Numerical Simulations of Bow Echo Formation Following a Squall Line Supercell Merger

Transcription

1 DECEMBER 2014 F R E N C H A N D P A R K E R 4791 Numerical Simulations of Bow Echo Formation Following a Squall Line Supercell Merger ADAM J. FRENCH Atmospheric and Environmental Sciences, South Dakota School of Mines and Technology, Rapid City, South Dakota MATTHEW D. PARKER Department of Marine, Earth, and Atmospheric Sciences, North Carolina State University, Raleigh, North Carolina (Manuscript received 8 November 2013, in final form 20 June 2014) ABSTRACT Output from idealized numerical simulations is used to investigate the storm-scale processes responsible for squall-line evolution following a merger with an isolated supercell. A simulation including a squall line supercell merger is compared to one using the same initial squall line and background environment without the merger. These simulations reveal that while bow echo formation is favored by the strongly sheared background environment, the merger produces a more compact bowing structure owing to a locally enhanced rear-inflow jet. The merger also represents a favored location for severe weather production relative to other portions of the squall line, with surface winds, vertical vorticity, and rainfall all being maximized in the vicinity of the merger. An analysis of storm-scale processes reveals that the premerger squall line weakens as it encounters outflow from the preline supercell, and the supercell becomes the leading edge of the merged system. Subsequent localized strengthening of the cold pool and rear-inflow jet produce a compact, intense bow echo local to the merger, with a descending rear-inflow jet creating a broad swath of damaging surface winds. These features, common to severe bow echoes, are shown to be a direct result of the merger in the present simulations, and are diminished or absent in the no-merger simulation. Sensitivity tests reveal that mergers in a weaker vertical wind shear environment do not produce an enhanced bow echo structure, and only produce a localized region of marginally enhanced surface winds. Additional tests demonstrate that the details of postmerger evolution vary with merger location along the line. 1. Introduction One of the challenges faced by severe weather forecasters is anticipating how storms will organize, and how that organization will change over time, as different storm types tend to produce different types of severe weather hazards. Widespread straight-line damaging winds tend to be most prevalent in quasi-linear convective systems, particularly those that organize as bow echoes (Fujita 1978), while damaging hail is most often found in supercell thunderstorms (e.g., Gallus et al. 2008; Smith et al. 2012). Tornadoes have been documented with both types of Corresponding author address: Adam J. French, South Dakota School of Mines and Technology, 501 E. St. Joseph St., Rapid City, SD adam.french@sdsmt.edu storm organizations; however, most significant [(enhanced Fujita) EF-2 or greater], and long-lived tornadoes occur with supercells (Trapp et al. 2005; Smith et al. 2012). The problem of forecasting convective mode has often been cast as a function of background environmental parameters (e.g., Weisman and Klemp 1982; Rasmussen and Blanchard 1998; Doswell and Evans 2003; Thompson et al. 2012) or forcing mechanisms (e.g., Jewett and Wilhelmson 2006; Dial et al. 2010; Houston and Wilhelmson 2012). The present work, however, seeks to better clarify the role that mergers between dissimilar storm types, specifically quasi-linear convective systems ( squall lines ) and isolated supercells, play in governing convective mode. Recent observational work by French and Parker (2012, hereafter FP12) has documented that these types of mergers frequently produce a spectrum of bow echoes (Fujita 1978) depending on DOI: /MWR-D Ó 2014 American Meteorological Society

