impact of varying landfall time and cyclone intensity on

Impact of varying landfall time and cyclone intensityon storm surges in the Bay of Bengal usingADCIRC model

V G SHASHANK1,2, SAMIRAN MANDAL1 and SOURAV SIL1,*

1Ocean Analysis and Simulation Laboratory, School of Earth Ocean and Climate Sciences, Indian Institute ofTechnology Bhubaneswar, Bhubaneswar, Odisha, India.2Department of Ocean Engineering, Indian Institute of Technology Madras, Chennai, Tamil Nadu, India.*Corresponding author. e-mail: [email protected]

MS received 12 January 2021; revised 27 May 2021; accepted 28 May 2021

This study focuses on the impact of varying landfall timing in a tidal cycle (i.e., the spring-neap phase)and varying wind speeds (i.e., cyclone intensities) on the surge tides for the tropical cyclone Fani in theBay of Bengal using a hydrodynamic Bnite element-based 2D (ADvanced CIRCulation) ADCIRC modelsetup. For atmospheric forcing, the Cyclostrophic Symmetric Holland wind Model (H80) and GeneralizedAsymmetric Holland Model (GAHM) model are used to estimate the wind Belds from the IMD best trackdata. Comparisons with in-situ winds from moored buoys within the proximity of cyclone track showedthat H80 simulated winds underestimate the observed winds in terms of magnitude followed by a mis-match in the wind directions as well. In contrast, the GAHM simulated wind Belds, which are statisticallybetter in terms of both magnitude and direction are used for different wind experiments. To understandthe impact of the varying landfall timing and varying wind speeds on storm surges, a series of sensitivityexperiments have been performed during a tidal cycle with modulated high and low winds along thecyclone track. The experiments considering varying landfall timing during a tidal cycle indicate thestrongest surge tides (1.99 m) during the spring high tide phase, whereas the lowest surge tide of 0.94 m isobserved during spring low tide. However, the surge tide at the actual time of landfall is 1.20 m which isduring the transition from low tide to high tide. On the other hand, the combined impact of wind speedsand varying landfall timing indicated the strongest surge tides of 2.25 m during high wind conditionsassociated with spring high tides. In contrast, the surge tides decrease significantly during low tide andlow wind conditions. This study conBrms the importance of both winds and landfall timing on the stormsurges, which will be crucial to forecast the storm surges associated with the tropical cyclones.

Keywords. ADCIRC; storm surge; tides; winds; cyclone Fani; GAHM.

1. Introduction

The northern Indian Ocean (NIO) is extremelyvulnerable to tropical cyclones (TC). Insights showthat about 15% of the worldwide tropical cyclonesare over the north Indian Ocean, and the averagenumber of TCs in the north Indian Ocean is about

5–6% of the global annual average of cyclones(source: http://www.imd.gov.in/), of which thefrequency of cyclones that make landfall over thecoastal regions of Bay of Bengal (BoB) is relativelyhigher than those over the Arabian Sea (Mohapa-tra et al. 2012; Sahoo and Bhaskaran 2015). TheintensiBcation of these cyclones in the BoB is

J. Earth Syst. Sci. (2021) 130:194 � Indian Academy of Scienceshttps://doi.org/10.1007/s12040-021-01695-y (0123456789().,-volV)(0123456789().,-volV)

mostly dependent on the sea surface temperature,ocean heat content, distribution of the eddies in theopen ocean, and stratiBcation in the upper ocean(Ali et al. 2007; Tory et al. 2013; Mandal et al.2018a). These TCs generate surges, extreme windwaves, and coastal Coods during their landfall,thereby causing a fundamental threat to thehuman population and damage to the coastalecosystem. These destructions are primarilydependent on the cyclone intensities, the maximumradius of curvature of cyclonic winds, geometry,and bathymetry of the landfall region (Dube et al.1982; Johns et al. 1983; Peng et al. 2004; Peng andReynolds 2006; Zhong et al. 2010).The storm surges of higher amplitudes induced

by the cyclones were reported in the past, mostlyalong the north-western and head bay. The genesisand propagation of these surges are predominantlyinCuenced by the basin characteristics (coastalgeometry and width of the continental shelf),approach angle and translation speed of animpinging cyclone, and local astronomical tides(Zhang and Chen 2012; Poulose et al. 2018; Pandeyand Rao 2019). Since the tidal range variations arequite large from south to north in the western BoB(Murty and Henry 1983; Sindhu and Unnikrishnan2013), one can expect the interaction among sur-ges, tides, and waves to be prominent along thenorthern part of the coast during extreme weatherevents. Similarly, another study by Poulose et al.(2018) has analyzed the role and dependence ofvarying continental shelf characteristics on thenon-linear interaction of storm surges and windwaves along the west coast of India. The resultsrecommend an ampliBcation of the highest surgeof nearly 12 cm for every 10 km increase in thecontinental shelf width.The BoB has a purely semi-diurnal tidal regime,

with the M2 tides having the highest amplitudes(Murty and Henry 1983; Sindhu and Unnikrishnan2013). The tides and the tidal current amplitudesof the major constituents (M2 and S2) increasenorthward in the BoB, with the highest of theamplitudes along the northern and north-westernBoB (Mandal et al. 2018b, 2020). The interactionof the storm surges along the northern BoB duringthe high, and low tides may significantly alter thewater level elevations, which may lead to severedamage and devastation of the coastal ecosystem.Li et al. (2019) conducted an idealized experimentto study the impact on storm surges by varying themaximum wind speed and radius of maximumwind associated with a cyclone along the northern

East China Sea and concluded that both thecyclone intensity and radius of maximum windplay a significant role in determining the stormsurge. Bennet and Mulligan (2017) used the para-metric wind Belds from Symmetric Holland Model(H80) and Generalized Asymmetric Holland Model(GAHM) to compute the significant wave heights,which showed higher correlations of 0.82 and 0.75,respectively (averaged over nine sites), whencompared with the observed wave heights fromwave rider buoys. The accuracy of surge tideforecast depends not only on the cyclonic track andits intensity, but also on the spatial distribution ofwinds which includes its speed and direction(Pandey and Rao 2018).The impact of varying landfall time and inten-

sity on the generation of storm surges, and non-linear interaction of surges and tides are importantfor coastal management. Thus, it is necessary tounderstand the variations in the hydrodynamiccharacteristics associated with tropical cycloneintensity on the surge tides. The accurate presen-tation of the spatial wind Beld is necessary. Hence,the objective of this study is to adopt an accuratewind scheme for atmospheric forcing to theADCIRC model, which ensures better spatial dis-tribution of the wind Beld relating to the observedwind Beld to understand the surge variationsinduced by the tropical cyclone Fani. A series ofsensitivity experiments using the ADCIRC modelis carried out by allowing the cyclone to makelandfall at different phases of a tidal cycle, andfurther investigate the role of varying wind speedson the storm surges. A brief discussion on thecyclone Fani is given in section 2. Model experi-ments, data, and methodology are explained insection 3. Results are discussed in detail in section4, followed by a summary and conclusions insection 5.

