Chiang Mai J. Sci. 2010; 37(3) 429

Mesoscale Simulation of a Very Heavy Rainfall Eventover Mumbai, Using the Weather Research andForecasting (WRF) ModelSukrit Kirtsaeng* [a, b], Somporn Chantara [b] and Jiemjai Kreasuwun [b][a] Thai Meteorological Department, Bangkok 10260, Thailand.[b] Environmental Science Program and Center for Environmental Health, Toxicology and Management of

Chemicals, Faculty of Science, Chiang Mai University, Chiang Mai 50200, Thailand.*Author for correspondence; e-mail: sukritk@hotmail.com

Received: 17 September 2009Accepted: 2 March 2010

ABSTRACTThe simulation of a severe weather phenomenon, in this case the unprecedented

heavy rainfall over Mumbai in India on July 26, 2005, was selected for this study. The mesoscalenumerical weather prediction model used here utilized the Advanced Research Weather ResearchForecast model (version 3.0.1), as developed at the National Center for Atmospheric Research(NCAR) in the USA. The study used the Kian-Fritsch (KF), Betts–Miller–Janjic (BMJ) andGrell–Devenyi ensemble (GD) cumulus parameterization schemes across three nested domainconfigurations. The precipitation simulation results were compared with rainfall observationdata from the Tropical Rainfall Measuring Mission. The NCEP analyses, with a 1 1 degreeresolution and 26 levels, were utilized to verify the simulation’s resulting large-scale circulationpattern, moisture content and relative humidity fields. The 24-hour simulated cumulative rainfalldata was created from the different measurements taken at 0300 UTC on July 27, 2005 overMumbai. It can be observed that the maximum rainfall simulated from the KF, BMJ and GDschemes at 0000 UTC on July 25, 2005, under the initial conditions, was 48, 64 and 32 cmrespectively, while the TRMM shows a maximum rainfall of 32 cm at that time. The centre ofmaximum rainfall was reduced drastically for all 0000 UTC measurements taken on July 26(the following day), where the initial condition experiments simulated a rainfall amount ofonly 16 cm. The specific location of the intense rainfall around Mumbai was very-well simulatedin the BMJ for 0000 UTC July 25 initial conditions.

Keywords: Heavy rainfall; Cumulus parameterization; WRF model.

1. INTRODUCTION Severe weather systems are generally

associated with strong, gusty winds and heavyprecipitation. The numerical prediction ofsuch events remains one of the mostchallenging problems in the field of

meteorology. Most of the global modelsdeveloped thus far generally underestimate thetotal rainfall produced in any heavyprecipitation event, and also contain errors interms of their prediction on the timing and

Chiang Mai J. Sci. 2010; 37(3) : 429-442www.science.cmu.ac.th/journal-science/josci.htmlContributed Paper

430 Chiang Mai J. Sci. 2010; 37(3)

location of the event. [1-3] For a betterprediction of flash floods, it is necessary tounderstand the dynamics and physicsassociated with isolated heavy precipitationand the dynamic features associated withthunderstorms and tornados etc. Heavyrainfall occurs frequently around Mumbaiduring the summer monsoon season.The rainfall over the northern parts ofMumbai on July 26, 2005 was extremely heavy.

On that day, within a span of a few hours,northern Mumbai received record rainfall[4, 5], with Santacruz recording 94.4 cm ofrainfall for the day; whilst heavy rainfall of104.9 cm was recorded at Vihar Lake, locatedaround 15 kilometers northeast of Santacruz.Bhandup, located southeast of Vihar Lake,received 81.5 cm of rainfall, whilst Colaba insouthern Mumbai, on the other hand, receivedonly 7 cm.

Figure 1. A Map of India and of Mumbai.

Figure 2 shows 24 hours of rainfall validat 0000 UTC on July 27, 2005 in Mumbai andin neighboring weather stations, as recordedby the India Meteorological Department(IMD) and observed by the Tropical RainfallMeasuring Mission (TRMM). Jenamani et al.[4] studied the daily rainfall statistics forMumbai using observation data spanningmore than 100 years, and showed that theheavy rainfall recorded on July 26, 2005 wasthe highest ever recorded in Mumbai.