2 4792 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 the background environmental conditions (see Fig. 7 of FP12). These results are in line with past studies showing that storm mergers in general are a common avenue to bow echo formation (Klimowski et al. 2004). As reviewed by FP12, several past studies have examined individual cases of squall line supercell mergers (e.g., Fujita 1978; Goodman and Knupp 1993; Sabones et al. 1996; Wolf et al. 1996; Wolf 1998; Calianese et al. 2002; Sieveking and Przybylinski 2004). These studies have documented a number of features common to these events, including tornado development proximal to the merger time (e.g., Goodman and Knupp 1993; Sabones et al. 1996; Wolf et al. 1996; Wolf 1998), evolution toward bow echoes (Fujita 1978; Calianese et al. 2002; Sieveking and Przybylinski 2004), and an overall sustenance of the supercell postmerger (Sabones et al. 1996; Wolf et al. 1996; Wolf 1998). However, for the most part, these works, as with FP12, were unable to fully determine the storm-scale processes at work in these merger events. Goodman and Knupp (1993) observed distortions to the squall line s gust front during the merger case that they investigated. They hypothesized that the mesohigh associated with the supercell s outflow may have locally blocked the advance of the squall line s cold pool, slowing its forward progress in the vicinity of the merger. However, their analysis focused on tornado development associated with the merger rather than bow echo formation so it is unclear what role this process may play in the bow echo evolution observed by FP12. In short,it remains unclear how the supercell merger may influence the development of important bow echo structures such as rearinflow jets and bookend vortices. With this in mind, the present study has used a series of idealized model simulations to investigate the stormscale dynamics at work in cases of squall line supercell mergers. While these simulations focused on the specific situation of mergers between squall lines and supercells, the results are also broadly applicable to the role that mergers play in bow echo genesis more generally. In particular, the simulations and the subsequent analysis were designed to address four main questions: 1) How does the initial squall line s cold pool evolve during the merger, and how does this impact the subsequent morphology of the merged system? 2) Given a favorable background environment, what role does the merger play in forming the bow echo, particularly in terms of modulating the formation of key bow echo features such as rear-inflow jets, bookend vortices, and strong cold pools? 3) In light of the severe storm report data analyzed by FP12, does the merger location represent a region of particularly severe weather compared to the rest of the squall line? 4) Finally, what impact does the specific location of the merger along the squall line play on subsequent morphology and in producing the range of bow echo modes observed by FP12? By focusing on the bow echo aspect of these merger cases this study examines the merger from the squallline perspective (i.e., how the squall line changes in response to the merger with the supercell). A subsequent paper will discuss the storm-scale evolution of the vertical vorticity field in these cases, evaluating the merger from the supercell perspective. Section 2 details the experimental setup for our simulations, including a novel means of introducing two distinct modes of convection into an idealized cloud model simulation with a homogeneous base state. This is followed in section 3 by an overview of the basic simulations, and a comparison between simulations with and without a merger. In section 4 we provide a detailed analysis of the evolution of the cold pool, the development of the bow echo, and the severe wind generation mechanisms associated with our simulated squall line supercell merger. The sensitivity of the postmerger evolution to the background wind profile and merger location are explored in section 5. Finally, in section 6, we conclude by summarizing our results in relation to the observations presented by FP12 and other past works, and set the stage for a companion paper that will focus on the evolution of the supercell during these types of events. 2. Idealized simulation setup This work utilized 3D idealized numerical model simulations using version 1.16 of the Bryan cloud model (CM1) described by Bryan and Fritsch (2002). We used a horizontal grid spacing of 500 m in order to sufficiently resolve convective-scale processes while also keeping computing costs manageable given the 300 km km 3 20 km grid necessary to simulate a squall line, supercell, and merged system. The vertical grid spacing was stretched from 100 m at the surface to 250 m above z m. We employed open x and y lateral boundary conditions, free-slip upper and lower boundary conditions, and a Rayleigh damping layer above 14 km. In the interest of keeping the simulations as simple as possible, radiative effects, surface friction and surface fluxes were all neglected. The simulations include Coriolis forcing, applied to perturbation winds only at a constant value of f s 21 across the entire domain (i.e., an f plane). This was included because initial tests revealed that it was necessary in order to produce the

3 DECEMBER 2014 F R E N C H A N D P A R K E R 4793 FIG. 1. Skew T logp diagrams and hodographs depicting (a) the horizontally homogeneous background environment for the initial 3 h of simulation time and (b) the far-field, pre-squall-line environment including the updated wind profile introduced by BSS and squall-lineinduced perturbations to the environment 3 h into the simulation. Insets display values for CAPE, CIN, 0 3-km helicity, and stormrelative helicity, and 0 1, 0 3, and 0 6 km AGL bulk vector wind difference for each environment. Color scheme on the hodographs identifies the 0 1- (green), 1 3- (blue), 3 6- (red), and.6-km (black) layers. asymmetric structures (i.e., a dominant cyclonic line-end vortex at the north end of the squall line) observed following real-world mergers. This is not surprising as the convergence of planetary vorticity has been shown by a number of studies to be important to the development of cyclonic mesoscale vortices over a wide range of scales (i.e., Weisman 1993; Skamarock et al. 1994; Trapp and Weisman 2003; Atkins and St. Laurent 2009b). The present simulations used a horizontally homogeneous background environment (Fig. 1a), based on the idealized environment of Weisman and Klemp (1982) that has been widely used in the simulation of convective storms. The squall line was triggered using a 200-km-long (y dimension), 20-km-wide (x dimension), and 2.8-km-deep (z dimension) linear warm bubble with a potential temperature perturbation of 12 K. The potential temperature excess was maximized along the bubble s center line and decreased to 0 following a cosine function over a 10 km (1.4 km) horizontal (vertical) radius. Random noise of 60.1 K was added to the thermal to help develop three-dimensional structures along the line. Trial and error revealed that the main challenge in simulating a squall line supercell merger in the desired idealized setting lies with producing both convective modes simultaneously within a single simulation: in many cases a simulation that produced a reasonable supercell storm would not produce a persistent squall line, and vice versa. This stems from the long-understood concept that convective organization is strongly tied to the background environment, particularly the wind profile (e.g., Weisman and Klemp 1982, 1984; Rotunno et al. 1988). In nature, the presence of environmental heterogeneity and strong linear forcing are often important to producing multiple modes in a localized region (e.g., Richardson et al. 2007; French and Parker 2008). However, trying to include such heterogeneity in our idealized model would limit our ability to run controlled tests focused on the role that the storm merger is playing in convective evolution. To address this issue, we employed the base-state substitution (BSS) technique of Letkewicz et al. (2013) to utilize two different wind profiles during the simulation. First, we initiated a squall line in an environment characterized by a favorable, unidirectional wind profile (Fig. 1a) and let it mature for 3 hours, essentially the time necessary for the line to mature to a quasi-steady

