Indian Journal of Radio & Space Physics Vol 43, February 2014, pp 103-123

Effect of cumulus and microphysical parameterizations on the JAL cyclone prediction

M Venkatarami Reddy1,2, S B Surendra Prasad1,2, U V Murali Krishna1 & K Krishna Reddy1,$,* 1Semi-arid-zonal Atmospheric Reseach Centre (SARC), Yogi Vemana University, Kadapa 516 003, Andhra Pradesh, India 2Department of Physics, Acharya Nagarjuna University, Guntur 522 510, Andhra Pradesh, India $E-mail: [email protected]

Received 19 August 2013; revised 7 November 2013; accepted 8 January 2014

Weather Research and Forecasting (WRF) model is used to predict the track and intensity of JAL cyclone, which formed during 04-08 November 2010 over the . The model has been simulated with numerous experiments using the logical/scientific combination of convection and micro-physics schemes. The model simulations have been conducted with different initial conditions to know the effective track and intensity prediction of JAL cyclone. In addition, the effect of cumulus parameterization schemes at different resolution (27 and 9 km) on the cyclone track and intensity is reported. The model simulated results showed the importance of cumulus schemes and their role at 9 km horizontal resolution. The results indicate that the track predicted by Kain-Fritsch (KF) scheme is in good agreement with the observed track in all the experiments and the land fall error is minimum (~11 km) for the combination of Ferrier and KF scheme with 0000 hrs UTC on 04 November 2010 as initial condition. Strong intensity is produced by KF, New Grell (NG) schemes and weak intensity is produced by Betts-Miller-Janjic (BMJ) scheme with all microphysics parameterization (MP) combinations. Further, the dependency of intensity of cyclone has been studied in terms of surface latent heat flux, divergence and vorticity fields. To validate the model performance, different meteorological parameters are derived from the model simulations over three different regions and are compared with the observed meteorological parameters. The model results are in good agreement with the observed parameters, but variations are observed at the landfall /dissipation of the cyclone.

Keywords: JAL cyclone, Weather Research and Forecasting (WRF model), Cumulus parameter, Microphysical parameters, Vorticity, Divergence, Convergence PACS Nos: 92.60.Aa; 92.60.hb; 92.60.Wc

1 Introduction Initially, TC forecast was carried out using a Tropical cyclones (TC) form over Bay of Bengal computer oriented half persistence and half (BOB) with strong winds, torrential rains and tidal climatology technique in India4. Later on, Mohanty & waves1,2 during pre- and post-monsoon season and Gupta5 have used multi-level primitive equation cause a severe damage to lives and property over the models with parameterization of physical processes east coast of India and Bangladesh. The tropical for the prediction of cyclone track. Prasad & Rama storms form and develop over the warm tropical Rao6 have performed numerous sensitive experiments oceans under favourable environmental conditions3. on Quasi-Lagrangian model (QLM) to forecast the The BOB in the is a potentially track of cyclones. In the last few years, the rapid energetic region for the development of TCs and it advancement of computing power and high resolution contributes about 7% of the global annual tropical mesoscale models are being used to know the TC storms. There is a need to develop a regional weather structure, intensification and movement7-12. Srinivas forecasting system to give the lead time forecast et al.13 explained the importance of boundary layer warnings and to mitigate the disasters. Hence, and convective parameterization in TCs and reported accurate prediction of the track and intensity of the model sensitive studies for the Andhra severe cyclones is essential to minimize the loss of human cyclone in 2003. lives and property damage. Several models and Globally, several studies were carried out using methods have been developed to predict the accurate different numerical models to simulate the structure, position of the cyclone to some extent. The improvement intensity, track of cyclones, hurricanes and typhoons14-21. in prediction helps to issue the appropriate cyclone The high resolution mesoscale models were being warnings to the society and disaster management. used extensively for hurricane sensitivity studies to 104 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

know the structure, intensification and movement may not be suitable for other regions. Most of the with different physical processes22,23. Fovell & Su24 studies mainly focused on the prediction of track and showed that the logical combination of microphysics intensity of the cyclones over BOB with resolutions parameterization (MP) and cumulus parameterization greater than 10 km using different physical processes (CP) schemes significantly influenced the track and and few of them performed less than 10 km resolution land fall of Hurricane Rita using WRF model at 30 with different MP and PBL processes. The present and 12 km resolution. They also reported the impact study focused mainly on the prediction of track and of cloud MP schemes on track forecasting. Different intensity of JAL cyclone using WRF model with parameterization schemes have been developed by different CP and MP schemes to find the best set several researchers but all of them have certain of physics options. During the last decade, JAL limitations25-28. cyclone alone passed over the semi-arid region The TC intensity forecast varied with different of India. Hence, the best logical combination of microphysical processes in numerical models. Zhu & parameterization schemes for JAL cyclone was Zhang29 examined the sensitivity of the hurricanes reported and the results compared with observational intensity with different microphysical processes in the meteorological parameters over three regions, viz. MM5 model. The intensity and track forecast of Oceanic (12.8-13.8°N, 80-90.2°E), inland (13-14°N, cyclones depend on the evolution and the distribution 79-80°E) and semi-arid (14-15°N, 78-79°E) at 3 km of the heat rates30, which depends on the release of resolution with and without CP schemes. latent heat flux by the condensation within the system. Deshpande et al.31 studied the track and intensity 2 Brief description of JAL cyclone prediction for cyclone Gonu using MM5 model at A severe cyclonic storm, JAL developed 90 and 30 km resolutions and concluded that (04-08 November 2010) over BOB. It moved track and intensity prediction were highly sensitive west-northwestwards and intensified into severe for convective parameterization schemes than cyclonic storm on 06 November 2010. India other physical parameterization schemes. Loh et al.32 Meteorological Department (IMD) reported41 that the also reported that the planetary boundary layer JAL cyclone moved to the southwest of BOB having parameterization (PBL) schemes did not show much lower ocean thermal energy and high vertical wind impact on the prediction of track and intensity of shear in association with the strong easterlies in the near equatorial typhoons. However, the intensity upper tropospheric level. The strong wind shear led forecast still remains a challenging problem in both to westward shearing of convective clouds from the operational and research communities17,33,34. centre of system and the lower Ocean thermal Even though the CP schemes are important for the energy caused to convection un-sustainability over simulation of TCs, there is a limitation in usage of the region. The severe cyclonic storm, JAL gradually CP schemes in the numerical simulations at high weakened into a deep depression and crossed north resolutions (e.g. less than 10 km grid spacing). Liu – South Andhra Pradesh coast, close to et al.35 and Braun et al.36 simulated the TC without the north of near 13.3°N and 80.3°E around any CP scheme in the high resolution because it is 1600 hrs UTC on 07 November 2010. It continued to not worthy to use CP schemes in the numerical move west–northwestwards and further weakened simulations where the grid horizontal resolution is into well marked low pressure area. The convective less than 10 km. Even though the domain horizontal clouds were sheared to the large extent to west on the resolution is less than 10 km, CP scheme plays date of landfall, i.e. on 07 November 2010. During a key role for the accurate prediction of TC37-39. the last decade, JAL cyclone is the only cyclone, McFarquhar et al.40 used 6 km horizontal resolution which passed over interior parts of Andhra Pradesh domain and reported the change in intensity and track (semi-arid region) and it caused heavy rainfall over prediction of TC with different CP schemes. the coastal region as well as the interior parts of semi- TCs track and intensity simulation are sensitive to arid regions. model resolution, CP, MP and PBL schemes used in the model. However, the best set of parameterization 3 Model description schemes in the model is not same for all the cyclones. The non-hydrostatic compressible WRF model The logical combination of schemes for one region was developed by National Center for Atmospheric REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 105

