2838 MONTHLY WEATHER REVIEW VOLUME 142

Origin and Maintenance of the Long-Lasting, Outer Mesoscale Convective System in (2008)

BUO-FU CHEN Department of Atmospheric Sciences, National University, Taipei, Taiwan

RUSSELL L. ELSBERRY Department of Meteorology, Naval Postgraduate School, Monterey, California

CHENG-SHANG LEE Department of Atmospheric Sciences, National Taiwan University, and Taiwan Typhoon and Flood Research Institute, National Applied Research Laboratories, Taipei, Taiwan

(Manuscript received 27 January 2014, in final form 21 March 2014)

ABSTRACT

Outer mesoscale convective systems (OMCSs) are long-lasting, heavy rainfall events separate from the inner-core rainfall that have previously been shown to occur in 22% of western North Pacific tropical cyclones (TCs). Environmental conditions accompanying the development of 62 OMCSs are contrasted with the conditions in TCs that do not include an OMCS. The development, kinematic structure, and maintenance mechanisms of an OMCS that occurred to the southwest of Typhoon Fengshen (2008) are studied with Weather Research and Forecasting Model simulations. Quick Scatterometer (QuikSCAT) observations and the simulations indicate the low-level TC circulation was deflected around the terrain and caused an elongated, north–south moisture band to be displaced to the west such that the OMCS develops in the outer region of Fengshen rather than spiraling into the center. Strong northeasterly vertical wind shear contributed to frictional convergence in the boundary layer, and then the large moisture flux convergence in this moisture band led to the downstream development of the OMCS when the band interacted with the monsoon flow. As the OMCS developed in the region of low-level monsoon westerlies and midlevel northerlies associated with the outer circulation of Fengshen, the characteristic structure of a rear-fed inflow with a leading stratiform rain area in the cross-line direction (toward the south) was established. A cold pool (Du ,23 K) associated with the large stratiform precipitation region led to continuous formation of new cells at the leading edge of the cold pool, which contributed to the long duration of the OMCS.

1. Introduction August 2009 due to Typhoon Morakot has been related to the interaction between the tropical cyclone (TC) While the heavy rainfall associated with the eyewall circulation and the southwest monsoon (Chien and Kuo region of a tropical cyclone is a primary focus for flood 2011; Lee et al. 2011). Lee et al. (2011) examined several forecasting, long-lasting heavy rainfall may also occur in factors involved in the Morakot disaster, such as the outer regions. For example, Typhoon Morakot (2009), moist and unstable air brought by the southwest mon- which was the deadliest typhoon to impact Taiwan in soon, steep topography that provided rapid lifting, and recorded history, produced record-breaking rainfall the slow movement of Morakot. .3000 mm well to the south of the center. The record One of the great challenges in forecasting the Morakot accumulated rainfall over southern Taiwan during 6–10 rainfall was the contribution from an east–west-oriented, quasi-stationary, and long-lasting convective band over the about 300 km south of the TC center. Corresponding author address: Prof. Cheng-Shang Lee, De- partment of Atmospheric Sciences, National Taiwan University, Notably, a mesoscale convective system (MCS) developed No.1, Sec. 4, Roosevelt Road, Taipei 10617, Taiwan. within this rainband (Fig. 1a), and the subsequent inter- E-mail: [email protected] action of the rainband with the steep terrain produced

DOI: 10.1175/MWR-D-14-00036.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2839

FIG. 1. Infrared satellite images of OMCSs embedded in the outer circulations of (a) Typhoon Morakot (2009), (b) Typhoon Mindulle (2004), (c) Typhoon Kalmaegi (2008), (d) (2008), and (e),(f) Typhoon Fengshen (2008). Thick red 3s indicate the TC centers and the three white circles indicate radii of 150, 450, and 750 km. extremely intense rainfall of approximately 1500 mm region. Predicting rainfall due to OMCSs when they are from 1200 UTC 8 August to 0300 UTC 9 August 2009. also interacting with topography is a great challenge. Because of its long duration and the orographic en- Consequently, understanding the initiation processes, hancement, this MCS accounted for a substantial frac- kinematic structure, and the maintenance of OMCSs is tion of the total precipitation during the slow passage of important. Morakot. Numerous studies (Willoughby et al. 1982, 1984; Lee et al. (2012) defined outer MCSs (OMCSs) as Barnes et al. 1983; Powell 1990b; May and Holland 1999; convective systems that develop in a distant rainband of Wang 2002, 2009; Moon and Nolan 2010) have shown a TC, have a large cold cloud shield (area of the 208-K that the TC rainbands play an important role in the cold cloud shield must exceed 72 000 km2), and persist rainfall distribution, dynamics, size, and intensity of for more than 6 h. Based on hourly infrared channel-1 TCs. Houze (2010) categorized the rainband complex of (IR1) cloud-top temperatures and passive microwave TCs as consisting of a principal rainband, secondary (PMW) images, Lee et al. (2012) documented 109 OMCSs rainbands, and distant rainbands. A principal rainband in 22% of the TCs that occurred from 1999 to 2009 in the may develop due to the convergence between the vortex western North Pacific. In addition to Typhoon Morakot flow and the environment (Willoughby et al. 1984). (Fig. 1a), other OMCSs such as in Typhoons Mindulle Several studies (Barnes et al. 1983; Willoughby et al. (2004; Fig. 1b), Bilis (2006), and Kalmaegi (2008; Fig. 1c) 1984; Powell 1990a; Hence and Houze 2008; Didlake have hit Taiwan and produced ‘‘unexpected’’ torrential and Houze 2013a,b) based on aircraft observations have rainfall because they occurred remote from the inner-core shown that the cloud structure in the upwind portion of

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2840 MONTHLY WEATHER REVIEW VOLUME 142 the principal rainband is more convective, but the clouds in the downwind portion typically consist of decaying convective cells and tend to be dominated by stratiform precipitation. These studies also reported that an overturning cir- culation with inflows originating from the convex (outer) side is associated with the principal rainband, and the convective cells are distributed near the concave (inner) side of the rainband axis. Furthermore, a sec- ondary horizontal wind maximum is often observed in the midlevels and along the principal rainband (Willoughby et al. 1984; May et al. 1994; Samsury and Zipser 1995; Hence and Houze 2008). However, Ishihara et al. (1986) and Tabata et al. (1992) have shown inflows that origi- nate from the concave side, and Li and Wang (2012) have shown that convective cells may develop on the convex side of the spiral band. Whereas the major portion of the principal rainband and secondary rainband are located in the inner-core region of the TC, the distant rainbands develop in the ‘‘outer region.’’ Cecil and Zipser (2002) suggested that the outer rainband region typically begins from about 150 to 200 km from the cyclone center and is typically bounded on the inside by a precipitation-free lane ad- jacent to an inner rainband. Cecil and Zipser (2002) defined a minimum outer rainband radius of 100 km and a maximum radius of 350 km. Corbosiero and Molinari (2002, 2003) defined ‘‘outer band regions’’ as being 100– 300 km from the center of hurricanes. Houze (2010, p. 324) stated that ‘‘distant rainbands are composed of buoyant convective cells aligned with confluence lines in the large-scale, low-level wind field spiraling into the TC vortex and are radially far enough from the eye of the FIG. 2. (a) Passive microwave image as observed by the SSMIS and (b) QuikSCAT oceanic surface wind observations for the OMCS storm that the vertical structure of the convection within embedded in Typhoon Fengshen (2008). Thick black lines indicate them is relatively unconstrained by the dynamics of the the contour of zero relative vorticity based on the QuikSCAT wind inner-core vortex of the cyclone.’’ observations, and red dashed lines indicate the 2758Ccoldcloud The OMCS in this study is another type of convective shield of the outer MCSs from the IR1 image at 1000 UTC 22 Jun system that occurs in the outer region of some western 2008; (a) is a modified version from a Naval Research Labora- tory (NRL) TC_PAGES website. North Pacific TCs (Lee et al. 2012). Compared with inner-core rainbands and typical distant bands described above, these OMCSs typically have larger stratiform indicated the presence of a surface wind jet under the precipitation regions than those in typical rainbands. stratiform region, which is indicated by a zero relative Moreover, the growth of the stratiform precipitation vorticity line (Fig. 2b, thick black line). Note also that region is typically accompanied by a surface wind jet the convective cells are on the cyclonic shear side of (Lee et al. 2012). Specifically, the OMCS (Figs. 1f, 2) the jet. that developed within the outer circulation of Typhoon While the Fengshen OMCS was selected in part be- Fengshen (2008) had a large stratiform precipitation cause it occurred within a region of synoptic-scale ob- region with a moderate (215–230 K) PMW 91-GHz servations around the South Sea, observations polarization-corrected temperature (PCT) brightness were not available to analyze the mesoscale features of temperature TB and a convective precipitation region the OMCS or the mechanism(s) that maintains the (approximately the area of very low PMW TB , 215 K) convection for the extended duration of the OMCS. In with linearly arranged convective cells (Fig. 2a). Fur- this study, Weather Research and Forecasting Model thermore, the Quick Scatterometer (QuikSCAT) satellite (WRF) simulations of the OMCS embedded in Typhoon

