2476 JOURNAL OF APPLIED AND VOLUME 52

Real-Time Track Prediction of Tropical over the North Indian Ocean Using the ARW Model

KRISHNA K. OSURI AND U. C. MOHANTY School of Earth, Ocean and , Indian Institute of Technology Bhubaneswar, Odisha, India

A. ROUTRAY National Centre for Medium Range Forecasting, Noida, India

M. MOHAPATRA India Meteorological Department, New Delhi, India

DEV NIYOGI Purdue University, West Lafayette, Indiana

(Manuscript received 18 November 2012, in final form 9 July 2013)

ABSTRACT

The performance of the Advanced Research version of the Weather Research and Forecasting (ARW) model in real-time prediction of tropical cyclones (TCs) over the north Indian Ocean (NIO) at 27-km res- olution is evaluated on the basis of 100 forecasts for 17 TCs during 2007–11. The analyses are carried out with respect to 1) basins of formation, 2) straight-moving and recurving TCs, 3) TC intensity at model initialization, and 4) of occurrence. The impact of high resolution (18 and 9 km) on TC prediction is also studied. Model results at 27-km resolution indicate that the mean track forecast errors (skill with reference to per- sistence track) over the NIO were found to vary from 113 to 375 km (7%–51%) for a 12–72-h forecast. The model showed a right/eastward and slow in TC movement. The model is more skillful in track prediction when initialized at the intensity of severe or greater than at the intensity stage of cyclone or lower. The model is more efficient in predicting landfall location than landfall time. The higher-resolution (18 and 9 km) predictions yield an improvement in mean track error for the NIO Basin by about 4%–10% and 8%–24%, respectively. The 9-km predictions were found to be more accurate for recurving TC track pre- dictions by ;13%–28% and 5%–15% when compared with the 27- and 18-km runs, respectively. The 9-km runs improve the intensity prediction by 15%–40% over the 18-km predictions. This study highlights the capabilities of the operational ARW model over the Indian region and the continued need for operational forecasts from high-resolution models.

1. Introduction the premonsoon season (April–early June). Because of growing population and development, damage from The north Indian Ocean (NIO), including the Bay of landfalling TCs over the NIO has shown a steady in- Bengal (BoB) and the Arabian (AS), experiences crease. For disaster warnings and mitigation efforts, two tropical-cyclone (TC) , with a primary maxi- forecasting TC tracks is a critical component; track mum in TC frequency during the postmonsoon season forecast errors over the NIO are high relative to those (October–December) and a secondary maximum during over the Atlantic and Pacific Oceans (Mohapatra et al. 2013b). It is evident that synoptic and statistical methods have limitations in predicting track and intensity beyond Corresponding author address: Prof. U. C. Mohanty, School of Earth, Ocean and Climate Sciences, Indian Institute of Technology 24 h over the NIO (Mohanty and Gupta 1997; Gupta Bhubaneswar, Satya Nagar, Bhubaneswar–751 007, India. 2006). With advancements in resolution, physics, and E-mail: [email protected] initializations, global and mesoscale models have the

DOI: 10.1175/JAMC-D-12-0313.1

Ó 2013 American Meteorological Society Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2477 potential to provide forecast guidance for TC genesis, baseline for mesoscale modeling capability over this re- intensity, and movement for a period of 72 h for disaster gion. A semioperational effort of this kind may provide mitigation and warning systems. a basis for improving track and intensity prediction skills The Advanced Research version of the Weather Re- in this region. This semioperational effort with ARW search and Forecasting system (ARW), developed at the shall become the benchmark for evaluating any future National Center for Atmospheric Research (NCAR), is developments over the NIO region for TC predictions. one of the two distinct dynamical cores of the Weather Research and Forecasting (WRF) model. The other 2. Modeling system and configuration core version, the Nonhydrostatic Mesoscale Model (NMM), was developed at the Environmental Modeling The model configuration including domain and physi- Center of the National Centers for Environmental cal parameterization schemes follows the sensitivity study Prediction (NCEP). Over the Indian monsoon region, of Osuri et al. (2012a). The initial and lateral boundary and indeed globally, the ARW model is being widely conditions for the ARW model are obtained from the used for the simulation of a variety of weather events, analysis and forecast fields of the NCEP GFS. The lateral such as heavy rainfall (Niyogi et al. 2006; Routray et al. boundary conditions are updated in 6-h intervals with 2010; Dodla and Ratna 2010; Hong and Lee 2009) and a fixed sea surface throughout the model TCs (Osuri et al. 2012a; Pattanaik and Rama Rao 2009; integration, with no regional used in this Davis et al. 2008). The ARW model has been used for study. The land surface boundary conditions are taken real-time TC forecasting since 2007. The application of from the U.S. Geological Survey with a horizontal grid such mesoscale models for TC forecasting over the NIO spacing of 10 min. A single experimental domain is fixed is a recent development. Osuri et al. (2012a) demon- between 778 and 1028Eandbetween38 and 288Noverthe strated the promising ability of the ARW model for real- BoB, and between 488 and 788E and between 58 and 308N time prediction of TC track and intensity over the NIO over the AS, with 51 vertical layers and 27-km horizontal during the 2008 season. This study also demonstrated grid resolution with an Arakawa C grid. This study used that the performance of the ARW model was reasonably a time step of 90 s with the Kain–Fritsch cumulus pa- good in comparison with other global models. The current rameterization scheme, the slab model for the land sur- TC operational model in use at the India Meteorological face representation, the WRF single-moment 3-class Department (IMD), New Delhi, is the Quasi-Lagrangian microphysics scheme, and the Yonsei University plane- Model (QLM), apart from global and regional models tary boundary layer scheme. Details of the model equa- like the (GFS) and the ARW. tions, physics, and dynamics are available in Skamarock The QLM has shown little skill over the NIO Basin be- et al. (2005; see also online at http://wrf-model.org). cause it runs at 40-km horizontal resolution with 16 ver- tical levels (see online at http://www.imd.gov.in/section/ 3. Method nhac/dynamic/RSMC-2011.pdf; Roy Bhowmik and Kotal 2010; Kotal and Roy Bhowmik 2011). As compared with On the basis of criteria adopted by the IMD, a TC is other operational centers, NCEP in the United States designated as a cyclonic if the associated maxi- uses the high-resolution hurricane WRF (HWRF) mum sustained surface (MSW) is 34–47 kt (1 kt ’ 2 model, which is based on the NMM dynamical core of 0.51 m s 1), as a severe cyclonic storm if it has MSW of WRF; operating at -permitting (3 km) resolution, 48–63 kt, as a very severe cyclonic storm if it has MSW of it has achieved significant skill in hurricane track and 64–119 kt, and as a supercyclonic storm if it has MSW intensity forecasts (Gopalakrishnan et al. 2006, 2012; of 120 kt or more (IMD 2011). It is considered a de- Tallapragada et al. 2008) over the Atlantic and Pacific pression if the MSW is 17–33 kt. Seventeen TCs during Ocean basins. The recent case studies by Mohanty et al. 2007–11 over the NIO are considered in this study. Of (2013) and Pattanayak et al. (2012) reported better re- these 17 TCs, 13 were formed over the BoB and 4 formed sults using the HWRF model for BoB cyclones as well. over the AS. A real-time TC forecast up to 120 h was The prime objective of this study is to evaluate the performed 2 times per day (0000 and 1200 UTC) performance of the ARW model for track predictions throughout the TC life with effects from the genesis over the NIO on the basis of 100 forecast cases involving stage (formation of depression). Table 1 shows the period 17 TCs that occurred over the NIO between 2007 and of simulations as well as the time of landfall for each 2011. Note that this 5-yr period considered is comparable TC. The numerical simulations result in 100 prediction to the period typically used in other operating centers ‘‘cases.’’ The details of synoptic situations and best- such as NCEP. This is an effort to document prediction track data of the 17 TCs were obtained from the IMD skills over the NIO, because there is no well-established Regional Specialized Meteorological Centre (RSMC)

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2478 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

TABLE 1. Details of the model simulations and observed landfall time of each TC. For intensity, CS 5 cyclonic , SCS 5 severe cyclonic storms, VSCS 5 very severe cyclonic storms, and SuCS 5 supercyclonic storms. For nature, SM 5 straight-moving TCs and RC 5 recurving TCs.

