Monsoon convection in the Himalayan region as seen … · merow et al., 1998) required remapping to...

QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETYQ. J. R. Meteorol. Soc. 133: 1389–1411 (2007)Published online 3 August 2007 in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/qj.106

Monsoon convection in the Himalayan region as seen by theTRMM Precipitation Radar

Robert A. Houze Jr.,* Darren C. Wilton and Bradley F. SmullUniversity of Washington, Seattle, USA

ABSTRACT: Three-dimensional structure of summer monsoon convection in the Himalayan region and its overallvariability are examined by analyzing data from the Tropical Rainfall Measuring Mission (TRMM) Precipitation Radarover the June–September seasons of 2002 and 2003. Statistics are compiled for both convective and stratiform componentsof the observed radar echoes.

Deep intense convective echoes (40 dBZ echo reaching heights >10 km) occur primarily just upstream (south) of andover the lower elevations of the Himalayan barrier, especially in the northwestern concave indentation of the barrier.The deep intense convective echoes are vertically erect, consistent with the relatively weak environmental shear. Theysometimes extend above 17 km, indicating that exceptionally strong updraughts loft graupel to high altitudes. Occasionally,scattered isolated deep intense convective echoes occur over the Tibetan Plateau.

Wide intense convective echoes (40 dBZ echo >1000 km2 in horizontal dimension) also occur preferentially justupstream of and over the lower elevations of the Himalayas, most frequently in the northwestern indentation of the barrier.The wide intense echoes have an additional tendency to occur along the central portion of the Himalayas, and they seldomif ever occur over the Tibetan Plateau. The wide intense echoes exhibit three mesoscale structures: amorphous areas,lines parallel to the mountain barrier, and arc-shaped squall lines perpendicular to and propagating parallel to the steepHimalayan barrier. The latter are rare, generally weaker than those seen in other parts of the world, and occur when amidlevel jet is aligned with the Himalayan escarpment.

Deep and wide intense convective echoes over the northwestern subcontinent tend to occur where the low-level moistlayer of monsoon air from the Arabian Sea meets dry downslope flow, in a manner reminiscent of severe convectionleeward of the Rocky Mountains in the central USA. As the low-level layer of moist air from the sea moves over the hotarid northwestern subcontinent, it is capped by an elevated layer of dry air advected off the Afghan or Tibetan Plateau.The capped low-level monsoonal airflow accumulates instability via surface heating until this instability is released byorographically induced lifting immediately adjacent to or directly over the foothills of the Himalayas.

Broad (>50 000 km2 in area) stratiform echoes occur in the eastern and central portions of the Himalayan region inconnection with Bay of Bengal depressions. Their centroids are most frequent just upstream of the Himalayas, in the regionof the concave indentation of the barrier at the eastern end of the range. The steep topography apparently enhances theformation and longevity of the broad stratiform echoes. Monsoonal depressions provide a moist maritime environment forthe convection, evidently allowing mesoscale systems to develop larger stratiform echoes than in the western Himalayanregion. Copyright 2007 Royal Meteorological Society

KEY WORDS convective echoes; deep convection; orographic precipitation; Bay of Bengal; Arabian Sea

Received 3 July 2006; Revised 26 January 2007; Accepted 2 March 2007

1. Introduction

The South Asian monsoon provides an ideal settingto observe orographically influenced precipitation undervery warm and highly unstable conditions. During theboreal summer, warm moisture-laden low-level windsbring air from the Bay of Bengal and the ArabianSea over the hot South Asian continent (Figure 1(a);Rao, 1976; Krishnamurti, 1985; Webster, 1987a,b; John-son and Houze, 1987; Das, 1995; Pant and Kumar,1997; Webster et al., 1998). Upon making landfall, thelow-level winds encounter the Himalayas, the world’s

* Correspondence to: Robert A. Houze, University of Washington, 604ATG Bldg., Box 351640, Seattle, Washington 98195, USA.E-mail: [email protected]

largest mountain barrier (Figure 2(a); Murakami, 1987).During the monsoon, large amounts of rain fall inthis mountain-influenced regime over the northern Asiansubcontinent, where the upper-level winds are weak,between strong easterlies to the south and westerlies tothe north (Figure 1(b)). Most of the rain in this regionfalls during the monsoon season of June–September(Nayava, 1974; Shrestha, 2000). The physical interac-tion between the southwesterly monsoon flow and theSouth Asian topography affects the climatological dis-tribution of precipitation in the region (Figure 2(b)).Studies dating back to the colonial period in India(Hill, 1881) have examined the distribution of surfacerainfall in South Asia and its relation to orographicfeatures.

Copyright 2007 Royal Meteorological Society

1390 R. A. HOUZE JR, ET AL.

40°N

35

(a)

(b)

30

25

20

15

10

5

0

40°N

40°E

35

45 50 55 60 65 70 75 80 85 90 95 100 105 110 115

30

25

20

15

10

5

0

10 ms-1

10 ms-1

elev.

6.05.55.04.54.03.53.02.52.01.51.00.50.10.0

Figure 1. Mean wind at (a) 1000 mb and (b) 200 mb superimposed on topography. Wind data were obtained from the NOAA-CIRES ClimateDiagnostics Center, Boulder, Colorado (http://www.cdc.noaa.gov/) and plotted in MountainZebra (James et al., 2000). The colour scale shows

topography elevation in kilometres.

Rather than surface rain mapping, the goal of ourresearch is to gain insight into the physical mechanismsby which the heavy monsoon precipitation is produced.To gain this insight, we examine the three-dimensionalstructure of the storms producing intense monsoon pre-cipitation. For this we turn to meteorological radar.Narayanan (1967) foresaw radars as providing ‘betterinsight into the intricacies of the hitherto hidden phenom-ena of monsoon clouds. . .’ Specifically, we analyze datafrom the satellite-borne Precipitation Radar (PR) on theTropical Rainfall Measuring Mission (TRMM) satellite(Kummerow et al., 1998; Kummerow et al., 2000). Thesatellite-borne radar is especially well suited for deter-mining storm structure in mountainous regions since itpoints downwards from high altitude. It can detect pre-cipitation (with a horizontal resolution of ∼5 km) bothover peaks of terrain and within valleys, which would beblocked from the view of a ground-based radar (Joss andWaldvogel, 1990). The TRMM satellite’s orbit (whichis confined between 36 °N and 36 °S) results in multipledaily samples of the South Asian region, including theHimalayas. Since its launch in late 1997, the TRMM PR

has accumulated a climatological sample of radar dataover the monsoon region. Our objective is to use theTRMM PR data to assess the three-dimensional struc-tures of the radar echoes in the Himalayan region inrelation to the details of the topography and proximityto surrounding oceans.

2. Data acquisition, processing, and analysis

2.1. Acquisition of the PR dataset

The TRMM data products obtained for this study werethe qualitative rain characteristics field (TRMM prod-uct 2A23) and the gridded, attenuation-corrected three-dimensional reflectivity field (product 2A25). Version5 of the 2A23 and 2A25 products (Awaka et al.,1997; Iguchi et al., 2000a,b) was used. The rain-type field includes the categorization of the echo asconvective or stratiform (see Section 2.3 below). Wecorrected the 2A23 version 5 data to eliminate thepotential misclassification of shallow isolated precipita-tion, in a manner consistent with one of the updates

Copyright 2007 Royal Meteorological Society Q. J. R. Meteorol. Soc. 133: 1389–1411 (2007)DOI: 10.1002/qj

MONSOON CONVECTION IN THE HIMALAYAN REGION 1391

35°N

30

(a)

(b)

25

20

35°N

30

25

20

6.5

elev.

6.0

5.5

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

2.0mm hr-1

1.8

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

Figure 2. (a) South Asian topography in kilometres of elevation. (b) Rain rate in millimetres per hour averaged over all time, including non-rainingperiods, for 0.1° × 0.1° grid squares derived from TRMM overpasses from 1998 to 2004.

employed in the current version 6 of this product,described at the Japanese Space Agency (NASDA)TRMM website (http://www.eorc.nasda.go.jp/TRMM/document/pr manual/pr manual v6.pdf). Since weapplied this correction, updating to version 6 would havehad no effect on the results of this study. The 2A23and 2A25 data contain the orbital time and geoloca-tion data (latitude and longitude coordinates) used in thisstudy. For further discussion of these TRMM products,see Schumacher and Houze (2003a,b) and Houze et al.(2004).

The TRMM PR dataset is obtainable via the NASADistributed Active Archive Center (DAAC) website(http://disc.gsfc.nasa.gov/). However, in order to reducethe volume of data, we obtained reprocessed subsets ofthe 2A23 and 2A25 data directly from NASA (ErichStocker, personal communication) that included partialorbit files containing only the data from our region ofstudy.

2.2. Remapping the PR data

To analyze the precipitation processes in the monsoonconvection documented in this study, we used softwaredesigned to easily display and assess the TRMM PRdata. The primary visualization tool was MountainZebra(James et al., 2000), a specialized version of the NCARZebra software (Corbet et al., 1994), which permitsvisualization of the data overlaid on detailed topography.

MountainZebra requires a Cartesian grid structure asinput, thus the non-Cartesian native PR data (Kum-merow et al., 1998) required remapping to a regularlatitude–longitude grid. The data were remapped toCartesian grid points separated by 250 m in the verti-cal and 0.05° (approximately 5 km) in both latitude andlongitude. Prior to remapping, small corrections wereapplied to the geolocation of upper-level data as fol-lows. The PR has a beamwidth of 0.71° and a scan angleof ±17°, resulting in a swath width of 247.25 km anda horizontal ground resolution of 5 km at the orbital



height of 402.5 km. (The TRMM altitude was boostedfrom 350 km to 402.5 km in August 2001.) The PRdata were collected in 49 angle bins separated by 0.71°

(Figure 3(a)). The observable vertical range of the PRextends from the surface (with surface clutter removedby the rain profiling algorithm) to a height of 19.75 kmabove the earth ellipsoid (surface of an idealized ocean-covered earth). Both the 2A23 and 2A25 algorithmsassign the geolocation for each ray to the position ofthe range bin at the earth ellipsoid for that ray. Thus, allrange bins of the reflectivity data in each angle ray arereferenced to the same coordinate pair at the earth ellip-soid, despite the angular offset (Toshio Iguchi, personalcommunication). At nadir, there is no offset; however, theoffset increases as the angle from nadir increases, withthose bins at the highest altitude having the most sig-nificant offset. Our data processing involved shifting thenominal latitude and longitude coordinates of the binsabove the earth ellipsoid to allow for this offset, suchthat the coordinates in space correspond to the surfacecoordinates vertically below the bin (Figure 3(b)). Thesecorrected locations were the first step in remapping thedata into a regularly spaced three-dimensional Cartesiangrid for viewing in Zebra and for other calculations.

