1. Introduction
One of the main objectives of the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE; Webster and Lukas 1992) is to understand the processes that control and organize convection in the warm pool region at the different scales of motion (from synoptic to convective scale) and the interaction of deep convection with the larger-scale atmospheric circulation. In order to address these broad TOGA COARE objectives, the scientific motivations of this paper are (i) to document the precipitation, and kinematic and thermodynamic structure of one of the most impressive mesoscale convective systems (MCSs) observed during the TOGA COARE field phase (the 15 Dec 1992 cloud cluster) at both the mesoscale and convective scale, and (ii) to examine the scale interactions involved in the internal organization of this MCS. To study these scale interactions, a detailed comparison of the precipitation, and kinematic and thermodynamic fields characteristic of the synoptic-scale, mesoscale, and convective-scale circulations is conducted. The synoptic-scale mechanisms were examined Protat and Lemaître (2001, denoted Part I hereafter). This synoptic-scale study was performed using the European Centre for Medium-Range Weather Forecasts (ECMWF) model outputs (from which diagnostic parameters have been derived) and Geosynchronous Meteorological Satellite-4 fields of brightness temperature in the infrared channel (providing a picture of the convective activity every 15 min).
Let us first briefly summarize the results of Part I. From the ECMWF analyses at 0600 and 1200 UTC (all times are UTC), a particular region of the synoptic-scale domain was identified as strongly favorable to the development of deep convection, with (i) a strong low-level convergence at 850 hPa between a northwesterly flow (corresponding to the beginning of a westerly wind burst period) and a southwesterly flow, (ii) strong convective available potential energy (CAPE) from 1200 to 1800, and (iii) strong production of CAPE at 1200, when the MCS starts intensifying. This favorable region actually corresponds to the area where the studied MCS (hereafter referred to as system S) developed and intensified, which indicates that the synoptic-scale circulation and dynamic features provided the initial conditions for the formation and general evolution of S. The synoptic-scale dynamic and thermodynamic mechanisms leading to the observed initiation and general evolution of S were therefore examined in detail. It was shown that initially, north of S, an easterly equatorial jet at 500 hPa generated positive ζ and potential vorticity (PV) anomalies on its southern border by 0600 UTC, in response to the important horizontal wind shear in this region. Then, these ζ and PV anomalies propagated westward with the flow at 500 hPa, leading to the generation of an anticyclonic circulation C, vertically extending from 850 up to 250 hPa and with a peak magnitude at 1200 at 500 hPa and at 1800 at 700 hPa. Vortex stretching and vertical tilting of the horizontally oriented vortex tubes along the y axis were evidenced as the main processes involved in the intensification of this circulation C at 700 hPa from 0600 to 1800. This intensification of vertical vorticity at 700 hPa accelerated the northwesterly flow at 850 hPa from 0600 to 1800, which promoted in turn a strengthening of the synoptic-scale convergence line responsible for the intensification of S. From 1800, the magnitude of vertical vorticity associated with circulation C decreased, which diminished the synoptic-scale low-level convergence, and CAPE was not produced anymore in the S area. These two major changes in the synoptic-scale configuration associated with system S led to its decay. Again, this showed that the synoptic-scale circulation governed both the intensification and decaying periods of this MCS.
In the present paper, the dynamic and thermodynamic characteristics of the mesoscale and convective-scale circulations within S are derived from airborne Doppler radar data. These data were collected from 1700 to 2045 UTC by the two radars that operated aboard the National Oceanic and Atmospheric Administration P3-42 and P3-43 aircraft. The objectives of this study are to (i) come to a better understanding of the morphological structure and internal organization of the studied system at these scales of motion, and (ii) diagnose the downscale interaction processes (i.e., from synoptic scale to convective scale) involved in the internal organization of the system.
In section 2, the Doppler radar installed on the P3 research aircraft is briefly described, as well as the analysis methods used to access the mesoscale and convective-scale dynamic characteristics. In section 3, the mesoscale dynamic and thermodynamic characteristics are interpreted, and comparison is made with the larger-scale atmospheric context. Section 4 presents the retrieved dynamics and thermodynamics within selected convective-scale domains, along with a detailed comparison between the convective-scale, mesoscale, and synoptic-scale features. Finally, in section 5, concluding remarks are given.
2. Airborne Doppler radars and description of analysis methods
a. Description of the radars and antenna scanning methodology
The Doppler radar installed on the tails of the P3-42 and P3-43 aircraft is described in detail in Jorgensen et al. (1983) and Hildebrand and Mueller (1985). The radar is an X-band Doppler radar (3.2-cm wavelength). The scanning methodology is the fore–aft scanning technique, which collects two conical scans of data, one pointing ∼25° forward from a plane normal to the flight track and one scan pointing aft ∼25°. Using this scanning procedure, a three-dimensional region of space surrounding the aircraft track is swept out. Along the beams, the range resolution of data is 75 m for ranges less than 19.2 km, 150 m for ranges between 19.2 and 38.4 km, and 300 m for ranges between 38.4 and 76.8 km. As with most TOGA COARE airborne Doppler radar sampling, the flight strategy used on this day was to fly parallel to the leading convective line of the system, using flight paths long enough that the fore and aft scans fully sampled the storm. The tracks flown on 15 December are described in detail in section 3.
A lower-fuselage non-Doppler C-band radar (5-cm wavelength) was installed as well on both P3 aircraft. This radar collects within a range of about 300-km quasi-horizontal scans around the altitude of the aircraft. However, it must be noted that the wide beamwidth in the vertical (around 4°) corresponds to a 4-km vertical resolution of the measurements at 60-km range. The precipitation field derived from these measurements should therefore be considered as an overview of the precipitating activity.
b. Retrieval of the 3D wind field on the mesoscale and convective scale
In this study, we will analyze the 3D airflow structure on the mesoscale and convective scale. The three-dimensional wind field is retrieved in the present study using the Multiple Analytical Doppler (MANDOP) analysis [Scialom and Lemaître (1990, referred to as SL in the following) see also Dou et al. (1996) for the extension of the method to the processing of airborne Doppler radar data]. This analysis, based on a variational concept, lies upon the expression of the three wind components as products of expansions in series of orthonormal functions along the three axes of the coordinate system. The orthonormal functions (Legendre polynomials in the present case) are expanded up to a given order that partly determines the scale of motion resolved by the analysis. Then, the analytical form of the radial wind is variationally adjusted to the radial winds measured by the Doppler radar. The anelastic approximation of the airmass continuity equation is introduced in the minimization procedure, as well as the condition w = 0 at ground level. These physical constraints are mandatory to access an accurate estimate of the vertical wind component.
As already discussed in SL, Dou et al. (1996), and Protat et al. (1998), the analytical approach used in this analysis benefits from several advantages. In the context of this multiscale approach, the main advantage of the analytical formulation results in the natural filtering and interpolating properties of such methods. The cutoff wavelength λc2 of the analytical formulation in a given direction x is a function of the grid resolution Δx, the domain size Dx, and the order of expansion n of the analytical functions. As explained in SL, no structure with a wavelength smaller than λc1 = 2Δx can be resolved according to Shannon's theorem. On the other hand, a wave along each axis defined by a Legendre polynomial expanded up to order n reaches zero at most (n − 1) times along the domain size, which implies a minimal wavelength λc2 of 2Dx/(n − 1). Parameters n, Dx, and Δx must therefore be chosen in the three directions of space such that λc1 < λc2. As shown in SL, the corresponding minimum observable scale is lc = 0.26λc2. The airborne Doppler radar data have proven to be of a spatial and temporal resolution compatible with convective-scale studies. In order to perform a 3D wind field retrieval on the mesoscale, the MANDOP analysis can be applied with a smaller order of expansion than on the convective scale, which will filter out the convective-scale structures resolved in the measurements. As an example, for a domain size Dx = 300 km along a direction x and an order of expansion n = 5 for the analytical functions, the minimum observable scale is of about 40 km, which will filter out the convective-scale structures.