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2841

the methodology of identifying OMCSs in Lee et al. (2012), this second OMCS (Figs. 1f, 2) developed at ap- proximately 0600 UTC 22 June and terminated at ap- proximately 1500 UTC when Fengshen was just moving off Luzon. The OMCS in Fengshen was a ‘‘south type’’ that typically has a monsoon flow to the south (Lee et al. 2012). Some environmental factors at the closest 6-hourly analysis to the south-type OMCS initiation time associated with the 62 TCs during 1999–2009 are compared with a control sample of 1192 times that the TCs had no OMCS and with those environmental fac- tors existing prior to the OMCS in Fengshen (Fig. 4). These environmental factors are composited relative to FIG. 3. The JTWC best track of Typhoon Fengshen (2008) and 8 8 the accumulated rainfall from the TRMM 3B42 dataset from 16 to the TC centers west of 145 E and south of 26 N from 23 Jun 2008 from the National Aeronautics and Space Adminis- June to September and are extracted from the final tration (NASA) Goddard Space Flight Center website (http:// operational global (FNL) analyses of the National trmm.gsfc.nasa.gov/trmm_rain/Events/fengshen_rain_16-23june08.jpg). Centers for Environmental Prediction (NCEP). Dashed ellipses indicate the rainfall maxima that are associated Compared to the control sample with no OMCS (Fig. with the OMCS that is the focus of this study (upper left) plus an earlier OMCS (lower left). 4a), a larger size and more intense TC outer circulation, plus the presence of stronger westerly winds to the southwest of the center, are indicated for the sample of Fengshen (2008) (Figs. 1f, 2) are analyzed. These simu- TCs with an OMCS to the south (Fig. 4d). The asym- lations provide the three-dimensional wind field asso- metry of the environmental flow is even more evident ciated with convective cells, the multicellular cycle of for the Fengshen case (Fig. 4g) with the strongest winds in west-southwesterlies that originated in the Southern the convective system, and the development of the 21 OMCS. Descriptions of the case, WRF, and the verifi- Hemisphere. The average wind speed of about 11 m s cations are provided in sections 2 and 3, respectively. in the black boxes in Figs. 4d and 4g indicates the im- The initiation of the OCMS is described in section 4, and portance of strong environmental winds to the south and the kinematic structure and the multicellular cycle of the wrapping around the east side of the TCs with a south- OMCS are in section 5. Discussion and conclusions are type OMCS. Notice also a band of maximum northerly provided in section 6. winds (Fig. 4g) that is displaced well to the west of the center, which is a factor in the development of the OMCS in this Fengshen case. Vertical wind shear (VWS) is hypothesized to have 2. Overview of Typhoon Fengshen (2008) and the a role of the azimuthal distribution of the convection synoptic environment associated with Fengshen. Corbosiero and Molinari The track of Fengshen and the accumulated rainfall (2002, 2003) indicated lightning strike maxima are con- estimated by the Tropical Rainfall Measuring Mission centrated in the downshear right quadrant in the outer (TRMM) from 16 to 23 June 2008 are shown in Fig. 3. rainband region. Lee et al. (2012) also indicated that the Tropical Storm (TS) Fengshen had rapidly intensified OMCSs were concentrated in the downshear right into a typhoon (TY) before making its first landfall on quadrant. Compared to the control sample (Fig. 4b)orthe Samar Island in the . On 0000 UTC 21 June, sample of the south-type OMCSs (Fig. 4e), the 850–200-mb TY Fengshen turned to the northwest and passed Metro (1 mb 5 1 hPa) deep-layer mean VWS calculated in Manila at approximately 0000 UTC 22 June with winds a radial ring between 28 and 88 latitude of Fengshen (Fig. 2 of 45.8 m s 1. After TY Fengshen left Luzon Island, it 4h) was also toward the southwest, but had a very large 2 moved northward and later made its second landfall magnitude (14.3 m s 1). Notice also the area of strong at Shenzhen, Guangdong Province, at approximately easterly 850–500-mb local VWS over the South China 2200 UTC 24 June. The intense rainfall near the TC Sea to the west of Fengshen (Fig. 4h). center and the rainfall enhancement due to the topogra- Two other environmental factors related to tropical phy were expected, but Fengshen also produced two MCSs are the lower-tropospheric moisture and the sta- OMCSs (Figs. 1e,f). It is the northern OMCS (Fig. 3,black bility, which is represented in Figs. 4c, 4f, and 4i by the dotted circle) that is the focus of this study. According to 850-mb minus 700-mb equivalent potential temperature

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2842 MONTHLY WEATHER REVIEW VOLUME 142

FIG. 4. Environmental variables plotted in latitude/longitude relative to TC center for (a)–(c) control case with no OMCS, (d)–(f) sample 2 of 62 TCs with south-type OMCS, and (g)–(i) Fengshen case at 0000 UTC 22 Jun 2008, with (left) 850-mb streamlines and wind speed (m s 1, 2 2 shading); (middle) 850–500-mb VWS vectors and magnitude (m s 1, shading) and 850–200-mb deep-layer mean VWS (m s 1, red thick arrow); and (right) 850-mb relative humidity (%, shaded) and the difference of equivalent potential temperatures between 850- and 700-mb surface (K, contours). The black box in (d) and (g) indicates the area 400–1200 km south and 1200 km east to 800 km west of the TC center. The color bars and units for (a)–(f) are under panels (d)–(f) and the color bars and units for panels (g)–(i) are along the bottom.

uE. The sample of south-type OMCSs (Fig. 4f) has gradients in the region, this very large DuE between 850 a more concentrated moisture band associated with and 700 mb in Fig. 4i implies the midlevel air is much the westerly winds of the East Asian summer monsoon drier than the low-level air in the outer regions of (recall Fig. 4d) than in the control (Fig. 4c) and has Fengshen. an extensive potentially unstable region to the north- It is hypothesized that the interaction of the outer west of the center. Given small horizontal temperature circulation of Fengshen with the topography of Luzon

