• Bolton, D., 1980: The computation of equivalent potential temperature. Mon. Wea. Rev., 108 , 10461053.

  • Bryan, G. H., 2002: An investigation of the convective region of numerically simulated squall lines. Ph.D. thesis, The Pennsylvania State University, 181 pp.

    • Search Google Scholar
    • Export Citation
  • Bryan, G. H., , and J. M. Fritsch, 2002: A benchmark simulation for moist nonhydrostatic numerical models. Mon. Wea. Rev., 130 , 29172928.

    • Search Google Scholar
    • Export Citation
  • Caracena, F., , and M. W. Maier, 1987: Analysis of a microburst in the FACE Meteorological Mesonetwork in southern Florida. Mon. Wea. Rev., 115 , 969985.

    • Search Google Scholar
    • Export Citation
  • Chappel, L., 2001: Assessing severe thunderstorm potential days and storm types in the tropics. Proc. Int. Workshop on the Dynamics and Forecasting of Tropical Weather Systems. Darwin, NT, Australia, Australian Meteorological and Oceanographical Society, 1–11.

    • Search Google Scholar
    • Export Citation
  • Clark, T. L., 1979: Numerical simulations with a three-dimensional cloud model: Lateral boundary condition experiments and multicellular severe storm simulations. J. Atmos. Sci., 36 , 21912215.

    • Search Google Scholar
    • Export Citation
  • Colon, J. A., 1953: The mean summer atmosphere of the rainy season over the western tropical Pacific Ocean. Bull. Amer. Meteor. Soc., 34 , 333334.

    • Search Google Scholar
    • Export Citation
  • Gilmore, M. S., , and L. J. Wicker, 1998: The influence of midtropospheric dryness on supercell morphology and evolution. Mon. Wea. Rev., 126 , 943958.

    • Search Google Scholar
    • Export Citation
  • Gilmore, M. S., , J. M. Straka, , and E. N. Rasmussen, 2004: Precipitation and evolution sensitivity in simulated deep convective storms: Comparisons between liquid-only and simple ice and liquid phase microphysics. Mon. Wea. Rev., 132 , 18971916.

    • Search Google Scholar
    • Export Citation
  • Jordan, C. L., 1958: Mean soundings for the West Indies area. J. Meteor., 15 , 9197.

  • Klemp, J. B., 1987: Dynamics of tornadic thunderstorms. Annu. Rev. Fluid Mech., 19 , 369402.

  • McCaul Jr., E. W., , and M. L. Weisman, 2001: The sensitivity of simulated supercell structure and intensity to variations in the shapes of environmental buoyancy and shear profiles. Mon. Wea. Rev., 129 , 664687.

    • Search Google Scholar
    • Export Citation
  • McCaul Jr., E. W., , C. Cohen, , and C. Kirkpatrick, 2005: The sensitivity of simulated storm structure, intensity, and precipitation efficiency to environmental temperature. Mon. Wea. Rev., 133 , 30153037.

    • Search Google Scholar
    • Export Citation
  • Rotunno, R., 1981: On the evolution of thunderstorm rotation. Mon. Wea. Rev., 109 , 577586.

  • Weisman, M. L., , and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110 , 504520.

    • Search Google Scholar
    • Export Citation
  • Weisman, M. L., , and J. B. Klemp, 1984: The structure and classification of numerically simulated convective storms in directionally varying wind shears. Mon. Wea. Rev., 112 , 24792498.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Wind profiles as defined by Eq. (2) with zs = 3 km for Us = (15, 25, 35, 45) m s−1.

  • View in gallery

    Skew-T diagram depicting temperature and moisture profiles used in the (a) qυ0 = 14 g kg−1 midlatitude (dashed lines) and Darwin (solid lines) model experiments and in the (b) Jordan (solid lines) and Colon (dashed lines) model experiments. The Darwin profile plotted here is an average profile over days when severe and nonsevere storms occurred (Davg; YYMMDD = 051114, 031126, 060207).

  • View in gallery

    Vertical cross sections of the model-generated storm for the Us = 0 m s−1 experiment for (a) the midlatitudes and (b) the tropics at a time when the cloud-top height is at a maximum. Vectors represent wind, the thick line is the 0.1 g kg−1 contour of cloud water, and the mixing ratio qυ is represented by the thin lines contoured at 4, 8, 12, and 16 g kg−1.

  • View in gallery

    Horizontal cross section through the supercell propagating to the right of the mean wind (a) 47 and (b) 70 min after model initialization. A mirror image storm propagates to the left (not shown). The model is initialized with the Colon profile and Us = 30 m s−1. The vertical velocity at midlevels (4.6 km) is contoured every 2 m s−1. Vectors represent storm-relative low-level (175 m) horizontal winds (a mean storm speed of 17 m s−1 was subtracted from the initial field). The surface rain field is indicated by the shaded region and represents the +0.1 g kg−1 perturbation contour, whereas the surface gust front is denoted by the single thick line and represents the −0.5-K temperature perturbation contour.

  • View in gallery

    As in Fig. 4, but initialized with the Jordan profile.

  • View in gallery

    Plotted are (a) CAPE vs shear magnitude Us, (b) CAPE8 km vs Us, and (c) wmax vs Us. Squares (diamonds) represent the midlatitude (tropical) cases in which no splitting or an incomplete split occurs. Hourglasses (triangles) represent the midlatitude (tropical) cases in which complete splitting with or without intensification occurs.

  • View in gallery

    Speed of the gust front υgf (in the north–south direction) vs wind shear Us for models in which the microphysics of (a) warm and (b) cold clouds (Bice) is used. The large circles represent the cases in which splitting occurred. The lowermost three curves in each plot are the midlatitude cases.

  • View in gallery

    Time of gust front appearance tgf minus the time of the occurrence of the maximum updraft velocity twmax vs wind shear Us for Bryan’s model.

  • View in gallery

    Time of gust front appearance tgf minus time of splitting tsplit (in minutes) vs wind shear Us for all midlatitude and tropical cases simulated with Clark’s model. White dots represent the incomplete splits; gray dots represent the complete splits without intensification; black dots represent splitting with intensification.

  • View in gallery

    Maximum of total liquid water vs peak minimum downdraft speed wmin taken at cloud base for all the midlatitude and tropical cases simulated with Bryan’s model. Within every model, Us is varied giving different points in the max(qr + qc) − wmin phase space, which are connected. The cases with large midtropospheric relative humidity (Colon, mid-12, mid-14, mid-16, DavgWK, D011120WK) are marked with large symbols.

  • View in gallery

    Maximum of total liquid water vs water vapor mixing ratio averaged over the lowermost 2 km when the models are run with (a) Bryan’s and (b) Clark’s model.

  • View in gallery

    Maximum of total liquid water vs wind shear magnitude Us for the results obtained with Bryan’s model.

