 3.4° elevation, east of the hook echo, downstream of the rear-flank downdraft (RFD; Figure 4(c2)).

At 1442 LST, the hook echo wrapped around itself, and the updraft width narrowed, as indicated by the narrowing of the Weak Echo Region (WER, Figure 4(a3)). The new vortex had strengthened into a mesocyclone, M2, while M1 exhibited a distinct backward movement relative to the WER. According to Meng et al. [27], the tornado’s damage swath narrowed significantly after this moment. By 1448 LST, M1 had weakened to submesocyclone strength, while the hook echo merged with the storm’s main reflectivity core and became undiscernible. However, the progress of occlusion slowed subsequently.

From 1448 to 1459 LST, the storm updraft widened, as indicated by a gradually expanding WER. Rather than propagating rearward, M1 remained relatively stationary near the hook-echo root, in close proximity to the WER. The WER also produced a small notch in the main reflectivity core toward M1 at 1.5° and 2.5° elevation (Figure 5), suggesting that the updraft continued to partially feed M1. This may reflect a transient rebalancing between inflow and outflow during this period. M1 persisted until 1459 LST, consistent with the end of Tornado 1 at approximately 1500 LST based on damage surveys.

Another indication of this reorganization is the persistence of a cyclonic–anticyclonic vortex pair on either side of the RFD at 0.5° elevation during the same period, suggesting that the RFD maintained a relatively steady orientation, with a possible tendency toward the northeast (Figure 4(b4–b6)). During this period, the RFD reflectivity became elongated northwest–southeast within the confluence zone between the RFD and the inflow. Notably, it evolved into a narrow tendril-shaped echo at 1459 LST (Figure 4(b6)), closely resembling the conceptual model of Beck et al. [8].

At 1505 LST, the RFD shifted towards a more southward direction (Figure 4(b7)), facilitating reformation of a new hook echo, and increased reflectivity also appeared at the forward-flank downdraft (FFD; Figure 4(a7)). At 1511 LST, the new hook echo had fully reformed, and tornado 2 struck SY.

3.3. Transition Between M2 and M3

The occlusion of M2 and the genesis of M3 occurred from 1516 to 1545 LST (Figure 6). At 1511 LST, Tornado 2 formed at the tip of the hook echo (Figure 4(a8)). At 1516 LST, two mesocyclones coexisted: one to the north in the FFD and the other to the south in the WER (Figure 6(a1)). At 1.5° (or 2.4°) elevation, the TVS radial velocity couplet attached to the northern mesocyclone (Figure 7a) and moved with it in subsequent volume scans (Figure 7b–d). Therefore, it is plausible that M2 moved northward towards the FFD, consistent with an intensifying inflow during this period. Meanwhile, a new mesocyclone, M3, formed in the WER downstream of the RFD, near the original location of M2.

From 1516 to 1528 LST, M2 moved westward along the FFD and reached the “pocket” between the FFD and RFD, weakening to submesocyclone strength (Figure 6(a1–a3)). In terms of their relative positions, the mesocyclones rotated counterclockwise about each other during this period, with their relative displacement exhibiting an average angular velocity of 4.1° min⁻¹. M2 later dissipated at 1533 LST.

The hook echo deformed and reformed from 1516 to 1545 LST. It extended beyond the TVS at 1522 LST (Figure 6(a2)) and deformed north–southward at 1528 LST (Figure 6(a3)), assumed a recurved hook shape at 1533 LST (Figure 6(a4)), and reformed into a new hook echo at 1545 LST (Figure 6(a5)). The recurved hook echo and the mutually rotating mesocyclones in this case show similarities to the cyclic mesocyclones reported by French et al. [9].

At 1528 LST, a north–south-oriented appendage developed ahead of the hook echo (Figure 8(a1)), and a possible cyclonic–anticyclonic vortex pair was collocated with the appendage (Figure 8(b1,c1)). However, the anticyclonic vortex on the southern portion of the appendage was only detectable above the 2.4° elevation angle because lower-level observations were affected by range folding; its existence should therefore be interpreted with caution (Figure 8(c1)). The appendage resembles the “hammerhead”-shaped echo documented by Bluestein et al. [34] and Houser et al. [15], suggesting that an eastward-moving RFD diverged upon encountering the inflow. The northern branch of the RFD may have partially disrupted the inflow to M2 but had not yet completed the occlusion.

