-
2179Bulletin of the American Meteorological Society
1. Traditions
“Stratiform precipitation” is a term often used inmeteorology,
yet the term is not defined in the Glos-sary of Meteorology
(Huschke 1959), and its mean-ing and usage continually evolve. This
article attempts
to clarify the meaning of the term stratiform precipi-tation as
it is currently used in the particular contextof tropical
meteorology and other regions whereclouds are generated by
atmospheric convection. Suchan article would have seemed
paradoxical, if not out-right heretical, 30 years ago. Stratiform
precipitationwas generally thought to occur almost exclusively
withfronts in midlatitude cyclones, where ice particlesgrow
predominantly by vapor deposition in a deep,largely buoyantly
stable nimbostratus cloud layer, driftdown from upper levels, melt,
and fall to the earth’ssurface as raindrops. Under certain cold
conditions, theparticles may never melt, and they reach the
surface
Stratiform Precipitation inRegions of Convection:
A Meteorological Paradox?
Robert A. Houze Jr.University of Washington, Seattle,
Washington
ABSTRACT
It was once generally thought that stratiform precipitation was
something occurring primarily, if not exclusively, inmiddle
latitudes—in baroclinic cyclones and fronts. Early radar
observations in the Tropics, however, showed large ra-dar echoes
composed of convective rain alongside stratiform precipitation,
with the stratiform echoes covering greatareas and accounting for a
large portion of the tropical rainfall. These observations seemed
paradoxical, since stratiformprecipitation should not have been
occurring in the Tropics, where baroclinic cyclones do not occur.
Instead it was fall-ing from convection-generated clouds, generally
thought to be too violent to be compatible with the layered, gently
set-tling behavior of stratiform precipitation.
In meteorology, convection is a dynamic concept; specifically,
it is the rapid, efficient, vigorous overturning of theatmosphere
required to neutralize an unstable vertical distribution of moist
static energy. Most clouds in the Tropics areconvection-generated
cumulonimbus. These cumulonimbus clouds contain an evolving pattern
of newer and older pre-cipitation. The young portions of the
cumulonimbus are too violent to produce stratiform precipitation.
In young, vigor-ous convective regions of the cumulonimbus,
precipitation particles increase their mass by collection of cloud
water,and the particles fall out in heavy showers, which appear on
radar as vertically oriented convective “cells.” In regions ofolder
convection, however, the vertical air motions are generally weaker,
and the precipitation particles drift downward,with the particles
increasing their mass by vapor diffusion. In these regions the
radar echoes are stratiform, and typicallythese echoes occur
adjacent to regions of younger convective showers. Thus, the
stratiform and convective precipita-tion both occur within the same
complex of convection-generated cumulonimbus cloud.
The feedbacks of the apparent heat source and moisture sink of
tropical cumulonimbus convection to the large-scaledynamics of the
atmosphere are distinctly separable by precipitation region. The
part of the atmospheric response deriv-ing from the areas of young,
vigorous convective cells is two layered, with air converging into
the active convection atlow levels and diverging aloft. The older,
weaker intermediary and stratiform precipitation areas induce a
three-layeredresponse, in which environmental air converges into
the weak precipitation area at midlevels and diverges from it
atlower and upper levels. If global precipitation data, such as
that to be provided by the Tropical Rainfall Measuring Mis-sion,
are to be used to validate the heating patterns predicted by
climate and general circulation models, algorithms mustbe applied
to the precipitation data that will identify the two principal
modes of heating, by separating the convectivecomponent of the
precipitation from the remainder.
Corresponding author address: Robert A. Houze Jr., Dept. of
At-mospheric Sciences, University of Washington, Box
351640,Seattle, WA 98195.E-mail: [email protected] final
form: 15 May 1997.©1997 American Meteorological Society
-
2180 Vol. 78, No. 10, October 1997
as snow rather than rain. In both cases, the upward airmotion
induces vapor deposition onto the growing iceparticles but is weak
enough to allow the particles todrift downward while they grow.
In a baroclinic cyclone, the widespread lifting pro-ducing the
growth of the precipitation particles occursin the regions of
large-scale warm advection, whichis concentrated in the vicinity of
fronts. But stratiformprecipitation also can occur in other
large-scale andmesoscale dynamical settings, as long as there is
as-cent of saturated air weak enough to allow ice particlesto fall
out. In other words, the term stratiform precipi-tation designates
a particular set of microphysical pro-cesses leading to the growth
and fallout of the precipi-tation within the context of relatively
gentle upwardair motion; it does not refer to the specific
dynamiccause of the vertical air motions within which the
par-ticles form, grow, and fall out.
Since the advent of weather radar in the late 1940s,the term
stratiform has been used to describe precipi-tation as it appears
in displays of radar data. Stratiformprecipitation is fairly
homogeneous in the horizontal,giving it a layered structure in
vertical cross sectionsof radar reflectivity. In particular, it
often exhibits apronounced layer of high reflectivity called the
“brightband,” marking the layer in which the downward settlingice
particles are melting (Battan 1973; Houze 1993).
Stratiform radar echoes contrast sharply with radarechoes from
“convective precipitation,” which appearson radar as “cells,”
another loosely defined term, ap-parently coined by Byers and
Braham (1949) in theirclassic report on the Thunderstorm Project in
Ohio andFlorida. Cells are horizontally localized patches orcores
of intense radar reflectivity. In a vertical crosssection, a cell
is a tall, thin column of high reflectivity.The orthogonality of
its structure to that of abrightband echo often makes the
convective “cell”quite recognizably distinct from “stratiform”
precipita-tion on radar, and radar meteorologists have developeda
tradition of speaking of radar echoes as convective orstratiform,
according to whether they form patterns ofvertically oriented
intense cores or horizontal layers.
2. Radar observations in tropical fieldexperiments: Emergence of
aparadox
Because of its ubiquitous presence in barocliniccyclones and
fronts, and because of the dearth of me-teorological observations
in the Tropics up to the sec-
ond half of this century, stratiform precipitation wasthought to
occur primarily, if not exclusively, inmiddle latitudes. As late as
1979, Herbert Riehl re-ferred to tropical rain as “a cumulus
regime, in con-trast to the climates with stratus precipitation in
higherlatitudes.” Riehl was correct in characterizing theTropics as
a “cumulus regime”; precipitating cloudsin the Tropics are entirely
convection-generated cumu-lonimbi, and baroclinic cyclones and
fronts, indeed,do not occur in the Tropics. However, his
statementdoes not recognize that stratiform precipitation is
notrestricted to higher-latitude clouds. It may also occurwithin
the “cumulus regime.”
In 1972, scientists aboard a Soviet research shipoperating in
the intertropical convergence zone (ITCZ)of the eastern tropical
Atlantic took photographs of aradar display showing strong
brightband echo(Shupiatsky et al. 1975, 1976a,b). In 1974, the
GlobalAtmospheric Research Programme Atlantic TropicalExperiment
(GATE) obtained more extensive radarobservations in this region.
One of the most signifi-cant findings of GATE was that the
stratiform radarechoes, seen earlier by the Soviet scientists, had
verystrong bright bands and covered large areas (Houze1975, 1977;
Leary and Houze 1979a,b). About 40%of the rain falling on the ocean
surface in GATE wasstratiform (Cheng and Houze 1979; Leary
1984).These observations showed definitively that both con-vective
and stratiform radar echoes occurred in a re-gime in which the
clouds are generated entirely byatmospheric convection.
Since GATE, radar studies have confirmed, manytimes over, that a
large component (~20%–50%) oftropical precipitation exhibits
stratiform radar-echostructure [see Houze (1989) for a summary].
More-over, the stratiform echoes usually coexist with cellswithin a
“mesoscale” precipitation area, in which therain covers
contiguously a region 10–1000 km inhorizontal dimension. Higher
rainfall rates in thesemesoscale precipitation areas are
concentrated in thecells, but lighter stratiform rain covers most
of themesoscale area. A portion of the area is covered by echoof
intermediate intensity, which may be either “convec-tive” or
“stratiform,” depending on its vertical structure.
In addition to the purely tropical regimes, in whichthere is no
chance of a baroclinic contribution to theprecipitation processes,
there have now been manystudies recognizing the large presence of
stratiformprecipitation in convection-generated clouds and
pre-cipitation in midlatitude regimes (e.g., Rutledge andMacGorman
1988; Houze et al. 1990). The discussions
-
2181Bulletin of the American Meteorological Society
in this paper apply to both midlatitude and
tropicalconvection.
