Inner-Core Lightning Outbreaks and Convective Evolution in Super Typhoon Haiyan (2013) T ⁎ Wenjuan Zhanga,B, , Steven A

Atmospheric Research 219 (2019) 123–139

Contents lists available at ScienceDirect

Atmospheric Research

journal homepage: www.elsevier.com/locate/atmosres

Inner-core lightning outbreaks and convective evolution in Super Typhoon Haiyan (2013) T ⁎ Wenjuan Zhanga,b, , Steven A. Rutledgeb, Weixin Xub, Yijun Zhangc a State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing, China b Department of Atmospheric Sciences, Colorado State University, Fort Collins, CO, United States c Institute of Atmospheric Sciences, Fudan University, Shanghai, China

ARTICLE INFO ABSTRACT

Keywords: Using lightning data from the World Wide Lightning Location Network, infrared satellite imagery, and micro- Lightning outbreak wave observations, this study investigates lightning outbreaks and convective evolution in the inner core Convective burst (0–100 km) of Super Typhoon Haiyan (2013), the strongest storm on record to make landfall in the northwest Inner core Pacific. This storm was characterized by intense lightning activity with half of the strokes occurring in the inner Rapid intensification core. Three major inner-core lightning outbreaks and convective bursts (CBs) were observed during rapid in- Super Typhoon Haiyan tensification (RI), maximum intensity (MI), and weakening stages. These outbreaks coincided with favorable large-scale environmental conditions for TC development with higher sea surface temperature (29–30 °C), higher − relative humidity (75–80%), and weaker deep-layer vertical wind shear (3–8ms 1), compared to the climatological averages for the month of November in the northwest Pacific. The RI lightning outbreak occurred primarily in the downshear quadrants and CBs were located inside the radius of maximum wind (RMW). The MI lightning outbreak occurred just after the eyewall replacement cycle, inducing marked depression of brightness temperature at 91-GHz. The lightning outbreak during Haiyan's weakening stage preferred the upshear–left quadrant outside the RMW. In contrast, relative lack of cloud-to- ground lightning in the rainbands was observed during all three main outbreaks. The radial and azimuthal distributions of lightning outbreak within the inner core provided indicative information on the relationships between convective structure and intensity changes of Haiyan.

1. Introduction TC intensification often occurs in environments with warm sea surface temperature (SST), high relative humidity (RH), and weak A recent tropical cyclone (TC) that highlights the difficulties in vertical wind shear (VWS) (Kaplan and DeMaria, 2003; Kaplan et al., forecasting TC intensity change, particularly rapid intensification (RI), 2010). Besides environment-scale parameters, internal processes also was Super Typhoon Haiyan in 2013. Haiyan formed in the northwest influence intensification (Marks and Shay, 1998; Hendricks, 2012; Pacific on 3 November 2013, and underwent RI 24 h after its formation. Wadler et al., 2018). For example the development of inner-core meso- In this work, we will refer to the definition of Kaplan and DeMaria vortices may play an important role in the RI process (Guimond et al., − (2003) to identify RI periods (a 15 m s 1 increase in the maximum 2010; Jiang, 2012; Rogers et al., 2013). Latent heat release due to sustained wind during 24 h). Haiyan intensified at RI rates for 66 h inner-core convective burst (CB) has been proposed to promote TC in- during which it evolved from a tropical storm to a category 5-equiva- tensification (Williams et al., 1992; Guimond et al., 2010; Rogers et al., − lent super typhoon (1-min maximum sustained wind = 70 m s 1). 2015). A CB is usually defined as deep convective period with high Haiyan made landfall in the Philippines at its peak intensity and caused cloud top, strong microwave ice scattering signature, increased light- catastrophic loss of life and property (6293 fatalities, 28,689 injured ning frequency, and/or high radar reflectivity (Cecil et al., 2002; and > 1.1 million homes destroyed) (National Disaster Risk Reduction Guimond et al., 2010; Rogers et al., 2013; Susca-lopata et al., 2015). and Management Council, 2014). Extremely strong winds, storm Observations showed that intensifying storms appeared to have more surges, and flash floods made it the deadliest typhoon on record in the CBs, preferentially located within the radius of maximum wind (RMW) Philippines (Lander et al., 2014). compared to non-intensifying storms (Rogers et al., 2013). When a CB is

⁎ Corresponding author at: State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing, China. E-mail address: [email protected] (W. Zhang). https://doi.org/10.1016/j.atmosres.2018.12.028 Received 27 June 2018; Received in revised form 26 December 2018; Accepted 31 December 2018 Available online 02 January 2019 0169-8095/ © 2018 Published by Elsevier B.V. W. Zhang et al. Atmospheric Research 219 (2019) 123–139 present within the RMW, axisymmetric vorticity in the eyewall will addition to serving as a proxy for the occurrence of vigorous updrafts, increase (Rogers et al., 2015) and warming from upper-level subsidence the lightning radial and azimuthal distributions can provide further around the burst could be enhanced (Braun et al., 2006), both leading information on TC convective structure. For example, in Tropical Storm to TC intensification. Gabrielle (2001), a highly asymmetric structure was observed during The development of global ground-based lightning detection net- the intensification, with the highest reflectivity and the strongest works, such as the World Wide Lightning Location Network (WWLLN; lightning outbreaks located downshear. When convective cells moved Rodger et al., 2004), and the Vaisala's world-wide network (GLD360; into the upshear region, lightning activity quickly ended (Molinari Demetriades et al., 2010; Said et al., 2010), has allowed continuous et al., 2006; Molinari and Vollaro, 2010). Hurricane Danny (1997) observation of lightning activity in TCs worldwide. Many studies have exhibited persistent downshear convective outbreaks, with inner-core noted the positive relationships between inner-core (or eyewall) light- lightning occurring consistently downshear during the development ning outbreaks and TC intensification. Lightning outbreaks within the phase (Molinari et al., 2004). Lightning distributions in Hurricane storm core preceding the onset of intensification were observed in many Edouard (2014) suggested that deep convection was prevalent TCs (e.g., Lyons and Keen, 1994; Molinari et al., 1994; Shao et al., 2005; throughout the intensification period of the storm, and primarily oc- Squires and Businger, 2008; Fierro et al., 2011). In the Atlantic Ocean, curred in the downshear quadrants inside the RMW (Rogers et al., TCs were more likely to intensify 24 h after an inner-core lightning 2016; Zawislak et al., 2016). Nevertheless, the inner-core lightning in burst if the storms were steady or intensifying before burst onset Hurricane Earl (2010) maximized in the upshear quadrant (Stevenson (Stevenson et al., 2018). A marked increase in inner-core lightning et al., 2014; Susca-lopata et al., 2015), possibly due to enhanced low- density during intensifying stages was also observed for TCs in the level RH and elevated convection in the upshear quadrant. These stu- northwest Pacific(Pan et al., 2010; Zhang et al., 2012; Zhang et al., dies suggest that lightning provides useful information on the evolution 2015). Lightning outbreaks in the eyewall led the TC peak intensity by of deep convection within TCs, intensity changes and their bulk dyna- several hours. Similarly, in the Indian Ocean, lightning outbreaks oc- mical/microphysical evolution. curred prior to the TC maximum intensity (MI) (Ranalkar et al., 2016) Modeling studies have explored in detail the electrical processes and lightning activity in the eyewall tended to decrease for weakening within the inner core and their relationships to TC evolution. In an TCs (Bovalo et al., 2014). idealized hurricane modeled by Fierro et al. (2007), the highest total In contrast, DeMaria et al. (2012) suggested that bursts of lightning lightning flash rates occurred within the eyewall where larger updrafts within the inner core may be an indicator of end of RI, while lightning produced a favorable charging environment. During the RI of Hurricane increase in the outer rainbands was an indicator of imminent in- Rita (2005), electrically-active CBs located near the eyewall. The ra- tensification. Similarly, Stevenson et al. (2016) found that enhanced pidly rotating electrical activities were found near the RMW during inner-core lightning was associated with non-intensifying TCs in the intensification phases (Fierro and Reisner, 2011). The simulated charge eastern North Pacific basin. Xu et al. (2017) further displayed the ne- structure in these studies indicated that both the eyewall and the gative relationship of TC intensity change to inner-core total lightning rainband convection exhibited a gross normal tripole charge structure and intense convection using long term (16 years) Tropical Rainfall (Fierro et al., 2013, 2015; Fierro and Mansell, 2017, 2018), consisting Measuring Mission (TRMM) lightning and precipitation radar data. of a main negative charge layer sandwiched in between two layers of Despite these findings, a clear relationship between lightning activity in positive charge (Williams, 1989). Akin to continental convection (e.g., the inner core or eyewall region and TC intensification has not yet Fierro et al., 2006; Kuhlman et al., 2006), the temporal and spatial emerged. evolution of total lightning flashes in both the eyewall and rainband Previous studies based on global ground-based networks, such as were shown to correlate well with updraft and graupel volume. Fierro WWLLN, have helped augment our understanding of the general pat- and Mansell (2017) further highlighted that because the eyewall con- tern of TC lightning. However, these networks operated in very low vection exhibited a more active warm rain trait than in the rainbands, frequency (VLF) and mainly detect cloud-to-ground (CG) lightning. As this could explain why lightning in the inner core of TCs is generally the ratio of intra-cloud (IC) to CG lightning averages 3:1 in garden- episodic. The relative peaks of these parameters were coincident with variety storms (Boccippio et al., 2001) to ratios as large as 10:1 in se- the end of an intensification phase (Fierro et al., 2015). The above vere continental storms (e.g., MacGorman et al., 1989; Schultz et al., modeling studies revealed some of the relationships between simulated 2011; Medici et al., 2017), there are many benefits of using total electrification and TC's kinematic and microphysical characteristics, lightning in contrast to CG-only data. In particular, total lightning was and suggested the utility of using total lightning to diagnose changes in shown a more comprehensive description of lightning activity within TC structure and intensity. TCs. Moreover, total lightning is shown to better correlate with storm's Super Typhoon Haiyan (2013) has ranked as the second-most in- bulk dynamical and microphysical properties (MacGorman et al., 1989; tense TC in the world since 1979. However, very few observational Lang and Rutledge, 2002). The recent launches of the Geostationary studies (Wang et al., 2016; Shimada et al., 2018) of Haiyan have been Lightning Mapper (GLM; Goodman et al., 2013; Rudlosky et al., 2018) made due to lack of dense observations in the northwest Pacific. Studies on board the Geostationary Operational Environmental Satellite 16 analyzing lightning in TCs in recent years have shown the correlation (GOES-16) and the Lightning Mapper Imager (LMI; Yang et al., 2017) between inner-core lightning and TC structure evolution, and indicated on board the Fengyun Satellite 4A (FY-4A) provide continuous ob- the potential use of inner-core lightning to aid TC intensity change servations of total lightning in the Atlantic and Pacific basins, respec- nowcasting (Zhang et al., 2015; Fierro et al., 2018; Stevenson et al., tively. Both datasets will enable a better understanding of TC total 2018). The results of this study aims to provide additional evidence on lightning and its relationship to intensity changes. this correlation and indication. Based on lightning and satellite ob- Lightning can provide useful information on the distributions of servations, this study investigates the inner-core lightning patterns and deep convection. Petersen et al. (1999) indicated the connection be- microwave-based convective properties throughout the life cycle of tween lightning and kinematics, precipitation structure, and micro- Super Typhoon Haiyan. The study intends to partially address the fol- physics in tropical convection. Black and Hallett (1999) provided the lowing questions: (1) Is there any inner-core lightning outbreak first conceptual model of hydrometeor and charge distribution within throughout Haiyan's life cycle? If so, during what stages do these out- TCs. Their observational study indicated that, in developing storms, breaks occur and what about lightning activity in the rainbands during − strong vertical velocities (> 10 m s 1) promote the presence of super- these stages? Are inner-core lightning outbreaks particularly related to cooled liquid water. When supercooled liquid water is present at tem- the RI of the storm? (2) What are the inner-core convective properties peratures < −20 °C, significant electrical activity can occur during various lightning outbreak periods, in terms of lightning rate, (Takahashi, 1978; Saunders and Peck, 1998; Saunders et al., 2001). In microwave ice scattering signature, distributions (radial and azimuthal)