Research (NCAR)42. It has features like a fully Miller convective adjustment scheme47,48 was commonly compressible, Eulerian non-hydrostatic control used in simulations35,40. The GD is a equation set, a terrain-following, hydrostatic pressure cloud ensemble scheme with 144 sub-grid members. vertical coordinate system with the constant pressure It uses 16 ensemble members derived from 5 popular surface at the top level of model. The staggered closure assumptions to obtain an ensemble‐mean grid, like Arakawa-C grid, used in the model and a realization at a given time and location49. The third-order Runge-Kutta time integration scheme MP schemes used in the model are WRF Single used for both horizontal and vertical directions. The Moment 6 class scheme (WSM6) [“6” stands for six WRF model incorporates several processes like MP, classes of hydrometeors]50, Thompson et al.51 CP, PBL, surface layer, land surface, and long wave (THOMP) scheme, Lin et al.52 (LIN) scheme, Ferrier and shortwave radiations, with multiple options for et al.53 (FERR) scheme. Dudhia54 scheme is used each process. The model details and equations are for the short‐wave radiation, Rapid Radiative Transfer available in the NCAR technical note. In the present Model (RRTM) scheme55 is used for the long-wave study, WRF-ARW model is used to simulate the radiation and Yonsei University (YSU) scheme track and intensity of JAL cyclone. The horizontal is used for the PBL. resolutions used are 27 km, 9 km and 3 km (Fig. 1) with 27 vertical levels. The input details of the model 4 Data and Experiment description are given in Table 1. In the present study, WRF-ARW model is WRF-ARW model offers user specified options, initialized with the National Center for Environmental number of physical parameterization schemes. Prediction Final Analyses (FNL) data from Global Simulation experiments have been conducted with Data (with every 6 hourly interval of time period four CP schemes such as Kain-Fritsch (KF), Betts- Table 1 — Model configuration and parameterization schemes Miller-Janjic (BMJ), Grell-Devenyi ensemble (GD) used in the WRF model and New Grell (NG). KF scheme, a deep and Model NCAR mesoscale model WRF shallow convection sub-grid scale scheme, uses a Dynamics Non-hydrostatic with 3-D coriolis mass flux approach with downdrafts and Convective force Available Potential Energy (CAPE) removal time Grid size/ Horizontal Domain1:(123×135)×27/ scale. The KF convective parameterization is based grid resolution (193×151)×27 43,44 on a simple cloud model of Kain & Fritsch . Domain2:(193×151)×27/ 9 km×9 km It includes the effects of detrainment, entrainment, Domain1:(154×139)×27/ 3 km×3 km simple microphysics and moist updrafts and Data NCEP Final Analyses (FNL) data 45,46 downdrafts. The BMJ scheme derived from Betts- Horizontal grid system Arakawa-C grid Vertical coordinates Terrain-following hydrostatic pressure vertical co-ordinate with 51 vertical levels Micro physics (i) Lin et al. (ii) Ferrier (new Eta) (iii) WSM 6-class Scheme graupel scheme (iv) Thompson Long wave radiation Rapid Radiative Transport Model scheme (RRTM) scheme Short wave radiation Dudhia scheme scheme Land surface scheme Noah land surface model Planetary Boundary Yonsei University scheme (YSU) Layer (PBL) scheme Cumulus parameterization (i) Kain- Fritsch (new Eta) scheme scheme (ii) Betts-Miller-Janjic (iii) Grell-Devenyi ensemble

Fig. 1 — Model domains used for the prediction of JAL cyclone (iv) New Grell Scheme (NG) with horizontal resolution of 27 km, 9 km (D02) and 3 km (D03) Diffusion option Simple diffusion 106 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

having the resolution of 1°×1°) Assimilation System Figure 2(a) shows the comparison of predicted tracks (GDAS)56 for the prediction of track and intensity with the India Meteorological Department (IMD) of tropical cyclone. The details of different types of track using the combination of LIN MP scheme with data undergoing into the analyses, can be obtained four CP schemes in 27 km resolution experiment from the Data Documentation for Data Set 6141B Exp-D1I3. According to the IMD observations, the (NCEP Model Output—FNL archive data). The cyclone JAL was formed over the South Andaman model results of track and intensity are compared Sea near 8.0°N and 92.0°E [Fig. 2(a)] and moved to with IMD observational data. For validation, the west-northwest, whereas in LIN+KF scheme the model results of meteorological parameters (sea cyclone formed over the South Andaman Sea near level pressure, temperature and relative humidity) 8.5°N and 91.5°E, moved to northwest and then are compared with the data collected from Automatic moved in the direction of west-northwest. In LIN+NG Weather Station (AWS) located in oceanic, in land scheme, the storm developed near 8.5°N and 91.5°E and semi-arid regions, respectively. For each region, and follows the observed track up to 1800 hrs UTC data is collected from seven AWS located nearby to on 06 November 2010 (12°N, 83.8°E) and deviated cover complete region of interest. thereafter. Also, the LIN+BMJ scheme shows more or A detailed study on the impacts of different less similar behaviour with genesis point near to combinations of CP and MP schemes on track and 8.5°N and 91.5°E and follows the path up to 0000 hrs intensity of JAL cyclone using WRF-ARW Model is UTC on 06 November 2010 (10.8°N, 86°E) and carried out. For this purpose, 64-core (8 cluster + 5TB then deviated more thereafter. In LIN+GD scheme, storage server) High Performance Computational the storm developed near 8.5°N and 91.5°E (HPC) facility established at Yogi Vemana University and followed the track up to 0000 hrs UTC on 06 (YVU), Kadapa is utilized. Six numerical experiments November 2010 (11°N, 85.8°E). The distance were designed and performed to investigate the between the IMD observed and model predicted characteristics of the simulated track and intensity. tracks are considered as track errors and are shown The details of the experiments are reported in Table 2. in Fig. 3(a). The track error lies in the range 35-140

5 Results and Discussion km, 55-280 km, 65-245 km and 20-165 km for KF, BMJ, GD and NG schemes, respectively. The track 5.1 Simulation of JAL cyclone at 27 km resolution error is minimum for LIN+KF scheme and maximum The JAL cyclone is simulated with different for LIN+BMJ. The track error is 87, 210, 170 and 112 CP schemes and MP schemes at 0000 hrs UTC on km for KF, BMJ, GD and NG schemes, respectively 03 November 2010 (Exp-D1I3) and 0000 hrs UTC on during the landfall of cyclone at 1600 hrs UTC on 07 04 November 2010 (Exp-D1I4) as initial conditions November 2010. Figure 2(b, c and d) shows the track for WRF model at 27 km resolution. prediction using FERR, WSM6 and THOMP MP 5.1.1 Track schemes with 4 CP schemes in Exp-D1I3 along with The track of the cyclone is predicted with different the track observed by IMD. In FERR scheme, all the CP and MP schemes using the WRF model. CP schemes except BMJ scheme follow the observed Table 2 — Description of the performed numerical experiments track well. Among these schemes, FERR+KF scheme Experiment Horizontal No of grid Initial time, followed the track well compared to the others. Also, name resolution, km points Date FERR+GD and FERR+NG schemes followed the Exp-D1I3 27 123×135 0000 hrs UTC, observed track well compared to LIN+GD and 03 November 2010 LIN+NG schemes, respectively. The track simulated Exp-D2I3 9 193×151 0000 hrs UTC, using WSM6 and THOMP MP schemes followed the 03 November 2010 track as in LIN MP scheme with all the CP schemes, Exp-D3I3 3 154×139 0000 hrs UTC, whereas WSM6+BMJ and THOMP+BMJ deviated 03 November 2010 more when compared to LIN+BMJ. Further, a small Exp-D1I4 27 123×135 0000 hrs UTC, variation is observed in the track error simulated 04 November 2010 using WSM6, and THOMP schemes except Exp-D2I4 9 193×151 0000 hrs UTC, 04 November 2010 WSM6+BMJ and THOMP+BMJ compared to LIN Exp-D3I4 3 154×139 0000 hrs UTC, MP scheme [Fig. 3(c and d)]. The track error 04 November 2010 is minimum for FERR MP scheme with KF, GD REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 107

