Downloaded 10/04/21 05:16 PM UTC 6674 JOURNAL of CLIMATE VOLUME 27 Directly Proportional to the Vertical Gradient of the Heat- More Elevated Divergence Proﬁles

1SEPTEMBER 2014 H O M E Y E R E T A L . 6673

Assessing the Applicability of the Tropical Convective–Stratiform Paradigm in the Extratropics Using Radar Divergence Proﬁles

CAMERON R. HOMEYER National Center for Atmospheric Research,* Boulder, Colorado

COURTNEY SCHUMACHER Department of Atmospheric Sciences, Texas A&M University, College Station, Texas

LARRY J. HOPPER JR. Department of Atmospheric Sciences, School of Sciences, University of Louisiana at Monroe, Monroe, Louisiana

(Manuscript received 16 September 2013, in ﬁnal form 15 May 2014)

ABSTRACT

Long-term radar observations from a subtropical location in southeastern Texas are used to examine the impact of storm systems with tropical or extratropical characteristics on the large-scale circulation. Clima- tological vertical profiles of the horizontal wind divergence are analyzed for four distinct storm classifications: cold frontal (CF), warm frontal (WF), deep convective upper-level disturbance (DC-ULD), and nondeep convective upper-level disturbances (NC-ULD). DC-ULD systems are characterized by weakly baroclinic or equivalent barotropic environments that are more tropical in nature, while the remaining classifications are representative of common midlatitude systems with varying degrees of baroclinicity. DC-ULD systems are shown to have the highest levels of nondivergence (LND) and implied diabatic heating maxima near 6 km, whereas the remaining baroclinic storm classifications have LND altitudes that are about 0.5–1 km lower. Analyses of climatological mean divergence profiles are also separated by rain regions that are primarily convective, stratiform, or indeterminate. Convective–stratiform separations reveal similar divergence characteristics to those observed in the tropics in previous studies, with higher altitudes of implied heating in stratiform rain regions, suggesting that the convective–stratiform paradigm outlined in previous studies is applicable in the midlatitudes. Divergence profiles that cannot be classified as primarily convective or stratiform are typically characterized by large regions of stratiform rain with areas of embedded convection of shallow to moderate extent (i.e., echo tops ,10 km). These indeterminate profiles illustrate that, despite not being very deep and accounting for a relatively small fraction of a given storm system, convection dominates the vertical divergence profile and implied heating in these cases.

1. Introduction vertical structure of heating play an important role in the large-scale circulation through the generation of poten- Atmospheric motions can occur, in part, because of tial vorticity anomalies. Houze (1982) observed that the the nonuniformity of diabatic heating processes associ- maximum heating (composed of latent, radiative, and ated with precipitating cloud systems. This concept is eddy sensible components) in stratiform rain regions oc- especially relevant in the tropics where variations in the curs at higher altitudes than in convective rain regions. Hartmann et al. (1984) used this observation to show that * The National Center for Atmospheric Research is sponsored the elevated heating proﬁle associated with a mesoscale by the National Science Foundation. convective system (MCS) composed of both convective and stratiform precipitation regions was able to produce a much more realistic tropical dynamical response than Corresponding author address: Cameron Homeyer, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO a convective-only proﬁle. Haynes and McIntyre (1987) 80305. and Mapes and Houze (1995) helped explain this result E-mail: [email protected] by showing that the generation of potential vorticity is

DOI: 10.1175/JCLI-D-13-00561.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6674 JOURNAL OF CLIMATE VOLUME 27 directly proportional to the vertical gradient of the heat- more elevated divergence profiles. The fidelity of the ing profile in convective systems in the tropics. Many model simulations was dependent on the organization of studies since Hartmann et al. (1984) have examined how the system, which was in turn dependent on the baro- the vertical structure of tropical heating influences the clinicity of the environment. Model results were also dynamical response (e.g., DeMaria 1985; Sui and Lau sensitive to the microphysics and convective parame- 1989; Wu et al. 2000; Chiang et al. 2001; Schumacher et al. terizations and whether convection was parameterized 2004). on an intermediate 9-km nested grid. Both studies also Diabatic heating variations also play a role in the suggested that MCSs could have a significant feedback large-scale circulation in the midlatitudes, but their con- with the large-scale circulation at higher latitudes in tributions are not as well quantified or understood. Be- both warm-season and weakly baroclinic environments. ginning with Ninomiya (1971a), case studies have shown This study uses long-term radar observations in that midlatitude MCSs that cover large regions for ex- southeastern Texas, which experiences a diverse spec- tended periods or that occur in weakly baroclinic envi- trum of precipitating systems common in the tropics and ronments can have an active dynamic feedback with the midlatitudes, to evaluate the dynamical characteristics larger-scale circulation through vertical variations in dia- of storms based on their organization and synoptic batic heating, especially in the trailing stratiform rain re- forcing. This study’s overarching goal is to provide gion of an MCS (e.g., Hertenstein and Schubert 1991). a more quantitative context for assessing mesoscale– Many other midlatitude studies indicate that large con- synoptic interactions and the role precipitating sys- vective systems affect the local large-scale environment tems may have in large-scale flow outside of the by creating a large anticyclonic flow perturbation aloft, tropics. We do this by utilizing a multiyear subtropical intensifying upper-level divergence, and enhancing the dataset to observationally constrain the divergence large-scale baroclinicity (Ninomiya 1971b; Maddox et al. (and implied heating profiles) that may be used in fu- 1981; Raymond and Jiang 1990; Zhang and Harvey 1995; ture studies. The results have implications for the Stensrud 1996). Stensrud and Anderson (2001) further modeling and forecasting of a variety of extratrop- argued that regions of persistent midlatitude convection ical storms while also evaluating how applicable the can affect the hemispheric circulation and Chang et al. convective–stratiform paradigm (see Houze 1997)isat (2002) described how diabatic heating from condensa- higher latitudes. tional processes strengthens midlatitude storm tracks. In addition, horizontal variations in diabatic heating within frontal storm systems have been shown to strengthen 2. Methods frontogenesis (Bryan and Fritsch 2000; Wakimoto and a. Storm classifications Murphey 2008). However, very little research exists be- yond individual case studies on the relationship between Four classifications are used to categorize storms precipitating systems and the large-scale circulation out- depending on the primary mechanism responsible for side of the tropics. initiating convective or nonconvective precipitation. Hopper and Schumacher (2009, 2012) attempted to Surface and upper-air maps, satellite images, and sur- rectify this void by using a larger set of cases and a com- face radar reflectivity images from the online archive bination of mesoscale modeling and observations to an- maintained by the National Center for Atmospheric alyze microphysical and dynamical differences between Research (NCAR) Mesoscale and Microscale Meteo- storms of varying organization [e.g., leading-line trailing- rology Division (online at http://www.mmm.ucar.edu/ stratiform (LLTS) MCSs versus less organized systems] imagearchive/) have been matched with the indepen- occurring in a range of synoptic conditions (i.e., warm dent radar data used in this study and analyzed to season, weakly baroclinic, and strongly baroclinic) in determine each storm’s classification. Archives of me- southeastern Texas. Microphysical comparison metrics soscale discussions and mesoanalyses from the National included stratiform rain production and vertical profiles Weather Service Storm Prediction Center and vertical of reflectivity, while dynamical metrics included vertical cross sections of radar data are also used in classifying velocity and divergence profiles. Divergence is especially some storms. In this study, we consider that stratiform useful because it can be linked back to diabatic heating rain can originate from 1) deep, vertically oriented profiles (e.g., Mapes and Houze 1995) and is more easily convective sources (deep convection) and 2) synoptic- observable than vertical velocity and diabatic heating. scale lifting without a deep convective source that does Hopper and Schumacher (2009, 2012) found that storms not exclude slantwise, elevated, or shallow convection occurring in less baroclinic environments have more (nondeep convection). The composite schematics illus- convective rain area, less stratiform rain production, and trated in Fig. 1 depict representative cases for these

