• Alexander, G. D., and G. S. Young, 1992: The relationship between EMEX mesoscale precipitation feature properties and their environmental characteristics. Mon. Wea. Rev.,120, 554–564.

  • Barnes, G. M., and K. Sieckman, 1984: The environment of fast- and slow-moving tropical mesoscale convective cloud lines. Mon. Wea. Rev.,112, 1782–1794.

  • Caniaux, G., J. L. Redelsperger, and J. P. Lafore, 1994: A numerical study on the stratiform region of a fast moving squall line. Part I: General description and water and heat budget. J. Atmos. Sci.,51, 2046–2073.

  • ——, ——, and ——, 1995: A numerical study on the stratiform region of a fast moving squall line. Part II: Relationship between mass, pressure, and momentum fields. J. Atmos. Sci.,52, 331–352.

  • Chong, M., and O. Bousquet, 1999: Anticyclonic mesovortex in a near-equatorial mesoscale convective system during TOGA COARE. Mon. Wea. Rev.,127, 1145–1156.

  • ——, P. Amayenc, G. Scialom, and J. Testud, 1987: A tropical squall line observed during the COPT81 experiment in West Africa. Part I: Kinematic structure inferred from dual-Doppler data. Mon. Wea. Rev.,115, 670–694.

  • Cuxart J., P. Bougeault, and J. L. Redelsperger, 2000: A turbulent scheme allowing for mesoscale and large-eddy simulations. Quart. J. Roy. Meteor. Soc.,126, 1–30.

  • Gao, K., D.-L. Zhang, M. W. Moncrieff, and H.-R. Cho, 1990: Mesoscale momentum budget in a midlatitude squall line: A numerical study. Mon. Wea. Rev.,118, 1011–1028.

  • Hildebrand, P. H., 1998: Shear parallel moist convection over tropical ocean: A case study from February 1993 TOGA COARE. Mon. Wea. Rev.,126, 1952–1976.

  • Hitschfeld, W., and J. Bordan, 1954: Errors inherent in the radar measurement of rainfall at attenuating wavelengths. J. Meteor.,11, 58–67.

  • Houze, R. A., Jr., 1977: Structure and dynamics of tropical squall line system. Mon. Wea. Rev.,105, 1540–1567.

  • Jorgensen, D. P., M. A. LeMone, and S. B. Trier, 1997: Structure and evolution of the 22 February 1993 TOGA COARE squall line: Aircraft observations of precipitation, circulation, and surface energy fluxes. J. Atmos. Sci.,54, 1961–1985.

  • Kessler, E., 1969: On the Distribution and Continuity of Water Substance in Atmospheric Circulations. Meteor. Monogr., No. 32, Amer. Meteor. Soc., 84 pp.

  • Kingsmill, D. E., and R. A. Houze, 1999: Kinematic characteristics of air flowing into and out of precipitating convection over the west Pacific warm pool: An airborne Doppler radar survey. Quart. J. Roy. Meteor. Soc.,125, 1165–1207.

  • Klemp, J. B., R. B. Wilhlemson, and P. S. Ray, 1981: Observed and numerically simulated structure of a mature supercell thunderstorm. J. Atmos. Sci.,38, 1558–1580.

  • Lafore, J. P., and M. W. Moncrieff, 1989: A numerical investigation of the organization and interaction of the convective and stratiform regions of a tropical squall line. J. Atmos. Sci.,46, 521–544.

  • ——, J. L. Redelsperger, and G. Jaubert, 1988: Comparison between a three-dimensional simulation and Doppler radar data of a tropical squall line: Transport of mass, momentum, heat and moisture. J. Atmos. Sci.,45, 3483–3500.

  • ——, and Coauthors, 1998: The Meso-NH atmospheric simulation system. Part I: Adiabatic formulation and control simulations. Ann. Geophys.,16, 90–109.

  • LeMone, M. A., E. J. Zipser, and S. B. Trier, 1998: The role of environmental shear and CAPE in determining the structure and evolution of mesoscale convective systems during TOGA-COARE. J. Atmos. Sci.,55, 3493–3564.

  • Lewis, S. A., M. A. LeMone, and D. P. Jorgensen, 1998: Evolution and dynamics of a late-stage squall line that occurred on 20 February 1993 during TOGA COARE. Mon. Wea. Rev.,126, 3189–3212.

  • Lipps, F. B., and R. S. Helmer, 1982: A scale analysis of deep moist convection and some related numerical calculations. J. Atmos. Sci.,39, 2192–2210.

  • Moncrieff, M. W., 1992: Organized convective systems: Archetypal dynamical models, mass and momentum flux theory, and parametrization. Quart. J. Roy. Meteor. Soc.,118, 819–850.

  • Montmerle, T., 1998: Validation et initialisation d’un modèle tri-dimentionnel méso-échelle non-hydrostatique par des données expérimentales issues de TOGA-COARE. Ph.D. thesis, Université Pierre and Marie Curie (Paris VI), 246 pp. [Available from Centre d’Étude des Environnements Terrestre et Planétaires, Velizy, France.].

  • ——, and Y. Lemaître, 1998: Three-dimensional variational data analysis to retrieve thermodynamical fields from various nested wind measurements. J. Atmos. Oceanic Technol.,15, 640–665.

  • Oury, S., J. Testud, and V. Marécal, 1999: Estimate of precipitation from the dual-beam airborne radar in TOGA COARE. Part I: The K–Z relationships derived from stereo- and quad-beam analysis. J. Appl. Meteor.,38, 156–174.

  • Pinty, J. P., and P. Jabouille, 1998: A mixed-phased cloud parameterization for use in a mesoscale non-hydrostatic model: Simulation of a squall line and of orographic precipitations. Preprints, Conf. on Cloud Physics, Everett, WA, Amer. Meteor. Soc., 217–220.

  • Protat, A., Y. Lemaître, and G. Scialom, 1998: Thermodynamic analytical fields from Doppler radar data by means of the MANDOP analysis. Quart. J. Roy. Meteor. Soc.,124, 1633–1669.

  • Redelsperger, J.-L., and G. Sommeria, 1986: Three-dimensional simulation of a convective storm: Sensitivity studies on subgrid parameterization and spatial resolution. J. Atmos. Sci.,43, 2619–2635.

  • ——, and Coauthors, 1999: Review on convection in TOGA-COARE. Proceedings of the Conference on TOGA-COARE, WMO/TD-No. 940, 16–42.

  • ——, and Coauthors, 2000: A GCSS model intercomparison for a tropical squall line observed during TOGA-COARE. Part I: Cloud resolving models. Quart. J. Roy. Meteor. Soc.,126, 823–864.

  • Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci.,45, 463–485.

  • Roux, F., 1998: The oceanic mesoscale convective system observed with airborne Doppler radars on 9 February 1993 during TOGA-COARE: Structure, evolution and budgets. Quart. J. Roy. Meteor. Soc.,124, 585–614.

  • Scialom, G., and Y. Lemaître, 1990: A new analysis for the retrieval of three-dimensional mesoscale wind fields from multiple Doppler radar. J. Atmos. Oceanic Technol.,7, 640–665.

  • Testud, J., and S. Oury, 1997: An algorithm to correct weather radar data for attenuation. C. R. Acad. Sci. Paris,324 (II.a), 705–710.

  • Thorpe, A. J., M. J. Miller, and M. W. Moncrieff, 1982: Two-dimensional convection in nonconstant shear: A model of mid-latitude squall lines. Quart. J. Roy. Meteor. Soc.,108, 739–762.

  • Trier, S. B., W. C. Skamarock, M. A. LeMone, D. B. Parsons, and D. P. Jorgensen, 1996: Structure and evolution of the 22 February 1993 TOGA-COARE squall line: Numerical simulation. J. Atmos. Sci.,53, 2861–2886.

  • ——, ——, and ——, 1997: Structure and evolution of the 22 February 1993 TOGA COARE squall line: Organization mechanisms inferred from numerical simulations. J. Atmos. Sci.,54, 386–407.

  • ——, M. A. LeMone, and W. C. Skamarock, 1998: Effect of three-dimensional structure on the stormwide horizontal accelerations and momentum budget of a simulated squall line. Mon. Wea. Rev.,126, 2580–2598.

  • Webster, P. J., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Soc.,73, 1377–1416.

  • Zipser, E. J., 1977: Mesoscale and convective scale downdrafts as distinct components of squall-line circulation. Mon. Wea. Rev.,105, 1568–1589.

  • View in gallery

    Horizontal cross section of reflectivity (dBZ) obtained from the LF radar of the P-3 N42 between 2215 and 2230 UTC. The domain analysis is 200 × 200 km2 (Guadalcanal Island is depicted by intense ground echoes east of the MCS and the flight track is in dashed line). Maximum measured reflectivity is 42 dBZ.

  • View in gallery

    Initial hodograph in the absolute and the relative framework. Arrows display the horizontal preconvective circulation in the system moving frame. The direction of the initial line direction is drawn for the three simulations.

  • View in gallery

    Horizontal cross sections of the simulated reflectivity (dBZ) at 1 km MSL deduced from an empirical formulation for ice every hour from 2 h of simulation. Shading intervals correspond to a 10-dBZ variation in the precipitation field, and the dense thin lines represent reflectivity greater than 40 dBZ.