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2843

FIG. 5. (a) Four domains of WRF. (b) JTWC best track positions (gray) and simulated (dashed black) track of Typhoon Fengshen (2008) each 12 h from 0000 UTC 21 Jun to 1200 UTC 23 Jun 2008. may account for the band of maximum northerly winds grid nudging technique in the first and second domains to the west of the center (Fig. 4g). This hypothesis and throughout the model simulation. the roles of the asymmetric distribution of moisture and The convection-allowing physics on all grids utilizes the dynamical role of the VWS on the formation of the the new Thompson et al. (2008) microphysics scheme, OMCS will be examined in section 4 using a high- which includes ice, snow, and graupel processes. The resolution simulation. It is anticipated that this combi- Grell and Devenyi (2002) ensemble convective param- nation of topographic interaction, asymmetric moisture eterization scheme is used in the outer two domains. The distribution, and VWS as in the Fengshen case will be Yonsei University (YSU) planetary boundary layer relevant to other cases of south-type OMCSs that have (PBL) scheme (Noh et al. 2003) is used to calculate the caused heavy rainfall over Taiwan. vertical fluxes of sensible heat, moisture, and momen- tum at the lower boundary. The Dudhia shortwave ra- 3. Model description and verifications of the diation scheme (Dudhia 1989) and the Rapid Radiative simulations Transfer Model (RRTM) longwave radiation scheme (Mlawer et al. 1997) are used. a. Model description To ensure a more reasonable TC structure and max- A 60-h simulation with the Advanced Research WRF imize agreement with the observed track (Fig. 5b), an (Skamarock et al. 2005), version 3.4.1, was initialized at artificial vortex to represent Fengshen in the initial 0000 UTC 21 June, with the innermost domain in- conditions is created with the WRF tropical cyclone tegrated from 0000 UTC 22 June to 0000 UTC 23 June to bogus scheme. After some experimentation, a bogus 2 simulate the mesoscale features of the OMCS in vortex with an intensity of 30 m s 1 at a radius of 200 km Fengshen. WRF is a fully compressible, Eulerian, and was adopted. A modified Rankine wind profile is spec- nonhydrostatic model. WRF uses terrain-following, ified beyond the radius of maximum winds that is similar hydrostatic pressure vertical coordinates with 35 levels between the surface and 600 mb and linearly decreases from the surface to 20 hPa. In this study, four nested to near zero at 100 mb. domains that are fixed geographically are employed with b. Verifications and preliminary analyses of two-way interaction between the inner grids (Fig. 5a). the OMCS These four domains have horizontal gridpoint spacings of 36, 12, 4, and 1.33 km and grid dimensions of 229 3 The simulated track and the Joint Typhoon Warning 142, 391 3 316, 646 3 631, and 400 3 400, respectively. Center (JTWC) best track for Fengshen agree well (Fig. The initial and lateral boundary conditions for the 5b), especially from 0000 UTC 22 June to 0000 UTC 23 model are from the NCEP FNL analyses, which are June that includes the period of the OMCS de- available on the surface and at 26 vertical levels from velopment. The track errors during the entire simulation 1000 to 10 mb, and these analyses are also used as the period are less than 120 km and are less than 55 km forcing in the WRF four-dimensional data assimilation during 22 June. In addition, the simulated intensities on

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2844 MONTHLY WEATHER REVIEW VOLUME 142

et al. 2004). These precipitation analyses at very high spatial and temporal resolution have been exclusively derived from low orbiter satellite microwave observa- tions with features that are translated via spatial prop- agation information based on geostationary infrared satellite data. Note the rainfall maximum of approxi- mately 300 mm located in the outer region southwest of the TC center. Although the simulated region of maxi- mum rainfall is shifted to the south by approximately 100 km, the rainfall amount agrees well with the CMORPH observations (Fig. 6b). A key convective feature in the visible imagery (Figs. 7d–f) and in the simulation (Figs. 7a–c) before the OMCS developed is a nearly east-to-west band of moderate convection plus a nearly north-to-south band of shallow convection. The initiation of this north-to- south band of shallow convection and its subsequent interaction with the east-to-west band is considered to be the crucial contribution to the formation of the long- lasting OMCS in Fengshen (Figs. 7c,f). In the simulation, the moisture band originates from an area of high water vapor mixing ratio between 168– 188N and 1188–1208E(Fig. 7g) at 2300 UTC 21 June and contributes to the formation of the north-to-south band of shallow convection at 0500 UTC 22 June (Fig. 7b, red ellipse, and Fig. 7h, letter A). At the downstream end of moisture band A, the surface convective available poten- 2 tial energy (CAPE) is approximately 3600–4000 J kg 1, and the level of free convection is less than 400 m (Fig. 7i), which are conditions favorable for an outburst of deep convection. Note also that a north–south-oriented region of dry air at a larger radius is also wrapped around the outer TC circulation (Fig. 7i). The simulation of the extensive east–west region of radar reflectivity .40 dBZ 3 h later at 1100 UTC (Fig. 8a) is taken to represent the OMCS development. Note the north–south band of shallow convection that is con- tinuing to feed into the OMCS. The first clue as to the source of this north–south convection is from the QuikSCAT oceanic surface winds (Fig. 8d). Note that

FIG. 6. (a) The JTWC track of Fengshen each 6 h during the surface winds northeast of Luzon at 300 km from the 0000 UTC 22 Jun–0000 UTC 23 Jun and the accumulated rainfall center of Fengshen are deflected around the topogra- from the CMORPH dataset from 0300 to 1500 UTC 22 Jun 2008. phy. Some distance to the northwest of Luzon, these (b) Corresponding simulated track and simulated accumulated easterly surface winds turn southward and accelerate. rainfall. However, the northerly surface winds are still at such a large radius at 168N to the west of Fengshen that they 22 June 2008 agree with those in the JTWC best track continue southward into the OMCS rather than being file. Whereas the minimum sea level pressures according able to reach the center. The simulated surface winds to the JTWC were 978 hPa at 0600 and 1800 UTC, the (Fig. 8b) have a similar westward flow past the northern simulated pressures are 982 and 981 hPa, respectively. tip of Luzon and then turn southward and accelerate, An estimate of the rainfall from 0300 to 1800 UTC 22 but similarly pass well to the west of the low pressure June (Fig. 6a) was obtained from the Climate Prediction center and continue toward the simulated OMCS. The Center morphing technique (CMORPH) dataset (Joyce north–south-oriented band of shallow convection south

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2845

FIG. 7. (a)–(c) Simulated mixing ratio of the combination of ice, snow, and graupel particles in the atmospheric column at 0200, 0500, and 0800 UTC 22 Jun 2008, respectively. (d)–(f) Satellite visible images of the typhoon at the times corresponding to (a)–(c), respectively. Thick red 3s indicate the TC center. Dashed red lines, red ellipses, and red arrows indicate the east-to-west convection band, north-to- south shallow convection band, and the convective cells of the initial OMCS, respectively. (g),(h) Water vapor mixing ratio at 0–1 km 2 (values ,15 g kg 1 indicated by red contours) and surface wind at 2300 UTC 21 Jun and 0500 UTC 22 Jun, respectively. (i) CAPE 2 (102 Jkg 1, shaded) at the surface and level of free convection of 400 m (contour) at 0500 UTC 22 Jun. of 178NinFig. 8a may be associated with a region of band of shallow convection. Deep-layer mean (28–88 confluence in the surface wind directions in Fig. 8b. radius) VWS from 850 to 500 mb and from 850 to 200 mb As indicated in section 2, the environmental VWS is in the Fengshen case are indicated in Fig. 8c. While hypothesized to be the dynamical origin of the north–south a reversal from low-level northerlies in the band of

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2846 MONTHLY WEATHER REVIEW VOLUME 142

FIG. 8. Simulated OMCS in its mature state at 1100 UTC 22 Jun 2008 in terms of (a) maximum radar reflectivity with 2 values larger than 40 dBZ (black dotted line) designating the OMCS, (b) surface wind speed (m s 1; shading bar on right), 2 and (c) vertical wind shear from 0- to 3-km height with magnitude .9ms 1 indicated by shading bar on the right and 850– 2 2 200 mb (red thick arrow, 13.8 m s 1) and 850–500 mb (blue thick arrow, 6.9 m s 1)(28–88 radius) VWS. (d) QuikSCAT oceanic surface winds at 1036 UTC 22 Jun when the east–west-oriented OMCS has developed to the southwest of Fengshen (typhoon symbol), which can be compared with the simulated surface winds in overlapping area with panel (b). shallow convection to upper-tropospheric easterlies is well simulated by the model. Therefore, this WRF sim- expected, even the 0–3-km local VWS is strong, and the ulation will be used in the following sections to further 2 850–500-mb mean VWS is easterly at 6.9 m s 1 (Fig. 8c, examine various mechanisms that are considered to have blue arrow). Whereas the OMCS develops in the down- led to the development, the kinematic structure, and the shear left side of the 850–200-mb mean VWS (Fig. 8c,red maintenance of the long-lasting OMCS in Fengshen (2008). arrow), the shallow convective band is on the downshear right side. This aspect of the formation of the OMCS in Fengshen will be discussed further in section 4c. 4. Mechanisms leading to the OMCS development To summarize, the model simulates the TC track and a. Effect of Luzon terrain intensity rather well, especially on the day the OMCS occurs. Although the simulated TC is larger and the As described in section 3b and first indicated by the region of maximum rainfall is shifted to the south, the QuikSCAT wind vectors in Fig. 8d, the low-level circu- mesoscale convective features of the OMCS are generally lation of Fengshen is deflected around the northern end