To complete the remapping, we emulated thethree-dimensional interpolation method employed byREORDER, a method often used with airborne radarsystems (Oye and Case, 1995). However, we also takeadvantage of the regularity in the retrieved data interms of the distance from the satellite and the smallalong-beam sample depth (range bin). Our approach isan adaptation of the Cressman (1959) method, whichemploys a radius of influence N . However, we use asmall radius of influence (N = 4.25 km), which is smallenough to retain most of the detail of the recorded data,while minimizing gaps in the Cartesian-gridded field. Theremapped value for radar reflectivity at a grid point at

a given height incorporates the data within a horizontalradius of 4.25 km for height bins both at and immediatelyabove and below the level in question. That is, data ina 0.75 km-deep layer centred on the height of the gridpoint are considered to determine the remapped valueof reflectivity at the grid point. The echo features weexamine in this study are of a sufficiently large scale thatany slight smoothing that results from this remappingprocedure has no effect on the conclusions we draw fromthe TRMM PR data. Details of our remapping method arein the Appendix.

2.3. Methods of analyzing the radar data

2.3.1. Using radar echoes to indicate precipitationprocesses

Houze (1997) has described the microphysical anddynamical differences between the convective and strat-iform components of a precipitating tropical cloud sys-tem, and methodology exists for objectively separatingradar echoes into convective and stratiform components(Churchill and Houze, 1984; Steiner et al., 1995; Awakaet al., 1997). The TRMM PR rain-type fields in product2A23 have been subjected to a separation algorithm basedon these methods, and we use the separation provided bythis product as our first-order subdivision of the data. Wefurther identify certain extreme degrees of convective andstratiform structure to select the best defined examples ofthese echo types in Himalayan monsoon convection.

2.3.2. Analysis of convective echoes: definition ofintense convection

To analyze the characteristics of convective radar echoes,we examine only those echoes identified by TRMMproduct 2A23 as convective. An intense convective echois defined as a subvolume of a convective echo consistingof an array of contiguous Cartesian bins with reflectivity

400

350

300

250

200

150

100

50

0-100 0

0

5

10

15

20(a) (b)

100 -125 -120 -115 -110 -105

Distance from nadir (km)

Bin

hei

ght (

km)

Alti

tude

(km

)

Figure 3. Approximate geometry of TRMM overpass. (a) TRMM satellite orbit altitude of approximately 400 km with 17° scan width (0.71°

beam width) results in 49 angle bins and 80 height bins in the lowest 19.75 km, indicated by near-vertical lines at the bottom of the figure.(b) Exploded view of the inset square in (a), showing off-nadir bins. Arrows indicate the corrected reference position for the three outer beamsat three different heights. Note that no correction is needed for the nadir position, and the amount of correction increases with angle from nadir

and altitude.



≥40 dBZ. Defined in this way, intense convective echoesmay comprise either single or multiple convective ‘cells’,as identified by discrete, vertically oriented reflectivitymaxima occurring on the scale of individual buoyantconvective updraughts. Often, the intense convectiveecho consisted of 40+ dBZ echo conjoined only atsome level, usually at lower levels. Since stratiform echowas eliminated by the 2A23 algorithm, stratiform bright-band returns did not affect the identification of intenseconvective echoes.

For purposes of analysis and discussion, two specialcategories of intense convective echoes are defined:deep and wide. Deep intense convective echoes arethose for which the top heights of the 40+ dBZ echovolume exceed 10 km. Wide intense convective echoesare defined as those with horizontally contiguous regionsof 40 dBZ echo exceeding 1000 km2 in area at somealtitude. The altitude of the areal maxima ranged from1.0 km to 4.5 km amsl.

Echoes with intense convective components classifiedas deep or wide were further examined with the aidof MountainZebra. The 117 deepest and 121 widestintense convective echo cases were subjected to detailedanalysis.

2.3.3. Analysis of stratiform echoes: definition of abroad stratiform area

The locations and sizes of stratiform areas were deter-mined by finding contiguous grid areas that were clas-sified as stratiform precipitation by the 2A23 algo-rithm. We applied no reflectivity thresholds to thedata. However, the PR has a rather low sensitivityand can detect only reflectivities in excess of approx-imately 17 dBZ (Kummerow et al., 1998). A broadstratiform area in the PR data was defined as a strat-iform echo region exceeding 50 000 km2 in horizontalextent.

2.4. Ancillary datasets

Both intense convective echoes and broad stratiformareas were analyzed visually with the aid of Moun-tainZebra to asses their vertical structures in relationto ancillary synoptic and satellite data. Climatologicalwind plots (Figure 1) were obtained though the NationalCenters for Environmental Prediction (NCEP)–NationalCenter for Atmospheric Research (NCAR) global reanal-ysis dataset (Kalnay et al., 1996). Individual synopticmaps were generated for us by the Indian NationalCentre for Medium Range Weather Forecasting (NCM-RWF) from their global T80 model. These reanaly-ses incorporated regional data exclusive to the NCM-RWF (Rajagopal et al. 2001; Goswami and Rajagopal2003). Sounding data were obtained through the Uni-versity of Wyoming (http://weather.uwyo.edu/upperair/).Visible and infrared satellite imagery from Meteosat-5 were obtained from the European Organization forthe Exploitation of Meteorological Satellites (EUMET-SAT, http://www.eumetsat.int/). Lightning flash data

from the Lightning Imaging Sensor (LIS; Christian et al.,1999) on board the TRMM satellite were obtainedvia the Global Hydrology Resource Center (GHRC;http://ghrc.msfc.nasa.gov/ghrc.html).

3. Locations of extreme precipitation events relativeto the Himalayas

Figure 4 shows the locations of deep and wide intenseconvective echoes and broad stratiform echoes over thetwo monsoon seasons of this study. The data are from allTRMM overpasses for which the PR swath included atleast a portion of the three rectangles shown in the figureand referred to here as the western, central, and easternsubregions.

35°N

35°N

30

(a)

(b)

(c)

25

20

30

25

20

35°N

30

25

20

67°E 78 89 100

Ocean Lowland Foothills Mountain

Figure 4. Regional distribution of (a) deep intense convective echoes(40 dBZ echo >10 km in height), (b) wide intense convective echoes(40 dBZ echo >1000 km2 in area), and (c) large stratiform areas(>50 000 km2 in area). The rectangles indicate locations of west-ern, central, and eastern Himalayan subregions. The terrain heightcategories are: lowland 0–300 m, foothills 300–3000 m, mountain

>3000 m.



3.1. Occurrence of intense convection as a function ofdistance from the barrier

The southwest monsoon undergoes intraseasonal varia-tions known as active and break periods (Webster, 1987b;Webster et al., 1998; Lawrence and Webster, 2002; Wanget al., 2005; Webster, 2006). It also undergoes day-to-dayfluctuations. However, the daily wind patterns throughoutthe season retain a persistent gross similarity to the clima-tological mean (Figure 1). When deep convection occursduring active periods in the regions shown in Figure 4,the low-level moist flow (Figure 1(a)) tends to underrundry flow coming off the Afghan or Tibetan Plateau.

Carlson et al. (1983) described how ‘a particular con-figuration of topography and air flow can produce alow-level restraining inversion or “lid” which focusesthe location and even enhances the intensity of severelocal storms’. While capping inversions had ‘oftenbeen attributed to subsidence’, the Carlson et al. model‘demonstrated that the lid may originate from differen-tial advection of a hot, dry mixed layer from an elevatedplateau over a cooler, moister layer’. Sawyer (1947) pre-sented a similar interpretation of the environmental fac-tors affecting monsoon convection occurring over north-western India. He analyzed soundings from ten days (23August–1 September) of the 1945 monsoon. and inferredthat the low-level moist layer of air entering India fromthe southwest (Figure 1(a)) underran dry continental airadvected off the Afghan or Tibetan Plateau (Figure 5(b)).According to this idea, the dry air flowing off the highplateau caps the moist layer and provides a lid, whichallows the low-level southwesterly flow with high mois-ture content to build up extreme buoyancy via sensibleheat fluxes as the air moves over the flat desert region ofthe northwestern subcontinent. The lid effectively delaysthe release of the deep convective instability. Sawyer andCarlson et al. pointed out that intense convection mayoccasionally break through the lid or break out along thelid edge (Figure 5(c)).

In the case of severe convection over the US HighPlains, as analyzed by Carlson et al. (1983), deep con-vective updraughts are frequently triggered near a ‘dry-line’, where the top of the moist layer intersects thesloping terrain (Bluestein 1993, pp. 282–290). By anal-ogy, we might expect intense deep convection to breakout just upstream of or over the lower portion of themountain barrier, as seems to be the case in Figures 4(a)and (b), and as will be shown statistically in Section 7.Dryline-type convection, however, breaks out only whenthe stable lid is subjected to organized lifting (e.g. theupward motion ahead of a short-wave trough). By anal-ogy, in the Himalayan region deep convection might thusbe expected to erupt where the potentially unstable col-umn defined by the previously capped moist monsoonallayer underrunning the warm dry layer is subjected to oro-graphic lifting. Such lifting would be expected directlyover the foothills (in connection with upslope flow) orimmediately upwind of the mountain barrier towardswhich the moist layer is flowing, by diurnal processes

interacting with the larger-scale flow, or as a result ofblocking effects extending well upstream of the barrier(Grossman and Durran, 1984).

