Characteristics of an Explosive Cyclone over Northeast China Revealed by Satellite Water Vapor Imagery

Extratropical cyclone is one of the major meteorological hazards affecting North China, often causing severe weather conditions such as strong windstorms, sand and dust storms, and heavy snowstorms. The rapidly deepening cyclone was first noted by Bergeron in 1954 ^[1]. Sanders and Gyakum firstly defined the explosive cyclone (EC, also known as "bomb"in the literature) as a cyclone with the central sea level pressure decreasing over 24 hPa within 24 h after the cyclone center moves to 60°N ^[2]. Compared with normal extratropical cyclones, the horizontal range of ECs is not significantly different, but the weather events related to ECs are much severer. The wind speed near these ECs can reach a maximum of 30 m s⁻¹, and ECs deepen much faster than normal extratropical cyclones (Bergeron ^[1]). Previous studies on extratropical cyclones focused on statistical analyses (Fu et al. ^[3]; Dacre ^[4]; Zhang et al. ^[5]; Zhang et al. ^[6]; Chen et al. ^[7]; Iwao et al. ^[8]), mechanism study (Kouroutzoglou et al. ^[9]; Hagay and Brenner ^[10]; Wang et al. ^[11]; Gao et al. ^[12]; Xiong et al. ^[13]) and diagnostic analyses (Gong et al. ^[14]; Jiang et al. ^[15]; Mary et al. ^[16]; Strahl and Smith ^[17]; Wang et al. ^[18]; Chu and Zhang ^[19]; Sun et al. ^[20]; Martin and Otkin ^[21]).

Fu et al. summarized the research status of ECs systematically ^[22]. Sun et al. analyzed the statistical characteristics of ECs over the North Atlantic Ocean during the cold seasons in 2000-2015 ^[23]. Zhang et al. indicated that in East Asia, ECs are active over the eastern coast of China, Japan, and the Northwest Pacific, but they rarely occur inland ^[24]. Furthermore, winter, spring, and autumn are the favourable seasons for ECs. By analyzing the extratropical cyclones during 1958-2001, Wang et al. found that the frequency of cyclone events in East Asia has significantly decreased in midlatitudes, but has increased in high-latitudes, and global warming has resulted in a northward shift of storm tracks ^[25]. Zhang and Fu investigated the structures and evolutions of ECs over the Northwest and Northeast Pacific, and found that the strong baroclinicity, abundant latent heat release at low levels, and strong upper-level forcing can provide a favourable environment for the generation of ECs over the Northwest Pacific ^[26].

Dry intrusion, coming down from high levels near the tropopause and characterised by high potential vorticity (PV) and low wet-bulb potential temperature (θ_w), is considered to be significant in accelerating the development of cyclones (Browning and Golding ^[27]; Browning ^[28]). Yao et al. explored the impacts of dry intrusion upon the Meiyu front, and described the dry intrusion as that the high PV descends from the upper levels to lower levels ^[29]. Therefore, the dry intrusion can be used as an important indicator of the high PV forcing. Lu and Sun ^[30] and Huang et al. ^[31] proved that the explosive development of EC is closely related to the upper-level jet stream (ULJ) and the downward extension of high PV zone by studying some cases of ECs. Zha et al. emphasized that the downward extension of high PV zone and the latent heat-releasing caused by the warm-wet jet stream can cause the explosive development of cyclones ^[32]. Zhu et al. diagnostically analyzed an extratropical cyclone by using the extended Zwack-Okossi equation, and found that the latent heat release and the horizontal warm-air advection are main forcing terms at the early developing stage, while the horizontal absolute vorticity advection is the main forcing later ^[33]. Miao et al. suggested that the jet stream could accelerate the development of cyclones in both thermal and dynamic aspects ^[34]. Xiong et al. found that the back-bent warm front and warm seclusion may exist in a fully developed extratropical cyclone over land, and they applied quasi-geostrophic theory to explain the anomalous development and movement of a cyclone ^[35-36]. Miao et al. carried out a comparative study of two types of Yellow River cyclones passing over the Bohai Sea in spring and summer by analyzing factors including ULJ, energy evolution, wet PV, and sea surface ^[37]. Sun et al. found that cyclones across the sea would reduce energy loss and provide extra sensible heat, radiation, and latent heat ^[38].