Different TC name No. of forecasts TC categories (intensity) Simulation period in 12-h intervals Obs landfall (LF) (nature) Arabian Gonu (SuCS) 0000 UTC 2–1200 UTC 5 Jun 2007 0300 UTC 6 Jun (over Oman) 8 (SM) Sea cyclones Yemyin (CS) 0000 and 1200 UTC 25 Jun 2007 0300 UTC 26 Jun 2 (SM) Phyan (CS) 1200 UTC 9–0000 UTC 11 Nov 2009 Between 1000 and 4 (RC) 1100 UTC 11 Nov Phet (VSCS) 1200 UTC 31 May–0000 UTC 6 Jun 2010 0000 UTC 4 Jun (LF 1), 12 (RC) 1200 UTC 6 Jun (LF 2) Akash (CS) 1200 UTC 13–1200 UTC 14 May 2007 0000 UTC 15 May 3 (SM) cyclones Sidr (VSCS) 1200 UTC 11–0000 UTC 15 Nov 2007 1500 UTC 15 Nov 8 (RC) Nargis (VSCS) 1200 UTC 27 April–0000 UTC 2 May 2008 1200 UTC 2 May 10 (RC) Rashmi (CS) 0000 UTC 25–1200 UTC 26 Oct 2008 0000 UTC 27 Oct 4 (SM) KhaiMuk (CS) 1200 UTC 13–1200 UTC 15 Nov 2008 0000 UTC 16 Nov 5 (SM) Nisha (CS) 1200 UTC 25–26 Nov 2008 0000 UTC 27 Nov 3 (SM) Bijli (CS) 1200 UTC 14–0000 UTC 17 Apr 2009 1500 UTC 17 Apr 6 (RC) Aila (SCS) 1200 UTC 23–0000 UTC 25 May 2009 0900 UTC 25 May 4 (SM) Ward (CS) 1200 UTC 10–1200 UTC 13 Dec 2009 0900 UTC 14 Dec 7 (RC) Laila (VSCS) 1200 UTC 17–19 May 2010 1200 UTC 20 May 5 (RC) Giri (VSCS) 1200 UTC 20–0000 UTC 22 Oct 2010 1400 UTC 22 Oct 4 (SM) Jal (VSCS) 0000 UTC 4–7 Nov 2010 1600 UTC 7 Nov 7 (SM) Thane (VSCS) 0000 UTC 26–1200 UTC 29 Dec 2011 0000 UTC 30 Dec 8 (SM) Total no. of forecast cases 100 reports from 2008, 2009, 2010, 2011, and 2012 (http://www. 700 hPa. Many global centers, including NCEP, determine imd.gov.in/section/nhac/dynamic/rsmc1.htm). TC positions by following the same procedure as that The model predictions were evaluated on the basis of documented in Marchok (2002). the calculation of different types of errors such as direct The DPE is the great-circle distance between the position error (DPE), longitudinal (zonal, or DX), and model forecast position and the observed TC position at latitudinal (meridional, or DY) errors. These errors a particular time. The systematic error DX (or DY) is were calculated with respect to the best-track estimates defined to be positive if the forecast position is right (or of the IMD. The IMD best-track estimates are mainly ahead) of the best-track position. The systematic errors based on satellite data and expert interpretation when can provide some information about the directionality the TC is in the midsea region. The satellite estimates are of the errors in the zonal or meridional directions. In the sometimes tweaked on the basis of the available mean sea case of northward- and westward-moving TCs, however, level (MSLP) and 10-m wind observations from there are additional difficulties in interpreting the sys- ships and buoys and of scatterometer from the tematic errors on the basis of DX and DY errors. To satellite in the region. When the TC approaches the coast determine whether the forecast position is left or right and is within range, the best-track estimates are and slower or faster, the cross-track (CT) and along- based on the radar observations followed by satellite track (AT) errors can be calculated relative to the ob- data. When the system lies close to the coast, then the served track. The calculation method of these errors is IMD gives higher weight to operational shown in Fig. 1. The full details on calculation of AT synoptic observations followed by radar and satellite. and CT errors can be found in Fiorino et al. (1993). The On a synoptic level, the location of MSLP values and the perpendicular distance from the model track forecast center of the 10-m wind circulations are considered to position to the observed track is the CT error. It is determine the best track of a TC. positive (negative) when the forecast position lies to the The TC positions have been identified by using an right (left) of the observed track. CT errors give the automated tracking scheme that is based on Marchok spatial spread of the simulated tracks. At a given fore- (2002). The center of the TC is determined on the basis cast time, AT errors are defined as the distance from the of the spatial distribution of seven parameters including observed position to the position along the observed the minimum MSLP and relative maxima, geo- track where the model forecast track meets perpendic- potential height minima, and minima at 850 and ular to the observed track. If the perpendicularity meets

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2479

FIG. 1. Schematic of the method used for calculating track-error metrics. the track ahead of (behind) the observed position, the (WMO 2009) because it is dependent on the climato- value of AT is positive (negative) and indicates faster logical behavior of TC tracks for a particular basin. (slower) movement of the simulated TC relative to ac- Therefore, the ideal situation would be to adopt a uni- tual movement of the TC. The information on CT and versal reference model that serves as a sole reference so AT errors is very helpful for TC disaster management, that the performance of the ARW model in different because the CT error will help in determining the area of ocean basins can be compared. In the absence of a better evacuation needed in case of a landfalling TC and the reference model, persistence can be considered as a ref- AT error will help in determining the time of evacua- erence model. In this study, the ‘‘persistence’’ method has tion. A relatively lower CT error is desirable by disaster been used to obtain the reference track. Although the managers because it will lower the cost of evacuation. ARW model has been run to provide forecast output Along with these errors, forecast-skill metrics have every 3 h, only the 12-h track forecasts up to 72 h were also been computed. In the analysis, a distinction is verified to compare with the 12-h track forecasts obtained made between accuracy and skill. Accuracy of the model by the persistence method. The gain in skill is given as is the actual error between the forecast track and the follows: observed best-track locations. The skill of the model is persistence DPE 2 model DPE the relative performance of the model with respect to gain in skill 5 3 100. some reference technique. The reference model can persistence DPE be persistence, which is the simplest forecast model, or a model that is a combination of climatological values Positive (negative) value represents gain (loss) in plus persistence (referred to as ‘‘CLIPER’’; Pike and model skill. Apart from this gain in skill, the confidence Neumann 1987). The persistence forecast is calculated intervals (CI) for mean errors are also calculated. on the basis of the principle that the future track (speed and direction of movement) of a TC for the next 72 h 4. Results and discussion at 12-h intervals will be same as the track (speed and a. Initial vortex position errors direction of movement) followed by the TC during the past 12 h. In a CLIPER model, different or equal weight Figure 2 shows the ‘‘observed’’ best tracks of all 17 is given to climatological and persistence methods. The TCs over the Arabian Sea (Fig. 2a) and Bay of Bengal CLIPER model can differ according to TC basins (Fig. 2b) during 2007–11 as provided by the IMD. Figure 3