2. Fani cyclone

The extremely severe cyclone storm Fani is one ofthe most powerful and destructive tropicalcyclones since the 1999 super cyclone that dwelledin BoB from 26th April to 3rd May 2019 (IMD2019). The system has experienced the lowestpressure drop to 934 hPa with a maximum 3 minsustained wind speed of 59 m/s. The low-pressuresystem originated as a depression on 26th April2019 under favourable oceanic and atmosphericconditions and started moving northwestwards

194 of 16 J. Earth Syst. Sci. (2021) 130:194

(Bgure 1). It further intensiBed into a deepdepression on 27th April 2019, followed by a quickintensiBcation to cyclonic storm over the south-eastern BoB on the same day. Moving northwest-wards, it escalated to a very severe cyclonic stormon 30th April 2019 over the southwestern BoB witha maximum wind speed of 41.15 m/s and 970 hPapressure drop. It moved northwards and furtherintensiBed into an extremely severe cyclonic stormon 1st May 2019 with a maximum pressure drop of934 hPa and the highest sustained wind speed of 59m/s. The system has undergone a rapid intensiB-cation to this stage due to the interaction with veryfavourable oceanic conditions with sea surfacetemperatures ranging between 30� and 31�C andlow vertical wind shear until the northwesternBoB. However, the system has comparativelyweakened in the oAshore regions Odisha coast justbefore landfall on 3rd May 2019. It continued to

move northeastwards, weakened to a severecyclonic storm over north Odisha of the same day,and ultimately dissipating by 4–5th May 2019(Bgure 1a).

3. Model, data, and methodology

This section gives a brief description of the Bnite-element based ADCIRC (ADvanced CIRCulation)hydrodynamic model, the various in-situ andbathymetry datasets used and the methodologyframed for the study.

3.1 ADCIRC model

ADCIRC is a two-dimensional depth-integratedshallow water hydrodynamic model that solves theshallow-water equations on an unstructured trian-gular mesh to calculate the water level and cur-rents at different scales (Kolar et al. 1994a, b;Luettich and Westerrink 2004). The water levelsare calculated by solving the vertically integratedgeneralized wave continuity equation (GWCE).The GWCE based on Bnite element solutions tothe shallow water equations provides stability tothe model (Kolar et al. 1994a, b; Luettich andWesterrink 2004; Dawson et al. 2006). The fullyparallelized version of ADCIRC using the message-passing interface (MPI) is used in the presentstudy. The MPI library calls to allow it to operateat high eDciency (typically better than 90%) onparallel computer architectures (source: https://adcirc.org). The model has been used in the BoBdomain in earlier studies to understand the waveand hydrodynamics characteristics associated withthe tropical cyclones, Phailin and Hudhud (Murtyet al. 2014, 2016) and further determined theextent of coastal inundation associated with thedifferent cyclones (Bhaskaran et al. 2014; Gayathriet al. 2015; Sahoo and Bhaskaran 2019; Rao et al.2020). Also, the model has been utilized to under-stand the tidal current characteristics in thenorthern Indian Ocean and the asymmetry hasbeen studied in the Hooghly estuary, the northernBoB by (Jena et al. 2018).

3.2 Atmospheric forcing models

The parametric cyclone model can be furtherdivided into a symmetric model and an asymmetricmodel. The idealized wind vortex is taken from thesymmetric Holland wind model 1980 (henceforth

Figure 1. (a) Best-estimated track (source: IMD) of thetropical cyclone Fani (black line) with centers at every 0000UTC from 26th April 2019 to 4th May 2019. Observationalsea-level data are taken from tide gauges (square dot) at Vizagand Paradeep. The colour bar denotes the bathymetry depthof the whole basin from ETOPO2. Buoys BD08, BD10, andBD13 locations are marked by black square dots. (b) Thetemporal variation of wind speed (in m/s) and surface pressure(in hPa) from the IMD track data during 26th April 2019–4thApril 2019 and, the red dot indicates landfall time.

J. Earth Syst. Sci. (2021) 130:194 of 16 194

denoted H80) as proposed by Holland (1980). Theasymmetric wind vortex is taken from the Gener-alized Asymmetric Wind Model (henceforth deno-ted GAHM) proposed by Gao et al. (2013).Further, the GAHM is adapted from the H80model by better taking into account the asymme-try of a cyclone. Here, we take different spatiallyvarying surface wind Belds from the parametricH80 and GAHM and compared them to windobservations using buoy winds for the recenttropical cyclone Fani formed in the BoB.

3.2.1 Symmetric Holland model 1980 (H80)

The wind Beld and atmospheric pressure are calcu-lated at each of the nodes internally by ADCIRCusing the symmetric Holland model, which hasbecome themost widely used cyclostrophic model inseveral hurricane-related studies (Mattocks andForbes 2008; Bhaskaran et al. 2013, 2014;Muis et al.2016; Wang et al. 2020). The radial pressure andwind proBles for the H80 expression are as follows:

P rð Þ ¼ Pc þ Pn � Pcð Þe�ðRmaxr ÞB ; ð1Þ

Vg rð Þ ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

V 2maxe

1� Rmaxrð ÞB Rmax

r

� �B

þ rf

2

� �2s

� rf

2

� �

; ð2Þ

where Pc is the minimum central atmosphericpressure, Pn is the ambient pressure (theoreticallyat inBnite radius), P rð Þ is the pressure at radius rfrom the center of the cyclone, Rmax is the radius ofmaximum wind, Vg is the gradient wind at radiusr, f is Coriolis force and, B is shape parameterwhich is given as:

B ¼ ½ðVm � VTÞ=WPBL�2qePn � Pcð Þ ; ð3Þ

where Vm is the maximum sustained wind speed,VT is the speed cyclone forward motion q is the airdensity (1.225kg/m3) and,WPBL is awind reductionfactor whose value is set to 0.90 (Powell et al. 2003).H80 requires the ‘Best Track’ parameters of thecyclone including storm eye location, Rmax, Vm, Pc

and Pn. In this study, the information on storm eyelocation, Vm, and Pc are extracted from the IndiaMeteorological Department (IMD) best-track datato be given as an input to the H80 model.

The poor approximation of the radial proBle inthe inner core of some cyclone cases in the H80model has been noted by Willoughby and Rahn(2004) amd Willoughby et al. (2006). Owing to thedrift motion, the wind Beld is not intensiBedon the right-hand side of the cyclone as it shouldbe in the northern hemisphere. However, thecyclostrophic approximation may induce inaccu-rate wind estimation away from the region ofmaximum winds. Thus, some asymmetric windBeld models were developed by including theelimination of cyclostrophic assumption, frictionalinCow corrections, and superimposing the transla-tion component of a cyclone to the H80 model(Zheng et al. 2017; Musinguzi et al. 2019; Qiaoet al. 2019).