This event disrupted lives, causing heavydamage to property as well as deaths andinjuries. Prior knowledge of such an extremeevent by even a day or just a few hours couldhave minimized the loss of life [6]. This severeweather phenomenon caused around 500

deaths and costs one billion US Dollars indamage [7]. Although the MeteorologicalOffice (UKMO) model was one of the fewmodels that predicted up to 80 cm of rainfallover a small area of Mumbai on July 26, - 24hours in advance, most of the numericalmodels, both global and mesoscale, failed tosimulate this catastrophic event in anoperational or hindsight mode [6]. Althoughsome modeling studies of this extreme rainfallevent [8-10] have been reported in a hindsightmode, none of the operational models gavean accurate real-time prediction. The study ofDeb et al. [8], examined the performance ofthe WRF modeling system (version 2.0.2) forthe simulation of this severe rainfall event,using varying horizontal resolutions and

Chiang Mai J. Sci. 2010; 37(3) 431

cumulus parameterization schemes. Themodel’s performance was evaluated byexamining the different predicted parameters,such as upper and lower level circulations,temperature, moisture content and rainfall.Their study found that the GD cumulusparameterization had a high predictingcapability when compared to the BMJ schemefor this event. However, Chang et al. [9]studied the impact of land surfaceparameterizations on a simulation of the July26 event using two meso-scale models: theMM5 and the WRF. This results showed thatthe rainfall simulations produced using theWRF model were better than thecorresponding MM5 model in terms of bothlocation and intensity. The WRF model wasalso used to study the various small- to large-scale dynamical interactions, such as theformation of convective cells and large-scalewind circulations, interactions that resulted in

the localized and heavy precipitation overMumbai on that day [10].

The purpose of this study is to identifythe impact of three cumulus parameterizationschemes in the WRF model, used for thesimulation of this extreme rainfall event, andalso to reveal the sensitivity of the initialconditions. Taking the observed features, suchas large-scale upper and lower levelcirculations, moisture content and the thermalstructures from the NCEP analyses, thegridded, observed rainfall from the TRMMwas analyzed. The performance of eachscheme is presented here, with respect to theupper- and lower-level circulations, thermalstructures and the contour of rainfall.

This scientific study is structured asfollows: Section 2 describes the model andthe experimental design, and after that,Section 3 describes the results and discussion.The conclusions are given in Section 4.

Figure 2. 24-hour cumulative rainfall (cm) for 0300 UTC on July 27, 2005; observed bya) IMD at different stations near Mumbai, and b) TRMM .

432 Chiang Mai J. Sci. 2010; 37(3)

2. MATERIALS AND METHODS2.1 Model

The mesoscale model used in this studyis the Advanced Research WRF (ARW) model(version 3.0.1), which has been available sinceApril 2008. The WRF is a limited area, non-hydrostatic primitive equation model, withmultiple options for various physicalparameterization schemes [11]. This versionemploys Arakawa C-grid staggering forhorizontal grids and a fully compressiblesystem of equations. A terrain following sigmacoordinate is used in the vertical grids. Thetime-split integration uses a third-orderRunge–Kutta scheme with a smaller time-stepfor acoustic and gravity wave modes. Incombination with multiple-nest capability, alarge number of physics options make themodel capable of performing simulations onany scale; limited only by data resolution, quality,and computer resources.

2.2 Experimental DesignThe heavy rainfall event over Mumbai

on July 26, 2005 was selected for the subjectof this study. For this research, three nesteddomains were configured as shown in Figure3. The domain of interest was centered onMumbai (18.56oN, 72.51oE) and contained 27vertical levels. Domain 1, the outer domain,had 100 75 grid points in the horizontal planewith east-west and north-south directionsrespectively, covering 52.67oE to 92.34oE and4.11oS to 31.88oN, with a horizontal resolutionof 45 km, and with the sub-domain having a15 km grid spacing with 178 133 grid pointsin the horizontal east-west and north-southdirections respectively (set in the ratio of 1:3),covering 60.63oE to 84.38oE and 10.00oN to26.70oN. The fine-mesh Domain 3, with a 5km grid spacing, had 208 202 grid points(set in the ratio of 1:3), covering 67.87oE to77.14oE and 14.18oN to 22.70oN.