c. Retrieval of the pressure and temperature perturbations
From the analytical representation of the 3D wind field, the pressure and “virtual cloud” potential temperature1 perturbations are retrieved under an analytical form by introducing the analytical form of the 3D wind field obtained by the MANDOP analysis in the nondissipative and stationary first-order anelastic approximation of the momentum equation (Protat et al. 1998). This analysis differs from that described in previous studies (e.g., Gal-Chen 1978; Hane and Ray 1985), in that the input is the analytical 3D wind field, not a 3D wind field numerically retrieved at each grid point of a regular grid mesh. In this method, the pressure and temperature perturbations are obtained under an analytical form, and the coefficients of expansion of the wind are those of the pressure and temperature perturbations as well. These perturbations are actually deviations from a reference state that is constant horizontally but varying with height (Protat et al. 1998). In order to access the full 3D pressure and virtual cloud potential temperature fields, a boundary condition provided by an in situ measurement (e.g., radiosounding, dropsounding) must be introduced in the method. Such a measurement is unfortunately not available around the studied system, which implies that vertical cross sections of dynamic perturbations cannot be interpreted. This also means that a −1 K perturbation does not mean that the flow is 1 K colder than the environment, because of the unknown horizontal constant that should be added to the perturbation. Then, only the horizontal gradients of dynamic perturbations will be interpreted throughout this paper.
The horizontal gradients of pressure and virtual cloud potential temperature analyzed in the present study have been cross-validated (Protat et al. 1998, see their Fig. 11) with those obtained using the single-beam airborne velocity azimuth display (SAVAD) analysis2 (Protat et al. 1997) within the S2 system. A very good agreement was obtained between the two independent estimates. This result indicated that the retrieved thermodynamic horizontal gradients could be reasonably trusted.
d. Air and water parcel trajectory analysis
The analytical representation of the 3D wind field can be used to recover three-dimensional trajectories of air and water parcels within a given retrieval domain, provided that the 3D wind field can be assumed as relatively steady state. In fact, from a given initial position of an air parcel in the retrieval domain, each new location of this parcel at a given time T can be calculated from its previous location at time T − ΔT using the analytical 3D wind field and the time interval ΔT (which can be chosen as small as needed since the wind field is analytically expressed) so as to determine the displacement vector of the parcel. This procedure allows for obtaining the displacement vector anywhere in the domain, since the analytical form can be calculated at any intermediate location and not only at given grid points.
The water parcel trajectories can be obtained by adding an estimate of the terminal fall velocity of the hydrometeors to the retrieved vertical component of the wind. This terminal fall velocity of the hydrometeors is generally estimated using a empirical relationship as a function of the radar reflectivity. In the present case, vertical profiles of terminal fall velocity have been recovered within the 15 December 1992 MCS, using the SAVAD analysis (Protat et al. 1997) described just above. This estimate derived from radar measurements is used in the following.
3. Mesoscale analysis of the 15 December 1992 MCS
The mesoscale description of the MCS remains a crucial problem despite the significant amount of instrumental means deployed during the TOGA COARE Intensive Observing Period (IOP). In fact, the radiosounding network is too coarse to recover the mesoscale flows, especially on this day of the IOP, characterized by poor radiosounding coverage and frequency. Therefore, in order to access the mesoscale characteristics associated with the selected MCS, it is chosen in the course of this study to process the airborne Doppler radar data in such a way that the largest retrieval zone (typically 250 × 250 km2) is covered.
Figure 1a shows a time-composite reflectivity map of the 15 December 1992 MCS for the whole aircraft mission (1700–2045 UTC), as sampled by the C-band radar mounted on the P3-42 aircraft. For each 5-km grid element of this map, the highest reflectivity value collected over the sampling time is retained. Let us recall that such a picture should be considered as an overview of the precipitation structure of the MCS (see section 2a). The trajectories of the P3-42 and P3-43 aircraft are the white and black lines, respectively. This overview of the precipitation field indicates that the MCS consists of two main areas of strong precipitation (denoted as S1 and S2, respectively, for the northwesternmost and southeasternmost subsystems). Their detailed structure will be discussed later using the high-resolution reflectivity data from the tail Doppler radars.
Figure 1a indicates that the P3-42 aircraft, which started collecting data at 1700 UTC, systematically flew at the rear of (i.e., north of) and along an almost linear convective structure associated with S1 (at a flight altitude of about 4 km). This sampling strategy enabled collection of data within S1 from 1700 to 1830 UTC. Meanwhile, the P3-43 aircraft systematically flew ahead of (i.e., south of) and along the linear convective structure associated with S1, at a flight altitude ranging from 200 m to 1 km. This scanning strategy was then repeated from 1915 to 2045 UTC to collect data within the second convective structure S2 (southeast of S1).
In order to access the mesoscale characteristics associated with this MCS in the largest retrieval domain, Doppler radar data collected during the time interval 1700–2045 UTC must be used. Nevertheless, this cumulation of data during this relatively large interval is possible only if the MCS can be assumed as reasonably steady state. The steady-state assumption is typically evaluated through inspection of the temporal evolution of the most active part of the convective systems, as revealed by the tail-Doppler radar observations. This evaluation is possible in the present case, since the research aircraft flew several straight-flight tracks parallel to each other within both subsystems. Figures 1b and 1c show the reflectivity fields at the 1-km altitude derived from observations collected 1 h apart within S1. It is seen that the general organization of precipitation looks reasonably the same in both figures, except in the stratiform region of the system where small differences appear. It also shows that the convective structures remain fairly fixed in location. These results indicate that a “quasi-” steady-state assumption can be reasonably assumed in the present case, since convection is not found to significantly change within S1. The same conclusions apply to S2 (not shown). In addition, it will be shown that the structure of the precipitation field on the mesoscale, which is obtained using a long time period of observations, is reasonably similar to that of the precipitation field derived from a much shorter time period of 10–12 min within selected precipitating areas. This validates indirectly the steady-state assumption.
In section 3a, a detailed description of the dynamic characteristics of the mesoscale circulation is presented, and comparison is made with the synoptic-scale circulation so as to examine the possible interactions between these scales of motion. The interpretation of the internal organization of the system using air parcel trajectory analysis on mesoscale is presented in section 3b. Momentum fluxes calculations are conducted in section 3c, in order to estimate the respective contribution of the synoptic-scale and mesoscale circulations to the vertical transport of horizontal momentum on mesoscale. Finally, the thermodynamic characteristics of the mesoscale flow are analyzed in section 3d.
a. Mesoscale dynamics
The 3D wind and precipitation fields are given in Fig. 2 at 0.5-, 1.5-, 3-, 5-, 8-, and 10-km altitude. The domain size is 270 × 170 km2, and the orders of expansion of the analytical functions in the x and y directions are 6 and 4, respectively. Following the calculations in section 2b, this corresponds to a minimum observable scale of about 30 km. The convective structures are therefore filtered out by the method. The reflectivity fields of Fig. 2 show a more detailed morphological structure than that seen in Fig. 1a. The existence of two subsystems, S1 and S2, within S appears more clearly. These subsystems are located in the northwestern and southeastern areas of the mesoscale retrieval domain, respectively. Subsystem S1 is broadly characterized by a typical structure of squall line (Fig. 2b), with (i) a very active bow-shaped precipitation area characterized by reflectivities stronger than 40 dBZ, denoted LS1 in what follows (see dashed line in Fig. 2a), oriented along a southwest–northeast axis, (ii) a region of quite homogeneous precipitation on the northwestern part of S1 (the stratiform part), within which exists a more intense precipitating area (reflectivities stronger than 30 dBZ) that will be discussed later on, and (iii) a region of weak reflectivity echoes, strongly reminiscent of the “transition zone” of weaker precipitation commonly observed within tropical and midlatitude squall lines (e.g., Lemaître and Testud 1986; Biggerstaff and Houze 1993; Braun and Houze 1994). Note that this transition region in the present case is clearly shifted to the northeast of LS1, instead of lying all along the rear part of the leading edge as is often observed within squall lines. This latter characteristic gives to S1 a three-dimensional structure. Subsystem S2 consists of a strong leading convective line, denoted LS2 in the following (see dashed line in Fig. 2a), and a stratiform region (much less developed than that of S1) at the rear of LS2.