As previous findings have suggested, dry intrusion is significantly related to the cyclone development, and it could be directly observed in water vapor (WV) imagery (Yao et al. ^[29]; Huang et al. ^[31]; Xiong et al. ^[36]; Olander and Velden ^[39]). Dry intrusion and high PV, which are associated with rapidly deepening cyclones has been stressed by Young et al. ^[40]. Georgiev has shown that model PV fields near the tropopause can be checked against the satellite images of WV ^[41]. Meanwhile, a lot of information related to the physical quantities of large-scale weather processes can be detected on WV imagery (Weldon and Holmes ^[42]), such as vertical motion (Kouroutzoglou et al. ^[43]). Nevertheless, the application of WV imagery in weather forecast and meteorological research still needs further popularity. ECs occur frequently on the surface of the North Pacific and the east and west coasts of the North Atlantic during cold seasons (October-April) (Bluestein ^[44]). Explosive cyclones with intensity above 2.0 Bergeron (B) mostly occur from December to March of the following year. ECs seldom occur in May and are usually of low intensity, so the related research is rare. However, during May 2-4, 2016 (GMT + 08, same below), an EC occurred in the Yellow Sea and Bohai Sea area of China. Its central pressure decreased by 17.5 hPa in 12h, its horizontal scale was over 2000 km, and it caused a very rare and extreme spring rainstorm in Northeast China. Therefore, this study analyzes the evolution and development process of this EC, and identifies important weather areas through water vapor images, vertical motions, and the ULJ, to provide reference for the forecast of such cyclones.

The data used in this study included conventional ground observational data, National Centers for Environmental Prediction (NCEP) reanalysis data (horizontal resolution 1°×1°, four times per day) and 6.7 mm water vapor imagery from the FY-2E meteorological satellite (30min temporal resolution).

In this paper, 12-h sea level pressure (SLP) change is analyzed to find an instance of the most rapid deepening in a cyclone's life. Therefore, the definition of EC given by Sanders and Gyakum is revised to be a cyclone whose central SLP decrease normalized at 60°N is greater than 12 hPa within 12 h ^[2]. The deepening rate of cyclone SLP is calculated with the following formula:

Deepening rate of cyclone SLP (hPa h^-1) = $\left(\frac{P_{t-6}-P_{t+6}}{12}\right) \times\left(\frac{\sin 60^{\circ}}{\sin \frac{\varnothing_{t-6}+\varnothing_{t+6}}{2}}\right) $

where t is time in hours, P is the central SLP, and $ \varnothing$ is the latitude of the cyclone center. Subscripts"t-6"and "t+6"represent variables of 6h before and after present time t, respectively.

In a free atmosphere, the PV equation in isobaric coordinates (P coordinates) is as follows (Yin et al. ^[45]):

where g is the acceleration of gravity, θ is the potential temperature, p is the pressure, ζ is the relative vorticity, and f is the Coriolis parameter. The units commonly used for the presentation of PV are 10^-6m²s^-1 Kkg^-1, termed as the PV unit (PVU). The region where the PV is greater than 1.5 PVU is defined as the high PV region.

During May 1-5, 2016, a severe rainstorm caused by an EC hit Northeast China, with the maximum amount of precipitation exceeding 100 mm and a strong wind over Force 7. The cyclone generated in the lower reaches of the Yellow River in Shandong Province on the afternoon of May 2 (Fig. 1a), passed through the Yellow Sea and moved northeastward along the boundary of Northeast China. After entering the Yellow Sea area at night, it strengthened rapidly and its central pressure dropped sharply, accompanied by an intense rainfall event. On May 3, the wind speed near the cyclone increased significantly, reaching a peak of more than 30 m s⁻¹ on the Liaodong Peninsula. It moved out of China from the eastern Heilongjiang Province on the night of May 4. The central pressure of the cyclone started to drop since 14:00 on May 2, reached 975.7 hPa at 14:00 on May 3, and then rose slowly. From 20:00 on May 2 to 08: 00 on May 3, the cyclone developed explosively, with deepening rate larger than 1B. Therefore, in this study, it is assumed that the period from 20:00 on May 2 to 08:00 on May 3 is the explosion stage of the cyclone, the period from 20:00 on May 1 to 14: 00 on May 2 is the pre-explosion stage, and the period from 20:00 on May 3 to 14:00 on May 4 is the post-explosion stage.

Figure 1.  (a) The track of the cyclone center from 14:00 (GMT+8) on May 2 to 14:00 on May 4, 2016. The numbers represent the time, e.g. 0214 represents 14:00 on May 2. (b) The amount of precipitation from 08:00 on May 2 to 08:00 on May 5 (units: mm).