  • View in gallery

    Storm depth vs wind shear magnitude Us for the results obtained with Bryan’s model.

  • View in gallery

    Time series cross section of (a) midlevel and (b) low-level vertical vorticity maximum (in 10−4 s−1) for the qυ0 = 14 g kg−1 midlatitude, Darwin (average), Jordan, and Jordan WK experiment simulated with Clark’s model using Us = 35 m s−1.

  • View in gallery

    Maximum updraft speed as a function of time for the qυ0 = 14 g kg−1 midlatitude, Darwin (average), and Colon experiments, in which Us is set to 35 m s−1. These cases are simulated with Bryan’s model using both the Kessler scheme (mid-14, Colon, Davg) and ice microphysics (mid-14 ice, Colon ice, Davg ice).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 87 87 20
PDF Downloads 68 68 19

A Comparison of Tropical and Midlatitude Thunderstorm Evolution in Response to Wind Shear

View More View Less
  • 1 Department of Physics, University of Munich, Munich, Germany
© Get Permissions
Full access

Abstract

The influence of vertical wind shear on storm development within a tropical environment is studied with the aid of two numerical models and compared with that in simulations of midlatitude storms. The simulations show that larger wind shears are required in a tropical environment than in a midlatitude environment for a storm of given updraft velocity to split. This finding is supported by the experience of forecasters at the Australian Bureau of Meteorology Regional Forecasting Centre in Darwin that the operational storm forecasting tools developed for midlatitude storms overforecast supercells within the tropics.

That tropical storms require higher shears to split can be attributed either to the larger gust front speed or to the earlier gust front occurrence compared to those in the midlatitudes. A fast gust front cuts off the storm from the warm moist inflow and the updraft has little or no time to split. In the cases where the midtropospheric relative humidity is larger in the tropics or comparable with that in the midlatitudes, the total liquid water and ice content within the deeper tropical storms is larger than in the midlatitude storms, causing a stronger downdraft. In other words, the main contribution to the negative buoyancy of the downdraft is the water loading rather than the evaporative cooling. When a tropical storm is simulated in an environment with smaller midtropospheric relative humidity than in the midlatitudes, the amount of liquid water and ice within the storm is comparable to that within the midlatitude storm. Intense evaporation within the tropical storm then leads to a stronger negative buoyancy than in the midlatitude storm, causing a stronger downdraft and thus an earlier or a faster-spreading gust front.

At higher shears in the tropics, entrainment reduces the storm depth and thus water loading, resulting in a delayed gust front initiation and/or reduction of the gust front speed, which then allows storm splitting to occur.

Corresponding author address: Ulrike Wissmeier, Department of Physics, University of Munich, Theresienstrasse 37, 80333 Munich, Germany. Email: ulrike@meteo.physik.uni-muenchen.de

Abstract

The influence of vertical wind shear on storm development within a tropical environment is studied with the aid of two numerical models and compared with that in simulations of midlatitude storms. The simulations show that larger wind shears are required in a tropical environment than in a midlatitude environment for a storm of given updraft velocity to split. This finding is supported by the experience of forecasters at the Australian Bureau of Meteorology Regional Forecasting Centre in Darwin that the operational storm forecasting tools developed for midlatitude storms overforecast supercells within the tropics.

That tropical storms require higher shears to split can be attributed either to the larger gust front speed or to the earlier gust front occurrence compared to those in the midlatitudes. A fast gust front cuts off the storm from the warm moist inflow and the updraft has little or no time to split. In the cases where the midtropospheric relative humidity is larger in the tropics or comparable with that in the midlatitudes, the total liquid water and ice content within the deeper tropical storms is larger than in the midlatitude storms, causing a stronger downdraft. In other words, the main contribution to the negative buoyancy of the downdraft is the water loading rather than the evaporative cooling. When a tropical storm is simulated in an environment with smaller midtropospheric relative humidity than in the midlatitudes, the amount of liquid water and ice within the storm is comparable to that within the midlatitude storm. Intense evaporation within the tropical storm then leads to a stronger negative buoyancy than in the midlatitude storm, causing a stronger downdraft and thus an earlier or a faster-spreading gust front.

At higher shears in the tropics, entrainment reduces the storm depth and thus water loading, resulting in a delayed gust front initiation and/or reduction of the gust front speed, which then allows storm splitting to occur.

Corresponding author address: Ulrike Wissmeier, Department of Physics, University of Munich, Theresienstrasse 37, 80333 Munich, Germany. Email: ulrike@meteo.physik.uni-muenchen.de

1. Introduction

The city of Darwin lies within the tropics of northern Australia and records an average of 80 days of thunder each year. Although tropical thunderstorms are not usually expected to be severe1 due to the generally weak vertical wind shear present, the Darwin area records an average of 12 severe storm events during the “build-up” and wet season (October to May) each year (Chappel 2001). During the four wet seasons in 2002/03–2005/06, the probability of detection of severe storms was lower than 25% with the exception of the season 2005/06 (see Table 1). The false alarm ratio, indicating overprediction of severe storms, was over 50% for each season.

One reason for these poor forecasts is that forecasters in Darwin, and perhaps elsewhere in the tropics, currently use conceptual models of storms developed for the midlatitudes because such models do not exist for tropical environments. Observations, theoretical studies, and numerical simulations of convective midlatitude storms (e.g., Weisman and Klemp 1982, hereafter WK82) have shown that certain thresholds for convective available potential energy (CAPE) and wind shear determine which of the three storm types will be produced: single cell, multicell, or super cell. CAPE and wind shear are often combined to form a Richardson number
i1520-0469-66-8-2385-e1
where u is weighted shear between z = 0 and 6 km.

WK82 considered a unidirectional wind shear with a range of shear magnitudes, as well as buoyant energies representative of midlatitude thunderstorms. Ice microphysics were not included in their model. Weisman and Klemp (1984) extended their work to examine the effect of directionally varying wind shear profiles on the modeled convective storm spectrum. They found that for a clockwise-turning hodograph, supercellular growth is confined to the right flank of the storm system because of the enhancement of favorable vertical pressure gradients. Gilmore et al. (2004, hereafter GSR04) extended the work of Weisman and Klemp (1984) by including ice microphysics in the calculations. The presence of ice produced both stronger updrafts as a result of the additional latent heat released during fusion and colder low-level downdrafts due to the melting of hail. The sensitivity of storm structure and intensity to variations in the shapes of the environmental buoyancy profile, shear profile, and environmental temperature was studied by McCaul and Weisman (2001) and McCaul et al. (2005) within the framework of the Convection Morphology Parameter Space Study (COMPASS). It was found that when storms grow in an environment where the buoyancy profile has a maximum at low levels, typical of a midlatitude sounding, a strong downdraft is produced. In comparison, when the buoyancy is evenly distributed through the depth of the troposphere, such as in tropical environments, the downdrafts produced are not as strong. This dependence of the downdraft strength on the height of maximum buoyancy was shown to be reduced once CAPE values were equal to or exceeded 2000 J kg−1. Furthermore, when the low levels have a large water vapor mixing ratio typical of a tropical environment, the subsequent high condensation loading in the updraft can produce a strong downdraft regardless of the buoyancy profile.