3. Consistent terminology: The purposeof this paper
The fact that tropical precipitation, when observedby radar, has
a substantial stratiform component hasled to a terminology that
seems contradictory, sincethe tropical atmosphere is clearly a
region of convec-tion. The term convection refers to the
overturning ofa fluid under gravity, and in the Tropics, nearly
allprecipitation (except part of that associated with tropi-cal
cyclones and possibly some orographic rain) is theproduct of
atmospheric convection. However, the termstratiform, as defined
above, is an adjective describ-ing a set of microphysical processes
that occur inweaker vertical air motions. Similarly, the term
“con-vective” describes a radar echo from precipitationgrowing by
the cloud-microphysical processes thatattend the strong vertical
air motions of young, activeconvection. Since stratiform radar
echoes may be seen
in an atmosphere of older convection, the terms strati-form (an
adjective) and convection (a noun) are notmutually exclusive;
stratiform describes the nature ofsome of the precipitation growth
processes occurringin the region of convection, while the term
convectionalludes to the fluid dynamical origin of the air
motions.
Henceforth, this paper, which is about tropical andother
convection-generated precipitation, will use thenoun convection to
refer to an overturning fluid andthe adjective convective to
describe the precipitation(or radar echo) associated with young,
active convec-tion. The adjective stratiform will refer to
precipita-tion occurring in older, less active convection
andpossessing radar echoes that have weak horizontalgradients
and/or a bright band. Describing the radarechoes as being either
convective or stratiform thusimplies that the echoes depict
precipitation in regionsof stronger or weaker air motions,
respectively, whileat the same time recognizing that both the
weaker andstronger air motions may arise from the process
ofconvection. Convection can lead to cumulonimbusclouds, whose
precipitation is partly convective andpartly stratiform.1 Figures 1
and 2 illustrate, in ideal-
FIG. 1. Conceptual models of vertical cross sections through
(a)–(c) young, vigorous precipitating convection and (d)–(f) old
convection.
1The stratiform precipitation could be considered to be falling
from nimbostratus cumulonimbogenitus (precipitating stratiform
cloudarising from cumulonimbus). Since such nimbostratus is usually
a cloud deck currently or previously extruding out of the
cumulo-nimbus, it seems artificial to call it a separate cloud. The
author prefers to call the whole, single precipitating cloud entity
a cumulonimbus.
-
2182 Vol. 78, No. 10, October 1997
ized form, the terminology of this paper. The remain-der of this
paper delves further into this idealized pic-ture in order to reach
a deeper understanding of con-vection-generated precipitation and
how to observe,describe, and interpret it.
4. What is convection?
To understand how both convective and stratiformprecipitation
can occur in response to convection, weshould first review what we
mean by convection in thecontext of cloud-formation processes. We
say convec-tion occurs when a fluid under gravity is heated
from
below or cooled from the top at such a rate that mo-lecular
diffusion cannot redistribute the modifieddensity field fast enough
so as to maintain equilibrium.The fluid layer then becomes
buoyantly unstable andoverturns macroscopically to stabilize the
stratifica-tion of the density. One of the simplest examples
isBenard convection in an incompressible fluid layer,where the
fluid overturns by means of a field of geo-metrically simple
vertically circulating elements. Theatmospheric convection that
produces cumulus andcumulonimbus clouds has the additional
complica-tions that air is a compressible fluid and contains wa-ter
vapor, so overturning acts to neutralize the stratifi-cation of
moist-static energy. Figure 1a shows atmo-spheric convection as a
pattern of idealized verticallycirculating elements, which are
geometrically sym-metric (like Benard cells) in the drawing,
although inreality, the shapes of the overturning elements are
highlymodified by wind shear and by the interactions of thecloud
and precipitation fields with the air motions.Asymmetries in the
vertical structures of the cells donot affect the present
discussion, which aims only todistinguish the terms “convective”
and “stratiform” indescribing precipitation. At the most basic
level, itserves to keep to the symmetric idealization in Fig.
1a.
In atmospheric convection, water may condenseand fall out. A
saturated layer of air that is convectivelyoverturning and
precipitating, therefore, differs froma layer of Benard convection
in that there must be anet upward transfer of air between the upper
and lowerboundaries of the fluid in order to account for the
netcondensation of water in the layer (as long as the rateof rise
of temperature is less than the heating rate). Theupdrafts of the
overturning cells must, therefore, trans-port more mass on average
than do the downdrafts.Figures 3a and 3b summarize the effects of
verticalmass transport in a region of active convective cells:net
upward mass transport produces net latent heatingat all levels
(Fig. 3a), and mass continuity requires nethorizontal convergence
at low levels and divergencein the upper troposphere (Fig. 3b, open
arrows in Fig.1a). In the real atmosphere, a region of young,
vigor-ous convection may not be completely saturated in
itsdowndraft regions, but measurements confirm that thenet mass
divergence profiles are like that in Fig. 3b(e.g., Mapes and Houze
1995).
5. What is convective precipitation?
In deep precipitating atmospheric convection, the
FIG. 2. Idealization of a horizontal map of radar reflectivity
(a)divided into convective and (b) stratiform regions.
-
2183Bulletin of the American Meteorological Society
vertical velocities in the cores of the updrafts are sev-eral
meters per second or greater. These strong updraftscondense vapor
rapidly, producing large concentra-tions of cloud liquid water.
Houghton (1968) pointedout that, in the presence of such strong
updrafts, thedominant growth mechanism for precipitation par-ticles
is the collection of this cloud water by largerdrops and/or ice
particles sweeping out the cloud wa-ter in their fall paths. For
growing water drops, thisprocess is termed “coalescence”; in the
case of collec-tion by ice particles, it is known as “riming.”
More-over, the strong updrafts allow for the larger particlesto
grow for a long period of time because they are car-ried upward
relative to the earth, even though they arefalling relative to the
smaller cloud water droplets.This upward advection of the growing
particles in-creases their residence time in cloud and thus
theiropportunities to collect cloud droplets. Figure 4 illus-trates
how, as a parcel of updraft air rises, the grow-ing particles
within it move up until they become largeenough to fall relative to
the air. At each successiveheight, some particles fall out of the
parcel, while theremainder, which are not as heavy, are spread out
lat-
erally over an increasingly greater area by the diver-gent air
flow as the rising, buoyant air parcel expands.
Since the bulk of the precipitation mass falls outwithin a few
kilometers of the updraft centers, the ra-dar reflectivity pattern
associated with the convectiveair motions in Fig. 1a is a set of
concentrated peaks ofreflectivity, as shown in Fig. 1c. In plan
view, the con-vective region appears in the radar echo as a field
oflocalized reflectivity maxima (Fig. 2).
6. Stratiform precipitation in oldconvection
When atmospheric convection is young and vigor-ous, the
precipitation occurs in cells, as indicated con-ceptually in Fig.
1a. However, as a region of convec-tive cells weakens, the
precipitation associated withthe cells takes on a layered structure
resembling thatfound in extratropical cyclonic precipitation. The
up-ward vertical air motions in the region of weakenedcells are
strong enough to allow precipitation particlesto grow by vapor
diffusion but too weak to supporthigh concentrations of cloud
liquid water; therefore,growth by riming is not so effective as
when the up-drafts are stronger. In this respect, then, the region
ofweakened cells in a tropical cloud system resemblesan
extratropical cyclone.
Thus, when we divide the precipitation area of
aconvection-generated cloud system into convectiveand stratiform
components (Fig. 2), we imply that the
FIG. 3. Characteristic profiles of latent heating and
horizontalmass divergence in convective and stratiform regions of
tropicalprecipitation.
FIG. 4. Conceptual model of an updraft behaving like a
buoyantbubble with different entrainment rates. Dots indicate
hydro-meteors suspended by updraft; downward-pointing
arrowsindicate particles heavy enough to fall through updraft;
horizontalarrows indicate lateral spreading of bubble. Open arrows
representthe vector field of the buoyancy pressure gradient force
(as in Fig.7.1 of Houze 1993). [From Yuter and Houze (1995c).]