4 4794 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 intensity. At this point, the BSS technique was employed to replace the background wind profile with one more representative of a supercell environment [i.e., strong (25 m s 21 ) deep-layer (0 6 km AGL) bulk layer vector wind difference and a low-level shear vector that veers with height (Fig. 1b)]. As detailed by Letkewicz et al. (2013), this was done by separating the original basestate wind profile from the perturbations that had developed in the course of running the 3-h squall-line simulation, introducing the new base-state wind profile, and then adding the original storm-induced perturbations back on to the new wind profile. In doing this we were able to create an environment that would support an isolated supercell, while still maintaining the physical perturbations to the wind and thermodynamic fields embodied by the squall line. This change in environment was completed over a single model time step (e.g., the instant BSS of Letkewicz et al. 2013) in order to limit the amount of time necessary for the new wind profile to take effect. Only the wind profile was changed in this manner; the base-state thermodynamic profile was left untouched. 1 Once the modifications were complete, the simulation was restarted using the new background environment and the supercell was triggered 60 km ahead of the squall line at y km using a 11-K spheroid warm bubble with horizontal and vertical radii of 10 km and 1.4 km, respectively. The y position of the bubble was chosen to simulate a merger near the northern end of the squall line, similar to the system-scale bowing paradigm of FP12, which was their most common postmerger evolution in cases of weak synoptic forcing. Care was taken to ensure that there were no lingering effects of the BSS on our analysis of the simulated squall line supercell merger. Aside from small, discrete changes to the domain-total mass, energy, and cloud water fields at the time of the BSS, as documented by Letkewicz et al. (2013) there was little detrimental impact of the BSS on the simulated squall line (Fig. 2). The line gradually intensified and grew more organized over the min following the introduction of the new environment, consistent with the expected behavior of a squall line encountering stronger vertical wind shear (e.g., Weisman et al. 1988; Weisman 1993). More importantly, the simulation was configured so that the merger occurred 1 While the base-state thermodynamic variables (e.g., u 0, p 0, q, y 0, etc.) do not change throughout the simulation, the full variables (e.g., u 0 1 u 0 ) do respond to the presence of the squall line, which has perturbed the far-field environment. This accounts for the changes in CAPE and CIN between Figs. 1a and 1b. If the basestate variables alone were plotted in Fig. 1b, the thermodynamic profile would be identical to Fig. 1a. approximately 90 min after the application of the BSS, so any adjustments to the new environment would not impact the analysis of the merger itself. It should be emphasized that the purpose of the unidirectional shear pre-bss environment was to facilitate the development of a well-organized squall line that could then be introduced to an environment that supported supercells. Our analysis will focus on the impact of the merger within the highly sheared, post-bss environment, rather than the effects of change in environment. A sensitivity test simulating a squall line with the environmental modifications applied gradually over an hour (e.g., the gradual BSS of Letkewicz et al. 2013) was not substantially different from that run with the instant BSS once the environmental modification was complete. The gradual method did, however, require an additional hour of simulation time, so the instant method was chosen in favor of computational efficiency. We ran two primary simulations using the instant BSS method. The first simulation includes the wind profile modification after 3 h, but does not trigger the supercell. It is intended to serve as a baseline for how the squall line evolves in the absence of the merger and will be referred to as the NOMERGER simulation. In the second simulation, the modified wind profile and initiating bubble for the supercell are added 3 h into the simulation in order to simulate the squall line supercell merger. This will be referred to as the MERGER simulation. An additional simulation was run that included the supercell in isolation in the higher shear environment in order to compare its evolution with the supercell in the MERGER case. Finally, we ran a series of sensitivity tests using different wind profiles and merger locations. The details of these simulations will be discussed when they are introduced in section Overview and comparison of the basic simulations We begin our analysis with an overview of the NOMERGER simulation as a baseline example of how a squall line evolves in the simulated environment. Following the introduction of the stronger vertical wind shear environment (t min into the simulation), the NOMERGER squall line gradually strengthens and organizes into a broad bow echo characterized by counterrotating bookend vortices and a well-organized rearinflow jet between y and y km (Figs. 3a f). This is the expected result, as bow echo structures are favored in environments characterized by strong deeplayer wind shear (e.g., Weisman 1993; Evans and Doswell 2001). The bow echo becomes increasingly asymmetric over the duration of the simulation, with the northern,

5 DECEMBER 2014 FRENCH AND PARKER 4795 FIG. 2. Simulated radar reflectivity (dbz, shaded), 22-K surface u0 (black contour), and 1 km AGL wind vectors (m s21, scale vector below color bar) for the (a) (e) base-state simulation with no BSS and (f) (j) NOMERGER simulation between (left to right) 120 and 240 min into the simulation. The new environment is applied 180 min into the simulation in (g) (j). cyclonic, bookend vortex eventually becoming dominant (Figs. 3d f). All told, the bowing and asymmetric structure of the simulated squall line are in line with what is expected given the environment and presence of Coriolis accelerations. The evolution of the squall line in the MERGER simulation parallels that of the NOMERGER simulation through approximately t min into the simulation (Fig. 4a). At this point, the northern end of the line begins to decline in intensity as it interacts with outflow from the supercell ahead of the line. By t min the supercell is located just ahead of the north end of the line (Fig. 4b), and the structure of the squall line differs considerably from the NOMERGER simulation. The convective line west of the supercell has weakened, and the rear-inflow jet is weaker and much smaller in areal extent than that produced by the NOMERGER squall line (c.f., Figs. 3b and 4b). As the supercell begins to merge with the line, simulated reflectivity associated with the northern end of the squall line weakens further (Fig. 4c), and eventually the supercell becomes the dominant feature. A strong rear-inflow jet develops just south of the merged supercell by approximately t min (Fig. 4d), leading to a bow echo developing between y and y km (Figs. 4e,f). This rear-inflow jet is generally narrower (in along-line extent) and is displaced farther south than that in the NOMERGER squall line. Throughout the NOMERGER simulation a cyclic pattern of line-end vortex development, occlusion and redevelopment is observed near the north end of the line (Figs. 3a f). In the MERGER run, however, the initial line-end vortex moves rearward and occludes as the

