Airborne Radar Observations of a Cold Front during FASTEX 1

Transcription

1 JULY 2000 WAKIMOTO AND BOSART 2447 Airborne Radar Observations of a Cold Front during FASTEX 1 ROGER M. WAKIMOTO AND BRIAN L. BOSART Department of Atmospheric Sciences, University of California, Los Angeles, Los Angeles, California (Manuscript received 16 March 1999, in final form 22 October 1999) ABSTRACT A detailed analysis using airborne Doppler radars of an oceanic cold front associated with precipitation core and gap regions is presented. The precipitation cores appear to form as a result of the combined effects of horizontal shearing instability and the advection of hydrometeors by the core-relative winds. In contrast to previous schematic models, it is shown that a strong surface discontinuity does not exist along the entire length of the precipitation core. The southern section of the core can be accompanied by an abrupt discontinuity and lighter precipitation while the northern section can be associated with more slowly changing variables but heavier precipitation. The peak updrafts at the leading edge of the front appeared to be primarily driven by frictional convergence and the acceleration of the vertical vorticity in the boundary layer. The overall motion of the cold front was not well predicted using density current theory even though the kinematic structure of the front resembled classical studies of these types of flows. Local regions of the cold front in the vicinity of the precipitation cores, however, did appear to propagate as a density current in a direction perpendicular to the major axis of the cores. A large gap region ( 10 km) within a narrow cold frontal rainband is examined. While small gaps are generated by shearing instability, the large gap is created by differential movement of two segments of the front. 1. Introduction The complete treatise of the general characteristics of cold and warm fronts within the context of an evolving extratropical cyclone was first elucidated by the Bergen School in Norway. The identification of fronts as a focal point for precipitation within cyclones has led to a number of subsequent studies. It is now well known that cold and warm fronts nominally differ in their direction of motion, speed of propagation, vertical structure, and their accompanying cloud patterns. The conceptual model of the structure of fronts, however, has continued to evolve with the advent of more sophisticated observational tools (e.g., Shapiro et al. 1985; Keyser 1986). The cold front that propagates into a neutrally rather than unstably stratified air mass is often characterized by line convection or narrow cold frontal rainbands (NCFRs) and has garnered the most attention in recent 1 While this paper was in press, Brian Bosart was killed in an automobile accident. He was a graduate student nearing the completion of his Ph.D. thesis. This and future papers will be published as a memorial to him. Corresponding author address: Prof. Roger M. Wakimoto, Dept. of Atmospheric Sciences, University of California, Los Angeles, 405 Hilgard Ave., Los Angeles, CA roger@atmos.ucla.edu years (e.g., Browning and Harrold 1970; Hobbs and Biswas 1979; James and Browning 1979; Matejka et al. 1980; Hobbs and Persson 1982; Bénard et al. 1992; Chen et al. 1997). James and Browning (1979) and Hobbs and Biswas (1979) were the first to show that line convection often consists of an alternating series of precipitation core and gap regions. The precipitation cores are typically aligned at a clockwise angle with respect to the synoptic-scale cold front (e.g., Hobbs and Persson 1982). Horizontal shearing instability developing along the cold front has frequently been advanced as the generating mechanism leading to the core gap structure (e.g., Carbone 1982; Hobbs and Persson 1982; Parsons and Hobbs 1983; Moore 1985). More recent work by Locatelli et al. (1995), Browning and Roberts (1996), and Brown et al. (1998) has questioned whether this instability alone can account for this structure. Locatelli et al. (1995) hypothesized that the advection of precipitation by the core-relative winds can influence the shape of the precipitation core. Unfortunately, their wind syntheses, based on data collected by ground-based Doppler radars, were insufficient to calculate trajectories. Moreover, Brown et al. (1999) recently proposed that a trapped gravity wave rather than shearing instability was responsible for the precipitation core gap regions examined in their study. Fewer than 20% of gap regions are characterized with 2000 American Meteorological Society

2 2448 MONTHLY WEATHER REVIEW VOLUME 128 length scales greater than km (James and Browning 1979). Locatelli et al. (1995) have referred to these regions as large gaps and have suggested that they are dynamically distinct from the smaller gaps. To the authors knowledge, the only documentation of these large gaps has been by tracking radar reflectivity patterns. No detailed kinematic discussion has been presented in the literature. Quantitative analyses of the three-dimensional structure of intense NCFRs synthesized from multi-doppler radars have been conducted by Carbone (1982) and Roux et al. (1993). These two studies provided a plethora of information concerning the structure of these systems; however, the precipitation cores in their study tended to be much larger than is normally the case and the axis of the cores did not exhibit any rotation with respect to the surface front as is commonly observed. Locatelli et al. (1995) and Braun et al. (1997) appear to be the only observational studies that have investigated the finescale structure of a classical NCFR analogous to those documented by Hobbs and Biswas (1979) and James and Browning (1979) based on multi-doppler wind syntheses with relatively coarse spatial resolution. This paper focuses on the detailed analysis of a cold front based on data collected by airborne Doppler radars during the Fronts and Atlantic Storm Track Experiment (FASTEX; Joly et al. 1997) second intensive observation period (IOP2) on 12 January The front was characterized by well-defined precipitation core and gap regions. One of the airborne radars flew a parallel track to the cold front at a distance 10 km providing greater spatial resolution of the wind syntheses than those presented by Locatelli et al. (1995) and Braun et al. (1997). This improved resolution reveals details of the kinematic structure of an NCFR that have not been previously available. In addition, coordination with a second aircraft equipped with a Doppler radar resulted in greater areal and temporal coverage of the frontal zone than would have been possible with a single aircraft. A brief overview of FASTEX and a description of the airborne platforms are presented in section 2. Section 3 provides a description of IOP 2, the flight track of the Electra, and the overall echo structure along the cold front. A finescale analysis of the precipitation core and gap region of an NCFR is shown in section 4. Trajectories are calculated in order to test the hypothesis advanced by Locatelli et al. (1995) that the core structure is largely determined by the advection of hydrometeors by the core-relative winds. Section 5 documents a large gap region between two consecutive precipitation cores. Although discussed in previous studies, the present case is the first to reveal the kinematic structure within this region. A summary and discussion are presented in section FASTEX and the aircraft platforms The field phase of FASTEX occurred in January and February The major goal of the experiment was TABLE 1. Characteristics of the ELDORA and P-3 radars. Descriptions ELDORA P-3 Antenna rotation rate ( s 1 ) No. of samples Pulse repetition frequency (Hz) Gate length (m) Sweep-angle resolution ( ) Along-track resolution (m) Maximum range (km) Maximum unambiguous velocities ( ms 1 ) / to improve forecasts of end-of-storm-track cyclogenesis over the eastern Atlantic Ocean in the h range (Joly et al. 1997). Primary scientific objectives were to test hypotheses on frontal cyclogenesis, to understand and improve the predictability of cyclones, and to document the meso- and microscale organization of cyclone cloud systems. Major observing facilities for FASTEX included four ships capable of releasing soundings, six research aircraft, and a number of buoys deployed over the Atlantic Ocean. The two primary platforms used in the present study are the National Oceanic and Atmospheric Administration P-3 and the National Center for Atmospheric Research Electra Doppler Radar (ELDORA). Both of these aircraft were deployed from Shannon, Ireland. The P-3 and ELDORA are equipped with X-band tail radars and a suite of probes capable of recording in situ measurements (see Jorgensen et al. 1983; Hildebrand et al. 1994, 1996; Wakimoto et al. 1996). In addition, the P-3 is also equipped with a lower fuselage C-band radar that records horizontal scans of radar reflectivity to extended ranges. The scanning parameters for the radars are shown in Table 1. ELDORA is capable of greater along-track and sweep-angle resolutions than the P-3. This difference is primarily attributable to the multifrequency transmission of the EL- DORA transmitters that allow for a faster scanning rate of the tail antenna (see Hildebrand et al. 1994, 1996). Discussion of the wind synthesis technique is presented in the appendix. 3. IOP 2 and overall echo structure An open wave associated with a weak low pressure system was identified in the infrared satellite imagery and the surface analyses at 1800 UTC (hereafter, all times are in UTC LST in Ireland) in the evening of 12 January 1997 (Fig. 1). No resolvable closed circulation is shown west of Ireland along this open wave. A separate low pressure center at the surface and accompanying trough was located west of the open wave at approximately 55 N and 28 W. The analysis at the 850-mb level at 1200 is shown in Fig. 2. The upperlevel trough in Fig. 2 was accompanied by a potential vorticity anomaly (not shown) and was forecast to enhance the development of the open wave. Note the weak