-
2184 Vol. 78, No. 10, October 1997
stratiform region is dominated by weaker air motions.These
vertical air motions could be horizontally uni-form across the
stratiform region; however, they areusually variable in the
horizontal. The requirementsare only that there be net upward mass
transport (toallow net condensation of vapor) in the cloud layer
inwhich particles are growing on average and that theensemble of
vertical air motions in the cloud layercontain few if any upward
velocities > 1 m s−1 (so thatcloud liquid water production, and
hence riming, isminimal and precipitation-sized particles are not
ad-vected upward, so as to overly disrupt the layeredreflectivity
pattern).
In a deep precipitating tropical cloud system, the“stratiform”
precipitation region is typically a regionof older convection (Fig.
1d). It is well known that inthe older portions of a cumulonimbus,
the lower tro-posphere is dominated by downdrafts, while the up-per
levels are dominated by weak updrafts. The upperlevels are not
wholly occupied by updrafts, nor are thelower levels wholly
occupied by downdrafts. In theolder part of the precipitation area
of a Florida cumu-lonimbus, the vertical mass transport at upper
levelswas net upward (Fig. 5b), such that there was netgrowth of
precipitation particles. However, the netupward mass transport at
upper levels was the com-bined effect of an ensemble of vertical
velocities, in-cluding some negative (downward) values, as shownby
the distribution of the mass flux as a function ofthe vertical air
velocity and height in Fig. 5a. How-ever, updrafts < 2 m s−1
accounted for most of the up-ward mass transport; that is, nearly
all of the volumeof the storm at upper levels was occupied by air
mov-
ing upward too slowly to overcome the downward fallvelocities of
precipitation particles. At these verticalair velocities, the
condensed water is transferred to thegrowing ice particles almost
entirely by vapor diffu-sion (Rutledge and Houze 1987), and the
particles driftdownward, as suggested by Fig. 1e. Below the
0°Clevel, the older precipitation regions are dominated bynet
downward motion (Houze 1989). Figure 5 showsthat updrafts, mostly
< 1.5 m s−1 in intensity, are in-terspersed among more
predominantly downward airvelocities. Zipser and LeMone (1980) and
LeMoneand Zipser (1980) described how research aircraft inGATE
encountered both up- and downdrafts in osten-sibly stratiform parts
of tropical cumulonimbus.
When growing ice particles drift downward withinthe mid- to
upper altitudes of a volume of air within aregion of old convection
such as that portrayed in Figs.1d–f, distinct layers become evident
via contrastingmicrophysical processes (Fig. 1e). At upper
levelsvapor diffusional growth is prevalent, and it is too coldfor
aggregation to occur. Empirical evidence indicatesthat aggregation
occurs at ambient temperatures of 0°to ∼ −15°C with the highest
probability of aggrega-tion in the temperature range of about 0 to
∼ −5°C(Hobbs 1974, 641). Houze and Churchill (1987) con-firmed this
layering in tropical convection: they foundlarge aggregates in this
temperature layer among theparticle images collected by aircraft in
the stratiformregions of tropical convection over the Bay of
Bengalin the Global Atmospheric Research Programme’sMonsoon
Experiment.
Displays of radar data enhance the layered appear-ance of
stratiform precipitation because the reflectivity
FIG. 5. (a) Vertical mass transport distribution by vertical
velocity and (b) mass flux in a Florida cumulonimbus. Units of
contoursare [mass transport/(dw dz)], where dw = 1 m s−1 and dz =
0.4 km. Contours are at intervals of 25 × 106 kg s−1. [Adapted from
Yuterand Houze (1995c).]
-
2185Bulletin of the American Meteorological Society
of the melting snow is especially high for two reasons:1) some
of the melting particles are very large aggre-gates of ice
crystals, and reflectivity is proportional tothe sixth power of the
particle dimension, and 2) theindex of refraction of melting
snowflakes is ~5 timesgreater than nonmelting ice. In addition,
when the iceparticles melt, their fall speeds increase by a factor
ofabout 5, and the particles evacuate the melting layerrapidly in a
vertically divergent fashion, producing adecreased concentration of
particles below the melt-ing layer. These factors combine to
produce a pro-nounced bright band of enhanced reflectivity in a
shal-low layer centered just below the 0°C level (Fig. 1f).
The bright band is an unambiguous indicator of thepresence of
stratiform precipitation. However, theabsence of a bright band does
not imply the absenceof stratiform precipitation structure. A
bright bandwill not be observed unless the vertical resolution
ofthe radar is sufficiently fine to delineate the band.Hence, the
bright band is a property of the radar data;it is not a unique
measure of the magnitude of themelting. A strong bright band only
will appear if someof the melting particles are in the form of
large aggre-gates. Temperature, ambient humidity, and
crystalstructures must all meet certain criteria for aggrega-tion
to occur. Aggregation of precipitating ice particlesto form
snowflakes does not add any mass to the pre-cipitation, nor do the
particles fall any faster once theyhave aggregated. Just as much
latent heat is consumedin the melting, and the vertical flux of
precipitationmass remains the same, regardless of whether or notthe
precipitating ice particles happen to aggregate toform large highly
reflective particles and hence abright band. Braun and Houze (1995)
showed that themelting layer of a midlatitude mesoscale
convectivesystem was much more horizontally extensive than theradar
bright band, which was present only in the partof the melting layer
in which melting ice particles hadaggregated to form very large
snowflakes.
The presence of a bright band thus indicates thepresence of a
particular type of stratiform precipitation,namely that in which
large aggregate snowflakes areamong the melting particles. Since
the ability to de-tect a bright band is a function of the
characteristicsof the radar rather than the storm, further caution
isrequired, since the absence of a bright band does notindicate the
absence of a layer of melting aggregates.Typically, a radar detects
a bright band only at closerange because the radar beam broadens
with distancefrom the antenna and the melting layer is only ~0.5
kmor less in depth.
The idealized bright band sketched in Fig. 1f ishorizontally
inhomogeneous, broken into severalpatches, and in one patch
“fallstreaks” extend down-ward from the melting layer. Stratiform
precipitationin tropical cumulonimbus usually occurs in
regionswhere previously intense convection (like that de-picted in
Figs. 1a–c) has weakened. Since the previ-ous convection was
concentrated in cells, the result-ing stratiform structure remains
somewhat patchy,with highest reflectivities in the regions where
convec-tive cores once were vigorous; inspection of the
high-resolution airborne radar data obtained in TOGACOARE2 shows
fallstreaks consistently in the loca-tions of previously active
convective cells (e.g., Yuterand Houze 1997a). The fallstreaks
appear to occurwhere the remnant precipitation cores have not
com-pletely lost their identities, and since they are no longerpart
of an active convective cell, the remnant showersare distorted into
tilted, bent fallstreaks by the ambi-ent wind shear.
An alternative explanation for the fallstreaks is thatan
unstable layer produced by the latent cooling of theair by the
melting leads to the formation of small, shal-low convective cells
in the melting layer. In this case,the fallstreaks would be similar
to the “generatingcells” seen in extratropical cyclonic
precipitation(Marshall 1953). It is possible that the fallstreaks
intropical convection are a combination of remnant pre-cipitation
cores and overturning induced by meltingcooling.
7. Dynamic implications of thestratiform component of
tropicalprecipitation
Although vertical air motions are relatively weakin a stratiform
region, the area covered by the strati-form precipitation can be
much larger than that occu-pied by the active convective cells. A
large portion ofthe time- and space-integrated vertical mass
transportof the convection thus occurs in the older convection,that
is, in the stratiform regions of cumulonimbus. Forthis reason, the
stratiform regions of tropical convec-tion have important dynamic
implications.
2The Tropical Ocean Global Atmosphere Coupled Ocean–Atmo-sphere
Response Experiment (TOGA COARE) was a large fieldexperiment
employing aircraft and ships to study the behavior ofthe atmosphere
and ocean over the western tropical Pacific Ocean.The experiment
ran from 1 November 1992 to 28 February 1993.
-
2186 Vol. 78, No. 10, October 1997
The large-scale environment responds to a convec-tive system in
a manner analogous to the way thewater responds to a rock dropping
into a pond. Thewater has to make way for the negatively buoyant
rockaccelerating downward through the fluid environment.The fluid
adjusts to the sudden displacement of thewater by a spectrum of
gravity waves, which rippleaway from the spot where the rock falls
in the water.