3.2. West–east variability in the occurrence of intenseconvection

Figure 4(a) shows the locations of deep intense convec-tive echoes. They occur occasionally over the extremelyhigh terrain of the Tibetan Plateau, in the eastern andcentral subregions. However, both deep and wide intenseconvective echoes are most frequent in the western sub-region, especially within the concave indentation at thenorthwestern end of the Himalayan range (Figures 4(a)and (b)). The occurrence of the most intense convec-tion in the northwestern indentation of the Himalayas isconsistent with the results of Barros et al. (2004), whoshowed a strong maximum of lightning frequency in thatregion (Figure 6). The concentration of the deep and wideintense convective echoes in the northwestern indentationof the Himalayan barrier suggests further that the com-plex shape of the mountain barrier may play a role in thedistribution of intense convection. The concave indenta-tion in the western subregion may be the natural preferredend point of trajectories of the Arabian Sea air heatedby passage over the desert. Alternatively, the concaveindentation in the barrier may affect the flow in a waythat concentrates low-level convergence in the region ofthe indentation, as in the case of precipitation maximain concave indentations on the Mediterranean side of theEuropean Alps (Frei and Schar, 1998).

Figure 4(b) shows the locations of the wide intenseconvective echoes, which consist of groups of intensecells comprising portions of mesoscale convective sys-tems (MCSs; Houze, 2004). Similar to deep intense con-vective echoes, the wide ones concentrate in the north-western indentation (cf. Figures 4(a) and (b)). However,the wide intense convective echoes tend also to occurin the central subregion, in front of the central, convex,steep, wall-like Himalayan barrier.

3.3. Occurrence of broad stratiform events relative toterrain

Figure 4(c) shows that the broad stratiform precipitationevents occur primarily in the eastern and central subre-gions, roughly at the terminus of the axis of mean south-westerly low-level flow, in the zones of rapidly risingterrain immediately northeast and northwest of the Bay ofBengal (Figure 1(a)). They are clustered upstream of themountain barrier and are grouped within a concave inden-tation in the mountain barrier. Compared to the westernsubregion, where low-level flow from the Arabian Seaexperiences a long passage over the desert, the mountainbarrier and the sharp indentation of lower terrain extend-ing northeastwards from the Bay of Bengal between theHimalayan Barrier and the Arakan Mountains of Myan-mar, are much closer to where the monsoon flow makeslandfall. The preferential occurrence of well-defined strat-iform regions in the northeastern subcontinent suggests



Figure 5. Sawyer’s (1947) conceptual model of the monsoon environment over the northwestern subcontinent, as observed during 23 August–1September 1945. (a) Synoptic and orographic context. The wide arrow indicates the general direction of the low-level monsoon flow. The

cross-sections in (b), along AB, and in (c), along CD, are traced from the original paper.

35°N

60°E

10

70 80 90

0.3

0.25

0.2

flashes/km-2/day-1

0.15

0.1

0.05

0100

15

20

25

30

Figure 6. TRMM Lightning Imaging Sensor (LIS) data (flashes persquare kilometre per day) averaged for June–September 1998–2004.

that convective cloud systems upstream of the mountainsin this region retain a maritime character. Schumacherand Houze (2003a) showed that the stratiform compo-nent of tropical rain tends to be substantially greater overoceans than over land.

In addition, the convection in the eastern region occursprimarily in conjunction with synoptic-scale Bay of Ben-gal monsoon depressions (Ramage, 1971; Rao, 1976;Godbole, 1977; Sikka, 1977; Shukla, 1978; Sanders,1984; Johnson and Houze, 1987; Douglas, 1992a,b).These depressions occur in conjunction with the advanceof the active monsoon into the northernmost Bay of Ben-gal (Lawrence and Webster, 2002; Wang et al., 2005).The highly stratiform character of the clouds and precip-itation in one such depression was documented by aircraft



radar and microphysical measurements in the SummerMonsoon Experiment of 1979 (Houze and Churchill,1987).

4. Vertical dimensions of intense convective echoes

Altogether, 117 deep intense convective echoes (asdefined in Section 2.3.2) were observed in our TRMMPR dataset (see Wilton, 2005, Appendix B, Table B.1,for details on the times and locations).

4.1. Example of a deep intense convective elementover lower terrain

An example of a deep intense convective echo forming atthe boundary of the low-level moist flow from the Ara-bian Sea and large-scale leeside downslope dry flow fromthe Afghan Plateau (Section 3.1) occurred in northwest-ern India at ∼0900 UTC 14 June 2002. The flow resem-bled the climatological patterns in Figure 1 (Figures 7(a)and (b)). The low-level (10 m) winds show southwesterlyflow entering northwestern India/Pakistan from the Ara-bian Sea, coming ashore and crossing the inland desert ina strong pulse. The deep intense convective echo occurrednear the terminus of this moist low-level southwesterlyjet (star in Figure 7), and at the southern edge of themidlatitude westerly jet (Figure 7(b)). The 10 m windsindicate that northwesterly flow (on a dry downslopetrajectory from the Afghan Plateau) met the southwest-erly moist monsoon flow near the deep intense convec-tive echo. Since these maps are for the afternoon (1200UTC = 1730 local standard time, LST), it is unlikely thatthe downslope flow was due to the diurnal heating cycleover the plateau (Murakami, 1987).

Soundings at 0000 and 1200 UTC (9 h prior to and3 h after the TRMM overpass in which the deep intenseconvective echo was observed) were consistent withthe observed convective development (Figure 8). Thesoundings in Figure 8 display the two-layered structuredescribed in Section 3.1, as a low-level moist layeroriginating over the Arabian Sea underran dry warm airadvected downstream from the Plateau, in the mannerdescribed by Sawyer (1947) and Carlson et al. (1983).The sounding at Patiala at 0000 UTC (Figure 8(a))showed relatively large convective available potentialenergy (CAPE; Houze 1993, chapter 8) of 2700 J kg−1.

Figure 9 shows the TRMM PR reflectivity associatedwith the deep intense convective cell at 0900 UTC (1430LST) 14 June 2002. The vertical section shows thatthe cell was over 18 km in height, and that the 40dBZ echo reached 16 km. Also notable is the verticalalignment (erect structure) of the deep intense convectiveecho, as opposed to the more tilted structure oftenassociated with robust thunderstorms and squall lines(Houze, 1993, chapters 8 and 9). The erect cell structurein Figure 9(b) is consistent with the wind data showinglittle shear at the convective location well south ofthe core of the midlatitude westerly jet (Figure 7(b)).We have found this vertically upright structure of deep

Figure 7. NCMRWF reanalysis of (a) 10 m winds and (b) 200 mbwinds for 1200 UTC 14 June 2002. The star shows the location ofdeep intense convective echo. The locations of sounding stations are

indicated by P (Patiala), D (Delhi), and J (Jodhpur).

intense convective cells to be a ubiquitous aspect ofTRMM PR data in northern India, consistent with thefact that the convection over and just upstream ofthe Himalayan barrier occurs generally in a region ofweak shear, south of the upper-level westerly jet andnorth of the upper-level easterly jet (Figure 1(b)). Thetopography of the subtending terrain, shown in red atthe bottom of Figure 9(b), further confirms that thisconvective event was clearly not triggered directly overthe mountains.

4.2. A deep intense convective element over highterrain

While most monsoon convection occurs just upstream ofor over the lower slopes of the Himalayas, deep intenseconvective echoes do occur occasionally over the highterrain of the Tibetan Plateau (Figure 4(a)). Wide intenseconvective echoes were generally not observed over thePlateau (Figure 4(b)), indicating that the convection overthe Plateau tends to be unable to develop mesoscaleorganization, possibly because of insufficient moisture inthe environment (generally <5 g kg−1 of water vapour;Wang and Gaffen, 2001). We suggest that the semi-arid conditions over the southern portion of the Plateau(the part included in our study) are adequate to support



500(a) (b)

(c) (d)

(e) (f)

Patiala 00 UTC Patiala 12 UTC

Delhi 00 UTC

Jodhpur 00 UTC

Temperature (°C)

Jodhpur 12 UTC

Delhi 12 UTC

600

700

800

900

500

500

600

600

700

700

800

800

900Pre

ssur

e (m

b)

900

10 20 30 40 10 20 30 40

Figure 8. Rawinsonde data in skew T –log p format for Patiala, India on 14 June 2002 at (a) 0000 UTC and (b) 1200 UTC. (c)–(d) and(e)–(f) are as (a)–(b), but for Delhi and Jodhpur, respectively. Adapted from University of Wyoming Department of Atmospheric Science Upper

Air Data website (http://weather.uwyo.edu/upperair/).

32°N

74°E

31

(a)

(b) 2070

dBZ

62

54

46

38

30

22

14

6

Hei

ght (

km)

16

0 25 50 75 100

Distance (km)

12

8

4

0

30

29

2875 76 77

Figure 9. TRMM PR reflectivity for 0900 UTC 14 June 2002. (a) Horizontal pattern at 4 km. (b) Vertical cross-section running left to right alongthe red line in (a). The red line in the cross-section plot indicates the height of the subtending terrain. The colour scale indicates reflectivity in

dBZ. Note that dBZ values <17 are the result of interpolation of the gridded data. The lowest measured value is ∼17 dBZ.

the occurrence of occasional single deep intense isolatedconvective cells but inadequate to support a cloud systemof greater breadth. Convection over the Plateau tendsto be scattered cells, mostly isolated but sometimesaligned. An example of convective echoes over thePlateau (Figure 10(a)) shows that intense convectiveechoes (including deep ones, with 40 dBZ extendingabove 10 km) occur over the Plateau and look similarto the echoes over the lowlands and foothills except

that their lower portions are truncated (Figures 10(b) and(c)).

4.3. Statistics of heights of intense convective echoes

Figure 11(a) shows the cumulative frequency distributionof the heights of all the identified intense convectiveechoes for the 2002 and 2003 monsoon seasons. Thebulk of the intense convective echoes in all subregions



33°N

79°E

31

29

(a) (b)

(c)

27

25

81 83 85 87 89 0 50

0 200 400 600 800

67dBZ

59

51

43

35

27

11

3

19

16

12

8

4

0

16

12

8

4

0

100 150

Distance (km)

Hei

ght (

km)

Figure 10. TRMM PR reflectivity for 1227 UTC 14 July 2002. (a) Horizontal pattern at 6.5 km. (b) and (c) show vertical cross-sections runningleft to right along the red lines in (a). Other details are as in Figure 9.

are below the 0 °C level, which is generally locatedbetween 5.5 and 6.0 km over the study region. Since40 dBZ is not the highest reflectivity found in theintense convective echoes, this result is consistent withthe findings of Maheshwari and Mathur (1968) whofound that the maximum reflectivity in Indian monsoonconvective storms was generally at or below the 0 °Clevel. This figure also demonstrates that the easternsubregion has more frequent occurrence of 40 dBZ echotop below the 0 °C level. The occurrence of the echomaximum at low levels is a typical characteristic ofconvective precipitation in airmasses of maritime-tropicalorigin (Caracena et al., 1979; Szoke and Zipser, 1986;Szoke et al., 1986). The eastern subregion, being closerto the sea, evidently had convection resembling that overtropical oceans.