  • View in gallery

    Horizontal cross section of relative wind field (arrows) and vertical vorticity (contoured) at 2 km MSL after 5 h, 30 min of simulation for ice within a 100 × 130 km2 subdomain. Arrows are plotted every two grid points, with vector length between these two grid points corresponding to 6 m s−1. The vertical vorticity contour increment is 1 × 10−3 s−1, positive values with dark shading starting at 1 × 10−3 s−1, and negative values with light shadings starting at −1 × 10−3 s−1.

  • View in gallery

    Horizontal cross sections at 2 km MSL of reflectivity (contoured) and relative wind field (arrows every grid points) for (a) ice after 4 h of simulation and (b) leg A. The horizontal domain is 63 × 63 km2 in both case. Grayscales are shown to the right of each panel.

  • View in gallery

    As Fig. 5 but for (a) ice after 5 h, 30 min of simulation and (b) leg B.

  • View in gallery

    Vertical cross sections of averaged vertical velocity w2d, of averaged line-normal velocity u2d and of averaged line-parallel velocity υ2d for, respectively, (a), (c), (e) ice after 4 h of simulation and (b), (d), (f) observation during leg A. These values are averaged within the boxes defined in Fig. 5a for ice and Fig. 5b for the radar. The contour intervals are 0.5, 2, and 1 m s−1 for w2d, u2d, and υ2d, respectively. Dashed lines present negative values.

  • View in gallery

    Vertical cross sections of the square root of the spatial variance σ2dw of the averaged vertical velocities w2d plotted in Figs. 7a and 7b, for (a) ice and (b) the observation. Isocontours are plotted every 0.5 m s−1.

  • View in gallery

    Vertical cross sections of averaged vertical velocity w2d and averaged line-normal velocity u2d relative to the system for (a), (c) ice after 5 h, 30 min of simulation and (b), (d) the observations during leg B, respectively. Contours are plotted every 1 and 2 m s−1, respectively; negative values are dashed. These values are averaged within the boxes defined in Fig. 6a for ice and Fig. 6b for the radar, across the northern part of the line during the bow stage.

  • View in gallery

    As for Fig. 8 but relatively to Figs. 9a and 9b, respectively.

  • View in gallery

    Vertical cross sections of averaged line-normal horizontal momentum flux ρuw2d in the system moving frame for the radar observations during (a) leg A and (c) leg B, and ice after (b) 4 h and (d) 5 h, 30 min of simulation (contour interval of 8 kg m−1 s−2, negative values dashed). The values plotted in (a), (b) are averaged within the boxes defined in Fig. 5a for ice and Fig. 5b for the radar, and those plotted in (c), (d) within boxes drawn in Figs. 6a and 6b, respectively.

  • View in gallery

    Profiles of (a) mean relative line-normal u-component wind speed, (b) mean relative line-parallel υ-component wind speed, and (c) mean vertical velocity. The bold continuous line corresponds to ice after 4 h of simulation; the bold dashed line to ice after 5 h, 30 min;the dashed line to leg A; and the dotted line to leg B. For the simulation, only the points situated in the precipitating area of E1 are taken into account.

  • View in gallery

    Vertical profiles (a), (b) of line-normal u-momentum flux and (c), (d) of line-parallel υ-momentum flux in the system moving frame for ice after 5 h, 30 min of simulation and leg B, respectively. For the simulation, only the points situated in the precipitating area of E1 are taken into account.

  • View in gallery

    Horizontal cross sections of the virtual potential temperature perturbation θ*vl at (a) 0.5 and (c) 4.3 km MSL with contour interval of 0.5 K, and of the pressure perturbation P* at (b) 0.2 and (d) 2.5 km MSL with contour interval of 10 Pa for ice after 5 h of simulation. Negative values are dashed. Arrows represent the relative wind circulation at the same levels every two grid points, with the vector length of these two grid box equivalent to 12 m s−1.

  • View in gallery

    Vertical cross sections across E1 of simulated (a) averaged vertical velocity w2d with contour interval of 0.5 m s−1, (b) averaged line-normal velocity u2d (2 m s−1), (c) averaged cloud virtual potential temperature perturbation θ*vl (0.5 K), and (d) averaged pressure perturbation P* (10 Pa). These values are averaged within the box A defined in Fig. 16a. The positions of the horizontal cross sections plotted in Fig. 16 are shown in (c) and (d).

  • View in gallery

    As Fig. 15 but for box B plotted in Fig. 16a across E2.

  • View in gallery

    Horizontal cross sections of (a) the virtual cloud potential temperature perturbation θ*c at 4.5 km MSL (0.2 K) and (b) the pressure perturbation P* at 2.5 km MSL (10 Pa) observed during leg C. Arrows represent the relative wind circulation at the same levels every two grid points.

  • View in gallery

    As in Fig. 3 but for the warm simulation.

  • View in gallery

    As in Fig. 3 but for the ice30 simulation. The simulation has been stopped after 6 h, as the simulated system has reached the domain frontiers.

  • View in gallery

    Evolution at 3, 4, and 5 h of the synthesis of the squall-line system simulated by the warm, ice, and ice30 experiments for first, second, and third rows, respectively. Active part of the cold pool leading edge is drawn as a cold front. Convective part is identified by the mesolow at 3 km (gray area) and the trailing stratiform precipitation by the highest density air at the 0°C level (hatched area). See text for details on the synthesis definition.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 13 13 1
PDF Downloads 2 2 0

A Tropical Squall Line Observed during TOGA COARE: Extended Comparisons between Simulations and Doppler Radar Data and the Role of Midlevel Wind Shear

View More View Less
  • 1 Centre d’Étude des Environnements Terrestre et Planétaires, Velizy, France, and Centre National de Recherches Météorologiques, Météo-France, and CNRS, Toulouse, France
  • | 2 Centre National de Recherches Météorologiques, Météo-France, and CNRS, Toulouse, France
© Get Permissions
Full access

Abstract

Results from a three-dimensional cloud model are extensively compared with airborne Doppler radar data in the case of a tropical oceanic squall line observed during the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment. The comparison is based on the precipitation patterns, the dynamical and thermodynamical distributions, and the vertical transport of horizontal momentum.

The model simulates the evolution of the mesoscale convective system (MCS) frontal convective line from a quasi-linear to a broken pattern. The area located south of the “break,” which designates the region where the MCS leading edge reorientates from the N–S to the E–W direction, is composed of a pronounced bow-shaped structure with two vortices located on both sides of a strong rear inflow.

The vertical circulation is characterized by a jump updraft and an overturning downdraft. Both structures exhibit a vertical, intense updraft in the break zone, whereas the jump updraft is more sloped and less intense in the bow region. Front-to-rear momentum is injected mainly by the jump updraft. Both observations and simulation indicate the major role played by convective eddies in the vertical transport of cross-line and parallel-line horizontal momentum.

A synthesis summarizes the complex three-dimensional structure of the simulated system, based on three salient features and their relative locations: the deep convection region, the leading edge of the cold pool, and the melting area. The relative positions between the two last mentioned explains the observed asymmetric structure and the existence of more upright and narrow updrafts in the northern part of the system. Numerical experiments suggest that the wind profile at midlevel is mainly responsible for the location of the melting area relative to the cold pool. The system tends to generate new convective elements organized along the direction that reduces the angle between the convective line and the midlevel shear vector.

Corresponding author address: Thibaut Montmerle, Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke Street West, Montreal, PQ H2H 2C2, Canada.

Email: montmerl@cumulus.meteo.mcgill.ca

Abstract

Results from a three-dimensional cloud model are extensively compared with airborne Doppler radar data in the case of a tropical oceanic squall line observed during the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment. The comparison is based on the precipitation patterns, the dynamical and thermodynamical distributions, and the vertical transport of horizontal momentum.

The model simulates the evolution of the mesoscale convective system (MCS) frontal convective line from a quasi-linear to a broken pattern. The area located south of the “break,” which designates the region where the MCS leading edge reorientates from the N–S to the E–W direction, is composed of a pronounced bow-shaped structure with two vortices located on both sides of a strong rear inflow.

The vertical circulation is characterized by a jump updraft and an overturning downdraft. Both structures exhibit a vertical, intense updraft in the break zone, whereas the jump updraft is more sloped and less intense in the bow region. Front-to-rear momentum is injected mainly by the jump updraft. Both observations and simulation indicate the major role played by convective eddies in the vertical transport of cross-line and parallel-line horizontal momentum.

A synthesis summarizes the complex three-dimensional structure of the simulated system, based on three salient features and their relative locations: the deep convection region, the leading edge of the cold pool, and the melting area. The relative positions between the two last mentioned explains the observed asymmetric structure and the existence of more upright and narrow updrafts in the northern part of the system. Numerical experiments suggest that the wind profile at midlevel is mainly responsible for the location of the melting area relative to the cold pool. The system tends to generate new convective elements organized along the direction that reduces the angle between the convective line and the midlevel shear vector.

Corresponding author address: Thibaut Montmerle, Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke Street West, Montreal, PQ H2H 2C2, Canada.

Email: montmerl@cumulus.meteo.mcgill.ca

1. Introduction

The study of the different physical processes responsible for the development of the convection over the western Pacific warm pool and its effects on large-scale budgets of heat, moisture, and momentum fluxes were the two main objectives of The Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE; Webster and Lukas 1992). The airborne Doppler radars aboard the two National Oceanic and Atmospheric Administration (NOAA) WP-3Ds and the National Center for Atmospheric Research Electra sampled several mesoscale convective systems (MCSs) during the intensive observation periods conducted during the 4-month field phase. The Doppler measurements allow retrieval of the three-dimensional air circulation, the precipitation distribution, and, indirectly, the thermodynamical fields of precipitating systems.