In a similar way, a buoyant parcel of air pushesambient air out
of its path, and the surrounding atmo-sphere adjusts by sending out
a spectrum of gravitywaves, whose propagation speeds are
proportionalto their vertical wavelengths (Bretherton
andSmolarkiewicz 1989; Mapes 1993). In a region of pre-cipitating
convection, there is a net positive buoyancyproduced by the latent
heat gained by air when pre-cipitation falls out. The surrounding
atmosphere ad-justs to this generation of buoyancy by a spectrum
ofgravity waves (or more accurately bores) that have thenet effect
of displacing mass outside the convectiveregion downward (Mapes
1993; Mapes and Houze1995). The open arrows in Fig. 1a show the net
resultof this compensating downward displacement as a nethorizontal
transport of air from the environment intothe active convective
region (mass convergence) atlower levels and out of the region
(mass divergence)at upper levels.
By the time the convective region is older, the netvertical mass
transport in the lower levels of the pre-cipitation region is
downward (the negatively buoy-ant downdrafts begin to dominate over
the updrafts).Net cooling occurs as a result of melting and
evapo-ration at mid- to low levels during this stage (Fig. 3c),and
horizontal divergence of air out of the storm mustcompensate the
negative buoyancy in the lower tro-posphere of the rain areas (Fig.
3d, and the open ar-rows at low levels in Fig. 1d). Gravity waves
that re-arrange the environmental mass field around the strati-form
rain areas must produce a net upward displace-ment of the
environment to compensate the net hori-zontal mass divergence out
of the old convective (i.e.,stratiform) region at low levels.
The open arrows in Fig. 1d also show a net hori-zontal transport
of air into the old convective region(mass convergence) at
midlevels. This midlevel con-vergence compensates both the (net)
positive buoy-ancy of the upper levels of the stratiform region
andnet negative buoyancy of the lower levels.
The most direct way to measure the adjustment ofthe large-scale
environment to a region of convectionis to measure the horizontal
mass divergence into and
out of the region. Accurate measurement of the out-ward normal
component of the wind along the bound-ary of the volume of
atmosphere containing the con-vection gives us this divergence, and
in tropical fieldexperiments, airborne Doppler radar data and
rawin-sondes have been used for this purpose. In TOGACOARE,
aircraft Doppler radar provided 143 profilesof divergence around
regions of precipitation. An ex-ample of the average of the
profiles obtained on a flightsampling primarily convective
precipitation showedconvergence at low levels and divergence at
upper lev-els (Fig. 6a), consistent with the open arrows in Fig.1a.
A flight sampling primarily old convection, indi-cated by the
stratiform character of the radar echo,indicated strong convergence
at midlevels (Fig. 6b),consistent with the open arrows in Fig. 1d.
Figure 7shows the grand mean profile of divergence, obtainedby
averaging all 143 samples obtained on flights inTOGA COARE. It
represents the combination of ac-tive and old (stratiform)
convection sampled on all theflights, over a 4-month period. The
solid curve in Fig.8 shows the profile of divergence obtained by
averag-ing TOGA COARE rawinsonde data obtained aroundregions of
satellite-observed active deep convection.It is consistent with the
profile of divergence from theairborne Doppler radar (compare the
profile in Fig. 7with the solid curve in Fig. 8).
To determine how the large-scale atmosphere ad-justs to a
disturbance of its mass field by a region ofdeep convection, Mapes
and Houze (1995) used themeasured divergence profiles in Figs. 7
and 8 as inputto a linear spectral primitive-equation model. In
thismodel the divergence profile serves as a mass distur-bance for
the large-scale flow. The time-dependentflow predicted by the
model, using the observed massdivergence profiles as input,
indicates the modes ofmotion that develop when the large-scale
atmosphereis perturbed by a divergence profile similar to
thatmeasured in the vicinity of the deep convection occur-ring in
TOGA COARE. The calculation applies theobserved divergence profile
impulsively over a 140-km diameter region (a typical size of a deep
convec-tive precipitation area in TOGA COARE).
Two distinct modes dominate the model’s large-scale response to
the convection. The solutions of theequations are decomposed into
vertical wavelengthcomponents, which correspond to the vertical
wave-length components of the spectrally decomposed ob-served
profile of divergence. The wave speed is di-rectly proportional to
the vertical wavelength, and Fig.9a shows the magnitudes of the
spectral coefficients
-
2187Bulletin of the American Meteorological Society
of the divergence for each wave speed. The bigger
thecoefficient, the more that component of the divergenceprofiles
accounts for the large-scale response to the
convection. Figure 9a shows that the motions aredominated by
bands of wave speeds centered on 52and 23 m s−1.
FIG. 6. Mean profiles of divergence measured by airborne Doppler
radar in tropical precipitation over the western tropical
Pacificduring TOGA COARE: (a) from a flight in which most of the
sampled precipitation was in a vigorously convective state (average
of13 samples obtained 15 December 1992); (b) from a flight in which
most of the sampled precipitation was in a stratiform state
(averageof 15 samples obtained on 6 November 1992). All samples are
for a region 30 km in diameter. The three lines in each plot are
threeindependent radar measures for negligibly different geometry.
[From Mapes and Houze (1995).]
FIG. 8. Divergence measured by rawinsondes when extensive,deep
convection-generated cloud systems affected the intensiveflux array
(centered near 2°S, 156°E) over the western tropicalPacific during
TOGA COARE. Sixteen cases were used tocompute the net divergence
(solid curve). The components of thenet divergence constituted by
the 52 and 23 m s−1 gravity waveresponses to the mass disturbance
are shown by the dashed anddashed–dotted curves, respectively.
[Derived from calculations ofMapes and Houze (1995) and provided by
B. Mapes.]
FIG. 7. Mean of all divergence profiles measured by
airborneDoppler radar in tropical precipitation over the western
tropicalPacific during TOGA COARE. The average is based on
143samples, obtained on 10 different aircraft missions. [From
Mapesand Houze (1995).]
-
2188 Vol. 78, No. 10, October 1997
Figure 8 shows the computed 52 and 23 m s−1 com-ponents of the
divergence profile in comparison to thetotal observed profile of
divergence. The 52 m s−1 com-ponent has a two-layer structure,
consisting of conver-gence at low levels and divergence at upper
levels,while the 23 m s−1 component has a three-layer struc-ture,
consisting of convergence at midlevels and di-vergence at upper and
lower levels. Evidently, thesetwo dominant modes of the observed
divergence pro-
file correspond respectively to the convective andstratiform
portions of the precipitation pattern. Addedtogether, the
convective (52 m s−1) and stratiform (23m s−1) curves in Fig. 8
account for nearly the entireobserved divergence (solid curve).
The atmospheric adjustment to the active convec-tive region is a
fast-moving gravity wave “bore” thatdisplaces environmental mass
downward, to mass-balance the convergent inflow to the convective
regionat low levels and the divergent outflow aloft. Since thisbore
moves away at 52 m s−1 (over 100 mph), the ad-justment is very
quick. The adjustment to the strati-form region’s older convection
is less than half as fast,and this bore displaces mass upward in
the lower tro-posphere and downward in the upward troposphere
toadjust for the stratiform region’s convergence atmidlevels and
divergence at upper and lower levels.
Figure 9b shows the effect of the fast (convective)and slow
(stratiform) bores 6 h after the impulsive startof the heating. The
convective bore has produced amass-compensating downward
displacement throughthe depth of the troposphere at 600–1300 km
from thecenter of the disturbance. This compensating down-ward
motion suppresses convection at these distances.At 150–600 km from
the source, the stratiform bore,which has moved out only about half
as far as the con-vective bore, has produced a net upward
displacementof mass in the lower troposphere and a net
downwarddisplacement aloft. An important implication is that adeep
convective tropical disturbance may actuallyencourage new
convection in its immediate vicinityvia the compensating upward
motion in the lower tro-posphere. This result led Mapes (1993) to
dub tropi-cal convection “gregarious.”
8. Convective and stratiformprecipitation: A dichotomy or
theends of a spectrum?