By 1533 LST, the “hammerhead” echo had intensified and reconnected with the main reflectivity core via a west–east high-reflectivity band (Figure 8(a2)), indicating an intense reflectivity surge. A convergence line in the radial velocity field collocated with the strong precipitation surge (Figure 8(b2)) suggests that a northward moving rear-flank gust front was converging with the FFD, closing the “pocket” that contained M2 and completing the occlusion. Consequently, M2 had dissipated by this time.

3.4. Cyclic Evolution of Storm Intensity

The storm-scale reflectivity core intensity was quantified by first remapping the radar data onto a three-dimensional Cartesian grid with a resolution of 250 m using the Python ARM Radar Toolkit version 2.2.1 [35], and then counting the number of grid points exceeding 60 dBZ within the entire storm (Sensitivity tests were performed using a 500 m grid spacing and a 55 dBZ threshold, and the results remained consistent).

Figure 9 illustrates the evolution of the storm-wide reflectivity core intensity from 1328 to 1630 LST, together with the 50 dBZ echo-top height and the lifetimes of the mesocyclones and tornadoes. During this period, the storm underwent three distinct intensity cycles that exhibited a temporal correspondence with the mesocyclone lifecycles. In each cycle, the storm initially intensified, as indicated by an enhanced reflectivity core and increasing echo-top height, and subsequently weakened, accompanied by decreases in both quantities. This cyclic behavior ceased after 1556 LST, when the storm moved offshore and the sharp decrease in low-level vertical wind shear (Figure 1b,c) likely contributed to the dissipation of M3. Notably, the reflectivity core experienced rapid weakening and lowering toward the surface at 1408–1419 LST, 1448–1459 LST and 1545–1601 LST. The genesis of M2 and M3 occurred with similar time lags relative to the weakening and lowering phase of the reflectivity core: approximately 24 min for M2 and 18 min for M3.

4. Statistics of Mesocyclones and TVSs

4.1. Mesocyclone and TVS Features

Figure 10a shows the top and bottom heights of the three mesocyclones, along with the intensity and height of maximum shear. Due to range folding at the 0.5° elevation angle, the bottom-height estimates may be biased at 1408 LST and throughout the lifetimes of M2 and M3. From 1408 to 1459 LST (Tornado 1), M1 exhibited an average maximum shear of approximately 61.2 × 10⁻³ s⁻¹, which is more than four times higher than the mean value of 14.4 × 10⁻³ s⁻¹ reported for six tornadic supercell storms (EF1–EF3) in the Yangtze–Huaihe region from 2003 to 2010, as documented by Zhou et al. [36]. Figure 10b illustrates the evolution of the top, bottom, and maximum shear heights of the TVSs, together with the maximum shear values summarized in the accompanying table. During Tornado 1 (1414–1454 LST), the TVSs exhibited an average maximum shear of 73.1 × 10⁻³ s⁻¹, which is considerably higher than the mean value of 30.8 × 10⁻³ s⁻¹ reported by Zhou et al. [36].

4.2. Maximum Wind Speed Estimation for Tornadoes

At present, mobile radars are regarded as among the most reliable instruments for measuring wind speeds in tornadoes. Doppler On Wheels radar [37] observations and WSR-88D data were statistically compared by Toth et al. [38], who established a linear relationship between the maximum tornado wind speeds derived from the two systems. The relationship yielded a coefficient of determination (R²) of 0.83 across 14 tornadic events. Given the close similarity between the CINRAD and WSR-88D radar, their measurements are treated as broadly comparable in this study. Accordingly, the same relationship is applied to convert CINRAD observations into equivalent DOW maximum velocities, providing a first-order estimate of tornado wind speeds. The conversion is expressed as:

$$v_{\mathrm{D}\mathrm{O}\mathrm{W}}=1.4v_{\mathrm{C}\mathrm{I}\mathrm{N}\mathrm{R}\mathrm{A}\mathrm{D}}+0.4$$

(1)

where $v_{\mathrm{D}\mathrm{O}\mathrm{W}}$ and $v_{\mathrm{C}\mathrm{I}\mathrm{N}\mathrm{R}\mathrm{A}\mathrm{D}}$ denote the maximum wind speeds observed by DOW and CINRAD, respectively. The $v_{\mathrm{C}\mathrm{I}\mathrm{N}\mathrm{R}\mathrm{A}\mathrm{D}}$ term is estimated as:

$$v_{\mathrm{C}\mathrm{I}\mathrm{N}\mathrm{R}\mathrm{A}\mathrm{D}}={\displaystyle \frac{1}{2}}\mathrm{L}\mathrm{L}\mathrm{D}\mathrm{V}+v_{\mathrm{s}}$$