3.2.2 Generalized asymmetric Holland model(GAHM)

Gao et al. (2013) developed GAHM, which differsfrom H80 with modiBcations that include elimi-nating the assumption of cyclostrophic balance atRmax (i.e., where the Coriolis force is negligiblecompared to the pressure gradient and centripetalforce in the gradient wind equation). This generates amore accurate representation of the wind Beld for awider range of cyclones and ensures that the pre-dicted winds match all available isotach information(Gao 2018). The radial pressure and wind proBle forthe GAHM expression are as follows:

P rð Þ ¼ Pc þ Pn � Pcð Þe�uðRmaxr ÞBg ; ð4Þ

Vg rð Þ

¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

V 2max 1þ 1

R0

� �

e1�Rmax

rð ÞBg Rmax

r

� �Bg

þ rf

2

� �2s

� rf

2

� �

; ð5Þ

where R0 is the Rossby number at r = Rmax, givenas R0 ¼ Vmax=Rmaxf . Moreover,Vmax is deBned asVmax ¼ ½ðVm � cVTÞ=WPBL�, where, c is the damp

factor ¼ Vg=Vmax and, Bg is the shape parameter(same as H80, without assuming the cyclostrophic

balance), given as, Bg ¼ B 1þ 1=R0ð Þ eu�1=u, whereu is the scaling parameter introduced in GAHM,given as u ¼ 1þ ð1=R0Þ=Bg 1þ 1=R0ð Þ.The GAHM is integrated within the ADCIRC

source code and uses the ‘Best Track’ informationin the Automated Tropical Cyclone Forecast


(ATCF) format. The information on the location ofthe eye (latitude and longitude), maximum sus-tained wind speed, and minimum sea-level pressureis extracted from the best track details obtainedfrom IMD (Shankar and Behera 2021), the Bg andRmax are pre-computed in four storm quadrants forthe available isotaches in the Asymmetric WindInput Preprocessor (ASWIP) program (in-built inADCIRC) and appended to the input Ble beforerunning an ADCIRC simulation.In addition, the storm translational speed cVT

is added to the vortex winds and tangential windsare rotated by frictional wind inCow. Angle (b)approximation were applied to the GAHM windBeld according to Harper et al. (2001), in whichapproximating the inCow angle increases linearlyfrom 0 at the storm center to 10� at Rmax and thento 25� at 1.2Rmax, and remains at 25� beyond 1.2Rmax (Lin and Chavas 2012)

b ¼10� for r\Rmax

10�þ 75ðr � RmaxÞ=Rmax for Rmax � r � 1:2Rmax

25� for r � 1:2Rmax:

8

<

:

ð6Þ

3.3 Data

The best-estimated track of the tropical cycloneFani and other relevant information on the atmo-spheric parameters are obtained from the IndiaMeteorological Department (IMD 2019). Thegridded bathymetry of the study domain isextracted from the NOAA ETOPO2 (NOAA 2006)database with a resolution of a 2-min grid. To val-idate the tides from the model, tide gauges at Par-adeep and Vizag are used. The winds from deep-water moored buoys (BD08, BD10, and BD13) areused to compare with the model-simulated winds.The observed data were obtained from INCOIS andNIOT, Ministry of Earth Sciences (MoES). Threebuoys are considered as they are in the proximity ofthe outer core region of the tropical cyclone Fani(Bgure 1a). The locations of buoys are mentioned intable 1. The details of the sensors used in the buoyare given in Venkatesan et al. (2013).

3.4 Methodology

A Bnite-element unstructured mesh was con-structed using the scalar paving density methodfrom a software package surface modeling system

(SMS) using the ETOPO2 bathymetry data. Theunstructured mesh constructed considering geo-graphical longitude/latitude projection and datumset for Indian mean value, which consists of 116,822elements and 229,914 nodes (Bgure 2). Theunstructured grid resolution, i.e., the node spacingvaries between 3 km very near the coast and amaximum of 60 km in the open ocean, as thecontinental shelf width increases from the southernto the northwestern BoB (source: http://www.aquaveo.com/software/sms-surface-water-modeling-system-introduction).In the present study, the standalone ADCIRC

model simulations use a hybrid bottom frictionformulation with a drag coefBcient of 0.0028 tocompute the storm surges. The values corre-sponding to the horizontal eddy viscosity coefB-cient and weighting factor (s0) in GWCE areconsidered as 4 m2/s and 0.05, respectively. Thetime step in the model is set to 20 s. This time stepsatisBes the Courant–Friedrichs–Lewy (CFL) sta-bility condition which avoids numerical instabilityassociated with the explicit numerical methodsused in the model. The Ramp function for all thesimulations is about 1 day. Ramping is a way bywhich terms can be steadily increased over sometime in a simulation. This is most often done formodel forcing terms like tides or winds, to avoidapplying a shock to the model. Also, the wettingand drying algorithm is activated in these simula-tions at a minimum depth of 0.5 m to delineate thewet and dry elements, and wind drag formulation isapplied (Garratt 1977). Le Provost’ tidal databaseis used to compute tidal forcing implicated to thedomain (Le Provost et al. 1998). The major tidalconstituents, K1, O1, Q1, P1, M2, N2, S2, K2, L2,2N2, MU2, NU2, and T2, are used as openboundary forcings to force the model. Two mainforces that generate storm surges on the sea surfaceare the surface wind stress and pressure gradientforce.In the present study, the wind Beld at the sea

surface for the cyclone Fani is derived by using H80and GAHM models from the IMD best track data.For the accuracy of surge tide prediction, the

Table 1. Details of the buoy locations.

Buoy Latitude Longitude

BD08 18�N 89�EBD10 16.3�N 88�EBD13 11�N 86�E


spatial wind distribution proBle obtained fromboth wind models has been analyzed. A time series(three hourly) comparison of wind speed and winddirection is made with the wind observation atBD08, BD10, and BD13. It is found that the windsin the outer core region of the cyclone are under-estimated by H80, whereas the GAHM producedstrong and broader wind Belds giving fewer errorsrelated to mean bias and root mean square near thebuoy locations (see section 4.2). Hence, furtherexperiments of predictions of surge tide for variousforcing phenomena (impact of landfall time andcyclone intensity) consisted of 21 simulations car-ried out using GAHM for atmospheric forcing inthe ADCIRC.The impact of varying winds and non-linear

interactions with the bathymetry during the dif-ferent phases of the tidal cycle may cause signifi-cant modulations in the surge levels. With theabove-mentioned model conBgurations, an attempthas been made to investigate the storm surgevariations by means of a series of sensitivityexperiments and thereby understand the impact ofvarying landfall timing during the differentphases of the tidal cycle in combination with themodulations in wind speeds.In the Brst part (see section 4.3), multiple sim-

ulations have been performed during the differentphases of a tidal cycle, namely, the nearest neaplow tide (NLT), the nearest neap high tide (NHT),the nearest spring high tide (SHT), the nearestspring low tide (SLT), nearest same day high tide(SDHT), the nearest same day low tide (SDLT),and the actual surge landfall time (AS). Therefore,

the study consists of seven simulations duringdifferent phases of a tidal cycle at the observed(actual) wind speed (refer to table 3). In the secondpart (see section 4.4), another 14 sets of experi-ments have been done to quantify the tide surge byincreasing the wind speed (high wind) anddecreasing the wind speed (low wind) by a constantmagnitude of value 7.7 m/s (15 Knots) to theactual wind speed value for the same tropicalcyclone track, Fani (see Bgure 3). And thus,imposing modulated wind intensity (high wind andlow wind) forcing on seven sets of different phasesof the tidal cycle.