3 JULY 2000 WAKIMOTO AND BOSART 2449 FIG. 1. (left) Infrared satellite image at 1800 UTC 12 Jan The flight track of the Electra is shown by the black line. The Electra and P-3 were based in Shannon, Ireland. (right) Surface analyses at 1800 UTC superimposed on the corresponding infrared satellite image. Surface wind reports are also plotted (full barb, 5ms 1 ; half barb, 2.5 m s 1 ). The black dots represent station locations where thermodynamic information was recorded but where there were no wind reports. baroclinic zone at this level in the vicinity of the open wave. The cold front near 20 W was targeted by both the P-3 and Electra. The purpose of the mission was to collect a high-resolution dataset to investigate the kinematic variability along the cold front. The flight track of the Electra is shown in Figs. 1 and 3 and was flown at 600 m mean sea level (MSL, hereafter all heights are MSL). Figure 3 reveals that the Electra executed eight legs along the front providing continuous data collection for nearly 3.5 h. Most of the tracks were flown in an elongated box pattern with the cold front centered within the box. The longer legs ( 100 km) were parallel to the front in order to ensure sampling of a significant portion of the front. Synthesis of the Doppler winds at high resolution was achieved by flying these long transects at a distance from the front that was not to exceed 10 km. Extensive radio communication between the mission scientists on board the two aircraft resulted in the execution of an analogous box pattern FIG. 2. Upper-level analysis at the 850-mb level at 1200 UTC 12 Jan Geopotential height and isotherms are shown by the black and dashed lines, respectively. Notation for wind reports is the same as in Fig. 1. FIG. 3. Flight track of the Electra on 12 Jan The white area represents the west coast of Ireland. The airport at Shannon is denoted by the star.

4 2450 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 4. Time series of radar echoes at 400 m MSL illustrating the movement of the precipitation cores associated with the narrow cold frontal rainband on 12 Jan Reflectivity contours are based on the tail radar data recorded by ELDORA. The gray lines represent radar reflectivity contours with values greater than 35 dbz shaded gray. The black line represents the axis of the maximum in vertical vorticity associated with the cold front. The black dots represent individual point maximum as shown in the inset. The thin dashed line represents the movement of the individual precipitation cores. The thick dashed line denotes the end of the wind syntheses, which also marked the end of the flight legs. (not shown) by the P-3 that ranged between 25 and 50 km to the west of the Electra tracks shown in Fig. 3. The parallel tracks by the two aircraft resulted in merged wind syntheses for several passes that were nearly 100 km long in a direction perpendicular to the front. A time series of the echo structure along the leading edge of the cold front at 400 m based on the ELDORA syntheses is presented in Fig. 4. The classical picture of a narrow cold frontal rainband associated with precipitation core and gap regions is shown. The black line in the figure represents the axis of maximum vertical vorticity. This line was the best discriminator of the kinematic location of the surface cold front. The vorticity axis also illustrates that on this scale the frontal boundary can depart significantly from linearity, consistent with past studies (e.g., Carbone 1982; Hobbs and Persson 1982; Locatelli et al. 1995). The black dots drawn along the axis of vertical vorticity denote the locations of individual point maxima along the front. Although there are several exceptions in Fig. 4, these points are highly correlated with the location of the precipitation cores. The results support the conclusions proposed by Hobbs and Persson (1982), based on single- Doppler measurements, suggesting that the strongest horizontal shears develop within the core regions while weaker shear values are found within the gap regions. The results shown in Fig. 4, however, suggest that vertical vorticity maxima can occasionally develop within gap or weak reflectivity regions. In general, these gap regions did not appear to feature a prominent displacement in the front, which might support a local maximum in vorticity. The segment of the front labeled B is one notable exception. In addition, these maxima within the gaps did not appear to be historically linked to precipitation cores that subsequently dissipated. An example is shown by the second black dot from the bottom during the and passes shown in Fig. 4. The generation of strong vorticity in a gap is important to understand since Carbone (1983) has presented evidence of intense rotation leading to tornadogenesis in a weak-echo region of an NCFR. There is a clockwise rotation ( 20 ) between the average line defining the cold front and the long axes of the precipitation cores. These characteristics have been noted by other investigators (e.g., Hobbs and Persson 1982). The motion of the individual precipitation cores ( 16ms 1 from 235 ) is shown by the short dashed lines in Fig. 4 and illustrates the difference between the average motion of the cores and the cold front (7.8 m s 1 from 290 in the front-normal direction). The frontal movement was determined by the successive positions of the maximum in vorticity derived from all of the Doppler wind syntheses (a total of six passes by the Electra). It should be emphasized that the frontal speed is an along-line average since the movement shown in Fig. 4 depicts specific locations along the front where the speeds are sometimes greater or less than the overall motion of the front. An example of a segment of the front that experiences an acceleration is shown by the label A on the figure. 4. Analysis of three precipitation cores and gap regions The flight leg by the Electra from 1600:00 to 1619: 34 was chosen to illustrate the high-resolution analyses of the cold front and accompanying precipitation cores (Fig. 5). This was one of several passes when the P-3 was executing a parallel track 30 km to the west and within 10 min of the Electra. The combination of the wind syntheses from these two airborne platforms resulted in a dataset that encompassed a 180 km by 90 km area in the parallel and perpendicular directions to the front, respectively. The location of the cold front in