124 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

− of lightning and microwave-indicated deep convection? Are inner-core with an intensification rate of 44 kt (24 h) 1. convective properties during the RI stage special different as compared to other lightning outbreak periods? The detailed analysis on this in- 2.3. Satellite data dividual case would provide additional insight for the application of FY- 4A LMI lightning data in the monitoring and forecasting of typhoon in Multifunctional Transport Satellite (MTSAT) infrared (IR) bright- the northwest Pacific. ness temperature (TBB) at 0.05° spatial and 1 h temporal resolution were used to infer the convective characteristics of the storm. The 2. Data and methodology MTSAT-1R meteorological sensor contains one visible and four infrared channels and operates in the region 20°S–70°N, 70–160°E. Data from 2.1. Lightning data the IR1 (10.3–11.3 μm) channel were used in this study. Passive microwave data from the Special Sensor Microwave Imager/ The lightning data used in this study were from the WWLLN (http:// Sounder (SSMIS) were also used. Microwave observations have been wwlln.net). Many studies have used this dataset to study the temporal known to better relate to convective intensity compared to IR mea- and spatial structure of TC lightning activity (e.g., Pan et al., 2010; surements (Cecil et al., 2002; Liu et al., 2011; Solorzano et al., 2016). Abarca et al., 2011; Bovalo et al., 2014; Ranalkar et al., 2016; Zhang The SSMIS data were from the NOAA's National Centers for Environ- et al., 2018); to help distinguish between intensifying and non-in- mental Information (https://www.ncei.noaa.gov/data/ssmis- tensifying TCs (e.g., Price et al., 2009; Thomas et al., 2010; DeMaria brightness-temperature-csu/). SSMIS measures 91-GHz brightness et al., 2012; Pan et al., 2014); and to quantitatively indicate the dis- temperatures in both vertical and horizontal polarizations. The 91-GHz tributions of intense convection in TCs (e.g., Stevenson et al., 2014; polarization corrected temperatures (PCTs) were calculated using the Susca-lopata et al., 2015; Solorzano et al., 2016). method by Spencer et al. (1989), to remove the contrast between land The WWLLN monitors lightning around the globe in real time using and ocean emissivity. VLF (3–30 kHz) receivers (Rodger et al., 2004). It detects both CG and IC lightning strokes, as long as peak current of the stroke exceeds the 2.4. Environmental analysis detection threshold of 30 kA (Jacobson et al., 2006). Since CG lightning flashes typically have larger peak currents than IC lightning, the The National Centers for Environmental Prediction (NCEP) global WWLLN detection efficiency (DE) of CG strokes is much larger than that Final (FNL) analysis data were used to quantify the wind shear and of IC strokes (Abarca et al., 2010). As of January 2013, the WWLLN humidity conditions. The NCEP FNL data are provided at 6 h intervals consists of over 70 sensors around the world (Hutchins et al., 2013). at a horizontal resolution of 1°. Comparisons of NCEP FNL and ECMWF The increasing number of stations has led to an improvement in net- ERA-Interim for the time period of analysis revealed relatively con- work DE. Rudlosky and Shea (2013) found that WWLLN DE in the sistent wind shear profiles and relative humidity fields (not shown). western hemisphere improved from 6% in 2009 to 9.2% in 2012, with a Deep-layer VWS was calculated between 850 and 200 hPa, by averaging mean location accuracy of 11 km. Studies (Hutchins et al., 2013; Virts over a 0–500 km radius from the best-track center to eliminate the et al., 2013) showed that the network has an average accuracy of 5 km symmetric vortex (Hanley et al., 2001). Low-level RH was calculated as and an estimated DE of 11%. the average value over 850–700 hPa within the 200–800 km annulus relative to the best-track positions from JTWC (Kaplan et al., 2010). SST 2.2. Best-track data data were taken from the NCEP Real Time Global high-resolution analysis. The dataset is a daily SST product at 1/12° horizontal re- TC track and intensity data were obtained from the Joint Typhoon solution. The mean SST within 300 km of the storm center was calcu- Warning Center (JTWC) best-track dataset (http://www.usno.navy. lated and used as the TC SST. mil/JTWC). This dataset provides, at 6-h intervals, the storm central latitude and longitude, maximum 1-min mean sustained wind speed 3. Overview of Haiyan (kt), minimum sea level pressure (mb), as well as the RMW. The de- termination of RMW from JTWC is based on estimates of intensity, Haiyan formed within the Asian monsoon trough (Shu and Zhang, latitude and information contained in satellite IR images. The main 2015) in the western Pacific on 2 Nov 2013, and developed into a source of intensity estimates is the Dvorak model. Using satellite data, tropical depression around 0600 UTC 3 Nov. At 0000 UTC 4 Nov, the the distance between the coldest cloud top temperature and the system reached tropical storm strength, followed by a RI period lasting warmest temperature within the eye is one of the methods used to over 60 h. The system achieved typhoon strength by 0000 UTC 5 Nov compute the RMW (Mueller et al., 2006). Shimada et al. (2018) used and moved west toward the Philippines. By 1200 UTC 6 Nov, the the ground-based velocity track display (GBVTD) technique to retrieve system was designated as a category-5-equivalent super typhoon Haiyan's wind field from Guiuan Doppler radar (2.5-h landfall period). (Fig. 1b). Meanwhile, the storm underwent an eyewall replacement The estimated range of the maximum 1-min sustained wind speed was cycle (ERC; Willoughby et al., 1982) until the end of 6 Nov (determined quite consistent with that from the JTWC's best-track data. The esti- using the Morphed Integrated Microwave Imagery MIMIC-TC product), mated RMW from 2- to 6-km altitude at 2020 UTC 7 Nov 2013 was marking the end of Haiyan's RI. At 1200 UTC 7 Nov, Haiyan reached its 25.5 –28.5 km. The interpolated RMW from JTWC at this time was peak intensity with a minimum sea level pressure of ~895 hPa and 1- 31.3 km. Therefore, we inferred that the uncertainty in RMW from min maximum sustained winds of 170 kt (Fig. 1b). Haiyan made its first JTWC compared to radar measurements was < 6 km for Haiyan. landfall in the Philippines at Guiuan at 2300 UTC 7 Nov (National Storm center location and intensity at hourly intervals were then Disaster Risk Reduction and Management Council, 2014). Following the obtained via spline interpolation to estimate the rainfall and large en- first landfall, Haiyan made four more landfalls as it crossed the central vironmental parameters relative to TC center. When analyzing the ra- Philippines. On the morning of 9 Nov, the motion of Haiyan shifted dial and azimuth distribution of lightning to the storm, the best-track toward the northwest, and the storm began to weaken. data were interpolated to the occurrence time (to minute) of lightning stroke. The JTWC best-track data for Haiyan were recorded from 0600 4. Results UTC 3 Nov to 0600 UTC 11 Nov, 2013 (Fig. 1a). Based on the RI definition of 24 h TC wind change > 30 kt (Kaplan and DeMaria, 2003), a 4.1. Lightning evolution and large-scale environmental conditions 66-h RI period (i.e., from 0000 UTC 4 Nov to 1800 UTC 6 Nov) was identified, during which the wind speed increased from 35 to 155 kt Fig. 1c presents lightning stroke density of Haiyan detected by

125 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

(caption on next page)

126 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Fig. 1. JTWC best tracks, TRMM 3B42 rainfall, and lightning density for the life cycle of Super Typhoon Haiyan in November 2013. (a) Accumulated rainfall (shading) and 0000 UTC best tracks (open circles). (b) Maximum sustained winds (blue) and minimum central pressures (orange). (c) Lightning stroke density (stroke − − (100 km) 2 h 1) detected by WWLLN. The red line segment in (a) and the shaded area in (b) indicate the RI period (between 0000 UTC 4 Nov and 1800 UTC 6 Nov). (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

− Fig. 2. Time series of the hourly lightning rate (str h 1) within inner core (100 km) and rainbands (100–500 km) detected by (a) the WWLLN and (b) the GLD360. Haiyan's landfall was made at 2040 UTC 7 Nov, ~2 h after the MI. The RI period lasted 66 h from 0000 UTC 4 Nov to 1800 UTC 6 Nov. ERC started at 0800 UTC 6 Nov and ended at 2000 UTC 6 Nov. The gray shadings indicate three periods of inner-core lightning outbreaks.