and NG schemes compared to all other MP resolution using the same MP and CP schemes. In this combinations [Fig 3(a-d)]. Among all the CP experiment (Exp-D1I4), the track simulated with LIN schemes, KF scheme followed the track well in all MP scheme [Fig. 2(e)] shows small variations MP schemes. The KF combination showed minimum compared to observed track simulated in Exp-D1I3. track errors compared to the other combinations. Here also, LIN+KF scheme followed the observed From the above experimental results, it is clear track well compared to other CP schemes. The track that in the combination of KF scheme with different error simulated using LIN MP scheme with 4 MP schemes, FERR and LIN has minimum track different CP schemes in Exp-D1I4 is shown in error. In the combination of BMJ with all MP Fig. 3(e). The track error in LIN+KF decreased schemes, LIN and WSM6, in GD scheme FERR initially and then increases compared to LIN+KF in and THOMP and in NG scheme FERR and LIN has Exp-D1I3, but the track error for LIN+BMJ, LIN+GD minimum track error. and LIN+NG are similar to Exp-D1I3. Simulated Similar to the above experiment, the cyclone is track and track error using FERR MP scheme with simulated in the experiment Exp-D1I4 with 27 km 4 CP schemes in Exp-D1I4 are shown in Fig. 2(f) and

Fig. 2 — Track simulated using: (a) LIN, (b) FERR, (c) WSM6, and (d) THOMP MP scheme, with four CP schemes (KF, BMJ GD and NG) in experiment (Exp-D1I3) along with the observed track; (e) LIN, and (f) FERR scheme, with four CP schemes (KF, BMJ GD and NG) in experiment (Exp-D1I4) along with the observed track 108 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

Fig. 3(f), respectively. Track simulated in Exp-D1I4 MSLP and MSW by IMD. The observed MSLP using FERR MP scheme followed the track simulated decreased from 1002 to 988 hPa; and then increased in Exp-D1I3. The track error decreased in FERR+KF to 1004 hPa during the passage of cyclone; thereafter, and FERR+GD schemes, whereas in FERR+NG, the the cyclone is weakened. LIN+BMJ scheme predicted track error increased in Exp-D1I4 compared to Exp- the MSLP very well up to 1800 hrs UTC on 06 D1I3. Track simulated using WSM6 and THOMP in November 2010 and deviated thereafter from IMD Exp-D1I4 are comparable with the track simulated in observations. Whereas the LIN+GD scheme predicted

Exp-D1I3 for the respective schemes (figure not shown). the MSLP well up to 0600 hrs UTC on 06 November The landfall error for all CP and MP combinations in 2010. The MSW are predicted well by LIN+BMJ and all the experiments is shown in Table 3. LIN+GD up to 1800 hrs UTC on 06 November 2010 and deviated thereafter. Among these two, LIN+GD 5.1.2 Intensity scheme followed the MSW well compared to The intensity of the tropical cyclone is discussed LIN+BMJ scheme. From these observational studies, in terms of minimum central sea level pressure LIN+KF and LIN+NG are not able to represent the (MSLP) and maximum surface wind (MSW). true intensity of the cyclone. The minimum variations Figure 4(a and b) shows the model predicted MSLP in predicted MSLP in BMJ, GD, KF and NG schemes and MSW using LIN MP scheme with 4 CP schemes are 3, 13, 18 and 28 hPa, respectively and the MSW in the experiment (Exp-D1I3) along with observed are 5, 9, 14 and 15 ms-1 for the respective schemes.

Fig. 3 — Simulated track error (km) using: (a) LIN, (b) FERR, (c) WSM6, and (d) THOMP MP scheme, with four CP schemes (KF, BMJ GD and NG) in experiment Exp-D1I3; (e) LIN, and (f) FERR scheme, with four CP schemes (KF, BMJ GD and NG) in experiment Exp-D1I4; (g) LIN, and (h) FERR scheme, with four CP schemes (KF, BMJ GD and NG) in experiment Exp-D2I3; (i) LIN, and (j) FERR scheme, with four CP schemes (KF, BMJ GD and NG) in experiment Exp-D2I4 REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 109

The experimental results show that strongest storm is ms-1 in both MP schemes, respectively. But in both produced by KF and NG schemes and the weakest MP schemes, the BMJ scheme shows less MSW storm produced by BMJ and GD schemes. compared to the observed value. Intensity of the cyclone is simulated using FERR, Figure 4(i and j) shows the intensity simulated WSM6 and THOMP MP Scheme with four CP by LIN MP scheme with 4 different CP schemes schemes in Exp-D1I3. In FERR MP scheme, BMJ in the experiment Exp-D1I4. The intensity simulated scheme predicted the MSLP and MSW well up to in Exp-D1I4 is similar to those simulated in Exp-D1I3 1800 hrs UTC on 05 November 2010 and decrease with a small variation in MSLP and MSW. In until 0000 hrs UTC on 07 November 2010 and then LIN+KF and LIN+GD schemes, the MSLP decreased increased thereafter compared to observed MSLP and in LIN+BMJ and LIN+NG schemes, the MSLP and MSW [Fig. 4(c and d)]. The FERR+GD scheme remains almost same compared to Exp-D1I3. followed the LIN+GD scheme until 0600 hrs UTC on Here, also the strongest storm was simulated by KF 07 November 2010 and then deviated. Whereas, KF and NG schemes, whereas weakest storm was and NG schemes shows similar intensity compared simulated by BMJ scheme. The intensity simulated with LIN+KF and LIN+NG schemes. The minimum using FERR MP scheme with 4 CP schemes in variations in the predicted MSLP in BMJ, GD, Exp-D1I4 is shown in Fig. 4(k and l). Initially, KF and NG schemes are 3, 19, 18 and 23 hPa, MSLP increased in FERR+KF scheme, decreased respectively and the MSW are -5, 9, 14 and 11 ms-1 in the remaining combinations and finally became for the corresponding schemes. WSM6 and THOMP almost same in all combinations compared to Exp- schemes predicted the intensity similar to LIN MP D1I3. Whereas, MSW initially decreased in KF scheme with 4 CP schemes [Fig. 4(e-h)]. The scheme and increased in GD scheme and then became minimum variations in predicted MSLP by BMJ, GD, almost equal in both the combination compared to KF and NG schemes are -2, 1, 13, 28 and -2, 16, 13, Exp-D1I3. But the variations in MSW in BMJ and NG 23 hPa and the MSW are 1, 1, 11, 15 and 3, 7, 11, 15 schemes are not uniform.

Table 3 — Predicted landfall error with different CP and MP schemes

Landfall errors Experiments Exp-D1I3 Exp-D2I3 Exp-D1I4 Exp-D2I4

Schemes Distance, km Time, h Distance, km Time, h Distance, km Time, h Distance, km Time, h

KF+ LIN 79 -1 79 -1 90 -1 90 -1 BMJ+ LIN 288 4 288 4 244 3 244 3 GD+ LIN 234 5 234 5 244 6 244 6 NG+LIN 123 1 123 0 90 0 90 0 KF+ FERR -22 -2 -22 -2 -11 -3 -11 -3 BMJ+FERR 400 7 400 6 385 8 385 8 GD+ FERR 57 1 57 1 22 0 22 0 NG+ FERR 90 5 90 5 -11 -1 -11 -1 KF+WSM6 90 0 90 0 134 0 134 0 BMJ+WSM6 370 4 370 5 341 0 341 0 GD+WSM6 244 6 244 6 223 6 223 6 NG+WSM6 178 3 178 3 134 2 134 2 KF+THOMP 79 0 79 0 101 1 101 1 BMJ+ THOMP 536 5 536 5 525 6 525 7 GD+ THOMP 211 3 211 3 189 6 189 6 NG+ THOMP 90 2 90 1 57 0 57 0 KF+ MMS 256 2 256 2 200 1 200 1 BMJ+ MMS 545 0 540 0 589 -1 589 -1 GD+ MMS 326 -1 326 -1 343 1 343 1 NG+ MMS 299 0 299 0 291 2 291 2 110 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

5.2. Simulation of JAL cyclone at 9 km resolution and became comparable afterwards for FERR+KF The effect of CP schemes on the cyclone track and FERR+GD combinations. Whereas the track and intensity predicted are investigated using error does not show much variation in FERR+BMJ Exp-D2I3 and Exp-D2I4 with 9 km resolution. and FERR+NG schemes compared to Exp-D1I3.