Although the intense convective echoes most fre-quently have tops below 5 or 6 km, a surprisingly largenumber of the 40 dBZ echo tops over northern India canreach extremely high levels. Seshadri (1963) and Ghosh(1967) studied echoes observed by ground-based radarat Delhi and found echoes occasionally reaching 17 km.To focus on the deeper echoes seen by the TRMM PR,Figure 11(b) shows only those 40 dBZ intense convectiveechoes with tops at or above 10 km. The occurrence ofthese high reflectivities in the upper-level ice region arelikely explained by the presence of larger ice particles,which may be too large to be in the Rayleigh scatteringregime for the 2 cm wavelength of the TRMM PR. Grau-pel particles of the order of 1 cm in diameter are evidentlylofted to high altitudes by extremely strong updraughtsin these cases. Rayleigh scattering requires the particlesize to be much less than the radar wavelength. It is fur-ther evident from Figure 11(b) that the highest frequencyof very deep 40 dBZ intense convective echoes occursin the western subregion, with some intense convectiveecho-top heights extending to over 16 km! This region is

Figure 11. (a) Normalized frequency distribution of 40+ dBZ intenseconvective echo. (b) Inset showing normalized frequency distributionof 40+ dBZ intense convective echo tops only for intense convective

echoes extending above 10 km amsl.

far from the ocean, and the convection is of a more con-tinental nature, with stronger updraughts. The convection



in this region is known for its intensity and high fre-quency of lightning, as noted by Barros et al. (2004)(Figure 6).

5. Horizontal dimensions of intense convectiveechoes

Altogether 121 of the intense convective echoes includedin this study qualified as wide intense convective echoes(as defined in Section 2.3.2). Generally these echoesare complexes of interconnected echoes, which can beregarded as MCSs.

5.1. Example of a wide but amorphous intenseconvective element

A wide intense convective echo occurred in the concaveindentation of the northwestern end of the Himalayas at∼1300 UTC 22 July 2002. Figure 12 shows the synopticconditions on that day. Similar to the climatologicalpatterns in Figure 1, a strong low-level southwesterlyflow of moist air from the Arabian Sea was crossingthe northwest coast of India and extended all the wayto the foothills of the Himalayas (Figure 12(a)). Thewide intense convective echo (star) occurred in theconvergent region north of the terminus of the 10 msouthwesterly flow, between the southwesterly flow anddry downslope flow from the north. Similar to the casedescribed in Section 4(a), this convection occurred in thelocal afternoon, and dryline-type behaviour is suggestedby the 10 m flow and its relation to the adjacent steeplysloped terrain. From the 200 mb winds (Figure 12(b)), itappears that the flow was again south of an upper-levelwesterly jet, and that the shear was not especially strong.

Soundings prior to the wide intense convective echoat Jodphur resembled those of a classic severe stormsituation over the Great Plains of the USA (Fawbushand Miller, 1954; Newton and Fankhauser, 1964; Palmenand Newton, 1969; Bluestein, 1993, pp. 448–451). Thesesoundings indicated a strong capping inversion atop themoist boundary layer (Figure 13). The strength of thecapping inversion lessened from 0000 to 1200 UTCand thus evolved towards a sounding where vigorousconvection was more likely to break out. These soundingsare consistent with upper-level warm dry flow from theAfghan Plateau overriding the low-level moist monsoonalflow and providing a cap that allows the low-levelwarm moist flow from the Arabian Sea to build upinstability as it moves farther inland until being releasedby convergence directly over or just upstream of theHimalayas.

At 1300 UTC (1830 LST), the TRMM PR detected awide intense convective echo over the plains of the north-western subcontinent, within the concave indentation ofthe Himalayan barrier (Figure 14(a)). The 40 dBZ echocovered 3475 km2 at an altitude of 3.5 km. It containedmultiple more intense cells in its interior. However, itshowed no elongation to form an echo band, nor was it

Figure 12. NCMRWF reanalysis of (a) 10 m winds and (b) 200 mbwinds for 1200 UTC 22 July 2002. The star shows the location ofwide intense convective echo. The location of the sounding station is

indicated by J (Jodhpur).

part of an organized line of similar echo structures. Inplan view, it was amorphous.

The vertical cross-section in Figure 14(b) shows aboutsix vertically oriented cells (best identified by theirprotruding tops in the 30–35 dBZ contour range). Thesecells were vertically erect and packed close togetherto form a contiguous 40+ dBZ echo over a horizontaldistance of ∼90 km. The vertically erect structure of thecells seen in this case was typical of all the embeddedcells of all the wide convective echoes that we haveexamined, consistent with the generally weak shear inthis region during the monsoon. The cells could havebeen still growing or beginning to collapse. In either casethey were likely headed toward formation of a stratiformregion later in the lifetime of the complex of cells, inthe manner normally expected as cells weaken in a MCS(Gupta et al., 1955; Houze, 1993, 1997, 2004).

Of the 121 wide intense convective echoes observedin this study, 20 (including the example above) werealso classified as deep convective echoes. Fourteen ofthose doubly intense cases (i.e. both deep and wide) werelocated in the western subregion; only six were locatedin the central subregion, while no echoes in the easternsubregion were simultaneously classified as both deepand wide.



500(a) (b)

600

700

800P

ress

ure

(mb)

900

10

Jodhpur 00 UTC Jodhpur 12 UTC

Temperature (°C)

Soundings for 22 July 2002

403020 10 403020

Figure 13. Rawinsonde data in skew T –log p format for Jodhpur, India, on 22 July 2002 at (a) 0000 UTC and (b) 1200 UTC, from the samesource as Figure 8.

36°N

35

(a)

(b)34

33

32

31

30

29

28

27

71°E 72 74 77

16

12

8

4

0

Hei

ght (

km)

0 50 100 150 200

Distance (km)

70dBZ

62

54

46

38

30

14

6

22

Figure 14. As Figure 9, but for 1309 UTC 14 June 2002.

5.2. Examples of wide intense convective echoes withline organization

On some occasions, the wide intense convective echoesshowed line organization. One recurring type is illustratedby Figure 15. A contiguous line of high-reflectivity echoroughly paralleled the steep Himalayan barrier. A cross-section along the convective line showed vertically erectcells (consistent with the weak shear) interconnected bystrong (>40 dBZ) echo (Figure 15(c)). Stratiform echo islocated along the northern red line in Figure 15(a). Thevertical cross-section in Figure 15(b) shows bright-bandecho surrounding a convective cell (at 85 km on the hor-izontal scale). The stratiform echo was likely formed ascells similar to the one at 85 km collapsed. (The historyof the echoes cannot actually be determined because theTRMM PR provides snapshot coverage only 2–3 timesper day over the region of study.) The stratiform precipi-tation likely formed as the vertically erect cells collapsed.Collapse of convective echo cores is a well-known pro-cess of generating areas of stratiform precipitation inMCSs in low-shear environments (Houze, 1993, 1997)and had been noted in India by Gupta et al. (1955).

The central subregion is the only zone of our SouthAsian study area in which we have noted propagat-ing arc-shaped leading-line/trailing-stratiform squall-line

MCS structure. (Squall lines may be more common in thepre-monsoon season, when the westerly jet is positionedfarther south, and the shear is greater; however, theseearly-season storms would be excluded from our studysince we are considering only the monsoon months.)These squall lines are not as well developed as those overthe central USA (Houze et al., 1989, 1990) or west Africa(Hamilton and Archbold, 1945; Fortune, 1980; Houzeand Betts, 1981; Hodges and Thorncroft, 1997; Schu-macher and Houze, 2006), where stronger shear prevails.An example in northern India is shown in Figure 16. Itexhibits an arc-shaped line roughly perpendicular to thesteep wall of the main Himalayan barrier. The bow of theline in Figure 16(a) was pointing generally to the south-east. Time-lapse looping of Meteosat-5 infrared satelliteimagery shows that this system was propagating south-eastwards, in the direction of the convex bow of the line.The PR data show that a stratiform echo was trailingbehind, giving the echo a classic squall-line structure(Figure 16(b)). This day was unusual in that a strongmidlevel jet lay over and parallel to the Himalayan bar-rier (Figure 17(b)). This jet lent the convection a shearedenvironment and apparently played a role in organizinga rear-inflow jet (Smull and Houze, 1987), in the waythat the midlevel African Easterly Jet helps to organizesquall lines in West Africa. The midlevel jet seen in



35°N

71°E

34

33

(a)

(b)

32

3172 73 74 75 76

0 50

0 50 100 150

70dBZ

62

54

46

38

30

14

6

22

16

20

12

8

4

0

(c)

16

20

12

8

4

0100 150 200

Distance (km)

Hei

ght (

km)

Figure 15. TRMM PR reflectivity for 2208 UTC 3 September 2003. (a) Horizontal pattern at 4 km. (b) and (c) contain vertical cross-sectionsrunning left to right along northern and southern red lines in (a), respectively. Other details are as in Figure 9.

29°N

85°E

28

27

26

25

86 87 88 89

(b)

(a)

16

12

8

4

0

Hei

ght (

km)

0 50 100 150

Distance (km)

67dBZ

5951433527

113

19

Figure 16. As Figure 9, but for 2017 UTC 5 June 2003.

Figure 17(b) persisted for three days, and squall linesoccurred in the pre-Himalayan channel on each day. Inour data sample, we found one additional instance of asimilar squall line occurring along the Himalayas in ananalogous synoptic regime, with a midlevel jet overlyingthe Himalayan escarpment. Apparently the convectionfavoured in the region just upstream of and over thefoothills of the Himalayas can become organized intoa propagating squall line in conjunction with these rela-tively rare synoptic events exhibiting strong shear.