6 4796 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 3. Summary plot of 1 km AGL simulated radar reflectivity (dbz, gray shading), 22-K surface u 0 (dashed purple contour), and 2.5 km AGL wind vectors and wind speed [contoured at 25 m s 21 (blue) and 30 m s 21 (red)] at (a) 240, (b) 250, (c) 265, (d) 280, (e) 295, and (f) 325 min into the NOMERGER simulation. Curved green arrows denote approximate locations of bookend vortices based on vertical vorticity magnitude s 21, Obuku Weiss number, 0, and general cyclonic or anticyclonic pattern of the wind vectors.

7 DECEMBER 2014 F R E N C H A N D P A R K E R 4797 FIG. 4. As in Fig. 3, but for the MERGER simulation. The curved yellow and red arrows denote, respectively, the approximate locations of the supercell mesocyclone and postmerger cyclonic vortex described in the text. Criteria for identifying these are as in Fig. 3.

8 4798 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 merger begins (Figs. 4a c), and a new vortex does not take its place until after the merger (Figs. 4c f). This cyclonic vortex is farther to the south and stronger than those seen in the NOMERGER simulation (cf. curved arrows in Figs. 3c f and Figs. 4c f). It initially represents the remnants of the supercell mesocyclone; however, additional vorticity production within the merged system contributes to its growth and intensification, the details of which will be presented in a future manuscript. The rear-inflow jet, strong cyclonic line-end vortex, and bow echo structure remain for the duration of the MERGER simulation (Figs. 4e,f). The simulated reflectivity structures evident during and following the supercell merger are very reminiscent of those observed for cases exhibiting the system-scale bowing evolution identified by FP12. This gives us confidence that the simulation is capturing the salient details of this squall line supercell merger archetype. If the supercell is simulated in this environment without the squall line, it remains isolated through t min (Fig. 5), and does not develop into a bow echo on its own, as is sometimes observed for high precipitation supercells (e.g., Moller et al. 1990; Finley et al. 2001). The differences in convective evolution between the MERGER and NOMERGER simulations correspond to differences in model proxies for severe weather in the respective squall lines (Fig. 6). The MERGER simulation produces a swath of strong winds, enhanced near-surface vertical vorticity, and enhanced rainfall along the portion of the squall line influenced by the MERGER, all of which are absent in this region in the NOMERGER simulation (Fig. 6). However, the two simulations appear quite similar for all three fields south of approximately y km. The supercell-only simulation also produced enhanced surface winds along its path (Fig. 5f); however, the values are not as intense as those that developed within the MERGER simulation. Simply put, the merger with a supercell appears to promote a locally more severe, damaging squall line; however, the impact is limited to the vicinity of the merger. This is in line with observations of severe weather (particularly damaging wind and tornado) reports associated with merged systems (FP12), and suggests that the merger may in fact be a preferred location for severe weather production. To better understand why this may be the case, the next section will explore the processes responsible for the evolution of the MERGER simulation. 4. Storm-scale processes a. Cold pool evolution In their observations of squall line supercell mergers, FP12 repeatedly observed a weakening of the squall line near the onset of the merger, and the key features of the supercell remained identifiable well after the completion of the merger. This is in line with mesonet data analyzed by Goodman and Knupp (1993) that revealed an apparent distortion of the squall line s gust front during a merger with a nearby supercell. These observations suggest that the supercell in some way alters the characteristics of the squall line s gust front; however, the details of this evolution are not clear. To better understand these interactions, we will explore the evolution of the squall line s cold pool in the present simulations. The evolution of the cold pool in the MERGER simulation is summarized in Figs. 7 and 8.Byt min the squall line is characterized by a largely homogeneous cold pool with an average potential temperature perturbation of 29K (Fig. 7a), and depth of 2000 m (Fig. 8a) that is producing an unbroken region of slabular (James et al. 2005) gust front lifting (e.g., w. 5ms 21 at 1 km AGL; Fig. 7f). As the supercell develops and matures, it produces a shallow (average depth of 500 m; Fig. 8b) cold pool in its wake, characterized by an average potential temperature perturbation of 24K (Fig. 7a). As this cold pool expands, it begins to interact with the squall line s gust front by approximately t min (Fig. 7b), leading to a rapid removal of the lowlevel potential temperature gradient along the squall line s gust front (Figs. 7c e). Additionally, prior to disappearing completely, the strongest potential temperature gradient appears to lag to the west slightly (dashed oval in Fig. 7c) compared to the more continuous arc present at earlier times (e.g., Fig. 7b). This would appear to be similar to the distortion of the squall line s gust front observed in the merger case studied by Goodman and Knupp (1993). The removal of the low-level gust front potential temperature gradient brings about a concurrent decline in near-surface lifting along the squall line s gust front in this region as well (Figs. 7h j). Farther aloft (between approximately 500 and 2000 m AGL), the squall line initially continues to ingest high-u e air (Figs. 8b,c) from a layer above the supercell s cold pool. However, despite the presence of an elevated layer of favorable inflow, the squall line continues to decline in intensity (e.g., Fig. 4b). This appears to be due to a decline in the strength of the line-normal vertical wind shear over the depth of the gust front as a result of the supercell perturbing the local environment. This can be seen through a visual examination of the wind vectors immediately ahead of the squall line (e.g., at approximately x 5 190, 210, and 215 in Figs. 8a, 8b, and 8c, respectively) and is quantified in Table 1. Vertical wind shear is reduced by nearly half over the depth of the initial squall-line cold pool (0 2 km AGL; Table 1) as the squall line encounters

