Tropical cyclones (TCs) are the most destructive weather systems on the earth, producing intense winds, resulting in high surges, meteotsunamis, torrential rains, severe floods and usually causing damage to property and loss of life. In the northern Indian Ocean, both the Bay of Bengal (BOB) and the Arabian Sea (AS) are potential genesis regions for cyclonic storms. Intense winds associated with TCs, blowing over a large water surface, cause the sea surface to pile up on the coast and leads to sudden inundation and flooding of the vast coastal regions. Also, the heavy rainfall causes flooding of river deltas in combination with tides and surges. A number of general reviews and description of individual cyclones and associated surges in the BOB and the AS have been published previously by several investigators (Murty et al., 1986; Dube et al., 1997; Sundar et al., 1999; Fritz et al., 2010; Joseph et al., 2011). Developments in storm surge prediction in the Bay of Bengal and the Arabian Sea have been highlighted by Dube et al. (2009) and references therein (e.g. Das, 1994; Chittibabu et al., 2000, 2002; Dube et al., 2006; Jain et al., 2007; Rao et al., 2008).

Apart from the studies carried out with a view to assessing the coastal vulnerability, few studies concentrated on the variations in characteristics of different oceanographic parameters in response to tropical cyclones. Joseph et al. (2011) examined the response of the coastal regions of eastern Arabian Sea (AS) and Kavaratti Island lagoon to the tropical cyclonic storm “Phyan”, during 9–12 November 2009 until its landfall at the northwest coast of India, based on in situ and satellite-derived measurements. Mehra et al. (2012) reported similarities in the spectral characteristics of sea level oscillations in the Mandovi estuary of Goa in the eastern Arabian Sea due to cyclones (June 2007 and November 2009) and the Sumatra geophysical tsunami (September 2007). Wang et al. (2012) reported the variations in the oceanographic parameters due to the tropical Cyclone Gonu, which passed over a deep autonomous mooring system in the northern Arabian Sea and a shallow cabled mooring system in the Sea of Oman. Near-inertial oscillations at all moorings from thermocline to seafloor were observed to be coincident with the arrival of Gonu. Sub-inertial oscillations with periods of 2–10 days were recorded at the post-storm relaxation stage of Gonu, primarily in the thermocline of the deep array and at the onshore regions of the shallow array. Antony and Unnikrishnan (2013) used hourly tide gauge data at Chennai, Visakhapatnam and Paradip along the east coast of India and at Hiron Point, at the head of Bay of Bengal, to analyse statistically the tide-surge interaction. Recently, Rao et al. (2013) simulated surges and water levels along the east coast of India using an advanced 2-D depth-integrated circulation model (ADCIRC-2DDI).

It is necessary that the problem of storm surge must be seriously addressed by the countries of the various regions through collective efforts and in an integrated manner. In the present study, the objective is to examine the characteristic of the sea level oscillations at different topographic locations in the AS and the BOB due to the meteorological events. Our interest is confined to a few minutes to days and analysis of the spectral features of sea level oscillations in the two basins.

In the present study, we report the response of the sea level to the episodic meteorological events at various coastal and Island locations of India from 1 September 2011 to 31 January 2012. Study encompasses two episodic meteorological events: (i) deep depression in November 2011 (E1) in the AS and (ii) the tropical Cyclone Thane (E2) in the BOB as shown in Fig. 1. Summary of observations is given in Table 1. The radar gauge (RG), which measures sea level, is described in detail by Prabhudesai et al. (2006, 2008) and the evaluation and comparative studies have been reported by Mehra et al. (2013). RG acquires samples over 30 s window at 1 min interval and the average over 5 min is recorded at 5 min interval. The surface meteorological variables are collected by autonomous weather station (NIO-AWS). AWS samples (wind, air temperature, air pressure and relative humidity) data every 10 s over a window of 10 min, averaged and then recorded at every 10 min interval. In the present study, we have used time-series data at 5 (10) min interval from the RG (AWS). Both the systems have been designed and developed in the Marine Instrumentation Division, CSIR-NIO, Goa. Summary of observations from different coastal and Island locations of India are provided in Table 1 and the periods covered for different events are as follows:

Event 1 (E1): 26 November to 1 December 2011, occurrence of deep depression in the Arabian Sea.

Sea level data are de-tided using TASK tidal analysis and prediction program (Bell et al., 2000) to obtain sea level residual (SLR). A multi-linear regression model linking sea level and atmospheric parameters has been established. The model can be described in general as follows: η=B0+B1τU+B2τV+B3AP+∈, In the above expression, sea level residual (η) is the dependent variable and the independent variables are cross-shore (along-shore) wind stress τU(τV) and atmospheric pressure (AP). Likewise B0,B1,B2 and B3 are the coefficients of regression and ϵ is the difference between the measured SLR and estimated SLR using multi-linear regression. The cross-shore (along-shore) wind stress τU(τV) is estimated using cross-shore (U) and along-shore (V) component of winds as follows: τU=ρACDUU2+V2τV=ρACDVU2+V2. ρA=1.3 kg m-3 is the density of air and CD=1.2×10-3 is the drag coefficient. The regression is performed using daily mean SLR, τU,τV and AP. For each month, coefficients of regression of the daily data are obtained to estimate the SLR, which is then merged to generate the time series of estimated SLR for the duration of September 2011 to January 2012.