5 JULY 2000 WAKIMOTO AND BOSART 2451 FIG. 5. Horizontal plot at 400 m MSL of radar reflectivity and ground-relative winds at 1600: :34 UTC on 12 Jan The location of the cold front was determined by the ridge of vertical vorticity. The gray shade represents radar reflectivities greater than 30 dbz. The solid and dashed lines represent the flight tracks of the Electra and P-3, respectively. The boxed-in area is enlarged in Fig. 6. compared with the precipitation cores, which has been suggested in past studies but not clearly shown. It has been common to invoke the presence of horizontal shearing instability of the along-front winds to explain the formation of the precipitation core gap regions (e.g., Carbone 1982; Parsons and Hobbs 1983; Moore 1985; Braun et al. 1997). The raw single-doppler velocities from the fore antenna at 400 m were averaged over the entire pass shown in Fig. 6 to assess this instability (Fig. 9). The velocities in Fig. 9 represent components of the actual wind velocities and are used to provide a better estimate of the dimension of the wind shift rather than using the filtered wind syntheses. The data were adjusted such that the maximum in vertical vorticity was centered on the origin of the abscissa. The shear zone was determined to be 2.35 km based on the mean gradient of radial velocities in the figure. This length is similar to other case studies of cold fronts (e.g., Parsons and Hobbs 1983; Bond and Fleagle 1985). Linear theory indicates that the fastest growing mode will be at a wavelength 7.5 times the transition zone width although larger wavelengths can be unstable but are associated with smaller growth rates (Haurwitz 1949; Miles and Howard 1964). As a result, theory predicts that the most unstable wavelength for the present case is 17.6 km. This finding is close to the observed wavelength shown by the single-doppler velocities in Fig. 7 providing support that shearing instability may have been the dominant mechanism that led to the core gap structure along the cold front. Fig. 5, separating strong south-southwesterly flow in the warm air from weaker southwesterly flow in the cold air, was determined by the axis of maximum vertical vorticity. An enlargement of the three northern precipitation cores is presented in Figs. 6 and 7. The elliptically shaped precipitation cores with maximum reflectivity values located in the northern section of the echo can be seen in Figs. 6 and 7a. a. Horizontal structure 1) VERTICAL VORTICITY The structure of the ribbon of positive vertical vorticity is shown in Fig. 6a. The maxima in vorticity are s 1 and are situated near the precipitation cores, as previously mentioned in Fig. 4. The vorticity pattern also exhibits a wavelike structure 20 km in length. This sinusoidal pattern is also evident in the single-doppler velocity plot shown in Fig. 7b. An enlargement that focuses in on the details of the vertical vorticity pattern between the northernmost and middle cores is presented in Fig. 8a. Closer inspection of the vorticity pattern at this resolution reveals that the maximum occurs at the leading edge of the precipitation core rather than within the main echo. The horizontal gradient of vorticity is weaker within the gap region 2) VERTICAL VELOCITY The updraft pattern associated with the three northern cores is shown in Fig. 6b. The high quality of the Doppler radar synthesis in the present case allows for an estimate of the component of frontal updraft attributed to frictional forces. Fleagle and Nuss (1985) have outlined a methodology for calculating the vertical velocity at the top of the boundary layer using the following equation: 1 B wb 0 k f f B [ ] B (V V) k, (1) t B f where w B vertical velocity at the top of the boundary layer, B density at the top of the boundary layer, 0 surface stress, f Coriolis parameter, f/ y, and B depth of the boundary layer. The averages shown in Eq. (1) are over the depth of the boundary layer. It is the first term, in many synoptic situations, that dominates (e.g., Fleagle and Nuss 1985) and it is often referred to as Ekman pumping. The surface stress can be estimated by 0 C d (V V s )(V V s ), (2) B

6 2452 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 6. Enlargement of the three northern precipitation cores shown in Fig. 5. Radar reflectivity and ground-relative velocities superimposed on (a) vertical vorticity, (b) vertical velocity, (c) isogons, (d) magnitude of the horizontal velocity, where C d drag coefficient ( over the ocean), V mean horizontal velocity, and V s velocity of the ocean surface (assumed to be negligible). The winds at 400 m shown in Fig. 6 were used to estimate 0. The vertical velocity at the top of the boundary layer owing to Ekman pumping was calculated using Eq. (1) and found to range between 4 and 5ms 1 along the cold front (not shown). In the present case, however, the large values of vertical vorticity leads to significant contributions from the terms within the bracket in Eq. (1). Estimates of the vertical velocity near the cold front (not shown) from these two terms were comparable to those derived purely from Ekman pumping. Therefore, both frictional convergence and the mean acceleration