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 1SEPTEMBER 2014 H O M E Y E R E T A L . 6675

FIG. 1. Composite figures of representative cases for (a) cold frontal, (b) warm frontal, (c) deep convective upper- level disturbance (ULD), and (d) nondeep convective ULD storm classifications. All images are adapted from an online image archive maintained by the Mesoscale and Microscale Meteorology Division of NCAR. The blue and red frontal boundaries in (a),(b) represent surface cold and warm fronts, respectively. classifications whose broader spectrum of background associated with a midlatitude cyclone (Matejka et al. environments are described below: 1980; Herzegh and Hobbs 1980; Houze et al. 1981). Embedded convection is typically of a slantwise or d Cold frontal (CF; Fig. 1a) or trough precipitation elevated nature, but deep convection may occur near initiates along a surface cold front associated with and parallel to the surface warm front. Isentropic a midlatitude cyclone, dryline (Schaefer 1974), pre- lifting and propagating shortwave troughs are also frontal trough or wind shift (Schultz 2004), baroclinic often present in these cases but are not a necessity. surface trough (Sanders 2005), or convectively in- d Deep convective upper-level disturbance (DC-ULD; duced outflow boundaries associated with a dissipating Fig. 1c) precipitation initiates in the presence of cold front. Deep convection along the leading edge of a stationary or propagating midlevel circulation these boundaries is driven by strong surface conver- (700–500-hPa closed low or shortwave trough) that gence, but regions of nondeep convective precipita- does not have an associated surface cold or warm tion associated with frontal lifting may also occur front. Precipitation in these cases may also form below along the upper boundary of the frontal zone (e.g., jet streak circulations, but deep convection must be Matejka et al. 1980; Hobbs et al. 1980). present for this classification that typically includes d Warm frontal (WF; Fig. 1b) precipitation typically mesoscale, warm-season circulations. includes widespread precipitation resulting from d Nondeep convective upper-level disturbance (NC- synoptic-scale frontal lifting or mesoscale updrafts ULD; Fig. 1d) precipitation includes the subset of ULDs on the cool side of an advancing surface warm front whose stratiform precipitation does not originate from

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6676 JOURNAL OF CLIMATE VOLUME 27

a deep convective source and may include storms with across the climatology have been crosschecked for in- slantwise or elevated convection. In these cases, pre- ternal consistency. cipitation results from synoptic-scale lifting associated b. Radar observations and analysis with positive differential vorticity advection induced by midlevel circulations or jet streaks and occasionally in Climatological mean velocity–azimuth display (VAD) combination with weak warm air advection at middle wind profiles and horizontal wind divergence are esti- levels. mated for each storm classification using 4 yr of data (June 2006–July 2010) from Texas A&M University’s A few possible storm classifications have been com- S-band Aggie Doppler Radar (ADRAD) located in bined with other storm types or omitted from analysis College Station, Texas (30.68N, 96.38W). ADRAD typi- in this study. Named tropical cyclones are excluded cally performs full azimuthal scans at 24 unique elevation because of their infrequency relative to other storm angles ranging from 0.48 to 29.58 every 12 min. The azi- 2 types. Discrete and/or weakly forced storms initiating muthal scan rate for ADRAD is 188 s 1, so each 3608 from forcing mechanisms not described above (e.g., elevation scan takes 20 s to complete. An extensive de- supercell thunderstorms, airmass thunderstorms, and scription of ADRAD’s scan strategies, calibration shifts, sea-breeze convection) have also been omitted because and quality control methods utilized during this period their echo coverages are typically too low to generate are provided in Hopper and Schumacher (2012). reliable divergence estimates. Excluding tropical cy- Hourly VAD wind and divergence profiles are com- clones and discrete/weakly forced storms should not puted using the technique of Browning and Wexler significantly alter climatological divergence profiles (1968) and following the methods in Mapes and Lin because they each account for only 2%–3% of annual (2005) with some modifications outlined here. The re- rainfall over the radar domain, based on a climatology sults in this study are computed at a vertical resolution of from March 2002 to February 2010 (Hopper 2011). In 25 hPa, twice that of Mapes and Lin (2005), whose addition, stationary frontal storms are classified as ei- methods are subject to biases from nonuniformly dis- ther warm or cold frontal, depending on whether their tributed echo coverage in azimuth and at close ranges convective elements propagate away from the cold from the radar. In the method, divergence profiles are front into the warm sector (i.e., cold frontal) or move estimated at select ranges from the radar using radial parallel to the front or into the cool sector (i.e., warm velocity observations within 8-km bins centered on each frontal). Splitting stationary frontal storms and merg- annulus. These divergence calculations are representa- ing several types of surface boundaries into the cold tive of the dynamical characteristics of the entire storm frontal category simplifies the climatology while com- system within the farthest range considered from the bining storm types that have similar dynamical and mi- radar. In this study, we use divergence estimates at three crophysical properties indicated by detailed analyses of the aforementioned 8-km-wide annuli spanning 32– of radar divergence estimates shown in this study and 56 km in range in order to avoid biases in the divergence disdrometer-based observations of drop size distributions calculation at close ranges from the radar and to provide (not shown). an estimate of the uncertainty of each hourly calcula- In cases where multiple forcing mechanisms are tion. To avoid errors resulting from incomplete azimuthal present, frontal classifications are given precedence coverage, we retain hours with storm area fractions over ULD classifications unless there is a spatial break $80% in the 32–56-km range space and at all altitudes up (.100 km) between precipitation associated with a to 8 km for analysis. A threshold of 80% in storm area ULD and the location where the frontal precipitating coverage was found to ensure near-concentric echo cov- system initiated downstream. One example illustrating erage from inspection of individual cases. this occurrence is the NC-ULD depicted in Fig. 1d It is important to note here that adiabatic contribu- whose precipitation associated with large-scale lifting tions to the divergence profiles (i.e., dry dynamics), east of an upper-level low is clearly separated from which could impact their vertical structure and implied precipitation that initiated along a warm front down- heating, are not diagnosed in this study. However, based stream from Louisiana to the Florida Panhandle (not on the scaling arguments given in Mapes and Houze shown). ULD classifications are also used if evapora- (1995), such contributions to radar-observed divergence tive cooling associated with precipitation initially pro- are likely negligible in most cases. In particular, signifi- duced by an ULD induces the formation of a mesoscale cant adiabatic contributions will occur for MCSs with baroclinic frontal zone that is not connected to a sur- tropospheric temperature perturbations comparable to face cyclone. Although some degree of subjectivity is their latent heating rates (i.e., strongly baroclinic frontal inherent in qualitative classification systems, storms systems) and low precipitation amounts (,1 cm), but the