5.2.1 Track The track simulated by WSM6 and THOMP The track and track error simulated by LIN MP schemes in Exp-D2I3 are comparable to Exp-D1I3 scheme with four CP schemes in Exp-D2I3 is shown (figure not shown). in Fig. 5(a) and Fig. 3(g). The track simulated in The experiment Exp-D2I4 was simulated with Exp-D2I3 is similar to those simulated in Exp-D1I3. 9 km resolution using all the MP and CP The track error simulated using LIN+KF scheme combinations. The track simulated using LIN MP decreased initially, then increased and finally became scheme in Exp-D2I4 is shown in Fig. 5(c). In LIN MP comparable to Exp-D1I3. Whereas, the tracks for scheme, the track simulated is comparable to the LIN+BMJ, LIN+GD and LIN+NG schemes are observed track as in Exp-D1I4. The track error comparable to the respective combination in Exp- in LIN+KF scheme [Fig. 3(i)] decreased, whereas D1I3. The simulated track and track error using FERR in LIN+BMJ, LIN+GD and LIN+NG schemes, no MP scheme with 4 CP schemes are shown in Fig. 5(b) much variation in track error is observed. Similarly, and Fig. 3(h). The track simulated by FERR MP the track simulated using FERR MP scheme followed scheme in Exp-D2I3 is similar to those simulated the observed track well as in Exp-D1I4 [Fig. 5(d)]. in Exp-D1I3. The track error decreased initially Also, the track error decreased in FERR+NG schmes

Fig. 4 — Estimated MSLP (hPa) and estimated MSW (ms-1) of the cyclone: (a and b) LIN, (c and d) FERR, (e and f) WSM6, and (g and h) THOMP MP scheme, with four CP schemes (KF, BMJ GD and NG) in experiment (Exp-D1I3); (i and j) LIN, (k and l) FERR scheme, with four CP schemes (KF, BMJ GD and NG) in experiment Exp-D1I4 REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 111

[Fig. 3(j)] and the variation are not uniform in the However, in the present study, when the initial remaining schemes. Even in the 9 km resolution conditions change from 03 November to 04 November, experiment also, variation in track error is comparable the landfall error is decreased in some MP and CP to variations in track error in 27 km resolution combinations and increased in other MP and CP experiment when the initial condition is changed from combinations both in 27 and 9 km resolutions.

0000 hrs UTC on 03 November to 0000 hrs UTC 5.2.2 Intensity on 04 November 2010. The land fall error for all Figure 6(a and b) shows the model simulated the schemes in Exp-D2I3 and Exp D2I4 is given in MSLP and MSW using LIN MP scheme with four CP Table 3. From the observations, it is clear that there is schemes in Exp-D2I3 along with observed MSLP and no much variation in the land fall error and track error MSW by IMD. The minimum variations in predicted simulated in 27 km and 9 km resolution experiments. MSLP using BMJ, GD, KF and NG schemes are 6, From these experiments, one can notice that 15, 18 and 37 hPa, respectively and the MSW are 11, all the schemes followed the observed track well until 13, 14 and 27 ms-1 for the corresponding schemes. In 0000 hrs UTC on 07 November 2010 and deviated this experiment, the strongest intensities are shown by thereafter. But the track simulated with FERR+KF LIN+NG and LIN+KF; and weakest intensities by schemes initially moved to west until 1200 hrs UTC LIN+BMJ and LIN+GD schemes. In FERR MP on 06 November 2010 and then to north-west, scheme, the simulated intensity followed the observed whereas the observed track moved to north-west from intensity as in Exp-D1I3 [Fig. 6(c and d)]. But the beginning. In all experiments, the combination in Exp-D2I3, all the MP and CP combinations of KF scheme with all the MP schemes followed show lower MSLP and higher MSW compared to the observed track well compared to other CP Exp-D1I3. Hence, the intensity simulated using LIN schemes because of its moist updraft and downdrafts, and FERR MP schemes is stronger in Exp-D2I3 detrainment, entrainment and simple microphysics44 compared to Exp-D1I3. The intensity simulated using along with the widespread triggering of convection37. LIN and FERR MP schemes with all CP schemes in Pattanaik & Rama Rao57 observed a decrease in Exp-D2I4 is shown in Fig. 6(e-h). The intensity landfall error when the initial conditions change simulated by both MP schemes are stronger compared from 30 April to 01 May for the cyclone Nargis. to Exp-D1I4 in all the CP schemes.

Fig. 5 — Ttrack simulated using: (a) LIN and (b) FERR MP scheme, with four CP Schemes (KF, BMJ, GD and NG) in experiment Exp-D2I3 along with observed track; (c) LIN and (d) FERR MP scheme, with four CP Schemes (KF, BMJ, GD and NG) in experiment Exp-D2I4 112 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

The prediction of track and intensity of tropical 06 November 2010, the latent heat flux at 1200 hrs cyclone depends on choice of MP schemes used in the UTC on 06 November 2010 in Exp-D1I3 and model23. Deshpande et al.31 observed that the track Exp-D2I3 is considered. The experiments, Exp-D1I3 and intensity of TC are very sensitive to the choice of and Exp-D2I3, are chosen because the intensity convective parameterization schemes compared predicted in Exp-D1I3 is low and in Exp-D2I3, it is to other physical parameterization schemes using high in all schemes. In Exp-D1I3 using LIN MP 90 and 30 km resolutions in the MM5 model. It is scheme, all CP schemes produce stronger surface clear from the present study also that the predicted latent heat flux where the wind speed is maximum. track and intensity are sensitive to CP schemes rather The stronger surface latent heat flux is produced by than MP schemes even at high resolutions (9 km). KF scheme and the weaker surface latent heat flux Hence, for the accurate prediction of TC, high by GD scheme. The NG and BMJ schemes show resolution models with a better choice of CP schemes intermediate values for surface latent heat flux are essential. To investigate further, the impact of [Fig. 7(a-d)]. The MSW and MSLP predicted at the the different MP and CP schemes on numerical eye-wall are weaker in all the schemes. Similarly, simulations of intensity of the cyclone, surface latent predicted latent heat flux also shows its minimum heat flux, convergence and divergence and vertical at the eye-wall region in all the schemes. But the structure of wind speed and temperature over the intensity predicted by KF and NG schemes storm core region are examined for the simulations is stronger, whereas the intensity predicted by BMJ at 1200 hrs UTC on 06 November and 0000 hrs UTC and GD schemes is weaker. To illustrate further, on 07 November 2010, when the cyclone was the surface latent heat flux at 0000 hrs UTC and 1200 intensified. hrs UTC on 07 November 2010 is considered. At 0000 hrs UTC on 07 November 2010, the intensity as 5.3 Predictions of surface latent heat fluxes well as surface latent heat flux decreased in KF The observations show that the intensity of cyclone scheme, whereas in NG scheme, the intensity of is strong from 1200 hrs UTC on 06 November to the cyclone remains the same but the surface latent 0000 hrs UTC on 07 November 2010. To explain heat flux increases. But in BMJ and GD schemes, the strong intensity observed at 1200 hrs UTC on the surface latent heat flux and intensity of the