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2847 of Luzon and flows westward some distance before due to stretching deformation and shearing deformation, turning southward. Although the OMCS is too far south and z is the vertical relative vorticity. In regions where in the simulation (Fig. 8), the surface wind vectors (Fig. vorticity dominates strain (Q , 0), trajectories of two 8b) are deflected around the northern tip of Luzon. neighboring particles do not separate exponentially in Backward tracers at low levels from the simulated OMCS time (Schubert et al. 1999) and coherent structures such also confirm the role of the Luzon terrain in deflecting the as mesovortices can survive. In the region where strain TC circulation (not shown). Thus, the associated moisture dominates vorticity (Q . 0), the vorticity gradient is band passes well to the west of the center of Fengshen intensified to form long, thin vorticity streamers. In such rather than being drawn toward the inner core. regions, fluid elements are stretched and the exponen- This role of Luzon terrain in the development of the tial divergence of nearby particles leads to the so-called OMCS is further investigated with a model sensitivity chaotic stirring in two-dimensional turbulence (Okubo test with a 30% reduction (TER30) and an increase to 1970; Weiss 1991; Elhmaidi et al. 1993; Kevlahan and 150% (TER150) of the terrain height. No special data Farge 1997). assimilation was used as the 6-hourly NCEP fields were In the region where the TC inner-core circulation is used in the nudging technique for both altered terrains. interacting with the Luzon terrain, the Q , 0 region in In the TER150 sensitivity test, the TC moved offshore TER30 (Fig. 10a) is much larger than that of TER150 within 12 h and the accumulated rainfall was primarily (Fig. 10b). Thus, the TC simulated in TER30 maintains offshore with a maximum .400 mm in the OMCS a stronger, more axisymmetric, inner-core structure (Fig. 9b). The deflection of low-level flow around the and thus may be less influenced by the environmental enhanced Luzon topography, and the westward dis- VWS. In the TER30 (Fig. 10a), the regions with large Q placement and broader moisture band (Fig. 9d,green are smaller and are closer to the inner-core region of dashed line) are much more pronounced than for mois- the TC. Note that the vorticity-dominated region in ture band A in Fig. 7h. Another effect of the terrain may TER150 (Fig. 10b) extends farther to the west of northern have been leeside subsidence given the indications of dry Luzon such that the strain-dominated region that is as- air downstream of where the airflow passes over Luzon sociated with the north–south moisture band is pushed (note red contours indicating mixing ratios less than westward to the outer region of the TC. Both this elon- 2 15 g kg 1 in Fig. 9d as well as in Figs. 7g and 7h. Note also gated moisture band and the east–west band where the the maximum rainfall in the OMCS in Fig. 9b is associ- OMCS forms have larger Q values and have larger con- ated with a concentrated region of ice, snow, and graupel vergence in the boundary layer (Fig. 10b). Therefore, that is downshear from the center (Fig. 9f). leeside subsidence may have also contributed additional By contrast, the maximum rainfall in the TER30 moisture convergence to the elongated moisture band. sensitivity test with the reduced terrain is associated with These sensitivity tests confirm the important role of the outer TC circulation (Fig. 9a). Note that the surface the Luzon terrain in deflecting the low-level flow in such streamlines are minimally affected by the reduced terrain a way that the north–south-elongated moisture band (Fig. 9c), and the TC circulation is more symmetric about passes well to the west of the TC center. Rather than the center (in a flat terrain test the circulation is sym- feeding moist air into the inner core of the TC, this metric; not shown). Although there is a relative maximum elongated moisture band extends to the intersection in the water vapor mixing ratio to the west of Luzon, it is with the monsoon airstream to the southwest and con- associated with a rainband that spirals around the inner tributes to the development of the OMCS. core rather than leading to an OMCS. The combination of b. Role of the elongated moisture band ice, snow, and graupel in Fig. 9e has isolated maxima along that rainband that are more representative of con- After the deflection of the low-level, moist monsoonal vective cells rather than the OMCS in Fig. 9f. air around Luzon and its turning southward well to the To diagnose these altered terrain effects on the TC west of Luzon, a key question is how the moisture be- structure and the flow characteristics between TER30 comes concentrated into an elongated north–south band and TER150, the Okubo–Weiss quantity Q after Okubo of shallow convection. The moisture budget equation is (1970) and Weiss (1991) is calculated: written in the flux form: › › y › ›q 52 qu 2 q 2 qw 1 › › E 2 C , (1) 5 1 2 1 2 2 z2 |{z}›t |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl}x y |ffl{zffl}›z |fflfflffl{zfflfflffl} Q (S1 S2 ) 4 TEND HMFC VMFC S where S1 5 [(›u/›x) 2 (›y/›y)], S2 5 [(›y/›x) 1 (›u/›y)], where q is specific humidity, u (y) is the eastward (north- and z 5 [(›y/›x) 2 (›u/›y)] are the strain rates S1 and S2 ward) horizontal velocity, w is the vertical velocity, and

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2848 MONTHLY WEATHER REVIEW VOLUME 142

FIG. 9. Luzon terrain sensitivity tests with (left) 30% and (right) 150% topography heights indicating (a),(b) the TC track each 6 h during 0000 UTC 22 Jun to 0000 UTC 23 Jun and the accumulated rainfall from 0300 to 1500 UTC 22 2 2 Jun 2008; (c),(d) the water vapor mixing ratio (g kg 1, values ,15 g kg 1 indicated by red contours) at 0–1 km with moisture band axis (green dashed line) and surface streamlines at 0500 UTC; and (e),(f) simulated mixing ratio 2 (g kg 1) of the combination of ice, snow, and graupel particles in the atmospheric column and 850–200-mb deep-layer mean VWS (red thick arrow) at 1000 UTC.

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2849

29 22 24 21 FIG. 10. The Okubo–Weiss quantity Q (10 s , shaded) and contours of convergence exceeding 3 3 10 s averaged from 0 to 1 km for the sensitivity test with (a) 30% terrain height and (b) 150% terrain height. The moisture band axis is indicated with a black dashed line in (b).

E (C) is evaporation (condensation). From left to right, moisture convergence represented by HDIV is the key the terms in Eq. (1) are the moisture tendency (TEND), factor that concentrates the moisture into an area that is the horizontal moisture flux convergence (HMFC), the already moist. vertical moisture flux convergence (VMFC), and the The time series of terms in the moisture budget from sources and sinks (S). The HMFC may be rewritten as 0.1 to 1 km in the upstream region of the OMCS (box in the terms of horizontal advection (HADV) and horizontal Fig. 11a between 14.78 and 16.78N and 116.58 and divergence (HDIV): 118.58E) indicates the variations of the TEND term are dominated by the HADV term on a short temporal scale,   but the HDIV 1 VMFC combination contributes to the › › › ›y 52 q 2 y q 2 u 1 positive TEND before the moisture maximum occurs at HMFC u › › q › › . (2) |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl}x y |fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}x y approximately 0715 UTC (Fig. 11b). After the OMCS HADV HDIV develops, both the HADV and the S have minimum values, so the low-level moisture begins to decrease due These moisture budget terms are calculated from a to the vertical flux to the convection of the OMCS. 5-min model output. A temporal and vertical–spatial A second region of high surface moisture flux is as- average is defined as sociated with the interaction between the TC circulation ð ð T2 Z2 and the monsoonal westerlies in the southern to eastern 5 () ()dz dt, quadrants of the TC (Fig. 11c). Therefore, it is proposed T1 Z1 that the southwesterly monsoon flow has an important with T2 2 T1 5 4h,Z15 0.1 km, and Z2 5 1 km. role in supplying moisture to the TC as well as via its role The combination of HMFC and VMFC averaged of convergence with the north–south band in providing between 0.1 and 1 km and from 0300 to 0700 UTC in the favorable thermal and moisture conditions for the control simulation is plotted in Fig. 11a. The positive OMCS development. total moisture flux convergence (MFC; Fig. 11a) is large c. Role of vertical wind shear near both the strong moisture band A (Fig. 7h) and the region of the decaying convective band (Figs. 7a,b). The The environmental VWS in the outer circulation of negative total MFC is mainly in the region of the de- the TC may be the dynamical origin of the shallow caying convective band from 1168 to 1188E and from 138 convection in the elongated moisture band based on to 148N. Near moisture band A, positive (negative) idealized numerical experiments of Riemer et al. (2010), HADV is found to the west (east) side of the moisture who showed that the wavenumber-1, asymmetric, up- bands, and the HDIV is collocated with the band. ward, vertical motion is on the downshear right side in Whereas the horizontal advection term is clearly related the outer region of the TC. Riemer et al. (2010) illus- to the westward displacement of the moisture band, the trated the resulting convective asymmetry was due to the