The aim of our study is to use two numerical models to investigate how storms in a tropical environment are influenced by unidirectional vertical wind shear, and whether the thresholds of the Richardson number used to classify the storm type in the midlatitudes apply for storms in the tropics. Essentially this study is an extension of the work of WK82 to storms that occur within a tropical environment. Section 2 describes the modeling framework and methodology used. Results obtained by simulating midlatitude and tropical storms with two different numerical models and two types of microphysics schemes are presented and discussed in section 3, and the conclusions are made in section 4.

2. The numerical model

a. General settings

The two numerical models used in this study are the three-dimensional cloud-scale model of Bryan and Fritsch (2002) and Bryan (2002) and the three-dimensional Clark–Hall cloud-scale model (Clark 1979). Both models use a Kessler-type parameterization for the microphysics; that is, only vapor and liquid processes are considered. Ice processes within the models are omitted initially and those results are presented in section 3a. In addition, Bryan’s model is initialized with ice microphysics (hereafter referred to as Bice). This scheme is identical to Gilmore’s Li scheme, in which cloud water, rainwater, cloud ice, snow, and hail/graupel are predicted (GSR04). The results obtained with the Bice model are presented in section 3b.

One reason for carrying out the simulations with two different models is to assess the sensitivity of the results to a particular model. Further, it was found, that Clark’s model is less suitable for simulations of deep convection including the ice phase because in some cases no gust front, or only an extremely weak one, was observed, which was deemed to be unrealistic.

Various parameters pertaining to the model configurations are given in Table 2. The horizontal domain size used is 60 km × 60 km with a constant grid interval of 1 km. The vertical domain extends to a height of z = 36 km, with the vertical grid interval stretching smoothly from either 300 or 350 m at the lowest grid point up to 1 km for z ≥ 26 km. A sponge layer is implemented in the uppermost 10 km to inhibit the reflection of waves from the upper boundary. The lower boundary is free slip.

Convection is initiated in both models using a symmetric thermal perturbation with a specified horizontal radius and temperature excess. The temperature excess is at a maximum at the center of the perturbation and decreases smoothly to 0 K at its edge. The depth of the perturbation is set to 1 km so that the perturbation lies within the boundary layer (see below). The sensitivity of the results to bubble radius and temperature excess was studied, but no qualitative differences in the storm evolution were found.

The wind profile is defined as in WK82 as a straight-line hodograph:
i1520-0469-66-8-2385-e2
where z is the height and zs is a constant (Fig. 1). The magnitude of the shear is varied by altering the parameter Us (in m s−1). In our calculations the shear layer has a depth of 6 km and Us is varied from 0 to 50 m s−1 in increments of 5 m s−1, giving a maximum shear of 8.3 × 10−3 s−1. To keep the storm in the center of the domain, a mean wind speed is subtracted from the wind profile.

b. Initialization of the model for the midlatitudes and tropics

Midlatitude storms are produced here by initializing each model with the vertical temperature and moisture profile from WK82 (Fig. 2a). The buoyancy (CAPE) is varied by defining surface mixing ratio values qυ0 of 12, 14, and 16 g kg−1, giving CAPE values of 840, 1893, and 2917 J kg−1, respectively.2

Tropical storms are produced by initializing each model with the vertical temperature and moisture profile from the 0000 UTC Darwin sounding (0930 local time) on days when either severe or nonsevere storms occurred. Profiles for three days are considered here (051114, 011120, 041217; dates throughout identified with number showing YYMMDD), together with an average of three profiles (Fig. 2a). To represent the midafternoon conditions when the storms on these days developed, the lowest 1 km of each sounding is modified to give a convectively mixed boundary layer. The boundary layer depth is based on radiosonde observations during the Tropical Warm Pool International Cloud Experiment (TWP-ICE), which took place in the Darwin area in 2006. The midlevel relative humidity for most of the Darwin soundings is lower than that used in WK82. To examine the sensitivity of the model results to midlevel relative humidity, some of the tropical cases were initialized with the same relative humidity profile as in WK82.

To show that the results are not specific to the Darwin thermodynamic profile, the tropical soundings of Jordan (1958) and Colon (1953) are also examined (Fig. 2b). The Jordan sounding is a mean nighttime sounding for the West Indies, where the monthly aerological records (July–October) for the 10-yr period 1946–55 for Miami, Florida; San Juan, Puerto Rico; and Swan Island, Honduras, have been used. The Colon sounding is a Pacific mean sounding, where the nighttime observations from June to September at Kwajalein, Guam, and Palau for the period 1944–47 were used. The temperature profile for the Jordan and the Colon soundings are almost identical. However, the Jordan profile is considerably drier above about 850 hPa. As for the Darwin profile, the Jordan and Colon profiles are modified to give a convectively mixed boundary layer in the lowest 1 km. In the following, a storm simulated using a Darwin thermodynamic profile will be referred to as a “Darwin storm,” whereas a storm simulated in the other environments will be referred to as a “Colon storm,” “Jordan storm,” or a “mid-latitude storm” (see Table 3).

A summary of the magnitudes of CAPE, as calculated from the surface to the equilibrium level (i.e., total CAPE) and from the surface to a height of 8 km (CAPE8 km), for each of the environments considered here is presented in Table 3. The table shows also the maximum updraft speed wmax of the model storms initialized in an environment without vertical wind shear (Us = 0 m s−1).

3. Results and discussion

a. Warm clouds

The following results are obtained by simulations conducted with Bryan’s and Clark’s models using the Kessler scheme, when a thermal perturbation of 1-km depth, 8-km width, and an 8 K temperature excess3 is used.

Model-generated storms in a midlatitude and a tropical environment in the absence of an environmental flow are shown in Fig. 3. The tropical storm has a deeper and stronger updraft than the midlatitude storm because of the higher tropopause and larger CAPE. Moreover, the midlatitude storm has a shorter lifetime than the tropical storm (the lifetime is defined here as the time when the updraft strength w falls below 2 m s−1).