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2480 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 2. The ‘‘observed’’ best tracks of tropical cyclones (from IMD) over the (a) AS and (b) BoB during 2007–11. provides the mean initial vortex position error of each TC official agency for determination of best track for TCs and the ensemble mean of 100 cases over the NIO as based over the NIO. There is approximately a 30–70-km differ- on 27-km horizontal resolution. The mean initial vortex ence between the IMD and JTWC best-track positions position errors vary from 30 to 94 km for 17 TCs. The av- over the NIO (Falguni et al. 2004). Similar interagency erage error is ;61 km with 99% CI of 4 km over the NIO differences also exist for other ocean basins. Ahn et al. Basin. The mean initial position errors are ;58 and 68 km (2002) demonstrated that, when comparing the de- over the BoB and AS, respectively. The 99% CIs of the parture between JTWC and RSMC, Tokyo is greater mean initial vortex position error for the BoB and the AS (;50 km) when the system is at the depression stage and are 6 and 10 km, respectively. The initial position error may less (30 km) when the system is at the severe-intensity be due to poor data near and around the vortex over the stage. Thus the difference between the estimates of TC NIO and also to the coarser resolution of the global positions by JTWC and IMD might have contributed to analyses. Further, the NCEP considers Joint this higher initial error also. Warning Center (JTWC) TC positions for calculation of b. Mean track forecast errors over the NIO positional errors and TC vortex relocation in the GFS model. IMD best-track positions were used for calcula- Figure 4 shows the model-simulated tracks with 27-km tion of ARW positional errors because the IMD is the horizontal resolution at different initial conditions along

FIG. 3. Mean initial vortex position error (km) for each TC and for the overall mean, as based on 27-km resolution.

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2481

FIG. 4. Model-predicted tracks as based on 27-km horizontal resolution (in gray) of (a) Akash, (b) Gonu, (c) Yemyin, (d) Sidr, (e) Nargis, (f) Rashmi, (g) KhaiMuk, (h) Nisha, (i) Bijli, (j) Aila, (k) Phyan, (l) Ward, (m) Laila, (n) Phet, (o) Giri, (p) Jal, and (q) Thane at different initial times (shown in Table 1) along with the IMD best track (black line with bigger cyclone symbol). The numbers in the legend represent the lead forecast hours. The dotted tracks in (n) represent tracks with 132-h (short dashes) and 144-h (long dashes) lead times. The small cyclone symbols over gray color tracks represent TC position at each 24-h interval.

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2482 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52 with the IMD best track for all TCs for 2007–11. The TABLE 2. Mean DX, DY, CT, and AT errors (km) of predicted model could realistically predict the tracks for these TCs tracks from the ARW model of 27-km horizontal resolution and with most of the initial conditions. The exception is TC from the persistence method for up to 72-h forecast length for NIO TCs. The 99% CI is given in parentheses. Gonu, where the model predicted the initial movement toward Oman and the first landfall over Oman with Forecast ARW-based errors Persistence-based errors some southward displacement. The model could not length DY DX CT AT DY DX CT AT predict the second landfall of Gonu over Iran. Hence, 12 245 20 22 2106 223 227 22 2150 the lead time of the model-predicted tracks is repre- (24) (21) (24) (30) (25) (26) (32) (41) sented with respect to the first landfall (0000 UTC 6 June 24 271 33 10 2114 259 275 8 2138 2007) in Fig. 4b. In a similar way, the model could not (33) (31) (32) (35) (46) (49) (44) (53) 2 2 2 2 2 2 simulate the realistic movement in some cases of Sidr 36 86 41 27 145 73 142 11 148 (44) (43) (43) (47) (72) (80) (72) (84) and Laila. The TC Sidr moved northward under the 48 294 44 51 2150 2100 2213 83 2182 influence of the upper-tropospheric trough to the west of (71) (69) (70) (68) (116) (118) (122) (119) the system and the upper-tropospheric located to 60 279 69 29 2132 2147 2331 77 2195 the east. The deep-layer mean wind (850–200 hPa) ob- (95) (103) (86) (111) (148) (155) (172) (173) 2 2 2 2 2 tained from the ARW model was analyzed for TC Sidr 72 65 67 56 146 163 457 0 194 (125) (158) (146) (133) (192) (223) (241) (215) for the initial conditions at 1200 UTC 11 November– 1200 UTC 13 November 2007, and it could not predict the steering circulations. With better representation of the deep-layer mean wind from 0000 UTC 14 November, the 12-h forecast, leading to a higher standard devia- however, the model could simulate a realistic track that tion. The CT error becomes positive for higher forecast was close to what was observed. This appears to be due lengths because of rightward movement of the simu- to the fact that the ARW model could not pick up the lated TCs by the ARW model, however. Similar to the correct boundary conditions because the domain was analyses of DY, the analyses of mean AT errors of the limited to 788–1038E but the system was actually near ARW model reveal that track positions are generally 898–908E longitude. It is obvious that the domain size behind the observed tracks as seen by the result that AT plays an important role in providing better boundary errors are negative for all forecast lengths. In comparing conditions (Landman et al. 2005) along with steering the magnitudes of AT and CT errors of the ARW model, forcing. The TC Nargis moved eastward for about 3 days it is seen that the AT errors are larger for all forecast (from 29 April to 1200 UTC 1 May 2008) mainly because lengths. That is, the track error is elliptical in nature with it was under the joint influence of an upper-level anti- its major axis along the track. In other words, the spread cyclone lying to the southeast and midlatitude upper- of the track relative to the observed track is less. The 99% tropospheric westerlies (http://www.imd.gov.in/section/ CI of mean error from the ARW forecasts at all forecast nhac/dynamic/RSMC-2009.pdf). The ARW model could intervals is smaller when compared with those of per- resolve and capture the presence of the upper-tropospheric sistence track errors suggesting that the ARW model anticyclonic circulation to the south of the system as forecasts are in general more consistent for all forecast observed and the low-level (850 hPa) vorticity intervals. reasonably well (not shown). The mean DPE (km) of TC forecast positions as The mean error statistics (km) of the model at 27-km based on 27-km horizontal resolution is shown in Fig. 5a resolution, such as DX, DY, CT, and AT errors, as based over the NIO. The mean 12-, 24-, 48-, and 72-h forecast on 100 cases, for the NIO systems are given in Table 2. errors of the ARW (persistence) tracks are 113 (121), The mean DX of the ARW model is positive for all 140 (230), 248 (500), and 375 (770) km, respectively. forecast periods. The average ARW forecast track thus The 99% CI of mean DPE of the ARW model for the lies to the right of the best-track position for all of the 12-, 24-, 48-, and 72-h forecast is approximately 18, 23, simulations. In other words, the ARW model shows a bias 49, and 94, respectively. There is a gain in forecast skill to predict right or eastward movement of TCs over the that varies between 7% and 51% for the 12- to 72-h study domain. In contrast to the ARW model, persistence- forecast lengths (Fig. 5a) over the persistence track. The based TC forecast positions to the left (negative DX) of gain in skill is generally greater with an increase in the the observed positions have a left or westward bias. The forecast period. mean CT error of the ARW model is slightly negative at c. Mean track forecast errors over the BoB the 12-h forecast, and it is found that the slightly negative value of CT is mainly due to higher negative values in Table 3 provides the mean errors (similar to Table 2) a few cases [e.g., TCs Sidr (Fig. 4d) and Bijli (Fig. 4i)] at for 13 BoB TCs as based on 75 forecast cases. The mean

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2483

FIG. 5. Mean DPE (km; shaded bars) and gain in skill (%; line) of the model with 27-km resolution for the TCs over the (a) NIO, (b) BoB, and (c) AS. Open bars give mean DPE of persistence tracks.