Figure 2.  Temporal evolution of sea level pressure (units: hPa) and its deepening rate (units: Bergeron) in the center of the cyclone from May 1 to May 4, 2016.

Usually, a favorable vertical and horizontal atmospheric circulation is required for the explosive development of ECs. At 08: 00 on May 1, there was a longwave trough moving eastward near 40° N at 200 hPa. On both sides of the trough, there were two ULJ streams gradually merging with the subtropical westerly jet. At the explosion stage (Fig. 3), the longwave trough at 200 hPa moved eastward to the Bohai Sea, showing a northwest-southeast direction. The surface cyclone, controlled by the positive vorticity advection, was located at the right-back of the entrance area of the southwest jet in front of the trough, the left-exit region of the westerly subtropical jet, and the shunt area of the ULJ streams. At 20: 00 on May 3, the surface cyclone stopped developing as the ULJ streams gradually moved away. At 500 hPa, correspondingly, a cold trough progressed eastward from the Hexi Corridor on May 1, and the thermal trough was lagged behind the height trough. The cold trough was deepened gradually during the eastward movement. At the explosion stage, the cold trough at 500 hPa evolved into a low vortex over southern Liaoning Province, and meanwhile the surface cyclone was located at the east side of the vortex. After 8: 00 on May 3, the vortex tended to coincide with the surface cyclone center, and the vorticity advection over the cyclone disappeared without deepening.

Figure 3.  Geopotential height and wind field for the EC event at 08:00 on May 3, 2016 at (a) 200 hPa, and (b) 500 hPa and 850 hPa. Blue contours represent geopotential height (units: dagpm); wind barbs represent wind speed (units: m s⁻¹; a full barb represented 4 m s⁻¹, same below) and direction; shading represents the jet streams where wind speed exceeds 30 m s⁻¹ in (a), and shading represents sea level pressure (units: hPa) in (b).

Corresponding to the 500 hPa cold trough, there was a frontal zone at 850 hPa. At 08:00 on May 2, a low vortex formed and strengthened rapidly, and the warm and cold advections reached their peak values. Correspondingly, the surface cyclone deepened and frontogenesis proceeded to form a warm front when the cold front moved to the lower reaches of the Yellow River. After the cyclone entered the sea, the friction of the underlying surface decreased obviously. Previous studies have shown that there is little seasonal change in the dynamic effects of the sea surface (Miao et al. ^[37]; Yin et al. ^[45]). As a result, the angle between the isobaric and the wind direction near the cyclone center became very small, even approaching 0, and the wind speed increased significantly to approach the geostrophic wind gradually (figure omitted). While the cyclone progressed northeastward, it strengthened rapidly, and the horizontal scale extended to be more than 2000 km (Fig. 3b). The isobars around the cyclone were very dense, with the central pressure lower than 980 hPa, accompanied by the strong rainstorm.

To sum up, the left-exit region of the 200-hPa jet was conducive to upper-level divergence. The continuously deepening cold troughs at 500 hPa strengthened the positive vorticity advection (PVA) and cold advection, which jointly provided favorable conditions for the explosive development of the surface cyclone. Moreover, compared to the land surface, the lower friction of sea surface also contributed to the EC development.

WV, as a tracer, can present atmospheric motion when there is no cloud. Due to the continuity of the WV, the boundaries on the WV imagery are clearer and smoother than those on the satellite infrared imagery. Moreover, the information on the WV imagery is easy to understand, where the bright white zone represents WV or cloud and the dark zone reflects the dry zone of the troposphere. Humidity changes in the troposphere are related to the ascending or descending motions, so the changes of brightness on the WV imagery can reflect vertical motions of the airflow.

Closely related to the distribution of atmospheric vorticity, the ULJ is a key dynamic factor to induce the explosive development of the cyclone (Lu and Sun ^[30]; Huang et al. ^[31]; Yin et al. ^[45]). The vorticity advection can directly determine the distribution of divergence and vertical motions (Tao et al. ^[46]), and it is one of the main forces of the cyclone development. The PV theory is very effective in explaining the development of synopticscale weather systems in the mid-latitude regions, providing a concise view of atmospheric dynamics for weather analysis (Zhou et al. ^[47]). PV can be conserved under adiabatic and frictionless conditions. Therefore, the evolution of weather systems is influenced by the stretching and compression of the PV in the atmosphere. As a result, tracking PV anomalies is a good way to investigate the evolution of atmospheric disturbances. The downdraft from the high PV region in the stratosphere is a crucial trigger for the explosive development of the cyclone (Browning and Golding ^[27]; Browning ^[28]; Heo et al. ^[48]; Pang and Fu ^[49]; Pang et al. ^[50]). The evolution of the weather system on WV imagery and the related dynamic factors at different stages of this EC event are analyzed in detail in the following parapraghs.