7 JULY 2000 WAKIMOTO AND BOSART 2453 FIG. 6.(Continued) and (e) retrieved perturbation pressure. Radar reflectivity is drawn as gray lines with values greater than 30 dbz shaded gray. The flight track of the Electra is shown. The boxed-in areas are enlarged in Figs. 8 and 10. of vertical vorticity appear to contribute to the observed updrafts in the figure. The observed updraft structure shown in Fig. 6b provides a clearer picture of the vertical velocity precipitation core relationship than has been previously documented. The three updrafts are elliptically shaped and are located at the leading edge and the southern section of each core. Accordingly, these updrafts are located to the south of the maximum reflectivity and do not extend over the entire length of the precipitation core. Locatelli et al. (1995) also proposed that the precipitation cores were not characterized by updraft along their entire length. These observations suggest that hydrometeors first form in the southern section of the cores and are subsequently advected to the northeast before descending to the lowest levels. The elliptical shape of the precipitation cores has been frequently attributed to similarly shaped updraft regions. These updraft regions have been proposed to form owing to the development of the aforementioned horizontal shearing instabilities along the cold front creating perturbations of the convergence zone along the length of a narrow convective line. These perturbations can amplify becoming line elements that often orient themselves at an angle to the original line (Matejka 1980; Carbone 1982; Hobbs and Persson 1982; Parsons and Hobbs 1983; Moore 1985). Browning and Roberts (1996), however, presented a case of precipitation cores along a cold front that did not appear consistent with a pure shearing instability mechanism. They argue that the orientation of the cores appeared to be largely due to the changing orientation of the surface cold front. Locatelli et al. (1995) and Brown et al. (1998, 1999) have recently proposed a new hypothesis to explain the shape of these cores. The former study used streamline analyses of Doppler radar data and the latter used numerical simulations to conclude that the advection of precipitation by the core-relative winds is most likely responsible for the distribution of precipitation into elliptically shaped cores oriented at an angle to the cold front. Trajectories of air parcels that originate within the updraft of the southern core shown in Figs. 6 and 7a were calculated using the full three-dimensional wind field in order to approximate the path of the hydrometeors carried by the core-relative winds (Fig. 10). Previous observational studies have been unable to perform a similar analysis owing to the coarse temporal and/or spatial resolution. The parcels were released at the lowest grid level and along a line that corresponds to the axis of maximum updrafts. The time tendency term was included by incorporating the wind synthesis for the subsequent pass by the Electra. The long time delay between successive passes by the precipitation core ( 15 min) did not appear to bias the results since the trajectories shown in Fig. 10 were nearly identical to the steady-state solutions obtained from the two individual syntheses (not shown). The trajectories all indicate a northeastward path suggesting that the elliptical shape of the precipitation core is strongly influenced by the advection and fallout of hydrometeors by the corerelative winds. The results shown in Figs. 6 and 10 suggest that the previous schematic model of the formation of the elliptical shape of the precipitation cores should be modified. It is still likely that barotropic shearing instability produces amplifying disturbances along the cold front. These instabilities result in updraft regions that form at distinct intervals along the cold front. The final shape and orientation of the precipitation cores relative to the cold front, however, are largely determined by the corerelative winds. This conclusion is consistent with the results of Locatelli et al. (1995), although their wind syntheses did not have sufficient resolution to calculate trajectories. Their study also proposed that the updraft regions within the cores are circular. The syntheses shown in Figs. 6 and 10 reveal that the updrafts are primarily elliptical. 3) WIND DIRECTION AND WIND SPEED The characteristics of the wind direction and wind speed discontinuity across the cold front are shown in Figs. 6c and 6d, respectively. The former exhibits a wavelike pattern that largely mimics the vertical vorticity analysis. This is best illustrated by the enlargement shown in Fig. 8. The wavelengths for the vertical vorticity and wind direction shift are approximately in

8 2454 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 7. Enlargement of the three northern cores encountered during the 1600: :34 UTC pass by the Electra at 400 m MSL: (a) Radar reflectivity and (b) single-doppler velocities from the fore antenna. The solid and dashed black lines represent the viewing angles of the antenna and the flight track, respectively. The color scales for each of the fields are shown. phase and the amplitudes are comparable. The character of the shift in wind direction is different than the model proposed by Hobbs and Persson (1982). Their schematic model places the wind shift zone primarily along the entire length and at the leading edge of the precipitation core (see Fig. 11a). The results presented in Fig. 6 reveal that this zone may cross through the strongest echo region within a core. Accordingly, rainfall may precede or occur behind the surface wind shift zone depending on whether an observer s position is in the northern or southern sector of the precipitation core, respectively. The modification of the precipitation core wind shift discontinuity relationship as well as the preferred location of the updraft zone are summarized in Fig. 11b. The modification of the NCFR model shown in Fig. 11b again highlights that a pure shearing instability mechanism does not fully account for the precipitation core gap structure as Fig. 11a suggests. The new schematic also illustrates the differences in strength of the surface discontinuity experienced along the precipitation core. The northern sector of the core is accompanied by more slowly changing variables, such as the wind shift, while the southern section is characterized by a more abrupt discontinuity. Hobbs and Persson (1982) suggest that an abrupt discontinuity should be encountered along the entire length of the core. Moreover, the position of the maximum radar reflectivity within the precipitation cores in Fig. 6 indicates that the heavier and lighter precipitation will tend to fall in the region where the surface discontinuity is the weakest and strongest, respectively, against initial expectations. The gap region has been defined as an area devoid of precipitation in previous studies. The current study suggests that a gap should be defined as an area devoid of strong updraft and accompanied by weak surface discontinuities. The latter definition will, hereafter, be applied in this paper. The degree to which the schematic diagrams in Fig. 11 will differ will depend on the strength of the along-front component of the wind between the two cases and, hence, the degree to which hydrometeors are carried along the front downstream of the primary updraft. The isotach analysis in Fig. 6d also reveals a wave

9 JULY 2000 WAKIMOTO AND BOSART 2455 FIG. 8. Enlargement of the northern boxed-in area shown in Fig. 6. (a) Radar reflectivity (gray lines) and vertical vorticity (black lines) superimposed on ground-relative wind vectors. Radar reflectivities values greater than 30 dbz are shaded gray. (b) Magnitude of the horizontal velocity (gray lines) and isogons (black lines). structure along the cold front but its amplitude is less than the isogon field. Figure 8b highlights this amplitude difference by the superimposition of these two fields. These figures suggest that the time between the arrivals of the wind speed and wind shift discontinuities as well as the total time elapsed during which each perturbation affects a surface observing station will depend on the location of the site along the front. In order to quantify FIG. 9. Mean single-doppler velocities at 400 m MSL of the frontal zone for the syntheses area shown in Fig. 6. The shear profile shown represents an average over a horizontal distance of 43 km. The origin along the abscissa represents the location of the maximum in vertical vorticity. the intensity of the discontinuities in the core and gap regions, plots of the magnitude of the horizontal gradient of wind speed and direction were plotted (not shown). The former was associated with 8 m s 1 km 1 and 12 km 1 while the latter was 4 m s 1 km 1 and 4 km 1. 4) PERTURBATION PRESSURE Gal-Chen (1978) first proposed the use of wind syntheses based on multi-doppler analyses to retrieve the perturbation pressure using a least squares method. The retrieval treats the pressure and temperature as unknown variables in the anelastic momentum equations and solves the derived Poisson equation by inputting the Doppler wind fields. The reader is referred to Gal-Chen (1978) for details of the scheme. The retrieval of perturbation pressure for the northern three precipitation cores is presented in Fig. 6e. A measure of the quality of the retrieved fields can be performed by using a momentum check. These values, discussed in the appendix, suggest that the perturbation pressure shown in the figure are reliable. Brown et al. (1999), using a numerical model, hypothesize that trapped gravity waves can develop and produce significant sea level pressure patterns just behind the cold front. Their model results depict alternating couplets of high and low pressure in the core and