DX and DY values are also similar to those for all TCs negative at all forecast times. When comparing the CT taken together over the NIO. Thus, similar to the results and AT errors of the model, it is seen that the shape of the found for the NIO, model-based TC positions are biased errors is again elliptical with the major axis along the to the right and behind observed TC positions for all track for all forecast lengths for TCs over the BoB, similar forecast lengths over the BoB. This is further noted in to those over the NIO. The mean DPE of the ARW CT and AT errors (Table 3). The ARW-based CT error model for the BoB TCs (Fig. 5b) at 12-, 24-, 48-, and 72-h values are positive at all forecast hours except for the forecast lengths is 108, 137, 243, and 353 km, respectively. 12-h forecast over the NIO, and the AT errors are largely The 99% CI for the mean errors at the same forecast

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2484 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

TABLE 3. As in Table 2, but for BoB TCs. TABLE 4. As in Table 2, but for Arabian Sea TCs.

Forecast ARW-based errors Persistence-based errors Forecast ARW-based errors Persistence-based errors length DY DX CT AT DY DX CT AT length DY DX CT AT DY DX CT AT 12 237 16 25 2106 219 227 9 2154 12 268 30 7 2104 238 229 67 2134 (29) (24) (28) (36) (29) (31) (39) (49) (48) (44) (52) (54) (54) (49) (55) (75) 24 268 28 13 2103 253 273 1 2139 24 280 50 1 2145 277 280 31 2132 (40) (36) (36) (42) (56) (59) (52) (63) (59) (70) (73) (65) (86) (100) (88) (109) 36 278 40 16 2137 258 2129 220 2125 36 2108 46 58 2168 2114 2179 88 2213 (54) (48) (51) (55) (85) (91) (90) (95) (84) (110) (86) (96) (155) (183) (119) (198) 48 287 49 52 2143 291 2207 74 2123 48 2111 31 49 2166 2123 2229 175 2330 (86) (76) (83) (85) (143) (138) (165) (137) (135) (168) (148) (124) (221) (252) (138) (240) 60 270 76 43 2164 2167 2364 73 2127 60 297 53 0 263 2108 2264 85 240 (124) (115) (115) (128) (182) (174) (236) (225) (170) (253) (144) (256) (315) (376) (273) (325) 72 233 81 60 2130 2186 2533 117 2153 72 2120 44 250 2174 2128 2339 2182 323 (180) (174) (201) (153) (226) (242) (336) (254) (174) (370) (244) (300) (425) (513) (347) (393)

lengths is 22, 22, 57, and 124, respectively. The ARW in skill (26%) for AS systems at the 12-h forecast, unlike model shows a similar gain in skill for BoB TCs as that the NIO and BoB TCs. The 99% CI of mean DPEs for TCs over NIO for all forecast periods. For individual corresponding to the ARW model is small in compari- TCs, the error is higher for recurving TCs Bijli and son with that of persistence track errors, suggesting that Ward. This may be due to the fact that the initial vortex the ARW model is more consistent, similar to that over position error is more for TC Bijli (94 km, with a lower the BoB. intensity of 996 hPa) and TC Ward (91 km, with a lower e. Mean track forecast errors for straight movers and intensity of 996 hPa). recurving TCs d. Mean track forecast errors over the AS All of the TCs over the NIO are classified into two From Table 4, the DX errors are positive for all TC categories on the basis of characteristic movement (irre- forecast lengths over the AS. It shows that the ARW spective of the location of genesis) as 1) straight-moving trackpositionsarealsobiasedtowardtherightoverthe or 2) recurving TCs. Sidr, Nargis, Bijli, Phyan, Ward, AS Basin. The DY errors are negative for all forecast Laila, and Phet come under the category of recurving TCs lengths, suggesting that the model track positions lie (contributing to 52 forecast cases) and the remaining TCs behind the actual positions as also observed over the (Akash, Gonu, Yemyin, Rashmi, KhaiMuk, Nisha, Aila, NIO and BoB. The CT errors of the ARW model for Giri, Jal, and Thane) are classified as straight-moving the AS TCs are positive up to 60 h and become negative TCs (contributing to 48 cases). The mean track errors afterward, which suggests that the model positions and associated gain in skill of the model at 27-km reso- are to the right of the observed track up to 60 h and shift lution for the above two categories of TCs at different left of the observed track, similar to that found in the forecast hours are shown in Figs. 6a and 6b, respectively. persistence method. The CT and AT error values from The 99% CI for the mean error is also shown in Fig. 6a. the ARW forecast are also smaller when compared It is clear that the mean errors and the 99% CI associ- with the persistence errors. The AT errors are con- ated with straight-moving TCs are significantly less than siderably higher in comparison with CT errors for all those of recurving TCs. The mean track errors vary from forecast lengths, demonstrating that the error is ellipti- 55 to 250 km from 0- to 72-h forecast lengths for straight- cal in shape with its major axis along the track, similar to moving TCs and from 74 to 440 km for recurving TCs. that over the NIO and BoB. Figure 5c shows the mean The 99% CI of mean error at initial time, 12-, 24-, 48-, DPE of the ARW model on the basis of 25 cases for the and 72-h forecast lengths are 11, 23, 33, 61, and 101 km 12-, 24-, 48-, and 72-h forecasts at 125, 142, 270, and for straight movers and 16, 27, 36, 86, and 186 km for 413 km, with 99% CI of 33, 49, 99, and 166 km, re- recurving TCs, respectively. The gain in skill of the spectively. For the same forecast lengths, the persistence model in simulating these two categories is calculated tracks are 118, 227, 503, and 831 km, respectively. This with reference to persistence tracks. The skill of the result shows a positive gain in ARW skill of approxi- model is increasingly positive for straight-moving TCs mately 37%, 46%, and 50% at 24-, 48-, and 72-h forecast for all forecast lengths (30%–67% for 12–72-h forecast). time, respectively. The model has a negative gain (loss) The gain in skill is higher in the case of straight-moving

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2485

FIG. 6. (a) Mean DPEs (km) and (b) gain in skill (%) of the model with 27-km resolution for straight movers and recurving TCs over the NIO.