Figure 3: WRF model domains used in this study.

Chiang Mai J. Sci. 2010; 37(3) 433

The different dynamic and physicaloptions used for this study in terms ofcumulus parameterization schemes (that is theKF, BMJ and GD schemes) included the WRFsingle-moment 6-class Graupel (WSM-6)microphysics scheme, used along with theRapid Radiative Transfer Model (RRTM)scheme for long waves. The Duhia schemewas used for short-wave physics, the Monin-Obukhov scheme for the surface layer, theNoah Land-Surface Model for surfacephysics and the MRF scheme for treatmentof the Planetary Boundary Layer (PBL). Sixexperiments were carried out with the abovedomain configurations, as tabulated inTable 1.

The analysis data, with a resolution of1 1 degrees and taken at 6-hourly intervals,was obtained from the Global DataAssimilation System (GDAS) of the NationalWeather Service National Center for Environ-mental Prediction (NCEP), and was used forthe preparation of initial and boundary

conditions.This analysis was then interpolated to the

WRF-model grid to provide two initialconditions; the first, an initial condition takenat 0000 UTC on July 25 2005, and a secondinitial condition taken at 0000 UTC on July26, 2005. The model was run for 54 hoursand 30 hours respectively (up to 0600 UTCon July 27, 2005). The model utilizes variousterrestrial datasets for terrain, land-use, soiltype, soil temperature, vegetation fractions andmonthly albedo etc., taken from the WRF userwebsite [12]. To verify the results from themodel, TRMM was used. TRMM is a jointmission between NASA and the JapanAerospace Exploration Agency (JAXA). Itspurpose is to monitor tropical and subtropicalprecipitation and to estimate the associatedlatent heating. TRMM provides systematicvisible, infrared and microwave measurementsof rainfall in the tropics, as key inputs toweather and climate research.

Table 1. Design of the experiments.

Start-time End-timeCumulus

Experiment nameParameterization

EXP1 0000 UTC 0000 UTC Modified EXP25KFJuly 25, 2005 July 27, 2005 Kain-Fritsch

EXP2 0000 UTC 0000 UTC Betts-Miller-Janjic EXP25BMJJuly 25, 2005 July 27, 2005 scheme

EXP3 0000 UTC 0000 UTC Grell-Devenyi EXP25GDJuly 25, 2005 July 27, 2005 ensemble

EXP4 0000 UTC 0000 UTC Modified EXP26KFJuly 26, 2005 July 27, 2005 Kain-Fritsch

EXP5 0000 UTC 0000 UTC Betts-Miller-Janjic EXP26BMJJuly 26, 2005 July 27, 2005 scheme

EXP6 0000 UTC 0000 UTC Grell-Devenyi EXP26GDJuly 26, 2005 July 27, 2005 ensemble

434 Chiang Mai J. Sci. 2010; 37(3)

Figure 4.The simulated 850 hPa wind (m/s) along with magnitude, as valid at 1200 UTC onJuly 26, 2005 from the NCEP and using the different experiments: a) NCEP, b) EXP25KF,c) EXP25BMJ, d) EXP25GD, e) EXP26KF, f) EXP26BMJ, and g) EXP26GD. (Contourlevels 3, 6 and 12; winds of 20 and 25 m/s are shaded).

Figure 5. The simulated 300 hPa wind (m/s) along with its magnitude, as valid at 1200 UTCon July 26, 2005 from the NCEP and using the different experiments: a) NCEP, b) EXP25KF,c) EXP25BMJ, d) EXP25GD, e) EXP26KF, f) EXP26BMJ, and g) EXP26GD (Contourlevels 3, 6 and 9; winds of 12, 15 and 20 m/s are shaded).