4. Results

This section presents the results of this study infour subsections as the comparison of the observedsea level from the tide gauges with the modelsimulated sea level from the standalone ADCIRCmodel (in section 4.1), comparison of atmosphericwind models (in section 4.2), the impact of varyinglandfall timing on the surge heights (in section 4.3),and the combined eAect of both the varying windspeeds and the varying landfall timings during atidal cycle on the surge heights (in section 4.4).

4.1 Comparison with tide gauges

The sea level data from the ADCIRC model (withastronomical tidal forcing only) is compared withthe observed sea level data from the tide gaugesinstalled at Paradeep and Vizag (see Bgure 4).

Figure 2. The unstructured mesh using the Bnite element method (a) for the entire study domain and (b) enlarged view of theOdisha coast. BAN and SL indicate Bangladesh and Sri Lanka, respectively. The red dots denote the tide gauges installed atVizag and Paradeep.


Owing to the lack of tide gauge observed sea leveldata during the cyclone period, the comparisonshave been made during normal conditions for aperiod of about 25 days. Both the model simulatedand tide gauge records at Paradeep and Vizagindicate the significant signatures of two springand neap tides.The sea level at Paradeep varies within the range

of –1 to 1 m, with the evident signatures of a tidalcycle during January 2012, well captured fromboth the model and tide gauge (Bgure 4a). A strongcorrelation of 0.99 is observed with a low root-mean-square error (RMSE) of 7 cm (Bgure 4b).Similarly, the statistics at Vizag are also signifi-cant, with a correlation coefBcient of 0.99 and aslightly low RMSE of 5 cm (Bgure 4c and d). TheRMSEs vary in the range 5–7 cm, which is

attributed to the lack of wind forcing in the initialconditions of the standalone ADCIRC simulation.Furthermore, the harmonic analysis indicates thedominance of the M2 tides, followed by the othersemi-diurnal tidal constituents (S2 and N2) anddiurnal tidal constituents (K1 and O1), which is inwell accordance with the earlier results (Murty andHenry 1983; Mandal et al. 2018b, 2020).

4.2 Comparison of atmospheric wind models

The maximum wind speeds (in m/s) at all nodallocations in the domain, spatial wind distribution,and water levels in H80 and GAHM are repre-sented in Bgure 5(a and b), which shows the extentof high wind speeds caused by tropical cycloneFani. These Bgures clearly illustrate higher winds

Figure 3. The time series of the wind speeds associated with the cyclone tracks with different modulations. The red dotrepresents the landfall time.

Figure 4. Time series comparison (a and c) and scatter plots (b and d) of sea levels from the tide gauge and ADCIRC model atParadeep and Vizag. ‘R’ and ‘D’ denote correlation coefBcient and RMSE, respectively.


along the right side of the cyclone track (blue line)than the left side. However, the wind Beld fromGAHM is much stronger and wider than that ofH80. The H80 has a noticeably weaker and nar-rower wind Beld near the landfall area than that ofGAHM.The distance from the center of the cyclone to

the 10 ms�1 wind contour (shown in Bgure 5c andd) indicates that GAHM wind formulation leads toa reduction in exponential decay and an increase inthe radial extent of the wind in the outer coreregion of the cyclone, which is about 20% of theH80 wind proBle. The GAHM wind proBle showedthe radial extent of winds to be stronger and moredirected towards the coast on the right side of thecyclone track due to the asymmetric nature of thewind proBle. It is adopted by considering theelimination of the assumption of cyclostrophicbalance at the radius of maximum wind, imposingforward motion and frictional inCow angle

correction that allows for a better representation ofa wide range of cyclones (Gao 2018), which plays asignificant role in the development of surges alongthe coast.A closer investigation shows that H80 (Bgure 5e

and f) causes a peak surge in a narrower geo-graphical area and more localized near landfalllocation along the coast, whereas the GAHM cau-ses a peak in surge broader geographical area alongthe coast mostly on the right side of the track.Also, the simulated surge tide at the landfalllocation (85.67�E, 19.70�N) is 0.97 m for the sim-ulated winds from H80, and 1.20 m for the windsfrom GAHM, which leads to significantly lowerwater surface elevations for H80 than that ofGAHM.An inter-comparison of winds between H80 and

GAHM is performed with the buoy winds. Thewind Beld for the entire domain is constructedusing the GAHM model based on the IMD cyclone

Figure 5. Maximum wind speed track from (a) H80 and (b) GAHM for the tropical cyclone Fani. Spatial distribution of cyclonicwind Belds just before the landfall time (2nd May 17:00 hrs) from (c) H80 and (d) GAHM. The spatial maps of model-simulatedwater level elevation (m), the contours depict range from –0.2 to 1.2 m for every 0.2 m rise of water level and at the landfall time(3rd May 0300 hrs) using (e) H80 and (f) GAHM. Square black dots indicate the stations along the coast which have got aAecteddue to storm, namely, Srikakulam (SK), Ganjam (GJ), Puri (PU), Jagatsinghpur (JP) and Kendrapara (KP) and, thick blackline denotes tropical cyclone track for Fani.


track information and it is compared with theobserved data recorded from the moored buoysBD08, BD10, and BD13 (Bgure 6). The compar-isons of the model computed wind speeds(Bgure 6a, b, and c) with the observed data indi-cate that the GAHM has captured the wind speedsand proBles very well along with the maximumwind intensities (as the cyclones progressedtowards the coast) as compared to H80 and, theresults are in good accordance with the earlierresults (Bennett and Mulligan 2017).Also, a comparison of observed wind direction

has been performed with that of H80 and GAHMsimulated wind directions. The wind direction fromH80 shows a mismatch with the observed winddirection. In contrast, the GAHM simulated winddirections match better with the observed winddata from the moored buoys. The considerationfrictional inCow angle approximation according toHarper et al. (2001) in GAHM (as presented inequation 9) gives a better result. The H80 does notprovide wind inCow angle correction and correctionof the forward motion of the cyclone to account forthe asymmetry (Shen et al. 2006), hence the H80model results are deviating from the true windBelds in terms of wind direction.The validations results are estimated in terms