At 20: 00 on May 1, there was a longwave trough and three jet centers with the northwesterly wind, westerly wind, and southwesterly wind at the west, south, and east sides of the trough, respectively. The speed of the strongest wind in the jet center exceeded 55 m s⁻¹ on the south of the trough. There was an extensive and bright baroclinic leaf cloud ahead of the trough in the northeast-southwest direction (the black arrow in Fig. 4a), and behind it was the dark zone (the white arrow in Fig. 4a), which was related to the 500-hPa descending motion behind the trough. To the south of the Hetao area, the dark zone presented a triangular shape at 14: 00 on May 2. As the northwesterly and westerly jet, accompanied by the long-wave trough, moved eastward rapidly, the triangular dark zone accelerated, moved eastward, and became darker. As a result, the cloud ahead of the trough adjusted to lie in a north-south direction. It showed a more typical S-shaped baroclinic leaf cloud in Fig. 4b with clear convex (the black arrow in Fig. 4b) and concave (the white arrow in Fig. 4b) leading edges. It indicates that the cyclone would develop near the concave edge and the dark zone would expand southwestward.

Figure 4.  WV imagery of FY-2E 6.7 mm channel (units: K; wind barbs: 200-hPa jet where the wind speed exceeds 30 m s⁻¹; contours: 500-hPa descending motion regions, units: 10^-2 Pa s⁻¹; white arrow: dry zone; black arrow: water vapor plume) at (a) 20:00 on May 1 and (b) 14:00 on May 2. Vertical cross-section plot of horizontal divergence (rainbow contours, units: 10⁻⁶ s⁻¹) and flow field (black streamlines, units of v: m s⁻¹, and units of ω: 10^-2 Pa s⁻¹) along the yellow line in (a) and (b) (the triangle on the x-axis represents the latitude of the cyclone center) at (c) 20:00 on May 1 and (d) 14:00 on May 2. Vertical cross-section plot of relative humidity (filled contours, units: %), PV (red contours, units: 10^-6 m² s⁻¹ K kg⁻¹), and PT (black contours, units: K) along the yellow long dashed line in (a) and (b) (the triangle on the x-axis represents the longitude of the cyclone center) at (e) 20:00 on May 1 and (f) 14:00 on May 2.

On the east side of the cyclone center, a southerly low-level jet (LLJ) formed with the wind speed exceeding 20 m s⁻¹ at 850 hPa (figure omitted). The cyclone center at the right side of the entrance zone of the southwesterly ULJ and the left side of the 850-hPa southerly LLJ, which was a favorable position for its development. By analyzing the horizontal divergence and vertical motions before the explosion moment (Fig. 4c and 4d), we found that the divergence field was weak, although a relatively strong updraft appeared below 200 hPa over the cyclone. On the south of the cyclone, the streamlines near 32° N were denser and extended up to 150 hPa, indicating that the convergent updraft was stronger. At 14:00 on May 2 (Fig. 4d), a divergent center of 4 × 10⁻⁵ s⁻¹ appeared around 300 hPa. The low-level convergence below 500 hPa also exceeded −6 × 10⁻⁵ s⁻¹, mainly corresponding to the strong LLJ. It means the updraft center was not yet above the cyclone at that moment.

At 20:00 on May 1 (Fig. 4e), the isentropes below 200 hPa and above the surface low were relatively evenly distributed. The gradient of the isentropes was not steep, and the high PV zone was far from the cyclone. At 14:00 on May 2 (Fig. 4f), the high PV zone extended downward to around 400 hPa and moved eastward rapidly to about 116 ° E, approaching the cyclone. The isentropes above the cyclone center were mainly distributed horizontally. PT increased with height $ \left(-\frac{\partial \theta}{\partial p}>0\right)$, suggesting a stable atmosphere. It should be noted that a small range of high PV began to appear near the surface cyclone.