5.3. Statistics of the horizontal dimensions of intenseconvective echoes

Figure 18(a) shows the cumulative distribution functionof the horizontal area of 40+ dBZ intense convectiveechoes. This sample was extracted from the set ofall intense convective echoes observed by the TRMMPR over the western, central, and eastern subregionsduring the 2002 and 2003 summer monsoon seasons.The average height at which the areal maximum waslocated was approximately 2.5 km. The highest altitudeat which the areal maximum occurred was 4.5 km. These

heights are below the melting level and consistent withthe low-level echo centroid heights found by Szoke et al.(1986) and Szoke and Zipser (1986) in tropical maritimeconvection and with the results of the analysis of ground-based radar in India by Maheshwari and Mathur (1968).The wide areas of convection were all classified asconvective precipitation by the 2A23 algorithm, thuseliminating the possibility that the horizontal scale ofthe echoes could have been determined by radar bright-band echoes. The portion of Figure 18(a) includingechoes >1000 km2 is replotted in Figure 18(b), whichshows that widest areas of convection exhibited a higherfrequency of occurrence in the western subregion, witha tendency to decrease eastwards. This same regionalbehaviour was noted for intense convective echo height(cf. Figure 11(b)).

6. Characteristics of stratiform echo regions in themonsoon

We have intensively investigated the 10 largest broadstratiform regions (as defined in Section 2.3.3) in our



Figure 17. NCMRWF reanalysis of (a) 10 m winds and (b) 500 mbwinds for 1200 UTC 5 June 2003.

dataset. Of these, six were associated with well-definedmonsoon depressions. The remaining four were associ-ated with what we would call quasi-depressions, whichhad all the characteristics of a Bay of Bengal depres-sion except that the vortex in the low-level wind was notcompletely closed.

6.1. Example of a broad stratiform region in a Bay ofBengal depression

Figure 19 is a composite showing how the Bay of Bengaldepression is connected with the monsoon intraseasonaloscillation (Webster, 2006). The four panels show theprogression over a 13-day period. Day 0 is defined asthe occurrence of 10 mm day−1 of rain at 90 °E. Thecyclonic gyre over the head of the Bay of Bengal on day8 (Figure 19(d)) has propagated northeastwards into theBay of Bengal as part of an equatorial Rossby wave. Thearrival of the gyre in the northernmost portion of the Baymarks the onset of an active monsoon period.

On 11 August 2002, the PR detected a broad stratiformregion in connection with such a Bay of Bengal depres-sion. The large-scale flow (Figure 20) broadly resemblesthe climatological patterns in Figure 1. In the low-levelwinds (Figure 20(a)), the Bay of Bengal depression wascentred at the head of the Bay (about 22 °N, 90 °E). Thesounding for Tengchong (Figure 21, station location inFigure 20(a)) shows saturated moist adiabatic conditions

1.0(a)

(b)

.95

.90

.85

.80

.75

.70

.65

.60

.55

.50

1.00

.999

.998

.997

.996F

requ

ency

.995

.994

.993

.992

.991

.99050002000 3000 4000

Area of 40 dBZ core (km2)

1000

250020001000 15005000

Western sub-region

Central sub-region

Eastern sub-region

Western sub-region

Central sub-region

Eastern sub-region

Figure 18. Cumulative distribution of 40 dBZ convective areal maxima(a) between 25 and 2500 km2 and (b) between 1000 and 5000 km2.

through the troposphere. This sounding typifies the upper-air conditions in the depressions we have examined.

Figure 22(a) shows the radar echo sampled during aTRMM PR overpass of the Bay of Bengal depression.The PR echo (250 km × ∼500 km) represents only asmall portion of the overall cloudy region associated withthe depression. The areas of enhanced upper-level cloudand their associated radar echoes in the depressions nevertake on the synoptic-scale dimension of the depression.They appear to be MCSs embedded within and proba-bly abetted by the parent Bay of Bengal depression. Thewidespread upward motion of the synoptic system evi-dently destabilizes the air and provides a moist environ-ment favourable for MCSs. The distribution of MCS-likestructures within the depression was reminiscent of thehighly stratiform MCSs analyzed by Houze and Churchill(1987). The flow over the terrain likely enhances thesemesoscale precipitation areas. Figure 22(b) exemplifieshow the mesoscale precipitation features on the northside of the depression often tend to nestle into the easternconcave indentation of the Himalayan range, where largestratiform echoes occur most frequently (Figure 4(c)).The stratiform precipitation echo appears to have beencomposed of weakened convective cells (Figure 22(c)).This structure was typical of the mesoscale precipita-tion areas in all the depressions we have examined in



30°N

0

30°S

5 m s-1

8 10 12 14 16 18

mm day-1

Composite 925 mb wind vectors and precipitation (>8mm day–1)

(a) (b)

30°N

0

30°S30°E 60 90 120 60 90 12030°E

(d)(c)

Figure 19. Evolution of the rainfall and 925 mb wind field of the monsoon intraseasonal oscillation. Rainfall is based on Microwave SoundingUnit data. Winds are based on NCEP reanalysis. Composite of 39 events from 1985 to 1995. (a)–(d) show composite distributions relative today 0, defined as the occurrence of >8 mm day−1 rainfall at 80–90 °E/5 °N–5 °S persisting for three days. The numbers in the upper left ofeach panel indicate the number of days before (negative) and after day 0. The annual cycle has been removed. Adapted from Webster and Tomas

(1997).

the PR data, whether the precipitation features were overthe sea, the lowlands, or the foothills of the Himalayas.Occasionally they contained active, deep convective echocells, which apparently later collapsed to form the broaderbackground of stratiform precipitation (Houze, 1997), andactive deep cells only occupy a small fraction of the rainareas at any given time. (The existence of ‘generatingcells’ associated with a layer of convective overturningaloft, as is frequently seen in midlatitude cyclones, cannotbe ruled out as an alternative explanation for the observedecho structure. In the tropical oceanic environment, wherethe convection in the depression takes the form of dis-crete MCSs, it seems more likely that the echo pattern isa manifestation of collapsed deep penetrative convectivecells.)

Figure 19 shows how the cyclonic gyre within theoverall low-level wind pattern of the Rossby wave struc-ture that propagates into the Bay of Bengal typicallyinduces strong southwesterly winds across the Myan-mar coast. Zuidema (2003) showed that clouds tend toconcentrate upstream of the coast, as they do upstreamof the western Ghats and other mountain ranges of theregion (Grossman and Durran, 1984; Xie et al., 2006).

A large number of the mesoscale patches of enhancedhigh cloudiness in Figure 22(a) are south of the depres-sion centre within the low-level southwesterly regimeupstream of the Myanmar coast. The strong upper-leveleasterlies prevailing at this comparatively low latitude areproducing streamers of westward anvil blow-off from theprecipitating cloud lines off Myanmar. A PR observationof some of these precipitation echoes over the northernBay of Bengal is shown in Figures 23(a) and (b). A ver-tical cross-section shows convective echo cells of rathermoderate depth, some of which have collapsed into strat-iform echo (Figure 23(c)). Zuidema (2003) also foundthe clouds upstream of the Myanmar coast to be of smallto moderate depth and width in comparison to the over-all population of clouds seen in infrared imagery of theregion.

6.2. Statistics of stratiform area sizes

Figure 4(c) showed that broad stratiform regions, i.e.the largest stratiform regions, preferentially occur inthe eastern subregion. The cumulative distribution inFigure 24 further shows that stratiform areas of all sizes



Figure 20. NCMRWF reanalysis of (a) 10 m winds and (b) 200 mbwinds for 11 August 2002. The location of the sounding station is

indicated by T (Tengchong, China).

200

300

400

500

600

700800900

Pre

ssur

e (m

b)

-40 -30 -20 -10 0 10 20 30 40

Temperature (°C)

Figure 21. Rawinsonde data in skew T –log p format for Tengchong,China, on 11 August 2002 at 0000 UTC, from the same source as

Figure 8.

tend to be more frequent in the eastern subregion,where synoptic forcing by Bay of Bengal depressionsand proximity to the ocean should both favour thedevelopment of MCSs with larger stratiform precipitationareas.

7. Probability distributions of Himalayan regionradar echoes

Figure 25 is a Contoured Frequency by Altitude Dia-gram (CFAD; Yuter and Houze, 1995) containing the

non-interpolated TRMM PR 2A25 radar reflectivity val-ues for the entirety of the 2002 and 2003 monsoon sea-sons and inclusive of all three subregions (west, central,and east). The contours represent the frequency of occur-rence of a given reflectivity at a given height, f (dBZ,z).The entire distribution was normalized by dividing bythe maximum absolute frequency of the samples withinthe region of analysis, fmax(dBZ,z). Thus, the contoursare labelled in values of f/fmax. This method permits thecomparison of CFADs between regions despite the dif-ferent absolute frequencies. The contours are truncated atlower reflectivity values as a result of the reduced sensi-tivity to low rain rates of the TRMM PR.

Figure 25 is characterized by three statistically distinctregimes, viz. those located above, at, and below theclimatological 0 °C level (∼5.5–6). Below the 0 °C level,the echoes are produced by raindrops and have a medianreflectivity of ∼25 dBZ (corresponding to a rain rate ofabout 1–2 mm h−1). The wide spread of the distributionbelow the 0 °C level is consistent with rain falling fromconvection in various stages of development. Above the0 °C level, in the ice region, the median reflectivity valueis ∼20 dBZ. At (or actually just below) the 0 °C level, thetendency for a melting-layer bright band to occur in theechoes is evident in the statistics in the form of a distinctrightward shift in the CFAD contours toward higher dBZvalues at ∼5 km. Subsequent CFADs add detail to theregional trends evident in Figure 4 by breaking down theCFAD statistics by subregion, rainfall type, and terrainheight.

Figures 26(a)–(c) are generally consistent with moreintense convection in the west. They further show thatthe western subregion had higher reflectivity at upperlevels, indicating a tendency for the deepest convectionto occur in the western subregion and for the maximumecho heights to be less in the eastern subregion. Theeastern subregion CFAD exhibited a more pronouncedrightward shift in reflectivity in the melting layer thanin the central and western subregions, consistent with agreater tendency for stratiform precipitation in the easternsubregion.

When the CFADs are further subdivided accordingto precipitation type (Figure 27), the regional differ-ences become even more evident. The convective echoes(Figure 27(a) and (b)) showed a tendency to have bothhigher overall reflectivity values and higher reflectivitiesat upper levels in the western region. The CFAD con-tours indicate the occurrence of up to 40 dBZ echo at12 km in the west, while 40 dBZ echoes are statisticallyevident only to about 9 km in the eastern subregion. Thestratiform partitions (Figures 27(c) and (d)) show a morepronounced bright-band signature in the more maritimeregime of the eastern subregion.