During TOGA COARE, numerous types of MCSs were observed and studied (e.g., Jorgensen et al. 1997, hereafter JLT97; Roux 1998; Hildebrand 1998; Lewis et al. 1998; Montmerle and Lemaître 1998; Chong and Bousquet 1999; Kingsmill and Houze 1999) and are summarized in Redelsperger et al. (1999). Their structure and evolution mainly depend on the vertical structure of the troposphere, characterized by the wind, relative humidity profiles, and thermal stratification (LeMone et al. 1998).

The 22 February 1993 MCS examined herein is a fast, eastward-moving oceanic squall line characterized by a 100-km-long convex north–south leading edge. During its more linear stage it was oriented almost perpendicular to the low-level wind shear, which was characterized by a strong westerly jet at 2 km above mean sea level (MSL). Figure 2 of JLT97, displaying horizontal cross sections of the mesoscale precipitation structure deduced from the lower fuselage (LF) radar aboard the two P-3s, shows an initially quasi-linear NNW–SSE-oriented MCS of heavy precipitation between 2100 and 2115 UTC turning into a pronounced bow-shaped leading convective line 45 mn later. During the next hour, the system evolves into a broken line of enhanced reflectivity (>40 dBZ) at its leading edge. At this time, the squall line is composed of two distinct elements, each approximately 70 km long, situated on both sides of this “break,” which coincides with an intense mesoscale convective vortex (MCV) and an associated region of intense precipitations. The main element located south of the break and the secondary one situated to the north will be referred as E1 and E2, respectively (Fig. 1). Doppler observations focused only on the evolution of the E1 element. JLT97 used these observations to show that, during its most linear stage, this part was self-maintained by the lifting and the rearward tilting of conditionally unstable moist air caused by a downdraft-induced rear inflow.

In the first of three papers, Trier et al. (1996, hereafter T96) studied the effect of ice parameterization and surface fluxes on the evolution of this squall line using a cloud resolving model (CRM). They showed that these physical processes do add realism but are not necessary to reproduce many of the features observed by the two P3s in the main element (E1) of the MCS. That included its evolution from a quasi-linear band of deep convection into a pronounced bow-shaped structure coinciding with a low- to midlevel MCV, and the development of elongated transverse bands of heavy precipitation in its southern region. In their following study, Trier et al. (1997, hereafter TSL97) illustrated how the front line of convective updrafts is modulated by the orientation of the low-level cold pool relative to the preconvective low-level vertical wind shear. They found that this MCV results of the tilting of the low-level horizontal vorticity generated by the horizontal gradients of buoyancy near the cold pool–updraft interface. They also demonstrated that the local differences in the convective structure and related mesoscale circulation were associated with differences in the MCS leading edge orientation relative to the ambient midlevel vertical shear. Trier et al. (1998) focused on the effect of the three-dimensional structure of the same simulated squall line on MCS-scale horizontal accelerations by analyzing momentum over different parts of the system. They found that the global vertical structure of the acceleration along the environmental low-level vertical shear has a weaker effect than the two-dimensional central region of the MCS, where the system tends to increase the midtropospheric vertical shear, as reported in past studies. Finally, a working group of the (Global Energy and Water Experiment (GEWEX)) Cloud System Study (GCSS; Redelsperger et al. 2000, hereafter R2000) choose this TOGA COARE case for a detailed intercomparison of CRMs. The organization, budgets, time series, and vertical profiles of settled quantities were compared with Doppler radar observations.

The validation of the CRM outputs owing to comparisons with observations, in particular with Doppler radar analysis, is an important step before using them to improve our physical understanding of convection. This exercise is difficult, mainly because of the transient character of the convection in both time and space and the difficulty of representing the inhomogeneities of the environment. Klemp et al. (1981) successfully compared the evolution of the precipitation and dynamical fields for a supercell thunderstorm. Lafore et al. (1988, hereafter LRJ88) found good general agreement between a three-dimensional simulation and observations of a continental tropical squall line observed during the Convection Profronde Tropicale experiment. They used a two-dimensional analysis to compare the mean circulation and the along-line temporal fluctuations as well as thermodynamic budgets and vertical transports of mass and momentum. Although good agreement between simulations and radar observation was obtained, improvements in both the model and radar analysis were needed, particularly those concerning the ice microphysical processes. Concerning TOGA COARE, T96 compared aspects of simulations with radar observations including the thermodynamic characteristics of the cold pool, updraft structure during the linear stage, and the precipitation and vortex structure during the bow-shaped stage of the same MCS.

While the importance of the low-level shear to determine the orientation and organization of squall lines has long been recognized (Thorpe et al. 1982; Barnes and Sieckman 1984; Rotunno et al. 1988; Lafore and Moncrieff 1989, among others), the role of midlevel shear has only been recently addressed by Alexander and Young (1992) for the Equatorial Mesoscale Experiment (EMEX) experiment and by LeMone et al. (1998) for TOGA COARE. Although LeMone et al. (1998) proposed a relationship between environmental shear and convective structure based on observations, basic studies are required to validate this conceptual model. The present case exhibits a strong westerly low-level shear up to 3 km, and a moderate ESE midlevel shear above the jet level at 2 km; it is therefore appropriate in this regard.

The main goal of this paper is twofold: first, to compare a simulation of the 22 February TOGA COARE system with analysed Doppler radar observations and, second, to quantify the role of the midlevel shear. The observational techniques and the simulation configuration are described in section 2. Section 3 presents the main characteristics of the reference simulation (ice) and comparisons with the horizontal structure deduced from Doppler radar observations. Comparison of the vertical velocity circulation and of the momentum transport within the main structure of the squall line are presented in sections 4 and 5, respectively. Finally in section 6, the role of midlevel shear is addressed by analyzing the differences in behavior between E1 and E2, explained in terms of the impact of the upper-air wind profile on hydrometeor trajectories.

2. Observation and simulation methodology

a. Corrections on Doppler radar data

The flight track and the sampling strategy used by the two NOAA WP-3D radars (designated as N42 and N43) are extensively described in JLT97. Oury et al. (1999) found that N43 underestimates the reflectivity by 1.2 dBZ relative to the N42 for the 9 February 1993 case. Such a study is not possible for the 22 February 1993 case, because no microphysical measurements exist. A similar comparison between the N43 and the ELDORA-ASTRAIA radar of the NOAA Electra shows that the reflectivities of the N43 were 1.2 dBZ larger than the ASTRAIA ones, which were also undercalibrated by 4 dBZ. Thus, the reflectivity field measured by the N42 and the N43 were underestimated by 1.6 and 2.8 dBZ, respectively.

A second problem occurs in the radar data analysis at X band, when the radar wave is strongly attenuated through heavy precipitation. Thus the radar reflectivity may be severely negatively biased. We chose the dual-beam analysis (Testud and Oury 1997), based on the Hitschfeld and Bordan (1954) algorithm, in order to correct attenuation. This enables the retrieval of new cells of heavy precipitation, hidden by intervening high reflectivity cells. The most attenuated regions exhibit reflectivity increases of up to 7 dBZ. The dual-beam analysis was systematically applied only underneath the 0°C level (4.5 km MSL), as the attenuation is smaller for the ice phase regions.

These two corrections had a significant impact on the values of the retrieved reflectivity and gave confidence in the precipitation distribution. For comparison with observations, the simulated rain mixing ratio qr (g kg−1) has been converted in reflectivity Z (mm6 m−3) using an empirical formulation deduced from ASTRAIA microphysical data (V. Marecal 1996, personal communication):
i1520-0493-128-11-3709-e1

b. Three-dimensional dynamical and thermodynamical radar retrieval

The three-dimensional wind field is retrieved with the analytical technique MANDOP (for multiple analytical Doppler), developed by Scialom and Lemaître (1990), which allows the three wind components to be expressed in analytical form. Briefly, this method variationally adjusts the analytical wind to the observed one constrained by the anelastic continuity equation and a lower kinematic boundary condition. In addition to quantities directly related to the wind field such as vorticity, the method has been used to retrieve pressure and virtual “cloud” potential temperature perturbations (P* and θ*c, respectively) to within a constant depending on the altitude (see Protat et al. 1998 for more details). The analytical formulation of the wind field allows the direct derivation of an analytical form for the thermodynamic perturbation fields without discretization.

A horizontal mesh size of 1500 m and a vertical resolution of 500 m are chosen for each of the three legs presented herein: leg A corresponds to data sampled approximately between 2110 and 2120 UTC, leg B between 2158 and 2208 UTC, and leg C between 2218 and 2230 UTC. Leg A matches the most linear, mature stage of the squall line and the other two the evolution of the bow stage.

c. Numerical model and simulation configuration

We used the three-dimensional nonhydrostatic model Meso-NH (Lafore et al. 1998) jointly developed by the Centre National de Recherches Météorologiques [Météo-France and Centre National de la Recherche Scientifique (CNRS)] and the Laboratoire d’Aérologie (CNRS). This model is based on the Lipps and Helmer (1982) modified anelastic system incorporating a three-dimensional turbulent scheme (Cuxart et al. 2000) based on the turbulent kinetic energy equation of Redelsperger and Sommeria (1986). The microphysics is represented by a bulk scheme combining a three-class ice parameterization with the Kessler (1969) scheme for warm processes (Pinty and Jabouille 1998; Caniaux et al. 1994).