When the environmental wind shear is increased, a value Us is reached at which the initial updraft splits. The time at which the initial updraft splits is defined as the time when, at a height level of 4.6 km, the innermost vertical velocity contour splits into two parts, representing a weakening of the updraft and eventual downdraft development along the east–west axis of the storm. The 4.6-km height level was chosen by WK82 as a representative height for midlevels and is retained here, although it is possible that for the deeper tropical storms a greater height might be more appropriate. After splitting, the separated updrafts move to the left and right of the shear vector, which here points to the east. These left- and right-movers are classified as supercells because of their rotating updrafts (WK82). For the purpose of the following analysis, four types of storm evolution are considered: no split, incomplete split, splitting without intensification, and splitting with intensification. A no split situation occurs when the initial updraft decays without splitting. An incomplete split occurs when the storm decays just as the initial updraft begins to split and a downdraft w ≤ 0 fails to develop between the two updrafts. A split without intensification occurs when the updraft splits and the maximum updraft speed of either supercell remains smaller than the wmax of the initial updraft. If the maximum updraft speeds of the supercells become larger than wmax of the initial updraft, the split is defined as a split with intensification.

To show one example of an incomplete split, horizontal cross sections through a Colon storm initialized with the WK82-relative humidity profile and Us = 30 m s−1 are presented in Fig. 4. After 22 min, the updraft reaches a maximum vertical velocity of 38 m s−1. By 41 min, rain reaches the ground and the gust front begins to spread out. The leading edge of the gust front is defined here as the −0.5 K surface potential temperature perturbation contour. Splitting occurs after 47 min, and the cyclonically rotating “right mover” moves to the south, whereas the anticyclonically rotating “left mover” moves to the north (not shown). Even after 70 min the two updrafts have not separated completely and the system subsequently decays.

Figure 5 shows horizontal cross sections through a Jordan storm initialized with Us = 30 m s−1. A maximum updraft speed of 26.4 m s−1 is reached after 21 min. By 41 min, rain reaches the ground and the gust front begins to spread out. Splitting occurs 3 min later. Because the speed of the gust front is similar to the propagation speed of the supercell, the gust front continues to lift the warm inflow from the east into the supercell as shown in Fig. 5b. The supercell strengthens further, reaching a vertical velocity of 28.5 m s−1 after 90 min. This case is classified as splitting with intensification.

All cases listed in Table 3 are initialized with a range of values of Us defined in section 2a and up to 70 storm cases are examined for each model. Of the cases examined with Clark’s model, 18% are classified as incomplete splits, 23% as splits without intensification, and 59% as splits with intensification. The corresponding distribution for Bryan’s model is 4% incomplete splits, 32% splits without intensification, and 64% splits with intensification. An explanation for the difference is presented later on.

1) CAPE and updraft velocities

In previous studies (e.g., WK82) the dependence of the modeled storm structure on environmental buoyancy and wind shear was generalized in terms of the Richardson number [Eq. (1)]. The variation of CAPE with Us for each storm case examined with Bryan’s model is presented in Fig. 6a. The square roots of the thresholds of the Richardson number given in WK82 are shown by the two straight lines. These lines represent the limits when supercells are expected to occur. The right line indicates the lower boundary of the supercell range (R = 10), whereas the line on the left indicates the upper boundary where R = 50 (see WK82). Squares (diamonds) represent the midlatitude (tropical) cases in which no splitting or an incomplete split occurs. The hourglasses (triangles) represent the midlatitude (tropical) cases in which complete splitting with or without intensification occurs. From Fig. 6a, complete splitting occurs in the midlatitudes for Us ≥ 10 m s−1, whereas in the tropics Us ≥ 30 m s−1 is required for splitting. The simulations conducted with Clark’s model show that complete splits occur for the midlatitude storms if Us ≥ 15 m s−1, whereas for the tropical storms Us ≥ 30 m s−1 is required. Thus, both models show that a larger wind shear is required for a storm to split in a tropical environment than in a midlatitude environment. On first inspection of Eq. (1), it may be expected that the large CAPE in the tropical environments (Fig. 6a) would automatically mean that a larger shear is required to produce storm-splitting compared to midlatitude environments. However, when Fig. 6a is replotted with the maximum updraft speed wmax from the model as the ordinate (Fig. 6c), the values of wmax for storms in both environments are similar. In fact, for a given wind shear and relative humidity profile, there are cases in which the midlatitude CAPE (2917 J kg−1) is smaller than the Darwin CAPE (4890 J kg−1), but the storm updraft for the midlatitude case (39 m s−1) is stronger than the Darwin storm updraft (32 m s−1). These lower values of wmax, despite occurring in environments of larger CAPE than that in midlatitudes, highlight the known deficiencies of CAPE [e.g., it does not consider entrainment, precipitation loading, water vapor deficit, or vertical pressure gradient forcing; see Gilmore and Wicker (1998), among others]. Thus, uncertainties exist regarding use of the Richardson number as a criterion to define the regime where supercells are expected. The Richardson number R predicts that large shear is required for storms to split in tropical environments that have large CAPE, as is found here. However, it seems that a physical interpretation in terms of wmax, which in some cases is lower for the tropical cases compared to the midlatitude cases, suggests that a lower shear would be required to split the tropical storms.

Previous studies (e.g., McCaul and Weisman 2001) have shown that the shape of the buoyancy profile influences the characteristics of convective storms. A midlatitude sounding with large CAPE typically has large buoyancy at low levels, which is where most of the wind shear is located. In contrast, a tropical sounding with large CAPE typically has a thin buoyancy profile that is distributed relatively evenly over a large depth. So, although a midlatitude and a tropical sounding may have the same CAPE, below a height level of, say, 8 km, the CAPE8 km would be typically larger for the midlatitude case than for the tropical case. McCaul and Weisman (2001) showed further that the effect of the buoyancy profile shape on convection is large for small CAPE (800 J kg−1) but decreases as CAPE increases (2000 J kg−1) and “must become unimportant in environments featuring extremely large values of CAPE.”

To examine whether CAPE8 km is a better representation of the buoyancy, and whether the corresponding Richardson number has consistent values for split storms in the tropics and midlatitudes, CAPE8 km is determined for every model and presented in Table 3. Figure 6b is similar to Fig. 6a but shows CAPE8 km versus Us. Note, that the ordinate axis in Fig. 6b has been rescaled. For the tropical environments, the values of CAPE8 km, and thus CAPE8 km and R8 km are lower than the original values and are now comparable to those of the mid-16 case (the abbreviations of the model names can be found in Table 3). For example, the Richardson number R8 km calculated using CAPE8 km and Us = 25 m s−1 (u = 15.2 m s−1) for the Darwin average profile (Davg) and mid-16 case is nearly identical at 16. However, the mid-16 storm does split whereas the tropical storm does not, showing that the Richardson number so defined cannot be used to differentiate the different thunderstorm behaviors. Furthermore, counting the number of cases successfully predicted4 in Figs. 6a and 6b shows that when using R, 49 of 81 cases are correct, whereas with R8 km only 36 cases are correct. In general, complete splitting for the midlatitude cases occurs for R8 km < 100, whereas for the tropical cases R8 km < 13 must be satisfied for complete splitting to occur. Because of this lack of consistency with the Richardson number, we will focus on the maximum updraft speed of the cell wmax in the following sections.