When the CFADs are stratified according to subtend-ing terrain, they show that convection over the lowlandstended to be slightly deeper than that over the foothillsin all three subregions (Figure 28(a)–(f)). Evidently, thenet effect of the Himalayan range in triggering convectionwas most pronounced over the lowlands just upstream of



350340330320310300290280270260250240230220210200190

ir

30°N

25

30°N

2520

15

1080°E 85 90 95 100 105 90°E 95

(a) (b)

12

8

4

16

0100 200 300 4000 500

Distance (km)

Hei

ght (

km)

(c)dBZ

67595143352719113

Figure 22. TRMM PR reflectivity for 0252 UTC 11 August 2002. (a) Horizontal swath of PR reflectivity (green and yellow shades) at 4 kmis superimposed on the Meteosat-5 infrared temperature field. Grey shades show the infrared temperature in K, with red and yellow shadeshighlighting the coldest cloud tops. (b) Zoomed-in view of the PR reflectivity in (a). (c) Vertical cross-section running left to right along the

black line in (a) and (b). Meteosat-5 data were obtained from http://www.eumetsat.int. Other details are as in Figure 9.

40°N

35

30

25

20

15

20°N

15

10

70°E 90°E75 80 85 90 95 95100 105

350340330320310300290280270260250240230220210200190

(b)(a)

12

8

4

16

075 150 2250 300

Distance (km)

Hei

ght (

km)

(c)dBZ

67595143352719173

Figure 23. TRMM PR reflectivity for 0455 UTC 11 August 2002. (a) Horizontal swath of PR reflectivity (green and yellow shades) at 2 km issuperimposed on Meteosat-5 infrared temperature field. Other details are as in Figure 22.



Figure 24. Cumulative distribution of area of stratiform precipitation.

14 0.91.0

0.80.70.60.50.40.30.20.10.0

12

10

8

6

4

2

015 20 25 30 4035 45

Alti

tude

(km

)

Reflectivity (dBZ)

CFAD for all regions

Figure 25. Contoured frequency by altitude diagram (CFAD) of TRMMPR reflectivity data for June–September 2002–2003 for western,central, and eastern subregions combined. Contours represent thefrequency of occurrence relative to the maximum absolute frequency.Altitudes are geopotential height (km amsl) relative to the surface ofthe earth ellipsoid. The ordinate of the CFAD is altitude (in 250 mincrements, or bins) and the abscissa is reflectivity in dBZ (in 1 dBZ

bins).

the mountains. This result suggests that the formation ofthe most intense convection was focused at lower ele-vations immediately upstream of the mountains, i.e. nearthe leading edge of the warm, dry air advected downslopefrom the Tibetan Plateau, in a manner possibly similarto dryline convective triggering observed in the centralUSA (Section 3.1). Convection located directly over thefoothills, though still strong, already has (on average) lostsome buoyancy – conceivably because the near-surfacemoisture has been removed from the airstream flowingup the terrain. The availability of low-level moisture andability to form deep convection is apparently nearly com-pletely absent over the higher terrain. The echoes seen inthe mountain CFADs (Figure 28(g)–(i)) were composedpredominantly of stratiform echo, probably advectedfrom the tops of the lowland and foothill convection.

8. Conclusions

The ability of the TRMM PR to provide vivid anddetailed images of the vertical structure of convection

14

12

10

8

6

4

2

0

14

12

10

8

6

4

2

0

14

12

10

8

6

4

2

0

Alti

tude

(km

)

15 20 25 30 35 40 45 50

Reflectivity (dBZ)

(a)

(b)

(c)

Western sub-region

Central sub-region

Eastern sub-region

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

Figure 26. Contoured frequency by altitude diagram (CFAD) of TRMMPR reflectivity data for June–September 2002–2003 for (a) western,(b) central, and (c) eastern subregions separately. Contours and other

details are as in Figure 25.

in remote regions of complex terrain has given us freshinsight into two longstanding questions:

(1) the behaviour of highly convective clouds in a moistflow impinging on a major mountain barrier, and

(2) the particular structure and organization of summermonsoon convection over the subcontinent of SouthAsia.

The first question has general applicability to under-standing the role of orography in affecting precipitation,while the second is vital to understanding and predictingmonsoon clouds and precipitation. The schematic draw-ing in Figure 29 suggests some of the conclusions wehave drawn from an extensive examination of the verticalstructure of Himalayan region convection as seen in thethree-dimensional TRMM PR data. The schematic doesnot depict every observed scenario but rather illustratesscenes typical of each of the subregions.



1614121086420

1614121086420

Alti

tude

(km

)

Western sub-region (convective) Eastern sub-region (convective)

Western sub-region (stratiform) Eastern sub-region (stratiform)

(a)

(c)

(b)

(d)

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

15 20 25 30 35 40 45 5550 15 20 25 30 35 40 45 5550

Reflectivity (dBZ)

Figure 27. Contoured frequency by altitude diagram (CFAD) of TRMM PR reflectivity data for June–September 2002–2003 for (a) westernsubregion convective echoes, (b) eastern subregion convective echoes, (c) western subregion stratiform echoes, and (d) eastern subregion

stratiform echoes. Contours and other details are as in Figure 25.

1614121086420

1614121086420

1614121086420

Alti

tude

(km

)

Western lowlands Central lowlands

Reflectivity (dBZ)

Eastern lowlands

Western foothills Central foothills Eastern foothills

Western mountains Central mountains Eastern mountains

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

15 20 25 30 35 40 45 50 15 20 25 30 35 40 45 50 15 20 25 30 35 40 45 50

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

Figure 28. Contoured frequency by altitude diagram (CFAD) of TRMM PR reflectivity data for June–September 2002–2003 for categories ofsubregion and height of subtending terrain. Contours and other details are as in Figure 25. Terrain height categories are defined in Figure 4.

Our analysis of PR echoes identified by the TRMM2A23 algorithm as being convective in nature showsthat convective echoes >40 dBZ in intensity in the

Himalayan region most commonly have tops belowthe 0 °C level, consistent with early studies of radarreflectivity structure in the region (Maheshwari and



Mathur, 1968). Also consistent with early investigations(Seshadri, 1963; Ghosh, 1967), the TRMM PR dataoccasionally show much more extreme echo heights, andwe term echoes of 40 dBZ or more extending above10 km ‘deep intense convective echoes’. The Himalayanregion of this study lies between the upper-level westerlyjet to the north and the easterly jet to the south. Inthis setting of weak winds aloft and weak shear, thedeep intense convective echoes usually exhibit no tilt.These vertically erect 40+ dBZ intense convective echoessometimes extend above 17 km. Such high reflectivityvalues extending to such extreme altitudes suggestsgraupel particles of substantial size being lofted to greatheights by intense convective updraughts.

20

15

10

5

20

15

10

5

20

15

10

5

Hei

ght (

km)

(a)

(b)

(c)

Figure 29. Schematic illustrations of typical extreme precipitationevents in each subregion of the study: (a) A deep intense convectiveecho in the western subregion, (b) a wide intense convective echo(mesoscale convective system) over the lower terrain of the centralsubregion and deep intense convective echoes over the plateau, and(c) a broad stratiform precipitation area over the lower terrain of theeastern subregion and a deep intense convective echo over the plateau.

The TRMM PR data further show the spatial distribu-tion of extreme convective events. Deep intense convec-tive echoes (40 dBZ echo tops >10 km) occur especiallyfrequently in the western subregion, primarily within theconcave indentation in the western end of the Himalayanrange. The frequent occurrence of deep intense convec-tive echoes lofting graupel to high altitude in the con-cave indentation of the Himalayan barrier in the westernsubregion is consistent with the frequent occurrence oflightning in this region (Barros et al., 2004).

The deepest intense convective echoes tend to belocated just upwind of the Himalayas or over thefoothills (Figures 4(a), 28). These extremely deep con-vective events appear to occur where the advance edgeof the low-level moist southwesterly monsoon flow fromthe Arabian Sea meets dry air descending from the adja-cent Afghan or Tibetan Plateau (Figure 29(a)). The loca-tion of the extremely deep convective outbreaks suggestsa possible similarity to dryline convection. The moistsouthwesterly low-level flow coming from the IndianOcean is capped by an associated inversion as it under-runs dry air advected from the Afghan or Tibetan Plateau.Sawyer (1947) and Carlson et al. (1983) suggested thatintense convection may occasionally penetrate this ‘cap-ping inversion’ or (more frequently) break out near theleading edge of the warm, dry air advected in fromthe elevated plateau, where the lower moist layer isnot capped. Our results are consistent with these ear-lier studies. This scenario is particularly applicable overthe northwestern subcontinent, where the moist low-levelflow from the Arabian Sea must traverse the hot aridplain of the northwestern subcontinent before reachingthe vicinity of the Himalayan barrier. The similarity tosome aspects of severe convective formation over the USGreat Plains in the region east of the Rocky Mountainssuggests a behaviour that is general and repeatable fromone region to another.

The prevalent occurrence of the deepest and widestintense convective echoes just upstream of and over thelower elevations of the Himalayas – not over the higherterrain – has been seen in other precipitation regimes,e.g. the western Ghats (Grossman and Durran, 1984)and the European Alps (Houze et al., 2001). However,in the Himalayan region considered here, we also findthat the most intense of the convection occurs upstreamof and not over the higher terrain. The processes leadingto this behaviour in the Himalayan region are likelyat least partially different from the processes governingthe precipitation in other regions. Volumetric statisticalsummaries (CFADs) compiled from all the TRMM PRoverpasses of the study area show that the convectiveechoes tend to be slightly deeper and more intense overthe lowlands than over the foothills of the Himalayas,indicating that the low-level conditions favouring deepconvection are maximized just upstream (i.e. south) ofthe mountains. Apparently, the full complement of low-level moisture and heat content present over the plains isnot present over the foothills and higher terrain, so thatthe greatest orographic effect on convective precipitation



is felt just upstream of the mountain barrier. The dryair off the Afghan or Tibetan Plateau overriding theleading portion of the low-level monsoon flow caps theconvection in the region possessing this stratification,and a triggering mechanism is favoured immediatelyupstream of or over the lower foothills of the mountainrange. The triggering might involve upstream lifting(Grossman and Durran, 1984). However, diurnal heatingand cooling over the mountain barrier likely play a rolein the triggering near the barrier. The exact triggeringmechanism is a question to be resolved in future studies.