At 20: 00 on May 2 (Fig. 5a), the convex edge of the S-shaped baroclinic leaf cloud further developed westward and gradually evolved into a comma cloud system. The dark zone had expanded significantly and continued to extend to the southeast quadrant of the surface cyclone. At 08:00 on May 3 (Fig. 5b), the leafshaped cloud developed into the shape of a full comma, and the dry zone expanded from the southeast to the northeast of the comma cloud near the cyclone center. At the head of the comma cloud system, there was a hook cloud rotating around the cyclone center. As the dry zone invaded into the comma cloud, the comma cloud gradually transformed into a spiral cloud (figure omitted).

Figure 5.  WV imagery of FY-2E 6.7 mm channel (units: K; wind barbs: 200-hPa jet where the wind speed exceeds 30 m s⁻¹; contours: 500-hPa descending motion regions, units: 10^-2 Pa s⁻¹; white arrow: dry zone; black arrow: water vapor plume) at (a) 20:00 on May 2 and (b) 08:00 on May 3. Meridional vertical cross-section of horizontal divergence (rainbow contours, units: 10^-6 s^-1) and flow field (black streamlines, units of v: m s⁻¹, and units of ω: 10^-2 Pa s⁻¹) along the yellow line in (a) and (b) (the triangle on the x-axis represents the latitude of the cyclone center) at (c) 20:00 on May 2 and (d) 08:00 on May 3. Meridional vertical cross-section of relative humidity (filled contours, units: %), PV (red contours, units: 10^-6m²s^-1 Kkg⁻¹) and PT (black contours, units: K) along the yellow long dashed line in (a) and (b) (the triangle on the x-axis represents the longitude of the cyclone center), at (e) 20:00 on May 2 and (f) 08:00 on May 3.

From 20: 00 on May 2 to 14: 00 on May 3, the central sea level pressure of the cyclone continued to decrease and the isobaric contours became denser (figure omitted). The surface cyclone, once located in the right-entrance region of the southwesterly ULJ on the north, gradually shifted to the left-exit region of the westerly-southwesterly ULJ on the south (Fig. 5d). The ULJ expanded downward to around 700 hPa (figure omitted). In general, the cyclonic shear with positive relative vorticity was on the left side of the ULJ axis, so the largest PVA appeared at the left-exit zone of the jet. However, the anticyclonic shear with negative relative vorticity was on the right side of the ULJ axis, so the largest PVA appeared at the right-entrance zone of the jet as well. PVA on high levels can create divergent circulations on high levels. As a result of the mass adjustment, there was a pressure reduction below it, which was beneficial to the cyclone development. Meanwhile, the LLJ at 850 hPa expanded and strengthened, with the central wind speed exceeding 30 m s⁻¹. The cyclone was located at the front-left of the LLJ core, which was of significant convergence. The configuration of the high - and low-level systems was conducive to the cyclone development.

During the explosive development, the streamlines above the cyclone gradually became denser, and the airflow ascended obliquely from the surface to 200 hPa, indicating that the ascending motions were significantly strengthened. Fig. 5d shows a strong divergence at 250 hPa, the strongest divergence occurred around one latitude north of the cyclone, reaching the peak value of 10^-4 s^-1, and the divergence center was formed by the overlap of the right-rear entrance region of the westerly ULJ on the north and the left exit region of the southwesterly ULJ on the south. At 500 hPa, there was a convergence center reaching - 6 × 10^-5 s^-1. Below 700 hPa, there was a relatively weak convergence center and a divergence center. The convergence was more intense than the divergence. These two pairs of convergence and divergence centers maintained the strong updrafts and accelerated the explosive development of the surface cyclone. At this time, the strength and the horizontal scale of surface cyclone reached their peak, respectively.

At 20:00 on May 2 (Fig. 5e), the high humidity and high PV region from the stratosphere extended downward to around 600 hPa on the west side of the cyclone center; high PV began to expand near the surface cyclone. At 08:00 on May 3, the high PV region further spread eastward and extended downward (Fig. 5f). Below 200 hPa, a high PV center over 8 PVU appeared, gradually approaching the cyclone center from the west side. At the same time, a high PV region exceeding 1.5 PVU also appeared below 600 hPa and above the cyclone center, with the peak value over 3 PVU around 900 hPa. The high PV region can also be found at 2 longitudes east of the cyclone center at a low level. The high PV region in the low levels corresponded to the dense isentropes. As a result, the gradient of the isentropes above and below that region became smaller, indicating that the atmosphere tended to be unstable. Given the PV conservation (Tao et al. ^[46]), the absolute PV below it must increase and the cyclonic vorticity on the surface would increase as well. In conjunction with the tilted and dense isentropes, it provided a favorable condition for the increase of cyclonic vorticity. The denser isentropes on both sides of the cyclone center below 800 hPa indicated the strengthening of the frontal zone. To sum up, during the explosively developing stage of the EC, the high PV region in the upper troposphere extended downward to a lower altitude, while the isentropes in the lower troposphere became sparse, leading to a decrease of the atmospheric stability and an increase of absolute vorticity. Accordingly, the cyclone experienced its explosive development.