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 1SEPTEMBER 2014 H O M E Y E R E T A L . 6677

TABLE 1. The number of hours, number of storms, convective rain fractions, and the 0–8-km mean standard deviations of hourly divergence estimates (shr) and climatological mean divergence proﬁles (sm) for each storm classiﬁcation and RRP: convective (C), stratiform (S), indeterminate (I), and the total (All).

24 21 24 21 Class RRP Hours Storms Convective rain (%) shr (10 s ) sm (10 s ) CF C 83 32 63.6 0.34 0.66 S 61 25 11.5 0.25 0.67 I 20 13 16.0 0.23 0.49 All 164 47 38.4 — — WF C 16 7 59.6 0.34 0.55 S 22 7 10.4 0.15 0.44 I 10 4 11.5 0.30 0.47 All 48 13 27.0 — — DC-ULD C 44 20 58.7 0.21 0.58 S 49 18 12.6 0.15 0.32 I 19 8 19.3 0.24 0.45 All 112 22 31.8 — — NC-ULD C — — — — — S 23 7 7.7 0.30 0.51 I 12 6 12.0 0.22 0.45 All 35 9 9.2 — —

diabatic component of the divergence profile typically from the radar. Only rain periods with at least 100 2 remains dominant in such cases (e.g., Hopper and drops and rain rates of at least 0.1 mm h 1 are used to Schumacher 2012). For the storm classifications in this compute the Z–R regression. study, NC-ULD systems would have the highest likeli- The storm coverage criteria outlined above yield 359 h hood of meeting the low precipitation criteria. However, of ADRAD data for analysis. CF systems account for the storm coverage criteria used to limit the analysis of nearly half of the hourly observations (164), followed by divergence profiles likely ensures that precipitation DC-ULD (112), WF (48), and NC-ULD (35) systems. In amounts are sufficiently large in all cases and thus fur- addition, convective–stratiform separations and associ- ther limits contributions from dry dynamics. ated radar reflectivity-based rainfall estimations show Radar-derived wind and divergence profiles are sep- that CF, DC-ULD, and WF systems have the highest arated by the storm classifications given in section 2a climatological convective rain fractions (27%–38%), and by the fraction of convective rainfall for analysis. whereas NC-ULD systems show convective rain frac- Separating profiles by the fraction of convective area tions ,10% (see also Table 1). Because of decreasing coverage rather than rainfall produces similar results. storm coverage at higher altitudes from the radar, un- To identify convective and stratiform rain regions at certainties in the hourly estimates of winds and di- each radar observation time, we follow methods out- vergence often become larger than their observed lined in Steiner et al. (1995) and Yuter and Houze (1997) magnitudes. Therefore, the following analyses are re- and updated for southeastern Texas in Hopper and stricted to altitudes below 10 km. Schumacher (2012). Rainfall estimates for each radar It is well known that divergence profiles for convec- observation are obtained using a well-documented method tive and stratiform rain regions can be related to atmo- of regressing low elevation radar reflectivity Z against spheric heating and vertical motion through mass ground-based observations of rain rate R to develop a Z–R continuity. In particular, positive vertical gradients of relationship (e.g., Marshall and Palmer 1948). The Z–R divergence typically represent heating and ascent while relationship has the form negative gradients represent cooling and descent. Con- sequently, peak diabatic heating and cooling rates in Z 5 aRb , (1) the vertical are often observed near a level of nondivergence (LND, where divergence is 0) because LNDs where the factors a and b are 177.3 and 1.66 in this study, are representative of the central altitude of a gradient respectively. We derive a and b using observations layer. Convective rain regions are typically character- during the entire period of this study from ADRAD ized by low-level convergence that transitions to di- and ground-based drop size distributions from two vergence at upper levels in the detraining anvil layer Joss–Waldvogel (J-W) impact disdrometers (Joss and (e.g., Mapes and Houze 1993), representing significant Waldvogel 1967) at about 4.9 and 165.9 km in range heating in the middle troposphere. Stratiform rain

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6678 JOURNAL OF CLIMATE VOLUME 27 regions are characterized by low-level divergence, midlevel convergence coincident with the melting (freezing) level, and upper-level divergence. This vertical divergence structure for stratiform rain is associated with significant cooling in the lower troposphere and heating in the upper troposphere, which typically occurs at higher altitudes than in convective rain regions. Because of these significant differences in divergence (and heating), it is desirable to develop an understanding of the characteristic contributions from convective and stratiform rain regions. Although complex methods to wholly separate convective and stratiform contributions to divergence are useful [e.g., linear regressions of divergence profiles versus reflectivity-estimated surface rain rates in Mapes and Lin (2005)], contributions from convective and stratiform rain regions can be characterized through identification of the primary rain region (RRP)foreach hourly observation.

We identify the RRP using a simple approach that compares the altitude of maximum convergence with the fraction of total rainfall from convective rain regions within 56 km in range of the radar for all analyzed hours FIG. 2. Scatterplot of the estimated altitudes of peak conver- and storm classifications (Fig. 2). Although this ap- gence vs the fraction of convective rain between 32 and 56 km in range from the radar for all hours and storm classifications. Gray proach does not result in a pure separation of each rain contours show a joint frequency distribution of the fraction of the region, it does isolate times in the life cycle of a storm total number of observations in 20% 3 1 km bins (1% contours that are dominated by convective or stratiform rain and ranging from 2% to 5%). Red horizontal and vertical lines de- thus presumably heating (e.g., Johnson 1984; Lin et al. marcate rain regions characterized as primarily convective, pri- 2004). The contoured joint frequency distribution in marily stratiform, or indeterminate. Fig. 2 clearly illustrates these convective and stratiform modes of rainfall and divergence. In particular, profiles altitudes of maximum convergence and low convective dominated by stratiform rain correspond to a frequency rain fractions that are not consistent with either pre- maximum at altitudes of peak convergence from 2.5 to dominately stratiform or convective processes, and they 6 km and convective rain fractions from 0% to 25%. are therefore classified as indeterminate here.