Horizontal flows on the mesoscale have been compared with the synoptic-scale ECMWF wind analysis (not shown). It has been clearly found that they were comparing fairly well at every altitude, but that a spatial shift of about 200 km appeared between the mesoscale and synoptic-scale circulations. This shift is particularly obvious for the area of convergence between the westerly and southwesterly flows at the 1.5-km altitude. Indeed, this convergence area is located within the mesoscale domain (Fig. 2b), according to the radar observations, while inspection of the ECMWF analysis of the horizontal wind at 850 hPa shows that this convergence area is located about 200 km west of the mesoscale domain (see Fig. 9c of Part I and the location of the mesoscale box in the analysis domain). This shift is not really surprising. First, there is a temporal shift between the analysis and the radar observations, and more importantly the scale resolved by the ECMWF model (around 200–300 km) and by the mesoscale 3D wind field (30 km; see section 2b) are very different. This spatial shift is therefore of a scale smaller than the scale resolved by the ECMWF model. In addition, it will be shown in section 4 that condensation and evaporation processes play an important role in the dynamic of the MCS on the mesoscale. These local processes (with respect to the synoptic scale) are not resolved by the ECMWF model, but are parameterized. An underestimation of the effect of these mesoscale processes on the local intensification of the synoptic-scale flows would certainly lead to a spatial shift. In fact, as shown in section 4, evaporation is responsible for a significant cooling of the northwesterly rear inflow. The resulting density current is propagating toward the southeast/east. This displacement, if not well represented by the parameterizations in the model, could explain why the convergence line is shifted in the ECMWF analysis slightly west of where it is observed by the airborne radars.
We now turn to a detailed description of the mesoscale airflow circulation. The retrieved 3D wind field at the 1.5-km altitude (Fig. 2b) is characterized by a relatively homogeneous southwesterly flow, within which two regions of strong convergence appear, coinciding with the leading convective lines of both subsystems. This mesoscale circulation at 1.5-km altitude is consistent with the synoptic-scale ECMWF analysis of the horizontal wind at the 850-hPa level (see Fig. 9b of Part I). The convergence zone associated with S1 is better defined than that associated with S2, with a convergence between a southwesterly flow ahead of LS1 and a relatively homogeneous westerly flow at the rear part of LS1, within the stratiform part of S1. This westerly flow extends vertically from the ground up to 3-km altitude (see Figs. 2a–d). Likewise, strong convergence is found within S2 between a southwesterly flow ahead of LS2, and a westerly flow at the rear (within the stratiform part). This westerly flow is however much more shallow at the rear of LS2 than at the rear of LS1, roughly less than 1.5 km deep (see again Figs. 2a–d). This suggests that this westerly flow at the rear of S2 results from processes (and in particular microphysical processes) within the stratiform part of S2, while the much deeper westerly flow associated with S1 may be linked to the synoptic-scale circulation. This difference in the low-level rear inflows will be studied in section 4 on the convective scale.
At 3-km altitude (Fig. 2c), the westerly flow still generates convergence within S1, while convergence is clearly less intense within S2, due to a veering of the flow with height from westerly to southwesterly between 1- and 3-km altitude. This relatively homogeneous southwesterly flow associated with S2 at 3-km altitude appears around the 5-km altitude within S1 (Fig. 2d). In the upper part of the troposphere, strong divergence occurs within S1. The horizontal wind is much weaker in magnitude and more perturbed in the 7–10-km layer (Figs. 2e,f). Strongly divergent flows are found, with northwesterlies and south-southeasterlies in the southeastern and northwestern parts of the domain, respectively. This strong divergence signature at higher levels is found as well at the synoptic scale within and in the vicinity of the mesoscale retrieval domain, as shown by the ECMWF analysis at the 250-hPa level (Fig. 3).
Figure 4 shows the hodograph of the horizontal mean wind in the mesoscale domain (profile 1) and individual hodographs of the retrieved 3D wind field within the different precipitating entities of the mesoscale domain. This figure shows that the orientation of the two leading convective lines, LS1 and LS2, is almost perpendicular to the mean vertical shear of horizontal wind between 0- and 3-km altitude (see profile 1). This behavior would put this convective system in the “shear perpendicular” category defined by LeMone et al. (1998) for the TOGA COARE convective systems, and in the “fast movers” category of Barnes and Sieckman (1984), Alexander and Young (1992), and Keenan and Carbone (1992). However, these categories all correspond to a relatively fast propagation speed of the systems (roughly >7 m s−1, or close to the low-level wind maximum, depending on the classification), which is not the case here since S1 and S2 remain fairly fixed in location during the sampling time of these entities by the airborne radars. LeMone et al. (1998) categorized therefore this 15 December system as one of the two “unclassifiable” systems out of the 20 cases examined. This may be due to the presence of a significant midlevel vertical shear associated with the inflow ahead of the leading convective lines, as shown by the local hodograph taken ahead of LS2 (profile 2 in Fig. 4). This midlevel shear is 180° to the low-level shear. With a shear-perpendicular configuration in the low levels and a reversed shear at midlevels, the low-level shear should determine the orientation of the primary convective lines (which is indeed the case here), and the midlevel shear should lead to the formation of secondary bands extending rearward from the primary lines in the later stages of evolution (LeMone et al. 1998, and earlier investigators). Such secondary bands are not observed at the time the radars sampled the MCS.
The hodographs performed within the stratiform regions of S1 and S2 mainly confirm that the westerly flow at the rear of S1 (profile 4) is deeper than at the rear of S2 (profile 3). The hodograph representative of the horizontal wind within the stratiform part of S1 (profile 4 of Fig. 4) shows that the orientation of the most active precipitating area within this stratiform part is oriented along the horizontal airflow at 5-km altitude, and parallel to the vertical wind shear between 5- and 7-km altitudes. This latter observation is consistent with the “quasi-universal” conceptual scheme of tropical squall line dynamics, with precipitation in ice phase (which, in this case, would be generated within the southwestern part of LS1) evacuated toward the rear part of the leading convective line by a midtropospheric front-to-rear outflow (e.g., Chong et al. 1987; Fovell and Ogura 1988; Houze 1993). The existence of such a midtropospheric outflow and its potential importance in the mesoscale dynamics of the system is discussed in what follows using vertical cross sections and air parcel trajectory analysis.
Such a vertical cross section of reflectivity and wind along the plane of the cross section is given in Fig. 5a. Contrary to the other TOGA COARE case studies documented by Jorgensen et al. (1997) and Hildebrand (1998), the vertical structure of precipitation does not exhibit the rearward tilting typical of the leading convective lines of quasi-2D squall lines. Figure 5a shows that upward vertical motions are initiated in the lowest tropospheric layers within S2, and at about the 2–3-km altitude within S1. This difference is probably due to the different depth of the low-level westerly rear inflow of S1 and S2, respectively. It is indeed clearly seen from Fig. 5b, presenting the isotaches of the u component of the wind in the same vertical cross section as Fig. 5a, that the low-level westerly flow is progressively deepening to the northwest (left). The vertical cross section of Fig. 5a shows that the deep westerly rear inflow associated with S1 is capped by the 0°C isotherm (around the 4.5-km altitude in the present case). This capping effect of the 0°C stable layer has also been noted in the TOGA COARE systems studied by Jorgensen et al. (1997) and Hildebrand (1998). This westerly rear inflow, which results from the synoptic-scale circulations, is found to be accelerated within both stratiform parts of S1 and S2 (see Fig. 5b). Previous investigations of tropical squall lines documented the existence of a pressure deficit in the low levels, at the rear of the leading convective line. This pressure deficit is induced hydrostatically by the rearward spreading of warm and moist air from the updraft and dynamically by the interaction of the updraft and the ambient wind shear (e.g., Rotunno and Klemp 1982; Weisman 1992; Szeto and Cho 1994). This local pressure deficit in the stratiform part of the squall line induces a negative horizontal pressure gradient from the rear part of the stratiform area toward the rear of the leading convective line. According to the horizontal momentum equations, this negative horizontal pressure gradient must induce a local acceleration of the horizontal airflow in the direction of the horizontal pressure gradient force, acceleration well observed in the present case at the rear of LS1 and LS2 (Fig. 5b).