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2850 MONTHLY WEATHER REVIEW VOLUME 142

tilt of the outer vortex by showing that the wavenumber- 1 vorticity asymmetry above the boundary layer ex- tended out to a 150–200-km radius, which produces frictional convergence in the boundary layer and thus forces upward motion based on Ekman pumping theory. Whereas Riemer et al. (2010) imposed a simple wind profile with easterly VWS, the VWS direction in the Fengshen case rotated anticlockwise and increased with height. In the Fengshen simulation at 0200 UTC 22 June, a large azimuthal wavenumber-1 vorticity asymmetry at 1.5-km height is located to the downshear right side of 850–200-mb VWS (Figs. 12a,b). This vorticity asymme- try is also collocated with an elongated region of con- vergence from 158 to 188N and near 1188E. Thus, the interaction of the TC and VWS may have had an im- portant role in concentrating the moisture and thus in- creasing the CAPE at the downstream end of the band. Note also that another band structure of convergence with higher rain rate is associated with the decaying east–west convective band (Fig. 12b). The shallow convection develops from 0200 UTC 22 June as the moisture is concentrated in the elongated band region (see streamlines in Figs. 12c,d). The 0–6-km 2 local VWS greater than 15 m s 1 is associated with the shallow convection in the region from 158 to 188N and from 1168 to 1188E at 0200 UTC and then moves downstream and intensifies at 0500 UTC (Fig. 12d). 2 Note also the large (.21 m s 1) 0–6-km VWS that is associated with the decaying east–west convective band between 138 and 148N. Since the 0–6-km VWS vector within this elongated band of shallow convection has a direction shift of about 908 to the low-level winds, these conditions are favorable for developing a large stratiform precipitation region on the convex side. The VWS may also have had an important role in the OMCS devel- opment via separating the region of upward motion from the region of precipitation and thereby preventing new convective cells from being suppressed by the rainfall.

5. Kinematic structure and the maintenance mechanisms To analyze possible maintenance mechanisms of the OMCS, the kinematic structure and multicellular cycle of it will be examined in this section. a. Analysis of the simulated OMCS

FIG. 11. (a) Average moisture flux convergence in the 0.1–1-km The mesoscale convective features of the OMCS are layer from 0300 to 0700 UTC with average water vapor mixing ratios well simulated by the model (Fig. 13). The atmospheric 2 greater than 20 g kg 1 indicated by contours. (b) Time series of column ice, snow, and graupel hydrometer mixing ratio 8 8 moisture budget terms in the 0.1–1 km in the area from 14.7 to 16.7 N (Figs. 13a,b) is distributed similarly to the PMW 91-GHz and 116.58 to 118.58E [black box in (a) and (c)]. (c) Average surface moisture flux and surface streamlines from 0300 to 0700 UTC with PCT TB observed by the Special Sensor Microwave 2 average rainwater mixing ratios in the 0.1–1 km greater than 3 g kg 1. Imager/Sounder (SSMIS) (Figs. 13c,d). The convective

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2851

24 21 FIG. 12. (a) Asymmetric relative vorticity (10 s , shaded) at 1.5 km and wavenumber-1 asymmetric vorticity 2 2 2 2 2 contours of 1.5 3 10 5 and 3.5 3 10 5 s 1 at 0200 UTC 22 Jun. (b) Divergence (10 5 s 1, shaded) at 1.5 km and 2 contours of 0.1–1-km average rainwater mixing ratio of 3 g kg 1 at 0200 UTC. (c) Surface streamlines and 0–6-km 2 2 VWS (arrows, m s 1, shaded) greater than 15 m s 1 at 0200 UTC. (d) As in (c), except for 0500 UTC 22 Jun.

cells in both the simulation and the PMW TB are located 8b), which is therefore in an environment of large VWS to the north or in the central area of the stratiform (see section 4c). precipitation region. Previous studies (McCaul and Weisman 1996; Houze Recall from Fig. 8a that the simulated OMCS in the 2010; Akter and Tsuboki 2012) that have addressed mature stage at 1100 UTC 22 June 2008 consists of distant rainbands have generally assumed that cold a convective line (maximum radar reflectivity .50 dBZ) pools in the TC are relatively weak due to small evap- and a stratiform precipitation region (35- to 45-dBZ orative cooling rates associated with the stratiform region) to the south of the convective line. This OMCS is precipitation. In this simulation, the large stratiform approximately 275 km southwest of the TC center and is precipitation region that developed in association with collocated with the 850-hPa positive vorticity band that the OMCS has a cold pool with surface potential is independent to the region of large positive vorticity temperatures 2–3 K colder than in the upstream region associated with the TC inner-core circulation. Recall (Fig. 14a). The vertical velocity at 1.5-km height also from Fig. 8b that a surface jet is collocated with smoothed to 25 km (Fig. 14b) has an area of upward the stratiform precipitation region of the simulated motion on the northern edge of the OMCS that is as- OMCS, but is slightly shifted cyclonically inward to- sociated with the convective precipitation region and ward the TC center. Furthermore, the OMCS has de- an area of downward motion under the stratiform veloped in the confluence region between the northerly precipitation region that is specifically associated with wind relative and the monsoonal westerly wind (Fig. the maximum cold pool.

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2852 MONTHLY WEATHER REVIEW VOLUME 142

FIG. 13. (a),(b) Mixing ratio of the combination of ice, snow, and graupel particles in the atmospheric column at 0900 UTC 22 Jun and 1200 UTC 22 Jun in the simulation and (c),(d) corresponding PMW images from SSMIS.

The time evolution of the OMCS is described on monsoonal flow. As described in section 4b, the central a western region between 117.78 and 118.08E and in an core of the north–south band has a high moisture con- 2 eastern region between 118.28 and 118.58E as the west- tent (.18 g kg 1, Fig. 15e). Because this band is more ern and central portions of the elongated north–south moist than the monsoonal flow to the south, horizontal moisture band converges with monsoonal westerly flow moisture convergence occurs and contributes to the (Fig. 15). In the western portion of the OMCS, a series of development of the convective cells at the intersection convective cells (defined by column mixing ratio of ice, of the north–south band with the monsoonal flow. 2 2 snow, and graupel .25 3 10 4 kg kg 1) form and The key difference in the eastern region (Fig. 15, right propagate southeastward (Fig. 15a) as the surface winds column) is that the central core of the north–south 2 within the north–south band exceed about 16 m s 1 (Fig. moisture band with more northerly wind components 15c). Local enhancement of these northwesterly winds is (Fig. 15d, top) is interacting with the monsoonal flow simulated in conjunction with the convective cells and to produce a continuous, and more stationary, line of localized maxima in the monsoonal flow occur where convective cells that then form the long-lasting OMCS the enhanced northwesterly winds converge with the (Fig. 15b, along the thick, black dashed line). Except for