Fig. 6 — Estimated MSLP (hPa) and estimated MSW (ms-1) of cyclone: (a and b) LIN, (c and d) FERR MP scheme, with 4-different CP schemes (KF, BMJ GD and NG) in experiment Exp-D2I3; (e and f) LIN, (g and h) FERR MP scheme, with 4-different CP schemes (KF, BMJ GD and NG) in experiment Exp-D2I4 REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 113

cyclone increases and is more in GD scheme flux [Fig. 7 (e-h)]. Further, at 1200 hrs UTC on (figure not shown). Moreover, at 1200 hrs UTC on 07 07 November 2010, the stronger intensity is predicted

November 2010, the intensity and surface latent by NG and KF; whereas the stronger surface latent heat flux are stronger in NG scheme and weaker in heat flux is predicted by KF and NG schemes, GD scheme, whereas it is moderate in both KF and respectively. However, BMJ and GD schemes show BMJ schemes [Fig. 8 (a-d)]. the same intensity and surface latent heat fluxes Similar to Exp-D1I3, the surface latent heat flux [Fig. 8 (e-h)]. The surface latent heat flux in Exp- is simulated in Exp-D2I3 using LIN MP scheme with D2I3 is more in all schemes compared to Exp-D1I3. all CP schemes at 1200 hrs UTC on 06 November Previous studies by Zhu & Zhang29 and Li & Pu23 2010. In Exp-D2I3, the stronger intensity is predicted showed that the intensity of cyclone depends on the by NG and KF schemes, but the strong surface latent latent heat flux released. From the observations of JAL heat flux is predicted by KF and NG scheme, cyclone, it is clear that the intensity of the cyclone respectively. However, BMJ and GD schemes show depends not only on the current surface latent heat flux weakest intensity along with surface latent heat but also on the preceding surface latent heat flux.

Fig. 7 — Predicted surface latent heat flux (Wm-2, shaded), Fig. 8 — Predicted surface latent heat flux (Wm-2, shaded), wind wind speed (ms-1, thick black line) and sea level presure speed (ms-1, thick black line) and sea level presure (hPa, solid red (hPa, solid red lines) at 1200 hrs UTC on 06 November 2010 lines) at 1200 hrs UTC on 07 November 2010 using LIN MP using LIN MP scheme with: (a) KF, (b) BMJ, (c) GD, and scheme with: (a) KF, (b) BMJ, (c) GD, and (d) NG CP scheme, in (d) NG CP scheme, in experiment Exp-D1I3; (e) KF, (f) BMJ, experiment Exp-D1I3; (e) KF, (f) BMJ, (g) GD, and (h) NG CP (g) GD, and (h) NG CP scheme, in experiment Exp-D2I3 scheme, in experiment Exp-D2I3 114 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

5.4 Prediction of cyclone structure using divergence and convergent inflow between 950 and 850 hPa levels is vorticity fields 58 observed in GD scheme compared to KF scheme According to Willoughby , air flows in a tropical [Fig. 11(a-i)]. FERR+KF, FERR+NG and FERR+GD cyclone works in an in-up-and-out pattern. At low schemes predicted the same intensity at 0000 hrs levels of the atmosphere, the air flows toward the UTC on 07 November 2010. Because the convergent storm center. This inward air brings heat and moisture field and vorticity integrated between 950 and from the ocean surface into the storm. At the upper 850 hPa levels at 0000 hrs UTC on 07 November troposphere, outflow exists to compensate the inflow 2010 show almost similar values for FERR+KF, at the low troposphere. It is a critical factor for FERR+NG and FERR+GD schemes [Fig. 11(d-l)]; maintenance and development of a storm eye-wall. To so, a strong convergent inflow in the lower troposphere elucidate this, the divergence field and vorticity of the and a strong divergent outflow in the upper troposphere TC are considered, integrated between 950 and 850 hPa and strong vorticity at the eye of the cyclone explain levels and divergence field integrated between the strong intensity of the cyclone. Similarly, weaker 200 and 100 hPa levels in Exp-D1I3 and Exp-D2I3. convergent flows in the lower troposphere, less Figure 9 shows the divergence field integrated divergent outflow in the upper troposphere and weak between 950 and 850 hPa levels in Exp-D1I3 using vorticity at the eye of the cyclone correspond to LIN MP scheme with all CP schemes at 1200 hrs weaker storm intensity. UTC on 06 November 2010. The negative values in the figure represent convergent flows; and positive 5.5 Vertical cross sections of wind speed and temperature values represent divergent flows. A strong convergent Environmental vertical wind shear is another inflow between 950 and 850 hPa levels and a strong important factor that could influence cyclone structure, and well organized divergent outflow between 200 development and maintenance. Figure 12(a-d) shows and 100 hPa and strong vorticity at the eye of the the vertical cross sections of meridional wind speed cyclone explains the more intense storm produced and temperature at 1200 hrs UTC on 06 November by the KF and NG schemes. Meanwhile, weaker 2010 for LIN MP scheme with different CP schemes convergent flows between 950 and 850 hPa, less in the experiment Exp-D1I3. The temperature field divergent flows between 200 and 100 hPa and weak has a warm core above the cyclone because of vorticity at the eye of the cyclone corresponds to compensating descending inside the cyclone. Among weaker storm intensity predicted by BMJ and GD all the schemes, KF and NG schemes show maximum schemes. Also, the intensity predicted in Exp-D2I3 temperature at the eye of the cyclone because KF and is stronger than predicted in Exp-D1I3 at 1200 hrs NG schemes show strong intensity (MSW). A vertical UTC on 06 November 2010. This is due to the strong column of low speed is observed in all the schemes. convergent inflow, divergent outflow at the eye-wall The distribution of the wind is symmetric around the region and strong vorticity observed at the eye of eye-wall region in KF and NG schemes. Also, the the cyclone in the Exp-D2I3 [Fig. (10)] compared to maximum wind is observed in the low levels in all the Exp-D1I3. schemes. A minimum wind is observed at the eye of To explain further the dependency of intensity of the cyclone (width of the eye) in BMJ, KF and GD cyclone on the vorticity, low level convergence and schemes and it is more pronounced in BMJ scheme. upper level divergence, the following three cases are This is because BMJ scheme shows weak intensity considered with different MP and CP schemes. In and the MSLP predicted by KF is less than GD scheme. THOMP+NG scheme, the MSLP and MSW are Figure 12(e-h) shows the vertical cross sections of smaller compared to THOMP+KF scheme at 1200 hrs meridional wind speed and temperature at 1200 hrs UTC on 06 November 2010. This can be explained by UTC on 06 November 2010 in the experiment the stronger vorticity predicted by NG scheme Exp-D2I3 using LIN MP scheme with different CP compared to KF scheme. A strong convergent inflow schemes. The temperature and meridional wind speed between 950 and 850 hPa levels is observed in KF observed at the eye of the cyclone is less in Exp-D2I3 scheme compared to NG scheme. In THOMP+GD compared to Exp-D1I3. The vertical wind is scheme, MSLP is almost equal to KF scheme; hence, more symmetric about eye wall region in Exp-D2I3 the vorticity predicted by GD scheme is comparable compared to Exp-D1I3. Hence, the intensity (MSLP to KF scheme. Whereas, MSW predicted by GD and MSW) of the cyclone depends on vertical cross scheme is smaller than KF scheme. Hence, a weak section of wind speed and temperature. REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 115

Fig. 9 — Divergence (10-5 s-1) field integrated between lower level (950-850 hPa), upper level (200-100 hpa) and lower level (950-850 hPa) vorticity (10-5 s-1) using LIN MP scheme with: (a, b and c) KF; (d, e and f) BMJ; (g, h and i) GD; (j, k and l) NG CP scheme, at 1200 hrs UTC on 06 November 2010 in experiment Exp-D1I3

6 Validation parameters. For this, simulation of JAL cyclone is To validate the model results, the meteorological carried out using different microphysical parameterization parameters like sea level pressure (slp), temperature schemes with and without CP schemes in WRF model and relative humidity are compared with the observed using Exp-D3I3 at 3 km resolution.