The results of Mapes and Houze (1995) in Figs. 8and 9 suggest
that there is no reason, from the view-point of large-scale
dynamics, to distinguish catego-ries of air motion in a tropical
precipitation regionother than the categories convective and
stratiform—that is, a dichotomous structure of convection. The
im-pression formed from looking at radar-echo maps oftropical
precipitation, however, is that a significantportion of the
precipitation field is difficult to charac-terize as convective or
stratiform. In an earlier paper,Mapes and Houze (1993) classified
one-third of the
FIG. 9. Results of a linear spectral model of a
large-scaletropical atmosphere perturbed by a mass divergence
profile. Thesymbol δ
d stands for the “diabatic divergence,” which is the mass
disturbance produced by the net heating of a
mesoscaleprecipitating cloud system embedded in the large-scale
flow. Theobserved divergence profile used for δ
d is based on airborne
Doppler radar and sounding observations obtained over thewestern
tropical Pacific in TOGA COARE in regions containinglarge
cumulonimbus cloud systems. This observed divergenceprofile is
applied impulsively at the initial time t over a region 140km in
horizontal dimension with a horizontal cos2 profile. Themodel
computes the response of the large-scale environment tothe cloud
disturbance at all times after the impulsive massdisturbance. The
response is a spectrum of gravity waves withvarying vertical
wavelengths. Each wavelength corresponds to awave speed c
n. The symbol δ
dn in (a) represents the spectral
coefficient (magnitude) of the response at each wave
speed.Temperature perturbation T shown by the contours in (b)
isproduced by the fast (convective) and slow (stratiform)
gravity-wave bores 6 h after the impulsive start (time t) of the
heating.[Adapted from Mapes and Houze (1995).]
-
2189Bulletin of the American Meteorological Society
radar echoes observed with the WP-3D radars over thetropical
ocean north of Australia as “intermediary,”because the echo
structure appeared to be neither ob-viously convective nor
obviously stratiform, and be-cause these echoes appeared to be in
the process ofconverting from convective to stratiform
structure.Mapes and Houze’s (1993) findings thus suggest thata
spectrum of structures exists between the convectiveand stratiform
extremes, an apparent contradiction tothe 1995 results.
The key to resolving this apparent contradiction isto keep in
mind that radar-echo structure is not asingle-valued function of
the vertical air motions. Ifthe intermediary radar-echo structures
represented asignificantly different kinematic category, the
mass-divergence spectrum (Fig. 9a) would not be so decid-edly
bimodal. Each instantaneous two-dimensionalcross section of radar
data (horizontal or vertical) con-tains only a selected subset of
the data in a three-di-mensional volume. A natural tendency is to
draw crosssections through extrema, which give the cross sec-tions
a rougher (more “convective”) appearance. Amore representative and
complete way of viewing theechoes is to combine all the data in the
three-dimen-sional volume of space scanned by a radar in a
singleplot and inquire about the statistics of the radar ech-oes in
that volume. Yuter and Houze (1995a–c) presentthe radar data from a
three-dimensional volume ofspace as a joint probability
distribution, which theyrefer to as a “contoured frequency by
altitude dia-gram,” or CFAD. The CFAD of radar reflectivity
con-sists of contours showing the range and relative nor-malized
frequency of occurrence of radar reflectivityvalues as a function
of reflectivity and height aboveground.
Figure 10a shows a CFAD of the radar reflectivityvalues observed
in a convective precipitation regionof a Florida cumulonimbus. The
distribution ofreflectivities at all altitudes is broad, and even
at highaltitudes a few rather high reflectivities occur,
indicat-ing the presence of graupel or small hail, which is
con-sistent with growth by riming in strong updrafts. Incontrast,
the stratiform region CFAD of a squall linein Kansas (Fig. 10c)3
shows a narrower distribution ofrelatively uniform values of
reflectivity at all altitudes,and at upper levels, the mode is a
much lower
reflectivity, consistent with a dominance of vapordeposition as
the primary microphysical growth modeof the particles. The mode of
the distribution increasesas the altitude decreases, indicating the
microphysi-cal growth and aggregation of the falling
precipitationparticles as they approach the melting level. Just
be-low the 0°C level (~4 km), the mode jumps to a highvalue,
corresponding to the typical reflectivity of themelting aggregates
falling through the layer.
As the storm within the three-dimensional volumescanned by a
radar ages, the CFAD of reflectivitychanges from a more convective
to a more stratiformstructure. Yuter and Houze (1995b) found
thatchangeover occurred quickly, even while the stormstill
contained a few intense up- and downdrafts andbefore the storm
appeared obviously stratiform in in-dividual vertical cross
sections of reflectivity. TheCFAD in Fig. 10b is for a time in the
mid- to late stagesof the same storm whose early convective stage
is rep-resented in Fig. 10a. Although the storm still containeda
wide range of vertical velocities, the reflectivity fieldwas
already statistically more stratiform, with a nar-rowing of the
distribution. This result of Yuter andHouze (1995b) suggests that
the intermediary radar-echo structures identified by Mapes and
Houze (1993)would have been categorized as stratiform had theybeen
viewed as CFADs.
Yuter and Houze’s (1995b,c) results further indi-cate that even
while some strong up- and downdraftswere still active, and the
radar echo pattern seen inhorizontal and vertical cross sections
was nonuniform,most of the mass flux of the storm was in updrafts
in-sufficiently strong to prevent ice particles from fall-ing or to
produce much liquid water to be collected bythe falling ice
particles. In a statistical sense, then, mostof the volume of the
storm was occupied by ice par-ticles growing primarily by vapor
deposition and fall-ing relative to the ground; that is, the
stratiform pre-cipitation process dominated in most of the volume
ofthe storm even though some spots within the storm stillhad strong
updrafts.
Mapes and Houze (1993) used a large set of air-borne
Doppler-radar measurements to determine thevertical profiles of
divergence in the convective, in-termediary, and stratiform radar
echo structures theyobserved over the tropical ocean north of
Australia
3Figure 10c is not from the same case as (a) and (b), because
the observations in the Florida case illustrated in (a) and (b)
ended beforethe fully developed stratiform region could be
observed. The convective region of the Kansas case illustrated in
(c) could not be usedfor (a) and (b) because the spatial resolution
of the radar data in the Kansas case was too coarse to show the
detailed structure of theconvective region. Hence, Fig. 10 had to
be pieced together from the two cases.
-
2190 Vol. 78, No. 10, October 1997
(Fig. 11). Around intermediary echo structures, theyfound
divergence in low to midlevels overlain by alayer of convergence
(Fig. 11b).4 This behavior is fur-ther evidence that the
intermediary echo regions were
kinematically more like stratiform than convectiveprecipitation
regions. However, the layer of conver-gence was at a somewhat
higher altitude than in thestratiform regions (cf. Figs. 11b,c).
This differencemay have been a result of sampling (the data in
Fig.11b were not necessarily obtained in the same stormsas the data
in Fig. 11c), or there may be a slight dif-ference between
intermediary and stratiform dynamics.
Figure 12 suggests possible relationships betweenthe pattern of
up- and downdrafts and the correspond-ing divergence profiles
during the various stages ofconvection. In regions dominated by
active convection(Fig. 1a), strongly buoyant, undiluted elements
rise,and mass continuity requires that the buoyant elementsbe
surrounded by a pressure-gradient force field, whichpushes air out
of the path of the top of the cell and intoward the base of the
buoyant updraft in the lower
FIG. 10. Examples of CFADS of reflectivity for (a) region of
young, vigorous convection; (b) a region of intermediary
convection;and (c) a stratiform precipitation region. Units of
contours are (frequency of occurrence, in percent of observations
at a given altitude)/(dz ddBZ), where dz = 0.4 km and ddBZ = 5.
Contours are drawn at intervals of 2.5% dBZ−1 km−1. The 5% dBZ−1
km−1 contour ishighlighted. [Adapted from Yuter and Houze
(1995b).]
4All the profiles in Fig. 11 should have a layer of divergence
atupper levels. The sensitivity of the radar used to obtain the
diver-gence profiles in Fig. 11 was, however, not sufficient to
observethe low intensity of the reflectivity at these high levels.
Other datashowed that the actual cloud tops were several kilometers
abovethe 200-mb level (~12 km), which is the highest level at
whichthe divergence could be reliably measured by the radar. All
themean profiles showed a net convergence below 12 km. On thelinear
pressure scale used in Fig. 11, the areas under the diver-gence
profiles are proportional to mass divergence. Since mass
con-vergence and divergence must balance over a full vertical
column,the net convergence seen in Fig. 11 implies that the higher
levelsmust have been characterized by net divergence in all three
cases.