10 2456 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 11. (a) Schematic diagram of the surface wind shift and updraft zones across two precipitation cores and the gap region between them, based on a figure by Hobbs and Persson (1982). (b) Same schematic diagram but modified based on the analyses presented in this paper. The wind shift zone is indicated by the shaded region. The updraft regions are shown by the dashed lines. FIG. 10. Trajectories of air parcels (thick black line) released in the updraft for the southern precipitation core shown in Figs. 6 and 7a. Black dots represent 5-min intervals. The analysis area was shown by the southern boxed-in area shown in Fig. 6. gap regions, respectively. This pressure pattern was hydrostatically produced and proposed to modify the wind and thermal structure (instead of horizontal shearing instability) and aid in the organization of the precipitation core and gap regions along the cold front (see their Fig. 13). The pressure pattern suggested by Brown et al. (1999) is also apparent in Fig. 6e although the magnitudes are reduced. Examination of the fluid shear and extension terms in the diagnostic pressure equation (not shown) reveals that the relatively higher pressure associated with the southern section of the precipitation cores is confined to the regions of horizontal convergence at low levels (i.e., it is a result of fluid extension). Hence, the pressure perturbations in the present case are dynamically produced. No evidence of the presence of trapped gravity waves could be found. b. Vertical structure 1) CORE REGION The high-resolution wind syntheses for the cold front on 12 January provided an opportunity to construct separate mean vertical cross sections through the core and gap regions. A cross section perpendicular to the cold front and averaged along the long axes of the updrafts associated with the three precipitation cores shown in Fig. 6 is presented in Fig. 12. This mean profile was created by averaging a total of 62 individual cross sections through the precipitation cores. The maximum vertical vorticity ( s 1 ) is situated within the precipitation core in Fig. 12a. To the west (left) of the precipitation core is a weak-echo trench reminiscent of the transition region documented with mesoscale convective systems (e.g., Biggerstaff and Houze 1993). Braun et al. (1997) also noted similarities between the vertical structure of NCFRs and midlatitude squall lines. Farther to the west is a broad band of light-to-moderate rain associated with mainly stratiform clouds extending behind the cold front (e.g., Browning 1990). The frontal boundary does not suggest a hydraulic head structure characteristic of density currents in this mean cross section. Numerous individual cross sections (not shown), however, did reveal a hydraulic head feature. The kinematic location of the cold front was determined by the magnitude of the total vorticity vector shown in Fig. 12b. It has been a common practice, based on dual-doppler wind syntheses, to locate fronts at the position where the component of relative flow perpendicular to the front is zero (e.g., Carbone 1982; Locatelli et al 1995). This method of positioning the front is dependent on an accurate determination of the frontal motion and orientation. No such restrictions are necessary if the vorticity field is used. Indeed, the isopleth of zero-relative flow shown in Fig. 12c is close to but displaced from the frontal boundary identified by the vorticity field. The velocity component parallel to the front shown in Fig. 12b is closely related to the strongest gradients in vorticity presented in Figs. 12a and 12b. The relative flow in the plane of the mean vertical cross section (Fig. 12c) reveals an updraft at the leading edge of the cold front, which is tilted slightly over the cold air. There is front-relative flow at low levels within the cold air toward the leading edge of the front. This type of flow has not always been present in past studies of cold fronts (e.g., see review by Smith and Reeder 1988). The environmental flow has a component of mo-

11 JULY 2000 WAKIMOTO AND BOSART 2457 FIG. 12. Mean vertical cross sections through the precipitation cores. (a) Radar reflectivity (gray lines) superimposed on vertical vorticity (black lines) and the ground-relative horizontal wind field. (b) The component of the horizontal velocity parallel to the front (u, plotted as gray lines) superimposed on the magnitude of the three-dimensional vorticity vector (black lines) and the ground-relative horizontal wind field. (c) The relative component of the horizontal velocity perpendicular to the front (, plotted as gray line) superimposed on the front-relative wind field in the plane of the cross section. The location of the cold front was determined by the maximum in the three-dimensional vorticity vector. Wind vectors are plotted with the half barb, full barb, and flag representing 2.5, 5.0, and 25.0 m s 1, respectively.

12 2458 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 13. A time series of in situ data collected at flight level during a low-level penetration ( 600 m) of one of precipitation cores from 1619:40 to 1621:40 UTC. The approximate location of the precipitation core based on liquid water measurements is shown. The black, dotted line is the derived surface pressure. The minimum in pressure is shown by the black arrow. The total distance traversed by the aircraft during this time was 13 km as indicated. tion toward the front at low levels ( 1 km) and away from the front between 1 and 2 km. The suggestion of a component of the transverse motion flowing ahead of the front at this elevated level is consistent with the schematic model proposed by Browning and Harrold (1970) and Roux et al. (1993). A time series of in situ data collected at flight level during a low-level penetration of the northernmost precipitation core (presented in Fig. 6) is presented in Fig. 13. The virtual and equivalent potential temperature both fall 1.3 K during the penetration from the warm to the cold air side. The pressure trough 2 is situated at the leading edge of the cold air and ahead of the precipitation core. The pressure rises as the air becomes progressively cooler to the west. The updraft peaks at 4ms 1 within the core and is consistent with the maximum updrafts derived from the wind syntheses shown in Figs. 6b and 12c. There is a slight decrease in wind speed at the same time as the initial falls in temperature 2 This pressure minimum could be a result of deviations in the flight-level altitude; however, it is characteristic of the pressure traces during two subsequent penetrations of the cold front and pressure. This decrease is more rapid through the precipitation core. The minimum speed is reached well before the coldest temperatures are attained. The wind direction shift occurs 1.5 km behind the pressure trough, well to the rear of the leading edge of the cold pool. The latter observation is different than the typical sequence of events associated with a gust front (e.g., Charba 1974; Goff 1976; Wakimoto 1982). Indeed, the shift in wind direction is nominally among the first discontinuities experienced at an observing site during the passage of a gust front. The relationship between the discontinuities in wind speed and direction will vary depending on the location of the aircraft penetration of the precipitation core. This dependence is a direct result of the higher-amplitude variation of the wind direction along the cold front supported by the analyses shown in Figs. 6c and 6d, and 8b. Several studies have shown that the vertical structure of these type of fronts exhibit a characteristic shape of a density current with a hydraulic head at its leading edge (e.g., Carbone 1982; Hobbs and Persson 1982; Parsons 1992; Locatelli et al. 1995). Although the kinematic structure appears to be similar to a density current, there has been considerable debate over whether the propagation speed of this type of cold front can be adequately predicted based on density current dynamics. Carbone (1982), Hobbs and Persson (1982), Parsons and Hobbs (1983), Bond and Fleagle (1985), Shapiro et al. (1985), Parsons et al. (1987), and Lamaître et al. (1989) have provided arguments that the dynamics between the two phenomena are similar while Smith and Reeder (1988), Dudhia (1993), Browning and Reynolds (1994), and Chen et al. (1997) suggest that these fronts cannot be considered density currents even though they may structurally appear the same. Indeed, it is not clear that the movement of a cold front should be solely governed by the density difference across the front. Although a front is defined as a density discontinuity, it is constrained by along-boundary geostrophic and thermal wind balance. These constraints are not present in laboratory experiments of density currents. The following equation was used to estimate the propagation speed of a density current: 0.5 c k g h, (3) where k internal Froude number, g acceleration of gravity, mean difference in virtual potential temperature at flight level, vc mean virtual potential temperature in the cold air, h height of the cold pool, and c propagation speed of the front. The calculation of c based on the in situ data collected during the penetration of a precipitation core at 1621 is shown in Table 2. The Froude number was assumed to vary between 0.7 and 1.1 (e.g., Koch 1984) and h was estimated to range between 1.5 and 2.7 km based on the mean vertical cross section (Fig. 12). The temper- vc