Maximum upward motions associated with LS1 and LS2 are located in the upper part of the troposphere (above 12 km from S1, and around 12 km for S2; see bold contours in Fig. 5a), with a peak magnitude of about 1.5 m s−1. This characteristic is also observed at the synoptic scale, with maximum upward motions located around the 8-km altitude at that scale, as shown by the vertical profile of vertical wind component deduced from the ECMWF analysis at the location of the mesoscale domain and at 1800 UTC (Fig. 6). This observation of an elevated vertical velocity maximum for tropical convection is in agreement with previous studies (e.g., Jorgensen et al. 1997; Hildebrand 1998; Roux 1998). The mesoscale updrafts associated with LS1 and LS2 are both diverging in the upper part of the troposphere (see Figs. 2e,f). This upper-level divergence signature produces a symmetric structure of the upper-level outflow, with a first part of the outflow that is advected rearward of the leading lines, and a second outflow that backs ahead of the leading lines [denoted as the “overturning updraft” in Lemone and Moncrieff (1994)]. Other vertical cross sections through the mesoscale domain (not shown) confirm this characteristic of the vertical structure of the updrafts, with maximum upward motions of about 1–2 m s−1 located at about 12–13-km altitude for S1, and about 11 km for S2.
It is noted that the frontward upper-level outflow was not observed in the 22 February 1993 fast-moving shear-perpendicular squall line observed during TOGA COARE (documented in Jorgensen et al. 1997), while it is clearly observed in the 18 February 1993 shear-parallel non-propagating rainbands observed during TOGA COARE (documented in Hildebrand 1998). The 15 December 1992 subsystems and the rainbands studied by Hildebrand (1998) seem therefore to share two characteristics: a fixed location and the existence of a frontward upper-level outflow.
b. Interpretation of the obtained mesoscale dynamics
As explained in section 2d, three-dimensional air parcel trajectories can be retrieved using the 3D wind field under a steady-state assumption. In order to reconstruct a general picture of the 3D airflow circulation on mesoscale, we choose initial positions ahead of the leading convective lines of S1 and S2 every 10 km in the horizontal and every 500 m in the vertical. The individual 3D air parcel trajectories retrieved are then subjectively grouped into “most representative” flows. The same procedure is applied to the rear part of the leading convective lines of S1 and S2. A conceptual scheme of the 3D airflow circulation associated with the studied MCS is then constructed from these individual most representative flows. This conceptual scheme is given in Fig. 7. Also drawn schematically in this figure are the two low-level flows representative of the larger-scale atmospheric context (see Part I), that is, a westerly flow linked to the beginning of a westerly wind burst period, overlayed by a southwesterly flow. In the mesoscale area, the synoptic-scale westerly flow is characterized by a southern border that progressively deepens toward the north. This slantwise slope is also found on the mesoscale, as discussed in section 3a (Fig. 5b). Figure 5b also shows that vertical variations are superimposed to this general deepening across the domain. These variations likely result from the local intensification of the synoptic-scale westerly flow, in response to the presence of S1 and S2, as discussed previously.
The air trajectory composite associated with S1 (Fig. 7) shows the existence of a mesoscale westerly flow denoted 1 in Fig. 7 (within the synoptic-scale westerly flow), which initially extends vertically from the ground up to 2-km altitude. This flow is oriented along the precipitation structure located in the southwestern part of LS1, slightly overruns the northernmost part of the westerly flow (denoted 2 in Fig. 7), and ends up in the 2–3.5-km layer in the rear part of the stratiform region. This northernmost part of the westerly rear inflow 2 is deeper than flow 1 (located in the 0–3.5-km layer), remains quasi-horizontal, and deflects horizontally to the northeast as it approaches the northeastern part of LS1. Within the synoptic-scale southwesterly flow, two mesoscale ascending flows (denoted 3 and 4 in Fig. 7) having distinct upper-level characteristics are identified. Both flows are located initially in the 1.5–4-km layer. As those flows rise up, they progressively deflect westward and eastward, respectively, and end up in the 6–14-km layer. Updraft 3 is oriented along the most intense precipitating region within the stratiform part of S1 (see the schematic representation of the precipitation field of Fig. 7 and the reflectivity field of Fig. 2a), while updraft 4 is oriented along the northeastern part of LS1. A part of each flow 3 and 4 backs toward the front side of the leading line LS1. These “overturning updrafts” are denoted 3′ and 4′ in Fig. 7 and are discussed in section 3a. Interestingly, the advection of precipitating particles within the overturning updraft 4′ may in particular explain the existence of an anvil cloud ahead of the northeastern part of LS1 [at around (x, y) = (60, 0) in Fig. 2a]. Finally, the last flow characteristic of the mesoscale circulation within S1 (denoted 5 in Fig. 7) is the southernmost part of the synoptic-scale westerly flow, located within the 0–1.5-km layer. This flow (5) slightly ascends along and just ahead of LS1 (rising up to the 2–2.5-km altitude in the northeastern part of LS1), indicative of a moderate convective instability.
The air trajectory analysis within S2 indicates that the mesoscale rear inflow 6 (located initially in the 0–1-km layer) within the synoptic-scale westerly flow and the mesoscale southwesterly flow 7 (whose initial depth is difficult to assess, since there is a lack of radar targets ahead of the leading line LS2) both contribute to the upward motion within S2. Also noteworthy is the existence of an overturning updraft (denoted 7′ in Fig. 7) backing ahead of the leading convective line LS2, as discussed previously. It is found that the air parcels within flow 7 systematically reach higher levels than those within flow 6, which reach heights less than 6–7 km (not shown). This stronger convective instability of the flow ahead of LS2 may be explained by the presence of a synoptic-scale area of CAPE production ahead of the MCS from 1800 UTC (see Part I). This result highlights again the major role played by the synoptic-scale context in the mesoscale organization of the subsystems. This significant role is examined quantitatively in what follows in terms of momentum fluxes on mesoscale and their mean and eddy contributions.
c. Momentum fluxes on mesoscale
Momentum fluxes are calculated in this section to examine more quantitatively the respective contribution of the synoptic-scale and mesoscale circulations to the vertical transport of horizontal momentum. Figure 8 shows the vertical profiles of mean wind components and averaged u- and υ-momentum fluxes (terms 〈ρ〉 〈uw〉 and 〈ρ〉 〈υw〉) computed over the whole mesoscale domain. Both momentum flux profiles exhibit positive values from the ground up to 10-km altitude (Figs. 8d,e), with a maximum at 3.5- and 5-km altitude, just above the level of maximum winds (Figs. 8a,b), with peak values of about 1.0 and 1.4 kg m−1 s−2, respectively for u and υ, which is consistent with the magnitudes observed by Caniaux et al. (1995) at the same scale. Positive fluxes in both directions (Figs. 8d,e), owing to positive wind components (Figs. 8a–c) from the ground up to 10-km altitude are typically encountered in the literature for both tropical and midlatitude squall lines (LeMone 1983; Lafore et al. 1988; Matejka and LeMone 1990; Lin et al. 1990; LeMone and Jorgensen 1991; Caniaux et al. 1995; Hane and Jorgensen 1995, and others), as well as in the recently published TOGA COARE study of the 18 February 1993 tropical rainbands (Hildebrand 1998). Above 10-km altitude, the υ-momentum flux is negative, owing to the contribution of the upshear-tilted updrafts (upshear with respect to low-level shear) above the 10-km altitude in the present case (see Fig. 5a) that change the sign of the υ component (Fig. 8b). This importance of the upshear-tilted updrafts in the vertical transport of momentum was also noted for the TOGA COARE rainbands studied by Hildebrand (1998), who found negative line-normal momentum fluxes in the upper levels associated with upshear-tilted updrafts. The cases with upshear-tilted updrafts were also noted by LeMone and Moncrieff (1994) as being the largest exception to upper-level positive momentum flux of all the convective systems they considered.