127 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

WWLLN during its life cycle and Fig. 2a the evolution of WWLLN − lightning stroke rates (str h 1) within a 100 km (inner-core lightning) and 100–500 km (rainband lightning) radius from the JTWC-estimated center of Haiyan. To facilitate comparisons, the above radii criteria were purposively chosen to be similar than those used in many studies on TC lightning (Corbosiero and Molinari, 2002, 2003; Abarca et al., 2011; DeMaria et al., 2012; Zhang et al., 2015; Xu et al., 2017; Fierro et al., 2018). The tropical depression stage (i.e., before 0000 UTC 4 Nov) was characterized by low lightning rates. As the system developed into a tropical storm and entered the RI phase, both the inner-core and rainband lightning increased substantially. When Haiyan achieved typhoon intensity (24 h after the RI onset), lightning in both the inner core and outer rainbands became frequent. This behavior continued for the first half of the RI phase until 1200 UTC 5 Nov. When the intensity of Haiyan reached super typhoon strength, lightning was largely confined in the outer rainbands. Meanwhile, the rainband lightning underwent a decrease until the occurrence of an ERC. When Haiyan reached its maximum intensity, the rainbands and the inner core si- multaneously produced high lightning activity. After the landfall in the Philippines, the rainband lightning began to notably increase with a similar increase of the inner-core lightning seen 4 h later. During this period, Haiyan experienced a sharp weakening and lightning activity ceased 12 h later. Fig. 2b presents the evolution of lightning rates detected by the GLD360. A comparison between Fig. 2a and b shows same patterns in lightning rates for both the inner-core and the rainbands. The two networks are consistent in the relationship between lightning rate and TC intensity. The lightning peaks and outbreaks observed by WWLLN were also detected by the GLD360. Inner-core lightning outbreaks detected by WWLLN were de- − termined as a period with high lightning stroke rate (> 300 str h 1, i.e., the upper 10% of all hourly rates) within 100 km of the storm center that persisted several hours. Three inner-core lightning outbreaks were identified during the life cycle of Haiyan (Fig. 2a). The first outbreak (OB1, termed the RI outbreak) took place during the RI period, from 1200 UTC 4 Nov to 1500 UTC 5 Nov, when the storm intensity was upgraded from tropical storm to typhoon status. This stage was comprised of several very short periods of enhanced lightning activity, highlighting the transient nature of the deep convection during this outbreak (Squires and Businger, 2008; Fierro et al., 2011). The inner-core lightning dropped rapidly after OB1, and there were no large values of inner-core lightning rate during the following 35 h. A second outbreak (OB2, termed the MI outbreak) occurred after the ERC, from 1200 UTC to 1700 UTC 7 Nov, coincident with the storm's MI. The third outbreak (OB3, termed the weakening outbreak), from 1400 UTC to 2200 UTC 8 Nov, occurred during the weakening stage after the storm's landfall. The first two outbreaks occurred during the stages when the system intensified rapidly from a tropical storm to a super typhoon. This is consistent with lightning behaviors in Atlantic Hurricanes Rita and Katrina (2005) (Squires and Businger, 2008; Fierro et al., 2011), and Earl (2010) (Stevenson et al., 2014; Susca-Lopata et al., 2015). Although the third outbreak occurred during the weakening stage, the storm maintained typhoon intensity (category 4 equivalent on the Saffir-Simpson scale) at that time. Lightning outbreaks during the weakening stage were also observed in Typhoons Saomai (2006) and Jangmi (2008) (Pan et al., 2010). Relatively few inner-core lightning strokes occurred after this point in time. Fig. 3 shows the time series of large-scale environmental parameters for Haiyan. The large-scale environment was very favorable for the Fig. 3. Time series of maximum sustained winds from Haiyan and large-scale storm's rapid development (Shu and Zhang, 2015). Haiyan developed in environmental parameters from the NCEP analysis. (a) Sea surface temperature a warm and moist environment under weak wind shear. All the out- (°C) averaged within 300 km of the storm center; (b) relative humidity (%) breaks occurred under the conditions of elevated SST (29–30 °C, averaged in the 850–700 hPa layer at radii of 200–800 km from the storm – −1 ff Fig. 3a), enhanced RH (75–80%, Fig. 3b), and moderate shear center; and (c) 850 200 hPa vertical wind shear (m s ,di erence in wind − between 850 and 200 hPa) averaged within 500 km of the storm center. The (3–8ms 1, Fig. 3c). Based on the NCEP monthly gridded climate da- gray shading indicates the RI period. OB indicates the inner-core lightning tasets, the climatological mean values (starting in 1979) of SST and outbreak. wind shear for November within the Haiyan region [5–25°N, − 105–160°E] were 27.9 °C and 12.3 m s 1, respectively. The values of

128 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Table 1 Radial distribution of average lightning density, percentage of stroke number (in brackets in the right-most columns), and peak intensity for the life cycle of Haiyan (2013), Rita (2005) and Katrina (2005).

− − TC name (year) Peak intensity Maximum lightning rate in Lightning density (str (100 km) 2 h 1) and percentage (%) in radial regions − inner core (str h 1) Minimum Maximum wind Eyewall Inner core Inner rainband Outer rainband pressure (hPa) (kt) (0–50 km) (0–100 km) (100–200 km) (200–500 km)

Haiyan (2013) 895 170 936 86.2 (38%) 27.9 (49%) 0.2 (1%) 1.4 (50%) Rita (2005)a 895 155 475 17.6 (3%) 8.3 (5%) 8.0 (14%) 6.3 (81%) Katrina (2005)a 902 150 86 4.3 (2%) 1.7 (3%) 2.7 (13%) 2.3 (84%)

a As the detection eﬃciency of WWLLN varied considerably from 2005 to 2013, WWLLN data for Rita and Katrina were multiplied by an adjustment factor that made the detection eﬃciency equal to that of Haiyan-era. The High Resolution Full Climatology (HRFC) data (http://lightning.nsstc.nasa.gov/data/) from TRMM LIS/OTD were used to calculate the adjustment factor (3.32), following the procedures by DeMaria et al. (2012), Bovalo et al. (2014) and Zhang et al. (2015). The best-track data for Rita and Katrina were from the NHC best-track dataset.

SST (wind shear) of Haiyan were clearly higher (lower) than the No- data, found this ratio was 30:1 for Hurricane Maria (2017), a category-5 vember climatology. Thus, these combined conditions provided a very storm in Atlantic evolving in a similar environment to Haiyan and also favorable environment for Haiyan's development and intensification. A undergoing an ERC. The high ratios contrast with previous studies using recent storm in Atlantic, Hurricane Maria (2017), which produced CG lightning that outer rainbands tended to dominate the lightning several lightning outbreaks during intensification, was also embedded density distribution (e.g., Molinari et al., 1994, 1999). − in a similar environment (SST of 29–30 °C, wind shear < 3 m s 1 at the Fig. 4 shows the lightning density within the inner core of Haiyan as beginning of intensification) than Haiyan and underwent two dee- a function of time and distance from the center, with RMW from the pening phases to a category-5 hurricane (Fierro et al., 2018). JTWC overlaid for comparison. The radial coverage of frequent lightning within the inner core stems for the presence of intense convection 4.2. Inner-core lightning structure near the center. Generally, the radial lightning distribution indicates a shift in the maximum lightning density toward the RMW during the fi Haiyan was characterized by vigorous lightning activity in the inner intensi cation stage, and a shift outwards during the weakening stage. fi – core. Table 1 shows the radial distribution of average lightning density The rst outbreak of inner-core lightning began to develop at 30 40 km for Haiyan compared with Rita and Katrina, two category-5 hurricanes radius from the RMW. As the storm strengthened to typhoon intensity in 2005 that also presented intense inner-core lightning (Shao et al., and continued to intensify rapidly, the maximum density of inner-core 2005; Squires and Businger, 2008; Fierro et al., 2011). Lightning data lightning moved gradually close to the RMW and the minimum distance for Rita and Katrina were multiplied by an adjustment factor (3.32) between the maximum lightning density and the storm center was re- derived from TRMM LIS/OTD that made the WWLLN DE in 2005 equal duced to 10 km. The inner-core lightning rate was reduced during the fi to that of the Haiyan-era. A significant proportion of lightning in ERC after the rst outbreak. After the ERC, the second outbreak began. Haiyan (49%) occurred within the inner-core region. The difference in Frequent lightning was present during this MI outbreak, indicating deep lightning density between the inner core and the outer rainbands was convection within the inner core. The inner-core lightning extended to – more obvious in Haiyan than Rita and Katrina (Table 1). The inner-core a larger radial area with the radii distance of 50 60 km. However, the − lightning rates were as high as 936 str h 1 in Haiyan and lightning maximum of lightning density (deep convention) was still located in- − − density in the inner core was 27.9 str (100 km) 2 h 1. For comparison, side the RMW during this outbreak. In contrast, the lightning density fi Rita and Katrina exhibited lightning density of 8.3 and 1.7 str maximum (core of the deep convection) during the nal outbreak was − − (100 km) 2 h 1, respectively, in the inner-core region. Haiyan showed located outside the RMW. Dense lightning activity gradually shifted – a large lightning density ratio (20:1) between the inner core and outer outward from the storm center and moved to the distance of 70 80 km fi rainbands during the life cycle. Fierro et al. (2018), using total lightning at the end of the nal outbreak.