-
2191Bulletin of the American Meteorological Society
troposphere. In a region containing one or more active,deep
convective elements, both positively buoyantupdrafts and negatively
buoyant downdrafts arepresent in the lower troposphere (Fig. 12a).
Over theregion of convection, the upward vertical mass flux ofthe
updrafts dominates over the downward mass flux
of the downdrafts. In the boundary layer, some of theconvergence
into the base of the buoyant updrafts iscanceled, in an
area-averaged sense, by divergenceassociated with downdrafts. Thus,
the maximum con-vergence into the convective region of a
mesoscaleconvective system is usually found at some heightabove the
boundary layer. Even so, the convective-region divergence profile
is two layered, with thelower troposphere characterized by net
convergenceand the upper troposphere by net divergence.
As a strongly buoyant element rises in a cumulon-imbus cloud, it
typically gets cut off from its supplyof high moist-static energy
air from below. When itreaches the stable tropopause, it spreads
out, collaps-ing into a flattened buoyant element (Lilly 1988).
Theconvergence produced by the pressure-gradient forceat the base
of the buoyant element is then elevated into
FIG. 11. Mean divergence observed by airborne Doppler radar in
tropical precipitating cloud systems over the ocean north
ofAustralia. [Adapted from Mapes and Houze (1993).]
FIG. 12. Conceptual models explaining the vertical profiles
ofdivergence in regions of precipitation dominated by (a)
young,vigorous convection; (b) intermediary convection; and
(c)stratiform precipitation. In (a) and (b), the rectangles
encloseregions of positive (+) and negative buoyancy (−) associated
withup- and downdrafts, respectively. The arrowheads indicate
thedirection of the pressure-gradient force field induced by
thebuoyancy field [see Fig. 7.1 of Houze (1993).] The
divergenceprofiles on the right-hand sides of (a) and (b) reflect
the net effectof these force fields. In (c), no attempt is made to
identifyindividual buoyant or negatively buoyant elements, as
theprecipitation (outlined by the radar-echo boundary) contains
theremains of previously vigorous buoyant and negatively
buoyantelements, which intermingle such that the upper levels
aregenerally positively buoyant, and the lower levels are
generallynegatively buoyant. The heavy shading in (c) indicates the
brightband just below the 0°C level and the remnants of old
cellsdistorted into fall streaks by the wind shear.
-
2192 Vol. 78, No. 10, October 1997
the upper troposphere. Figure 12b suggests that anintermediary
echo region, of the type analyzed byMapes and Houze (1993), is
dominated at upper lev-els by buoyant elements at this stage of
developmentand at lower levels by negatively buoyant downdrafts.The
net profile of divergence in a region of interme-diary echo
structure would then look like that on theright-hand side of Fig.
12b, with a layer of maximumconvergence in the upper troposphere
correspondingto the convergence at the base of the elevated,
hori-zontally spreading buoyant elements. This hypotheti-cal
profile of divergence is consistent with the observeddivergence
profile in intermediary regions (Fig. 11b).
In the stratiform region, negatively buoyantdowndrafts dominate
the lower troposphere, whileweakened, but still positively buoyant
updrafts, domi-nate the upper troposphere, and a strong melting
layerproduces a radar bright band in the middle troposphere(Fig.
1b). By this time, a lot of individual buoyant el-ements have risen
into the upper troposphere, reachedtheir level of zero buoyancy,
and spread laterally. Somereach the tropopause, but others entrain
and come torest at lower levels. The net result is that the whole
mid-to upper troposphere above the melting layer is filledwith
intermingled old buoyant elements (Fig. 12c). Atthis stage, no
strongly buoyant elements dominate themean buoyancy; rather, the
conglomeration of old el-ements dominates the picture. So Fig. 12c
shows themid- to upper levels as generally positively buoyantbut
does not attempt to identify individual elements.
A general region of negative buoyancy dominatesthe stratiform
region from the 0°C level down to theearth’s surface (Fig. 12c).
This region of negativebuoyancy, in a mature stratiform region, is
a combi-nation of old negatively buoyant convective
elements,cooling of the air in the melting layer, and entrainmentof
low moist-static-energy air from midlevels of theenvironment. The
pressure force field required by theresidual positive buoyancy
aloft and the negative buoy-ancy of both the downdrafts and the
melting layer leadto a maximum of convergence just above the 0°C
level.
By comparing the divergence profiles in Figs. 12band 12c, we
infer that they differ only in the altitudeof the maximum
convergence into the region, whichis slightly higher in the
intermediary case. Thus, whileon the order of one-third of the
reflectivity patternshave an intermediary appearance in
two-dimensionalcross sections of reflectivity fields (Mapes and
Houze1993), the air motions in the intermediary and strati-form
stages have qualitatively similar divergence pro-files,
characterized by a mid- to upper-level layer of
convergence sandwiched between upper and lowerdivergence layers.
This three-layered structure differsfundamentally from the
two-layered structure seen inthe convective region (Fig. 12a). The
three divergenceprofiles sketched in Figs. 12a–c are entirely
consistentwith the three observed profiles in Figs. 11a–c. Fromthis
we infer that the intermediary and stratiform pro-files both
contribute to the slow mode of atmosphericresponse identified by
Mapes and Houze (1995, Fig.9), while the convective region profiles
(Fig. 11a) con-tribute to the fast mode. To put it another way, the
first-order dynamical distinction to be made in evaluatinga
tropical convective precipitation system is betweenactively
convective portions (like Fig. 11a) and allother intermediary and
stratiform portions, which con-tain old convective elements and
well-developedstratiform rain.
9. Implications for identifyingconvective and
stratiformprecipitation
Because precipitation is a measure of the latent heatof
condensation released into the air, and since heat isthe result of
upward air motion, the precipitation fieldis diagnostic of the
rearrangement of the mass field bytropical convection. This
relationship of the heatingand mass fields to the precipitation
field is a primarymotivation of the Tropical Rainfall Measuring
Mis-sion (TRMM; Simpson 1988). It is now evident thatthe
large-scale mass field responds primarily to thetwo-layered mass
divergence profile in convectiveprecipitation regions and the
three-layered profile instratiform and intermediary precipitation
regions. Touse the precipitation field diagnostically, it is
neces-sary to separate the convective component of the
pre-cipitation from the remainder. This exercise, whileoften
referred to as “convective/stratiform separation,”is simply to
separate the convection portion of the pre-cipitation from the
remainder of the precipitation. Forthis reason,
convective/stratiform separation methodsthat seek the stratiform
component first and assign theremainder of the precipitation to the
convective cat-egory are risky since they are likely to combine
theintermediary precipitation (with its three-layered
massdivergence profile) with the convective (two-layeredmass
divergence profile) to produce a kinematicallynonuniform category
of rainfall.
Figure 13 illustrates a method (Churchill and Houze1984; Steiner
et al. 1995; Yuter and Houze 1997a,b)
-
2193Bulletin of the American Meteorological Society
that seeks the convective precipitation first and assignsthe
remainder to the “stratiform” category (whichshould really be
called stratiform plus intermediary).Each pixel of a Cartesian
reflectivity field is examinedand declared a convective center if
it exceeds a pre-scribed threshold of reflectivity or if it exceeds
thereflectivity in a region surrounding the pixel (~ 10 kmin
diameter) by a specified factor, which increases asthe reflectivity
at the convective center decreases. If apixel is declared a
convective center, then a 4–5-kmdiameter region surrounding the
convective center isdefined as convective precipitation. There is
no at-tempt to define a convective region smaller than 4–5km in
diameter. Often several convective centers areclose together and
the convective regions overlap, andthe union of these convective
areas may define a con-tiguous region of convective precipitation
tens to evenhundreds of kilometers across. Thus, the
separationmethod does not attempt to identify individual up-
anddowndraft regions, which may be smaller in scale than4–5 km, but
rather attempts to delineate general re-gions in which active
convection, like that idealizedin Fig. 1a, predominates and hence
has a mass diver-gence profile like that in Fig. 12a.
The method illustrated by Fig. 13 is very practicalbecause it is
applied to the reflectivity field alone. Itdoes not assume any
direct knowledge of the in-cloudair motions. The basic premise of
the method is thatif the reflectivity field is either very intense
or pock-marked with cores of intense reflectivity, then the
airmotions must be more like those in Fig. 1a than thosein Fig. 1d.