9 DECEMBER 2014 F R E N C H A N D P A R K E R 4799 FIG. 5. (a) (e) As in Fig. 3, but for supercell-only simulation. (f) Maximum wind speed (m s 21, shaded) at the lowest model level accumulated over the duration of the supercell-only simulation. The black 3 in (f) denotes the point when the right-moving supercell splits from the initial storm and the dashed black oval indicates the region of enhanced wind associated with the right-moving supercell.

10 4800 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 6. Swaths of (a),(b) maximum wind speed (m s 21, shaded) and (c),(d) vertical vorticity (s 21, shaded) at the lowest model level, and (e),(f) rainfall (mm, shaded) accumulated between 3 and 6 h into the (a),(c),(e) MERGER and (b),(d),(f) NOMERGER simulations. Dashed black ovals denote the regions of enhanced fields in the MERGER simulation. The black 3 in each panel denotes the relative location of the merger, defined as the onset of the permanent union of the 40-dBZ contour between the squall line and supercell, which occurred at t min into the MERGER simulation.

11 DECEMBER 2014 F R E N C H A N D P A R K E R 4801 FIG. 7. (a) (e) Surface u 0 (K, shaded contours) and (f) (j) 1 km AGL w (m s 21, shaded) and 22-K u 0 (black contour) between 225 and 265 min into the MERGER simulation. Dashed black ovals denote the weakened squall-line gust front in (c),(d),(h), and (i). the supercell, and even more substantially if just the elevated inflow layers are considered (0.5 2 and 1 2 km AGL; Table 1). These layers likely become increasingly important as the near-surface layer stabilizes in response to the supercell s cold pool (French and Parker 2010). Over time the squall-line updraft tilts more dramatically rearward over its cold pool (e.g., Fig. 8c). This is consistent with the cold pool overwhelming the lowlevel shear, which is detrimental to low-level lifting and squall-line maintenance (Rotunno et al. 1988). Thus, the weakening low-level shear ahead of the squall line likely contributes to an overall decline in the gust front updraft associated with the premerger squall line, leading to the observed weakening. In short, the interaction between the squall line and the low-level outflow and local wind perturbations associated with the supercell lead to a decline in convective intensity along the northern part of the line (e.g., Fig. 4b). This is consistent with FP12 s frequent observations of the squall-line weakening or breaking prior to the merger. During this same period, a new unbroken region of low-level ascent becomes established farther east, extending south from the supercell and connecting with the squall line s gust front south of y km (Figs. 7h j). In other words, the supercell becomes the new leading edge of the northern portion of the squall line. Thus, instead of the supercell merely being overtaken by the squall line with little change in overall system structure (as will be shown for mergers in a weaker wind shear environment in section 5), it instead promotes a considerable change in squall-line intensity and organization, as was observed repeatedly by FP12. b. Bow echo development The development of the bow echo in the present simulations is a result of processes common to bow echo formation; what is unique is that these processes occur in direct response to the supercell merger. Following the

12 4802 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 8. (a) (d) Vertical cross section of u e (K, shaded), u 0 (K, colored contours), w (white contours every 5 m s 21 starting at 5 m s 21 ), and wind vectors [sample vector to the right of (c)] averaged from y to 175 km at t 5 (a) 225, (b) 245, (c) 255, and (d) 265 min into the MERGER simulation. merger, the subsequent development of the bow echo bears many similarities to the well-documented evolution from high-precipitation (HP) supercells to bow echoes (e.g., Moller et al. 1994; Finley et al. 2001; Klimowski et al. 2004). As the weakening squall line approaches the rear flank of the supercell there is a rapid increase in convective intensity and upward vertical velocity along the rear-flank gust front extending southwestward from the supercell (often referred to as the flanking line ; Figs. 9a c). This intensification occurs as the cold pool initially associated with the squall line overtakes the supercell s cold pool and deepens to approximately 3 km deep (Figs. 9d f). This is deeper than the initial squall line (2-km depth; Fig. 9d) and supercell (1.5-km depth; Fig. 9d) cold pools, and it is also deeper than the cold pool farther south along the squall line (e.g., y km, approximately 1.5 km deep, not shown). This deeper cold pool facilitates strong low-level lifting (Houston and Wilhelmson 2011) as it encounters favorable lowlevel wind shear and high u e air ahead of the now merged