The model configuration and the initialization procedure are as defined in R2000 for the GCSS model intercomparison. The horizontal simulation domain extends 150 km along and 100 km across the squall line. Open lateral boundary conditions are used. The horizontal resolution is 1250 m and a vertically stretched grid interval is used, varying from 70 m at ground level to 700 m at the model top. The simulation period is 7 h with a time step of 7.5 s. A gravity wave absorbing layer is situated between 17- and 22-km altitude, below a rigid upper boundary. The initial conditions are horizontally homogeneous and composed of rawinsonde observations (Honiara site) and P3 aircraft data at low levels. This sounding is characterized by moderate instability (convective available potential energy around 1500 J kg−1) in a moist environment (see Fig. 2 of R2000). The environmental wind profile exhibits moderately strong low-level shear with a jet (12 m s−1) at 2 km above MSL (Fig. 2). The observed squall line is oriented nearly perpendicular to the low-level shear and it propagates at about 12 m s−1.

The squall line is initiated in the middle of the domain by an artificial density current, emulated by four mid-oval-shaped cooling and drying sources applied along the line over a 2.5-km depth during the first 20 min of the simulation, with maximum intensities of 6.7 × 10−3 °C s−1 and of 2.4 × 10−6 kg kg−1 s−1, respectively. A constant system velocity of 12 and −2 m s−1 along the x axis and the y axis, respectively, was added to keep the system inside the simulation domain. The above initialization proposed by R2000 differs from that used by T96. Besides the minor differences concerning the size and shape of the initial cold pool, its 330°–150° orientation corresponds to the observed orientation and differs by 30° from the present study. In addition to the reference simulation called ice, two supplementary experiments were performed to examine the role of the microphysics and the initial cold pool orientation. The warm simulation uses only the warm microphysics scheme, whereas ice30 is identical to the ice simulation except for the initial cold pool oriented as in T96 (see Fig. 2).

3. Comparison of horizontal behaviors

a. Structure and evolution of the precipitation field

Figure 3 presents the evolution of the simulated radar reflectivity at 1 km for the ice experiment. During the first 2–3 h, the line structure is obviously strongly affected by the initialization procedure, both in terms of length and orientation. During this stage the convective line retreats, as the initial cold pool weakens and the precipitation pattern is stretched rearward by the wind in the northwest direction.

After 3 h the simulated system develops an asymmetric bow-shaped structure in the southern part, whereas the northern part develops a line oriented 330°–150° as observed. The salient features of ice are the intense break formation and the splitting of the leading edge of the MCS into elements E1 and E2 consistent with the radar observations (Fig. 1). The mesoscale reflectivity field measured by the P3 lower fuselage radar between 2215 and 2230 UTC (Fig. 1) presents the best fit with the ice simulation between 5 and 6 h, with a broken line of enhanced reflectivity at the leading edge and intense precipitation located near the break region.

The simulated squall line is composed of a 80–100-km-long band of heavy precipitation (>40 dBZ) and of an extended stratiform region that becomes fully developed after 5 h. However, the simulated system is smaller than observed (200 km), arguably due to the simulation domain and also the initialization procedure. After 5 h, the elongated bands of high reflectivity (>40 dBZ) perpendicular to the main structure in the southern portion are realistic. T96 realized similar structures and showed that these reflectivity bands become elongated as the front-to-rear flow intensifies and progressively extend rearward in the southern portion of the line. As in T96, the simulated convective system continues to intensify after 6 h, whereas observations indicate a decay approximately 45 min after the bow stage. T96 attributed this disagreement to the change in ambient low-level conditions not included herein.

b. Convective circulation pattern

The horizontal wind circulation at 2 km MSL for ice at 5 h, 30 min is displayed in Fig. 4, during the bow stage in addition with the vertical vorticity field. The E1 element contains two vortices, referred to as VN and VS, about 20–30 km in diameter situated at the northern and southern flanks, respectively. The two vortices drives an intense relative rear inflow locally exceeding 6 m s−1 that pushes forward E1 giving it a bow shape. Similar features were obtained by T96. The maximum vertical vorticity (5 × 10−3 s−1 for VN and −3 × 10−3 s−1 for VS) is located between 2 and 3 km. Observations during legs B and C display these two mesovortices, with weaker values located approximately 1 km higher (1.4 and −0.6 × 10−3 s−1 in VN and VS, respectively; not shown).

As observed, a small clockwise vortex (CVN in Fig. 4) coupled to VN is simulated on the northern side of the strong NE flow that rushes into E1, contributing to the rear inflow associated with E2. Finally, another “end of the line” vortex (VE hereafter) comparable with VN appears on the northern flank of E2. The maximum intensity of these two vorticies is almost as strong as VN and VS, but their diameters do not exceed 6 km. Like E1, they straddle the rear inflow, which is weaker. The following comparison with Doppler radar analysis focuses on E1 over the subdomain outlined on Fig. 4, which is the only region where Doppler observations were available.

c. Comparison of precipitation and horizontal circulation with observations

Figure 5 compares observations with simulation ice during the linear stage for the reflectivity and circulation pattern at 2 km. The same basic features occur but on a smaller scale than observed (20 vs 40 km). Both horizontal cross sections exhibit the northern vortex VN still in its formative stage and the growing albeit smaller southern vortex VS. At this time the break formation is indicated by the intensification of the NE flow at the northern line end but the rear inflow does not extend sufficiently rearward compared to observations. Other differences between observations and the simulation concern the lower intensity of the observed radar reflectivity (3–5 dBZ).

Even starting with such differences in size and intensity but similar structure, the simulation generates 1h, 30 min later a convective element E1 that closely resembles the radar observations during the bow stage (Fig. 6). The main features are realized with the occurrence of the break characterized by a strong intensification of precipitation and of the NE ascending flow (from 10 to 15 m s−1) into the system and the rotation and elongation of the convective cells in the east–west direction. The associated vortex is weaker and of larger scale in observations (Fig. 4). Although the size of the bow-shape element E1 is similar for both numerical realizations, the rear inflow and VS are more pronounced at this altitude, with VS clearly evident only near 4 km in the observations (not shown). The simulated precipitation overestimates the reflectivity (about 5 dBZ). The elongated band of precipitation perpendicular to the southern portion of the line is reproduced. Nevertheless the simulated rear to front flow extends up to ∼30 km behind the system’s leading edge, rather than the 50 km seen in observations. The different large-scale forcing and the system duration, which are not expected to be well reproduced with this idealized simulation over a limited domain size, can explain these differences.

4. Comparison of vertical structure

In order to objectively compare the 3D simulated wind fields with those retrieved from Doppler radar data, the 2D statistical analysis proposed by LRJ88 is used with the same definition of averaging operators except that temporal fluctuations are omitted as they showed that it does not basically change the results. Due to the large data volume and the complexity of the system, comparison is performed in two specific regions, namely the linear and bow stages. The choice of box width (15 km) results from a compromise. The regions needs to be sufficiently homogeneous (thus not too large) to reveal their physics and to be statistically representative. The choice of the box orientation and propagation speed is not obvious due to the 3D structure of the system and its time evolution, contrary to previous studies concerning stationary quasi-2D squall lines. The moving frame speed is taken as the one averaged for the whole system and over the entire simulation period.

a. Quasi-two-dimensional structure during the linear stage

The 2D statistical analysis is applied to the central part of the squall line (see boxes in Fig. 5) for the wind fields of leg A and of ice after 4 h of simulation. The mean vertical circulation (Fig. 7) exhibits the typical structure of a fast-moving squall line characterized by two main branches: a rearward tilting updraft and an overturning downdraft following the classification based on stationary dynamical models of convection (Moncrieff 1992). This classical circulation pattern was obtained in previous studies of the same squall line (Trier et al. 1996, 1997; JLT97) and has been extensively observed (Houze 1977; Zipser 1977; Chong et al. 1987, among others).

As discussed in T96, the jump updraft is characterized by a pronounced rearward tilt (about 25° from horizontal) for both simulation and observations during the linear stage with the same maximum vertical velocity w2d (5 m s−1). Nevertheless, the observed line-averaged updraft is broader and smoother than its simulated counterpart, arguably due to the 10-min duration of leg A compared to the instantaneous simulated fields. Doppler radar cannot detect the shallow but intense updraft forced by the precipitation-free leading edge of the cold pool. The second core maximum is nevertheless higher in the radar observations than for ice. The jump updraft corresponds to an intense injection of front-to-rear wind (up to −16 m s−1) for both realizations (Figs. 7c and 7d).

The overturning downdraft attains similar intensity (−1 m s−1) and is connected to a moderate rear-to-front flow (maximum of 4 m s−1 in both cases). Differences between observations and ice reveal a deeper rear inflow for observations (5 vs 4 km) and a wider horizontal extension, suggesting that the rear inflow was partly forced at a larger scale.