Figure 6c shows that as Us is increased, wmax within the modeled tropical and midlatitude storms decreases. Such a decrease in storm intensity has also been reported in numerical simulations by WK82 and has been shown to result from entrainment into the storm reducing the positive buoyancy of the updraft, which increases as the shear increases.

2) The gust front

To investigate why a larger vertical wind shear is required for tropical storms to split compared with midlatitude storms, we will focus on the cold-air outflow and subsequent gust front of the storm to examine how the gust front is affected by the vertical wind shear. Previous research on midlatitude storms (e.g., WK82; GSR04) has shown that the gust front is important for the evolution of the storm. If the cold-air outflow produces a gust front that moves too fast, the supply of warm air to the updraft and its flanks will be cut off by the spreading cold surface air.

Factors relating to the shape of the buoyancy profile that affect the updraft and subsequent downdraft have been investigated extensively in the COMPASS project (e.g., McCaul and Weisman 2001; McCaul et al. 2005). McCaul and Weisman (2001) define an altitude of maximum parcel buoyancy zb and show that as zb decreases, wmax increases along with the temperature deficit of the subsequent cold pool. As mentioned previously, this response was observed to be stronger for cases in which the total CAPE is relatively small (800 J kg−1), and it decreases as CAPE increases to and above 2000 J kg−1. Given that all but two of the cases examined here (Table 3) have CAPE values exceeding 2000 J kg−1, this effect is expected to be small.

McCaul et al. (2005) also investigated how variations of temperature at the lifted condensation level, which is a good proxy for the environmental precipitable water, affects storm structure. McCaul et al. found that for the warmer and moister sounding, more condensation was produced in the updrafts because of the higher available water vapor mixing ratio in the subcloud layer qυ0, which led to a larger precipitation fallout. The downdraft of this storm was found to be stronger and the outflow colder than that of a storm produced in a cooler, drier environment. It will be shown later (see Fig. 10, large symbols) that tropical storms growing in environments with the same midtropospheric relative humidity as the midlatitude storms, but where CAPE, qυ0 and zb are larger, have higher amounts of water loading and comparable or stronger downdrafts than the midlatitude storms.

We will now investigate further the experiments described in the previous sections (see Table 3). Even though several of the key parameters addressed by McCaul et al. (2005), such as zb, qυ0, CAPE, and relative humidity, vary from case to case, it will be shown that general statements can be made for the midlatitude and tropical thunderstorms regarding the speed or timing of the gust front and the mixing ratio of hydrometeors.

The gust front speed υgf is determined 12 min after the gust front appears, perpendicular to the shear vector, and thus is not directly affected by the environmental wind. The speed perpendicular to the shear vector is chosen here because the strongest low-level convergence occurs along the northern and southern flanks of the initial updraft (not shown). Figure 7a shows the speed of the gust front υgf as a function of Us for Bryan’s model. On the whole, for a particular storm, the speed of the gust front decreases as the shear parameter Us increases. For the storms examined here, the midlatitude storms have smaller gust front speeds than those in the tropical cases. Clark’s model shows no obvious correlation between the gust front speed and the wind shear, and this will be addressed later. The large circles represent those storms in which splitting (with and without intensification) occurred; they coincide with gust front speeds less than about 7.5 m s−1.

The time of gust front appearance relative to the time when the storm attains its maximum updraft speed in the model is shown in Fig. 8 as a function of Us for all storms examined here. Bryan’s and Clark’s models show that for a particular storm, the time of gust front occurrence relative to the time of maximum updraft strength is independent of Us. This means that the time that the storm has to develop before being cut off by the gust front is governed only by how fast the gust front expands outwards. Because the speed of the gust front decreases with increasing Us (Fig. 7a) a storm in a high–wind shear environment will have more time to evolve and possibly split before it becomes cut off from the warm inflow by the gust front. Because the tropical storms examined here exhibit a faster gust front than the midlatitude storms, a greater wind shear is required for the tropical storms for the gust front speed to be sufficiently reduced to allow the updraft to split.

To clarify the role that the gust front plays in determining whether storms split in the Clark model, only the split storms are examined here. The time of gust front occurrence is now calculated relative to the time at which splitting occurs. Figure 9 shows the difference in time between gust front appearance and when the updraft begins to split, as a function of the shear parameter Us. Incomplete split cases (white dots) coincide with relatively low shears; they occur when the gust front appears before splitting starts (tgftsplit < 0). Complete split cases (gray and black dots) coincide with relatively high shears and occur primarily when the gust front appears, either at the time of splitting or after splitting has commenced. Thus, the later the gust front appears, the more time the updraft has to split and develop.

The exact reason for the different development time and speed of the gust front between Bryan’s and Clark’s models is not the focus of this paper, since both models produce the consistent results that a higher shear is needed for tropical storms than for storms in the midlatitudes and that the gust front speed is slower or occurs later for the midlatitude than for the tropical cases. However the differences in the microphysics schemes between the models would, without doubt, influence the gust front development (see, e.g., GSR04).

The dependence of the gust front properties on wind shear is investigated now by examining the factors that control the downdraft, namely, the liquid water loading and the cooling from the evaporation of the liquid water. Figure 10 shows a measure for the liquid water content max(qr + qc) plotted versus the downdraft strength wmin measured at cloud base for every midlatitude and tropical storm simulated with different values of Us. Rain water is designated by qr and cloud water by qc. The results for a particular storm case in which Us is varied are connected by straight lines. If the environmental relative humidity is identical in all cases, one would expect the strength of the downdraft to be controlled primarily by the amount of liquid water loading within the storm. This can be seen by focusing on one storm case in Fig. 10, in which |wmin| generally decreases as max(qr + qc) decreases. This decrease occurs as the shear Us increases. However, because the environmental relative humidity varies from case to case, the strength of the correlation between max(qr + qc) and wmin changes as the contribution from the evaporative cooling varies. Thus, for storms forming in relatively moister environments (DavgWK, midlatitudes, Colon), a smaller |wmin| occurs compared to that in the storms forming in drier environments. “WK” indicates that the model is initialized with the relative humidity profile of WK82 [Eq. (2)].

Figure 10 shows that the midlatitude cases have weaker downdrafts than the tropical cases, a factor that would lead to a slower gust front speed. One reason for the stronger downdrafts in the tropical storms can be found by examining the liquid water content within the storm. The amount of liquid water loading max(qr + qc) versus the water vapor mixing ratio qυ averaged over the lowest 2 km for the midlatitude and tropical storm cases is shown in Fig. 11. The water vapor mixing ratio in the lowest 2 km for the midlatitudes ranges between 11 and 14 g kg−1 (squares and hourglasses), whereas for the tropics the near-surface qυ ranges from 14.5 to 17 g kg−1 (diamonds and triangles). The higher values for qυ in the tropics account for the large CAPE values (see Fig. 2a) and for the higher liquid water content within tropical storms as compared with those in the midlatitudes. The split cases for the tropics (triangles) and midlatitudes (hourglasses) have smaller values of max(qr + qc) for a given surface qυ than for the cases in which no split or an incomplete split occurred.