TCs by approximately 20%–30% when compared with From Table 5, the analyses of mean zonal and me- recurving TCs for different forecast lengths. Even the ridional errors associated with the straight-moving and skill is negative for the 12-h forecast period for recurving recurving TCs over the NIO reveal that the track po- TCs. Ramarao et al. (2006) also demonstrated that the sitions are to the right of (behind) the observed TC forecast errors increase for recurving cyclones and that position given that DX (DY) is positive (negative) for they are difficult to predict on the basis of the QLM straight movers as well as for recurving TCs. The model for NIO TCs. Mohapatra et al. (2013b) demon- analyses of CT and AT errors support the elucidation strated that the straight movers along the southern pe- for straight-moving TCs, however, whereas for re- riphery of a subtropical ridge have higher predictability curving TCs the CT errors are positive up to the 36-h than recurving systems, which have a low predictability. forecast only. After that, the CT errors become nega- This analysis showed that numerical models such as the tive up to 72 h, which implies that the model positions ARW still have difficulty in predicting recurving TCs, are to the right of the observed position up to 36 h and and these difficulties can be mainly attributed to the shift to left of the observed position from 36 h onward. following factors. Large-scale steering can play a prin- In comparing the CT and AT errors, it is seen that the cipal role in deciding storm track, especially in the case 99% CI is considerably less for straight-moving TCs of recurving TCs. Sometimes the ARW model may not when compared with recurving TCs. Like NIO TCs, the be able to capture the large-scale steering ridge and errorisellipticalinshape,withitsmajoraxisalongthe westerly trough in the upper , leading to track, for 12–72-h forecast lengths for these two cate- recurvature of the system because of the limited domain gories of TCs. (778–1028E). It is obvious that mesoscale models are mainly driven by lateral boundary conditions available TABLE 5. Mean DX, DY, CT, and AT errors of the ARW model from global analysis/forecast products. In this case, the for up to 72-h forecast lengths for straight-moving and recurving model could not pick up the correct boundary conditions TCs as based on 27-km horizontal resolution over the NIO. The 99% CI is given in parentheses. because of either the limited domain size or the lack of representation of such steering features in the global Forecast Straight-moving TCs Recurving TCs model initial and/or lateral boundary conditions, result- length DY DX CT AT DY DX CT AT ing in poor predictability of recurving systems. Landman 12 234 35 5 277 254 6 3 2131 et al. (2005) also demonstrated the role of domain size on (35) (22) (29) (31) (35) (35) (38) (49) simulation of TCs as vortices using a regional climate 24 253 62 4 2129 286 8 15 2157 model. Apart from this, model initialization is also an- (44) (32) (36) (40) (46) (48) (50) (55) other important factor. Osuri et al. (2012b) showed that 36 268 76 14 288 2100 13 37 2191 assimilation of satellite-derived sea surface winds im- (62) (36) (47) (54) (65) (73) (70) (70) 48 263 85 19 2130 2115 16 272 2163 proved the forecast of recurving TC Nargis. Therefore, (99) (53) (88) (63) (100) (111) (103) (108) high-resolution assimilation of quality-controlled ob- 60 244 125 1 2143 2100 34 247 2168 servations into initial conditions as well as coupling with (140) (77) (98) (119) (133) (162) (131) (170) ocean models for better air–sea interaction will help to 72 27 126 24 2136 2103 36 273 2152 improve predictions of recurving systems. (204) (120) (143) (182) (162) (234) (213) (185)

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2486 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 7. (a) As in Fig. 6, but with respect to cyclonic storms (maximum sustained wind speed of 34–47 kt) and severe cyclonic storms (maximum sustained wind speed is 48 kt or more). f. Mean track forecast errors relative to intensity gain in skill in track prediction when compared with of TCs persistence track when initialized at the SCS stage (varies between 22% and 61%) rather than the CS stage (6%– The performance of the ARW model for track pre- 43%) for all forecast lengths (Fig. 7b). The 99% CI of the diction is further extended by considering TC forecasts mean CT and AT errors is smaller for forecasts issued at initialized at different intensity stages such as de- the SCS stage in comparison with those at the CS stage. pression, cyclonic storm (CS), and severe cyclonic storm That is, the stronger the storm is, the lower is the track (SCS) stages. The analysis is carried out with respect to forecast error range. These results are comparable the stage of intensity at the time of model initialization. over the western Pacific Ocean using the ARW model Of 17 TCs considered in the study, 8 TCs reached CS (Ryerson et al. 2007). Mohapatra et al. (2013b) also intensity and the remaining 9 reached SCS intensity demonstrated the above fact by verifying the opera- (Table 1). Accordingly, there are 38, 28, and 34 forecast tional TC forecasts of IMD for the CS and SCS cases issued with depression, CS, and SCS stages at the categories. time of model initialization, respectively. It is observed that there is no significant difference in track errors up g. Mean track forecast errors relative to TC season to 36 h, when predictions are conducted from the de- of occurrence pression and CS stages. There is an improvement of about 20–25 km in 48–72-h forecast times, however, when pre- The model performance is also evaluated for the two dictions are carried out from the CS stage in comparison cyclone seasons (pre- and postmonsoon) on the basis of with those at the depression stage (not shown). A con- 48 and 50 forecast cases over the NIO Basin (Fig. 8). siderable difference in track forecasts errors is noticed Overall, the mean initial vortex position error is less for with respect to initializations at the CS and SCS stages postmonsoon TCs (52 km with a 99% CI of 11 km) when (Fig. 7a). The 99% CI is also shown in Fig. 7a. The mean compared with the premonsoon TCs (65 km with a 99% initial vortex position error (60 km at the CS stage and CI of 18 km). Note that in this comparison TC Yemyin is 51 km at the SCS stage with 99% CI of 23 and 10 km, not included in either of the seasons because it formed respectively) and the track forecast errors for all forecast during the active monsoon period (25–26 June 2007). lengths are smaller in the case of forecasts initialized at The model shows fewer errors with a minimum 99% CI the SCS stages. The 99% CI is also significantly less for for postmonsoon TCs as compared with premonsoon the SCS initializations (12 km at 12-h forecast and 90 km TCs (Fig. 8a). The 99% CI at 12-, 24-, 48-, and 72-h for the 72-h forecast) when compared with that of CS forecast lengths is approximately 27, 39, 89, and 174 km initializations (47 and 187 km, respectively). This result and 22, 31, 71, and 123 km, respectively, for pre- and can be due to better representation of the TC vortex at postmonsoon TCs. The gain in skill is greater for post- the SCS stage in terms of horizontal and vertical structure monsoon TCs at all forecast lengths up to 72 h (Fig. 8b) as a result of stronger intensity. The model performance over persistence tracks. The relatively higher track er- when initialized at the SCS stage (i.e., for stronger rors for premonsoon TCs may be due to the fact that storms) is consistent with findings over the Atlantic basin most of the premonsoon systems are recurving systems. (Gopalakrishnan et al. 2012). The model shows a better Yang et al. (2011) studied the seasonal variability of

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2487

FIG. 8. As in Fig. 6, but for premonsoon and postmonsoon TCs. a number of TCs between 1980 and 2009 and found that 5. High-resolution retrospective forecast of TCs westward- and northward-moving TCs (that come under the straight-mover category) occur more often (over To further investigate the potential improvement for 85%) during September–January (which covers the future seasons, the study also investigated the role of postmonsoon season), with a peak in October. Recurving horizontal resolution on the model forecast by running TCs are fewer in number during the postmonsoon sea- all of the cases listed in Table 1 again but with 18- and son, however, occurring more frequently during the 9-km resolution. Except for the higher resolution, the months of April and May (premonsoon season). From configuration of the ARW model such as domain, model 2003–11 TCs, Mohapatra et al. (2013b) also demon- physics, and initial and boundary conditions was the strated that 60% of the total TCs in the premonsoon same as that of the real-time setup. season are recurving TCs. In general, a deep trough in a. TC track prediction at 18-km resolution the mid- and upper-tropospheric westerlies is predom- inant during the premonsoon season over the Indian Relative to 27-km resolution, the 18-km horizontal region (Yang et al. 2011). If a system lies to the right of resolution improved track predictions. Table 6 shows this upper-level westerly trough, it is expected to recurve the mean error statistics for the 18-km runs for all of the to the east or northeast. This situation can be observed subcategories as discussed for the 27-km runs. For the more frequently during the premonsoon season than in NIO TCs, the forecast errors vary from 106 to 359 km for the postmonsoon season. Such recurving TCs would be the 12–72-h forecast. The 99% CI for the mean error expected to have higher errors relative to straight movers varies from 20 to 92 km for the same forecast length. The (Pike and Neumann 1987; Zhang et al. 2013; Mohapatra 18-km runs yield an improvement of 10% when com- et al. 2013b). pared with the 27-km runs for the NIO Basin. For the 75

TABLE 6. Mean DPE (km) of ARW simulations at 18-km resolution for different TC categories at various forecast lead times. The values in the parentheses represent the 99% CI.