Analysis of the mean circulation parallel to the line (Figs. 7e and 7f) is more difficult. Basically a northerly relative wind prevails at low levels up to 3 km as confirmed by Fig. 5 taken at 2 km, whereas southerly flow prevails above. Injection of northern wind is thus detected in the jump updraft but with weaker intensity and on a wider area for observations. In other words, the structures are similar although strongly smoothed for observations; this is consistent with the higher intensity and smaller size of the vertical vorticity in the simulation.

Figure 8 presents the intensity of along-line fluctuations measured by the standard deviation of vertical velocity σ2dw. The jump updraft coincides with intense fluctuations with intensity (up to 4 m s−1) comparable to the mean ascent w2d (5 m s−1), consistent with the three-dimensional nature of this region. The intensity of fluctuations obtained from the radar analysis (Fig. 8b) is half that of the simulated intensities. A possible reason is that shortest waves are filtered implicitly by the analytical formulation of the wind field. The upper part of the radar domain exhibits large variances of vertical velocity due to the lack of data resulting from the strong signal attenuation, with both aircraft flying near 1.5 km. The along-line fluctuations of horizontal wind (not shown) are intense (σ2du up to 3 m s−1) along the sloped interface separating the jump updraft from the overturning downdraft. For the same reasons argued for the vertical velocity, they are stronger than observed.

b. Bow stage

During this stage, corresponding to leg B and to ice after 5 h, 30 min of simulation, the mean 2D circulation retains the basic characteristics of the linear stage, except that the convective updrafts are weaker, and the ascent is slightly less sloped (not shown). The 2D statistical analysis obtains the circulation in the break region between convective elements E1 and E2 (Fig. 9). The boxes are oriented parallel to the intense northeasterly flows (see Fig. 6 for box locations) to better follow the mean flow, which is not locally normal to the leading edge. In contrast to the linear stage, the mean updraft ascends almost vertically, attaining more than 8 m s−1 in both observations and simulation. As for the linear stage, the observed structures are wider and smoother than the simulated ones. Behind this intense convective mean ascent both analyses display a transition zone characterized by descending motions. Observations suggest a broad (20 km), deep (over the depth of the troposphere), and intense downdraft (−1.5 m s−1) with ascending motion rearward. The ice simulation presents a narrower and weaker transition zone followed by a typical stratiform part with ascending motion above 6 km (2 m s−1) and subsidence below. The lateral fluctuations of vertical velocity, as measured by the standard deviation σ2dw, reach about 2 m s−1 (Fig. 10a) except at upper levels for ice. This suggests that the main updraft in the break zone has a moderate three-dimensional organization consistent with observations (Fig. 10b).

Concerning the horizontal wind component along the cross section, both analyses show a front-to-rear flow in the convective, transition, and stratiform upper regions, but it is stronger at midlevel for the simulation (about 10 m s−1 instead of 6 m s−1) (not shown). The intensity (up to 4 m s−1) and depth (4–5 km) of the rear inflow are comparable in observations and simulation. The observed rear inflow extends farther rearward as during the linear stage than in the simulation. Injection of NW wind at mid- and upper levels is noted albeit stronger, narrower, and shallower in the simulation.

5. Convective momentum transport

a. Structure of the line-normal horizontal momentum flux

As the system is characterized by intense jump updrafts injecting large front-to-rear momentum, it is interesting to analyze the attendant momentum transports and make comparisons with radar observations. Figure 11 displays cross sections of the averaged vertical flux ρuw2d of horizontal momentum approximately normal to the line in a moving frame of reference corresponding to the two vertical slabs (Figs. 5 and 6) previously analyzed (Figs. 7 and 9). During the linear stage (cf. Figs. 11a and 11b) the sloped ascent results in intense negative momentum transport (up to −32 kg m−1 s−2). Simulation ice and observations suggest similar structures. Differences are consistent with the previous discussions:the core of maximum transport is 2 km higher for observations and an additional core in the transition zone between the two main cells of vertical transport of u momentum occurs in the simulation. Transports are weak outside this zone. LRJ88 found a similar organization of the cross-line momentum transport for a fast-moving African squall line as deduced from radar observations and from a simulation during its mature stage, albeit of greater intensity (about 50 kg m−1 s−2).

Intense negative transport is also detected in the break zone during the bow stage (Figs. 11c and 11d), but it occurs in the convective zone. The flows in this region ascend almost vertically (Fig. 9). As detected in the mean updraft characteristics, the core of momentum transport is much stronger in the simulation (−40 compared to −16 kg m−1 s−2) and is narrower. Rearward of the main mean updraft, the transport patterns suggested by radars and ice differ. Whereas the simulation indicates negative transport above 6 km in the stratiform anvil and positive below, radar analysis exhibits weaker transport of the same sign over the entire depth of the system. Nevertheless, the transport remains weak outside the convective zone. This confirms that intense cross-line momentum transports occur in the convective region during the entire system life cycle.

b. Vertical profiles of mean velocities

Momentum transports are now horizontally averaged over similar domains, that is including data collected during each leg for radar observations and data of ice within the precipitating region E1. In the horizontal averages of the simulation, the columns situated north of the break were ignored, because these areas were not sampled by the radars. The difficulty of defining similar regions may impact the results. The resulting subdomains correspond to about a 65-km square corresponding to Figs. 5 and 6.

The vertical profiles of mean velocities are plotted in Fig. 12. Differences above 5 km are evident between radar and model analyses, arising mainly from the lack of radar sampling at higher altitudes due to strong attenuation. The effect is clear on the averaged vertical velocity w profile (Fig. 12c) where the maximum is greater than twice that simulated (1.3 vs 0.5 m s−1) during the linear stage. This difference is greatly reduced during the bow stage (0.9 vs 0.7 m s−1). The altitude of this maximum is located at about 8 and 10 km for simulation and observations, respectively.

The profiles of mean relative line-normal velocity u, plotted in Fig. 12a, show similar behavior below 4 km. The observed and simulated squall lines propagate at the speed of the mean speed at low levels. The altitude of the maximum rear inflow decreases with time, a characteristic more obvious for the simulation whose profile after 5 h, 30 min suggests an acceleration of the density current and a reduction of the front-to-rear outflow, indicating the onset of the system collapse. On the other hand, radar observations show a shear decrease above 8 km MSL, with leg B displaying values of u up to 5 m s−1 less than ice after 5 h, 30 min.

The slope and the intensity of the mean relative line-parallel velocity υ are similar, except in the boundary layer where the radar depicts 2 m s−1 stronger values (cf. Fig. 12b).

c. Vertical profiles of horizontal momentum transport

Momentum fluxes are analyzed by separating them into mean and perturbation parts. The mean flux is calculated as the product of horizontal averages u (υ) and w discussed above. Perturbation fluxes are computed as departure of local fluxes from mean fluxes and represent the contribution of convective eddies.

The line-normal u-momentum flux relative to the system is displayed in Figs. 13a and 13b for ice after 5 h, 30 min of simulation and leg B, respectively. As expected the profile of total vertical transport is negative and has the same structure and magnitude for both cases, with a double peak at 2 km and 7–8 km with maxima of about −2 and −2.7 kg m−1 s−2, respectively. Both analyses indicate that the convective eddy contribution is dominant below 4 km. At higher levels the mean ascent becomes the main contributor to the transport of u momentum especially in the simulation, whereas the observed contribution of convective eddies partly compensates the stronger contribution by the mean ascent above 8 km.

Concerning the transport of along-line υ momentum, both analyses suggest that the transport by the mean ascent is weak and that the main contribution comes from the convective eddies. Below 4 km negative transport of υ momentum occurs due to eddies, but is stronger for the simulation (−1.4 vs −0.7 kg m−1 s−2). Above 4 km the two analyses diverge, as intense positive υ-momentum transport is observed (up to 1.5 kg m−1 s−2) whereas the model suggests weak negative transport.

In short, this comparison shows that below 4 km observations and simulation agree rather well and stress the major role played by convective eddies in the vertical transport of both cross-line and parallel-line momentum. These results confirm previous findings stressing that the transport of cross-line momentum due to convective eddies can be important for fast-moving squall line (LRJ88; Gao et al. 1990; Caniaux et al. 1995) especially in the convective region. LRJ88 found stronger values (8 vs 2 kg m−1 s−2 for convective eddies) but over a smaller domain (30 km vs 65 km wide) in the convective part of an intense African squall line.

6. On the importance of midlevel shear

a. Differences of behavior between the E1 and E2 elements

To describe and compare the structure of E1 and E2 elements after 5 h of simulation, we will alternatively use horizontal cross sections (Fig. 14) and mean vertical cross sections (Figs. 15 and 16) taken across these elements computed over vertical slabs A and B whose locations are drawn in Fig. 14a.