(2)

where LLDV is the Low-Level Delta Velocity observed by the CINRAD system, and $v_{\mathrm{s}}$ is the storm-motion speed estimated from the temporal displacement of storm centroids identified by the storm cell identification and tracking algorithm of the CINRAD system, which is the same as that used in WSR-88D [39]. It should be noted that the storm-motion estimate may contain substantial uncertainty during periods of rapid storm-cell evolution, particularly during 1437–1442 and 1511–1516 LST in this case.

The estimated $v_{DOW}$ for Tornado 1 reached EF4 intensity (

Table 1. Estimated DOW maximum wind speed (m s⁻¹).

Tornado 1
Time (LST)	LLDV (m s−1)	vs (m s−1)	vDOW (m s−1)	EF scale
1414	36.5	16.6	49.2	1
1419	80.5	14.5	77.1	4
1425	77.5	18.9	81.1	4
1431	77.5	17.6	79.3	4
1437	84.5	18.3	85.2	4
1442	74	3.9	57.7	3
1448	65.5	11.5	62.4	4
1454	49	12.2	51.8	4
Tornado 2
Time (LST)	LLDV (m s−1)	vs (m s−1)	vDOW (m s−1)	EF scale
1511	20	10.2	28.7	--
1516	44.5	23.9	65	--
1522	51.5	10.2	50.7	--
1528	51	12.1	53	--

Bold text denotes the time associated with rapid storm cell evolution.

), which is broadly consistent with the available damage survey results. The estimated $v_{DOW}$ for Tornado 2 generally corresponded to EF0–EF2 intensity. Although the estimate briefly reached EF3 intensity at 1516 LST, this result is subject to considerable uncertainty in storm-motion estimation because of the rapid evolution of storm reflectivity during that period. No detailed damage survey has been reported in the literature to independently evaluate this estimate.

The estimated $v_{\mathrm{D}\mathrm{O}\mathrm{W}}$ values should be regarded as rough radar-based estimates and interpreted with caution. Uncertainties may arise from intrinsic errors in the CINRAD system, the application of a WSR-88D-based model to CINRAD data, and errors associated with the conversion model itself. Moreover, even direct DOW observations are not equivalent to EF-scale intensity estimates [40].

5. Summary

On 23 June 2016, a supercell storm swept eastward across the northern part of Jiangsu Province, China. The storm produced two tornadoes in FN and SY of Yancheng City, Jiangsu Province. The FN tornado reached EF4 intensity and caused the most severe damage and losses by one tornado in China since 1977. This study, based on data from an S-band stationary CINRAD, focuses on the cyclic occlusion and formation of mesocyclones within the storm.

Within the supercell, three mesocyclone life cycles were identified. The cyclic evolution resembled that of classical conceptual models: the old mesocyclones moved rearward relative to the storm inflow, new mesocyclones developed at mid-levels downstream of the RFD, and the hook echo deformed during occlusion.

The first occlusion was relatively prolonged. After the hook echo wrapped itself, submesocyclone-strength vortices associated with the decaying mesocyclone persisted for three additional volume scans (18 min), suggesting sustained inflow support. In contrast, the second occlusion evolved more rapidly, during which the old and new mesocyclones rotated cyclonically around each other.

The storm intensity exhibited a cyclic evolution. A temporal association was observed between the evolution of the reflectivity core and the occlusion and formation of mesocyclones. The new mesocyclone formed approximately 24 min after the first instance of reflectivity-core weakening and lowering toward the surface and 18 min after the second.

Due to the limitations of single, fixed-site Doppler radar observations, this study is subject to several constraints, including coarse temporal and spatial resolution, the absence of dual-Doppler wind retrievals, and a lack of polarimetric analyses. These limitations highlight the need for multiple, rapid-scan polarimetric radar systems in future investigations of cyclic storms in the region. In addition, comparison of environmental conditions associated with cyclic tornadic storms and non-cyclic events would be a valuable direction for future research and may provide reference information for operational warning applications.