Mean DPE (99% CI) for different forecast lead Initial position times TC categories error (km) 12 h 24 h 48 h 72 h NIO 57 (3) 106 (20) 129 (22) 222 (47) 359 (92) BoB 51 (4) 101 (15) 122 (21) 229 (38) 337 (103) AS 62 (6) 111 (27) 127 (42) 233 (85) 394 (141) TCs with characteristic movement Straight-moving TCs 49 (9) 89 (21) 101 (24) 139 (43) 226 (99) Recurving TCs 69 (12) 128 (22) 149 (29) 284 (70) 413 (158) Intensity of TC at the time Cyclonic storm 60 (12) 112 (33) 132 (46) 243 (79) 386 (156) of initialization Severe cyclonic storm 49 (9) 83 (16) 121 (36) 217 (42) 319 (85) Season Premonsoon 64 (11) 109 (21) 149 (29) 241 (71) 371 (151) Postmonsoon 51 (9) 91 (17) 124 (23) 207 (46) 337 (102)

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2488 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 9. Model forecast tracks with 9-km horizontal resolution for (a) Nargis, (b) Phet, (c) Jal, and (d) Thane at different initial times (shown in Table 1) along with the IMD best track. The labels are the same as in Fig. 4.

forecast cases over the BoB, the mean DPEs (99% CI) compared with the 27-km runs. These results demon- are 122 (21), 229 (38), and 337 (103) for the 24-, 48-, and strate that the ARW model would be able to improve 72-h forecast, with an improvement of about 11%, 6%, the track forecast using 18-km resolution. and 5%, respectively, over the 27-km runs. The increased b. TC track prediction at 9-km resolution resolution provided a better track forecast over the AS with an improvement in error of about 11%, 14%, and Figure 9 provides the track forecast positions for four 5% for the same forecast lengths. A similar improve- representative TCs: Nargis, Phet, Jal, and Thane at 6-h ment in track forecast errors was found for other cate- intervals. Table 7 summarizes the mean DPE, CT, and gories such as straight-moving TCs (11% for the 24-h AT errors up to 72 h for all cases. The gain in skill (%) of forecast, 13% for the 48-h forecast, and 10% for the 72-h the 9-km-resolution forecast over the 27- and 18-km- forecast), recurving TCs (9%, 8%, and 6%), CS initial- resolution forecasts is also calculated. The mean DPE of izations (10%, 15%, and 6%), SCS initializations (6%, the 9-km runs is 115, 204, and 329 km for the 24-, 48-, and 7%, and 3%), premonsoon systems (9%, 10%, and 7%), 72-h forecast, respectively. The mean DPE for the NIO and postmonsoon systems (10%, 10%, and 8%). The Basin is reduced considerably by 8%–24% over the directionality errors (CT and AT) of the 18-km runs are 27-km-resolution runs and by 4%–10% over the 18-km- very similar to those of the 27-km predictions, showing resolution runs for the 12–72-h forecast length. In other a consistent eastward bias with slow movement of the words, the impact of the higher-resolution forecast (in TC for all forecast periods. Note that the 99% CI of the this case 9- vs 27-km grid spacing) is such that it provided mean error is significantly less for the 18-km runs when almost 12 h of advantage in terms of the error. That is,

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2489

TABLE 7. Mean DX, DY, CT, and AT errors and mean DPE of the ARW model with 9-km horizontal resolution for different forecast lengths (up to 72 h) for NIO TCs. The corresponding 99% CIs are given in parentheses.

Forecast Gain in skill (%) over length (h) DY DX CT AT DPE 27-km-resolution runs 12 25 (23) 35 (20) 12 (26) 265 (28) 104 (15) 8 24 223 (26) 33 (26) 3 (32) 2103 (34) 115 (22) 18 36 234 (31) 44 (31) 0 (35) 2111 (40) 141 (25) 24 48 224 (46) 64 (64) 50 (57) 286 (58) 204 (44) 18 60 26 (67) 87 (100) 25 (88) 2102 (96) 274 (63) 13 72 277 (85) 53 (155) 34 (126) 2104 (122) 329 (88) 12 the 12-h forecast error of the 27-km resolution is now relocation, high-resolution regional data assimilation (for comparable to the 24-h forecast error in the 9-km- better genesis), sufficient domain size to capture the resolution runs. Other features like DX, DY, CT, and large-scale environmental forcing, ocean coupling for AT are similar to those found in the 27-km runs. The better analysis, improved model physics, and so on. mean AT errors—that is, the errors associated with the c. TC intensity prediction translation speed of the TC—are considerably smaller in the higher-resolution (9 km) runs, however. The mean DX Although this study mainly focuses on track predic- and CT errors became positive for all forecast lengths, tion, a brief description about intensity prediction with unlike in the 27-km-resolution runs. Given the difficulty 18- and 9-km resolution is also given here. Figure 10 gives in predicting recurving systems, the impact of high- the mean intensity error (model predicted 2 observed) resolution simulations for recurving TCs was analyzed. for both 18- and 9-km runs. Positive (negative) values In particular, the benefit of 9-km-simulations can be seen represent overestimation (underestimation) of inten- in the track spread of recurving TCs (e.g., Figs. 4e,n and sity by the model. The average intensity error of ARW- 2 9a,b for Nargis and Phet), and the track errors are sig- WRF for 18-km-resolution runs is 213, 210, and 26ms 1 nificantly reduced when compared with the 27- and 18-km for 24-, 48-, and 72-h forecast lengths, respectively, and 2 runs. The 12-, 24-, 48-, and 72-h track forecast errors that it is 28, 26, and 24ms 1 for 9-km-resolution runs. 2 are based on the 9-km runs are 109, 127, 244, and 389 km, An error that ranges from 28to210 m s 1 is noticed at respectively, with an improvement of 12%–28% over the initial time in the global model analyses. The higher 27-km runs and 4%–14% over the 18-km runs. The 99% errors in intensity forecast during the first 24-h period CI of mean error for the same forecast lengths is also may be due to the spinup problem associated with the lower. The 12-h DPE is still high (;104 km) in high- ARW model and the initial intensity error. The spinup resolution (9 km) runs, however, and it is expected that time can be reduced by adopting finer grid resolu- it can be further reduced by vortex initialization and tions and digital filter initialization (Lynch et al. 1997;

FIG. 10. Mean intensity errors in terms of 10-m maximum winds predicted by 18- and 9-km- resolution runs for the NIO cyclones. The error bars are at 99% CI.