Horizontal cross sections of the virtual potential temperature perturbation of the total cloud θ*vl are plotted in Fig. 14 [with θvl = θ(1 + 0.61 × qυqcqrqiqsqg); θ being the potential temperature and q∗ the mixing ratios of vapor, cloud, liquid rain, ice cloud, snow and graupel, respectively]. Figure 14a shows the storm-induced cold pool exhibiting an asymmetric structure. Whereas the E1 element is associated with an intense and widespread density current (θ*vl < −3 K), the north element E2 presents a weaker and narrower cold pool. This asymmetry also exists at the freezing level (4.3 km; Fig. 14c) with the E1 convective line showing a large area of positive buoyancy followed by negative values in the stratiform part, whereas a narrow line of buoyant cells is observed in its northern part followed by a zone of intense negative buoyancy of limited extension. Mean vertical cross sections (Figs. 15c and 16c) complete the description of the buoyancy field. Clearly E1 corresponds to a stronger, deeper, and wider density current, whereas E2 exhibits an intense melting layer whose leading edge is only 5 km behind the density current limit, contrary to the E1 configuration (20 km behind). This dramatic difference in the buoyancy field between the two regions is reflected in all other fields. In particular, the pressure deviation field, which is strongly dependent on the buoyancy field (Lafore and Moncrieff 1989; Caniaux et al. 1995; TSL97), also exhibits an asymmetric organization as displayed by the horizontal cross sections taken at 200-m and 2.5-km levels (Figs. 14b and 14d, respectively). Close to the surface the mesohigh induced by the density current is therefore stronger and wider in the central part of the system than to the north. At 2.5 km the situation is reversed: the convective mesohigh is much stronger to the rear of E2 due to a stronger melting layer, whereas the convective mesolow is wider and stronger for the E1 element due to stronger buoyant cells (Figs. 15d and 16d). The stronger resulting asymmetry is detected at 2.5 km from the pressure field (Fig. 14d) with two main centers of action: a mesolow band (−0.6 hPa) in the E1 convective part and a mesohigh center (0.3 hPa) located just at the rear of E2. The mesovortices are strongly connected to the pressure field at that level, in particular the northern and southern vorticies VN and VS wrap around the minima of pressure observed at the ends of the mesolow band, with the maximum rear-to-front flow in between.

Because Doppler radars sampled the central part corresponding to the E1 element, Fig. 17 shows the virtual cloud potential temperature θ*c and pressure perturbations P* fields retrieved from the wind field analyzed by the MANDOP technique, for comparison with simulated fields at the same levels (Figs. 14c and 14d). As previously noted except for the smaller horizontal extent of this region in the ice simulation, the structures are similar: a large area of positive buoyancy at 4.5 km with maxima at both ends of the band, and a mesolow stronger on the northern flank (−0.6 hPa) located at the edge of the break zone.

The horizontal gradients of pressure for E1 and E2 are thus quite different. The resulting rear-to-front acceleration at the rear of E1 is intense and concentrated at low levels. It is consistent with a strong rear-to-front flow (Fig. 15b) and a sloped updraft (Fig. 15a). On the contrary, the rear-to-front acceleration at the rear of E2 is higher (4 km) and weaker. It corresponds to a thin layer of rear-to-front flow (Fig. 16b) and an almost vertical updraft (Fig. 16b).

The above description has many similarities to the results of TSL97 who compared the structures of the north flank and of the central part (see their Figs. 9 and 10). The major difference with their simulations is that their north flank region is a short line of isolated convective cells whereas E2 is a longer continuous line in better agreement with observations.

b. Sensitivity of the system organization

In order to study the reasons for the break occurrence and for the E2 formation, we have performed two additional simulations warm and ice30, without modifying the initial profile.

1) Role of hydrometeor fall speeds: Warm simulation

The impact of the ice phase is illustrated by comparing the evolution of the reflectivity pattern from the warm experiment where the ice phase is not considered (Fig. 18) with those from ice (Fig. 3). The structures are very similar up to 3 h, but simulations subsequently diverge after. When only warm rain processes are considered, the system exhibits a rather symmetric pattern without E2, but with a longer bow-shaped element E1. The north part of the system consists of a weak east–west line of isolated convective cells. Table 1 indicates that warm generated more precipitation during the first simulation period but a similar total rain amount after 5 h.

2) Role of the initial cool pool orientation: Ice30 simulation

In experiment ice30, the orientation of the initial cold pool is similar to that used in the CONTRL experiment of T96. At 2 h (Fig. 19) the simulated orientation and extent are different from those previously simulated (Figs. 3 and 18). The system exhibits similarities up to 6 h with the T96 simulation, including the bow-shaped E1 element with similar structure, intensity, and orientation. As for the CONTROL experiment of T96 and the present warm experiment, ice30 does not develop the E2 element contrary to ice. Globally ice30 generates more precipitation (Table 1) and the length of the bow-shaped line is much longer, in better agreement with observations. Simulations ice and ice30 generate end vortices with the associated typical break structure of the reflectivity field as observed (Fig. 1). Both simulations reproduce the elongated band of enhanced precipitation normal to the main line, contrary to the warm simulation, which confirms the idea of T96 in the role of ice phase to structure these bands.

c. Synthesis of the simulated storm

The above sensitivity study suggests that both ice phase and line orientation play a crucial role on the system organization and in the E2 formation. To better compare these simulations and explain the differences, Fig. 20 gives a synthetic representation of the storm evolution as simulated by the three experiments. This synthetic picture is based on a selection of three salient features characterizing the system organization as suggested by the discussion of the 3D storm organization given in section 6a.

  1. The first feature corresponds to the cold pool whose strength and structure determine the triggering of new cells, the system orientation, and its propagation. It is outlined by the dashed lines corresponding to θ*vl values lower than −0.5 K at 500 m above MSL. The most intense parts of the leading edge of this density current (DC hereafter) (strongest θ*vl gradient) is drawn as a pseudo–cold front (gust front).
  2. The convective region is well correlated with the convective mesolow at 3 km just below the maximum of heat release. The envelope of positive buoyancy also corresponds to the convective part but presents a spotty pattern at the convective scale. Thus, the low pressure pattern has been chosen to identify the most active convection. Also the convective mesolow is strongly connected with the mesovortex as seen in observations and simulations (Figs. 17b and 14d). This second feature will be delineated by the pressure deviation lower than −0.2, −0.4, and −0.6 hPa (gray areas) at 3 km above MSL.
  3. The buoyancy field at the 0°C level (4.3 km) quantifies the melting and the amount of solid precipitation that strongly controls the mesohigh generated underneath the trailing stratiform part. This governs the midlevel rear inflow (Lafore and Moncrieff 1989;Caniaux et al. 1995), the system organization, and its interaction with larger scales. This last feature is sketched by buoyancy lower than −1.0°, −1.5°, and −2.0°C (hatched areas).

Figure 20 shows that the DC is stronger and wider, and exhibits a longer active part leading edge for warm than for ice (70 vs 40 km), as warm rain quickly reaches the surface, feeding a stronger DC. At 4 h both experiments exhibit the premise of the E2 north element. The major difference is the negative buoyancy area at the 0°C level just behind E2 of ice due to melting, which is assumed to be the reason for the E2 development in ice, compared to its collapse in warm.

After 3 h of ice, the DC extension is still moderate and the active part of its leading edge is much shorter than the length of the initial cold pool (30 vs 80 km). Nevertheless it retains its initial north–south initial orientation. The large amount of stratiform precipitation accumulated to the north generates an intense melting layer and a subsequent midlevel mesohigh that intensifies the E2 element and the break between the two parts of the system. After 5 h the simulated system confirms its asymmetric structure organized along the observed line direction, differing by 30° from the initial orientation.

The ice30 system development is very fast, suggesting that it is a privileged direction in agreement with observations. The convective part, DC, and gust front are longer (80 km) and more vigorous. For instance, the convective mesolow is already well formed at 3 h and its extent is in better agreement with the pressure field derived from the Doppler radar fields (Fig. 17b). Ice30 presents a symmetric structure where the melting area shifts to the system rear along the central axis of the main bow-shaped element. Also the width of the low pressure band progressively increases as the slope of the convective part decreases and the rear inflow intensifies, resulting in the system acceleration.

d. Relation between wind shear and the fall of hydrometeors

Previous results suggest the three-dimensional system organization is determined by the relative locations of the two layers of higher density (i.e., the density current and the melting layer), which directly follow the hydrometeor trajectories. The latter results from a combination of the particle fall speed and the mean wind prevailing in the crossed layers relative to the system.

The hodograph (Fig. 2) allows the hydrometeor trajectories to be estimated. It exhibits a strong westerly low-level shear up to 3 km and a moderate ESE midlevel shear above the jet maximum located at 2 km. Liquid precipitation below the 0°C level (4.3 km) is moved to the west of the leading edge, but not very far due to the high fall speed (∼7 m s−1) leading to a short precipitation cycle. In contrast, solid precipitation above 4.3 km travels WNW over longer distances due to weaker fall speeds (1.4 and 4.0 m s−1 for snow and graupel, respectively). We must distinguish the motion along and across the line depending on the angle between the system-relative wind and the orientation of the line leading edge. Cross-line displacements will change the distance between the DC and the mesohigh underneath the trailing stratiform part, and will thus govern the width of the system and the intensity of the rear inflow. Along-line displacements will influence the asymmetric organization of the system.

These rules can be used to help to answer to the two following questions: why is E1 shorter for ice, and why does E2 formation occur only for ice? Concerning E1 formation in the ice experiment, the distance covered by liquid and solid precipitation is increased across the line and reduced along it as compared with ice30. It allows for triggering new convective cells toward the north of E1 in the ice experiment after 4 h. On the other hand, larger precipitation fall speeds favor the DC feeding and the system intensification during the development stage of warm. The lack of stratiform precipitation in warm avoids the generation of the midlevel mesohigh at the rear of E2 and thus its intensification, contrary to ice. For ice30, both line rotation and rapid system development characterized by intense rear-to-front midlevel flow and acceleration, tend to advect solid precipitation farther to the rear and less to the north as compared with ice. The resulting more rearward position of the midlevel mesohigh is not favorable for the E2 formation.