Chiang Mai J. Sci. 2010; 37(3) 435

3. RESULTS AND DISCUSSION3.1 Upper and Lower-level Wind

The simulated 850-hPa winds valid at1200 UTC on July 26, 2005 over the Indiansubcontinent, and from the differentexperiments, are presented in Figure 4. All theexperiments predicted strong westerly ornorthwesterly winds over the western coastof India, with a maximum speed of 20 m/s.The experiment EXP26GD simulated windsof a magnitude 20 m/s over Mumbai (Figure4g), whereas in all the other simulations, windsof a magnitude 12-15 m/s were predicted inthis region. All of the schemes showed thepresence of a strong cyclonic circulationover Orissa and the eastern part of MadhyaPradesh, though the centre of the low wassimulated a bit further east as compared tothe analyses. It is clear from the figure that allthe experiments underestimated the winds inthe Arabian Sea, where a strong westerly flow

with a maximum speed of 20 m/s can beseen in the NCEP, not visible in the simulatedresults. Figure 5 shows the simulated 300-hPawinds valid at 1200 UTC on July 26, 2005,using the different experiments. All experi-ments simulated the upper level easterly ornortheasterly flow reasonably well, with aspeed of approximately 9 m/s over Mumbai.One interesting feature is that the EXP25GDsimulated 300-hPa winds of a magnitude of15 m/s over the Arabian Sea near Mumbai.On the whole, both the upper and lower levelwinds were simulated well using the model.

3.2 Thermal Structure and MoistureFields

Figure 6 shows the NCEP and thesimulated 850 hPa temperatures valid at1200 UTC on July 26, 2005, from all theexperiments. The simulated 850 hPa tempera-ture from all the experiments aligns well with

Figure 6. The simulated 850 hPa temperature (K) as valid at 1200 UTC July 26, 2005 takenfrom the NCEP and using the different experiments: a) NCEP, b) EXP25KF, c) EXP25BMJ,d) EXP25GD, e) EXP26KF, f) EXP26BMJ, and g) EXP26GD. (Contour level 1.5).

436 Chiang Mai J. Sci. 2010; 37(3)

Figure 7. Simulated 300 hPa temperature (K), as valid at 1200 UTC July 26, 2005 taken fromthe NCEP and using the different experiments: a) NCEP, b) EXP25KF, c) EXP25BMJ, d)EXP25GD, e) EXP26KF, f) EXP26BMJ, and g) EXP26GD. (Contour level 1).

the observed figure. It can be seen that thestructure of the isotherms is almost the samein all the experiments, except for some small-scale features simulated in synoptic resolution(45 Km) experiments (Figure 6b-g). Thesimulated 300 hPa temperatures valid for1200 UTC on July 26, 2005 (Figure 7) alsoshow warmer isotherms in the upper levelwhen compared with the NCEP analysis.Figure 8 represents the simulated latitude-height, cross-section of relative humidity fromthe different experiments around Mumbai

(average along longitude 72.5oE to 73.5oE)for 1200 UTC on July 26, 2005. All the experi-ments show a very high relative humidity,with a maximum as high as 95%, throughout400 hPa over Mumbai. Some small-scalefeatures are visible in the simulation, while theseare not seen in the NCEP analysis. As a wholeall the experiments simulated the observedthermal and moisture fields quite well, whencompared with the corresponding NCEPanalysis.

3.3 RainfallFigure 9 shows the simulated 24 hour

cumulative rainfall over Mumbai from theTRMM and the different experiments, validat 0300 UTC on July 27, 2005. It can beobserved that the maximum rainfall simulatedfrom the EXP25KF, EXP25BMJ andEXP25GD (Figure 9b, c and d) was 48, 64

and 32 cm respectively, while the TRMMshows a maximum total rainfall of 32 cm atthe same time.

The maximum rainfall from theexperiments EXP25BMJ and EXP25 KFwere very accurate when compared with theobserved position, though the position ofEXP25KF was a little south of the observed

Chiang Mai J. Sci. 2010; 37(3) 437

Figure 8. The latitude-height cross-section of relative humidity (%), as valid at 1200 UTCon July 26, 2005, taken from the NCEP and using the different experiments: a) NCEP, b)EXP25KF, c) EXP25BMJ, d) EXP25GD, e) EXP26KF, f) EXP26BMJ, and g) EXP26GD.(Contour level 5).

location. This level of maximum rainfalldecreased drastically with the experimentsEXP26KF, EXP26BMJ and EXP26GD, withsimulated rainfall amounts of only 16 cm(Figure 9d, e and f).