116 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

Fig.10 — Divergence (10-5 s-1) field integrated between lower level (950-850 hPa), upper level (200-100 hpa) and lower level (950-850 hPa) vorticity (10-5 s-1) using LIN MP scheme with: (a, b and c) KF; (d, e and f) BMJ; (g, h and i) GD; (j, k and l) NG CP scheme, at 1200 hrs UTC on 06 November 2010 in experiment Exp-D2I3 REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 117

Fig. 11 — Divergence (10-5 s-1) field integrated between lower level (950-850 hPa) and lower level (950-850 hPa) vorticity (10-5 s-1) - using THOMP MP scheme with: (a and d) KF, (b and e) GD, and (c and f) NG CP scheme, at 1200 hrs UTC on 06 November 2010; using FERR MP scheme with: (g and j) KF, (h and k) GD, and (i and l) NG CP scheme, at 0000 hrs UTC on 07 November 2010 in experiment Exp-D1I3 118 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

Fig. 12 —Vertical cross sections meridional wind speed (ms-1, shaded) and temperature (°C, dark black line) at 1200 hrs UTC on 06 November 2010 using LIN MP scheme with: (a) KF, (b) BMJ, (c) GD, and (d) NG CP scheme, in experiment Exp-D1I3; (e) KF, (f) BMJ, (g) GD, and (h) NG CP scheme, in experiment Exp-D2I3

Figure 13 shows the meteorological parameters During the landfall, observed slp decreases and it (slp, temperature, relative humidity and rainfall) increased after landfall. The observed slp and the slp observed using Automatic Weather Station (AWS) predicted by all the schemes remain the same up to and meteorological parameters (slp, temperature and 0300 hrs UTC on 07 November 2010 and after 0000 relative humidity) simulated in Exp-D3I3 between hrs UTC on 08 November 2010. The predicted slp 0000 hrs UTC on 06 November 2010 (before landfall) varied much compared to the observed slp during the and 0000 hrs UTC on 09 November 2010 (after landfall of the cyclone. The variation is maximum for landfall) over Oceanic region using different MP FERR and LIN and is minimum for WSM6 MP schemes with KF CP scheme and without any CP scheme. Even in the same experiment, without schemes. Figure 13(a) shows the observed pressure any CP scheme [Fig. 13(b)] also shows the same and slp predicted in Exp-D3I3 in the Oceanic region. behaviour before and after landfall of the cyclone. But REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 119

during the landfall of the cyclone, the slp decreased as in Exp-D3I3 with CP scheme. Figure 13(e) shows considerably compared to the same experiment with the observed and simulated relative humidity for

CP scheme. Figure 13(c) shows the observed Exp-D3I3 in the Oceanic region. The observed temperature and temperature simulated in Exp-D3I3 relative humidity increased before landfall and over Oceanic region. Observed temperature increases decreased during landfall and again increased after first and then decreases until 0300 hrs UTC on 07 landfall. In this case, the relative humidity predicted November 2010. After this, the temperature increases in all the MP schemes deviated more compared to the slightly and remains constant up to landfall of the observed relative humidity. Also, relative humidity cyclone. After landfall, the temperature decreases simulated in Exp-D3I3 without CP scheme [Fig. 13(f)] (0000 hrs UTC on 08 November 2010) and again shows similar behaviour as in Exp-D3I3 with CP increases following the diurnal pattern. The predicted scheme. temperature followed the observed temperature Figure 14 shows the comparison of observed and before 0300 hrs UTC on 07 November 2010 and after simulated meteorological parameters in Exp-D3I3 0000 hrs UTC on 08 November 2010. The between 0000 hrs UTC on 06 November 2010 and temperature simulated in Exp-D3I3 without CP 0000 hrs UTC on 09 November 2010 over in land and scheme [Fig 13(d)] also shows the same pattern semi-arid region using different MP schemes without

Fig. 13 — Observed and simulated (a) slp (hPa), (b) temperature (°C), and (c) relative humidity (%), of the cyclone using 4 MP schemes (LIN, FERR, WSM6 and THOMP) with KF CP scheme from 0000 hrs UTC on 06 November 2010 to 0000 hrs UTC on 09 November 2010 in the Oceanic region; (d) slp (hPa), (e) temperature (°C), and (f) relative humidity (%), of the cyclone using 4 MP schemes (LIN, FERR, WSM6 and THOMP) with KF scheme from 0000 hrs UTC on 06 November 2010 to 0000 hrs UTC on 09 November 2010 in the Oceanic region [vertical bars represent the observed rainfall (mm)] 120 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

Fig. 14 — Observed and simulated: (a) slp (hPa), (b) temperature (°C), and (c) relative humidity (%) of the cyclone using 4 MP schemes (LIN, FERR, WSM6 and THOMP) without CP scheme from 0000 hrs UTC on 06 November 2010 to 0000 hrs UTC on 09 November 2010 in the inland region; (d) slp (hPa), (e) temperature (°C), and (f) relative humidity (%) of the cyclone using 4 MP schemes (LIN, FERR, WSM6 and THOMP) without CP scheme from 0000 hrs UTC on 06 November 2010 to 0000 hrs UTC on 09 November 2010 in the semi-arid region [vertical bars represent the observed rainfall (mm)]

any CP schemes. Figure 14(a) shows the observed and dissipation time, whereas the remaining schemes simulated slp using Exp-D3I3 without any CP scheme deviated during the dissipation time. Figure 14(c) in the inland region. The slp predicted in the inland shows the observed and simulated relative humidity region shows higher values compared to the oceanic for Exp-D3I3 in the in land region. The simulated region. The simulated slp decreased 6-hours before relative humidity deviated a little as compared to the the dissipation of the cyclone in Exp-D3I3. All the observed relative humidity. schemes deviated from the observed slp between Figure 14(d) shows the observed and simulated slp landfall and dissipation time and followed well in the using Exp-D3I3 without any CP scheme in the remaining time. The variation is maximum for FERR semi-arid region. The observed slp decreased during and LIN and is minimum for THOMP MP scheme. the landfall of the cyclone and then increased Figure 14(b) shows the observed and simulated after landfall. Again the slp decreased during the temperature in Exp D3I3 in the inland region. During dissipation of the cyclone and increased afterwards. 07 November 2010, the temperature remained The decrement in the slp was more during dissipation constant throughout the day and increased after the compared to landfall time. All the schemes deviated dissipation of cyclone. The FERR scheme deviated from the observed slp between landfall and dissipation from observed temperature between landfall and time and followed well in the remaining time. The REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 121