root-mean-square error (RMSE) and mean biasError (MBE) for all the three buoys, as shown intable 2. The errors corresponding to wind speed at

all the buoy locations are considerably less forGAHM compared to H80. The averaged RMSE hasreduced from 53.07� to 24.16� for wind directionsand 6.08 to 2.63 ms�1 from wind speeds usingGAHM. Hence, further experiments (followingfrom section 4.3 to 4.4) consisting of a total of 21simulations to quantify the tide surge on variousforcing (impact of landfall time and varyingcyclone intensity) are carried out using GAHM.To quantify the impact of cyclonic winds over

the tidal forcing during 30th April 2019 03:00 hrs to4th May 2019 06:00 hrs, two distinct simulationscarried out are: (i) only tidal forcing, and (ii)combined eAect of tidal and atmospheric forcing(using GAHM). The cyclone was on its intensiBedstage near the landfall location (85.67�E, 19.70�N)near Puri coast, post the neap tide phase on 3rdMay 2019 (Bgure 7a). Predominant signatures ofthe tropical cyclone are significantly observed fromthe different physical parameters (water level ele-vation, wind speed, and depth-integrated velocity).A significant difference is observed for the waterlevel elevations from the two simulations indicat-ing a storm-induced surge of 1.19 m (Bgure 7a andb), which is underestimating but nearer to thevalue issued by the IMD (about 1.5 m). The stormsurge is also associated with a high wind speed of 41m/s and strong currents of 1.33 m/s (Bgure 7c andd). Furthermore, the spatial distribution of boththe parameters also indicates the increased wind

Figure 6. Comparison of H80 and GAHM wind speed (a–c) and wind direction (d–f) with the observed data for tropical cycloneFani at different buoy locations BD07, BD10, and BD13. The red dot represents the landfall time.


speeds and high-water levels in the nearshoreregions (Bgure 5e and f).

4.3 Impact of varying landfall timing in a tidalcycle

This section focuses on the quantiBcation of theactual surge due to the cyclone and further inves-tigates the impact of varying landfall timing on theinduced surges during the different phases of a tidalcycle. The standalone ADCIRC model with tideforcing captures the observed tidal signals during21st April 2019 to 19th May 2019.In order to quantify the eAect of this varying

landfall timing on storm surges, seven simulationswith tidal forcing using cyclone winds associatedwith the cyclone track, namely, landfall at (i) ac-tual surge (AS), (ii) neap low tide (NLT), (iii) neaphigh tide (NHT), (iv) spring high tide (SHT),(v) spring low tide (SLT), (vi) same day high tide(SDHT), and (vii) same day low tide (SDLT) (red

dots in Bgure 8). The actual landfall happened on3rd May 2019 at 0300 hrs near the Puri coast, andan induced surge tide of 1.20 m is observed. Thelandfall timing has been shifted to the high tide andlow tide on the same day (SDHT and SDLT), andrespective surge tides are 1.61 and 1.06 m (table 3).Also, the signatures of spring and neap tides arepredominantly observed from the model and thus,a significant change in the water levels is observedwhen the landfall timings are shifted to the highand low tides during the spring and neap phases(Bgure 8). A storm tide of 1.14 m is observed duringNLT, whereas, the same is 1.49 m during NHT. Onthe other hand, the water levels are quite highduring the spring tides, which are nearly 1.99 and0.94 m during SHT and SLT, respectively. Thus, itcan be concluded that the tide–surge interactionsdepend predominantly on the landfall timing in atidal cycle.On the other hand, the difference between the

water levels observed during the storm high tide

Table 2. Mean bias error (MBE) and root-mean-square error (RMSE) at different buoy locationsfor the tropical cyclone Fani.

Wind model

BD08 BD10 BD13

MBE RMSE MBE RMSE MBE RMSE

H80 wind speed (m/s) 7.20 7.27 6.00 6.19 4.52 4.80

GAHM wind speed (m/s) 2.59 3.25 1.03 2.53 1.69 2.12

H80 wind direction (degree) 27.61 32.39 44.05 47.99 78.01 78.85

GAHM wind direction (degree) 4.09 10.23 2.98 29.95 21.12 32.31

Figure 7. (a) Model simulated water level elevation (in m) with only tides (blue line) and with winds (red line) during 30th April2019 12:00 hrs – 4th May 2019 06:00 hrs. (b) The storm-induced surge (in m) due to the actual wind speeds (i.e., the water leveldifference between storm tide and tide). The temporal variations of model-simulated (c) wind speed (in m/s) and (d) depth-averaged velocity (in m/s) during the same time. The red dot on temporal plots (a–d) indicates the landfall time (3rd May 201903:00 hrs) at the location 85.67�E, 19.70�N.


and astronomical high tide, similarly storm lowtide and astronomical low tide conditions distinctlyduring the actual, spring, and neap phase timingsare calculated from the two simulations (table 4).The difference in water level differences duringdifferent phases of the tidal cycle for the respectivehigh tide and low tides that values are close andrange from 1.21 to 1.48 m. These minute sea leveldifferences indicate the non-linear interaction per-formed by the model during the different phases ofa tidal cycle.

The Hovm€oller diagram for water level (in m) forthe different landfall time showed that the waterlevel is more surging towards the right side ascompared to the left side along the coast fromlandfall location for all the tidal phases (Bgure 9).The highest water level of 1.99 m is observedduring the spring high tide phase (SHT) and thelowest water level of 0.94 m is observed during thespring low tide phase (SLT) for the combined tidaland wind forcing along the coast. Moreover, thewater level gradually decreases on the left side from

Figure 8. The sea level variability near Puri coast from ADCIRC model (with only tide forcing) during 21st April 2019 to 19thMay 2019, denoting the different phases of a tidal cycle. The red dots indicate the landfall timings for the experiments.

Table 3. QuantiBcation of storm-induced tide surge (in m) at varying landfall times in a tidal cycle assuming the cyclone track isnot changed.

Tidal phase Cyclone start time Cyclone end time Tide surge (m) Landfall time

Actual surge (AS) 30/4/2019, 12:00 04/5/2019, 06:00 1.20 03/5/2019, 03:00

Same day high tide (SDHT) 30/4/2019, 15:00 04/5/2019, 09:00 1.61 03/5/2019, 06:00

Same day low tide (SDLT) 30/4/2019, 21:00 04/5/2019, 15:00 1.06 03/5/2019, 12:00

Neap low tide (NLT) 26/4/2019, 12:00 30/4/2019, 06:00 1.14 29/4/2019, 03:00

Neap high tide (NHT) 26/4/2019, 18:00 30/4/2019, 12:00 1.49 29/4/2019, 09:00

Spring high tide (SHT) 04/5/2019, 18:00 08/5/2019, 12:00 1.99 07/5/2019, 09:00

Spring low tide (SLT) 05/5/2019, 00:00 08/5/2019, 18:00 0.94 07/5/2019, 15:00

Table 4. Water levels simulated with actual wind during high and low tide of spring, neap, and actual phase ofthe tidal cycle.