The particular location of the heavyprecipitation around Mumbai was very wellsimulated with the EXP25BMJ simulation,though the magnitude of rainfall was less thanthe actual observation. Though all theexperiments simulated a maximum rainfallcomparable to the TRMM rainfall aroundMumbai, the actual observed rainfall from theIMD showed a much higher rainfall whencompared to the TRMM rainfall. The rainfallrecorded by IMD is a point measurement,while the resolution in the TRMM rainfall is

0.25o both for scan and for pixel, and thiscould be one of reasons for their differencesin magnitude. Though TRMM rainfall is amerged, derived product from both IR andmicrowave measurements, sometimes TRMMunderestimates rainfall near the surface, dueto warm clouds.

The simulated 6-hourly cumulative rainfallfrom the most detailed resolution (innermostdomain: 5 km resolution) experimentsEXP25KF, EXP26KF, EXP25BMJ,EXP26BMJ, EXP25GD, and EXP26GDwere analyzed and a qualitative comparisonwith the TRMM merged rainfall was carriedout. The simulated 6-hourly cumulative rainfallfrom the EXP25KF, EXP26KF, EXP25BMJ,EXP26BMJ, EXP25GD and EXP26GD

438 Chiang Mai J. Sci. 2010; 37(3)

Figure 9. Simulated 24-hour cumulative rainfall (cm) from the TRMM and other experiments- 0300 UTC July 27, 2005: a) TRMM, b) EXP25KF, c) EXP25BMJ, d) EXP25GD,e) EXP26KF, f) EXP26BMJ, and g) EXP26GD (Contour levels 1, 2, 4, 8, 16, 32, 48, 64and 80).

Figure 10. Six-hourly accumulated rainfall (cm) from the TRMM: July 25, 2005 a) 0600 UTC,b) 1200 UTC, c) 1800 UTC July 26, 2005; d) 0000 UTC, e) 0600 UTC, f) 1200 UTC, andh) 0000 UTC July 27, 2005 respectively. (Contour levels 1, 2, 4, 8, 16, 32, and 48).

Chiang Mai J. Sci. 2010; 37(3) 439

Figure 11. Simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC July 26, 2005 to0000 UTC July 27, 2005 from EXP25KF and EXP26KF (Contour levels 1, 2, 4, 8, 16, 32,and 48).

Figure 12. Simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC July 26, 2005 to0000 UTC July 27, 2005 from EXP25BMJ and EXP26BMJ (Contour levels 1, 2, 4, 8, 16, 32,and 48).

440 Chiang Mai J. Sci. 2010; 37(3)

Figure 13. Simulated 6-hourly cumulative rainfall (cm) valid at 0600 UTC July 26, 2005 to0000 UTC July 27, 2005 from EXP25GD and EXP26GD (Contour levels 1, 2, 4, 8, 16, 32,and 48).

experiments, valid at 0600 UTC, 1200 UTCand 1800 UTC on July 26, 2005, and 0000UTC on July 27, 2006, are shown in Figures11, 12 and 13 respectively. The structure andpattern of the simulated rainfall using the initialcondition 0000 UTC on July 25, 2005 closelyresembled the TRMM observation, while theinitial conditions at 0000 UTC on July 26, 2005did not simulate the observed pattern ofrainfall accurately across all the experiments.The structure of rainfall in the EXP25GDsimulation was more accurate when comparedto EXP25KF and EXP25BMJ, where thecentre of maximum was shifted slightlynorthward (over Mumbai). This might be dueto the interaction of the detailed resolutionwith Domain 1 and Domain 2 in the nestedexperiment, where the error from the 45 and15 km resolution transferred to the 5 kminnermost domain during the model

integration. At 1200 UTC on July 26, 2005(Figures 11b, 13b), both EXP25KF andEXP25GD realistically simulated a localizedarea of heavy precipitation around southernMumbai. The quantum of rainfall predictedby both experiments (EXP25KF andEXP25GD), around 16 cm, was morerealistic than the 4 cm for EXP25BMJ. TheEXP26KF, EXP26BMJ and EXP26GDexperiments also simulated unrealistic rainfallalong the Western Ghats both at 1200 UTCand 1800 UTC on July 26, 2005 (Figure 11f,12f and 12f). At 1800 UTC on July 26, 2005,the EXP25BMJ experiment simulated rainfallof around 16cm. The experiments EXP25KF,EXP25BMJ and EXP25GD simulated alocalized area of heavy rainfall aroundMumbai at 0000 UTC on July 27, 2005, withthe magnitude ranging from 16-24 cm, andthe TRMM also showed rainfall of around