Steiner et al. (1995) and T.-C. Chen (1996,personal communication)
have tested this convective/stratiform separation method on
datasets for whichhigh-resolution dual-Doppler radar measurements
pro-vided simultaneous measurements of the vertical ve-locity and
radar reflectivity so that the vertical veloc-ity data could be
used as an independent test of theconvective/stratiform separation.
They first deter-mined the convective and stratiform regions by
apply-ing the method to the reflectivity field. Then they ex-amined
CFADs of the vertical velocity in these re-gions. The statistics of
the vertical velocity showedthat the air motions in the regions
designated convec-tive had a wide distribution of vertical
velocities, withpeaks > ~ 10 m s−1. In the regions that the
algorithmdesignated stratiform, the distribution was narrow,with
practically no vertical velocities exceeding2 m s−1 in absolute
value, with most of the values< 1 m s−1. These statistics
confirm that the air motionsin the regions identified as convective
by the reflec-
tivity algorithm were like those of Fig. 1a, while theair
motions in the stratiform regions were like thoseof Fig. 1d. More
tests of this type are needed.
10. Conclusions
In extratropical cyclones, lifting of air is generallywidespread
and weak, the precipitation produces lay-ered radar echoes, and the
precipitation is traditionallycalled “stratiform.”5 This conceptual
model contrastswith that of convective “cells,” which are
verticallyoriented cores of high reflectivity, embedded in
cumu-lonimbus, which form as a product of free convectiondriven by
buoyancy. When radar observations in theTropics in the early 1970s
showed stratiform radarechoes alongside convective cells, a paradox
arose:stratiform precipitation, thought previously to be aproduct
of widespread stable lifting, was falling fromtropical cumulonimbus
clouds as a product of buoy-ant convection. Stratiform
precipitation and convec-tion could no longer be viewed as mutually
exclusive.
This paper suggests resolving this paradox by us-ing the noun
convection to describe the overturning ofthe atmosphere that is
required to neutralize the verti-cal distribution of moist static
energy—just as convec-tion in an incompressible fluid neutralizes
an unstablevertical distribution of density—and by using the
ad-jectives convective and stratiform to describe the dif-ferent
types of precipitation seen within a given regionof
convection-generated cumulonimbus.
Within a region of precipitation where the convec-tion is young
and vigorous, strong updrafts producecells of heavy rain, seen on
radar as locally vertically
5There are several exceptions to the generally stratiform
natureof frontal precipitation. The stratiform precipitation often
containsweak embedded convective cells, often confined to a
potentiallyunstable layer aloft. Release of conditional symmetric
instabilitywithin the stratiform cloud mass may lead to embedded
lines ofsomewhat heavier precipitation. Embedded cells aloft and
linesof release of conditional symmetric instability do not change
thefundamentally stratiform nature of the precipitation but,
rather,give it a texture. Sometimes, however, the frontal cloud
systemcontains a “narrow cold frontal rainband,” in which strong,
con-vective-like air motions are forced by the convergence along
thesurface wind-shift line. The narrow cold-frontal band is
usuallyembedded in a much broader region of predominantly
stratiformclouds and precipitation. Sometimes the frontal
convergence trig-gers a squall line, in which a deep layer of
instability in the warmair is released by the frontal lifting and
leads to convective airmotions that dominate over and replace the
frontal air motions.For further details, see chapter 11 of Houze
(1993).
-
2194 Vol. 78, No. 10, October 1997
oriented cells of high reflectivity (Figs. 1a–c). Withina region
of precipitation where the convection is olderand much less
vigorous, a residual population ofweakened updrafts and downdrafts
produces precipi-tation that appears stratiform on radar (Figs.
1d–f). Asvigorous cells weaken, they pass through an interme-diary
stage of development, when two-dimensionalcross sections of radar
data are ambiguous but statis-tics of the reflectivity and vertical
air velocity in athree-dimensional volume of space indicate that
theradar echoes are basically stratiform; that is, they havea
narrow distribution of reflectivity with a vertical dis-tribution
suggestive of precipitation particles grow-ing mainly by vapor
deposition and vertical veloci-ties < ~ 1 m s−1. In contrast,
the statistics of the ech-oes in a three-dimensional volume
containing young,vigorous convection show a broad distribution
ofreflectivity. Particles there appear to be able to growby
collection (coalescence and riming), which ac-counts for the
existence of high reflectivities in theoutliers of the reflectivity
distribution. Statistics of the
air motions confirm that while most of the vertical ve-locities
in the convective regions are relatively small,there are a few
large vertical velocities (> ~ 2 m s−1)capable of carrying
particles upward and generatinghigh cloud liquid water contents for
the growing par-ticles to collect.
Large-scale dynamics of the tropical atmosphere areprofoundly
affected by cumulonimbus clouds. How-ever, the effects of the
clouds upon the atmosphere aredistinctly subdivided such that part
of the atmosphericresponse derives from the young vigorous
convectiveprecipitation areas, characterized by radar echoes in
theform of cells, while another part derives from the olderweaker
part of the convection, where the precipitationechoes are
intermediary and stratiform. The young,vigorous cells induce an
atmospheric response that istwo layered, with air converging into
the active con-vection at low levels and diverging aloft. The
older,weaker intermediary and stratiform precipitation ar-eas
induce a three-layered response, in which the en-vironment
converges into the weak precipitation areaat midlevels and diverges
out from it at lower and up-per levels.
These results suggest that it is important to sepa-rate the
convective precipitation from all the rest fortwo reasons:
1) The microphysical growth processes at work in theconvective
precipitation areas are profoundly dif-ferent from those in the
intermediary and stratiformprecipitation areas. In convective
precipitation ar-eas, precipitation particles increase their mass
pri-marily by collection of cloud water (coalescenceand/or riming).
In intermediary and stratiform pre-cipitation areas, the
precipitation particles increasetheir mass by vapor diffusion. If
high-resolutionnumerical prediction models are to forecast
accu-rately the location, amount, and type of precipita-tion, they
must represent (either parametrically orexplicitly) the
microphysical growth processesaccurately. Radar data can validate
these modelsonly if the echoes are accurately subdivided toseparate
the active vigorous convective echoesfrom the remainder of the
precipitation.
2) Global climate and circulation models must simu-late the role
of convection accurately. To validatesuch models, satellite
observational programs suchas TRMM (Simpson 1988) are attempting to
mapthe precipitation over the globe three dimension-ally.
Instrumentation in TRMM will determine thevertical as well as the
horizontal structure of pre-
FIG. 13. Schematic illustrating a method for separating
theconvective component of precipitation from the rest of
theprecipitation. The illustration assumes a 2-km Cartesian
gridspacing for the data points; however, the method may be
appliedat different resolutions and could be adapted to polar
coordinates.The algorithm examines the reflectivity at each grid
point. If thatreflectivity exceeds some prescribed high value, or
if it exceedsthe mean reflectivity in a “background region”
(lightly shadedregion, ~ 10 km in diameter) surrounding the point
by somespecified amount, the point is declared a convective center.
If thepoint is declared a convective center, then that point and
all thepoints in a small region (~ 4–5 km in diameter) surrounding
theconvective center constitute a “convective region” centered on
thepoint. This exercise is repeated for each data point in the grid
andthe union of all the convection regions makes up the
totalconvective precipitation region. [Adapted from Steiner et
al.(1995).]
-
2195Bulletin of the American Meteorological Society
cipitation throughout the Tropics. The primary goalis to use
this information to determine the four-di-mensional (i.e., temporal
and spatial) patterns oflatent heating over the whole Tropics.
Algorithmsapplied to the TRMM data will attempt to identifythe two
principal modes of heating, by separatingthe convective component
of the precipitation, andhence the associated heating, from the
remainder.The success of TRMM will hinge on the accuracyof this
separation.