High tide

(m)

High to high tide

difference

(m) Low tide (m)

Low to low tide

difference (m)

Spring

Tide 0.65 1.34 �0.54 1.48

Winds + Tide 1.99 0.94

Neap

Tide 0.13 1.36 �0.07 1.21


Actual

Tide 0.31 1.30 �0.33 1.39



the landfall locations as compared to the right sidefor all the landfall times with respect to a differentphase of the tidal cycle.

4.4 Combined impact of both the varying windspeeds and landfall timing

To further gain insights into the impact of varyingintensities of tropical cyclones along with thevarying landfall timings in a tidal cycle, 14 moresimulations have been carried out using two

manipulated wind speeds along the same cyclonetrack. The wind speeds associated with the tropicalcyclone have been modiBed from the actual windspeed by increasing the wind speed amount of 7.7m/s (+15 Knots), which is considered to be as highwind, and by decreasing by wind speed amount of7.7 m/s (–15 Knots) which is considered to be aslow wind. Moreover, the surge tides are the highestfor the high winds, followed by the actual windsand low winds. In the actual wind conditions and atactual landfall timing, the surge tide of 1.20 m isobserved. However, an increase in water level (1.45m) is observed due to higher winds, and the samedecreases to 0.96 m during the lower windconditions. Similar results are observed for all thephases of the tidal cycle for the respective windmodulations (Bgure 10a–g).The simulated surge tides at different tidal

phases with three different winds are shown inBgure 11(a). The maximum surge tide (2.25 m) isobtained in the case of spring high tide with highwind conditions. The minimum surge tide (0.78 m)is obtained from the spring low tide with low wind.The results showed that for a particular tidalphase, the tide surge increases with an increase inwind speeds. As Odisha coast is vulnerable tocoastal erosion due to the frequent occurrence ofthe tropical cyclone, so computation of alongshoresediment transport is an integral part of coastalscientists for sustainable management of its coastal

Figure 9. Hovm€oller diagram indicating water level (m) fordifferent landfall times (as shown on y-axis). The distancealong the coast (referring zero as landfall point (85.67�E,19.70�N). Negative (positive) values of x-axis representdistance on the left (right) of landfall.

Figure 10. The time series of water levels associated with the cyclone track for the tropical cyclone Fani with different windmodulations for the different phases of the tidal cycle (as presented in table 3).


resources, for which quantiBcation of the depth-averaged current Beld is also presented inBgure 11(b) in this study, which is essential for acyclonic event. It is evident that varying forcing byatmospheric wind and astronomical tide predomi-nantly acts on ocean currents, resulting in waterlevel changes.Thus, it can be concluded that not only the tides

and winds play a significant role in the intensiB-cation of the surge tides, but also the combinedimpact of both the higher winds and high tideconditions lead to an increased surge which couldcause potential damage to the coastal livingenvironment.

5. Summary and conclusions

The accuracy of storm surge forecast depends notonly on the cyclonic track and its intensity, butalso on the spatial distribution of winds whichinclude its speed and direction. Hence, in thisstudy, a two-dimensional hydrodynamic ADCIRCmodel has been conBgured for the BOB. The sim-ulated sea level data match well with the observedsea level data from the tide gauges (higher corre-lation and lower errors), suggesting that the relia-bility of the tide model studies the coastalhydrodynamics processes. Then, the recent tropicalcyclone Fani was simulated from the pressure datafrom IMD using symmetric H80 and asymmetricGAHM wind models. Comparisons with theobservations available at moored ocean buoysindicate that the H80 symmetric wind model

underestimates the cyclonic wind speeds in theouter core region and also shows a mismatch in itsdirection. The asymmetric winds from GAHMsimulated relatively better winds with less bias andRMSEs. The advantage of GAHM over H80 is theelimination of cyclostrophic assumption and, fric-tional inCow corrections, which generates anasymmetric and more accurate representation ofthe wind Beld for a wider range of cyclones forGAHM. Further, 21 simulations are carried outusing GAHM to quantify the impact of landfalltime and varying cyclone intensity on stormsurges.This study investigates the coastal hydrody-

namics associated with the tropical cyclone ‘Fani’to particularly focus on (i) the impact of variablelandfall timing on the surge tides in a tidal cycle,and also (ii) to elucidate the combined impact ofvarying winds and different landfall timings in atidal cycle on the surge tides. The major results ofthis study are summarized below:

• The signatures of the tropical cyclone are wellcaptured in the model with an asymmetric windspeed of 41 m/s, and a strong depth-integratedvelocity of 1.33 m/s. Also, a storm-induced surgeof 1.19 m is observed from the model, which isnearer to the surge values forecasted by the IndiaMeteorological Department (1.50 m).

• A series of sensitivity experiments consideringvarying landfall timing during a tidal cycleindicate the strongest tide surge (*1.99 m)during the spring high tide phase, followed by atide surge of 1.61 m during the same day hightide (SDHT). However, the tide surges arecomparatively less for landfall during the neaptidal phases.

• The combined impact of wind speeds and vary-ing landfall timing indicated the strongest surgetides of 2.25 m during high wind conditionsassociated with spring high tides. On the otherhand, the surge tides of 1.99 and 1.75 m areobserved during the actual wind and low windconditions, respectively. Surges are compara-tively less during the low wind and low tidecombinations during the neap tidal phase.

Forecasting the storm surges is very crucial toprevent the huge devastation of coastal ecosys-tems, habitat, and economic losses. However, theprediction of the surge during a high tide or a lowtide condition is very imperative, which is highlydependent on the cyclone landfall time. Thus, thetide-surge interactions in the prediction of storm

Figure 11. The bars indicate (a) the tide surge and (b) depth-averaged currents during different phases of a tidal cycle(according to ascending landfall time order), for low, actual,and high wind conditions (as presented in colour bars) at thelandfall location 85.67�E, 19.70�N.


surge are essential. This study presents insightsinto the stronger impacts of tides and winds on thesurges depending on the varying landfall timing(during high and low tides) during high and lowwind conditions. This particular result is quiteimportant and valuable to be incorporated in thedevelopment of an operational forecasting systemalong the different coasts around the world topredict the coastal inundation and surge tides todisseminate proper warnings to the coastal zoneauthorities to adopt evacuation policies.

Acknowledgements

The authors acknowledge the Bnancial assistancefrom the Science and Engineering Research Board(SERB), Government of India (CRG/2019/005842). They thank INCOIS and NIOT, Ministryof Earth Sciences (MoES) for providing the tidegauge and deep water buoy datasets free of cost.The authors thank the Editor and reviewers fortheir suggestions to improve the quality of thework. Finally, the Indian Institute of TechnologyBhubaneswar is acknowledged for providingBnancial and infrastructural support.

Author statement

VGS performed all analyses and wrote the initialmanuscript. SM contributed to the methodology,interpretation, and editing of the manuscript. SSsupervised the entire work. All authors VGS, SM,and SS have contributed to the discussions andwriting the manuscript.