For the tropical cases in which the relative humidity is lower than in the midlatitude cases (small symbols in Fig. 10), the liquid water content within the tropical storms is comparable to that in the midlatitude storms. However, stronger downdrafts occur in the tropical cases than in mid-12, mid-14, and mid-16 because of the greater entrainment of dry environmental air leading to more evaporative cooling. The variations of the liquid water content and storm top with wind shear Us are shown in Figs. 12 and 13, respectively. As Us increases, max(qr + qc) decreases owing to the increased entrainment into the storm. Furthermore, as Us increases, the storm top decreases as entrainment reduces the strength of the initial updraft (see Fig. 13). Although Figs. 12 and 13 show only the results from Bryan’s model, those from Clark’s model are similar. Thus, the role of the environmental shear is to reduce the liquid water content within the storm, which in turn reduces the downdraft strength.

3) Vertical vorticity

Time series of midlevel (z = 4.6 km) and low-level (z = 178 m) vertical vorticity for the Us = 35 m s−1 experiments for four different cases are presented in Fig. 14. The generation of vertical vorticity within the storm in the experiments is as described by Rotunno (1981) and WK82. During the first 20 min of storm growth, midlevel vertical vorticity develops as a vortex couplet associated with the initial updraft, which is consistent with vorticity generation through the tilting of horizontal vorticity. The vertical vorticity is proportional to the cross-shear horizontal gradient of vertical velocity and the magnitude of shear. Midlevel vorticity is largest in the Jordan WK case, with a peak value of 96.9 × 10−4 s−1, and smallest for the midlatitude case with a peak value of 72.8 × 10−4 s−1, which agrees with the peak midlevel vorticity of the initial updraft for the Us = 35 m s−1 experiment found in WK82 (their Fig. 8a).

After 20 min, the midlevel vorticity in the four cases in Fig. 14a decreases as the initial updraft weakens. The midlatitude, Darwin, and Jordan cases undergo splitting with intensification after approximately 44 min, whereas the Jordan WK case undergoes an incomplete split. At the time of splitting, the Jordan WK case exhibits the largest midlevel vorticity. However, after 55 to 65 min the vorticity of the split cells from the Darwin, Jordan, and midlatitude cases increases and becomes larger than that of the Jordan WK case.

A strong midlevel rotation on the updraft flanks acts to lower the pressure and thereby induce updraft growth on the flanks, enhancing the splitting (Klemp 1987). Because the Jordan WK case has the largest midlevel vorticity until 60 min after model initialization, one would expect that this case is more likely to split than the other three cases in Fig. 14a (i.e., Jordan, Davg, mid-14). This contradiction shows that in the cases studied here, the amount of midlevel vorticity at the time of splitting is not a good indicator for the likelihood of splitting and plays a minor role compared with the gust front (see previous section).

In WK82, significant surface vorticity does not develop until 40 min into the model simulations. However, Fig. 14b shows that in Clark’s model significant low-level vorticity is already present before 40 min because of the tilting and stretching of the horizontal vortex tubes by the rising warm bubble. The second peak in low-level vorticity at about 55 min in Fig. 14b coincides with the development of the cold surface outflow, as increased convergence along the gust front is important in creating and tilting the low-level vorticity (WK82).

The vertical vorticity in Bryan’s model (not shown) is larger on average than the mid- and low-level vorticity in Clark’s model. This difference can be explained by the stronger updrafts within the storms simulated with Bryan’s model than with Clark’s model (see Table 3). The stronger updrafts lead to a greater stretching of the vortex tubes. However, the evolution of the vorticity is similar in both models.

b. Cold clouds

We investigate now the effect of including the ice phase in Bryan’s model (Bice). As done previously for the warm clouds, about 70 storm cases with a range of Us values are considered here and summarized in Table 3. Table 3 shows also the maximum updraft velocities wmax attained by the storm. In the majority of cases considered, the value of Us at which the storms split was identical to that when ice was not included, and the tropical storms still required a higher shear to split than the midlatitude storms. However, three cases with ice did require a larger value of Us for splitting to occur. In general, with ice, 12.5% are classified as incomplete splits, 25% as splits without intensification, and 62.5% as splits with intensification. The corresponding percentages without ice are 4% incomplete splits, 32% splits without intensification, and 64% splits with intensification. Thus, when ice is included in the model, the likelihood of a split either with or without intensification is lower than without ice.

The tendency for these three cases with ice to split at higher shears than the storms simulated without ice can be attributed to the higher gust front speeds υgf observed with Bice. Figure 7b shows the variation of υgf with Us obtained with the Bice model, where the large circles represent the split cases and the three lowermost curves are those of the midlatitude experiments. As in Bryan’s model without ice (Fig. 7a), the split cases can be found at high Us and small υgf. However, the gust front speeds are on average 25% larger when ice is included. The larger gust front speeds are the result of more precipitation produced in Bice than when ice is omitted. Further, heat is extracted from the air in the melting and sublimation of hail/graupel (see GSR04), resulting in a downdraft wmin, which is on average twice as strong in the Bice model than in the model without ice (not shown).

To compare the evolution of storms simulated with Bryan’s model with and without ice, the time evolution of wmax for a midlatitude, a Darwin, and a Colon storm are plotted in Fig. 15. All cases in Fig. 15 are initialized with a wind shear of Us = 35 m s−1. The Colon and Darwin storms decay after approximately 50 min without splitting, independent of whether ice is included or not. The midlatitude storms with and without ice split after 38 and 46 min, respectively. The corresponding supercells then strengthen further and survive more than 2 h. The reason for the earlier time of split when ice is present is the stronger updraft speeding up development.

When ice is included, the updraft speed of all midlatitude storms is larger after a time of approximately 15–20 min than without ice. This result is due to the additional heat released during freezing and deposition, which increases the buoyancy of the updraft and has been found in other numerical studies of convection (e.g., GSR04). When ice is included for the Darwin storms, the updraft speed becomes larger than that without ice after 35 min. The reason why it takes longer for to become larger than in the Darwin ice case than in the midlatitude ice case is because the ice level5 lies approximately 1.1 km higher in the Darwin storms than in the midlatitude storms considered here. Because the updraft speeds within the Darwin storms are similar to that of the midlatitude cases (Fig. 15), a longer time is required to transport air above the ice level for it to have an effect on the updraft strength. When ice is included for the Colon storm, the updraft strength becomes larger than when ice is not included after approximately 18 min. This is earlier than for the midlatitude cases because of the ice level lying just above that of the midlatitudes and the initial updraft being stronger than those of the midlatitude cases.