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2853

21 FIG. 14. (a) Surface potential temperature (K, shading bar on right) and the 0.7 g kg contour of atmospheric column mixing ratio of ice, snow, and graupel in the simulation at 1100 UTC 22 Jun 2008. (b) Vertical velocity at 2 1.5-km height (m s 1, shading bar on right) and green contours indicate cold pools with surface potential tempera- tures of 300.5 and 300.8 K. All contours and the vertical velocity in (b) have been smoothed to 25-km resolution. the region of weaker northwesterly winds with some- While northerly flow in the (Northern Hemisphere) what drier air (Fig. 15f, dashed triangle), these more tropics is typically cooler and drier than a warm, moist northerly winds again are more moist than the mon- monsoonal flow, the airstream at the downstream end of soonal air and lead to horizontal moisture convergence the elongated moisture band is warmer (Fig. 15h) and along the convective line at the northern edge of the more moist (Fig. 15f) than the monsoonal flow. In the OMCS. This convective line is also the boundary be- cold pool region between 13.38 and 14.58N, the surface tween warmer air in the north–south band and the cold air temperatures are more than 2 K colder than that of pool air under the stratiform cloud region that has po- the upstream (northerly flow) environment (Fig. 16b). tential temperatures of 299–300 K (Fig. 15h). It is also Several studies (e.g., Barnes et al. 1983; Yamasaki 1983; under the stratiform cloud region that the monsoonal Powell 1990a; Eastin et al. 2012) have documented the westerly winds have a low-level jet with wind speeds of existence of cold pools in rainbands in which the surface 2 22 m s 1 (Fig. 15d). A momentum budget (not shown) air was more than 2 K colder than the surrounding en- confirms the conclusion of Trier et al. (1998) that such vironment. However, the area of cold pools identified in a convective line oriented perpendicular to the low-level those studies did not extend over as large an area as the VWS tends to increase the VWS by vertically trans- OMCS in this study. Note the gradient of u at the ferring momentum against the gradient. northern edge of the cold pool is large, which is favor- able for triggering new convective cells. b. Kinematic structure of the OMCS Viewed relative to this southward-moving OMCS, the The kinematic structure of the OMCS is displayed by low-level rear inflow from the north–south band is sup- taking a 2-h (0815–1015 UTC) average of the zonal-mean plying warm, moist air with a very high uE (Fig. 16e)thatis fields in the eastern portion of the OMCS (Fig. 16). then lifted in the convective region and forms the leading Because the orientation of the convective line along the stratiform precipitation region to the south of the convec- northern edge of the OMCS is roughly east to west, the tive region. Although the maximum northward cross-line zonal wind u and meridional wind y will be considered as velocities occur near the boundary of the convective region the along-line and cross-line velocities in Fig. 16. Note and the stratiform region (Fig. 16d), some of the outflow the convergence between the stronger (weaker) north- from the convective is back toward the north (Fig. 16b), erly (southerly) winds on the north (south) side of the which leads to the backward (northward) overhang in the convective line, which is consistent with the reflectivity reflectivity aloft (Fig. 16a). Thus, the kinematic structure (Fig. 16a) and vertical velocity (Fig. 16b). That is, of this OMCS is similar to the rear-fed, leading stratiform the convective precipitation region (14.28–14.78N) has archetype of Pettet and Johnson (2003),whofoundin large upward vertical velocity, and the stratiform pre- case studies that a considerable fraction of the leading cipitation region (13.58–14.28N) has downward vertical stratiform systems in the central United States are sus- velocity below the freezing level. tained by the inflow of high uE air from behind the system.

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2854 MONTHLY WEATHER REVIEW VOLUME 142

FIG. 15. Hovmoller€ diagrams of zonally averaged fields in the (left) western portion (117.78–118.08E) and the (right) eastern portion 2 2 (118.28–118.58E) of the OMCS. (a),(b) Atmospheric column mixing ratio of ice, snow, and graupel (10 4 kg kg 1); (c),(d) surface wind 2 2 2 speeds (m s 1, shaded); (e),(f) water vapor mixing ratio 10 3 kg kg 1 from 0- to 1-km height; and (g),(h) surface potential temperature (K). The latitude of the convective band is indicated by a thick, black dotted line, and the dashed, triangular region in (c)-(f) indicates a region of northwesterly winds that have drier air.

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2855

FIG. 16. Time mean (;0815–1015 UTC) height–latitude cross sections of zonally averaged fields in the eastern portion (118.28–118.58E) of the OMCS with colors for the various panels shown in the lower-right panel: (a) Re- 2 flectivity and mixing ratio of ice 1 snow 1 graupel (contour in 1 g kg 1 interval); (b) vertical velocity and potential 2 temperature (contours); (c) eastward horizontal velocity; (d) meridional wind speeds (m s 1, negative values are southward); and (e) equivalent potential temperature. Vectors in (a), (b), and (e) are (y, w 3 10). Solid and dotted 2 2 contours in (c) and (d) indicate upward motion (in 0.8 m s 1 intervals) and downward motion (in 20.04 m s 1 in- tervals), respectively.

The multicellular time evolution of the convective line cell B is in its mature stage between 0940 and 0950 UTC is displayed in Fig. 17 by taking 10-min averages of the (Fig. 17c). After another 10 min (Fig. 17d), a new cell (C) zonal-mean fields in the eastern portion of the OMCS. begins at the leading edge of the cold pool. During this Although the convective structure between 0930 and period (0930–1000 UTC), the downward vertical velocity 0940 UTC (Fig. 17a) is basically similar to the averaged within the stratiform precipitation region of the OMCS kinematic structure (Fig. 16a) with a mature cell (cell A) becomes more organized and stronger. The magnitude of and the stratiform region south of 148N, the initial stage the cold pool is also enhanced with minimum Du ,23K of cell B is also indicated above the leading edge of the near 148N. It is also suggested that the dry air that extends cold pool. After 5 min (Fig. 17b), cell A begins to dissi- around the western side of the OMCS at low levels to pate while cell B continues develop. As cell A dissipates, midlevels (Figs. 7i, 16e) may enhance the evaporation

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2856 MONTHLY WEATHER REVIEW VOLUME 142

the OMCSs is a critical issue for TC rainfall forecasting and disaster warning operations. This study extends the OMCS climatological analysis of Lee et al. (2012) and then uses a WRF model simulation to examine the de- velopment, kinematic structure, and the maintenance mechanisms of a ‘‘south-type’’ OMCS that occurred in- the outer circulation of Typhoon Fengshen (2008). The Fengshen OMCS had some similar conditions as a com- posite of 62 south-type OMCSs documented by Lee et al. (2012): (i) presence of extended monsoonal flow to the southwest of the TC center that wraps around the southern and eastern quadrants; (ii) a narrow moisture band that extends north to south well to the west of TC center (rather than spiraling into the center); and (iii) the OMCS forms where the southern end of the north–south moisture band interacts with the monsoonal flow. The track and minimum sea level pressure of Fengshen were well simulated, and the simulated rainfall distri- bution during the OMCS period (0300–1500 UTC) agreed well with the CMORPH rainfall, except that the location was shifted to the south by approximately 100 km. In addition, the mesoscale convective features including the shallow convection before the OMCS initiation, the convective and stratiform structure of the OMCS, and the surface jet under the stratiform region are well simulated. Thus, the simulation is the basis for an analysis of the development, kinematic structure, and maintenance of the Fengshen OMCS that is summarized in the conceptual model in Fig. 18. The role of Luzon topography was investigated with a model sensitivity test with 30% (TER30) and 150% (TER150) terrain height and demonstrated that the deflection of the TC low-level circulation by the terrain caused the development of an elongated moisture band well to the west of the center such that the OMCS de- velops in the outer region of the TC (Fig. 18b, black streamlines). The Okubo–Weiss calculation indicated that the strain-dominated region was associated with the elongated moisture band to the west of northern FIG. 17. Height–latitude cross sections of vertical velocity and po- tential temperature (K, contours) near the surface in the east portion Luzon and that leeside subsidence may have also con- (118.28–118.58E) to illustrate the multicellular (labeled A, B, and C) tributed additional moisture convergence to the elon- nature of the convective line of the OMCS. All fields are smoothed by gated moisture band. In the reduced 30% terrain height taking a 10-min average over the time intervals indicated. simulation, the moisture band/rainband formation spi- raled into the TC inner core (Fig. 18b, dashed stream- cooling under the stratiform precipitation region and thus lines). A similar deflection of the TC low-level flows by may have contributed to the formation of such a strong Taiwan terrain may have contributed to the formation of mesoscale cold pool in the OMCS. other OMCSs that have developed near Taiwan [e.g., TY Kalmaegi (2008) in Fig. 1c]. A moisture budget indicates moisture associated with 6. Summary and discussion the TC–monsoon interaction is transported cyclonically Since OMCSs are a special type of long-lasting rain- around the TC and becomes concentrated in the elon- band that occurs in the outer region of some western North gated north–south band to the west of the TC center Pacific TCs, forecasting the heavy rainfall associated with due to large moisture flux convergence. Thus, very large

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2857

FIG. 18. (a) Schematic diagram of radar reflectivity of a TC with a long-lasting OMCS well to the southwest of the center. (b) Conceptual model of the TC–monsoon flow interaction that is favorable for the OMCS development (star) with the streamlines of the TC circulation affected (not affected) by the Luzon terrain indicated by solid (dashed) lines. The outer circulation streamlines (heavy black lines with arrows) and the north–south moisture band (green shading) lead to the OMCS. The wavenumber-1 vorticity asymmetry that is consistent with the VWS (thick black arrow) is indicated by red, dashed lines. (c) Conceptual model of kinematic structure of OMCS in height– latitude cross section from A to B in (a) for radar reflectivity, u, uE, wind speed (arrows), and low-level jet. (d) Composite of the OMCS locations with respect to the TC center for nonterrain-influenced OMCSs (gray circles) and terrain-influenced OMCSs with the TC to the south (red triangles), north (blue triangles), and east (green triangles) of the terrain. The thick colored, dotted lines indicate the conceptual coastlines for different types of terrain-influenced OMCS.