References

Ali M M, Jagadeesh P V and Jain S 2007 EAects of eddies onBay of Bengal cyclone intensity; EOS 88(8) 93–95, https://doi.org/10.1029/2007EO080001.

Bennett V C and Mulligan R P 2017 Evaluation of surfacewind Belds for prediction of directional ocean wave spectraduring hurricane sandy; Coast. Eng. 125 1–5, https://doi.org/10.1016/j.coastaleng.2017.04.003.

Bhaskaran P K, Gayathri R, Murty P L, Bonthu S and Sen DA 2014 Numerical study of coastal inundation and itsvalidation for Thane cyclone in the Bay of Bengal; Coast.Eng. 83 108–118, https://doi.org/10.1016/j.coastaleng.2013.10.005.

Bhaskaran P K, Nayak S, Bonthu S R, Murty P N and Sen D2013 Performance and validation of a coupled parallelADCIRC–SWAN model for THANE cyclone in the Bay of

Bengal; Environ. Fluid Mech. 13(6) 601–623, https://doi.org/10.1016/j.csr.2006.07.021.

Dawson C, Westerink J J, Feyen J C and Pothina D 2006Continuous, discontinuous and coupled discontinuous–con-tinuous Galerkin Bnite element methods for the shallowwater equations; Int. J. Numer. Methods Fluids 52(1)63–88, https://doi.org/10.1002/Cd.1156.

Dube S K, Sinha P C and Rao A D 1982 The eAect of coastalgeometry on the location of peak surge; Mausam 33(4)445–450.

Gao J 2018 On the surface wind stress for storm surgemodeling. Doctoral dissertation, The University of NorthCarolina at Chapel Hill.

Gao J, Luettich R and Fleming J 2013 Development and initialevaluation of a generalized asymmetric tropical cyclone vortexmodel in ADCIRC; In: Proceedings of the ADCIRC UsersGroup Meeting, Vicksburg, MS, USA (Vol. 16).

Garratt J R 1977 Review of drag coefBcients over oceans andcontinents; Mon. Weather Rev. 105(7) 915–929.

Gayathri R, Bhaskaran P K and Sen D 2015 Numerical studyon storm surge and associated coastal inundation for 2009AILA Cyclone in the head Bay of Bengal; Aquat Procedia 4404–411, https://doi.org/10.1016/j.aqpro.2015.02.054.

Harper B A, Hardy T, Mason L, Bode L, Young I and NielsenP 2001 Queensland climate change and community vulner-ability to tropical cyclones: Ocean hazards assessment–Stage 1; Report prepared by Systems Engineering Aus-tralia in conjunction with James Cook University MarineModelling Unit, Queensland Government.

Holland G J 1980 An analytic model of the wind and pressureproBles in hurricanes; Mon. Weather Rev. 108(8)1212–1218, https://doi.org/10.1175/1520-0493(1980)108\1212:AAMOTW[2.0.CO;2.

IMD Report 2019 Extremely severe cyclonic storm Fani overeast central equatorial Indian Ocean and adjoining south-east Bay of Bengal, India Meteorological Department.

Jena B K, Sivakholundu K M and Rajkumar J A 2018Description of tidal propagation in Hooghly estuary usingnumerical and analytical solutions; Ocean Eng. 169 38–48,https://doi.org/10.1016/j.oceaneng.2018.09.009.

Johns B, Sinha P C, Dube S K, Mohanty U C and Rao A D1983 Simulation of storm surges using a three-dimensionalnumerical model: An application to the 1977 Andhracyclone; Quart. J. Roy. Meteorol. Soc. 109(459) 211–224,https://doi.org/10.1002/qj.49710945910.

Kolar R L, Gray W G, Westerink J J and Luettich R A Jr 1994Shallow water modeling in spherical coordinates: Equa-tion formulation, numerical implementation, and applica-tion; J. Hydraul. Res. 32(1) 3–24, https://doi.org/10.1080/00221689409498786.

Kolar R L, Westerink J J, Cantekin M E and Blain C A 1994Aspects of nonlinear simulations using shallow-water mod-els based on the wave continuity equation; Comput. Fluids23(3) 523–538, https://doi.org/10.1016/0045-7930(94)90017-5.

Le Provost C, Lyard F, Molines J M, Genco M L andRabilloud F 1998 A hydrodynamic ocean tide modelimproved by assimilating a satellite altimeter-derived dataset; J. Geophys. Res. Oceans 103(C3) 5513–5529, https://doi.org/10.1029/97JC01733.

Li J, Hou Y, Mo D, Liu Q and Zhang Y 2019 InCuence oftropical cyclone intensity and size on storm surge in the


Northern East China Sea; Remote Sens. 11(24) 3033,https://doi.org/10.3390/rs11243033.

Lin N and Chavas D 2012 On hurricane parametric wind andapplications in storm surge modeling; J. Geophys. Res.Atmos. 117(D9), https://doi.org/10.1029/2011JD017126.

Luettich R A and Westerink J J 2004 Formulation andnumerical implementation of the 2D/3D ADCIRC Bniteelement model version 44. XX; 74p.

Mandal S, Sil S and Gangopadhyay A 2020 Tide-current-eddyinteraction: A seasonal study using high frequency radarobservations along the western Bay of Bengal near 16�N;Estuar. Coast. Shelf Sci. 232 106523, https://doi.org/10.1016/j.ecss.2019.106523.

Mandal S, Sil S, Shee A and Venkatesan R 2018 Upper Oceanand subsurface variability in the Bay of Bengal duringcyclone ROANU: A synergistic view using in-situ andsatellite observations; Pure Appl. Geophys. 175 4605–4624,https://doi.org/10.1007/s00024-018-1932-8.

Mandal S, Sil S, Gangopadhyay A, Murty T and Swain D2018 On extracting high-frequency tidal variability fromHF radar data in the northwestern Bay of Bengal;J. Oper. Oceanogr. 11(2) 65–81, https://doi.org/10.1080/1755876X.2018.1479571.

Mattocks C and Forbes C 2008 A real-time, event-triggeredstorm surge forecasting system for the state of NorthCarolina; Ocean Model. 25(3–4) 95–119, https://doi.org/10.1016/j.ocemod.2008.06.008.

Mohapatra M, Nayak D P and Bandyopadhyay B K 2012Evaluation of cone of uncertainty in tropical cyclone trackforecast over North Indian Ocean issued by India Meteo-rological Department; Trop. Cyclone Res. Rev. 1(3)331–339, https://doi.org/10.6057/2012TCRR03.02.

Muis S, Verlaan M, Winsemius H, Aerts J and Ward P 2016 Aglobal reanalysis of storm surges and extreme sea levels;Nat. Commun. 7(1) 1–2, https://doi.org/10.1038/ncomms11969.