− − − Fig. 4. Lightning density (10 2 str km 2 h 1) within the inner core as a function of time and distance from the storm center. Bins on the distance axis are 10 km wide. The dashed line indicates hourly interpolated RMW from the JTWC best-track data. OBs indicate three outbreak periods of inner-core lightning.

129 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

distribution of deep convection in the inner core during this period (Shimada et al., 2018).

4.3. Inner-core convective properties by satellite

The evolution of lightning density and pattern described above indicates that there were three inner-core lightning outbreaks (i.e., the RI, MI, and weakening outbreaks) in Haiyan. In this subsection, using satellite data, we investigated the evolution and distribution of deep convection during these outbreaks. Fig. 6 shows hourly evolution of IR TBB and microwave 91-GHz PCT within the inner-core overlaid by lightning rates. All three inner-core lightning outbreaks were associated with significant decreases (10–20 K drops) in TBB. Cloud tops developed colder temperatures during the periods of lightning outbreak, suggesting deeper convection in the inner core. On the other hand, the 91-GHz PCT decreased persistently throughout the developing, intensifying, and mature stages of the TC, indicating continuous increase −2 −2 Fig. 5. Inner-core lightning density (10 str km ) as a function of time and of vertically integrated ice content within the inner core until the storm angle between lightning and the wind shear (measured clockwise from the became mature. The 91-GHz PCT exhibited the largest depression lightning direction). Lightning have been rotated with respect to the shear during the RI lightning outbreak period (i.e., change of ~20 K from the vector during each 6 h period. The labels ‘UL’, ‘DL’, ‘DR’, and ‘UR’ mark the upshear–left, downshear–left, downshear–right, and upshear–right quadrants, beginning to the end of the period), suggesting the production of large respectively, and the hatched area highlights the two most fully downshear amount of ice particles enhancing latent heating within the inner core, octants. The x-axis labels ‘left’ and ‘right’ are with respect to the shear vector. which helped to further promote TC intensification (Guimond et al., 2010; Rogers et al., 2015). As the storm weakened, the inner-core PCT at 91-GHz increased rapidly (due to the decrease of vertically integrated A radial distribution of convection located inside the RMW will be ice content). This is intriguing, given that an inner-core lightning out- more favorable for TC intensification. Satellite and airborne observa- break occurred during the same period. However, this could be com- tions have noted the potential role of precipitation and convection in- plicated by the influence from land on the inner-core convection during side the radius of RMW on TC intensification (Jiang, 2012; Rogers et al., landfall. 2013; Tao and Jiang, 2015). A recent observational study by Wadler Using TRMM passive microwave observations in hurricanes, Cecil et al. (2018) showed that deep convection in intensifying TCs were and Zipser (2002) constructed a conceptual model of microphysical preferentially located inside the RMW, with stronger updrafts and features for eyewall region. It is shown that hydrometeors were spread higher echo tops compared with steady-state TCs. Lightning outbreaks out over a great horizontal area and supercooled liquid water was inside the radius of RMW were also found to precede or accompany RI confined to a narrow region on the inward edge of the eyewall (Black in Hurricanes Rita and Katrina (2005) (Squires and Businger, 2008; and Hallett, 1999). In the passive microwave imageries, the coverage of Fierro et al., 2011), Earl (2010) (Stevenson et al., 2014; Susca-Lopata PCT < 250 K can serve as a proxy for the precipitation coverage and et al., 2015), and Maria (2017) (Fierro et al., 2018). Theoretically, PCT < 200 K for intense convection coverage (Cecil and Zipser, 1999; when convection occurs closer and especially within the RMW, the Cecil et al., 2002; Susca-lopata et al., 2015). In the inner-core region, convective heating is more easily retained allowing the vortex to in- the probability of lightning began to increase when 85-GHz PCTs were tensify further (Hazelton et al., 2017; Wadler et al., 2018). below 170 K (Cecil and Zipser, 2002). Figs. 7-9 show several satellite IR The azimuthal distribution of convection is important in TC asym- imageries overlaid with SSMIS 91-GHz PCT and lightning during the RI, metric structures and intensification. Previous studies have indicated MI, and weakening stages. Between 2 and 16 h before the onset of OB1 that the response of storm to deep-layer wind shear governs the azi- (Fig. 7a-b), a considerable fractional area of intense convection muthal distribution of convection (Fierro and Mansell, 2017; Fierro (PCT < 200 K) appeared in the core. Highly asymmetric cold cloud was et al., 2017; Wadler et al., 2018). Using lightning data as a proxy for evident in the rainbands. At the time of the onset of OB1 (Fig. 7c-d), a vertical motion and deep convection, Corbosiero and Molinari (2002, clear eyewall had formed and the storm center was characterized by a 2003) found that lightning occurrence showed a marked downshear ring of intense convection. Additionally, a strong outer rainband with and downshear-left preference. Fig. 5 shows the temporal evolution of large lightning rates (Fig. 7d and consistent with Fig. 2) developed lightning density in the inner-core region as a function of azimuthal during this period. Cold cloud in the rainbands exhibited an arc-like angle between lightning and wind shear. It is interesting to note that structure and was located in the downshear–right quadrant relative to Haiyan was dominated by lightning in the upshear–left quadrant of the the storm center. After the OB1 (Fig. 7e-f), Haiyan intensified to a ty- inner core at the very beginning of the RI stage. Such a relationship phoon (category 2 equivalent on the Saffir-Simpson scale). Wind shear between deep convection in the upshear-left quadrant and intensifica- − continued to increase (to 9.9 m s 1; Table 2), but strong axisymmetric tion was also shown in Hurricanes Earl (2010) (Stevenson et al., 2014; convection still persisted in the storm core. At this time, lightning ac- Rogers et al., 2015) and Edouard (2014) (Rogers et al., 2016), and the tivity in the rainbands ceased to near zero. The deep-layer shear or- composite study of airborne Doppler measurements of CBs by Wadler ientated easterly to northeasterly during this outbreak, and intense et al. (2018). As the storm intensified, intense convection and lightning convection developed preferentially in the downshear quadrant within tended to be more downshear. Table 2 shows that > 85% of inner-core the inner core (consistent with Fig. 5). These results agree with the lightning during the intensification period occurred downshear, which typical patterns of inner-core lightning relative to wind shear (e.g., is in agreement with previous studies of Atlantic storms. After Haiyan Corbosiero and Molinari, 2002, 2003; Molinari and Vollaro, 2010; made landfall, during the weakening stage, the inner–core convection Abarca et al., 2011). However, inner-core lightning could also peak in shifted to upshear–left quadrant. The Guiuan Doppler radar also ob- the upshear quadrant of a storm. This atypical pattern was observed in served strong reflectivity in the upshear side of the storm during Hurricane Earl (2010) (Stevenson et al., 2014; Susca-Lopata et al., Haiyan's landfall (Shimada et al., 2018). The strong tangential wind and 2015). relatively large translation speed (the averaged wind shear was − Strong TCs have often been observed to go through an ERC process: 7.1 m s 1 easterly at this time) may contribute to the upshear formation of a secondary eyewall, decaying of the original eyewall,

130 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Table 2 Vertical wind shear between 850 and 200 hPa, averaged within 500 km of the center of Haiyan; minimum central pressure and maximum surface wind; average lightning rate in the inner core and percentage of inner-core lightning in the downshear–left (DL), downshear–right (DR), upshear–left (UL), and upshear–right (UR) quadrants. Vertical shear angles were measured in the meteorological direction.