4. Summary and conclusions

The influence of vertical wind shear on deep convection in tropical and midlatitude environments has been investigated using Bryan’s and Clark’s cloud models. Three vertical temperature and humidity profiles from the midlatitudes, six Darwin profiles, and the Colon and Jordan mean profiles were used in the simulations and the vertical wind shear was varied between 0 and 0.008 s−1. Thus, up to 70 storm cases were examined with each model. Whereas Clark’s model was initialized only with a Kessler warm-rain parameterization, Bryan’s model was run with an ice microphysics scheme also.

The principal findings are as follows:

  • Storms that develop in tropical environments require a larger vertical wind shear to split (Us ≥ 30 m s−1) compared with storms in midlatitude environments (Us ≥ 10 m s−1). This result is independent of the cloud model used and whether or not ice microphysics is included.
  • The propensity for storms to split depends on the speed at which the gust front expands. A slower gust front allows the updraft more time to split and subsequently develop. The tropical storms modeled here exhibited a faster gust front than the midlatitude storms.
  • The speed of the gust front increases with the total water loading within the storm and with decreasing midtropospheric relative humidity in the environment. The greater water vapor mixing ratio in the tropical subcloud layer leads to more condensation in the tropical updrafts and thus to more water loading than in the midlatitude updrafts.
  • High wind shears lead to reduced water loading within the storm, which in turn reduces the speed of the gust front. Larger wind shears are required in the tropical cases than in the midlatitude cases to reduce the water loading and thus the speed of the gust front.
  • Although CAPE can be significantly larger in tropical environments than in midlatitude environments, the updrafts of the tropical storms are similar to, and in some cases weaker than, those of midlatitude storms. The Richardson number with its thresholds for supercells in the midlatitudes should be used with great caution when applied to the forecast of tropical supercells. Defining a CAPE quantity that excludes the upper tropospheric portion (i.e., considering only the CAPE between 0 and 8 km) did not improve the results.

The study of simulated tropical and midlatitude storms confirmed the experience of forecasters at the Bureau of Meteorology in Darwin that the operational storm forecasting tools developed for midlatitude storms overforecast supercells within the tropics. Thus, there is a need to develop new forecasting tools for severe storms valid in the tropics. This work is seen as a first step in such a development. Diagrams such as Fig. 6c may be helpful because they indicate both for midlatitude and tropical environments when split cells are to be expected for a particular wind shear and for the modeled updraft strength wmax. The combination of Figs. 10 and 11 might help to estimate the amount of water loading and thus the strength of the downdraft and gust front if the relative humidity and low-level moisture of the environment are known. If there is a way to estimate the gust front speed, a diagram such as Fig. 7 would aid to predict the likelihood of splitting for a given wind shear.

Of course, not all parameters used in this study are available to the forecasters. The Darwin Convective Analysis, which aims to provide guidance on the risk of severe storms in the Darwin area, is based on the 0000 UTC sounding. In other words, only parameters such as CAPE, the downdraft maximum available potential energy (DMAPE),6 and the maximum low–midlevel wind shear7 are available for forecasting.

Thus, more research is necessary to combine the findings of this work and the parameters that have been identified here as important, such as wmax, wmin, and max(qr + qc), with environmental parameters such as CAPE, DMAPE, midtropospheric relative humidity, and vertical wind shear.

Acknowledgments

We wish to express our gratitude to Roger K. Smith for the helpful discussions concerning this work and to George Bryan and Terry Clark for kindly making their models available. We also acknowledge the helpful comments of the reviewers. The first author is grateful to the Deutsche Forschungsgemeinschaft (DFG) for providing financial support for this study.

REFERENCES

  • Bolton, D., 1980: The computation of equivalent potential temperature. Mon. Wea. Rev., 108 , 10461053.

  • Bryan, G. H., 2002: An investigation of the convective region of numerically simulated squall lines. Ph.D. thesis, The Pennsylvania State University, 181 pp.

    • Search Google Scholar
    • Export Citation
  • Bryan, G. H., , and J. M. Fritsch, 2002: A benchmark simulation for moist nonhydrostatic numerical models. Mon. Wea. Rev., 130 , 29172928.

    • Search Google Scholar
    • Export Citation
  • Caracena, F., , and M. W. Maier, 1987: Analysis of a microburst in the FACE Meteorological Mesonetwork in southern Florida. Mon. Wea. Rev., 115 , 969985.

    • Search Google Scholar
    • Export Citation
  • Chappel, L., 2001: Assessing severe thunderstorm potential days and storm types in the tropics. Proc. Int. Workshop on the Dynamics and Forecasting of Tropical Weather Systems. Darwin, NT, Australia, Australian Meteorological and Oceanographical Society, 1–11.

    • Search Google Scholar
    • Export Citation
  • Clark, T. L., 1979: Numerical simulations with a three-dimensional cloud model: Lateral boundary condition experiments and multicellular severe storm simulations. J. Atmos. Sci., 36 , 21912215.

    • Search Google Scholar
    • Export Citation
  • Colon, J. A., 1953: The mean summer atmosphere of the rainy season over the western tropical Pacific Ocean. Bull. Amer. Meteor. Soc., 34 , 333334.

    • Search Google Scholar
    • Export Citation
  • Gilmore, M. S., , and L. J. Wicker, 1998: The influence of midtropospheric dryness on supercell morphology and evolution. Mon. Wea. Rev., 126 , 943958.

    • Search Google Scholar
    • Export Citation
  • Gilmore, M. S., , J. M. Straka, , and E. N. Rasmussen, 2004: Precipitation and evolution sensitivity in simulated deep convective storms: Comparisons between liquid-only and simple ice and liquid phase microphysics. Mon. Wea. Rev., 132 , 18971916.

    • Search Google Scholar
    • Export Citation
  • Jordan, C. L., 1958: Mean soundings for the West Indies area. J. Meteor., 15 , 9197.

  • Klemp, J. B., 1987: Dynamics of tornadic thunderstorms. Annu. Rev. Fluid Mech., 19 , 369402.

  • McCaul Jr., E. W., , and M. L. Weisman, 2001: The sensitivity of simulated supercell structure and intensity to variations in the shapes of environmental buoyancy and shear profiles. Mon. Wea. Rev., 129 , 664687.

    • Search Google Scholar
    • Export Citation
  • McCaul Jr., E. W., , C. Cohen, , and C. Kirkpatrick, 2005: The sensitivity of simulated storm structure, intensity, and precipitation efficiency to environmental temperature. Mon. Wea. Rev., 133 , 30153037.

    • Search Google Scholar
    • Export Citation
  • Rotunno, R., 1981: On the evolution of thunderstorm rotation. Mon. Wea. Rev., 109 , 577586.