Alternatively, proﬁles dominated by convective rain are Further inspection of indeterminate RRP proﬁles illustrated by a frequency maximum at convective rain from Fig. 2 illustrates that contributing hourly ob- fractions from 40% to 80% and altitudes of peak con- servations are often characterized by predominantly vergence from 0 to 3 km. These observed modes in the stratiform rain regions throughout the ADRAD relationship between convective rainfall and the altitude domain with embedded convection that the two- of maximum convergence allow us to characterize three dimensional convective–stratiform separation algo-

RRP: convective (convective rainfall .40%), stratiform rithm used in this study has difficulty identifying (convective rainfall ,40% and maximum convergence accurately. These indeterminate profiles represent an altitudes above 2.5 km), and indeterminate (convective often-complex mixture of convective and stratiform rainfall ,40% and maximum convergence altitudes rain regions, though convection typically dominates below 2.5 km). An altitude threshold of 2.5 km for con- the divergence structure, providing the low altitudes of vergence maxima here corresponds to the climatological maximum convergence observed. The vast majority altitude minimum of the melting level in southeastern of the convection during these indeterminate hours Texas determined from 10 yr of radiosonde data near is relatively shallow (i.e., less than 10 km deep and the radar domain (not shown). Analysis of each storm vertical extent of radar reflectivity .35 dBZ less than classification separately shows comparable altitude 2 km above the melting level). However, slantwise and transitions to that identified in Fig. 2 (not shown). Al- elevated convection that is also shallow but forms be- though the identified stratiform and convective RRP cause of large-scale lifting instead of or in addition to follow expected relationships from previous studies, gravitational instability may contribute in some cases there are a significant number of observations with low as well.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 1SEPTEMBER 2014 H O M E Y E R E T A L . 6679

Column-maximum radar reflectivity and a vertical cross section of radar reflectivity observations of a representative storm system with an indeterminate RRP are shown in Fig. 3. These radar observations are from three-dimensional composites of several radars from the Next Generation Weather Radar (NEXRAD) program Weather Surveillance Radar-1988 Doppler (WSR-88D) network (Crum and Alberty 1993) created following the methods outlined in Homeyer (2014). The vertical cross section (Fig. 3b), taken through the area within 56 km in range from ADRAD, shows the embedded shallow to moderate convection within the larger stratiform rain region. In addition to embedded convection in stratiform rain, there are a few indeterminate RRP profiles observed during hours of transition from convective to stratiform rain in LLTS MCS systems that show large fluctuations in divergence at lower altitudes. During these transition hours, convergence maxima at low altitudes are produced by descending midlevel rear inflow jets across the radar domain (e.g., see Fig. 5c in Hopper and Schumacher 2012). Although these profiles represent observations of systems with dynamical characteristics that differ from convection embedded in broad areas of stratiform rain, they account for less than 10% of the analyzed profiles and excluding them from the climatological analysis does not alter the results.

3. Results To examine the fidelity of the storm classifications used and the characteristic storm motion for each case, Fig. 4 shows VAD profiles of mean zonal, meridional, and total wind speeds and their direction. The near- surface wind directions for the CF (northwesterly) and FIG. 3. Three-dimensional composite NEXRAD WSR-88D ob- WF (southeasterly) classifications are representative of servations of (a) column-maximum radar reflectivity and (b) a ver- their characteristic surface wind speeds near their re- tical cross section along AB in (a) for a storm system at 1710 UTC 3 Jun 2010 that is identified as indeterminate in ADRAD observa- spective boundaries and propagation for the CF cases in tions in Fig. 2. The location of ADRAD and a contour of 56 km in the study region. In addition, all storm classifications are range from the radar location are shown by the black cross symbol characterized by southwesterly flow at middle and upper and circle in the map, respectively. In (b), the boundaries of the levels, except for the DC-ULD cases, which have nearly 56-km ADRAD range are shown by the thick black tick marks due southerly flow. There are also clear distinctions along the bottom axis. between the vertical structure of the wind profiles’ magnitudes for each storm classification. CF systems are unidirectional weak winds. NC-ULD systems show total characterized by strong upper-level winds (.8 km) and wind speed structures similar to WF systems with the deepest speed shear (from 0 to 8 km in Fig. 4c), weaker magnitudes at middle and upper levels and have whereas WF systems are characterized by the strongest the weakest lower-level winds (,3 km) of all storm winds and speed shear at middle levels (from 3 to 7 km). classifications. In addition, NC-ULD systems also have DC-ULD systems are characterized by strong speed the weakest boundary layer shear and instead display shear in the boundary layer like CF and WF storms strong speed shears at 3 km (;700 hPa). These con- but contain the weakest winds at middle and upper trasting features make sense considering most analyzed levels that are weakly veering with height, represen- NC-ULD events are associated with upper-level jet tative of weakly baroclinic systems or tropical systems streaks and/or midlevel shortwaves without organized that in some cases may be equivalent barotropic with surface fronts.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6680 JOURNAL OF CLIMATE VOLUME 27

FIG. 4. Mean (a) zonal wind speed, (b) meridional wind speed, (c) total horizontal wind speed, and (d) wind direction for cold frontal (blue lines), warm frontal (red lines), deep convective upper-level disturbance (black lines), and nondeep convective upper-level disturbance (green lines) storm classiﬁcations.

Seasonal distributions of the number of hourly ob- diagnosing the large-scale influence of each storm classi- servations for each storm classification are shown in fication. For each RRP and storm classification, the range Fig. 5, and provide further support for the accuracy of of LNDs span layers $5 km. In addition, the stratiform the storm classifications used. All storm classifications LND distributions for each storm classification (Fig. 6b) show distributions consistent with the seasonality in are at higher altitudes than convective RRP (Fig. 6a), their forcing mechanisms outlined in section 2a. CF and consistent with previous studies in the tropics (e.g., WF systems occur primarily in the transition and winter Mapes and Houze 1993). CF systems show the broadest seasons, although CF storms still occur during summer distribution of the LND for each RRP, frequently whereas WF storms do not. NC-ULD systems occur reaching lower altitudes than WF and DC-ULD sys- during the transition and winter seasons with peak oc- tems. The stratiform LND distribution for NC-ULD currence in fall and winter, when their primary forcing systems shows a distinct peak at altitudes much lower mechanisms are common in southeastern Texas. DC- than that observed for the remaining storm classifica- ULD systems are dominated by events in the summer tions, consistent with their typical occurrence in the cold months, with all remaining events distributed in the season when the melting level (and consequently the transition seasons. The seasonality for DC-ULD events LND) reaches its lowest altitude. LND distributions are is consistent with their more tropical (or barotropic) centered at the lowest altitudes for indeterminate RRP characteristics. Storm systems in midlatitudes and the environment they occur within show large seasonality not observed in the tropics. As identified in long-term radiosonde observations near southeastern Texas discussed in section 2b, the observed seasonality in the melting level generally spans altitudes from about 2.5 km in the winter to 5.5 km in the summer. The tropopause also shows significant range from near 10 km in the cold season to near 16 km in the warm season. This seasonality directly affects the vertical extent of storms and their divergence structures. Figure 6 shows one aspect of this seasonality: frequency distributions of the LND for convective and IG indeterminate RRP and the upper LND for stratiform F . 5. Seasonal distributions of the fraction of hourly observations used in divergence calculations for cold frontal (blue), warm RRP for each storm classification. Because the LND frontal (red), deep convective upper-level disturbance (black), and in these distributions often implies the altitude of max- nondeep convective upper-level disturbance (green) storm classi- imum diabatic heating, developing an understanding of fications. For each storm classification, the sum of all seasons its seasonality is an important consideration for accurately equals 100%.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 1SEPTEMBER 2014 H O M E Y E R E T A L . 6681