2 convective available potential energy (;3600–4000 J kg 1) height (Fig. 18b, red dashed line) is simulated on the and a low level of free convection (,400 m) was simu- downshear right side of 850–200-mb VWS and is consid- lated at the downstream end of the north–south mois- ered to have produced frictional convergence in the ture band where it interacted with the monsoonal flow boundary layer that then led to shallow convection along and produced the OMCS (Fig. 18b, star). the moisture band. Therefore, both the large moisture flux One of the environmental characteristics of OMCSs convergence and the frictional convergence in the elon- is the northeasterly vertical wind shear. In the Fengshen gated north–south band due to the VWS are considered to 2 case, the deep-layer VWS was particularly large (14 m s 1), have contributed to the moisture source for the OMCS. and even the lower-tropospheric VWS was large. A large The kinematic structure of the convection of the azimuthal wavenumber-1 vorticity asymmetry at 1.5-km OMCS is composed of a continuous line of convective

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC 2858 MONTHLY WEATHER REVIEW VOLUME 142 cells and a large stratiform precipitation region to the Lee is supported by the National Taiwan University south of the convective line as shown in the schematic in and the Taiwan Typhoon Flood Research Institute of Fig. 18c. Thus, this OMCS has a structure in the cross- the National Applied Research Laboratories. This re- line direction similar to the rear-fed inflow and leading search is supported by the National Science Council of stratiform type of MCS (Pettet and Johnson 2003). the Republic of China (Taiwan) under Grants NSC Furthermore, a strong cold pool (Du ,23 K) is associ- 98-2625-M-002-002 and NSC 99-2625-M-002-013-MY3. ated with the large stratiform precipitation region. New Professor Russell L. Elsberry is supported by the Ma- cells are continually formed when the air with high uE rine Meteorology section, Office of Naval Research. values from the north converges with the leading Mrs. Penny Jones provided excellent assistance in the (northern) edge of the cold pool. In this manner, the manuscript preparation. continuous formation of new cells may contribute to the longer duration of this OMCS than exists for the TC REFERENCES distant rainbands described in the introduction. In par- ticular, the large stratiform region of this OMCS is Akter, N., and K. Tsuboki, 2012: Numerical simulation of Cyclone unique compared to distant rainbands and the upwind Sidr using a cloud-resolving model: Characteristics and for- mation process of an outer rainband. Mon. Wea. Rev., 140, portion of primary rainbands that are generally lacking 789–810, doi:10.1175/2011MWR3643.1. in stratiform precipitation (Houze 2010; Akter and Barnes, G. M., E. J. Zipser, D. P. Jorgensen, and F. D. Marks Jr., Tsuboki 2012; Eastin et al. 2012), and the stratiform 1983: Mesoscale and convective structure of a hurricane region of primary rainbands typically develops in its rainband. J. Atmos. Sci., 40, 2125–2137, doi:10.1175/ , . downwind position (Hence and Houze 2008; Didlake 1520-0469(1983)040 2125:MACSOA 2.0.CO;2. Cecil, D. J., and E. J. Zipser, 2002: Reflectivity, ice scattering, and Houze 2013b; Li and Wang 2012). and lightning characteristics of hurricane eyewalls and rain- The development of the large stratiform precipitation bands. Part II: Intercomparison of observations. Mon. Wea. region to the south of the convective line leads to this Rev., 130, 785–801, doi:10.1175/1520-0493(2002)130,0785: OMCS having a large spatial scale. Yuter and Houze RISALC.2.0.CO;2. (1998) and Houze (2004) suggested that such large Chien, F.-C., and H.-C. Kuo, 2011: On the extreme rainfall of Typhoon Morakot (2009). J. Geophys. Res., 116, D05104, stratiform regions are sustained by the environmental doi:10.1029/2010JD015092. flow. Future studies will focus on the sensitivity of the Corbosiero, K. L., and J. Molinari, 2002: The effects of vertical environmental conditions that lead to the development wind shear on the distribution of convection in tropical and maintenance of the stratiform precipitation regions cyclones. Mon. Wea. Rev., 130, 2110–2123, doi:10.1175/ , . of OMCSs. 1520-0493(2002)130 2110:TEOVWS 2.0.CO;2. ——, and ——, 2003: The relationship between storm motion, Lee et al. (2011) found OMCSs occurred in 22% of all vertical wind shear, and convective asymmetries in trop- TCs in the western North Pacific. Among the 62 south- ical cyclones. J. Atmos. Sci., 60, 366–376, doi:10.1175/ type OMCSs that are associated with the summer mon- 1520-0469(2003)060,0366:TRBSMV.2.0.CO;2. soon, 6 of them were terrain-influenced cases similar to Didlake, A. C., Jr., and R. A. Houze Jr., 2013a: Convective-scale the Fengshen case with the TC passing to the south of the variations in the inner-core rainbands of a tropical cyclone. J. Atmos. Sci., 70, 504–523, doi:10.1175/JAS-D-12-0134.1. terrain (Fig. 18d). Six other terrain-influenced south-type ——, and ——, 2013b: Dynamics of the stratiform sector of OMCS cases were associated with a TC passing to the a tropical cyclone rainband. J. Atmos. Sci., 70, 1891–1911, north of the terrain, which are therefore similar to the TY doi:10.1175/JAS-D-12-0245.1. Morakot (2009) case in Fig. 1a. Another 13 terrain- Dudhia, J., 1989: Numerical study of convection observed during influenced south-type OMCS cases were associated the winter monsoon experiment using a mesoscale two- dimensional model. J. Atmos. Sci., 46, 3077–3107, doi:10.1175/ with a TC approaching the coast of southeast China. 1520-0469(1989)046,3077:NSOCOD.2.0.CO;2. Whereas the total number of these terrain-influenced Eastin, M. D., T. L. Gardner, M. C. Link, and K. C. Smith, 2012: OMCS is 26, they only account for 42% of 62 south-type Surface cold pools in the outer rainbands of Tropical Storm OMCSs. Further study is required of the other south-type Hanna (2008) near landfall. Mon. Wea. Rev., 140, 471–491, OMCS cases that were not terrain-influenced to gain doi:10.1175/MWR-D-11-00099.1. Elhmaidi, D., A. Provenzal, and A. Babinao, 1993: Elementary a better understanding and improve the capability of the topology of two-dimensional turbulence from a Lagrangian numerical models and thus improve forecasts of the viewpoint and single-particle dispersion. J. Fluid Mech., 257, OMCSs in TCs, which will lead to more accurate and 533–558, doi:10.1017/S0022112093003192. timely warnings of heavy rainfall events associated with Grell, G. A., and D. Devenyi, 2002: A generalized approach this type of rainband systems. to parameterizing convection combining ensemble and data assimilation techniques. Geophys. Res. Lett., 29 (14), doi:10.1029/2002GL015311. Acknowledgments. Buo-Fu Chen is supported by the Hence, D. A., and R. A. Houze, 2008: Kinematic structure of National Taiwan University, and Professor Cheng-Shang convective-scale elements in the rainbands of Hurricanes

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC AUGUST 2014 C H E N E T A L . 2859