Murty P L, Bhaskaran P K, Gayathri R, Sahoo B, Kumar T Sand Subba Reddy B 2016 Numerical study of coastalhydrodynamics using a coupled model for Hudhud cyclonein the Bay of Bengal; Estuar. Coast. Shelf Sci. 183 13–27,https://doi.org/10.1016/j.ecss.2016.10.013.

Murty P L, Sandhya K G, Bhaskaran P K, Jose F, Gayathri R,Nair T B, Kumar T S and Shenoi S S 2014 A coupledhydrodynamic modeling system for PHAILIN cyclone inthe Bay of Bengal; Coast. Eng. 93 71–81, https://doi.org/10.1016/j.coastaleng.2014.08.006.

Murty T S and Henry R F 1983 Tides in the Bay of Bengal; J.Geophys. Res. Oceans. 88(C10) 6069–6076, https://doi.org/10.1029/JC088iC10p06069.

Musinguzi A, Akbar M K, Fleming J G and Hargrove S K 2019Understanding hurricane storm surge generation and prop-agation using a forecasting model, forecast advisories andbest track in a wind model, and observed data – Case studyHurricane Rita; J. Mar. Sci. Eng. 7(3) 77, https://doi.org/10.3390/jmse7030077.

NOAA National Geophysical Data Center 2006 2-minuteGridded Global Relief Data (ETOPO2) v2; NOAANational Centers for Environmental Information, https://doi.org/10.7289/V5J1012Q.

Pandey S and Rao A D 2018 An improved cyclonic winddistribution for computation of storm surges; Nat. Hazards92(1) 93–112, https://doi.org/10.1007/s11069-018-3193-3.

Pandey S and Rao A D 2019 Impact of approach angle of animpinging cyclone on generation of storm surges and itsinteraction with tides and wind waves; J. Geophys. Res.Oceans 124(11) 7643–7660, https://doi.org/10.1029/2019JC015433.

Peng M S and Reynolds C A 2006 Sensitivity of tropicalcyclone forecasts as revealed by singular vectors; J. Atmos.Sci. 63(10) 2508–2528, https://doi.org/10.1175/JAS3777.1.

Peng M S, Ridout J A and Hogan T F 2004 RecentmodiBcations of the Emanuel convective scheme in theNavy operational global atmospheric prediction system;Mon. Weather Rev. 132(5) 1254–1268.

Poulose J, Rao A D and Bhaskaran P K 2018 Role ofcontinental shelf on non-linear interaction of storm surges,tides and wind waves: An idealized study representing thewest coast of India; Estuar. Coast. Shelf Sci. 207 457–470,https://doi.org/10.1016/j.ecss.2017.06.007.

Powell M D, Vickery P J and Reinhold T A 2003 Reduceddrag coefBcient for high wind speeds in tropical cyclones;Nature 422(6929) 279–283, https://doi.org/10.1038/nature01481.

Qiao W, Song J, He H and Li F 2019 Application of differentwind Beld models and wave boundary layer model totyphoon waves numerical simulation in WAVEWATCH IIImodel; Tellus A 71(1) 1657552, https://doi.org/10.1080/16000870.2019.1657552.

Rao A D, Upadhaya P, Pandey S and Poulose J 2020Simulation of extreme water levels in response to tropicalcyclones along the Indian coast: A climate change perspec-tive; Nat. Hazards 100(1) 151–172, https://doi.org/10.1007/s11069-019-03804-z.

Sahoo B and Bhaskaran P K 2015 Synthesis of tropical cyclonetracks in a risk evaluation perspective for the east coast ofIndia; In: International Conference on Water ResourcesCoastal and Ocean Engineering, (ICWRCOE) 15 389–396,https://doi.org/10.1016/j.aqpro.2015.02.052.

Sahoo B and Bhaskaran P K 2019 Prediction of storm surgeand inundation using climatological datasets for the Indiancoast using soft computing techniques; Soft. Comput.23(23) 12,363–12,383, https://doi.org/10.1007/s00500-019-03775-0.

Shankar C G and Behera M R 2021 Numerical analysis on theeAect of wave boundary condition in storm wave and surgemodeling for a tropical cyclonic condition; Ocean Eng. 220108371, https://doi.org/10.1016/j.oceaneng.2020.108371.

Shen J, Gong W and Wang H V 2006 Water level responseto 1999 Hurricane Floyd in the Chesapeake Bay; Cont.Shelf Res. 26(19) 2484–2502, https://doi.org/10.1016/j.csr.2006.07.021.

Sindhu B and Unnikrishnan A S 2013 Characteristics of tidesin the Bay of Bengal; Mar. Geod. 36(4) 377–407, https://doi.org/10.1080/01490419.2013.781088.

Tory K J, Dare R A, Davidson N E, McBride J L and Chand SS 2013 The importance of low-deformation vorticity intropical cyclone formation; Atmos. Chem. Phys. 13(4)2115–2132, https://doi.org/10.5194/acp-13-2115-2013.

Venkatesan R, Shamji V R, Latha G, Mathew S, Rao R R,Muthiah A and Atmanand M A 2013 In-situ oceansubsurface time-series measurements from OMNI buoynetwork in the Bay of Bengal; Curr. Sci. 10 1166–1177,https://www.jstor.org/stable/24092396.


Wang X, Yao C, Gao G, Jiang H, Xu D, Chen G and Zhang Z2020 Simulating tropical cyclone waves in the East ChinaSea with an event-based, parametric-adjusted model; J.Oceanogr. 76(6) 439–457, https://doi.org/10.1007/s10872-020-00555-5.

Willoughby H E, Darling RWR and RahnME 2006 Parametricrepresentation of the primary hurricane vortex. Part II: A newfamily of sectionally continuous proBles; Mon. Weather Rev.134 1102–1120, https://doi.org/10.1175/MWR3106.1.

Willoughby H E and Rahn M E 2004 Parametric representationof the primary hurricane vortex. Part I: Observations andevaluation of the Holland (1980) model; Mon. Weather Rev.132(12) 3033–3048. https://doi.org/10.1175/MWR2831.1.

Zhang D L and Chen H 2012 Importance of the upper-levelwarm core in the rapid intensiBcation of a tropical cyclone;Geophys. Res. Lett. 39(2), https://doi.org/10.1029/2011GL050578.

Zheng J H, Wang J C, Zhou C Y, Zhao H J and Sang S 2017Numerical simulation of typhoon-induced storm surgealong Jiangsu coast, Part II: Calculation of storm surge;Water Sci. Eng. 10(1) 8–16, https://doi.org/10.1016/j.wse.2017.03.011.

Zhong W, Lu H C and Zhang D L 2010 Mesoscale barotropicinstability of vortex Rossby waves in tropical cyclones; Adv.Atmos. Sci. 27(2) 243–252, https://doi.org/10.1007/s00376-009-8183-7.

Corresponding editor: C GNANASEELAN


impact of varying landfall time and cyclone intensity on

Documents