LND altitude (e.g., WF convective RRP and NC-ULD indeterminate RRP). However, these bimodal distributions have some of the smallest sample sizes of any

classification and RRP and may not represent physical characteristics other than the range of LND altitudes. Figure 7 shows climatological mean divergence profiles by storm classification for hours of convective,

stratiform, and indeterminate RRP and the mean proﬁle for all hours contributing to each classiﬁcation. Because the freezing level and the LND show large seasonality

for each storm classiﬁcation and RRP, we compute mean divergence proﬁles in relative altitude to the LND for

convective and indeterminate RRP and the upper LND for stratiform RRP. Computing mean profiles in relative altitude retains the characteristic structures and ampli- tudes of convergence and divergence, including vertical gradients through the LND, whereas computing them in native altitude broadens the otherwise sharp transitions, biases the mean LND, and increases the uncertainty in the climatological divergence profile. For individual profiles whose LND is higher than the mean LND, por- tions of the profile near the surface that are shifted below ground are removed.

The RRP proﬁles in Fig. 7 are presented in relative altitude to the mean LND for each storm type (Figs. 7a–c), whereas the climatological mean of all observations for each storm classiﬁcation shown in Fig. 7d is a frequency-weighted mean of the relative altitude means

for each RRP. Mean uncertainties for the 0–8-km layer of the climatological mean divergence proﬁles (sm) and hourly divergence estimates (shr), in addition to the number of hourly observations used and mean convective area fractions, are given in Table 1. Following the

identiﬁcation of RRP in Fig. 2, there are a similar number of stratiform and convective hours for each classiﬁcation, while indeterminate hours account for the fewest number

of hours in each case. In addition, shr values are typically half as large as their respective sm values in each case, suggesting that the hourly divergence estimates are ro-

bust. Although uncertainties in sm are larger than shr, they are generally smaller than the maxima in conver-

FIG. 6. Frequency distributions of the level of nondivergence for gence and divergence in each profile. (a) primarily convective, (b) above the midlevel convergence in For convective RRP divergence profiles (Fig. 7a), CF primarily stratiform, and (c) indeterminate rain regions identified systems have the lowest LND altitude, in agreement in Fig. 2 for each storm classification: cold frontal (blue lines), with a cold or transition season surface-forced system. warm frontal (red lines), deep convective upper-level disturbance (black lines), and nondeep convective upper-level disturbance The LND is higher in convective profiles for WF and (green lines). For each distribution, LND altitudes are binned at DC-ULD systems, with WF systems showing the largest a resolution of 1 km and centered at integer kilometers. observed magnitudes of peak convergence and divergence and DC-ULDs showing the smallest magnitudes. (Fig. 6c), in agreement with the identification of rela- The vertical gradient of divergence (or slope) through tively shallow convection as the primary source for their the LND (and implied heating rates) is largest for the low levels of maximum convergence. In addition, there WF and CF systems and smallest for DC-ULD systems. are some distributions in Fig. 6 that show two modes of Although the magnitude of the vertical gradient is

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6682 JOURNAL OF CLIMATE VOLUME 27

FIG. 7. Profiles of mean storm divergence for (a) primarily convective, (b) primarily stratiform, and (c) indeterminate rain regions identified in Fig. 2 and (d) all rain regions for each storm classification: cold frontal (blue lines), warm frontal (red lines), deep convective upper-level disturbance (black lines), and nondeep convective upper-level disturbance (green lines). Note that the divergence scale is slightly different for (a). The number of hourly observations contributing to these mean profiles and related uncertainties are given in Table 1. For (a)–(c), mean divergence profiles are computed in relative altitude to the LND and scaled to the mean LND in each case. The divergence profiles in (d) are computed by weighting the scaled LND-relative profiles by frequency for each storm classification. lowest for the DC-ULD systems, their larger depth maximum heating. As observed in the distributions of suggests that convective heating and the related dy- the stratiform upper LND in Fig. 6b, the NC-ULD sys- namical response extends to higher altitudes than in CF tems show the lowest upper LND (;5 km) and altitude and WF systems. of maximum convergence (;4 km) of the mean strati-

Climatological divergence profiles for stratiform RRP form RRP profiles. The strongest divergence aloft is also (Fig. 7b) also show distinct differences between storm observed for NC-ULD systems, in agreement with fre- classifications. WF and DC-ULD systems show the quent jet streak forcing. Similar to convective RRP highest upper LND (;7 km) and levels of maximum profiles, DC-ULD systems for stratiform RRP show the convergence (;6 km), whereas the upper LND and weakest vertical gradient in divergence near the upper maximum convergence for CF systems are at least 1 km LND. Although WF systems show a slightly larger ver- lower, implying similar displacements in the altitude of tical gradient near the LND, the vertical depth of the