13 JULY 2000 WAKIMOTO AND BOSART 2459 TABLE 2. Density current calculations using virtual potential temperature. k (K) c (K) h (km) c (m s 1 ) 0.62u af (m s 1 ) c (m s 1 f ) 0.62u ac (m s 1 ) c (m s 1 c ) ature deficit at flight level was estimated to be 1.3 K. Three other penetrations of the cold front that were examined revealed values ranging from 1.2 to 1.4 K. The minimum and maximum values of c shown in the table are for the smallest and largest values of k and h, respectively. The propagation speed using Eq. (3) is relative to the ambient flow perpendicular to the cold front. Bluestein (1993) states that the environmental flow should be obtained by vertically averaging the wind speeds up to the depth of the cold pool. This component of the ambient flow for the present case was estimated by taking areal averages of the Doppler wind syntheses over a 10-kmsquare area ahead of the cold front shown in Fig. 6 at the 0.4-, 0.8-, and 1.2-km levels. Brown et al. (1999) suggest that the density current equation may only apply in local regions of the cold front, especially in the region near the precipitation cores. To assess this effect, two values of the mean components of the environmental flow were derived one in a direction perpendicular to the overall orientation of the front and the other perpendicular to the major axis of the precipitation cores. These components were approximately 6.4 m s 1 (u af ) and 3.5 m s 1 (u ac ), respectively, and represent a numerical average over nearly 1900 points. Sensitivity experiments using larger boxes did not change the results significantly. Averages in ambient flow calculated for the other aircraft passes shown in Fig. 4 yielded values within 1 ms 1 of the value calculated for the present case. Simpson and Britter (1980) have proposed that 62% of the ambient flow perpendicular to the front should be used to calculate the density current speed in noncalm conditions. This effect has rarely been accounted for in past studies on cold fronts. The adjusted estimates of the density current speeds for the overall front motion ( c f ) and the motion perpendicular to the major axis of the precipitation cores ( c c ) are shown in Table 2. The range of numerical values of c f shown in the table are larger than the observed frontal motion of 7.8 m s 1. The values of c c, however, do encompass the observed frontal speed in a direction perpendicular to the precipitation core axes ( 7.3 m s 1 ). Accordingly, these segments of the front appear to satisfy density current theory. In contrast, the overall motion of the front was not well described by this theory as suggested by the numerical simulations of Brown et al. (1999). An alternative method for calculating the density current speed has been advanced by Seitter and Muench (1985) using the following equation: 0.5 p c k s (4) where p s the change in surface pressure across the front and w the mean density in the warm air. This method is advantageous since the change in surface pressure is representative of the integrated depth of cold air and would not be subject to errors related to the height of the cold pool and calculating a mean temperature within the postfrontal air. The estimates of c c and c f for the same pass as pre- sented in Table 2 are shown in Table 3. These values are consistent with those presented in Table 2 although the maximum estimates of c c and c f are lower. The latter speed suggests that the overall motion of the front was not similar to a density current. The former speed provides weaker support that density current theory describes the local frontal motion perpendicular to the major axis of the precipitation cores. Two other penetrations of the cold front were examined and led to similar estimates of the change in surface pressure. It should be noted that the pressure change of 0.75 mb shown in Table 3 represents an average. The actual peak to trough range is closer to 1 mb in Fig. 13. The larger pressure difference would lead to an even greater discrepancy of c f with the observed frontal speeds but would improve the estimate of c c. 2) GAP REGION The mean cross sections through the gap regions shown in Figs. 6 and 7 are presented in Fig. 14. A total of 24 individual cross sections were averaged to create the profile shown in this figure. The maximum vertical vorticity at the leading edge of the cold pool is comparable to the core region shown in Fig. 12a; however, the horizontal gradient of vorticity is weaker, analogous to the analysis shown in Fig. 8a. This weaker gradient near the surface is also apparent in the plot of the total vorticity vector in Fig. 14b. The overall echo structure is similar to the mean cross section through the precipitation core. Recall in Fig. 11b that the kinematic position of the gap, that is, a region w TABLE 3. Density current calculations using surface pressure. k p s (mb) c (m s 1 ) 0.62u af (m s 1 ) c (m s 1 f ) 0.62u ac (m s 1 ) c (m s 1 c )

14 2460 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 14. Mean vertical cross sections through the gap regions between precipitation cores. (a) Radar reflectivity (gray lines) superimposed on vertical vorticity (black lines) and the groundrelative horizontal wind field. (b) The component of the horizontal velocity parallel to the front (u, plotted as gray lines) superimposed on the magnitude of the three-dimensional vorticity vector (black lines) and the ground-relative horizontal wind field. (c) The relative component of the horizontal velocity perpendicular to the front (, plotted as gray line) superimposed on the frontrelative wind field in the plane of the cross section. The location of the cold front was determined by the maximum in the three-dimensional vorticity vector. Wind vectors are plotted with the half barb, full barb, and flag representing 2.5, 5.0, and 25.0 m s 1, respectively.