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2490 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

Brousseau et al. 2008). Xiao (2011) also mentioned that, predictions the achievement of which is only possible at because of the model spinup problem, the model ex- higher resolutions. hibits higher errors for MSW at 24 h over both 48 and 72 h. Osuri et al. (2012b) also showed higher TC intensity 6. Summary and conclusions errors during the first 24–30 h of the forecast, which is again a spinup issue. The error in the initial vortex (in The overall performance of the ARW model for track terms of position, strength, and structure) can be reduced and intensity prediction of TCs over the NIO and in- with a vortex initialization and relocation procedure dividual TC basins (BoB and AS) is assessed at different (Gopalakrishnan et al. 2012; Hsiao et al. 2010). Xiao horizontal resolutions (27, 18, and 9 km). This evalua- (2011) demonstrated that the use of (bogus) vortex ini- tion is based on 100 forecast cases of 17 landfalling TCs tialization can not only minimize the spinup period but that occurred during 2007–11. also improve the structure and intensity of the initial TC The ARW model has a good overall capability to vortex. This discussion highlights the need for the im- predict TCs over the NIO Basin as it exhibits a reason- provement in the initial vortex over the NIO Basin either able skill irrespective of the region of formation, nature through better observations or by vortex initialization of movement, intensity, and season of formation. The procedures. Hsiao et al. (2010) showed better intensity mean track forecast errors (skill with reference to per- and track predictions by adopting advanced vortex ini- sistence) over the NIO vary from 113 to 375 km (7%– tialization techniques. The intensity error from the 18-km- 51%) for 12–72-h forecast lengths at 27-km resolution. In resolution runs is greater at all forecast lengths. The 99% a comparison of the track forecast errors with respect to 2 CI for the mean intensity error varies from 6 to 11 m s 1 for TC intensity at initialization, the ARW model performed 6–72-h forecast intervals. At 9-km resolution, however, it better for track prediction if the model is initialized at the 2 varies from 4 to 7 m s 1 at the same forecast intervals. The SCS stage rather than at the CS/depression stage. With ARW model shows an overall underestimation in intensity respect to season of occurrence, the model exhibits less prediction with both 18- and 9-km resolutions when error and 3%–15% more gain in skill for postmonsoon compared with observed intensity. The 9-km resolution TCs than is found for premonsoon TCs. did improve the intensity prediction by 15%–40% over the The ARW model results showed a bias toward pre- 18-km predictions up to the 72-h forecast. It is noticed dicting eastward movement for the NIO TCs for both that the model could not predict the peak intensities of the the BoB and AS. This bias is particularly true for TCs and that the maximum error is observed for systems straight-moving TCs. In the case of recurving TCs, the with an intensity of very severe cyclonic storm (VSCS) or model showed a rightward bias up to the 36-h forecast 2 higher (MSW $ 64 kt) such as Gonu (65 m s 1), Nargis and thereafter a westward bias. The analyses of lat- 2 2 (46 m s 1), and Phet (43 m s 1) because of underestimation. itudinal systematic errors as well as AT errors show that The 72-h maximum intensity predicted by 18- and 9-km- the model forecast positions are biased to the south of 2 resolution runs for Gonu is 40 and 48 m s 1, for Nargis is (behind) the observed positions. The ARW forecasts are 2 2 36 and 42 m s 1, and for Phet is 34 and 45 m s 1. Both in general slower relative to the actual translation speed resolutions could predict the intensity of CS and SCS of the system for all forecast lengths, and they predict well, however. This result implies that the higher the a delayed landfall. The magnitude of CT errors is less in intensity is, the poorer is the peak intensity prediction by comparison with AT errors in the ARW model. Hence the model. In comparison with the operational intensity the ARW model is more accurate in predicting TC forecast errors of IMD over the NIO Basin, however, landfall location than landfall time. the error of the ARW model is higher at 24- and 48-h The higher the resolution of the model is, the better forecast lengths and is comparable at the 72-h forecast are the track and intensity predictions by the model. The length (Mohapatra et al. 2013a). high-resolution predictions (18 and 9 km) reduced the Better prediction at 9-km resolution is due to the fact track errors for NIO TCs and yielded an improvement of that the ARW model could resolve the convection bet- about 4%–10% and 8%–24% in mean DPE over 27-km ter than it could for runs at coarse resolution (27 and runs. The 9-km predictions are found to be better for re- 18 km). These results are also supported by recent curving TC track predictions by about 12%–28% as com- studies that demonstrated that TC predictions in terms pared with 27-km runs and 4%–14% as compared with of track and intensity can be improved by increasing the 18-km predictions. A comparison of 18- and 9-km predic- resolution (Gopalakrishnan et al. 2012; Torn and Davis tions showed that the intensity prediction at 9-km resolution 2012; Han and Pan 2011). Torn and Davis (2012) and was improved by ;15%–40% over the 18-km predictions Han and Pan (2011) reported that proper treatment of up to the 72-h forecast. The model underestimates the convection is an important aspect for realistic TC track peak intensities of systems of VSCS or higher intensity