Stage Date, time (UTC) Vertical wind shear TC intensity Inner-core lightning

− − Direction (°) Magnitude (m s 1) Pressure (hPa) Wind (kt) Rate (str h 1) % DL % DR % UL % UR

RI 4 Nov 0000 91 4.6 996 35 3 56 38 6 0 0600 87 4.8 993 40 1 17 0 83 0 1200 63 6.1 989 45 109 38 1 56 5 1800 51 8.0 982 55 26 58 20 7 15 5 Nov 0000 41 9.2 970 70 49 12 6 35 47 0600 66 7.7 967 75 123 34 1 57 8 1200 49 9.9 956 90 339 22 70 2 6 1800 33 5.9 941 110 67 2 93 1 5 6 Nov 0000 33 8.5 926 130 105 11 64 9 17 0600 51 7.9 922 135 50 45 42 4 9 1200 36 6.6 911 150 76 18 65 6 12 1800 49 7.5 907 155 47 31 17 18 35 MI 7 Nov 0000 48 7.1 907 155 178 25 31 15 30 0600 22 5.0 903 160 133 11 37 17 36 1200 44 3.2 895 170 472 12 74 2 13 1800 57 3.2 895 170 167 28 34 18 20 Weakening 8 Nov 0000 20 6.2 899 165 23 38 32 15 15 0600 63 7.6 914 145 17 56 3 37 4 1200 101 6.0 926 130 368 27 1 68 5 1800 95 7.3 933 120 254 42 0 58 0 9 Nov 0000 86 7.8 937 115 140 31 6 62 1 0600 66 7.8 944 105 15 58 26 3 13

− Fig. 6. The maximum sustained wind speed (kt), inner-core lightning rate (str h 1), median TBB (K) from MTSAT-1R and median 91-GHz PCT (K) from SSMIS within the inner-core. Gray shadings indicate three inner-core lightning outbreaks (OB1, OB2 and OB3) during the RI, MI, and weakening stages, respectively. contraction of the secondary eyewall, followed by the re-intensification was evident in Fig. 8a as a ring-like region of deep convection of the cyclone (e.g., Willoughby et al., 1982). The ERC process is often (PCT < 200 K) was observed outside the inner eyewall. An area of accompanied with temporary weakening of the storm, due to the strong updrafts, indicated by intense inner-core lightning, was located changes in the radial pressure gradient within the eyewall (Sitkowski much closer to the RMW than the during RI periods. Two convective et al., 2012). After the secondary eyewall envelops the original one, the cores with high lightning flash rates were present in the rainbands, new larger-diameter eyewall may contract in radius, causing re-in- located in the downshear-right quadrant of the storm center. The storm tensification of the storm (Houze et al., 2007). Based on data from the continued to intensify during the ERC, and reached category-5- Morphed Integrated Microwave Imagery (Wimmers and Velden, 2007), equivalent super typhoon intensity at 2043 UTC 6 Nov (Fig. 8b). Haiyan underwent a complete concentric ERC beginning at 0800 UTC Squires and Businger (2008) and Fierro et al. (2011) showed that and ending at 2000 UTC 6 Nov. The development of an outer eyewall eyewall lightning outbreaks occurred during the ERC in Hurricanes Rita

131 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

(caption on next page)

132 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Fig. 7. Lightning from WWLLN (black circles), SSMIS 91-GHz PCT (K, color shading) and MTSAT IR brightness temperature (gray shading) during the RI stage. The inner-core lightning outbreak during this stage (OB1) began at 1200 UTC 4 Nov and ended at 1500 UTC 5 Nov. The thick black circles mark the radii of 100 and 500 km from the storm center. Lightning within ± 15 min of the stated observation time are included. The arrow in the lower left corner of each image shows the direction of the shear vector for the 6 h bin nearest to the observation time. The white lines show the land-sea boundary.

(2005) and Katrina (2005). In the case of Haiyan, heightened lightning Satellite observations showed that the MI outbreak appeared to be as- activity was also observed in the inner core during the ERC, but with sociated with deeper convection (as indicated by lower median TBB and relatively lower rates compared to that during the RI period (Fig. 2). 91-GHz PCT, Fig. 6) than was the RI outbreak. A symmetric area of very The new eyewall was evident from the distinct lightning ring at a 20-km low PCT expanded around the inner core (Fig. 10b), suggesting an radius of the storm center (Fig. 8b). axisymmetric eyewall after the ERC. The fractional area with PCT < By the time of the second outbreak (OB2), the storm had developed 200 K could be used as an indicator of CB coverage (Susca-lopata et al., into a deep axisymmetric primary circulation (Fig. 8c-e). This MI out- 2015). During the MI phase in Haiyan, the inner core exhibited very break was observed with lightning in every azimuthal quadrant of the low brightness temperatures, with 87% of pixels falling below 200 K. inner core. Lightning density maxima were located in the downshear Significant differences in radial and shear-relative locations of CBs quadrants of the eyewall (Fig. 5), coincident with minimum 91-GHz existed during the weakening stage compared to the RI and MI periods. PCT values. It is noteworthy that deep convection with active lightning Summary graphic (Fig. 10c) showed an asymmetric distribution of low was observed in the inner-core region at the time of landfall (Fig. 8d-e). brightness temperature and CB activity during the weakening stage Deep convection was also observed in the outer rainbands to the north with more vigorous convection confined to the upshear-left quadrant of of the storm center. the storm. Furthermore, the radial distribution showed that the primary Lightning outbreak during the weakening stage (OB3) of Haiyan CB activity during this stage was located outside the RMW (r/ exhibited the highest inner-core lightning rate of the entire life cycle, RMW = 2.0, Fig. 10c). Both the asymmetric structure and the upshear- but was less persistent and of shorter duration than the previous two orientated deep convection outside the RMW were associated with the outbreaks (Figs. 2 and 4). Wind shear changed from northeasterly to weakening of the storm. easterly during this outbreak. Deep convection core was displaced to the southeast of the center with an upshear orientation at this time (Fig. 9a-b), compared to the CBs during the RI and MI stages. Eight 5.2. Possible mechanisms for lightning outbreak during the weakening stage hours after this final outbreak, Haiyan's motion changed from westward to northwestward. Relatively few lightning flashes occurred within the The most striking aspect of the convection occurring during the inner core and rainbands after this point of time (Fig. 9c). weakening stage were the new CBs that formed upshear (especially upshear–left) of the center. Compared to the CBs in the downshear quadrant in the RI and MI outbreaks, 60% of the inner-core lightning in 5. Discussion the weakening outbreak was in the upshear–left quadrant. Corbosiero and Molinari (2003) found that it was rare for lightning to peak in this 5.1. Convective bursts during the RI, MI, and the weakening stages quadrant (see their Fig. 7): only 4% of the cases in their study featured an upshear–left maximum. Lightning locations are related to both the Haiyan developed in the presence of high SST (29–30 °C), high low- direction of the deep-layer shear and the direction of motion of the − level humidity (75–80%), and weak wind shear (3–8ms 1)(Fig. 3). storm, but more precisely lightning distribution is determined by the The values of SST (wind shear) of Haiyan were clearly higher (lower) relative orientations of the shear and motion vectors (Corbosiero and than the November climatology in the northwest Pacific, providing Molinari, 2003). Fig. 11 shows time series of the wind shear and motion favorable large-scale environmental conditions for its RI. CBs within the vector directions, and the angle between the two vectors in Haiyan. The inner core during the RI stage began 12 h after the onset of RI. Elec- wind shear direction changed noticeably at 1200 UTC 8 Nov, at the trified TC inner cores were generally associated with strong updrafts beginning of the weakening outbreak. With the change in shear direc- − and deep mixed-phase layers, e.g., updraft speeds of 10–20 m s 1 in tion, the angle between the shear and motion vector changed by 77°. − Hurricane Karl (2010) (Reinhart et al., 2014) and exceeding 10 m s 1 in This likely represents one of the possible causes of lightning outbreak simulated hurricanes (Fierro and Reisner, 2011; Fierro et al., 2015; during the weakening stage. Land effects when Haiyan crossed the Fierro and Mansell, 2018). These microphysical environments were Philippines, notably frictional convergence could also have contributed favorable for the presence of graupel, ice particles and supercooled to this final outbreak of lightning in the inner core. water (Black and Hallett, 1999; Houze et al., 1992). Summary graphic After the outbreak of inner-core lightning during TC weakening, (Fig. 10a) shows that during the RI outbreak of Haiyan, deep convec- Haiyan changed direction from westward to northwestward. Previous tion and high coverage of azimuthal convection were maintained inside studies have noted increasing lightning rates accompanying or pre- the RMW. The location of the inner-core CBs relative to the RMW (r/ ceding an imminent change in TC track. Black and Hallett (1999) first RMW) was 0.9. Deep convection moved gradually inward as the storm suggested that lightning activity could be an indicator of TC track rapidly intensified (Fig. 4). Haiyan encountered moderate vertical shear changes. Molinari et al. (1999) found in Hurricane Bob (1991) an − (7.8 m s 1) during this RI period, with CBs and strong updrafts (in- outbreak of eyewall lightning occurred 3 h before the change of its dicated by lightning) appearing more systematically in the downshear- direction and motion speed. Molinari et al. (2006) found the in- left quadrant (Fig. 10a). The favorable environmental conditions, pre- tensification and track fluctuations of Tropical Storm Gabrielle (2001) sence of deep convection within the RMW, and contraction of the were initiated by wind shear in an environment with upward motion. eyewall allowed RI to be maintained for a long period of time (66 h). Such shear-induced changes are important in storms approaching the This prolonged period of intensification eventually allowed Haiyan to coastlines. In a study of northwest Pacific landfalling TCs, Zhang et al. become a category-5-equivalent super typhoon. (2012) also observed that lightning outbreaks in the eyewall were as- The second episode of intense inner-core convection and lightning sociated with changes in storm track. In tropical storms or weak hur- outbreak occurred just after the new eyewall formed (ERC) when the ricanes, outbreaks of asymmetric convection usually mark the start of storm reached its peak intensity. The shear at the time was overall quiescence. Asymmetric convection erupting near the center of weak − marginal (3.2 m s 1). CBs were more evident and more symmetrically hurricanes may cause the cyclone tracks to become irregular and cy- distributed around the center. Lightning rates were more frequent clone centers to move erratically (Willoughby, 1990). Numerical study within the inner core during the MI stage than during the RI stage. showed that the motion effect arises from asymmetric frictional forcing

133 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Fig. 8. As Fig. 7, but for the ERC and MI stages. The inner-core lightning outbreak during this stage (OB2) began at 1200 UTC and ended at 1700 UTC 7 Nov.