From 1) and 2) above, it is clear that the distinctionbetween
young convective precipitation and olderstratiform precipitation in
convection-generatedclouds is important in order to forecast
precipitationaccurately and to evaluate the effects of tropical
con-vection on the global circulation. These applicationsare of
great practical significance. The processes rep-resented in Fig. 1
remain, however, without muchquantitative documentation. If models
for numericalprediction and climate simulation are going to
im-prove, these processes must be measured and quanti-fied. The
only viable way to do this is to mount fieldprojects that can
determine the dynamical and micro-physical processes on space and
time resolutions ad-equate to resolve individual convective cells
and dis-tinguish unambiguously the major cloud microphysi-cal
mechanisms, especially at all altitudes above the0°C level. For
these projects, instrumentation and ob-servational platforms must
be developed to improveon our current capabilities to measure
vertical airmotion, water vapor, temperature, pressure, and iceand
liquid particle mass, shapes, and concentrations.Without such
empirical improvements in our knowl-edge base, it is hard to
foresee real progress in the abil-ity to model processes in the
atmosphere that are sen-sitive to convective and stratiform
precipitation pro-cesses in cumulonimbus clouds.
Acknowledgments. This paper is based on the author’s
presen-tation to the Tropical Rainfall Measuring Mission U.S.
ScienceTeam Meeting, Greenbelt, Maryland, July 1996. The author
ben-efited greatly from the comments of Dr. Pauline Austin,
Profes-sor Colleen Leary, Dr. Brian Mapes, Professor Bradley Smull,
andDr. Sandra Yuter. Candace Gudmundson edited the manuscriptand
Kay Dewar drafted some of the figures. This research has
beensupported by NOAA Cooperative Agreement NA67RJ0155(JISAO
Contribution 351), National Science Foundation GrantATM-9409988,
and NASA Award NAGS-1599.
References
Battan, L. J., 1973: Radar Observations of the Atmosphere.
Uni-versity of Chicago Press, 324 pp.
Braun, S. A., and R. A. Houze Jr., 1995: Melting and freezing
ina mesoscale convective system. Quart. J. Roy. Meteor. Soc.,121A,
55–77.
Bretherton, C. S., and P. K. Smolarkiewicz, 1989: Gravity
waves,compensating subsidence, and detrainment around
cumulusclouds. J. Atmos. Sci., 46, 740–759.
Byers, H. R., and R. R. Braham Jr., 1949: The Thunderstorm.
U.S.Government Printing Office, 287 pp.
Cheng, C.-P., and R. A. Houze Jr., 1979: The distribution of
con-vective and mesoscale precipitation in GATE radar echo
pat-terns. Mon. Wea. Rev., 107, 1370–1381.
Churchill, D. D., and R. A. Houze Jr., 1984: Development
andstructure of winter monsoon cloud clusters on 10 December1978.
J. Atmos. Sci., 41, 553–574.
Hobbs, P. V., 1974: Ice Physics. Oxford Press, 837 pp.Houghton,
J. G., 1968: On precipitation mechanisms and their
artificial modification. J. Appl. Meteor., 7, 851–859.Houze, R.
A., Jr., 1975: Squall lines observed in the vicinity of
the researcher during phase III of GATE. Preprints, 16th Ra-dar
Meteorology Conf., Houston, TX, American Meteorologi-cal Society,
206–209.
——, 1977: Structure and dynamics of a tropical squall-line
sys-tem. Mon. Wea. Rev., 105, 1540–1567.
——, 1989: Observed structure of mesoscale convective systemsand
implications for large-scale heating. Quart. J. Roy. Me-teor. Soc.,
115, 425–461.
——, 1993: Cloud Dynamics. Academic Press, 573 pp.——, and D. D.
Churchill, 1987: Mesoscale organization and
cloud microphysics in a Bay of Bengal depression. J. Atmos.Sci.,
44, 1845–1867.
——, B. F. Smull, and P. Dodge, 1990: Mesoscale organizationof
springtime rainstorms in Oklahoma. Mon. Wea. Rev., 118,613–654.
Huschke, R. E., 1959: Glossary of Meteorology. Amer.
Meteor.Soc., 638 pp.
Leary, C. A., 1984: Precipitation structure of the cloud
clusters ina tropical easterly wave. Mon. Wea. Rev., 112,
313–325.
——, and R. A. Houze Jr., 1979a: The structure and evolution
ofconvection in a tropical cloud cluster. J. Atmos. Sci., 36,
437–457.
——, and ——, 1979b: Melting and evaporation of hydromete-ors in
precipitation from anvil clouds of deep tropical convec-tion. J.
Atmos. Sci., 36, 669–679.
LeMone, M. A., and E. J. Zipser, 1980: Cumulonimbus
verticalvelocity events in GATE. Part I: Diameter, intensity, and
massflux. J. Atmos. Sci., 37, 2444–2457.
Lilly, D. K., 1988: Cirrus outflow dynamics. J. Atmos. Sci.,
45,1594–1605.
Mapes, B. E., 1993: Gregarious tropical convection. J. Atmos.
Sci.,50, 2026–2037.
——, and R. A. Houze Jr., 1993: Cloud clusters and superclus-ters
over the oceanic warm pool. Mon. Wea. Rev., 121, 1398–1415.
——, and ——, 1995: Diabatic divergence profiles in westernPacific
mesoscale convective systems. J. Atmos. Sci., 52,1807–1828.
-
2196 Vol. 78, No. 10, October 1997
Marshall, J. S., 1953: Frontal precipitation and lightning
observedby radar. Can. J. Phys., 31, 194–203.
Riehl, H., 1979: Climate and Weather in the Tropics.
AcademicPress, 611 pp.
Rutledge, S. A., and R. A. Houze Jr., 1987: A diagnostic
model-ing study of the trailing stratiform region of a midlatitude
squallline. J. Atmos. Sci., 44, 2640–2656.
——, and D. R. MacGorman, 1988: Cloud-to-ground lightning inthe
10–11 June 1985 mesoscale convective system observedduring
Oklahoma–Kansas PRE-STORM. Mon. Wea. Rev.,116, 1393–1408.
Shupiatsky, A. B., A. I. Korotov, V. D. Menshenin, R.
S.Pastushkov, and M. Jovasevic, 1975: Radar investigations
ofevolution of clouds in the eastern Atlantic. GATE Rep. 14,
Vol.II, International Council of Scientific Unions/World
Meteoro-logical Organization, 177–187. [Available from WMO,
CasePostale 2300, CH-1211, Geneva 2, Switzerland.]
——, ——, and R. S. Pastushkov, 1976a: Radar investigationsof the
evolution of clouds in the East Atlantic, in TROPEX-74. Atmosphere
(in Russian), Vol. 1, Gidrometeoizdat, 508–514.
——, G. N. Evseonok, and A. I. Korotov, 1976b: Complex
inves-tigations of clouds in the ITCZ with the help of satellite
andship radar equipment, in TROPEX-74. Atmosphere (in Rus-sian),
Vol. 1, Gidrometeoizdat, 508–514.
Simpson, J., 1988: TRMM: A satellite mission to measure
tropi-cal rainfall. NASA, 94 pp. [Available from Goddard
SpaceFlight Center, Code 912, Greenbelt, MD 20771.]
Steiner, M., R. A. Houze Jr., and S. E. Yuter, 1995:
Climatologi-cal characterization of three-dimensional storm
structure fromoperational radar and rain gauge data. J. Appl.
Meteor., 34,1978–2007.
Yuter, S. E., and R. A. Houze Jr., 1995a: Three-dimensional
ki-nematic and microphysical evolution of Florida cumulonim-bus.
Part I: Spatial distribution of updrafts, downdrafts,
andprecipitation. Mon. Wea. Rev., 123, 1921–1940.
——, and ——, 1995b: Three-dimensional kinematic and
micro-physical evolution of Florida cumulonimbus. Part II:
Fre-quency distributions of vertical velocity, reflectivity, and
dif-ferential reflectivity. Mon. Wea. Rev., 123, 1941–1963.
——, and ——, 1995c: Three-dimensional kinematic and
micro-physical evolution of Florida cumulonimbus. Part III:
Verti-cal mass transport, mass divergence, and synthesis. Mon.
Wea.Rev., 123, 1964–1983.
——, and ——, 1997a: Measurements of raindrop size distribu-tion
over the Pacific warm pool and implications for Z–R re-lations. J.
Appl. Meteor., 36, 847–867.
——, and ——, 1997b: Natural variability of precipitating
cloudsover the western Pacific warm pool. Quart. J. Roy.
Meteor.Soc., in press.
Zipser, E. J., and M. A. LeMone, 1980: Cumulonimbus
verticalvelocity events in GATE. Part II: Synthesis and model
corestructure. J. Atmos. Sci., 37, 2458–2469.