The intense convective echoes sometimes organizeupscale to form wide convective elements in whichseveral cells with vertically erect intense convectiveechoes become interconnected by weaker echo and cloud(Figure 29(b)). In extreme cases the interconnecting echois of high intensity (>40 dBZ) and the embedded cellseven more intense and very deep. We have called theseextreme cases ‘intense convective echoes’, defined asareas of echo exceeding 40 dBZ over a horizontal area>1000 km2. Such wide intense echoes are elements ofmesoscale convective systems (MCSs; Houze, 2004). Inseveral respects, the geographical pattern of occurrenceof the wide intense convective echoes detected by thePR resembles that of the deep intense convective echoes:viz. the preferred location was in the western subregion,and their occurrences were most frequent in the concaveindentation of the Himalayan barrier in that region.However, the pattern of occurrence of wide intenseconvective echoes differs in two respects from that of thedeep intense convective echoes as defined in this study:

(1) The wide intense convective echoes rarely, if ever,occur over the Tibetan Plateau, whereas isolated deepintense convective echoes do occasionally appearover the Plateau (Figures 29(b) and (c)). Wide MCSscontain large stratiform regions, which are favouredin more oceanic conditions (Schumacher and Houze,2003a). We suggest that the arid conditions over thePlateau (Wang and Gaffen, 2001) make it difficult forconvection to grow easily upscale to form MCSs.

(2) The wide intense convective echoes have a tendencyto appear in the central subregion, along the convexmid-portion of the steep Himalayan barrier. Theymost frequently occur along the foothills and justupstream of the steep terrain (Figure 29(b)).

The wide intense convective echoes in our datasetexhibit three repeatable types of structure:

(1) Multiple intense, vertically erect cells amassed withina horizontally amorphous mesoscale echo;

(2) Several intense, vertically erect cells grouped in aline parallel to the Himalayan barrier, either justupstream of the barrier or over the foothills. As thecells along these barrier-parallel lines weaken, theyform stratiform regions within the wide contiguousprecipitation feature; and

(3) Arc-shaped lines quasi-perpendicular to the Him-alayan barrier and propagating towards the southeastwith a trailing region of older stratiform echo. Theseoccur only rarely such as when anomalously strongmidlevel flow and associated shear becomes alignedwith the Himalayan escarpment.

Broad stratiform echoes occur primarily in the east-ern and central subregions (Figure 29(c)). They occurprimarily upstream of and over the lower mountainousterrain in association with Bay of Bengal depressions.Their centroids are most highly clustered within the con-cave indentation at the eastern end of the Himalayanbarrier. This suggests that the tendency of the depressionto promote convection is enhanced by moist flow overthe topography of the Himalayas. Vertical cross-sectionsin MountainZebra show that the broad stratiform echoesin the depression have widespread well-defined brightbands. The appearance of embedded cellular echoes inthe detailed cross-sections further suggests that broadstratiform echoes are, at least in part, the product ofthe weakening of earlier convective cells, in the mannerdescribed by Houze (1997). This latter supposition is dif-ficult to verify in this study since the TRMM PR providesonly snapshots. As pointed out by Houze (1993), strati-form precipitation may form in response to convection,if either the convection is tilted by shear or if convectivecells repeatedly die out and new cells form in the samemesoscale vicinity. In the latter case, the older, weak-ened cells coalesce to form a stratiform region in thevicinity of newer, active cells. Since the monsoon con-vection appears to have been predominantly non-tiltedand occurring in a generally weakly sheared environ-ment, it is likely that the stratiform regions were primarilyconglomerations of older, weakened cells. Cross-sectionssuch as Figures 22(c) and 23(c) are typical and qualita-tively suggest that the bright band is the net result ofdying cells.

CFADs show the bright-band signature to be muchsharper and better defined in the eastern subregion than inthe western subregion. This observation, together with themore frequent occurrence of large stratiform regions inthe eastern subregion, suggests that the deep convectionin the eastern subregion was of a more maritime char-acter. By contrast, convection in the western subregionwas of a distinctly continental nature. The large strati-form regions associated with Bay of Bengal depressionsobserved over land by the TRMM PR appear to be simi-lar to those previously observed over the Bay (Houze andChurchill, 1987). Apparently these depressions continueto provide a moist maritime monsoonal environment sup-porting the convection as their parent circulations moveover land. In the eastern subregion, the monsoonal flowassociated with depressions traverses only rather moistland surfaces before encountering high terrain, in contrastto the low-level southwesterly flow entering northwesternIndia from the Arabian Sea which must first traverse theextremely hot and extensive desert region of the north-western subcontinent.



The variations in cloud-system structure from west toeast in the Himalayan region described in this studypresent a useful target for high-resolution numericalmodels. Model output will provide deeper insight into theorographic, convective, and microphysical precipitationprocesses of the region than can be determined fromremote sensing from space or from the limited surfacedata in this rugged region. Ultimately, high-resolutionmodels will be tasked with precise operational forecastingof severe monsoon precipitation events.

Acknowledgments

The staff of NCMRWF under the direction of Dr A. K.Bohra generously provided model output for this study.Stacy Brodzik assisted with the data processing. BethTully executed the graphics. Harvey Greenberg assistedwith the production of Figure 2. This research was sup-ported by National Aeronautics and Space Administra-tion grants NAG5-13654, NAG5-9668-SO09 and NAG5-11685, and by National Science Foundation Grants ATM-0221843 and ATM-0505739.

Appendix

Algorithm for Remapping the Trmm Pr Data to aCartesian Grid

The Cressman method, as we have applied it here,employs a first guess, for which we use the nearestneighbour. The correction for the reflectivity Z at a gridpoint i is

Czi = − (WEzi) .

The subscript i indicates that this correction is basedon the reflectivity values at each data point i within the0.75 km layer centred on the grid point and within thehorizontal radius of influence N = 4.25 km centred onthe grid point. The variable W is the Cressman weightingfactor, given by

W = N2 − d2i

N2 + d2i

,

and di is the distance of point i from the grid point. Ezi isthe difference between the reflectivity value of the nearestneighbour, Znn, and the reflectivity, Zi , at point i. Theinterpolated value at the regular grid point is

Zgrid = Znn −∑

i

Czi .

References

Awaka J, Iguchi T, Kumagai H, Okamoto K. 1997. Rain typeclassification algorithm for TRMM precipitation radar. In Proc. of1997 Internat. Geoscience and Remote Sensing Symposium, 3–8August 1997, Singapore. IEEE Trans. Geosci. Remote Sensing 4:1633–1635.

Barros AP, Kim G, Williams E, Nesbitt SW. 2004. Probing orographiccontrols in the Himalayas during the monsoon using satelliteimagery. Nat. Hazards Earth Syst. Sci. 4: 29–51.

Bluestein HB. 1993. Synoptic-dynamic meteorology in midlatitudes.Volume II: Observations and theory of weather systems. OxfordUniversity Press: New York.

Caracena F, Maddox RA, Hoxit LR, Chappell CF. 1979. Mesoanalysisof the Big Thompson storm. Mon. Weather Rev. 107: 1–17.

Carlson TN, Benjamin SG, Forbes GS. 1983. Elevated mixed layers inthe regional severe storm environment: conceptual model and casestudies. Mon. Weather Rev. 111: 1453–1473.

Christian HJ, Blakeslee RJ, Goodman SJ, Mach DA, Stewart MF,Buechler DE, Koshak WJ, Hall JM, Boeck WL, Driscoll KT,Bocippio DJ. 1999. The Lightning Imaging Sensor. In Proceedingsof the 11th International Conference on Atmospheric Electricity,Guntersville, Alabama. 746–749.

Churchill DD, Houze RA Jr. 1984. Development and structure ofwinter monsoon cloud clusters on 10 December 1978. J. Atmos. Sci.41: 933–960.

Corbet J, Mueller C, Burghart C, Gould K, Granger G. 1994. Zeb:Software for geophysical data integration, display, and managementof diverse environmental datasets. Bull. Am. Meteorol. Soc. 75:783–792.

Cressman GP. 1959. An operational objective analysis system. Mon.Weather Rev. 87: 367–374.

Das PK. 1995. The Monsoons. Indian National Book Trust: New Delhi.Douglas MW. 1992a. Structure and dynamics of two monsoon

depressions. Part I: Observed structure. Mon. Weather Rev. 120:1524–1547.

Douglas MW. 1992b. Structure and dynamics of two monsoondepressions. Part II: Vorticity and heat budgets. Mon. Weather Rev.120: 1548–1564.

Fawbush EJ, Miller RC. 1954. The types of airmasses in which NorthAmerican tornadoes form. Bull. Am. Meteorol. Soc. 35: 154–165.

Fortune M. 1980. Properties of African squall lines inferred from time-lapse satellite imagery. Mon. Weather Rev. 108: 153–168.

Frei C, Schar C. 1998. A precipitation climatology of the Alpsfrom high-resolution raingauge observations. Int. J. Climatol. 18:873–900.

Ghosh BB. 1967. A radar study on thunderstorms and convectiveclouds around New Delhi during the southwest monsoon season.Indian J. Meteorol. Geophys. 18: 391–396.

Godbole RV. 1977. The composite structure of the monsoondepression. Tellus 29: 25–40.

Goswami BN, Rajagopal EN. 2003. Indian Ocean Surface winds fromNCMRWF analysis as compared to QuickSCAT and moored buoywinds. Indian Acad. Sci. (Earth Planet. Sci.) 112: 61–77.

Grossman RL, Durran DR. 1984. Interaction of low-level flow with theWestern Ghats Mountains and offshore convection in the summermonsoon. Mon. Weather Rev. 112: 652–672.

Gupta BK, Mani A, Venkiteshwaran SP. 1955. Radar observations ofrain at Poona. Indian J. Meteorol. Geophys. 6: 31–40.

Hamilton R, Archbold JW. 1945. Meteorology of Nigeria and adjacentterritory. Q. J. R. Meteorol. Soc. 3: 231–265.

Hill SA. 1881. The meteorology of the North-West Himalaya. Memoirsof the India Meteorological Department, 1: 377–426.

Hodges KI, Thorncroft CD. 1997. Distribution and statistics of Africanmesoscale convective weather systems based on the ISCCP Meteosatimagery. Mon. Weather Rev. 125: 2821–2837.