The “total” momentum flux (as defined by Hane and Jorgensen 1995) discussed previously may be decomposed as the sum of the mean and eddy components [e.g., for the u component 〈ρ〉 〈uw〉 = 〈ρ〉 (〈u〉〈w〉) + 〈u*w*〉) for u, where the star denotes the perturbations from the average]. For both components of the vertical fluxes of horizontal momentum, the mean fluxes contribute most strongly to the total flux (Figs. 8d,e). This indicates that the vertical transport of horizontal momentum is mostly carried out at a scale of motion larger than the mesoscale domain size (roughly 200 km). However, above 7.5-km altitude there is significant input to the total profile by eddy momentum transport for the u-momentum flux. The significant contribution of the eddy momentum transport above 7.5-km altitude is likely due to the strong upshear-tilted updraft eddies in the upper levels (Fig. 5a) in our case. This was noted as well by Hildebrand (1998) who computed such fluxes within the 18 February 1993 rainbands observed during TOGA COARE that were characterized by such updrafts as well.
d. Pressure and temperature perturbations on mesoscale
The analysis method used to recover the pressure and virtual cloud potential temperature perturbations from the analytical 3D wind field is described in section 2c. Figure 9 shows the
Figure 9b shows the existence of positive gradients of virtual cloud potential temperature from south to north, which progressively veer with height to positive gradients from east to west (Fig. 9c). This progressive rotation of the temperature gradients with height is consistent with the larger-scale horizontal gradients of potential temperature analyzed by the ECMWF model at 1800 UTC (not shown) at 700 and 400 hPa, respectively (around the 3.1- and 7.5-km altitudes). It must be recalled, however, that the horizontal gradients retrieved on the mesoscale include the effects of both temperature contrasts and cloud water loading, which disables any further quantitative comparison.
In what follows, convective-scale domains are selected within the mesoscale domain, and the subsystems' internal dynamics are examined. A more detailed interpretation of the retrieved thermodynamic fields (including pressure gradients) and processes involved in the organization of these thermodynamic gradients is carried out.
4. Convective-scale analysis of the 15 December 1992 MCS
In the previous section, it was shown that the mesoscale organization of S1 and S2 was largely determined by the synoptic-scale kinematic and thermodynamic characteristics. As explained in section 2, since the respective minimum observable scale is getting smaller in the present study from synoptic scale (about 200–300 km) to mesoscale (about 30 km; see calculations in section 2b) and convective scale (about 5 km), it is expected that convective-scale features will appear if convective-scale processes are involved in the internal organization of the studied MCS. For this purpose, the 3D fields of wind, precipitation, pressure, and virtual cloud potential temperature perturbations, and the air and water parcel trajectories are analyzed within selected convective-scale domains focusing on subsystems S1 and S2. In the present case, the radar data collected during a time period of 10–15 min are used, time during which the 3D wind field can reasonably be assumed as steady state. Indeed, as shown on the mesoscale, the disturbance does not experience a significant internal evolution during the sampling time of S by the airborne radars.
Figure 2a shows the location of the two selected convective-scale domains within S. Domains A and B are chosen in such a way that the leading convective lines of S1 and S2 can be studied individually. The horizontal domain size is 60 × 60 and 70 × 70 km2, respectively, for A and B. The horizontal grid resolution is 1.5 × 1.5 km2 and the order of expansion of the analytical functions is 7 in both cases, which (using the calculations of section 2b) corresponds to minimum observable scales of 5 and 6 km, respectively, for the retrievals within A and B.
a. Convective-scale study of S1
The first convective-scale study is carried out within S1, in the area A shown in Fig. 2a. Airborne radar data during the time interval 1800–1812 UTC are used to recover the convective-scale kinematic and thermodynamic fields.
1) Dynamic characteristics
Figure 10 shows the retrieved 3D wind and precipitation fields at 1-, 3-, 6-, and 8-km altitudes. The convective-scale structure of the precipitation field is fairly comparable to the mesoscale structure within the A domain (see Fig. 2). The retrieved 3D wind field is mainly characterized at low levels by a relatively homogeneous westerly rear inflow that enters in confluence with a southwesterly inflow ahead of LS1 (see Figs. 10a,b). This characteristic is in very good agreement with the airflows at synoptic scale (see Part I, Fig. 9) and mesoscale (see Figs. 2a,b). This suggests that the synoptic scale determines the organization of the convective-scale motions within S1. Moreover, it must be noted that the convergence line associated with LS1, already discussed with regard to the mesoscale and synoptic-scale (see section 3 and Part I, respectively), is also present on the convective scale, vertically extending from 0.5- to 3-km altitude. A rotation of the wind with height from westerlies at 3 km (Fig. 10b) to southerlies at 6 km (Fig. 10c) is observed. Above the 6-km altitude, a strong divergence of the flows appears (Fig. 10d), with the same characteristics as those discussed on the mesoscale and synoptic scale, despite the very different resolved scales.
The structure of the precipitation field appears to be linked to the structure of the 3D wind field. In particular, the horizontal wind field at 8 km (Fig. 10d) indicates that a parcel located initially at 8-km altitude in the southwesterly part of LS1 will be advected northeastward along LS1. Such a parcel will be finally drained off either southeastward or northeastward of the convective line, depending on its initial location within the different airflow at 8-km altitude. A detailed study of this relationship between the 3D wind and reflectivity fields is carried out in the next subsection, using air and water parcel trajectory analysis.
A vertical cross-section perpendicular to the orientation of LS1 is presented in Fig. 11 to evidence the strong similarities between the vertical kinematic characteristics on the mesoscale and convective scale. The westerly rear inflow at low levels discussed in section 3, as well as in Part I, is found at the rear of LS1, with a depth of 3 km in this region (Fig. 11). Another convective-scale retrieval performed in the northeastern part of LS1 (not shown) indicates that the westerly flow is deeper in this northeastern part (i.e., from 4 to 5.5 km as one moves northeastward along the line). This feature was also found on the mesoscale and at larger scale, which shows that this characteristic is not linked to convective-scale processes. The vertical cross section of Fig. 11 also shows that the maximum upward motions are located in the upper troposphere. This result is common to the three scales of motion (see Figs. 6, 5a, and 11, respectively). The convective-scale updraft associated with the more intense precipitating region is split into a rearward and a frontward upper-level outflow, as was found on the mesoscale.
2) Air and water parcel trajectory analysis
The air and water trajectory analysis is performed using the systematic procedure described in section 2d, but with initial positions of the air parcels every 5 km in the horizontal and 500 m in the vertical. Figure 12 gives the air parcel trajectories that are subjectively selected as the most characteristic of the retrieved convective-scale airflow circulation. Within the mesoscale westerly flow located at the rear of LS1, relatively stable flows are observed. Their stability depends on their location with respect to LS1. The part of the westerly flow that is closer to LS1 (Fig. 12a) slightly overlays the rearmost one (Fig. 12b). This configuration is consistent with the mesoscale observations (see conceptual scheme of Fig. 7, flows 1 and 2. Moreover, the updrafts associated with the synoptic-scale southwesterly flow are characterized by the same structure on the mesoscale and convective scale (see Fig. 7 and Figs. 12c,d), with the main updraft that is split into a frontward and a rearward component with respect to LS1.