variation was maximum for LIN MP scheme. Figure scheme predicted weak intensity and KF and NG 14(e) shows the observed and simulated temperature schemes predicted strong intensity with all MP in Exp-D3I3 in the semi-arid region. During 07 schemes in all the experiments. The KF and NG November 2010, the temperature remained constant schemes predicted stronger intensity with minimum throughout the day and increased after the dissipation track error. Whereas, BMJ scheme predicted weak of cyclone. The temperature increased slightly before intensity with maximum track errors. Intensity (MSLP landfall and then decreased during landfall. All the and MSW) of the cyclone depends on surface latent schemes deviated between landfall and dissipation heat flux, convergent inflow in the lower troposphere, time. Figure 14(f) shows the observed and simulated divergent outflow in the upper troposphere, vorticity, relative humidity for Exp-D3I3 in the semi-arid vertical cross section of wind speed and temperature. region. The simulated relative humidity deviated from The decrease in slp, temperature and increase in the observed relative humidity. relative humidity is observed during the landfall As the cyclone approached, the slp increased in the of the cyclone with AWS at oceanic, in land in land and semi-arid regions. However, during the and semi-arid regions. The model simulations well dissipation of cyclone, the observed temperature predicted the variation in meteorological parameters remained the same throughout the day in both in land before/after the depression/landfall of cyclone without and semi-arid regions, whereas in the oceanic region, using CP schemes, but significant variations have it varied during the landfall of the cyclone. The been observed during cyclone period using the CP observed rainfall was maximum over the oceanic Schemes. The model simulations clearly indicated the region (83 mm) compared to inland (79 mm) and importance of MP and CP schemes in the simulations semi-arid (65 mm) regions. The relative humidity of JAL cyclone. increased before the landfall of the cyclone in both oceanic and inland region. However, in semi-arid Acknowledgement region, relative humidity increased during dissipation The authors acknowledge the support of India of the cyclone. After dissipation, the relative humidity Meteorological Department (IMD), Govt of India, for decreased in all the regions at the same time. providing the JAL cyclone track information. They are thankful to Indian Space Research Organization 7 Summary and Conclusions (ISRO) for providing AWS data through MOSDAC The prediction of the tropical cyclone JAL formed website. They acknowledge the assistance of over Bay of Bengal is carried out using WRF model NOAA/OLR/ESRL PSD, Boulder, Colorado, USA, with different CP and MP schemes. The result shows for providing FNL data at their website at that the simulated track and intensity are sensitive to http://www.esrl.noaa.gov/. Two of the authors (MVR the choice of the CP and MP schemes used in the and SBSP) greatly acknowledge the financial support model. extended by Atmospheric Science Program (ASP), The track/landfall error decreased in some MP Indian Space Research Organization (ISRO), Govt of and CP combinations and increased in other India. One of the authors (UVMK) acknowledges MP and CP combinations when the initial conditions the support of Ministry of Earth Sciences (MoES), change from 0000 hrs UTC on 03 November 2010 to New Delhi, Govt of India for providing the fellowship 0000 hrs UTC on 04 November 2010. The track in carrying out this work. error is more sensitive to the choice of initial conditions, CP and MP schemes. Even, there is References no much variation in the track predicted using 27 km 1 Asnani G C, Tropical meteorology (Indian Institute of and 9 km resolutions, model simulates stronger Tropical Meteorology, Pune, India), 2005. 2 Pattanayak S & Mohanty U C, A comparative study on intensity in 9 km resolution compared to 27 km performance of MM5 and WRF models in simulation of resolution. The track predicted using KF scheme with tropical cyclones over Indian seas, Curr Sci (India), 95 (7) all MP schemes is in good agreement with the (2008) pp 923-936. observed track in all the experiments. The land fall 3 Gray W M, Global view of the origin of tropical disturbances error is minimum (-11 km) for the combination and storms, Mon Weather Rev (USA), 96 (1968) pp 669–700. of FERR+KF scheme in Exp-D1I4. 4 Sikka D R & Suryanarayana R, Forecasting the movement of tropical storms/depressions in the Indian region by computer The cyclone intensity simulations are performed in oriented technique wing climatology and persistence, terms of the MSLP and MSW in the model. The BMJ Indian J Meteorol Geophys, 23 (1972) pp 35–40. 122 INDIAN J RADIO & SPACE PHYS, FEBRUARY 2014

5 Mohanty U C & Gupta A, Deterministic methods for with the WRF-ARW model, Mon Weather Rev (USA), 137 prediction of tropical cyclone tracks, Mausam (India), 48 (2009) pp 3717-3743. (1997) pp 257-272. 20 Roacha R P & Caetano E, The role of convective 6 Prasad K & Rama Rao Y V, Cyclone track prediction by a parameterization in the simulation of a cyclone over the quasi-Lagrangian model, Meteorol Atmos Phys (Austria), 83 South Atlantic, Atmosfera (Mexico), 23 (1) (2010) pp 1-23. (2003) pp 173-185. 21 Jayaraman P, Oo S M, Raju P V S & Mohanty U C, 7 Trivedi D K, Sanjay J & Singh S S, Numerical simulation Performance of nested WRF model in typhoon simulations of a super cyclonic storm, Orissa (1999): Impact of initial over West Pacific and , Nat Hazards conditions, Meteorol Appl (UK), 9 (2002) pp 367-376, (Netherlands), 63 (2012) pp 1451-1470. doi: 10.1017/S1350482702003109. 22 Braun S A & Tao W –K, Sensitivity of high-resolution 8 Mohanty U C, Mandal M & Raman S, Simulation of Orissa simulations of Hurricane Bob (1991) to planetary boundary Super Cyclone (1999) using PSU/NCAR Mesoscale Model, layer parameterizations, Mon Weather Rev (USA), 128 Nat Hazards (Netherlands), 31 (2004) pp 373-390, doi: (2000) pp 3941–3961. 10.1023/B:NHAZ.0000023358.38536.5d. 23 Li X & PU Z, Sensitivity of numerical simulation of 9 Bhaskar Rao D V & Hari Prasad D, Sensitivity of tropical early rapid intensification of Hurricane Emily (2005) to cyclone intensification to boundary layer and convective cloud microphysical and planetary boundary layer processes, Nat Hazards (Netherlands), 41 (2007) pp parameterizations, Mon Weather Rev (USA), 136 (2008) pp 429-445, doi: 10.1007/s11069-006-9052-7. 4819-4838. 10 Bhaskar Rao D V, Hari Prasad D & Srinivas D, Impact of 24 Fovell R G & Su H, Impact of cloud microphysics on horizontal resolution and the advantages of nested domains hurricane track forecast, Geophys Res Lett (USA), 34 (2007) approach in the prediction of tropical cyclone intensification L24810, doi: 10.1029/2007/GL031723. and movement, J Geophys Res (USA), 114 (2009) D11106, 25 Frank W M, The cumulus parameterization problem, Mon doi: 10.1029/2008JD011623. Weather Rev (USA), 111 (1983) pp 1859-1871. 11 Reale O, Lau W K, Susskind J, Brin E, Liu E, Riishojgaard L P, 26 Emanuel K A & Raymond D J, The representation of Fuentes V & Rosenberg R, AIRS impact on the analysis and cumulus convection in numerical models, (Meteorology forecast track of tropical cyclone Nargis in a global data Monograph Series, 46) (Am Meteorol Soc, Boston, MA), assimilation and forecasting system, Geophys Res Lett 1993, 246. (USA), 36 (2009) L06812, doi: 10.1029/2008GL037122 . 27 Zhang D L, Kain J S, Fritsch J M & Gao K, Comments on 12 Shen B –W, Tao W –K, Lau W K & Atlas R, Predicting parameterization of convective precipitation in mesoscale tropical cyclogenesis with a global mesoscale model: numerical models: A critical review, Mon Weather Rev Hierarchical multi-scale interactions during the formation of (USA),122 (1994) pp 2222-2231. tropical cyclone Nargis (2008), J Geophys Res (USA), 115 28 Kuo Y –H, Bresch J F, Cheng M -D, Kain J, Parsons D B, (2010) D14102, doi: 10.1029/2009JD013140. Tao W –K & Zhang D –L, Summary of a mini workshop on 13 Srinivas C V, Venkatesan R, Bhaskar Rao D V & Hari cumulus parameterization for mesoscale models, Bull Am Prasad D, Numerical simulation of Andhra severe cyclone Meteorol Soc (USA), 78 (1997) pp 475-491. (2003): Model sensitivity to the boundary layer and 29 Zhu T & Zhang D L, Numerical simulation of Hurricane convective parameterization, Pure Appl Geophys (USA), 164 Bonnie (1998), Part II: Sensitivity to varying cloud (2007) pp 1465-1487. microphysical processes, J Atmos Sci (USA), 63 (2006) pp 14 Harr P A & Elsberry R L, Stucture of a mesoscale convective 109-126. system embedded in typhoon Robyn during TCM-93, 30 Wang Y, How do outer spiral rain bands affect tropical Mon Weather Rev (USA), 124 (1996) pp 634-652. cyclone structure and intensity? J Atmos Sci (USA), 66 15 Braun S A, A cloud-resolving simulation of Hurricane Bob (2009) pp 1250-1273, doi: 10.1175/2008JAS2737.1. (1991), storm structure and eye wall buoyancy, Mon Weather 31 Deshpande M, Pattnaik S & Salvekar P, Impact of physical Rev (USA), 130 (2002) pp 1573-1592. parameterization schemes on numerical simulation of super 16 Atlas R, Reale O, Shen B -W, Lin S -J, Chern J -D, Putman cyclone Gonu, Nat Hazards (Netherlands), 55 (2) (2010) pp W, Lee T, Yen K -S, Bosilovich M & Radakovich J, 211-231. Hurricane forecasting with the high-resolution NASA finite 32 Loh W, Juneng L & Tandang F, Sensitivity of Typhoon volume general circulation model, Geophys Res Lett (USA), Vamei (2001) simulation to planetary boundary layer 32 (2005) L03807, doi: 10.1029/2004GL021513. parameterization using PSU/NCAR MM5, Pure Appl 17 Krishnamurti T N, Pattnaik S, Stefenova L, Vijaykumar T S Geophys (USA), 168 (10) (2010) pp 1799-1811. V, Mackey B P, O’Shay A J & Pasch R J, The hurricane 33 Bender M A & Ginis I, Real case simulation of hurricane- intensity issue, Mon Weather Rev (USA), 133 (2005) pp ocean interaction using a high-resolution coupled model: 1886-1912, doi: 10.1175/MWR2954.1. Effects on hurricane intensity, Mon Weather Rev (USA), 128 18 Rogers R F, Black M L, Chen S S & Black R A, An (2000) pp 917-946. evaluation of microphysics fields from mesoscale model 34 Rogers R, Aberson S, Black M, Black P, Cione J, Dodge P, simulations of tropical cyclones. Part I: Comparisons with Dunion J, Gamache J, Kaplan J, Powell M, Shay N, Surgi N observations. J Atmos Sci (USA), 64 (2007) pp 1811-1834. & Uhlhorn E, The intensity forecasting experiment: A 19 Fierro A O, Rogers R F, Marks F D & Nolan D S, The NOAA multiyear field program for improving tropical impact of horizontal grid spacing on the microphysical and cyclone intensity forecasts, Bull Am Meteorol Soc (USA), 86 kinematic structures of strong tropical cyclones simulated (2006) pp 1523-1537. REDDY et al.: EFFECT OF CUMULUS AND MICROPHYSICAL PARAMETERIZATIONS ON JAL PREDICTION 123