13 DECEMBER 2014 F R E N C H A N D P A R K E R 4803 TABLE 1. Squall-line-normal, low-level bulk wind difference (DU) between t and t min into the MERGER simulation. Values represent averages between y and y km taken at the x positions indicated. Time (min) x location (km) km DU (m s 21 ) km DU (m s 21 ) km DU (m s 21 ) system (e.g., Fig. 8d), leading to the rapid intensification of deep convection. The heavy precipitation associated with this increase in convection locally enhances the cold pool even further (surface potential temperature perturbation 4 K colder than surrounding portions of the system s cold pool) by t min (Figs. 10a,c,e). As discussed by James et al. (2006), such a localized intensification of the cold pool is favorable for bow echo development, as it causes a portion of the squall line to locally overwhelm the ambient wind shear and accelerate, producing the bow shape in the system s gust front. Indeed, in the present simulations the region of colder outflow south of the supercell s updraft begins to accelerate eastward, signaling the onset of bow echo formation (Fig. 10e). This locally intense cold pool is also colder than the cold pool produced in the NOMERGER simulation during its period of bow echo development, which occurred earlier in the simulation (Figs. 10b,d,f). Another key feature in the development of the bow echo is the rapid acceleration of a narrow rear-inflow jet [RIJ; a common feature in bow echoes; e.g., Fujita (1978); Przybylinski (1995); Wakimoto (2001)] in the vicinity of the merger (Figs. 4d f). This RIJ is more localized in the along-line direction than the broad jet that characterizes the NOMERGER squall line and forms in response to a localized region of strongly depressed pressure that develops aloft (approximately 2.5 km AGL) following the merger, extending south from the merged supercell (Figs. 11a,c,e). This negative pressure perturbation develops in response to the vertical buoyancy gradient 2 between the strong cold pool and the strong convective heating aloft associated with enhanced convection in the vicinity of the merger (Fig. 12b). This is a common mechanism for RIJ 2 As can be shown using the diagnostic pressure equation [e.g., Eq. (2.131) of Markowski and Richardson (2010)], regions of buoyancy increasing with height are associated with a local minimum in perturbation pressure. development in squall lines (e.g., Lafore and Moncrieff 1989; Fovell and Ogura 1988; Weisman 1993; Grim et al. 2009); however, in the MERGER simulation the location and magnitude of the buoyancy gradient are directly related to the merger itself, as this drives the strengthening cold pool and deep convection. If we compare these results to the NOMERGER simulation, we find that the NOMERGER squall line maintains a broad RIJ (Fig. 3), which reflects a system-scale region of lower pressure near the leading edge of the line (Figs. 11b,d,f). Furthermore, while the perturbation pressure values are similar between the MERGER and NOMERGER simulations, there is a much stronger gradient in the MERGER case, resulting in a larger acceleration and stronger RIJ (Fig. 11). As the bow echo matures, it is further enhanced by a pair of counter-rotating vortices at the north and south ends of the bow (Figs. 4d f). The northern, cyclonic vortex appears collocated with the merged supercell, while the southern, anticyclonic vortex has persisted from the premerger squall line. The vortices serve to further enhance the RIJ as it expands and intensifies (e.g., Weisman 1993). As with the buoyant pressure perturbations, the bookend vortices in the MERGER simulation are stronger than those observed in the NOMERGER run, and are located closer together, implying that they would have a larger positive effect on the RIJ. To summarize the development of the bow echo postmerger, the role of the merger is to locally strengthen the cold pool of the merged system by increased precipitation. The strengthened cold pool causes the gust front to locally bow eastward and promotes the development of a strong RIJ via buoyant pressure perturbations associated with latent heat release in the convective line. This jet is further enhanced by a pair of counter-rotating vortices that intensify following the merger. The localized nature of the strengthened cold pool and RIJ produce a more compact bow echo than seen in the absence of the merger. These processes are common to bow echoes; however, in this case the merger instigates and amplifies the processes to produce a locally stronger bow echo. c. Damaging wind production In addition to the pronounced bowing that evolves, another key feature of the MERGER simulation is the production of a broad swath of very strong surface winds associated with the merged system (Fig. 6a). This finding compares well with the observations of FP12, who found that severe wind reports were maximized postmerger in squall line supercell merger cases.

14 4804 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 9. (a) (c) Simulated radar reflectivity (shaded, dbz), and w at 2.5 km AGL [colored contours, m s 21, color scheme shown on right side of (f)]. (d) (f) Along-line averaged u 0 (K, shaded) and w [contours as in (a) (c)] averaged between y and y km [horizontal black lines in (a) (c)]. Plots are at (left to right) t 5 255, 265, and 275 min into the MERGER simulation. An analysis of the trajectories of parcels 3 launched during the MERGER simulation reveals two primary source regions for air parcels that end up in the regions of strong winds at the lowest model level as the merger commences (Fig. 13). These include low- and midlevel parcels (Fig. 13, orange and blue trajectories, respectively) that descend while curving cyclonically into the system from the north, and midlevel parcels (Fig. 13, green trajectories) that descend toward the surface from the rear of the squall line. The first region represents parcels that are moving through the remnant supercell and descending through the remnant rear-flank downdraft (RFD), as the trajectories resemble the RFD parcels from past studies of simulated supercells (e.g., Wicker and Wilhelmson 1995; Adlerman et al. 1999; Xue 2004; Dahl et al. 2012). Early in the merger process, the majority of parcels originate in this region, and 3 A set of parcels were initialized within an 80 km 3 80 km 3 4 km box centered on the merged system at 10-min increments throughout the merger and integrated forward in time for 1 h during the simulation. Forward trajectories were chosen as they can be computed during the model run, greatly reducing errors in location compared to trajectories calculated from history files with a courser time resolution (e.g., Dahl et al. 2012). terminate either in a localized region of strong winds that is associated with a downburst (e.g., Fujita 1985) within the RFD of the merged supercell or in the broader swath of severe wind farther south (Fig. 13). The second region (green trajectories) is consistent with air parcels within the squall line s rear-inflow jet descending to the surface, and is a common feature in squall-line and bow echo flow fields. Early in the merger process (e.g., the trajectory window from min into the simulation), these parcels are primarily associated with a broad region of less intense strong winds generally south of the merger location (Fig. 13). What is significant during this time is that the trajectories are spread out over a comparatively broad region with minimal overlap, particularly between the orange and green trajectories. This implies that there are initially two mechanisms producing severe winds during the early evolution of the merger the remnant RFD within the supercell and a descending RIJ from the squall line both of which are producing strong near-surface winds in different portions of the merged system. The onset of severe wind production within the remnant RFD appears to be a result of an intensification of the downdraft during the merger. The remnant RFD intensifies in response to heavy precipitation developing