As explained in section 2d, the water parcel trajectories can be simply obtained from the air parcel trajectories by subtracting an estimate of the terminal fall velocity of the hydrometeors to the retrieved vertical wind component. In the present case, such an estimate has been obtained using the (SAVAD) analysis (Protat et al. 1997). The reflectivity field at 1-km altitude, displayed in Fig. 10a, shows the significant variability of the precipitation structure along LS1. In the southwestern part of the line, the structure is almost linear and roughly organized along the main ascending flow (see in particular Figs. 12c,d). In contrast, the northeastern part of LS1 exhibits a more complex structure that seems to be related to the diverging flows located in the upper troposphere, as discussed previously. The water parcel trajectory analysis (example given in Fig. 13a) shows that precipitation at the ground within the southwestern part of the line is generated in liquid phase. Indeed, water parcels advected by the main updraft between the ground and the 0°C isotherm altitude (Fig. 13a, for a water parcel located at 3.5-km altitude within the updraft) are falling to the ground in liquid phase within the southwestern “linear part” of LS1 and are not advected farther northeastward. It is seen from Fig. 13a that this parcel is falling almost vertically, in response to its large terminal fall velocity in liquid phase (around 5–10 m s−1).
In contrast, precipitation at the ground within the northeastern part of the line appears to be generated in ice phase. The water parcels located within the updraft at heights greater than the 0°C isotherm altitude do not fall within this linear part of LS1 (see example in Fig. 13b, for a parcel located initially by the 5-km altitude), but keep on slightly ascending, most likely in response to a compensation of their weak terminal fall velocity (typically between 0 and 2 m s−1) by the ascending air motions. These parcels are advected by the flow at 6–8-km altitude along the leading convective line LS1 and fall within the northeastern part of LS1 (Fig. 13b). Interestingly, the strong divergent flows at mid- to upper levels (Figs. 10c,d) act to spread the parcels horizontally at these heights, which likely explains the observed enlargement of the precipitation structure at the ground in the northeastern part of the line. It must be noted however that this simple interpretation does not account for the complex combination of microphysical processes that are known to play a significant role within tropical convective systems.
3) Thermodynamic study of the convective line of S1
Figures 14a–d show the virtual cloud potential temperature perturbations retrieved within area A of Fig. 2a. The field at 0.5-km altitude (Fig. 14a) exhibits a well-marked negative virtual cloud potential temperature gradient along the leading convective line LS1 from the southwestern to the northeastern part of the line. Also interesting is the local line of positive perturbations in the southwestern part of the leading convective line. These positive perturbations are associated with the slight ascent of the southwesterly flow at 0.5-km altitude (not shown). Comparisons (not shown) with the vertical velocity field at 0.5-km altitude indicate that the positive virtual cloud temperature perturbations in the left part of the domain are associated with slightly ascending motions (partly seen in Fig. 12a), while the negative perturbations in the northeastern part of LS1 are associated with downward motions in the low levels (below 1-km altitude), as shown in Fig. 12b at the end of the trajectory. This westerly rear inflow of Fig. 12b is progressively descending from 0.5- to 0.2-km altitude as it crosses the domain. Along this trajectory, the air parcel experiences a cooling of about 2–3 K. This result suggests that significant evaporation occurs during the advection of this flow across the A domain (especially at the end of the trajectory) within the northeastern part of LS1.
We now turn to a description of the virtual cloud potential temperature perturbation fields at higher levels (Figs. 14b,c). From 3-km altitude, the gradients rotate rapidly with height (see Fig. 14b), and positive temperature gradients are roughly oriented along the leading convective line LS1. This configuration of the temperature gradients is such that the coldest air mass in the lower layers (Fig. 14a) is overlayed by the warmest air in the right part of the domain. The presence of the warm air mass above 3-km altitude is likely resulting from the ascent of the mesoscale southwesterly flow (denoted 3 in Fig. 7) above the low-level westerly rear inflow. Above 6-km altitude, the positive gradients are progressively veering with height (see Figs. 14b,c). These positive gradients are oriented almost perpendicular to LS1 at the 10-km altitude (not shown). This result indicates the progressive advection of warm air toward the rear part of LS1, which is consistent with the rearward-tilting part of updraft 3 seen in Fig. 7 on the mesoscale within S1.
The well-marked transition in the orientation of the virtual cloud potential temperature gradients at 3-km altitude (Fig. 14b) also clearly appears in the pressure perturbation field (see Fig. 15b). The retrieved pressure gradients progressively veer with height (see Fig. 15b) and are oriented almost parallel to LS1 at 7-km altitude (Fig. 15c). The configuration of the horizontal pressure gradients appears therefore strongly favorable for the development of upward motions in the southwestern part of LS1, with the highest pressure perturbation in the lowest layers overlayed by the lowest pressure perturbation at 4 km, contrary to the other parts of the domain where the configuration of pressure perturbation is reversed. This rotation of the pressure gradients with height seems to be clearly related to the rotation with height of the vertical wind shear, which occurs very rapidly between the tropospheric layers located below and above 4-km altitude, as shown by the hodograph of the mean horizontal wind in the A domain (Fig. 16). This suggests that a dynamic process linked to this vertical wind shear likely plays a significant role in the retrieved structure of the pressure gradients. It is well known that the general organization of tropical convection within tropical systems is mainly governed by the vertical distribution of buoyancy. However, dynamic forcings can play a significant role as well in the detailed structure of the thermodynamic fields.
This relationship indicates that in the direction of the wind shear, positive and negative pressure perturbations should be located at the rear and ahead of the updraft (with respect to the direction of the shear), respectively. Inspection of the hodograph in Fig. 16 reveals that the vertical wind shear vector points to the north in the low levels. Therefore, we expect (using the approach of Rotunno and Klemp 1982) to find a negative pressure gradient from south to north, since there is a region of upward motions at the location of LS1. This is effectively observed on the pressure perturbation field at 1-km altitude (Fig. 15a). The retrieved Laplacian of pressure leads however to a −0.3 hPa pressure perturbation, which is much smaller than the perturbation of −1.3 hPa estimated using the Bernoulli theory. This is most likely due to the dominant contribution of the vertical gradient of buoyancy to the total Laplacian of pressure in this area (due to the presence of a density current at low levels). In the same way, above 4-km altitude, the vertical wind shear vector is pointing to the southwest (Fig. 16), which corresponds to the orientation of the negative pressure gradients at 4- and 7-km altitude (see Figs. 15b,c, respectively).
This study indicates that the structure of the pressure perturbation field and the organization of the areas favorable for the development of vertical motions are strongly related to the vertical distribution of vertical wind shear. Comparisons throughout section 4a between the convective-scale and synoptic-scale airflow circulations showed that the convective-scale flows were largely determined by the synoptic-scale circulation (although these flows are naturally modified locally by convective-scale processes). As a result, the vertical wind shear is also mostly determined by the synoptic-scale circulation, which clearly shows the major role played by the synoptic-scale characteristics in the internal structure of S1 observed at smaller scales.
b. Convective-scale study of S2
The second convective-scale study is conducted within subsystem S2, in area B of Fig. 2a. Airborne radar data collected during the time interval 2031–2041 UTC are processed so as to obtain the dynamic and thermodynamic fields within area B. Figure 18 shows the retrieved 3D wind and reflectivity fields at 0.5-, 1.5-, 5-, and 8-km altitude. The precipitation structure is found to be close to the mesoscale structure seen in Fig. 2, with a very active leading convective line LS2 ahead of S2, and a weakly developed stratiform precipitation area in the northwestern part of S2. The retrieved airflow within B is generally similar to the mesoscale retrieval for every altitude (see section 3, Fig. 2), with a well-defined confluence in the lowest levels ahead of LS2 (Fig. 18a) between a relatively homogeneous westerly rear inflow and an unstable south-southwesterly inflow ahead of the leading line. This confluence is progressively shifted rearward of LS2 in the upper levels (see Fig. 18b). Above 2-km altitude, a rotation of the wind with altitude is found, from westerly to south-southwesterly. This latter feature is again similar to the mesoscale flow.