134 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Fig. 9. As Fig. 7, but for the weakening stages. The inner-core lightning outbreak during this stage (OB3) began at 1400 UTC and ended at 2200 UTC 8 Nov. in the boundary layer and could force an asymmetry in the distribution 2011; Zhang et al., 2012), the ratio in lightning density between the of convection (Shapiro, 1983). Using lightning data from the National inner core and outer rainbands in Haiyan was 20:1. The temporal Lightning Detection Network (NLDN), Corbosiero and Molinari (2003) evolution of lightning rates showed that there were three outbreaks of examined the effects of wind shear and storm motion on convective inner-core lightning during the life cycle (Fig. 2). An examination of asymmetries in 35 Atlantic TCs. The study found that wind shear plays global analysis fields (Fig. 3) revealed lightning outbreaks within the a greater role in determining the distribution of TC lightning. The in- inner core developed under large-scale environmental conditions with fluence of asymmetric friction has little impact on convective asym- high SST (29–30 °C), high RH (75–80%), and weak to moderate deep- − metries unless wind shear is insignificant. Whether it is possible to use layer wind shear (3–8ms 1). lightning data as an indicator of TC track changes is an issue worthy of The first outbreak (RI outbreak) occurred during the RI period when further study. the storm's intensity increased from tropical storm to typhoon intensity. A second outbreak (MI outbreak) occurred at the time of the storm's MI, just after the ERC period. The final outbreak (weakening outbreak) 6. Conclusions occurred during the weakening stage when the storm crossed the central Philippines. The radial distribution of lightning indicated an inward Using a combination of lightning observations, microwave and in- shift in the lightning density maximum, and a preference for the frared satellite data, and global atmospheric analysis, this study in- downshear region, during the RI and MI outbreaks and an outward shift vestigated the environmental conditions, lightning outbreak char- and preference for the upshear–left quadrant during the weakening acteristics, and convective evolution within the inner core of Super outbreak (Figs. 4 and 5). Time series of satellite signatures showed that Typhoon Haiyan that occurred on November 2013. A significant pro- high lightning rates were associated with low 91-GHz PCT, and all three portion (49%) of lightning occurred within the inner core. Contrary to outbreaks occurred along with sharp decreases in the median TBB in the previous findings that the largest lightning density typically occurs in inner core (Fig. 6). the outer rainbands (e.g., Molinari et al., 1994, 1999; Abarca et al.,

135 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Fig. 10. Summary graphics highlighting the radial and azimuthal location of inner-core deep convection in Haiyan for (a) rapid intensiﬁcation, (b) maximum intensity, and (c) weakening stages. The SSMIS 91-GHz polarization corrected temperatures (K, color shading), radius of maximum wind (dashed black line), and the vertical wind shear (vector) are the median values of all overpasses at each stage. The thick black circles mark the radii of 100 km from − − − the storm center. Shear values are 7.8 m s 1, 3.2 m s 1 7.2 m s 1 for the RI, MI, and weakening stages, respectively.

CBs during the RI stage were located within the RMW in the downshear quadrant, and moved gradually closer to the center as the storm intensified. The favorable environmental conditions, location of deep convection, and the contraction of the eyewall combined to pro- long the RI stage for 66 h. The shear decreased to a weak magnitude − (< 5 m s 1) in the MI period when the second lightning outbreak occurred. CBs were apparent and lightning were more frequent during the MI stage than during the RI period. A symmetric distribution of very low brightness temperatures (< 200 K) (Fig. 8) due to increased ice scattering expanded around the inner core, suggesting that the coverage and strength of intense convection was greater after the ERC than before. The final CB was situated with an upshear–left orientation, in contrast to earlier bursts during the RI and MI stages. The asymmetric structure, the location of the deep convection outside the RMW, and the strong wind shear led to the weakening of the storm. In addition to the wind shear, interaction with the underlying land surface could also have contributed to the final outbreak of CBs and lightning in the inner core. The possible role of wind shear in the formation of lightning outbreak during the weakening stage was investigated. A sharp change in the orientation of the deep-layer wind shear was present at the time of the weakening outbreak, resulting in a decrease in the angle between the shear and motion vectors (Fig. 11). Haiyan changed direction just after the final outbreak during the weakening stage, consistent with previous studies that indicate the potential use of lightning data in the forecasting of track changes due to TC structure changes. To date a clear/systematic relationship between lightning activity, especially inner-core lightning and TC intensity changes has not yet been established. Several studies found that TC intensification was preceded or accompanied by an increase in inner-core lightning activity (Squires and Businger, 2008; Abarca et al., 2011; Stevenson et al., 2014; Susca-Lopata et al., 2015; Zhang et al., 2015). Others found a negative relationship and hypothesized lightning flash rate in the outer rainbands was more likely to increase prior to TC intensification (DeMaria et al., 2012; Stevenson et al., 2016; Xu et al., 2017). One possible reason for this inconsistency may be related to the different DEs of each of the networks used. The regional ground lightning detection networks (e.g., NLDN, LLDN) are confined to land and coastline with a high DE of CG lightning. The WWLLN, a global ground lightning detection network, detects both IC and CG lightning at a relatively low DE and most of the detected lightning are CG strokes. Total lightning can be measured with a high DE by LIS, but the observational region is temporally confined to a 90-s snapshot. In this study, using the WWLLN data, no correlation was found between inner-core lightning bursts and intensity changes in Super Typhoon Haiyan (2013), i.e., inner-core lightning bursts could occur during both the intensification and the weakening phases. However, it is worth noting that, the inner-core lightning bursts that occurred inside the RMW and downshear were preferentially associated with the RI of Haiyan. This is consistent with the findings by Stevenson et al. (2018) for a 10-year survey of inner-core lightning bursts in TCs in North Atlantic and eastern North Pacific basins. In this study, we found that lightning flash rates in the rainbands were much lower than or similar to those in the inner core during all three main outbreaks (Fig. 2). Thus, for this super typhoon, CG lightning burst in the rainband could not have been used to diagnose intensity changes contrary to the findings of DeMaria et al. (2012). Recent

136 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

Fig. 11. Vertical wind shear direction, storm motion, and the angle between the shear and motion vectors. Shear and motion angles are measured in the meteorological direction, and the angle between the shear and motion vectors is measured counter- clockwise from the vertical shear to the motion vector. Gray shadings indicate three inner-core lightning outbreaks during the RI, MI, and weakening stages, respectively.

works documented potential differences in the respective behavior of IC Lightning Location Network (WWLLN) using the National Lightning Detection and CG lightning flashes between the inner core and outer rainbands of Network (NLDN) as ground truth. J. Geophys. Res. 115, D18206. https://doi.org/10. ffi fl 1029/2009JD013411. TCs. Gri n et al. (2014) reported a higher ratio of IC ashes within the Abarca, S.F., Corbosiero, K.L., Vollaro, D., 2011. The World Wide Lightning Location eyewall compared to that in the outer rainbands in Tropical Storm Erin Network and convective activity in tropical cyclones. Mon. Weather Rev. 139, (2007). Xu et al. (2017) found total lightning density was much lower 175–191. Black, A.R., Hallett, J., 1999. Electrification of the hurricane. J. Atmos. Sci. 56, in the inner core than in the outer rainbands in RI TCs, suggesting a 2004–2028. lower IC:CG ratio in the inner core of RI TCs. A more recent study by Boccippio, D.J., Cummins, K.L., Christian, H.J., Goodman, S.J., 2001. Combined satellite- Fierro et al. (2018) further pointed out the difference of lightning types and surface-based estimation of the intracloud–cloud-to-ground lightning ratio over – between the inner core and outer rainbands in Hurricane Maria (2017). the continental United States. Mon. Weather Rev. 129, 108 122. Bovalo, C., Barthe, C., Yu, N., Bègue, N., 2014. Lightning activity within tropical cyclones They suggested that the IC:CG ratio differs between the two regions and in the South West Indian Ocean. J. Geophys. Res. Atmos. 119, 8231–8244. the inner core may have a larger fraction of IC flashes. Therefore, the Braun, S.A., Montgomery, M.T., Pu, Z., 2006. High-resolution simulation of Hurricane combination of CG-only and total lightning measurements is antici- Bonnie (1998). Part I: the organization of eyewall vertical motion. J. Atmos. Sci. 63, 19–42. pated to provide a more complete depiction of convective vigor and Cecil, D.J., Zipser, E.J., 1999. Relationship between tropical cyclone intensity and sa- updraft enhancement within TCs. The continuous, high DE coverage of tellite-based indicators of inner core convection: 85-GHz ice-scattering signature and – total lightning from spaceborne instruments such as the GLM aboard lightning. Mon. Weather Rev. 127, 103 123. Cecil, D.J., Zipser, E.J., 2002. Reflectivity, ice scattering, and lightning characteristics of GOES-16 or the LMI on FY-4A, could help researchers and forecasters to hurricane eyewalls and rainbands. Part II: Intercomparison of observation. Mon. better understand the evolution of deep, electrified convection within Weather Rev. 130, 785–801. TCs and, in turn, establish more meaningful relationships between Cecil, D.J., Zipser, E.J., Nesbitt, S.W., 2002. Reflectivity, ice scattering, and lightning characteristics of hurricane eyewalls and rainbands. Part I: Quantitative description. lightning activity and TC intensity changes. Mon. Weather Rev. 130, 769–784. Corbosiero, K.L., Molinari, J., 2002. The effects of vertical wind shear on the distribution Acknowledgments of convection in tropical cyclones. Mon. Weather Rev. 130, 2110–2123. Corbosiero, K.L., Molinari, J., 2003. The relationship between storm motion, vertical wind shear, and convective asymmetries in tropical cyclones. J. Atmos. Sci. 60, The authors wish to thank the World Wide Lightning Location 366–376. Network (http://wwlln.net) personnel for providing the lightning lo- DeMaria, M., DeMaria, R.T., Knaff, J.A., Molenar, D., 2012. Tropical cyclone lightning – cation data used in this study. The GLD360 lightning data were ob- and rapid intensity change. Mon. Weather Rev. 140, 1828 1842. Demetriades, N.W.S., Murphy, M.J., Cramer, J.A., 2010. Validation of Vaisala's Global tained from the Vaisala Global Lightning Dataset. Best-track data were Lightning Dataset (GLD360) over the continental United States. In: 29th Conf. On obtained from the Joint Typhoon Warning Center, satellite data from Hurricanes and Tropical Meteorology, Tuscon, AZ, Amer. Meteor. Soc, pp. 2. fi the Multifunctional Transport Satellite and NOAA's National Centers for Fierro, A.O., Mansell, E.R., 2017. Electri cation and lightning in idealized simulations of a hurricane-like vortex subject to wind shear and sea surface temperature cooling. J. Environmental Information, and global analysis data from the National Atmos. Sci. 74, 2023–2041. Centers for Environmental Prediction. The eyewall replacement cycle Fierro, A.O., Mansell, E.R., 2018. Relationships between electrification and storm-scale properties based on idealized simulations of an intensifying hurricane-like vortex. J. was determined from Morphed Integrated Microwave Imagery products – – Atmos. Sci. 75, 657 674. from University of Wisconsin Madison. We also thank the reviewers for Fierro, A.O., Reisner, J.M., 2011. High-resolution simulation of the electrification and their helpful comments on the manuscript. This work was supported by lightning of Hurricane Rita during the period of rapid intensification. J. Atmos. Sci. the National Key Research and Development Program of China (Grant 68, 477–494. Fierro, A.O., Gilmore, M.S., Mansell, E.R., Wicker, L.J., Straka, J.M., 2006. Electrification 2017YFC1501502), the National Natural Science Foundation of China, and lightning in an idealized boundary-crossing supercell simulation of 2 June 1995. China (Grant 41405004), and NASA award NNX16AD85G to Colorado Mon. Weather Rev. 134, 3149–3172. State University. Fierro, A.O., Leslie, L.M., Mansell, E.R., Straka, J.M., MacGorman, D.R., Ziegler, C., 2007. A high-resolution simulation of the microphysics and electrification in an idealized hurricane-like vortex. Meteor. Atmos. Phys. 98, 13–33. References Fierro, A.O., Shao, X.-M., Hamlin, T., Reisner, J.M., Harlin, J., 2011. Evolution of eyewall convective events as indicated by intracloud and cloud-to-ground lightning activity during the rapid intensification of hurricanes Rita and Katrina. Mon. Weather Rev. Abarca, S.F., Corbosiero, K.L., Galarneau Jr., T.J., 2010. An evaluation of the Worldwide