7. Conclusions

Three-dimensional simulations of a tropical squall line observed during TOGA COARE have been performed using a cloud-resolving model. Detailed comparisons between the simulation and data collected with the LF and the tail Doppler radar of the NOAA’s P3 show many similarities. The model reproduces the precipitation pattern with broadly the same intensity. This includes the evolution of the initial quasi-linear structure to a broken line of enhanced reflectivities, separating the system into two distinct elements, E1 and E2, situated north and south of the break, respectively. Reflectivity enhancement located near this break and bands of heavy precipitation perpendicular to the frontal convective line in the southern part of the main system E1 are also simulated.

General agreement is also found for the precipitation distribution and the horizontal and vertical air circulation in the convective element E1, the only part where Doppler data were available. During the bow stage of the system, the two observed clockwise and counterclockwise mesoscale vorticies situated on both sides of a strong rear inflow are realized, although the intensity of the observed southern vortex is weaker. The strong convective towers characterizing the bow-shaped northern flank of E1 are also reproduced.

Both radar observations and simulation exhibit the mean vertical circulation typical of a fast-moving squall line characterized by two main branches: a jump updraft and an overturning downdraft. Whereas the mean jump updraft is sloped and moderate (5 m s−1) during the linear stage, it rises almost vertically and is stronger (8 m s−1) in the break area during the bow stage. During both stages, a rear inflow zone is identified in both radar observations and simulation. Nevertheless, its vertical and rearward extent is less in the simulation. As a result of this vertical circulation, front-to-rear momentum is accelerated mainly by the main jump updraft. Simulation and radar observations also stress the major role played by convective eddies to vertically transport cross-line and transverse horizontal momentum.

A synthesis is proposed to summarize the complex three-dimensional structure of the convective system, by selecting three salient features: the cold pool leading edge, the deep convection region, and the melting area. The synthetic representation (Fig. 20) gives the evolution of these key features for simulations that differ in their initial orientation of the cold pool and in microphysical parametrizations. Their comparison allows the importance of midlevel shear and the fall of hydrometeors in the system structure to be quantify. In particular, only the reference simulation ice reproduces the north element E2. The relative locations of the two layers of high density (i.e., the density current and the melting layer) determine the three-dimensional system organization. Whereas E1 exhibits an intense cold pool and a large sloped area of deep convection, E2 corresponds to an intense melting layer overlying a weak cool pool with a narrow intense convective zone. We hypothesize that the wind profile at midlevel is mainly responsible for the location of solid precipitation and thus determines the location of the melting area relative to the cold pool. If at a given time the line orientation forms a critical angle relative to the midlevel shear, the system will be less intense in a first stage. Nevertheless, it will later generate new convective elements to organize itself along a new direction to reduce this angle.

This interpretation attributes an important role to the midlevel shear as already proposed by Alexander and Young (1992) and LeMone et al. (1998), whereas most previous studies stressed the role of the low-level shear. It also raises the question of the impact of the initial conditions in idealized simulations, since the observed evolution of the system can evidently depend on its own history. Work is under way to initialize more realistically the model with wind fields retrieved from the Doppler radar observations (Montmerle 1998).

Acknowledgments

We would like to express our gratitude to the support brought by the whole Meso-NH team and more especially to Drs. P. Jabouille, J. Stein, J. P. Pinty, V. Ducrocq, and Ms. J. Duron, at a time when the model Meso-NH was still under development. Pertinent comments and annotations of the manuscript by Drs. M. Moncrieff and S. Trier have been appreciated. The authors also thank N. Badrinath of McGill University for his helpful comments on the manuscript. This work at CETP and CNRM/GAME has been partly funded by the PATOM program of CNRS.

REFERENCES

  • Alexander, G. D., and G. S. Young, 1992: The relationship between EMEX mesoscale precipitation feature properties and their environmental characteristics. Mon. Wea. Rev.,120, 554–564.

  • Barnes, G. M., and K. Sieckman, 1984: The environment of fast- and slow-moving tropical mesoscale convective cloud lines. Mon. Wea. Rev.,112, 1782–1794.

  • Caniaux, G., J. L. Redelsperger, and J. P. Lafore, 1994: A numerical study on the stratiform region of a fast moving squall line. Part I: General description and water and heat budget. J. Atmos. Sci.,51, 2046–2073.

  • ——, ——, and ——, 1995: A numerical study on the stratiform region of a fast moving squall line. Part II: Relationship between mass, pressure, and momentum fields. J. Atmos. Sci.,52, 331–352.

  • Chong, M., and O. Bousquet, 1999: Anticyclonic mesovortex in a near-equatorial mesoscale convective system during TOGA COARE. Mon. Wea. Rev.,127, 1145–1156.

  • ——, P. Amayenc, G. Scialom, and J. Testud, 1987: A tropical squall line observed during the COPT81 experiment in West Africa. Part I: Kinematic structure inferred from dual-Doppler data. Mon. Wea. Rev.,115, 670–694.

  • Cuxart J., P. Bougeault, and J. L. Redelsperger, 2000: A turbulent scheme allowing for mesoscale and large-eddy simulations. Quart. J. Roy. Meteor. Soc.,126, 1–30.

  • Gao, K., D.-L. Zhang, M. W. Moncrieff, and H.-R. Cho, 1990: Mesoscale momentum budget in a midlatitude squall line: A numerical study. Mon. Wea. Rev.,118, 1011–1028.

  • Hildebrand, P. H., 1998: Shear parallel moist convection over tropical ocean: A case study from February 1993 TOGA COARE. Mon. Wea. Rev.,126, 1952–1976.

  • Hitschfeld, W., and J. Bordan, 1954: Errors inherent in the radar measurement of rainfall at attenuating wavelengths. J. Meteor.,11, 58–67.

  • Houze, R. A., Jr., 1977: Structure and dynamics of tropical squall line system. Mon. Wea. Rev.,105, 1540–1567.

  • Jorgensen, D. P., M. A. LeMone, and S. B. Trier, 1997: Structure and evolution of the 22 February 1993 TOGA COARE squall line: Aircraft observations of precipitation, circulation, and surface energy fluxes. J. Atmos. Sci.,54, 1961–1985.

  • Kessler, E., 1969: On the Distribution and Continuity of Water Substance in Atmospheric Circulations. Meteor. Monogr., No. 32, Amer. Meteor. Soc., 84 pp.

  • Kingsmill, D. E., and R. A. Houze, 1999: Kinematic characteristics of air flowing into and out of precipitating convection over the west Pacific warm pool: An airborne Doppler radar survey. Quart. J. Roy. Meteor. Soc.,125, 1165–1207.

  • Klemp, J. B., R. B. Wilhlemson, and P. S. Ray, 1981: Observed and numerically simulated structure of a mature supercell thunderstorm. J. Atmos. Sci.,38, 1558–1580.

  • Lafore, J. P., and M. W. Moncrieff, 1989: A numerical investigation of the organization and interaction of the convective and stratiform regions of a tropical squall line. J. Atmos. Sci.,46, 521–544.

  • ——, J. L. Redelsperger, and G. Jaubert, 1988: Comparison between a three-dimensional simulation and Doppler radar data of a tropical squall line: Transport of mass, momentum, heat and moisture. J. Atmos. Sci.,45, 3483–3500.

  • ——, and Coauthors, 1998: The Meso-NH atmospheric simulation system. Part I: Adiabatic formulation and control simulations. Ann. Geophys.,16, 90–109.

  • LeMone, M. A., E. J. Zipser, and S. B. Trier, 1998: The role of environmental shear and CAPE in determining the structure and evolution of mesoscale convective systems during TOGA-COARE. J. Atmos. Sci.,55, 3493–3564.

  • Lewis, S. A., M. A. LeMone, and D. P. Jorgensen, 1998: Evolution and dynamics of a late-stage squall line that occurred on 20 February 1993 during TOGA COARE. Mon. Wea. Rev.,126, 3189–3212.

  • Lipps, F. B., and R. S. Helmer, 1982: A scale analysis of deep moist convection and some related numerical calculations. J. Atmos. Sci.,39, 2192–2210.

  • Moncrieff, M. W., 1992: Organized convective systems: Archetypal dynamical models, mass and momentum flux theory, and parametrization. Quart. J. Roy. Meteor. Soc.,118, 819–850.

  • Montmerle, T., 1998: Validation et initialisation d’un modèle tri-dimentionnel méso-échelle non-hydrostatique par des données expérimentales issues de TOGA-COARE. Ph.D. thesis, Université Pierre and Marie Curie (Paris VI), 246 pp. [Available from Centre d’Étude des Environnements Terrestre et Planétaires, Velizy, France.].

  • ——, and Y. Lemaître, 1998: Three-dimensional variational data analysis to retrieve thermodynamical fields from various nested wind measurements. J. Atmos. Oceanic Technol.,15, 640–665.

  • Oury, S., J. Testud, and V. Marécal, 1999: Estimate of precipitation from the dual-beam airborne radar in TOGA COARE. Part I: The K–Z relationships derived from stereo- and quad-beam analysis. J. Appl. Meteor.,38, 156–174.

  • Pinty, J. P., and P. Jabouille, 1998: A mixed-phased cloud parameterization for use in a mesoscale non-hydrostatic model: Simulation of a squall line and of orographic precipitations. Preprints, Conf. on Cloud Physics, Everett, WA, Amer. Meteor. Soc., 217–220.