Vertical cross sections normal to LS2 (Fig. 19) indicate first that the westerly rear inflow is very shallow (less than 1 km deep), as was already documented on the mesoscale (Fig. 5a). Upward motions are triggered in the lowest tropospheric layers ahead of LS2, and the maximum vertical motions are located in the upper troposphere. These results were previously obtained at the synoptic scale and mesoscale (see Figs. 6 and 5a, respectively). It may be noted as well that the cross-line variability of the dynamic structures along LS2 is small (cf. Figs. 19a and 19b). Subsystem S2 is therefore characterized by a quasi-2D dynamic structure, despite the very “cellular” organization of the precipitation field. This suggests that the convective cells are not playing a major role in the dynamic structure of S2.
Air parcel trajectory analysis on the convective scale within S2 confirms the conceptual scheme of Fig. 7 on the mesoscale. More generally, a detailed study of the kinematic characteristics within S2 as was done within S1 in section 4a shows that the convective-scale processes involved in the dynamics of LS2 are identical to those governing LS1. The only slight differences are the existence of a shallower low-level westerly flow at the rear of LS2, and the less stable behavior of this rear inflow. This shallowness likely results from a smaller evaporative cooling of precipitation within the weakly developed stratiform part of S2. This explains also why the low-level westerly flow is more unstable at the rear of S2 than at the rear of S1, since precipitation generated within the stratiform part of S2 does not have time enough to reinforce the stability of this flow.
As shown on the mesoscale, a shallow density current establishes a dynamic link between S1 and S2, since the shallow density current comes from S1 and propagates toward S2. This suggests a mesoscale interaction between S1 and S2. Unfortunately, this mesoscale interaction cannot be studied in detail since the temporal evolution of S1 and S2 has not been sampled by the airborne radars (S1 and S2 have both been sampled during their mature stage). However, it may be speculated that the development of S2 had been initiated by the propagation of this density current, which likely arose within S1 from the well-known downward spreading of cold air near the ground. This density current is cooled and accelerated within the stratiform part of S2, as shown on the mesoscale in the vertical cross section of Fig. 5b. The fact that this density current gets more shallow as it approaches the leading convective line of S2 suggests that the amount of precipitation generated within the stratiform part of S2 is not enough at the time of the radar observations to strengthen this density current, leading to a diminution of the forcing of the upward motions.
In the same way, the retrieval of the thermodynamic fields of pressure and temperature perturbations within convective-scale domain B of Fig. 2a indicates strong similarities between the thermodynamic organization within S1 and S2. A detailed examination similar to that presented in section 4a(3) for S1—including hodographs at different locations within S2, the diagnostic equation of the Laplacian of pressure, and the retrieved thermodynamic perturbations—leads to the same overall conclusions for S2 as for S1. As for the kinematic properties, slight differences are nevertheless found, especially at 4-km altitude where the rotation with height of the temperature and pressure gradients occurs at lower heights within S2 than within S1, which is consistent with the shallower westerly flow at the rear of S2 and the veering with height of the horizontal wind that occurs at lower heights than within S1. As in the case of S1, it is found that the pressure gradients are mostly linked to the synoptic-scale vertical wind shear.
5. Summary and conclusions
In a first paper of this series, the successive synoptic-scale mechanisms involved in the initiation and evolution of the 15 December 1992 cloud cluster observed during TOGA COARE are inferred, using the ECMWF model outputs and derived kinematic and thermodynamic quantities. A summary of the inferred mechanisms is presented in section 1. This study showed in particular that the selected MCS appeared in a region particularly favorable at the synoptic scale for the development of convection. The temporal evolution of the system was found to be strongly determined by the supply of the MCS with CAPE by the synoptic-scale circulation and by the temporal evolution of the synoptic-scale low-level convergence area.
In the present paper, the kinematic and thermodynamic characteristics of the mesoscale and convective-scale flows are derived from airborne Doppler radar data collected within the selected MCS and compared with the synoptic-scale characteristics, in order to evaluate the downscale interactions involved in the internal organization of this system. The reflectivity fields on mesoscale and convective-scale detail the general morphological structure of the cloud cluster as seen in the satellite imagery. They evidence the existence of two subsystems, S1 and S2, within the MCS, characterized by distinct internal structures. In some aspects, the 15 December subsystems S1 and S2 share apparent similarities with archetypal tropical squall lines: 1) the presence of strong leading convective lines, stratiform areas in the rear part of the system, and transition zones of weaker precipitation in between the convective and stratiform parts, 2) the rearward evacuation of ice precipitation by a front to rear updraft, which generates the stratiform part of tropical squall lines, 3) an elevated vertical velocity maximum, often found in the literature within such systems, 4) the existence of strong positive vertical transport of horizontal momentum carried out by the mean flux component throughout the troposphere, and 5) the observation of leading convective lines perpendicular to the low-level vertical wind shear. However, as discussed throughout this study, significant departures are also evidenced. First, S1 is characterized by a 3D structure, with a transition zone of weak reflectivity echoes shifted to the northeast of the leading lines instead of lying just at the rear of the leading convective line. The vertical structure of precipitation and airflow is not sloping rearward with height, contrary to tropical squall lines (see Jorgensen et al. 1997). The rear inflow is not significantly colder than the front to rear inflow ahead of the leading convective lines (about 1 K colder). The system remains fairly fixed in location during the aircraft sampling, which does not put this system in any of the five categories of TOGA COARE convection proposed by LeMone et al. (1998). In addition, the existence of a strong forward and rearward splitting of the updraft above 10-km altitude into distinct upshear- and downshear-tilted outflows (which is not a typical feature) seems to explain several of the nonarchetypal kinematic and thermodynamic characteristics of this system, like the existence of negative transport of horizontal momentum above 10 km, carried out by these upshear-tilted updraft eddies.
The detailed comparison of the synoptic-scale, mesoscale, and convective-scale kinematic characteristics conducted throughout this paper showed the major role played by the synoptic-scale circulation in the internal organization of subsystems S1 and S2. In particular, the internal structure of the horizontal gradients of pressure and virtual cloud potential temperature on the convective scale was found to be mostly determined by a dynamic process related to the synoptic-scale vertical wind shear. It is also found that the front part of the westerly rear inflow exhibits characteristics consistent with the density current theory. Momentum flux calculations showed that the upward momentum transport on the mesoscale was mostly determined by the mean flux component. This indicates that the vertical transport of horizontal momentum is mostly carried out at a scale of motion larger than the mesoscale.
Internal processes on the mesoscale and convective scale are found to modulate the internal organization of the system defined by the synoptic-scale circulation. These processes are mostly related to the presence of precipitation within S1 and S2. In particular, it is shown that the cooling of the low-level inflow at the rear of S1 can be accounted for by evaporation of precipitation within the stratiform part of S1.
Comparisons between the kinematic and thermodynamic characteristics within S1 and S2 showed that the processes involved in the internal organization of these subsystems were relatively identical. The only differences were the existence of a shallower westerly rear inflow associated with S2, and the less stable behavior of this rear inflow. This smaller depth results from a smaller evaporative cooling of precipitation within the weakly developed stratiform part of S2. This shallow density current arising from S1 and propagating toward S2 establishes a dynamic link between the subsystems. This density current gets more shallow as it approaches the leading convective line of S2, which suggests that the amount of precipitation within the stratiform part of S2 is not enough at the time of the radar observations to strengthen and deepen this density current.