At this stage, both jets at 200 hPa moved eastward rapidly. The surface cyclone was not located at the leftexit zone of the ULJ. Consequently, the positive PV advection weakened, and the surface cyclone and the trough nearly coincided. LLJs were mainly located off the south and east sides of the cyclone center, and the convergence near the cyclone center weakened. The WV image dark zone became brighter than the previous stages, showing a gray zone (the white arrow in Fig. 6a) on the WV imagery. It rotated counterclockwise and was drawn into the bright white spiral cloud (the black arrow in Fig. 6b), indicating that the cold and dry air from upper levels had intruded into the cold vortex center. After that, the dark zone on the WV imagery gradually disappeared, and the surface cyclone stopped developing.

Figure 6.  WV imagery of FY-2E 6.7 mm channel (units: K; wind barbs: 200-hPa jet where the wind speed exceeds 30 m s⁻¹; contours: 500-hPa descending motion regions, units: 10^-2 Pa s⁻¹; white arrow: dry zone; black arrow: water vapor plume) at (a) 20:00 on May 3 and (b) 08:00 on May 4. Meridional vertical cross-section of horizontal divergence (rainbow contours, units: 10^-6 s^-1) and flow field (black streamlines, units of v: m s⁻¹, and units of ω: 10^-2 Pa s⁻¹) along the yellow line in (a) and (b) (the triangle on the x-axis represents the latitude of the cyclone center) at (c) 20:00 on May 3 and (d) 08:00 on May 4. Cross-section maps along the yellow long dashed line in (a) and (b) (the triangle on the x-axis represents the cyclone center) for relative humidity (filled contours, units: %), PV (red contours, units: 10^-6 m²s⁻¹ Kkg⁻¹), and PT (black contours, units: K) at (e) 20:00 on May 3 and (f) 08:00 on May 4.

From the flow field (Figs. 6c and 6d), above the cyclone center there were descending motions, while the updraft mainly appeared on the north of the cyclone beyond 45° N. The related high-level divergence weakened rapidly, but the weak convergence near the ground was maintained. The dynamic supports disappeared so that the cyclone tended to fill up. There were still low humidity and high PV regions in the upper layers of the troposphere (Figs. 6e and 6f), probably because the height of the troposphere decreases as the latitude increases. The high PV regions expanded down to about 700 hPa on the east side of the cyclone, and then disappeared in the lower layers of the troposphere. The distribution of isentropes became even, proving that the cold and warm fronts weakened or disappeared, and the cyclone gradually filled up.

As analyzed in Section 4, the patterns of the cloud on WV imagery varied from baroclinic leaf, comma to spiral, representing the evolution of the EC. In general, the WV imagery is analyzed based on the dry zone, moist zone, and the boundaries between them, especially the dry zone. However, the dry zone could be unclear sometimes. For instance, when the whole atmosphere is relatively dry, the WV imagery can be dim to some extent, leading to a misunderstanding of the crucial weather system. To identify the crucial dry zone on WV imagery, different types of dark zones on WV imagery are classified in the following part by analyzing the dynamic characteristics of the atmosphere.

At 20: 00 on May 1, the dark zone appeared in eastern Xinjiang Province behind the trough (the arrow in Fig. 4a), accompanied by a weak downdraft and the ULJ, ahead of which was the bright white cloud. The dynamic "dry band" signifies the dry intrusion development at the later stage. At 14: 00 on May 2, a triangular dark zone appeared at the west of Hetao area (arrow in Fig. 4b), and it is called the"dry delta", located ahead of the descending airflow, which was related to the jet break across the trough. The movement of the dry zone was closely related to the jet on the west side. At this moment, the surface cyclone was at the preexplosion stage.