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 1SEPTEMBER 2014 H O M E Y E R E T A L . 6683 gradient is less than DC-ULD systems and is the shal- In addition to examining the dynamical contributions lowest gradient of the four classifications (from 6 to from each RRP, the climatological mean divergence 7.5 km). CF and NC-ULD systems show the strongest profiles for each storm classification (Fig. 7d) show their and deepest vertical divergence gradients in the strati- bulk characteristic influence in southeastern Texas. form RRP profiles, extending from about 4.5 to 7.5 km. These profiles are computed by combining each RRP In addition to characteristics of the LND and implied profile weighted by their fraction of hourly observations. heating structures, there are unique differences in the Similar to the profiles separated by rain region, the cli- characteristics of the midlevel convergence layer within matological means for all profiles show that DC-ULD stratiform rain regions between storm classifications. CF systems are characterized by the highest altitude of the systems are characterized by the strongest midlevel LND and implied heating, with the LND for WF, CF, convergence with a sharp peak near 5 km, consistent and NC-ULD systems about 0.5–1 km below DC-ULD with additional dynamical contribution from strong rear systems. The climatological mean profile for NC-ULD inflow jets in LLTS MCSs that are common in CF events. systems shows two maxima in convergence: one below The maximum in midlevel convergence for WF systems 2 km from the indeterminate climatology and the other is the broadest of any classification, spanning altitudes above 4 km from the stratiform climatology. In addition, from 3.5 to 6 km. This large depth likely reflects in- the vertical gradient for the NC-ULD systems is larger stantaneous contributions from both classes of strati- than the remaining systems through the LND, likely form rain production in WF systems outlined in section representative of the limited seasonality for NC-ULD 2a: production from surface-based and/or slantwise systems compared to the remaining classifications. The convection near or along the WF surface boundary re- magnitude of the vertical divergence gradient through sponsible for the lower maximum and production from the LND in the total climatological mean for the re- elevated convection farther north from the WF surface maining classifications is largest in the CF systems, fol- boundary responsible for the upper maximum. Midlevel lowed by WF and DC-ULD systems, while the vertical convergence is concentrated at the highest altitudes in depth of the divergent layer is comparable. However, DC-ULD cases, implying that DC-ULDs have the the altitudes of the mean LNDs and their implied heat- deepest lower-tropospheric diabatic cooling and most ing maxima are still higher and concentrate more strati- elevated diabatic heating profiles. form heating at upper levels (and lower-tropospheric

Divergence profiles of indeterminate RRP (Fig. 7c) cooling) in the DC-ULD cases compared to their more show moderate low-level convergence, with maxima baroclinic counterparts. Therefore, these DC-ULD sys- between 1.5 and 2.5 km in altitude, transitioning to similar tems are likely most capable of producing adiabatically magnitudes of upper-level divergence first peaking be- driven, slow-moving gravity waves in the subtropics that tween 4 and 5 km in altitude. As in the convective and warm the upper troposphere and cool the lower tropo- stratiform RRP divergence profiles, DC-ULD systems sphere, thus destabilizing the lower atmosphere and en- show the highest LND at an altitude near 4 km, whereas couraging at least shallow expansive convection to develop the LNDs for the remaining storm classifications are nearby (e.g., Mapes 1993; Mapes and Houze 1995). more than 1 km lower. The low altitudes of peak convergence and the LND in the indeterminate profiles 4. Summary and discussion provide further evidence that they are primarily associated with shallow to moderate convection that is not Significant differences in the dynamical characteristics adequately resolved by the two-dimensional convective– of storms from four distinct forcing mechanisms are stratiform separation algorithm. However, contributions observed for a subtropical site in southeastern Texas. from stratiform rain regions may still be important con- These four storm classifications are used to isolate in- sidering that upper-level divergence estimates are rela- fluences from systems with midlatitude (baroclinic) and tively large and become nearly constant with altitude at tropical (barotropic) characteristics. Cold frontal (CF) middle levels. The vertical divergence gradients through storms, warm frontal (WF) storms, and nondeep con- the LND for all storm classifications in the indeterminate vective upper-level disturbances (NC-ULD) show asso-

RRP proﬁles are comparable to the largest observed in ciation with strong mid- and upper-level winds and occur any RRP, despite being limited to lower altitudes. This primarily during winter and transition seasons, consistent characteristic suggests that shallow to moderate convec- with characteristics of midlatitude systems. Deep con- tion may play an important role in the large-scale cir- vective upper-level disturbances (DC-ULD) that occur culation at low levels in the extratropics, where the during the summer and transition seasons display weak environmental circulations are commonly weaker than at horizontal wind speeds at all altitudes, characteristic of upper levels. weakly baroclinic or more tropical systems.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6684 JOURNAL OF CLIMATE VOLUME 27

A simple method was introduced to identify analysis comparable to the tropical cumulus congestus described times when divergence profiles are dominated by con- in Johnson et al. (1999). However, its identification is vective or stratiform rain regions. In addition to con- hampered because it is often embedded in large strati- vective and stratiform classifications, an additional form regions, and variations due to the seasonality in group of observations did not show characteristics in environmental baroclinicity and melting levels require agreement with either classification and were labeled as additional consideration. Therefore, further improve- indeterminate. Divergence profiles separated by storm ments to classifying convective and stratiform rain re- classification and primary rain region (RRP) show gions from radar reflectivity fields are likely warranted. characteristics similar to previous studies that fully sep- Finally, one limitation of this study is that only data arate convective and stratiform components and illustrate from a single radar and a limited period of record are that the convective–stratiform paradigm is generally ap- used. The extensive network of operational radars in the plicable at higher latitudes. In particular, stratiform RRP continental United States could provide a more detailed divergence profiles show higher altitudes of implied understanding of the seasonality and variability of these heating than convective RRP for all storm classifications, heating profiles in midlatitudes for baroclinic systems in agreement with previous studies in the tropics. with similar characteristics. In particular, additional ob- The observed differences and variability in divergence servational studies are needed to develop an under- profiles for all storm classifications are largely dependent standing of the role convective heating plays when on the seasonality of the melting level and associated convection is embedded within larger regions of strati- levels of nondivergence (LND) and convergence max- form rain. Identifying relative contributions from shallow ima. Climatological mean RRP divergence profiles were to moderate and deep convection in these storm systems computed in relative altitude to the LND in order to re- will improve our understanding of the vertical distri- move biases introduced by the large seasonality in the bution of divergence, diabatic heating, and related feed- LND in each case. In general, DC-ULD systems were backs to convection and the large-scale circulation. This shown to have the highest altitudes of implied heating for topic is especially relevant to the recently launched Na- each RRP, typically followed in order by WF, CF, and tional Aeronautics and Space Administration (NASA) NC-ULD systems. In addition, NC-ULD systems showed Global Precipitation Measurement satellite mission, which the largest divergence gradients through the LND in the extends the objectives of the NASA Tropical Rainfall stratiform RRP and total climatological mean profiles, Measuring Mission to higher latitudes, including the esti- reflective of their limited seasonality compared to the mation of four-dimensional heating associated with pre- remaining storm classifications. Despite having a higher cipitating systems (e.g., Shige et al. 2009; Tao et al. 2010). LND, the vertical gradient of divergence through the LND In addition, incorporating characteristic heating profiles in (and implied heating rates) in each case were smaller in regional and global climate models similar to those in this DC-ULD systems than the remaining baroclinic storm study will lead to the improvement of forecasted storm classifications because the upper-tropospheric heating and systems and facilitate analysis of their large-scale dynam- lower-tropospheric cooling is distributed over a greater ical influence in the extratropics. vertical depth. Perhaps one of the more important results of this Acknowledgments. We thank Brian Mapes and Jialin study is that divergence profiles often imply convective Lin for providing the radar VAD and divergence code heating when the overwhelming majority of the ob- and for helpful comments during the preparation of the served rain region is stratiform. This characteristic is manuscript; Aaron Funk at Texas A&M for providing observed for all storm classifications and seasons in this Z–R calculations for the ADRAD observations; and the study and is especially evident in mean profiles for in- three reviewers, whose comments helped to improve the determinate RRP. In particular, indeterminate profiles manuscript. The first author thanks the Advanced Study show a consistent structure across all storm classifica- Program (ASP) at NCAR for postdoctoral support. This tions that is representative of relatively shallow and/or research was also funded by National Science Founda- slantwise convection, with convergence maxima below tion Grant ATM-0449782 to Texas A&M University. 2.5 km in altitude and LND altitudes lower than both convective and stratiform RRP. In addition, the magnitude of divergence shows little variation from middle to REFERENCES upper levels, suggesting that stratiform rain regions may Browning, K. A., and R. Wexler, 1968: The determination of kine- also contribute significantly to the divergence (and matic properties of a wind field using Doppler radar. J. Appl. heating) profile in these situations. The convection of Meteor., 7, 105–113, doi:10.1175/1520-0450(1968)007,0105: moderate vertical extent in this subtropical study may be TDOKPO.2.0.CO;2.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 1SEPTEMBER 2014 H O M E Y E R E T A L . 6685