15 DECEMBER 2014 F R E N C H A N D P A R K E R 4805 FIG. 10. Surface u 0 (K, shaded), and 1 km AGL simulated radar reflectivity (black contours at 45 and 50 dbz)att 5 (a) 280, (c) 285, and (e) 290 min into the MERGER simulation and t 5 (b) 255, (d) 260, and (f) 265 min into the NOMERGER simulation. along the rear flank of the supercell as it merges with the squall line, and there also appears to be some phasing of the RFD with preexisting downdrafts associated with the initial squall line (Figs. 14a c). The link to the merger is apparent as the RFD observed at a similar time in the supercell-only simulation was weaker in intensity and smaller in areal extent (not shown). The severe winds associated with the RIJ during this period appear as a result of an intense, expansive downdraft (Fig. 14c) associated with the leading convective line transporting

16 4806 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 11. Pressure perturbation (hpa, shaded) and wind vectors (m s 21, scale vector at lower right) at 2.5 km AGL, and 45- and 50-dBZ simulated radar reflectivity contours at 1 km AGL at t 5 (a) 270, (c) 285, and (e) 290 min into the MERGER simulation and t 5 (b) 245, (d) 260, and (f) 265 min into the NOMERGER simulation.

17 DECEMBER 2014 F R E N C H A N D P A R K E R 4807 FIG. 12. Structure of pressure perturbations in MERGER simulation at t min. (a) Plan view of pressure perturbations (hpa, shaded) and winds at 2.5 km AGL. (b) Vertical cross section of buoyancy (m s 22, shaded), pressure perturbation (black contours every 1 hpa with 0 contour omitted and negative values dashed white), and winds. The dashed line in (a) denotes the plane of the cross sections in (b). In both panels, wind vectors.20,.25, and.30 m s 21 are colored black, blue, and red, respectively. the high-momentum RIJ air to the surface. The severity is likely due to the RIJ itself intensifying during this period, as discussed in the previous section. Over time, as the bow echo structure takes shape, the strongest near-surface winds become concentrated into a narrow swath as shown in Fig. 6a. The air parcels within this swath originate in the same two source regions discussed above; however, the details of the trajectory paths have changed (Fig. 15). First of all, there is considerable overlap between the trajectories from both regions (i.e., they are both contributing to the same swath of surface winds; Fig. 15). Second, a considerably larger fraction of the parcels at this point are sourced in the descending rear inflow, suggesting that the main mechanism for the elongated swath of surface winds is the descending rear-inflow jet. Finally, the path of the northern (orange and blue) parcels appears consistent with parcels wrapping around the developing line-end vortex and ascending the northern flank of the merged system s cold pool prior to descending in a system-scale downdraft toward the surface. This implies that the remnant RFD associated with the supercell is no longer a distinct feature as the supercell and squall line have merged into a coherent system. Since most of the parcels during this period are sourced from the RIJ region, this likely represents the primary mechanism responsible for the severe surface winds. The additional air wrapping into the system from the line-end vortex then serves to further accelerate the low-level winds owing to the additive effect of the rotational flow on the RIJ. This is akin to the enhancement of an RIJ by bookend vortices described by Weisman (1993) or the acceleration of near-surface winds by mesovortices (e.g., Trapp and Weisman 2003; Wakimoto et al. 2006; Atkins and St. Laurent 2009a). The result is a localized swath of very strong surface winds that follows the track of the descended RIJ and the southern flank of the line-end vortex associated with the merged system.

18 4808 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 13. (bottom left) Plan view, (top) x z cross section, and (bottom right) y z cross section of 30-min-forward parcel trajectories ending at t min into the MERGER simulation that passed through regions of near-surface winds. 30 m s 21 between t and t min. The starting point for each trajectory is denoted by an open circle, with the trajectory colors signifying three parcel source regions of interest: green parcels originated to the rear of the squall line (RIJ), and orange and blue parcels originated in the low levels (0 2 km AGL) and midlevels (2 4 km AGL) in the vicinity of the remnant supercell, respectively. Total counts of each parcel type are shown (by color) in the top right. Shading in the plan view plot denotes maximum surface wind speed during the min window (m s 21, shaded) and the 45-dBZ simulated radar reflectivity contour (purple). Only a subset of the parcels is plotted for clarity. The importance of the merger in creating the above conditions is underscored by the lack of a similar region of severe winds within the NOMERGER simulation (cf. dashed ovals in Figs. 6a,b), which is due to the RIJ remaining elevated in that run (Figs. 16a c). The descending RIJ in the MERGER simulation is located within a broad (across-line dimension), persistent region of heavy precipitation (Figs. 16d f) that develops in FIG. 14. (a) (c) As in Figs. 9a c, but for w, 0[ms 21 color scheme at right side of (c)], at t 5 (a) 260, (b) 275, and (c) 280 min into the MERGER simulation.