35 Liu Y, Zhang D –L & Yau M K, A multi-scale numerical 46 Janjić Z I, Comments on Development and evaluation of a simulation of Hurricane Andrew (1992), Part I: Explicit convection scheme for use in climate models, J Atmos Sci simulation and verification, Mon Weather Rev (USA), 125 (USA), 57 (2000) 3686. (1997) pp 3073–3093. 47 Betts A K, Saturation point analysis of moist convective 36 Braun S A, Montgomery M T & Pu Z, High resolution overturning, J Atmos Sci (USA), 39 (1982) pp 1484-1505. simulation of Hurricane Bonnie (1998), Part I: The 48 Betts A K & Miller M J, A new convective adjustment organization of eye wall vertical motion, J Atmos Sci (USA), scheme, Part 2: Single column tests using GATE wave, 63 (2006) pp 19-42. BOMEX, ATEX, and arctic air mass data sets, Q J R 37 Davis C A & Bosart L F, Numerical simulations of the Meteorol Soc (UK), 112 (1986) pp 693-709. genesis of Hurricane Diana, Part II: Sensitivity of track and 49 Grell G A & Devenyi D, A generalized approach to intensity prediction, Mon Weather Rev (USA), 130 (2002) pp parameterizing convection combining ensemble and data 1100-1124. assimilation techniques, Geophys Res Lett (USA), 29 (14) 38 Moncrieff M W & Liu C, Representing convective (2002) 1693, doi: 10.1029/2002GL015311. organization in prediction models by a hybrid strategy, 50 Hong S –Y, Dudhia J & Chen S -H, A revised approach to J Atmos Sci (USA), 63 (2006) pp 3404-3420, doi: ice microphysical processes for the bulk parameterizations of 10.1175/JAS3812.1. clouds and precipitations, Mon Weather Rev (USA), 132 39 LI X & PU Z, Sensitivity of numerical simulations of the (2004) pp 103-120. early rapid intensification of Hurricane Emily to Cumulus 51 Thompson G, Rasmussen R M & Manning K, Parameterization schemes in different model horizontal Explicit forecast of winter precipitation using an improved resolutions, J Meteorol Soc Jpn (Japan), 87 (3) (2009) pp bulk microphysical scheme, Part I: Description and 403-421, doi: 10.2151/jmsj.87.403. sensitivity analysis, Mon Weather Rev (USA), 132 (2004) pp 519-542. 40 McFarquhar G M, Zhang H, Heymsfield G, Hood R, 52 Lin Y ‐L, Farley R D & Orville H D, Bulk parameterization Dudhia J, Halverson J B & Marks F, Factors affecting of the snow field in a cloud model, J Clim Appl Meteorol the evolution of Hurricane Erin (2001), and the (USA), 22 (1983) pp 1065-1092. distributions of hydrometeors: Role of microphysical 53 Ferrier B S, Jin Y, Lin Y, Black T, Rogers E & Di Mego G, processes, J Atmos Sci (USA), 63 (2006) pp 127-150, doi: Implementation of a new grid-scale cloud and rainfall 10.1175/JAS3590.1. scheme in the NCEP Eta Model, Preprints, 15th Conference 41 Indian Meteorological Department, Cyclone Annual report on Numerical Weather Prediction (Am Meteorol Soc, San 2010 (Indian Meteorological Department, Delhi), 2011. Antonio, TX), 2002, pp 280-283. 42 Skamarock W C, Klemp J B, Dudhia J, Gill D O, Barker D 54 Dudhia J, Numerical study of convection observed during M, Duda M G, Huang X, Wang W & Powers J G, the winter monsoon experiment using a mesoscale A description of the advanced research WRF version 3, two-dimensional model, J Atmos Sci (USA), 46 (1989) pp NCAR Technical Note TN-475+STR, (NCAR, Boulder, 3077-3107. USA), 2008. 55 Malwer E J, Taubman S J, Brown P D, Iacono M J & Clough 43 Kain J S & Fritsch J M, A one-dimensional entraining/ S A, Radiative transfer for inhomogeneous atmosphere: detraining plume model and its application in convective RRTM, a validated correlated-k model for the long wave, parameterization, J Atmos Sci (USA), 47 (1990) pp J Geophys Res (USA), 102 (D14) (1997) pp 16, 663-16, 682, 2784-2802. doi: 10.1029/97JD00237. 44 Kain J S & Fritsch J M, Convective parameterization 56 Kanamitsu M, Description of the NMC global data for mesoscale models: The Kain-Fritsch scheme, The assimilation and forecast system, Weather Forecast (USA), representation of cumulus convection in numerical models, 4 (1989) pp 335-342. Am Meteorol Soc Meteorol Monogr (USA), 46 (1993) pp 57 Pattanaik D R & Rama Rao Y V, Track prediction of very 165-170. severe cyclone Nargis using high resolution weather research 45 Janjić Z I, The step-mountain eta coordinate model: Further forecasting (WRF) model, J Earth Syst Sci (India), 118 (4) developments of the convection, viscous sub layer and (2009) pp 309-329. turbulence closure schemes, Mon Weather Rev (USA), 122 58 Willoughby H E, The dynamics of the tropical cyclone core, (1994) pp 927-945. Aust Meteorol Mag (Australia), 36 (1988) pp 183-191.