As the ULJ center moved southward, the dark zone in deep black expanded from west to east at 20: 00 on May 2, (the arrow in Fig. 5a), forming an"expanding dry band"ahead of the ULJ and the downdraft. Then the surface cyclone began to develop. As descending airflow from the dry zone partly flowed into the surface low pressure at 08:00 on May 3, a dry slot was formed in the southwest of the low pressure (arrow in Fig. 5b). The dry slot continued to darken and expand towards the northeast side of the cyclone, and the cyclone developed explosively.

At 20:00 on May 3, the dry slot quickly expanded to higher levels upon the cyclonic circulation, spiraling around the cyclone center and forming a"dry spiral". At that time, the surface cyclone stopped deepening and development.

In summary, during the explosive development of the surface cyclone, the dry intrusion was obvious on the WV imagery. Its shape experienced the dynamic dry band, dry delta, expanding dry band, dry slot, and dry spiral. In this process, the descending motions in midlevels and the ULJ can always be found. This type of dynamic dry band may develop into dry intrusion at the later stage, which should be paid attention to on the WV imagery. Moreover, different shapes of dry zones on the WV imagery also have different forecasting indications. In this event, for instance, the dry delta appeared before the cyclone development, indicating that a favorable situation for the cyclone development was formed, so the dry delta could be considered as a signal of the cyclone development. The expanding dry band can mark the beginning of the cyclone development, and the dry slot signaled the key moment of the evolution of the dry intrusion, which could illustrate the explosive development of the cyclone. The spiral cloud implied the cyclonic flow in high altitude, and the surface cyclone stopped deepening at this stage. Overall, by jointly using the WV imagery, the wind field of high levels and the vertical motion plot, the dry zone could be efficiently analyzed.

In this study, an EC event occurred over Northeast China during May 2-4, 2016 has been investigated by analyzing the ULJ, vertical motions, PV, and the identification of dry zone on the WV imagery. The results can be concluded as follows.

EC events usually occurred in the left-exit region of the 200-hPa jet, and in the PVA region ahead of the 500-hPa trough. The decrease of the friction on the underlying surface also provided a favorable condition for the explosive development of the cyclone after it entered the sea.

At the pre-explosion stage of the EC event, with the ULJ moving eastward, the dry zone on the WV imagery spread into the south of the leaf cloud. It was characterized by the bright white S-shaped baroclinic leaf cloud. At that time, the divergence field above the cyclone was weak. The high PV region had just started to develop downward, showing a small-scale high PV zone in the lower layers near the cyclone. During the explosively developing stage, the S-shaped leaf cloud evolved into a comma cloud system, and the dry zone expanded to the north of the surface cyclone. The surface cyclone was located on the left-exit zone of the ULJ, where was of strong high-level divergence and updraft. The high PV region extended downward from the top of the troposphere to below 500 hPa, while the value of the PV around 900 hPa reached 3 PVU. The structure of the cold and warm fronts was clear at this stage. At the post-explosion stage, the dark zone brightened and disappeared as it was drawn into the spiral cloud. At the same time, the PVA at high altitude weakened, the sinking motion above the cyclone center appeared, and the high PV region at low altitudes disappeared.

The dry intrusion was a dynamic characteristic of the EC. On the WV imagery, the form of the dry zone changed from the dynamic dry band, dry delta, expanding dry band, and dry slot to dry spiral. The dynamic dry zone, accompanied by the sinking motion and ULJ, can develop into the dry intrusion at the later stage, which could be found on the composite map of WV imagery, the ULJ, and the vertical motion field.

This study analyzed the WV imagery and its dynamic mechanism of an East Asian EC case at different development stages. It is found that different shapes of dry zones on the WV imagery have different indications. The"dry delta"can signify the cyclone development. The"expanding dry band"showed the start of the cyclone development. The"dry slot"marked the key moment of the evolution of the dry intrusion. The"dry spiral"indicated the appearance of cyclonic flow at high levels and the end of surface cyclone development.

The combination of morphological characteristics of WV imagery and synoptic dynamic diagnosis makes the physical meaning of WV imagery clearer and improves the cognition of WV imagery morphology in the occurrence and development of East Asian ECs. It is worth noting that whether the features of the WV imagery suitable for most of ECs needs more cases to be verified in the future. To supplement the limitation of WV imagery qualitative analysis, further quantitative research is needed from the perspective of satellite blackbody brightness temperature, which will be carried out in the following work.