19 DECEMBER 2014 F R E N C H A N D P A R K E R 4809 FIG. 15. As in Fig. 13, but for 30-min trajectories ending at t min into the merger simulation. response to the storm merger. This heavy precipitation produces a persistent region of downward vertical velocity that penetrates nearly to the surface in the same region, facilitating the descent of the RIJ (Figs. 16d f). During this same time, the region of maximum bowing in the NOMERGER simulation is characterized by a much narrower (across line) region of heaviest precipitation, and it is less intense than that in the MERGER run (Figs. 16a c). Most of the downdraft activity remains above 1 km AGL, which results in the RIJ remaining elevated as well (Figs. 16a c). Differences in the static stability of the cold pool in the two simulations may also contribute to the differences in RIJ behavior (e.g., Weisman 1993). In the NOMERGER simulation, the cold pool is deeper on average, with the coldest potential temperature perturbations (u 0,210 K) extending well to the rear of the leading line (Figs. 17a c). The cold pool in the MERGER simulation, on the other hand, contains a localized region of deep, cold air near the leading line, with a shallow layer of comparatively warm air (u 0.26 K) extending toward the rear (Figs. 17d f). The warmer and shallower cold pool represents a less stable environment where there is less resistance to the descent of the RIJ. As a result, the increased precipitation forces the descent of the high-momentum RIJ air to the surface. This could also be thought of in terms of the buoyancy force opposing the downdraft within the squall line. In the absence of a deep, strong cold pool, the descending parcels remain cooler than their surrounding environment, resulting in less positive buoyancy to decelerate the downdraft, which allows the RIJ to descend. The difference in cold pool characteristics is a remnant of the initial weakening of the squall line discussed in section 4a even though the cold pool was locally stronger near the leading edge of the line (Fig. 10). To summarize, the descent of the RIJ in the MERGER simulation is driven by enhanced precipitation associated with the merger, occurring in the presence of a weakly stable cold pool. This is a common mechanism and condition for descending rear inflow in squall lines (e.g., Smull and Houze 1987; Yang and Houze 1995; Mahoney and Lackmann 2011). The unique element in this case is that the merger with the supercell produces the increase in precipitation that drives this process while also creating the condition leading to diminished cold pool stability. Without the merger, the RIJ does not descend in the same manner, and the resultant swath of severe surface winds does not occur.

20 4810 M O N T H L Y W E A T H E R R E V I E W VOLUME 142 FIG. 16. Vertical cross sections of along-line-averaged hydrometeor mixing ratio (sum of rainwater, graupel, cloud water, cloud ice, and snow mixing ratios, kg kg 21, shaded), horizontal wind speed (m s 21, gray contours with 30 m s 21 contoured in red), and w, 1ms 21 (vectors, scale vector at bottom right) at (left to right) t 5 280, 305, and 315 min into the (a) (c) NOMERGER and (d) (f) MERGER simulations. Values are averaged between y km in the NOMERGER simulation and y km in the MERGER simulation to capture the regions of strongest bowing and rear inflow. 5. Sensitivity tests Based on the above analysis, we have shown one basic pathway to bow echo development in cases of squall line supercell mergers. However, given the complexity of this scenario, it is worth exploring how sensitive these results are to our choices in model configuration. This section will explore the sensitivity of our results to 1) the background wind profile and 2) the location of the merger along the squall line. a. Sensitivity to background wind profile To explore the importance of having a background wind profile that favors right-moving supercells in promoting the development of the bow echo in the MERGER simulation, a set of additional simulations were run using unidirectional wind profiles with the vertical shear confined to the lowest 2.5 km. The first of these used a km AGL bulk wind differential of 17.5 m s 21 (hereafter, the MERGER17.5 and NOMERGER17.5 simulations), sufficient to maintain the organization of the squall line while limiting the preline convection to a multicellular mode. The second set used a km AGL bulk wind differential 25 m s 21 in order to organize the initial squall line into a bow echo, while producing splitting supercells ahead of the line that were much weaker and smaller than the well-organized right-moving supercell in the MERGER simulation. For both wind profiles, the squall lines weakened as they interacted with the preline convection s outflow, and the preline storms became the new leading edge of the merged system, similar to the original MERGER simulation. However, there was no enhancement to any bowing structures. In short, following the merger, the MERGER17.5 and MERGER25 systems maintained a strong resemblance to their nonmerger counterparts (cf. Figs. 18c,d and 18e,f). While these mergers had little impact on storm organization, both merged systems did produce an increase in severe winds compared to the non-merger systems. The MERGER17.5 produced a localized region of enhanced surface winds in the vicinity of the merger (Fig. 19a), and the MERGER25 produced a similar behavior, although it