Bryan, G. H., and J. M. Fritsch, 2000: Diabatically driven discrete 78, 2179–2196, doi:10.1175/1520-0477(1997)078,2179: propagation of surface fronts: A numerical analysis. J. Atmos. SPIROC.2.0.CO;2. Sci., 57, 2061–2079, doi:10.1175/1520-0469(2000)057,2061: ——, S. A. Rutledge, T. J. Matejka, and P. V. Hobbs, 1981: The DDDPOS.2.0.CO;2. mesoscale and microscale structure and organization of clouds Chang, E. K. M., S. Lee, and K. L. Swanson, 2002: Storm and precipitation in midlatitude cyclones. III: Air motions and track dynamics. J. Climate, 15, 2163–2183, doi:10.1175/ precipitation growth in a warm-frontal rainband. J. Atmos. 1520-0442(2002)015,02163:STD.2.0.CO;2. Sci., 38, 639–649, doi:10.1175/1520-0469(1981)038,0639: Chiang, J. C. H., S. E. Zebiak, and M. A. Cane, 2001: Relative roles TMAMSA.2.0.CO;2. of elevated heating and surface temperature gradients in driv- Johnson, R. H., 1984: Partitioning tropical heat and moisture ing anomalous surface winds over tropical oceans. J. Atmos. budgets into cumulus and mesoscale components: Im- Sci., 58, 1371–1395, doi:10.1175/1520-0469(2001)058,1371: plications for cumulus parameterization. Mon. Wea. Rev., RROEHA.2.0.CO;2. 112, 1590–1601, doi:10.1175/1520-0493(1984)112,1590: Crum, T. D., and R. L. Alberty, 1993: The WSR-88D and the PTHAMB.2.0.CO;2. WSR-88D Operational Support Facility. Bull. Amer. Meteor. ——, T. M. Rickenbach, S. A. Rutledge, P. E. Ciesielski, and , Soc., 74, 1669–1687, doi:10.1175/1520-0477(1993)074 1669: W. H. Schubert, 1999: Trimodal characteristics of tropical . TWATWO 2.0.CO;2. convection. J. Climate, 12, 2397–2418, doi:10.1175/ DeMaria, M., 1985: Linear responses of a stratified tropical at- 1520-0442(1999)012,2397:TCOTC.2.0.CO;2. mosphere to convective forcing. J. Atmos. Sci., 42, 1944–1959, Joss, V. J., and A. Waldvogel, 1967: Ein spektrograph für , . doi:10.1175/1520-0469(1985)042 1944:LROAST 2.0.CO;2. niederschlagstropfen mit automatischer auswertung (A Hartmann, D. L., H. H. Hendon, and R. A. Houze, 1984: Some im- spectograph for precipitation drops with automatic evalua- plications of the mesoscale circulations in tropical cloud clusters tion). Pure Appl. Geophys., 68, 240–246, doi:10.1007/ for large-scale dynamics and climate. J. Atmos. Sci., 41, 113–121, BF00874898. , . doi:10.1175/1520-0469(1984)041 0113:SIOTMC 2.0.CO;2. Lin, J., B. Mapes, M. Zhang, and M. Newman, 2004: Stratiform Haynes, P. H., and M. E. McIntyre, 1987: On the evolution of precipitation, vertical heating profiles, and the Madden– vorticity and potential vorticity in the presence of diabatic Julian oscillation. J. Atmos. Sci., 61, 296–309, doi:10.1175/ heating and frictional or other forces. J. Atmos. Sci., 1520-0469(2004)061,0296:SPVHPA.2.0.CO;2. 44, 828–841, doi:10.1175/1520-0469(1987)044,0828: Maddox, R. A., D. J. Perkey, and J. M. Fritsch, 1981: Evolution of OTEOVA.2.0.CO;2. upper tropospheric features during the development of a me- Hertenstein, R. F. A., and W. H. Schubert, 1991: Potential vor- soscale convective complex. J. Atmos. Sci., 38, 1664–1674, ticity anomalies associated with squall lines. Mon. Wea. doi:10.1175/1520-0469(1981)038,1664:EOUTFD.2.0.CO;2. Rev., 119, 1663–1672, doi:10.1175/1520-0493(1991)119,1663: Mapes, B. E., 1993: Gregarious tropical convection. J. Atmos. PVAAWS.2.0.CO;2. Sci., 50, 2026–2037, doi:10.1175/1520-0469(1993)050,2026: Herzegh, P. H., and P. V. Hobbs, 1980: The mesoscale and mi- GTC.2.0.CO;2. croscale structure and organization of clouds and precipitation ——, and R. A. Houze, 1993: An integrated view of the 1987 in midlatitude cyclones. II: Warm-frontal clouds. J. Atmos. Australian monsoon and its mesoscale convective systems. II: Sci., 37, 597–611, doi:10.1175/1520-0469(1980)037,0597: Vertical structure. Quart. J. Roy. Meteor. Soc., 119, 733–754, TMAMSA.2.0.CO;2. Hobbs, P. V., T. J. Matejka, P. H. Herzegh, J. D. Locatelli, and R. A. doi:10.1002/qj.49711951207. Houze Jr., 1980: The mesoscale and microscale structure ——, and ——, 1995: Diabatic divergence profiles in western and organization of clouds and precipitation in midlatitude cy- Pacific mesoscale convective systems. J. Atmos. Sci., , clones. I: A case study of a cold front. J. Atmos. Sci., 37, 568–596, 52, 1807–1828, doi:10.1175/1520-0469(1995)052 1807: . doi:10.1175/1520-0469(1980)037,0568:TMAMSA.2.0.CO;2. DDPIWP 2.0.CO;2. Homeyer, C. R., 2014: Formation of the enhanced-v infrared ——, and J. Lin, 2005: Dopper radar observations of mesoscale cloud top feature from high-resolution three-dimensional wind divergence in regions of tropical convection. Mon. Wea. radar observations. J. Atmos. Sci., 71, 332–348, doi:10.1175/ Rev., 133, 1808–1824, doi:10.1175/MWR2941.1. JAS-D-13-079.1. Marshall, J. S., and W. M. Palmer, 1948: The distribution of Hopper, L. J., 2011: Investigations in southeast Texas precipitating raindrops with size. J. Meteor., 5, 165–166, doi:10.1175/ , . storms: Modeled and observed characteristics, model sensi- 1520-0469(1948)005 0165:TDORWS 2.0.CO;2. tivities, and educational benefits. Ph.D. thesis, Texas A&M Matejka, T. J., R. A. Houze Jr., and P. V. Hobbs, 1980: Micro- University, 141 pp. [Available online at http://repository. physics and dynamics of clouds associated with mesoscale tamu.edu/bitstream/handle/1969.1/ETD-TAMU-2011-12- rainbands in extratropical cyclones. Quart. J. Roy. Meteor. 10341/HOPPER-DISSERTATION.pdf?sequence52.] Soc., 106, 29–56, doi:10.1002/qj.49710644704. ——, and C. Schumacher, 2009: Baroclinicity influences on storm Ninomiya, K., 1971a: Dynamical analysis of outflow from divergence and stratiform rain: Subtropical upper-level dis- tornado-producing thunderstorms as revealed by ATS III turbances. Mon. Wea. Rev., 137, 1338–1357, doi:10.1175/ pictures. J. Appl. Meteor., 10, 275–294, doi:10.1175/ 2008MWR2564.1. 1520-0450(1971)010,0275:DAOOFT.2.0.CO;2. ——, and ——, 2012: Modeled and observed variations in storm di- ——, 1971b: Mesoscale modification of synoptic situations from vergence and stratiform rain production in southeastern Texas. thunderstorm development as revealed by ATS III and aero- J. Atmos. Sci., 69, 1159–1181, doi:10.1175/JAS-D-11-092.1. logical data. J. Appl. Meteor., 10, 1103–1121, doi:10.1175/ Houze, R. A., 1982: Cloud clusters and large-scale vertical motions 1520-0450(1971)010,1103:MMOSSF.2.0.CO;2. in the tropics. J. Meteor. Soc. Japan, 60, 396–409. Raymond, D. J., and H. Jiang, 1990: A theory for long-lived me- ——, 1997: Stratiform precipitation in regions of convection: soscale convective systems. J. Atmos. Sci., 47, 3067–3077, A meteorological paradox? Bull. Amer. Meteor. Soc., doi:10.1175/1520-0469(1990)047,3067:ATFLLM.2.0.CO;2.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC 6686 JOURNAL OF CLIMATE VOLUME 27