137 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

139, 1492–1504. Petersen, W.A., Cifelli, R.C., Rutledge, S.A., Ferrier, B.S., Smull, B.F., 1999. Shipborne Fierro, A.O., Mansell, E.R., Ziegler, C., MacGorman, D.R., 2013. The implementation of an dual-Doppler operations during TOGA COARE: Integrated observations of storm ki- explicit charging and discharge lightning scheme within the WRF-ARW Model: nematics and electrification. Bull. Am. Meteorl. Soc. 80, 81–97. benchmark simulations with a continental squall line, a tropical cyclone, and a winter Price, C., Asfur, M., Yair, Y., 2009. Maximum hurricane intensity preceded by increase in storm. Mon. Weather Rev. 141, 2390–2415. lightning frequency. Nat. Geosci. 2, 329–332. Fierro, A.O., Mansell, E.R., MacGorman, D.R., Ziegler, C., 2015. Explicitly simulated Ranalkar, M.R., Pawar, S.D., Kumar, P.P., 2016. Characteristics of lightning activity in electrification and lightning within a tropical cyclone based on the environment of tropical cyclones developed over North Indian Ocean basin during 2010–2015. Hurricane Isaac (2012). J. Atmos. Sci. 72, 4167–4193. Atmos. Res. https://doi.org/10.1016/j.atmosres.2016.12.003. Fierro, A.O., Stevenson, S.N., Rabin, R.M., 2018. Evolution of GLM-observed total light- Reinhart, B., Fuelberg, H., Blakeslee, R., Mach, D., Heymsfield, A., Bansemer, A., Durden, ning in Hurricane Maria (2017) during the period of maximum intensity. Mon. S.L., Tanelli, S., Heymsfield, G., Lambrigtsen, B., 2014. Understanding the relation- Weather Rev. 146, 1641–1666. ship between lightning, cloud microphysics, and airborne radar-derived storm Goodman, S.J., et al., 2013. The GOES-R Geostationary Lightning Mapper (GLM). Atmos. structure during Hurricane Karl (2010). Mon. Weather Rev. 142, 590–605. Res. 125–126, 34–49. Rodger, C.J., Brundell, J.B., Dowden, R.L., Thomson, N.R., 2004. Location accuracy of Griffin, E.M., Schuur, T.J., MacGorman, D.R., Kumjian, M.R., Fierro, A.O., 2014. An long distance VLF lightning location network. Ann. Geophys. 22, 747–758. electrical and polarimetric analysis of the overland reintensification of Tropical storm Rogers, R.F., Reasor, P.D., Lorsolo, S., 2013. Airborne Doppler observations of the inner- Erin (2007). Mon. Weather Rev. 142, 2321–2344. core structural differences between intensifying and steady-state tropical cyclones. Guimond, S.R., Heymsfield, G.M., Turk, F.J., 2010. Multiscale observations of Hurricane Mon. Weather Rev. 141, 2970–2991. Dennis (2005): the effects of hot towers on rapid intensification. J. Atmos. Sci. 67, Rogers, R.F., Reasor, P.D., Zhang, J., 2015. Multiscale structure and evolution of 633–654. Hurricane Earl (2010) during rapid intensification. Mon. Weather Rev. 143, 536–562. Hanley, D., Molinari, J., Keyser, D., 2001. A composite study of the interactions between Rogers, R.F., Zhang, J.A., Zawislak, J., Jiang, H., Alvey, G.R.I.I.I., Zipser, E.J., Stevenson, tropical cyclones and upper-tropospheric troughs. Mon. Weather Rev. 129, S.N., 2016. Observations of the structure and evolution of Hurricane Edouard (2014) 2570–2584. during intensity change. Part II: Kinematic structure and the distribution of deep Hazelton, A.T., Hart, R.E., Rogers, R.F., 2017. Analyzing simulated convective bursts in convection. Mon. Weather Rev. 144, 3355–3376. two Atlantic hurricanes. Part II: Intensity change due to bursts. Mon. Weather Rev. Rudlosky, S.D., Shea, D.T., 2013. Evaluating WWLLN performance relative to TRMM/LIS. 145, 3095–3117. Geophys. Res. Lett. 40, 2344–2348. Hendricks, E.A., 2012. Internal dynamical control on tropical cyclone intensity varia- Rudlosky, S.D., Goodman, S.J., Virts, K.S., Bruning, E.C., 2018. Initial Geostationary bility. Trop. Cyclone Res. Rev. 1, 97–105. Lightning Mapper Observations. Geophys. Res. Lett. https://doi.org/10.1029/ Houze, R.A.Jr., Marks, F.D.Jr., Black, R.A., 1992. Dual aircraft investigation of the inner 2018GL081052. core of Hurricane Norbert. Part II: Mesoscale distribution of ice particles. J. Atmos. Said, R.K., Inan, U.S., Cummins, K.L., 2010. Long-range lightning geolocation using a VLF Sci. 49, 943–962. radio atmospheric waveform bank. J. Geophys. Res. 115, D23108. https://doi.org/ Houze, R.A.Jr., Chen, S.S., Smull, B.F., Lee, W.-C., Bell, M., 2007. Hurricane intensity and 10.1029/2010JD013863. eyewall replacement. Science 315, 1235–1239. Saunders, C.P.R., Peck, L.S., 1998. Laboratory studies of the in fluence of the rime ac- Hutchins, M.L., Holzworth, R.H., Virts, K.S., Wallace, J.M., Heckman, S., 2013. Radiated cretion rate on charge transfer during crystal/graupel collisions. J. Geophys. Res. VLF energy differences of land and oceanic lightning. Geophys. Res. Lett. 40, 103, 13949–13956. 2390–2394. Saunders, C.P.R., Peck, L.S., Varela, G.G.A., Avila, E.E., Castellano, N.E., 2001. A la- Jacobson, A.R., Holzworth, R., Harlin, J., Dowden, R., Lay, E., 2006. Performance as- boratory study of the influence of water vapour and mixing on the charge transfer sessment of the Worldwide Lightning Location Network (WWLLN) using the Los process during collisions between ice crystal and graupel. Atmos. Res. 58, 187–203. Alamos Sferic Array (LASA) as ground truth. J. Atmos. Ocean. Technol. 23, Schultz, C.J., Petersen, W.A., Carey, L.D., 2011. Lightning and severe weather: a com- 1082–1092. parison between total and cloud-to-ground lightning trends. Weather Forecasting 26, Jiang, H., 2012. The relationship between tropical cyclone intensity change and the 744–755. strength of inner-core convection. Mon. Weather Rev. 140, 1164–1176. Shao, X.M., Harlin, J., Stock, M., Stanley, M., Regan, A., Wiens, K., Hamlin, T., Pongratz, Kaplan, J., DeMaria, M., 2003. Large-scale characteristics of rapidly intensifying tropical M., Suszcynsky, D., Light, T., 2005. Katrina and Rita were lit up with lightning. Eos cyclones on the North Atlantic basin. Weather Forecasting 18, 1093–1108. Trans. AGU 86 (42), 398. Kaplan, J., DeMaria, M., Knaff, J.A., 2010. A revised tropical cyclone rapid intensification Shapiro, L.J., 1983. The asymmetric boundary layer flow under a translating hurricane. J. index for the Atlantic and eastern North Pacific basins. Weather Forecasting 25, Atmos. Sci. 40, 1984–1998. 220–241. Shimada, U., Kubota, H., Yamada, H., Cayanan, E.O., Hilario, F.D., 2018. Intensity and Kuhlman, K.M., Ziegler, C.L., Mansell, E.R., MacGorman, D.R., Straka, J.M., 2006. inner-core structure of Typhoon Haiyan (2013) near landfall: Doppler radar analysis. Numerically simulated electrification and lightning of the 29 June 2000 STEPS su- Mon. Weather Rev. 146, 583–597. percell storm. Mon. Weather Rev. 134, 2734–2757. Shu, S., Zhang, F., 2015. Influence of equatorial waves on the genesis of Super Typhoon Lander, M., Guard, C., Camargo, S.J., 2014. Super Typhoon Haiyan [in “State of the Haiyan (2013). J. Atmos. Sci. 72, 4591–4613. Climate in 2013”]. Bull. Am. Meteor. Soc. 95, S112–S114. Sitkowski, M., Kossin, J.P., Rozoff, C.M., Knaff, J., 2012. Hurricane eyewall replacement Lang, T.J., Rutledge, S.A., 2002. Relationships between convective storm kinematics, cycles and the relict inner eyewall circulation. Mon. Weather Rev. 140, 4035–4045. precipitation, and lightning. Mon. Weather Rev. 130, 2492–2506. Solorzano, N.N., Thomas, J.N., Hutchins, M.L., Holzworth, R.H., 2016. WWLLN lightning Liu, C., Cecil, D., Zipser, E.J., 2011. Relationships between lightning flash rates and and satellite microwave radiometrics at 37 to 183 GHz: Thunderstorms in the broad passive microwave brightness temperatures at 85 and 37 GHz over the tropics and tropics. J. Geophys. Res. Atmos. 