Katrina and Rita (2005). J. Geophys. Res., 113, D15108, ——, 1990b: Boundary layer structure and dynamics in outer hur- doi:10.1029/2007JD009429. ricane rainbands. Part II: Downdraft modification and mixed Houze, R. A., Jr., 2004: Mesoscale convective systems. Rev. Geo- layer recovery. Mon. Wea. Rev., 118, 918–938, doi:10.1175/ phys., 42, RG4003, doi:10.1029/2004RG000150. 1520-0493(1990)118,0918:BLSADI.2.0.CO;2. ——, 2010: Clouds in tropical cyclones. Mon. Wea. Rev., 138, 293– Riemer, M., M. T. Montgomery, and M. E. Nicholls, 2010: A new 344, doi:10.1175/2009MWR2989.1. paradigm for intensity modification of tropical cyclones: Ishihara, M., Z. Yanagisawa, H. Sakakibara, K. Matsuura, Thermodynamic impact of vertical wind shear on the inflow and J. Aoyagi, 1986: Structure of a typhoon rainband observed layer. Atmos. Chem. Phys., 10, 3163–3188, doi:10.5194/ by two Doppler radars. J. Meteor. Soc. , 64, 923–939. acp-10-3163-2010. Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. Xie, 2004: Samsury, C. E., and E. J. Zipser, 1995: Secondary wind maxima in CMORPH: A method that produces global precipitation esti- hurricanes: Airflow and relationship to rainbands. Mon. Wea. mates from passive microwave and infrared data at high spatial Rev., 123, 3502–3517, doi:10.1175/1520-0493(1995)123,3502: and temporal resolution. J. Hydrometeor., 5, 487–503, SWMIHA.2.0.CO;2. doi:10.1175/1525-7541(2004)005,0487:CAMTPG.2.0.CO;2. Schubert, W. H., M. T. Montgomery, R. K. Taft, T. A. Guinn, S. R. Kevlahan, N. K.-R., and M. Farge, 1997: Vorticity filaments in two- Fulton, J. P. Kossin, and J. P. Edwards, 1999: Polygonal eye- dimensional turbulence: Creation, stability and effect. J. Fluid walls, asymmetric eye contraction, and potential vorticity mixing Mech., 346, 49–76, doi:10.1017/S0022112097006113. in hurricanes. J. Atmos. Sci., 56, 1197–1223, doi:10.1175/ Lee, C.-S., C.-C. Wu, T.-C. C. Wang, and R. L. Elsberry, 2011: 1520-0469(1999)056,1197:PEAECA.2.0.CO;2. Advances in understanding the ‘‘perfect monsoon-influenced Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. typhoon’’: Summary from International Conference on Ty- Barker, W. Wang, and J. G. Powers, 2005: A description of the phoon Morakot (2009). Asia-Pac. J. Atmos. Sci., 47, 213–222, advances research. WRF version 2. NCAR Tech. Note doi:10.1007/s13143-011-0010-2. NCAR/TN-4681STR, 88 pp. [Available online at http://www. ——, B.-F. Chen, and R. L. Elsberry, 2012: Long-lasting convective mmm.ucar.edu/wrf/users/docs/arw_v2.pdf.] systems in the outer region of tropical cyclones in the western Tabata, A., H. Sakakibara, M. Ishihara, K. Matsuura, and North Pacific. Geophys. Res. Lett., 39, L21812, doi:10.1029/ Z. Yanagisawa, 1992: A general view of the structure of 2012GL053685. typhoon 8514 observed by dual-Doppler radar—From outer Li, Q., and Y. Wang, 2012: A comparison of inner and outer spiral rainbands to eyewall clouds. J. Meteor. Soc. Japan, 70, 897– rainbands in a numerically simulated tropical cyclone. Mon. 917. Wea. Rev., 140, 2782–2805, doi:10.1175/MWR-D-11-00237.1. Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008: May, P. T., and G. J. Holland, 1999: The role of potential vortic- Explicit forecasts of winter precipitation using an improved ity generation in tropical cyclone rainbands. J. Atmos. bulk microphysics scheme. Part II: Implementation of a new Sci., 56, 1224–1228, doi:10.1175/1520-0469(1999)056,1224: snow parameterization. Mon. Wea. Rev., 136, 5095–5115, TROPVG.2.0.CO;2. doi:10.1175/2008MWR2387.1. ——, ——, and W. L. Ecklund, 1994: Wind profiler observations of Trier, S. B., M. A. LeMone, and W. C. Skamarock, 1998: Effect Tropical Storm Flo at Saipan. Wea. Forecasting, 9, 410–426, of three-dimensional structure on the stormwide horizon- doi:10.1175/1520-0434(1994)009,0410:WPOOTS.2.0.CO;2. tal accelerations and momentum budget of a simulated McCaul, E. W., Jr., and M. L. Weisman, 1996: Simulations of shallow squall line. Mon. Wea. Rev., 126, 2580–2598, doi:10.1175/ supercell storms in landfalling hurricane environments. Mon. 1520-0493(1998)126,2580:EOTDSO.2.0.CO;2. Wea. Rev., 124, 408–429, doi:10.1175/1520-0493(1996)124,0408: Wang, Y., 2002: Vortex Rossby waves in a numerically simulated SOSSSI.2.0.CO;2. tropical cyclone. Part II: The role in tropical cyclone structure Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. and intensity changes. J. Atmos. Sci., 59, 1239–1262, doi:10.1175/ Clough, 1997: Radiative transfer for inhomogeneous atmo- 1520-0469(2002)059,1239:VRWIAN.2.0.CO;2. sphere: RRTM, a validated correlated-k model for the longwave. ——, 2009: How do outer spiral rainbands affect tropical cyclone J. Geophys. Res., 102, 16 663–16 682, doi:10.1029/97JD00237. structure and intensity? J. Atmos. Sci., 66, 1250–1273, Moon, Y., and D. S. Nolan, 2010: The dynamic response of the doi:10.1175/2008JAS2737.1. hurricane wind field to spiral rainband heating. J. Atmos. Sci., Weiss, J., 1991: The dynamics of enstrophy transfer in two-dimensional 67, 1779–1805, doi:10.1175/2010JAS3171.1. hydrodynamics. Physica D, 48, 273–294, doi:10.1016/ Noh, Y., W. G. Cheon, S. Y. Hong, and S. Raasch, 2003: Im- 0167-2789(91)90088-Q. provement of the K-profile model for the planetary boundary Willoughby, H. E., J. A. Clos, and M. G. Shoreibah, 1982: Con- layer based on large eddy simulation data. Bound.-Layer centric eyewalls, secondary wind maxima, and the evolution of Meteor., 107, 401–427, doi:10.1023/A:1022146015946. the hurricane vortex. J. Atmos. Sci., 39, 395–411, doi:10.1175/ Okubo, A., 1970: Horizontal dispersion of floatable particles in the 1520-0469(1982)039,0395:CEWSWM.2.0.CO;2. vicinity of velocity singularity such as convergences. Deep-Sea ——, F. D. Marks, and R. J. Feinberg, 1984: Stationary and moving Res., 17, 445–454, doi:10.1016/0011-7471(70)90059-8. convective bands in hurricanes. J. Atmos. Sci., 41, 3189–3211, Pettet, C. R., and R. H. Johnson, 2003: Airflow and precipitation doi:10.1175/1520-0469(1984)041,3189:SAMCBI.2.0.CO;2. structure of two leading stratiform mesoscale convective systems Yamasaki, M., 1983: A further study of the tropical cyclone determined from operational datasets. Wea. Forecasting, 18, 685– without parameterizing the effects of cumulus convec- 699, doi:10.1175/1520-0434(2003)018,0685:AAPSOT.2.0.CO;2. tion. Pap. Meteor. Geophys., 34, 221–260, doi:10.2467/ Powell, M. D., 1990a: Boundary layer structure and dynamics in mripapers.34.221. outer hurricane rainbands. Part I: Mesoscale rainfall and ki- Yuter, S. E., and R. A. Houze, 1998: The natural variability of nematic structure. Mon. Wea. Rev., 118, 891–917, doi:10.1175/ precipitating clouds over the western Pacific warm pool. Quart. 1520-0493(1990)118,0891:BLSADI.2.0.CO;2. J. Roy. Meteor. Soc., 124, 53–99, doi:10.1002/qj.49712454504.

Unauthenticated | Downloaded 10/04/21 04:30 AM UTC