15 JULY 2000 WAKIMOTO AND BOSART 2461 FIG. 15. A time series of in situ data collected at flight level during a low-level penetration ( 600 m) of a gap region from 1741:30 to 1743:10 UTC. The minimum in pressure is shown by the black arrow. The total distance traversed by the aircraft during this time was 13 km as indicated. of broader wind shifts and devoid of strong updrafts, can occur within the northern sector of the precipitation core. This is an important difference than the previous model shown in Fig. 11a. Accordingly, the mean vertical cross section in the gap region can be accompanied by significant radar reflectivity as shown in Fig. 14 even though pronounced updrafts are absent (see Fig. 14c). The presence of echo in Fig. 14a is a result of hydrometeors being advected into the gap region by the corerelative winds. In situ data collected at flight level during a low-level penetration of a gap region is presented in Fig. 15. It should be noted that the plot shown in Fig. 13 was recorded during a penetration of a core shown in Fig. 6. The time series presented in Fig. 15 was based on a penetration of a gap that occurred later in the flight and is the only example of the thermodynamic characteristics of this region during IOP 2. The timescale has been expanded in Fig. 15 so that the total horizontal distance traveled by the aircraft was 13 km, equivalent to the distance shown in Fig. 13. The reduced gradients of virtual and equivalent potential temperature during the penetration of a gap region compared to a core region are evident when comparing these two time series. The pressure trough is located at the leading edge of FIG. 16. The 1650: :40 UTC pass by the cold front where RHI cross sections at select locations were examined. Radar reflectivities (gray lines) and vertical vorticity (thin black lines) superimposed on ground-relative wind vectors. Reflectivities greater than 30 dbz are shaded gray. The thick black line segments pointed to the northwest indicate the positions of individual RHI cross sections shown in Fig. 17. The black line oriented approximately south-southwest to north-northeast denotes the Electra flight track. the cold air and rises a few tenths of a millibar after the aircraft penetrated the cold air. Subsequently, the pressure remains relatively constant even though the temperatures continue to drop before a positive trend is reached well behind the leading edge of the cold pool. The wind speed and direction monotonically decrease and increase, respectively, behind the front. The positive vertical velocities during the frontal passage are 2.5 ms 1, greater than what is shown in Fig. 14c. Recall that the in situ data were based on a penetration of a gap region that occurred later in the flight and may not be representative of the frontal structure shown in Fig. 14. The abrupt versus slowly changing variables for the core and gap regions, respectively, are consistent with the surface weather observations noted by James and Browning (1979) and Hobbs and Persson (1982). The sequence of events, however, is resolved in greater detail owing to the fast response by the aircraft probes. 3) RANGE HEIGHT INDICATOR CROSS SECTIONS Another view of the vertical structure of the updraft along the cold front was provided by the single-doppler range height indicator (RHI) cross sections (Fig. 16) during the third flyby past the cold front. This particular transect resulted in the fore antenna collecting radial velocities that were representative of the main flow fea-

16 2462 MONTHLY WEATHER REVIEW VOLUME 128 FIG. 17. RHI cross sections from the fore antenna at (a) 1710:39, (b) 1709:41, (c) 1709:33, (d) 1709:21, (e) 1708:54, and (f ) 1707:44 UTC. (top) Radar reflectivities and (bottom) single-doppler velocities, respectively. Positions of the cross sections are shown in Fig. 16. Black arrows are described in the text. Black dots in (e) denote the position of possible Kelvin Helmholtz waves. Color scale is shown at the bottom of the figure. Range rings are drawn every 5 km. tures resolved by the dual-doppler wind field. The other passes hinted at this structure but the viewing angle was not optimum for the single-doppler velocities. The RHI cross section in Fig. 17a reveals a structure that is reminiscent of a density current (e.g., Carbone 1982; Shapiro et al. 1985; Browning et al. 1997). The light-green colors near the surface represent the cold pool of air flowing toward the radar while the darkbrown colors (highlighted by the black arrow) indicate the ambient flow being undercut by the advancing front.

17 JULY 2000 WAKIMOTO AND BOSART 2463 FIG. 18. Analysis of the region near the large gap at 1650: :40 UTC at 800 m MSL. (a) Component of flow perpendicular to the front drawn as gray lines with values greater than 10 m s 1 shaded gray. (b) Isogon analysis drawn as gray lines with values greater than 240 shaded gray. The dashed black line represents the location of the axis of maximum vorticity. The echoes greater than 30 dbz are drawn as black lines. The Electra flight track is also shown. The letter B indicates the location of the large gap and is consistent with the notation used in Fig. 4. There is even a hint of a hydraulic head at the leading edge (e.g., Carbone 1982; Shapiro et al. 1985; Locatelli et al. 1995). This cross section was taken through a precipitation core (see Fig. 16). Figure 17b (1709:41) is located at a position where two cores overlap as shown in Fig. 16. The left arrow in the Doppler velocity image denotes the location of the ambient flow that is being undercut by the cold front and is a continuation of the updraft noted in the previous panel. The right arrow marks the position of an incipient updraft perturbation (note the upward bulging area of red) associated with the southern precipitation core. The next RHI cross section located farther to the south (Fig. 17c) reveals two areas of positive vertical motion associated with each precipitation core as marked by the two black arrows. The western updraft was weaker while the eastern was stronger. The weakening of the western updraft continues at 1709:21 (Fig. 17d) although it is still apparent. The updraft associated with the northern precipitation core has vanished at 1708:54 (Fig. 17e) and only the forced uplift accompanying the southern core is apparent. The hydraulic head feature is evident at this time. Also visible in this vertical cross section are well-defined velocity features marked by the black dots, which are likely the result of Kelvin Helmholtz (K H) waves with horizontal dimensions varying between 2.5 and 3 km. Other investigators have noted that the shear zone atop the cold pool and along the frontal boundary can support K H waves (e.g., Young and Johnson 1984; Chapman and Browning 1998). Unfortunately, it is not possible to calculate a Richardson number for the present case owing to a lack of thermodynamic data in the vertical. The RHI cross section at 1707:44 (Fig. 17f) is representative of the cold front near the gap region. Note that the hydraulic head feature and uplift at the surface location of the cold front are less apparent when compared with Figs. 17a and 17e. This structure supports the analyses presented by Locatelli et al. (1995). They show a strong correlation between the height of the hydraulic head/uplift and the intensity of precipitation falling behind the cold front into the cold air. The cold front shown in Fig. 17 provides additional evidence of this dependence. The discussion presented in this section suggests that the cold front resembles a density