Chiang Mai J. Sci. 2010; 37(3) 441

16cm at 0000 UTC on July 27, 2005 overMumbai. While the simulation using the BMJscheme showed a very heavy precipitationzone in both the EXP25BMJ (Figure 12d) andEXP26BMJ experiments (Figure 12h),EXP26KF (Figure 11d) also simulated alocalized area of heavy rainfall aroundMumbai. However, the centre of maximumrainfall for EXP25GD (Figure 13d),EXP26GD (Figure 13h) and EXP26KF(Figure 11h) were shifted slightly southward,eastward and southward respectively. One ofthe major limitations of all the simulations wasthat none of the schemes, either at the coarseror the finer domains, were able to simulatethe observed rainfall over the northwesternpart of India.

4. CONCLUSIONSA qualitative assessment of the simulation

of a heavy rainfall event on July 26, 2005 overMumbai, India was attempted with WRFmodel version 3.0.1 at three nested domainconfigurations, with horizontal resolutions of45, 15 and 5 km respectively, using the KF,BMJ and GD schemes. The sensitivity of thetwo different initial conditions: 0000 UTC onJuly 25, 2005 and 0000 UTC on July 26, 2005were also evaluated in this study. The large-scale circulation features in all the simulations,including moisture and temperature patterns,both at upper and lower levels, closelyresembled the NCEP analyses, and thoughthe localized heavy precipitation aroundMumbai was captured for both initialconditions, it was captured particularly wellfor 0000 UTC on July 25, 2005. Theexperiment EXP25BMJ simulated theobserved rainfall structure around Mumbai;however, the simulated rainfall in theEXP25KF and EXP25GD experiments wasreasonably accurate as well, though the rainfallwas shifted slightly southeast from the actualposition. All the experiments underestimated

the rainfall, when 0000 UTC on July 26, 2005was used as the initial condition. It may bethat from running a simulation period of lessthan 24 hours, this initial condition had anerror in precipitation quantity and position atperiods +06, +12, +18 hour (0600, 1200,1800 UTC on July 26, 2005). The BMJ schemeoutperformed the KF and GD schemes,while the results from the last version showedthat GD outperformed the BMJ scheme inthe same case [8]. Thus, all model users shoulddecide on the suitable physical parametersbefore utilizing an updated version of any ofthe models.

The BMJ scheme was derived from theBetts-Miller convective adjustment scheme[13]. In this scheme, the deep convectionprofiles and the relaxation time are variableand depend on the cloud efficiency; a non-dimensional parameter that characterizes theconvective regime. Cloud efficiency dependson the entropy change, precipitation and themean temperature of the cloud, as well as theshallow convection moisture profile which isderived from the requirement that the entropychange be small and non-negative. This schememay also be refined for higher horizontalresolutions, primarily through modificationsof the triggering mechanism. Thus it can beconcluded that the above factor was one ofthe main reasons responsible for the highvertical instability in the atmosphere aroundMumbai on this extreme event day.

This is a very preliminary attempt to usethe WRF model (version 3.0.1), and it isrecommended that further testing using morecase studies, as well as an improved versionof the model plus the incorporation ofdifferent satellite data and Doppler Radar data,through use of the WRF data assimilationsystem, is carried out. Although the aboveconclusions are based on a very limitednumber of experiments, this study providessome insights for WRF model users in the

442 Chiang Mai J. Sci. 2010; 37(3)

tropics and may prompt the modelingcommunity to pursue and evaluate real-timequantitative precipitation forecasting.