Houze RA Jr. 1993. Cloud Dynamics. Academic Press: San Diego.Houze RA Jr. 1997. Stratiform precipitation in regions of convection:

a meteorological paradox? Bull. Am. Meteorol. Soc. 78: 2179–2196.Houze RA Jr. 2004. Mesoscale convective systems. Rev. Geophys. 42:

DOI: 10.1029/2004RG000150.Houze RA Jr, Betts AK. 1981. Convection in GATE. Rev. Geophys.

Space Phys. 19: 541–576.Houze RA Jr, Churchill DD. 1987. Mesoscale organization and cloud

microphysics in a Bay of Bengal depression. J. Atmos. Sci. 44:1845–1867.

Houze RA Jr, Rutledge SA, Biggerstaff MI, Smull BF. 1989.Interpretation of Doppler weather radar displays in midlatitudemesoscale convective systems. Bull. Am. Meteorol. Soc. 70:608–619.

Houze RA Jr, Smull BF, Dodge P. 1990. Mesoscale organizationof springtime rainstorms in Oklahoma. Mon. Weather Rev. 118:613–654.

Houze RA Jr, James CN, Medina S. 2001. Radar observations ofprecipitation and airflow on the Mediterranean side of the Alps:Autumn 1998 and 1999. Q. J. R. Meteorol. Soc. 127: 2537–2558.



Houze RA Jr, Brodzik S, Schumacher C, Yuter SE, Williams CR.2004. Uncertainties in oceanic radar rain maps at Kwajaleinand implications for satellite validation. J. Appl. Meteorol. 43:1114–1132.

Iguchi T, Kozu T, Meneghini R, Awaka J, Okamoto K. 2000a. Rain-profiling algorithm for the TRMM Precipitation Radar. J. Appl.Meteorol. 39: 2038–2052.

Iguchi T, Meneghini R, Awaka J, Kozu T, Okamoto K. 2000b. Rain-profiling algorithm for TRMM Precipitation Radar data. Adv. SpaceRes. 25: 973–976.

James CN, Brodzik SR, Edmon H, Houze RA Jr, Yuter SE. 2000.Radar data processing and visualization over complex terrain.Weather Forecasting 15: 327–338.

Johnson RH, Houze RA Jr. 1987. Precipitating cloud systems of theAsian monsoon. In Monsoon Meteorology. Oxford Monographs onGeology and Geophysics No. 7. C-P Chang and TN Krishnamurti(eds.) Oxford University Press: New York; 298–353.

Joss J, Waldvogel A. 1990. Precipitation measurements and hydrology.Pp. 577–606 in Radar in Meteorology: Battan Memorial and 40thAnniversary Radar Meteorology Conference. D Atlas (ed.) AmericanMeteorol. Society: Boston.

Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L,Iredell M, Saha S, White G, Woollen J, Zhu Y, Leetmaa A,Reynolds B, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J,Mo KC, Ropelewski C, Wang J, Jenne R, Joseph D. 1996. TheNCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc.77: 437–471.

Krishnamurti TN. 1985. Summer Monsoon Experiment – a review.Mon. Weather Rev. 113: 1590–1626.

Kummerow C, Barnes W, Kozu T, Shiue J, Simpson J. 1998. TheTropical Rainfall Measuring Mission (TRMM) sensor package. J.Atmos. Oceanic Technol. 15: 809–817.

Kummerow C, Simpson J, Thiele O, Barnes W, Chang ATC, Stocker E,Adler RF, Hou A, Kakar R, Wentz F, Ashcroft P, Kozu T, Hong Y,Okamoto K, Iguchi T, Kuroiwa H, Im E, Haddad Z, Huffman G,Ferrier B, Olson WS, Zipser E, Smith EA, Wilheit TT, North G,Krishnamurti TN, Nakamura K. 2000. The status of the TropicalRainfall Measuring Mission (TRMM) after two years in orbit. J.Appl. Meteorol. 39: 1965–1982.

Lawrence D, Webster PJ. 2002. The boreal summer intraseasonaloscillation and the South Asian monsoon. J. Atmos. Sci. 59:1593–1606.

Maheshwari RC, Mathur IC. 1968. ‘Radar reflectivity studies of Indiansummer monsoon over Northwest India’. Pp. 98–103 in preprints for13th Radar Meteorology Conference, Montreal. American Meteorol.Society: Boston.

Murakami T. 1987. Orography and monsoons. Chapter 12 inMonsoons. JS Fein and PL Stephens (eds.) John Wiley and Sons:New York.

Narayanan V. 1967. Radar observation of a monsoon rain squall atBombay. Indian J. Meteorol. Geophys. 18: 397–402.

Nayava JL. 1974. Heavy monsoon rainfall in Nepal. Weather 20:443–450.

Newton CW, Fankhauser JC. 1964. Movement and developmentpatterns of convective storms, and forecasting the probability of stormpassage at a given location. US Weather Bureau: Washington DC.

Oye D, Case M. 1995. REORDER: A program for gridding radar data.Research Data program, Atmospheric Technology Division, NationalCenter for Atmospheric Research, Boulder, CO.

Palmen E, Newton CW. 1969. Atmospheric circulation systems: theirstructure and physical interpretation. Academic Press: New York.

Pant GB, Kumar KR. 1997. Climates of South Asia. John Wiley andSons: Chichester, UK.

Rajagopal EN, Harenduprakash L, Mitra AK, George JP. 2001.Description of the analysis and forecast data produced byNCMRWF for experimental ocean state forecast. Technical noteNCMRWF/OSF/1.

Ramage CS. 1971. Monsoon Meteorology. Academic Press: New York.Rao YP. 1976. Southwest Monsoon. Meteorological Monograph No.

1/1976. India Meteorological Department: Delhi.

Sanders F. 1984. Quasi-geostrophic diagnosis of the monsoondepression of 5–8 July 1979. J. Atmos. Sci. 41: 538–552.

Sawyer JS. 1947. The structure of the intertropical front over NWIndia during the SW monsoon. Q. J. R. Meteorol. Soc. 69:346–369.

Schumacher C, Houze RA Jr. 2003a. Stratiform rain in the tropicsas seen by the TRMM precipitation radar. J. Climate 16:1739–1756.

Schumacher C, Houze RA Jr. 2003b. The TRMM precipitation radar’sview of shallow, isolated rain. J. Appl. Meteorol. 42: 1519–1524.

Schumacher C, Houze RA Jr. 2006. Stratiform precipitation productionover sub-Saharan Africa and the tropical East Atlantic as observedby TRMM. Q. J. R. Meteorol. Soc. 132: 2235–2255.

Seshadri N. 1963. A radar study of heights of tops of cumulonimbusclouds around New Delhi. Indian J. Meteorol. Geophys. 14:46–49.

Shrestha ML. 2000. Interannual variation of summer monsoon rainfallover Nepal and its relation to Southern Oscillation Index. Meteorol.Atmos. Phys. 75: 21–28.

Shukla J. 1978. CISK-barotropic-baroclinic instability and the growthof monsoon depressions. J. Atmos. Sci. 35: 495–508.

Sikka DR. 1977. Some aspects of the life history, structure andmovement of monsoon depressions. Pure Appl. Geophys. 115:1383–1412.

Smull BF, Houze RA Jr. 1987. Rear inflow in squall lines with trailingstratiform precipitation. Mon. Weather Rev. 115: 2869–2889.

Steiner M, Houze RA Jr, Yuter SE. 1995. Climatological characteriza-tion of three-dimensional storm structure from operational radar andrain gauge data. J. Appl. Meteorol. 34: 1978–2007.

Szoke EJ, Zipser EJ. 1986. A radar study of convective cells inmesoscale systems in GATE. Part II: Life cycles of convective cells.J. Atmos. Sci. 43: 199–218.

Szoke EJ, Zipser EJ, Jorgensen DP. 1986. A radar study of convectivecells in mesoscale systems in GATE. Part I: Vertical profile statisticsand comparison with hurricanes. J. Atmos. Sci. 43: 182–198.

Wang B, Webster PJ, Teng H. 2005. Antecedents and self-induction ofactive-break south Asian monsoon unraveled by satellites. Geophys.Res. Lett. 32: L04704, DOI: 10.1029/2004GL020996.

Wang XL, Gaffen DJ. 2001. Late-twentieth-century climatology andtrends of surface humidity and temperature in China. J. Climate 14:2833–2845.

Webster PJ. 1987a. The elementary monsoon. Chapter 1 in Monsoons.JS Fein and PL Stephens (eds.) John Wiley and Sons: New York.

Webster PJ. 1987b. The variable and interactive monsoon. Chapter 11in Monsoons. JS Fein and PL Stephens (eds.) John Wiley and Sons:New York.

Webster PJ. 2006. The coupled monsoon system. Chapter 1 in TheAsian Monsoon. B Wang (ed.) Praxis: Chichester, UK. 1–66.

Webster PJ, Tomas RA. 1997. An atlas of precipitation and boundarylayer winds during intraseasonal oscillation events in the IndianOcean (1985–1995 extended summer season). Tech. Report No.2. Program in Atmospheric and Oceanic Sciences, University ofColorado: Boulder.

Webster PJ, Magana VO, Palmer TN, Shukla J, Tomas RA, Yanai M,Yasunari T. 1998. Monsoons: Processes, predictability, and theprospects for prediction. J. Geophys. Res. 103: 14451–14510.

Wilton DC. 2005. ‘South Asian monsoon convection as viewed by theTRMM precipitation radar’. MS thesis, University of Washington,USA.

Xie S-P, Xu H, Saji NH, Wang Y, Liu WT. 2006. Role of narrowmountains in large-scale organization of Asian monsoon convection.J. Climate 19: 3420–3429.

Yuter SE, Houze RA Jr. 1995. Three-dimensional kinematic andmicrophysical evolution of Florida cumulonimbus. Part II: Frequencydistributions of vertical velocity, reflectivity, and differentialreflectivity. Mon. Weather Rev. 123: 1941–1963.

Zuidema P. 2003. Convective clouds over the Bay of Bengal. Mon.Weather Rev. 131: 780–798.


Monsoon convection in the Himalayan region as seen … · merow et al., 1998) required remapping to...

Documents

Transcript of Monsoon convection in the Himalayan region as seen … · merow et al., 1998) required remapping to...