121, 12,298–12,318. subtropics. J. Geophys. Res. Atmos. 116, D23108. https://doi.org/10.1029/ Spencer, R.W., Goodman, H.M., Hood, R.E., 1989. Precipitation retrieval over land and 2011JD016463. ocean with the SSM/I: Identification and characteristics of the scattering signal. J. Lyons, W.A., Keen, C.S., 1994. Observations of lightning in convective supercells within Atmos. Ocean. Technol. 6, 254–273. tropical storms and hurricanes. Mon. Weather Rev. 122, 1897–1916. Squires, K., Businger, S., 2008. The morphology of eyewall lightning outbreaks in two MacGorman, D.R., Burgess, D.W., Mazur, V., Rust, W.D., Taylor, W.L., Johnson, B.C., category 5 hurricanes. Mon. Weather Rev. 136, 1706–1726. 1989. Lightning rates relative to tornadic storm evolution on 22 May 1981. J. Atmos. Stevenson, S.N., Corbosiero, K.L., Molinari, J., 2014. The convective evolution and rapid Sci. 46, 221–251. intensification of Hurricane Earl (2010). Mon. Weather Rev. 142, 4364–4380. Marks, F.D., Shay, L.K., 1998. Landfalling tropical cyclones: Forecast problems and as- Stevenson, S.N., Corbosiero, K.L., Abarca, S.F., 2016. Lightning in eastern North Pacific sociated research opportunities. Bull. Am. Meteor. Soc. 79, 305–323. tropical cyclones: a comparison to the North Atlantic. Mon. Weather Rev. 144, Medici, G., Cummins, K.L., Cecil, D.J., Koshak, W.J., Rudlosky, S.D., 2017. The intracloud 225–239. lightning fraction in the contiguous United States. Mon. Weather Rev. 145, Stevenson, S.N., Corbosiero, K.L., DeMaria, M., Vigh, J., 2018. A 10-year survey of tro- 4481–4499. pical cyclone inner-core lightning bursts and their relationship to intensity change. Molinari, J., Vollaro, D., 2010. Rapid intensification of a sheared tropical storm. Mon. Weather Forecasting 33, 23–36. Weather Rev. 138, 3869–3885. Susca-Lopata, G., Zawislak, J., Zipser, E.J., 2015. The role of observed environmental Molinari, J., Moore, P., Idone, V., Henderson, R.W., Saljoughy, A.B., 1994. Cloud-to- conditions and precipitation evolution in the rapid intensification of Hurricane Earl ground lightning in Hurricane Andrew. J. Geophys. Res. 99 (D8), 16,665–16,676. (2010). Mon. Weather Rev. 143, 2207–2223. Molinari, J., Moore, P., Idone, V., 1999. Convective structure of hurricanes as revealed by Takahashi, T., 1978. Riming electrification as a charge generation mechanism in thun- lightning locations. Mon. Weather Rev. 127, 520–534. derstorms. J. Atmos. Sci. 35, 1536–1548. Molinari, J., Vollaro, D., Corbosiero, K.L., 2004. Tropical cyclone formation in a sheared Tao, C., Jiang, H., 2015. Distributions of shallow to very deep precipitation–convection in environment: a case study. J. Atmos. Sci. 61 (21), 2493–2509. rapidly intensifying tropical cyclones. J. Clim. 28, 8791–8824. Molinari, J., Dodge, P., Vollaro, D., Corbosiero, K.L., 2006. Mesoscale aspects of the Thomas, J.N., Solorzano, N.N., Cummer, S.A., Holzworth, R.H., 2010. Polarity and en- downshear reformation of a tropical cyclone. J. Atmos. Sci. 63, 341–354. ergetics of inner core lightning in three intense North Atlantic hurricanes. J. Geophys. Mueller, K.J., DeMaria, M., Knaff, J.A., Kossin, J.P., Haar, T.H. Vonder, 2006. Objective Res. 115, A00E15. https://doi.org/10.1029/2009JA014777. estimation of tropical cyclone wind structure from infrared satellite data. Weather Virts, K.S., Wallace, J.M., Hutchins, M.l., Holzworth, R.H., 2013. Highlights of a new Forecasting 21, 990–1005. ground-based, hourly global lightning climatology. Bull. Am. Meteor. Soc. 94, National Disaster Risk Reduction and Management Council, 2014. Effects of Typhoon 1381–1391. ‘Yolanda’ (Haiyan). SitRep 107, 62. Wadler, J.B., Rogers, R.F., Reasor, P.D., 2018. The relationship between spatial variations Pan, L., Qie, X., Liu, D., Wang, D., Yang, J., 2010. The lightning activities in super ty- in the structure of convective bursts and tropical cyclone intensification as de- phoons over the Northwest Pacific. Sci China Earth Sci. 53, 1241–1248. termined by airborne doppler radar. Mon. Weather Rev. 146, 761–780. Pan, L., Qie, X., Wang, D., 2014. Lightning activity and its relation to the intensity of Wang, F., Qie, X., Liu, D., Shi, H., Srivastava, A., 2016. Lightning activity and its re- typhoons over the Northwest Pacific Ocean. Adv. Atmos. Sci. 31 (3), 581–592. lationship with typhoon intensity and vertical wind shear for Super Typhoon Haiyan

138 W. Zhang et al. Atmospheric Research 219 (2019) 123–139

(1330). J. Meterolr. Res. 30 (1), 117–127. Yang, J., Zhang, Z., Wei, C., Lu, F., Guo, Q., 2017. Introducing the new generation of Williams, E.R., 1989. The tripole structure of thunderstorms. J. Geophys. Res. 94, Chinese geostationary weather satellites, FengYun-4. Bull. Am. Meteorol. Soc. 98, 13151–13167. 1637–1658. Williams, E.R., Geotis, S.G., Renno, N., Rutledeg, S.A., Rasmussen, E., Rickenbach, T., Zawislak, J., Jiang, H., Alvey, G.R.I.I.I., Zipser, E.J., Rogers, R.F., Zhang, J.A., Stevenson, 1992. A radar and electrical study of tropical “hot towers”. J. Atmos. Sci. 49 (15), S.N., 2016. Observations of the structure and evolution of Hurricane Edouard (2014) 1386–1395. during intensity change. Part I: Relationship between the thermodynamic structure Willoughby, H.E., 1990. Temporal changes of the primary circulation in tropical cyclones. and precipitation. Mon. Weather Rev. 144, 3333–3354. J. Atmos. Sci. 47, 242–264. Zhang, W., Zhang, Y., Zheng, D., Zhou, X., 2012. Lightning distribution and eyewall Willoughby, H.E., Clos, J.A., Shoreibah, M.G., 1982. Concentric eyewalls, secondary wind outbreaks in tropical cyclones during landfall. Mon. Weather Rev. 140, 3573–3586. maxima, and the evolution of the hurricane vortex. J. Atmos. Sci. 39, 395–411. Zhang, W., Zhang, Y., Zheng, D., Wang, F., Xu, L., 2015. Relationship between lightning Wimmers, A.J., Velden, C.S., 2007. MIMIC: a new approach to visualizing satellite mi- activity and tropical cyclone intensity over the Northwest Paciﬁc. J. Geophys. Res. crowave imagery of tropical cyclones. Bull. Am. Meteor. Soc. 88, 1187–1196. Atmos. 120, 4072–4089. Xu, W., Rutledge, S.A., Zhang, W., 2017. Relationship between total lightning, deep Zhang, W., Zhang, Y., Zheng, D., Xu, L., Lyu, W., 2018. Lightning climatology over the convection, and tropical cyclone intensity change. J. Geophys. Res. Atmos. 122, Northwest Paciﬁc region: an 11-year study using data from the World Wide Lightning 7047–7063. Location Network. Atmos. Res. 210, 41–57.

139