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC NOVEMBER 2013 O S U R I E T A L . 2491 and experiences maximum errors in the case of these sys- . Soc., 7A.3. [Available online at https://ams.confex. tems; the intensity prediction of CS and SCS is reasonably com/ams/pdfpapers/107899.pdf.] simulated, however. Still, the ARW-based intensity errors ——, S. Goldenberg, T. Quirino, X. Zhang, F. Marks, K.-S. Yeh, R. Atlas, and V. Tallapragada, 2012: Toward improving high- are higher for 24 and 48 h when compared with the oper- resolution numerical hurricane forecasting: Influence of ational intensity forecast errors of the IMD. This can be model horizontal grid resolution, initialization, and physics. further reduced by improving the initial intensity and Wea. Forecasting, 27, 647–666. structure of the TC vortex by increasing the observational Gupta, A., 2006: Current status of track prediction network of buoy, ships, and aircraft reconnaissance over techniques and forecast errors. Mausam, 57, 151–158. Han, J., and H.-L. Pan, 2011: Revision of convection and vertical the NIO region or through advanced vortex initialization diffusion schemes in the NCEP Global Forecast System. Wea. techniques and ocean– coupling for better heat, Forecasting, 26, 520–533. moisture, and momentum exchanges. Hong, S.-Y., and J.-W. Lee, 2009: Assessment of the WRF model in reproducing a flash-flood heavy rainfall event over Korea. Acknowledgments. The Indian National Center for Atmos. Res., 93, 818–831. Hsiao, L.-F., C.-S. Liou, T.-C. Yeh, Y.-R. Guo, D.-S. Chen, K.-N. Ocean Information Services (INCOIS), Ministry of Earth Huang, C.-T. Terng, and J.-H. Chen, 2010: A vortex relocation Sciences, government of India, and the U.S. National scheme for tropical cyclone initialization in Advanced Re- Foundation (CAREER Grant 0847472) are search WRF. Mon. Wea. Rev., 138, 3298–3315. gratefully acknowledged for providing financial support IMD, 2011: Tracks of cyclones and depressions over North Indian to carry out this research. The authors also thank the Ocean (from 1891 onwards). India Meteorological De- partment Tech. Note Cyclone eAtlas—IMD, version 2.0. IMD for providing TC best-track positions for validation [Available online at http://www.rmcchennaieatlas.tn.nic.in/ of model tracks. The authors gratefully acknowledge Help/TechNote2011.pdf.] NCEP and NCAR for their input datasets as well the Kotal, S. D., and S. K. Roy Bhowmik, 2011: A multimodel en- MMM division of NCAR for making the ARW model semble (MME) technique for cyclone track prediction over available. We also thank Mrs. Ammaji for her help during the North Indian Sea. Geofizika, 28, 275–291. revision. Thanks are given to Dallas Staley for her usual Landman, A. W., A. Seth, and S. J. Camargo, 2005: The effect of regional climate model domain choice on the simulation of outstanding . We also sincerely thank the anony- tropical cyclone–like vortices in the southwestern Indian mous reviewers for their constructive suggestions and Ocean. J. Climate, 18, 1263–1274. comments to improve the quality of the paper. Lynch, P., D. Giard, and V. Ivanovici, 1997: Improving the effi- ciency of a digital filtering scheme for diabatic initialization. Mon. Wea. Rev., 125, 1976–1982. REFERENCES Marchok, T. P., 2002: How the NCEP tropical cyclone tracker works. Preprints, 25th Conf. on Hurricanes and Tropical Mete- Ahn, M.-H., Y.-K. Seo, H.-S. Park, and A.-S. Suh, 2002: De- orology, San Diego, CA, Amer. Meteor. Soc., P1.12. [Available termination of tropical cyclone center using TRMM Micro- online at https://ams.confex.com/ams/pdfpapers/37628.pdf.] wave Imager data. Geophys. Res. Lett., 29, doi:10.1029/ Mohanty, U. C., and A. Gupta, 1997: Deterministic methods for 2001GL013715. prediction of tropical cyclone tracks. Mausam, 48, 257–272. Brousseau, P., and Coauthors, 2008: A prototype convective-scale ——, K. K. Osuri, and S. Pattanayak, 2013: A study on high reso- data assimilation system for operation: The Arome-RUC. lution mesoscale modeling systems for simulation of tropical HIRLAM Tech. Rep. 68, 23–30. cyclones over the Bay of Bengal. Mausam, 64, 117–134. Davis, C. A., and Coauthors, 2008: Prediction of landfalling hur- Mohapatra, M., B. K. Bandyopadhyay, and D. P. Nayak, 2013a: ricanes with the Advanced Hurricane WRF model. Mon. Wea. Evaluation of operational tropical cyclone intensity forecasts Rev., 136, 1990–2005. over north Indian Ocean issued by India Meteorological De- Dodla, V. B. R., and S. B. Ratna, 2010: Mesoscale characteristics partment. Nat. Hazards, 68, 433–451. and prediction of an unusual extreme heavy ——, D. P. Nayak, R. P. Sharma, and B. K. Bandyopadhyay, 2013b: event over India using a high resolution mesoscale model. Evaluation of official tropical cyclone track forecast over Atmos. Res., 95, 255–269. north Indian Ocean issued by India Meteorological De- Falguni, P., C. M. Kishtawal, P. K. Pal, and P. C. Joshi, 2004: partment. J. Earth Syst. Sci., 122, 589–601. Geolocation of Indian Ocean tropical cyclones using 85 GHz Niyogi, D., T. Holt, S. Zhong, P. C. Pyle, and J. Basara, 2006: Urban observations from TRMM Microwave Imager. Curr. Sci., 87, and land surface effects on the 30 July 2003 mesoscale con- 504–509. vective system event observed in the southern Great Plains. Fiorino, M., J. M. Goerss, J. J. Jensen, and E. J. Harrison Jr., 1993: J. Geophys. Res., 111, D19107, doi:10.1029/2005JD006746. An evaluation of the real-time tropical cyclone forecast skill Osuri, K. K., U. C. Mohanty, A. Routray, A. K. Makarand, and of the Navy Operational Global Atmospheric Prediction M. Mohapatra, 2012a: Sensitivity of physical parameterization System in the western North Pacific. Wea. Forecasting, 8, schemes of WRF model for the simulation of Indian 3–24. tropical cyclones. Nat. Hazards, 63, 1337–1359. Gopalakrishnan, S. G., N. Surgi, R. Tuleya, and Z. Janjic, 2006: ——, ——, ——, and M. Mohapatra, 2012b: Impact of satellite NCEP’s two-way-interactive-moving-nest NMM-WRF mod- derived wind data assimilation on track, intensity and struc- eling system for hurricane forecasting. Preprints, 27th Conf. on ture of tropical cyclones over North Indian Ocean. Int. Hurricanes and Tropical Meteorology, Monterey, CA, Amer. J. Remote Sens., 33, 1627–1652.

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC 2492 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

Pattanaik, D. R., and Y. V. Rama Rao, 2009: Track prediction of Tallapragada, V., N. Surgi, Q. Liu, Y. Kwon, R. Tuleya, and very severe cyclone ‘Nargis’ using high resolution Weather W. O’Connor, 2008: Performance of the advanced operational Research Forecasting (WRF) model. J. Earth Syst. Sci., 118, HWRF modeling system during pre-implementation testing 309–329. and in real-time 2007 hurricane season. Recorded Presentation, Pattanayak, S., U. C. Mohanty, and S. G. Gopalakrishnan, 2012: 28th Conf. on Hurricanes and Tropical Meteorology, Orlando, Simulation of very severe Cyclone Mala over Bay of Bengal FL, Amer. Meteor. Soc., 4A.5. [Available online at https://ams. with HWRF modeling system. Nat. Hazards, 63, 1413–1437. confex.com/ams/28Hurricanes/webprogram/Paper138066.html.] Pike, C. A., and C. J. Neumann, 1987: The variation of track Torn, R. D., and C. A. Davis, 2012: The influence of shallow con- forecast difficulty among tropical cyclone basins. Wea. Fore- vection on tropical cyclone track forecasts. Mon. Wea. Rev., casting, 2, 237–241. 140, 2188–2197. Ramarao, Y. V., H. R. Hatwar, and G. Agnihotri, 2006: Tropical WMO, 2009: Standard format for verification of TC forecast. cyclone prediction by numerical models in India Meteoro- World Meteorological Organization TCM-VI/Doc. 2.4, 6 pp. logical Department. Mausam, 57, 47–60. [Available online at www.wmo.int/pages/prog/www/tcp/ Routray, A., U. C. Mohanty, S. R. H. Rizvi, D. Niyogi, K. K. Osuri, documents/Doc2.4_Verification.doc.] and D. Pradhan, 2010: Impact of Doppler data Xiao, Q., 2011: Improvements of hurricane forecast with vortex on simulation of Indian monsoon depressions. Quart. J. Roy. initialization using WRF Variational (WRF-Var) data assim- Meteor. Soc., 136, 1836–1850. ilation. Recent Hurricane Research—Climate, Dynamics, and Roy Bhowmik, S. K., and S. D. Kotal, 2010: A dynamical statistical Societal Impacts, A. Lupo, Ed., InTech, 297–318. [Available model for prediction of a tropical cyclone. Mar. Geod., 33, online at http://www.intechopen.com/books/recent-hurricane- 412–423. research-climate-dynamics-and-societal-impacts.] Ryerson, W. R., S. R. Russel, L. Elsberry, and J. Wegiel, 2007: Yang,L.,W.-W.Li,D.Wang,andY.Li,2011:Analysisof Evaluations of the AFWA Weather Research Forecast Model tropical cyclones in the South China Sea and Bay of Bengal tropical cyclone predictions. Preprints, 27th Conf. on Hurri- during monsoon season. Recent Hurricane Research— canes and Tropical Meteorology, Monterey, CA, Amer. Me- Climate, Dynamics, and Societal Impacts, A. Lupo, Ed., teor. Soc., 7A.5. [Available online at https://ams.confex.com/ InTech, 227–246. [Available online at http://www.intechopen. ams/pdfpapers/108856.pdf.] com/books/recent-hurricane-research-climate-dynamics-and- Skamarock, W. C., J. B. Klemp, J. Dudhia, O. G. David, D. M. societal-impacts.] Barker, W. Wang, and J. G. Powers, 2005: A description of the Zhang, W., Y. Leung, and J. C. L. Chan, 2013: The analysis of Advanced Research WRF version 2. NCAR/TN–4681STR, tropical cyclone tracks in the western North Pacific through 88 pp. [Available online at http://www.mmm.ucar.edu/wrf/ data mining. Part I: Tropical cyclone recurvature. J. Appl. users/docs/arw_v2.pdf.] Meteor. Climatol., 52, 1394–1416.

Unauthenticated | Downloaded 09/30/21 10:54 PM UTC