ACKNOWLEDGEMENTSThe authors are very grateful to Dr. Sanjib

Kumar Deb, Atmospheric Sciences Division,Space Applications Centre (SAC) for commentsand suggestions. The authors wish toacknowledge the Centre for Space Scienceand Technology Education in the Asia Pacific(affiliated to the United Nations) and SAC,The Indian Space Research Organization(ISRO) Ahmedabad for their knowledge,encouragement and help. The authors alsoacknowledge the use of the WRF model,which is available on Internet, by NCEP/NCAR and the WRF User Support Groupfor useful suggestions during the modelinstallation. The use of reanalyzed data by theNCEP/NCAR, the TRMM rainfall mapsfrom http://trmm.gsfc.nasa.gov, the rainfallobservation data from the IMD are thankfullyacknowledged. Partly financial support fromCenter for Environmental Health, Toxicologyand Management of Chemicals is gratefullyacknowledged.

REFERENCES

[1] Wang B. and LinHo., Rainy seasons ofthe Asian-Pacific monsoon. J. Climate,2002; 15: 386-398.

[2] Lin J.L., Kiladis G.N., Mapes B.E.,Weickmann K.M., Sperber K.R., Lin W.,Wheeler M.C., Schubert S.D., Del-GenioA., Donner L.J., Emori S., Gueremy J.F.,Hourdin F., Rasch P.J., Roeckner E. andScinocca J.F., Tropical intraseasonalvariability in 14 IPCC AR4 climatemodels. Part I: Convective signals. J.Climate, 2006; 19: 2665-2690.

[3] Lei M., Niyogi1 D., Kishtawal C., PielkeR.A., Beltr’ an-Przekurat A., Nobis T.E.and Vaidya S.S., Atmos. Chem. Phys. Discuss.,2008; 8: 8773–8816.

[4] Jenamani R.K., Bhan S.C. and Kalsi S.R.,Observational/forecasting aspects of themeteorological event that caused a record

highest rainfall in Mumbai. Curr Sci.,2006; 90(10): 1344–1362.

[5] Shyamala B. and Bhadram C.V.V., Impactof mesoscale-synoptic scale interactionson the Mumbai historical rain eventduring 26–27 July 2005. Curr Sci. 2006;91(12): 1649–1654.

[6] Bohra A.K., Basu S., Rajagopal E.N.,Iyengar G.R., Das G.M., Ashrit R. andAthiyaman B., Heavy rainfall episode overMumbai on 26 July 2005: assessment ofNWP guidance. Curr Sci. 2006; 90(9):188–1194.

[7] NCDC: National Climatic Data Center,Climate of 2005, http://www.ncdc.noaa.gov/ sotc?report=hazards&year=2005&month=jul# Flooding.

[8] Deb S.K., Kishtawal C.M., BongirwarV.S. and Pal P.K., The simulation of heavyrainfall episode over Mumbai: impact ofhorizontal resolutions and cumulusparameterization schemes. Nat Hazards.,2009; DOI 10.1007/s11069-009-9361-8.

[9] Chang H.I., Kumar A., Niyogi D.,Mohanty U.C., Chen F. and Dudhia J.,The role of land surface processes onthe mesoscale simulation of the July 26,2005 heavy rain event over Mumbai,India. Glob Planet Change., 2009; 64:87-103.

[10] Kumar A., Dudhia J., Rotunno R.,Niyogib D. and Mohanty U.C., 2008:Analysis of the 26 July 2005 heavy rainevent over Mumbai, India using theWeather Research and Forecasting (WRF)Model. Q. J. R. Meteorol Soc., 2008; 134:1897–1910.

[11] Skamarock W.C., Klemp J.B., Dudhia J.,Gill D.O., Barker D.M., Duda M.G.,Huang X.Y., Wang W. and Powers J.G.,A Description of the Advanced ResearchWRF Version 3. National Center forAtmospheric Research, Boulder,Colorado, 2008.

[12] The WRF user’s website, http://www.mmm.ucar.edu/wrf/users

[13] Betts A.K., A new convective adjustmentscheme. Part I: Observational andtheoretical basis. Quart J Roy Meteor Soc.1986; 112: 677–691.