Sanders, F., 2005: Real front or baroclinic trough? Wea. Fore- J. Climate, 14, 2222–2237, doi:10.1175/1520-0442(2001)014,2222: casting, 20, 647–651, doi:10.1175/WAF846.1. IMCAAO.2.0.CO;2. Schaefer, J. T., 1974: The life cycle of the dryline. J. Appl. Me- Sui, C.-H., and K.-M. Lau, 1989: Origin of low-frequency (intra- teor., 13, 444–449, doi:10.1175/1520-0450(1974)013,0444: seasonal) oscillations in the tropical atmosphere. Part II: . TLCOTD 2.0.CO;2. Structure and propagation of mobile wave-CISK modes Schultz, D. M., 2004: Cold fronts with and without prefrontal and their modification by lower boundary forcings. J. Atmos. wind shifts in the central United States. Mon. Wea. Rev., Sci., 46, 37–56, doi:10.1175/1520-0469(1989)046,0037: 132, 2040–2053, doi:10.1175/1520-0493(2004)132,2040: OOLFOI.2.0.CO;2. CFWAWP.2.0.CO;2. Tao, W.-K., S. Lang, X. Zeng, S. Shige, and Y. Takayabu, 2010: Schumacher, C., R. A. Houze, and I. Kraucunas, 2004: The tropical dynamical response to latent heating estimates derived from Relating convective and stratiform rain to latent heating. the TRMM precipitation radar. J. Atmos. Sci., 61, 1341–1358, J. Climate, 23, 1874–1893, doi:10.1175/2009JCLI3278.1. doi:10.1175/1520-0469(2004)061,1341:TTDRTL.2.0.CO;2. Wakimoto, R. M., and H. V. Murphey, 2008: Airborne Dopp- Shige, S., Y. N. Takayabu, S. Kida, W.-K. Tao, X. Zeng, ler radar and sounding analysis of an oceanic cold C. Yokoyama, and T. L‘Ecuyer, 2009: Spectral retrieval of front. Mon. Wea. Rev., 136, 1475–1491, doi:10.1175/ latent heating profiles from TRMM PR data. Part IV: Com- 2007MWR2241.1. parisons of lookup tables from two- and three-dimensional Wu, Z., E. S. Sarachik, and D. S. Battisti, 2000: Vertical structure of cloud-resolving model simulations. J. Climate, 22, 5577–5594, convective heating and the three-dimensional structure of doi:10.1175/2009JCLI2919.1. the forced circulation on an equatorial beta plane. J. Atmos. Steiner, M., R. A. Houze, and S. E. Yuter, 1995: Climatological Sci., 57, 2169–2187, doi:10.1175/1520-0469(2000)057,2169: characterization of three-dimensional storm structure VSOCHA.2.0.CO;2. from operational radar and rain gauge data. J. Appl. Me- Yuter, S. E., and R. A. Houze, 1997: Measurements of raindrop teor., 34, 1978–2007, doi:10.1175/1520-0450(1995)034,1978: size distributions over the pacific warm pool and implications CCOTDS.2.0.CO;2. Stensrud, D. J., 1996: Effects of persistent, midlatitude mesoscale for Z–R relations. J. Appl. Meteor., 36, 847–867, doi:10.1175/ , . regions of convection on the large-scale environment during 1520-0450(1997)036 0847:MORSDO 2.0.CO;2. the warm season. J. Atmos. Sci., 53, 3503–3527, doi:10.1175/ Zhang, D.-L., and R. Harvey, 1995: Enhancement of extratropical 1520-0469(1996)053,3503:EOPMMR.2.0.CO;2. cyclogenesis by a mesoscale convective system. J. Atmos. ——, and J. L. Anderson, 2001: Is midlatitude convection an active Sci., 52, 1107–1127, doi:10.1175/1520-0469(1995)052,1107: or a passive player in producing global circulation patterns? EOECBA.2.0.CO;2.

Unauthenticated | Downloaded 10/04/21 05:16 PM UTC