This multiscale approach, based on the analysis of the kinematic and thermodynamic fields at different scales of motion, is currently applied to other MCSs sampled during TOGA COARE. These cases are characterized by propagative and morphological characteristics significantly different from the 15 December case that allow examination of the distinct processes involved in the initiation, evolution, organization, and propagation of these tropical MCSs in relation to their specific larger-scale atmospheric context.
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.
Berry, F. A. Jr, E. Bollay, and N. R. Beers, 1945: Handbook of Meteorology,. McGraw-Hill, 1068 pp.
Biggerstaff, M. L., and R. A. Houze Jr., 1993: Kinematics and microphysics of the transition zone of a midlatitude squall line system. J. Atmos. Sci, 50 , 3091–3110.
Braun, S. A., and R. A. Houze Jr., 1994: The transition zone and secondary maximum of radar reflectivity behind a midlatitude squall line: Results retrieved from Doppler radar data. J. Atmos. Sci, 51 , 2733–2755.
Caniaux, G., J-P. Lafore, and J-L. Redelsperger, 1995: A numerical study of 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., P. Amayenc, G. Scialom, and J. Testud, 1987: A tropical squall line observed during the COPT 81 experiment in West Africa. Part I: Kinematic structure inferred from dual-Doppler radar data. Mon. Wea. Rev, 115 , 670–694.
Dou, X. K., G. Scialom, and Y. Lemaître, 1996: MANDOP analysis and airborne Doppler radar for mesoscale studies. Quart. J. Roy. Meteor. Soc, 122 , 1231–1261.
Fovell, R. G., and Y. Ogura, 1988: Numerical simulation of a midlatitude squall line in two dimensions. J. Atmos. Sci, 45 , 3846–3879.
Gal-Chen, T., 1978: A method for the initialization of the anelastic equations: Implications for matching models with observations. Mon. Wea. Rev, 106 , 587–606.
Hane, R. B., and P. S. Ray, 1985: Pressure and buoyancy fields derived from Doppler radar in a tornadic thunderstorm. J. Atmos. Sci, 42 , 18–35.
Hane, R. B., and D. P. Jorgensen, 1995: Dynamic aspects of a distinctly three-dimensional mesoscale convective system. Mon. Wea. Rev, 123 , 3194–3214.
Hardy, K. R., 1963: The development of raindrop-size distributions and implications related to the physics of precipitation. J. Atmos. Sci, 20 , 299–312.
Hildebrand, P. H., 1998: Shear-parallel moist convection over the tropical ocean: A case study from 18 February 1993 TOGA COARE. Mon. Wea. Rev, 126 , 1952–1976.
Hildebrand, P. H., and C. K. Mueller, 1985: Evaluation of meteorological airborne Doppler radar. Part I: Dual-Doppler analyses of air motions. J. Atmos. Oceanic Technol, 2 , 362–380.
Houze, R. A. Jr, . 1993: Cloud Dynamics. International Geophysics Series, Vol. 53, Academic Press, 572 pp.
Jorgensen, D. P., P. H. Hildebrand, and C. L. Frush, 1983: Feasability test of an airborne pulse-Doppler meteorological radar. J. Climate Appl. Meteor, 22 , 744–757.
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 structure, circulation, and surface energy fluxes. J. Atmos. Sci, 54 , 1961–1985.
Keenan, T. D., and R. E. Carbone, 1992: A preliminary morphology of precipitation systems in tropical northern Australia. Quart. J. Roy. Meteor. Soc, 118 , 283–326.
Lafore, J-P., 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 , 521–544.
Lemaître, Y., and J. Testud, 1986: Observation and modelling of tropical squall lines observed during the “COPT 79” experiment. Ann. Geophy, 4B , 21–36.
LeMone, M. A., 1983: Momentum transport by a line of cumulonimbus. J. Atmos. Sci, 40 , 1815–1834.
LeMone, M. A., and D. P. Jorgensen, 1991: Precipitation and kinematic structure of an oceanic mesoscale convective system. Part II: Momentum transport and generation. Mon. Wea. Rev, 119 , 2638–2653.
LeMone, M. A., and M. W. Moncrieff, 1994: Momentum and mass transport by convective bands: Comparisons of highly idealized dynamical models to observations. J. Atmos. Sci, 51 , 281–305.
LeMone, M. A., G. M. Barnes, and E. J. Zipser, 1984: Momentum fluxes by lines of cumulonimbus over the tropical oceans. J. Atmos. Sci, 41 , 1914–1932.
LeMone, M. A., E. J. Zipser, and S. B. Trier, 1998: The role of environmental shear and thermodynamic conditions in determining the structure and evolution of mesoscale convective systems during TOGA COARE. J. Atmos. Sci, 55 , 3493–3518.
Lin, Y., T. C. Wang, R. W. Pasken, H. Shen, and Z. S. Deng, 1990:: Characteristics of a subtropical squall line determined from TAMEX dual-Doppler data. Part II: Dynamic and thermodynamic structure and momentum budgets. J. Atmos. Sci, 47 , 2382–2399.
Marécal, V., S. Houée, P. Amayenc, T. Tani, and J. Testud, 1996: Radar attenuation in TOGA-COARE systems: Estimation and interpretation from microphysical data. Proc. 12th Conf. on Clouds and Precipitation, Zurich, Switzerland, Int. Commission on Clouds and Precipitation and Int. Association of Meteorology and Atmospheric Science, 204–207.
Marécal, V., T. Tani, P. Amayenc, C. Klapisz, E. Obligis, and N. Viltard, 1997: Rain relations inferred from microphysical data in TOGA COARE and their use to test a rain profiling method from radar measurements at Ku-band. J. Appl. Meteor, 36 , 1629–1646.
Matejka, T., and M. A. LeMone, 1990: The generation and redistribution of momentum in a squall line. Preprints, Fourth Conf. on Mesoscale Processes, Boulder, CO, Amer. Meteor. Soc., 196–197.
Protat, A., and Y. Lemaître, 2001: Scale interactions involved in the initiation, structure, and evolution of the 15 December 1992 MCS observed during TOGA COARE. Part I: Synoptic-scale processes. Mon. Wea. Rev, 129 , 1757–1778.
Protat, A., Y. Lemaître, and G. Scialom, 1997: Retrieval of kinematic fields using a single-beam airborne Doppler radar performing circular trajectories. J. Atmos. Oceanic Technol, 14 , 769–791.
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.
Rotunno, R., and J. B. Klemp, 1982: The influence of the shear induced pressure gradient on thunderstorm motion. Mon. Wea. Rev, 110 , 136–151.
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.
Schlesinger, R. E., 1984: Mature thunderstorm cloud-top structure and dynamics: A three-dimensional numerical simulation study. J. Atmos. Sci, 41 , 1551–1570.
Scialom, G., and Y. Lemaître, 1990: A new analysis for the retrieval of the three-dimensional wind field from multiple Doppler radars. J. Atmos. Oceanic Technol, 7 , 640–665.
Szeto, K. K., and H-R. Cho, 1994: A numerical investigation of squall lines. Part II: The mechanics of evolution. J. Atmos. Sci, 51 , 425–433.
Webster, P. J., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Sci, 73 , 1377–1415.
Weisman, M. L., 1992: The role of convectively generated rear-inflow jets in the evolution of long-lived mesoconvective systems. J. Atmos. Sci, 49 , 1826–1847.
APPENDIX
Diagnostic Equation Governing the Laplacian of Pressure
The virtual cloud potential temperature perturbation is defined as
The SAVAD analysis provides vertical profiles of mesoscale quantities (horizontal wind and its first-order derivatives, terminal fall speed of the hydrometeors, and the horizontal gradients or pressure and temperature using the horizontal momentum equations) from the data collected during the circular P3 aircraft tracks (Protat et al. 1997).