  • Protat, A., Y. Lemaître, and G. Scialom, 1998: Thermodynamic analytical fields from Doppler radar data by means of the MANDOP analysis. Quart. J. Roy. Meteor. Soc.,124, 1633–1669.

  • Redelsperger, J.-L., and G. Sommeria, 1986: Three-dimensional simulation of a convective storm: Sensitivity studies on subgrid parameterization and spatial resolution. J. Atmos. Sci.,43, 2619–2635.

  • ——, and Coauthors, 1999: Review on convection in TOGA-COARE. Proceedings of the Conference on TOGA-COARE, WMO/TD-No. 940, 16–42.

  • ——, and Coauthors, 2000: A GCSS model intercomparison for a tropical squall line observed during TOGA-COARE. Part I: Cloud resolving models. Quart. J. Roy. Meteor. Soc.,126, 823–864.

  • Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci.,45, 463–485.

  • Roux, F., 1998: The oceanic mesoscale convective system observed with airborne Doppler radars on 9 February 1993 during TOGA-COARE: Structure, evolution and budgets. Quart. J. Roy. Meteor. Soc.,124, 585–614.

  • Scialom, G., and Y. Lemaître, 1990: A new analysis for the retrieval of three-dimensional mesoscale wind fields from multiple Doppler radar. J. Atmos. Oceanic Technol.,7, 640–665.

  • Testud, J., and S. Oury, 1997: An algorithm to correct weather radar data for attenuation. C. R. Acad. Sci. Paris,324 (II.a), 705–710.

  • Thorpe, A. J., M. J. Miller, and M. W. Moncrieff, 1982: Two-dimensional convection in nonconstant shear: A model of mid-latitude squall lines. Quart. J. Roy. Meteor. Soc.,108, 739–762.

  • Trier, S. B., W. C. Skamarock, M. A. LeMone, D. B. Parsons, and D. P. Jorgensen, 1996: Structure and evolution of the 22 February 1993 TOGA-COARE squall line: Numerical simulation. J. Atmos. Sci.,53, 2861–2886.

  • ——, ——, and ——, 1997: Structure and evolution of the 22 February 1993 TOGA COARE squall line: Organization mechanisms inferred from numerical simulations. J. Atmos. Sci.,54, 386–407.

  • ——, M. A. LeMone, and W. C. Skamarock, 1998: Effect of three-dimensional structure on the stormwide horizontal accelerations and momentum budget of a simulated squall line. Mon. Wea. Rev.,126, 2580–2598.

  • Webster, P. J., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Soc.,73, 1377–1416.

  • Zipser, E. J., 1977: Mesoscale and convective scale downdrafts as distinct components of squall-line circulation. Mon. Wea. Rev.,105, 1568–1589.

Fig. 1.
Fig. 1.

Horizontal cross section of reflectivity (dBZ) obtained from the LF radar of the P-3 N42 between 2215 and 2230 UTC. The domain analysis is 200 × 200 km2 (Guadalcanal Island is depicted by intense ground echoes east of the MCS and the flight track is in dashed line). Maximum measured reflectivity is 42 dBZ.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 2.
Fig. 2.

Initial hodograph in the absolute and the relative framework. Arrows display the horizontal preconvective circulation in the system moving frame. The direction of the initial line direction is drawn for the three simulations.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 3.
Fig. 3.

Horizontal cross sections of the simulated reflectivity (dBZ) at 1 km MSL deduced from an empirical formulation for ice every hour from 2 h of simulation. Shading intervals correspond to a 10-dBZ variation in the precipitation field, and the dense thin lines represent reflectivity greater than 40 dBZ.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 4.
Fig. 4.

Horizontal cross section of relative wind field (arrows) and vertical vorticity (contoured) at 2 km MSL after 5 h, 30 min of simulation for ice within a 100 × 130 km2 subdomain. Arrows are plotted every two grid points, with vector length between these two grid points corresponding to 6 m s−1. The vertical vorticity contour increment is 1 × 10−3 s−1, positive values with dark shading starting at 1 × 10−3 s−1, and negative values with light shadings starting at −1 × 10−3 s−1.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 5.
Fig. 5.

Horizontal cross sections at 2 km MSL of reflectivity (contoured) and relative wind field (arrows every grid points) for (a) ice after 4 h of simulation and (b) leg A. The horizontal domain is 63 × 63 km2 in both case. Grayscales are shown to the right of each panel.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 6.
Fig. 6.

As Fig. 5 but for (a) ice after 5 h, 30 min of simulation and (b) leg B.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 7.
Fig. 7.

Vertical cross sections of averaged vertical velocity w2d, of averaged line-normal velocity u2d and of averaged line-parallel velocity υ2d for, respectively, (a), (c), (e) ice after 4 h of simulation and (b), (d), (f) observation during leg A. These values are averaged within the boxes defined in Fig. 5a for ice and Fig. 5b for the radar. The contour intervals are 0.5, 2, and 1 m s−1 for w2d, u2d, and υ2d, respectively. Dashed lines present negative values.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 8.
Fig. 8.

Vertical cross sections of the square root of the spatial variance σ2dw of the averaged vertical velocities w2d plotted in Figs. 7a and 7b, for (a) ice and (b) the observation. Isocontours are plotted every 0.5 m s−1.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 9.
Fig. 9.

Vertical cross sections of averaged vertical velocity w2d and averaged line-normal velocity u2d relative to the system for (a), (c) ice after 5 h, 30 min of simulation and (b), (d) the observations during leg B, respectively. Contours are plotted every 1 and 2 m s−1, respectively; negative values are dashed. These values are averaged within the boxes defined in Fig. 6a for ice and Fig. 6b for the radar, across the northern part of the line during the bow stage.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 10.
Fig. 10.

As for Fig. 8 but relatively to Figs. 9a and 9b, respectively.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 11.
Fig. 11.

Vertical cross sections of averaged line-normal horizontal momentum flux ρuw2d in the system moving frame for the radar observations during (a) leg A and (c) leg B, and ice after (b) 4 h and (d) 5 h, 30 min of simulation (contour interval of 8 kg m−1 s−2, negative values dashed). The values plotted in (a), (b) are averaged within the boxes defined in Fig. 5a for ice and Fig. 5b for the radar, and those plotted in (c), (d) within boxes drawn in Figs. 6a and 6b, respectively.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 12.
Fig. 12.

Profiles of (a) mean relative line-normal u-component wind speed, (b) mean relative line-parallel υ-component wind speed, and (c) mean vertical velocity. The bold continuous line corresponds to ice after 4 h of simulation; the bold dashed line to ice after 5 h, 30 min;the dashed line to leg A; and the dotted line to leg B. For the simulation, only the points situated in the precipitating area of E1 are taken into account.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 13.
Fig. 13.

Vertical profiles (a), (b) of line-normal u-momentum flux and (c), (d) of line-parallel υ-momentum flux in the system moving frame for ice after 5 h, 30 min of simulation and leg B, respectively. For the simulation, only the points situated in the precipitating area of E1 are taken into account.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 14.
Fig. 14.

Horizontal cross sections of the virtual potential temperature perturbation θ*vl at (a) 0.5 and (c) 4.3 km MSL with contour interval of 0.5 K, and of the pressure perturbation P* at (b) 0.2 and (d) 2.5 km MSL with contour interval of 10 Pa for ice after 5 h of simulation. Negative values are dashed. Arrows represent the relative wind circulation at the same levels every two grid points, with the vector length of these two grid box equivalent to 12 m s−1.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 15.
Fig. 15.

Vertical cross sections across E1 of simulated (a) averaged vertical velocity w2d with contour interval of 0.5 m s−1, (b) averaged line-normal velocity u2d (2 m s−1), (c) averaged cloud virtual potential temperature perturbation θ*vl (0.5 K), and (d) averaged pressure perturbation P* (10 Pa). These values are averaged within the box A defined in Fig. 16a. The positions of the horizontal cross sections plotted in Fig. 16 are shown in (c) and (d).

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 16.
Fig. 16.

As Fig. 15 but for box B plotted in Fig. 16a across E2.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 17.
Fig. 17.

Horizontal cross sections of (a) the virtual cloud potential temperature perturbation θ*c at 4.5 km MSL (0.2 K) and (b) the pressure perturbation P* at 2.5 km MSL (10 Pa) observed during leg C. Arrows represent the relative wind circulation at the same levels every two grid points.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 18.
Fig. 18.

As in Fig. 3 but for the warm simulation.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 19.
Fig. 19.

As in Fig. 3 but for the ice30 simulation. The simulation has been stopped after 6 h, as the simulated system has reached the domain frontiers.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Fig. 20.
Fig. 20.

Evolution at 3, 4, and 5 h of the synthesis of the squall-line system simulated by the warm, ice, and ice30 experiments for first, second, and third rows, respectively. Active part of the cold pool leading edge is drawn as a cold front. Convective part is identified by the mesolow at 3 km (gray area) and the trailing stratiform precipitation by the highest density air at the 0°C level (hatched area). See text for details on the synthesis definition.

Citation: Monthly Weather Review 128, 11; 10.1175/1520-0493(2001)129<3709:ATSLOD>2.0.CO;2

Table 1.

Mean precipitation rates over 1-h periods for all simulations (cm day−1).

Table 1.
Save