  • Weisman, M. L., , and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110 , 504520.

    • Search Google Scholar
    • Export Citation
  • Weisman, M. L., , and J. B. Klemp, 1984: The structure and classification of numerically simulated convective storms in directionally varying wind shears. Mon. Wea. Rev., 112 , 24792498.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Wind profiles as defined by Eq. (2) with zs = 3 km for Us = (15, 25, 35, 45) m s−1.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 2.
Fig. 2.

Skew-T diagram depicting temperature and moisture profiles used in the (a) qυ0 = 14 g kg−1 midlatitude (dashed lines) and Darwin (solid lines) model experiments and in the (b) Jordan (solid lines) and Colon (dashed lines) model experiments. The Darwin profile plotted here is an average profile over days when severe and nonsevere storms occurred (Davg; YYMMDD = 051114, 031126, 060207).

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 3.
Fig. 3.

Vertical cross sections of the model-generated storm for the Us = 0 m s−1 experiment for (a) the midlatitudes and (b) the tropics at a time when the cloud-top height is at a maximum. Vectors represent wind, the thick line is the 0.1 g kg−1 contour of cloud water, and the mixing ratio qυ is represented by the thin lines contoured at 4, 8, 12, and 16 g kg−1.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 4.
Fig. 4.

Horizontal cross section through the supercell propagating to the right of the mean wind (a) 47 and (b) 70 min after model initialization. A mirror image storm propagates to the left (not shown). The model is initialized with the Colon profile and Us = 30 m s−1. The vertical velocity at midlevels (4.6 km) is contoured every 2 m s−1. Vectors represent storm-relative low-level (175 m) horizontal winds (a mean storm speed of 17 m s−1 was subtracted from the initial field). The surface rain field is indicated by the shaded region and represents the +0.1 g kg−1 perturbation contour, whereas the surface gust front is denoted by the single thick line and represents the −0.5-K temperature perturbation contour.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 5.
Fig. 5.

As in Fig. 4, but initialized with the Jordan profile.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 6.
Fig. 6.

Plotted are (a) CAPE vs shear magnitude Us, (b) CAPE8 km vs Us, and (c) wmax vs Us. Squares (diamonds) represent the midlatitude (tropical) cases in which no splitting or an incomplete split occurs. Hourglasses (triangles) represent the midlatitude (tropical) cases in which complete splitting with or without intensification occurs.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 7.
Fig. 7.

Speed of the gust front υgf (in the north–south direction) vs wind shear Us for models in which the microphysics of (a) warm and (b) cold clouds (Bice) is used. The large circles represent the cases in which splitting occurred. The lowermost three curves in each plot are the midlatitude cases.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 8.
Fig. 8.

Time of gust front appearance tgf minus the time of the occurrence of the maximum updraft velocity twmax vs wind shear Us for Bryan’s model.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 9.
Fig. 9.

Time of gust front appearance tgf minus time of splitting tsplit (in minutes) vs wind shear Us for all midlatitude and tropical cases simulated with Clark’s model. White dots represent the incomplete splits; gray dots represent the complete splits without intensification; black dots represent splitting with intensification.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 10.
Fig. 10.

Maximum of total liquid water vs peak minimum downdraft speed wmin taken at cloud base for all the midlatitude and tropical cases simulated with Bryan’s model. Within every model, Us is varied giving different points in the max(qr + qc) − wmin phase space, which are connected. The cases with large midtropospheric relative humidity (Colon, mid-12, mid-14, mid-16, DavgWK, D011120WK) are marked with large symbols.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 11.
Fig. 11.

Maximum of total liquid water vs water vapor mixing ratio averaged over the lowermost 2 km when the models are run with (a) Bryan’s and (b) Clark’s model.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 12.
Fig. 12.

Maximum of total liquid water vs wind shear magnitude Us for the results obtained with Bryan’s model.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 13.
Fig. 13.

Storm depth vs wind shear magnitude Us for the results obtained with Bryan’s model.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 14.
Fig. 14.

Time series cross section of (a) midlevel and (b) low-level vertical vorticity maximum (in 10−4 s−1) for the qυ0 = 14 g kg−1 midlatitude, Darwin (average), Jordan, and Jordan WK experiment simulated with Clark’s model using Us = 35 m s−1.

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Fig. 15.
Fig. 15.

Maximum updraft speed as a function of time for the qυ0 = 14 g kg−1 midlatitude, Darwin (average), and Colon experiments, in which Us is set to 35 m s−1. These cases are simulated with Bryan’s model using both the Kessler scheme (mid-14, Colon, Davg) and ice microphysics (mid-14 ice, Colon ice, Davg ice).

Citation: Journal of the Atmospheric Sciences 66, 8; 10.1175/2009JAS2963.1

Table 1.

Severe thunderstorm warning statistics for the seasons 2002/03–2005/06. For each season, the number of warned events issued by the Bureau of Meteorology in Darwin, successful warnings, missed events, and false alarms is given. Also shown are the probability of detection (POD) calculated as the ratio of successful warnings to all events and the false alarm ratio (FAR) calculated as the ratio of false alarms to the sum of successful warnings and false alarms.

Table 1.
Table 2.

Comparison of various parameters set in Bryan’s and Clark’s models and chosen in WK82.

Table 2.
Table 3.

Values of total CAPE, CAPE8 km (calculated from the surface to a height of 8 km), and maximum updraft speed wmax for the different models initialized with Us = 0 m s−1. Here wmax is given in m s−1 and CAPE is given in J kg−1. The abbreviations of the model names can be found in parentheses.

Table 3.

1

The Bureau of Meteorology in Australia classifies as severe those storms that produce strong winds with speeds of 90 km h−1 or greater, tornadoes, flash flooding, and/or hail with a diameter of 2 cm or larger.

2

CAPE is calculated here using pseudoadiabatic process and Bolton’s formula [Eq. (43) in Bolton 1980] for the equivalent potential temperature.

3

In reality such a temperature excess within the mixed layer would not be observed. However, because surface heating is not included here, convective motions within the mixed layer that aid deep convective development are thus not modeled. By trial and error, the value of 8 K was chosen so that storms are consistently produced in both models.

4

“Successfully predicted” means that the value of R (or R8 km) for the split cases lies within the plotted thresholds for split storms in Fig. 6a or 6b and outside for the cases that do not split.

5

The ice level is defined here as the height z at which qi > 0.1 g kg−1.

6

A physically based technique for assessing potential downdrafts when significant precipitation reaches the ground (Caracena and Maier 1987).

7

The shear between a low-level (0–0.6 km) and a midlevel (2.1–5.6 km) wind vector. The maximum low–midlevel wind shear is assessed from the 0000 UTC sounding and may be modified for expected changes prior to storm initiation.

Save