18 2464 MONTHLY WEATHER REVIEW VOLUME 128 current in regions near the precipitation cores. This provides kinematic support that complements the density current calculations presented earlier. 5. Large gap The evolution of the echo structure shown in Fig. 4 is highlighted by a large gap (denoted by the letter B), a term first coined by Locatelli et al. (1995). While not as numerous as the smaller gap regions, this area devoid of precipitation leads to a front that appears segmented with the southern section km ahead of the northern section. Fewer than 20% of all gaps are typically of this length or larger (James and Browning 1979). The time sequence of echo positions shown in Fig. 4 reveals that the formation of this large gap is primarily a result of differential movement of the cold front. The section of the front located south of B was moving 3 ms 1 faster than the northern section between and An analysis of the third flight leg of the Electra is shown in Fig. 18. The location of the large gap is denoted by the letter B. The faster movement of the southern section of the cold front is supported by the stronger flow within the cold air that is perpendicular to the front (Fig. 18a). These stronger postfrontal winds are also associated with a more westerly wind direction (Fig. 18b). While small gaps appear to form in response to horizontal shearing instability, the results presented in Fig. 18 suggest a different mechanism for the development of a large gap. A larger-scale perspective of this area is presented in Fig. 19. The radar reflectivity is plotted out to a range of 80 km and at a height of 1.2 km in the figure. A significant area of enhanced stratiform precipitation is noted near and south of the large gap within the postfrontal air. Another view of the echo pattern is shown by the lower fuselage image recorded by the P-3 at an earlier time (Fig. 20). The NCFR is quasi-linear at the time shown in the figure with the section of the front south of B nearly aligned with the northern section. Clearly apparent in the image is the presence of a wide cold-frontal rainband (e.g., Parsons and Hobbs 1983) with its northern extent terminating at B. The analysis and images shown in Figs. 19 and 20 suggest that the acceleration of the southern part of the cold front was in response to the enhanced precipitation associated with a postfrontal rainband. The subsequent latent cooling of hydrometeors enhanced the rear-tofront flow (as suggested by Oliver and Holzworth 1953) leading to the differential movement of the northern and southern segments of the cold front and the formation of the large gap at B. In situ data collected at flight level (not shown) at 1700 revealed postfrontal air in the southern sector that was 1 K colder in than the air behind the northern part of the front. The preceding analyses provide evidence that density current dynamics can modify the movement of the cold front. The results FIG. 19. Horizontal plot of radar reflectivity at 1.2 km MSL for 1650: :40 UTC. Reflectivity values greater than 20 and 30 dbz are shaded gray and black, respectively. The short, dashed line denotes the location of the axis of maximum vorticity. The long, dashed line represents the farthest range of the radar data. The flight track of the Electra is shown. The letter B denotes the location of the large gap. in section 4, however, imply that this theory alone does not fully account for the overall propagation of this front. It is also possible that the larger differential speed of the southern section of the front was related to the descent of higher momentum air. This did not appear to be the case since the flow at upper levels was relatively uniform over the entire domain. A closer inspection of the large gap produced at point B is presented in Fig. 21. The vertical vorticity analysis shown in Fig. 21a reveals a broader area of vorticity s 1 in the region of the large gap suggesting that the frontal zone is also broad in this region. This was confirmed by examining the flight-level data (not shown) during the penetration of this gap. The wind shifts and thermal discontinuity were more diffuse compared to other locations along the front. The vorticity pattern also shows that the front is still continuous (kinematically) even though the gap region is nearly 15 km in length. The large gap region is associated with relatively weak vertical motions as shown in Fig. 21b. There is a tendency for the updrafts associated with the core regions to be primarily located on the southern part of the

19 JULY 2000 WAKIMOTO AND BOSART 2465 FIG. 20. (left) Surveillance scan recorded by the lower fuselage radar on board the NOAA P-3 at 1546:47 UTC 12 Jan. Precipitation echoes along the narrow cold frontal rainband and a wide cold frontal rainband are noted east of the aircraft. Radar reflectivity values are shown on the color scale. Range rings are every 50 km. (right) Schematic diagram illustrating the main echo features. The letter B denotes the location where the large gap, shown in Fig. 4, forms. echoes, consistent with the analyses shown in Fig. 6. The isopleths of the ground-relative wind speed in Fig. 21c reveal a discontinuity across the cold front. The gradient of this speed change is greatest at the location of the precipitation cores and weaker at the location of the large gap in agreement with observations by James and Browning (1979) and Hobbs and Persson (1982). The enhanced speeds in the postfrontal air south of the large gap, noted earlier, are apparent. There is also a suggestion in this plot that the maximum wind speed in the ambient air is slightly stronger ahead of the precipitation cores compared to the gap region (note that the speeds in the ambient air are 28 m s 1 ahead of the large gap region). Numerical simulations by Dudhia (1993) and dual-doppler observations by Braun et al. (1997) also present some evidence of stronger flow ahead of the cores. Crook (1987) and Dudhia (1993) hypothesize that the low-level jet ahead of surface cold fronts is a result of a Coriolis turning of the flow that is feeding into the updraft of a convective system. This could explain the slight enhancement of wind speeds shown in Fig. 21c east of the location of the precipitation cores. The isogon analysis presented in Fig. 21d indicates that the total shift in the wind direction through the gap region is greater than through the precipitation cores. The gradient of wind direction, however, does not appear to be significantly different in these two areas. 6. Summary and discussion There have been numerous studies of NCFRs in recent years. Much of this work has focused on the processes that lead to the formation of the precipitation core and gap regions along the cold front. Horizontal shearing instability or trapped gravity waves have been proposed as the primary generating mechanisms. The results presented in this study using detailed wind fields synthesized from an airborne Doppler radar, however, suggest that it is a combination of horizontal shearing instability and advection of hydrometeors by the corerelative winds that lead to the development of this structure. The former theory closely predicts the wavelength of the ribbon of maximum vertical vorticity observed at the leading edge of the cold front. Individual point maxima of vertical vorticity in this ribbon were correlated with the locations of the precipitation cores. This instability sets up preferred regions of enhanced convergence and updraft along the front. The precipitation cores, however, are not confined to the regions of positive vertical velocities. Detailed analysis of air parcel trajectories suggests that advection of hydrometeors by the core-relative winds determines the final shape and orientation of the cores. Trapped gravity waves did not appear to be present as has been suggested in past studies. The along-core structure documented in this study