The tracks of the meteorological event under study, which occurred in the AS (the BOB) are shown in Fig. 1. The data and information about these episodic meteorological events is taken from www.imd.gov.in.

The meteorological event (E1) in AS, developed on 26 November 2011 at 7.5∘ N, 76.5∘ E near the southern tip of the Indian sub-continent and moved north-westwards. By 28 November 2011 00:00 UTC, it intensified as a deep depression with maximum sustained surface winds reaching up to 15 m s-1 (Fig. 2a) and the minimum estimated central pressure (ECP) ∼ 998 mb (Fig. 2b). The average translational speed of this system remained steady to ∼ 6.5 m s-1. However, during the minimum ECP, the translation speed also decreased to ∼ 2 m s-1 on 29 November and increased to ∼ 6 m s-1 on 30 November (Fig. 2c). The system weakened into a well-marked low pressure area over the west central Arabian Sea.

The cyclonic system named “Thane” initially originated as a depression on 25 December 2011 at 8.5∘ N, 88.5∘ E and moved north-westwards (Fig. 1). Thane intensified into a very severe cyclonic storm with maximum sustained surface winds peaked up to ∼ 45 m s-1 as shown in Fig. 2d and ECP falling to ∼ 956 mb (Fig. 2e). The cyclone track turned westwards on 28 December, with an average translational speed of ∼ 4 m s-1 and then became steady at ∼ 3.5 m s-1 as shown in Fig. 2f. The translation speed of a storm can exert significant control on the intensity of storms by modulating the strength of the negative effect of the storm-induced sea surface temperature (SST) reduction on the storm intensification (Mei et al., 2012). Thane crossed the Tamil Nadu coast just south of Cuddalore between 01:00 and 02:00 UTC of 30 December 2011 and weakened into a well-marked low pressure area over northern Kerala and its neighbourhood.

Multi-linear regression analysis of SLR as the dependent variable with wind stress components (τU,τV) and atmospheric pressure (AP) as the independent variable is performed as explained in Sect. 2 (Eq. 1). How well the model describes the sea level residual is assessed by looking at the percentage of sea level variance explained (Vare) by the model.

Vare=1-variance(ε)variance(measuredSLR)×100 The multi-linear regression performed with 10, 60 min, 6, 12 and 24 h averaged data of Verem and Karwar is shown in the Supplement (Figs. S1–S5) along with the SLR variance (Table S1). The daily average of estimated SLR is comparable with the daily average of measured SLR, and it is able to account for the low frequency variations in the SLR during the study period of 5 months (September 2011 to January 2012). Therefore, in this section daily averaged data series of SLR and AWS is used to perform multi-linear regression. To begin with, multi-linear regression is performed with cross-shore (τU), along-shore (τV) components of winds stress and atmospheric pressure (AP) individually as independent variable to regress the SLR. This would enable us to know, the contribution to the SLR variability by the various surface meteorological variables individually. Then all the three independent variables (τU,τV,AP) together are used to regress the daily mean SLR. Results are listed in Table 4 and plotted in Figs. 6 and 7. The cross- and along-shore components are estimated using the local shoreline angles with respect to the north from Google Earth. In AS, the local shoreline angle estimated at Ratnagiri, Verem and Karwar with respect to the north is ∼ -12, -27 and -19∘, respectively. Similarly, in the BOB the local shoreline angle with respect to the north is ∼ 52, 45 and 50∘ at Gopalpur, Gangavaram and Kakinada, respectively. The variations in τU,τV and AP results in the SLR variability. The relation between atmospheric pressure and sea level is ∼ -1 cm mb-1, inverse barometric effect. The cross-shore wind (U) towards land (sea) will give rise to increase (decrease) in SLR due to wind stress. Since the study region is in the Northern Hemisphere, the along-shore northward winds with the coast at their right will increase the sea level due to Coriolis force. Therefore, in the AS the positive (negative) U and V will increase (decrease) the SLR (Fig. 6) and in the BOB the positive (negative) U and V will decrease (increase) the SLR (Fig. 7).

At all the three study locations in AS, τU,τV and AP individually could explain an average 45 % SLR variability; when τU,τV, and AP together are used to regress the daily mean SLR, the total Vare is ∼ 69 % (Table 4). Daily mean U(τU), V(τV) and estimated SLR obtained from independent variables (τU,τV,AP) for Ratnagiri is plotted in Fig. 6a. It is observed that the estimated daily mean SLR during E1 is able to peak at up to 96.3 % of the measured daily mean SLR (Fig. 6a.2 and Table 5). The monthly Vare is ∼ 50 % during October and December 2011 (Table 4 and Fig. 8a). However, when E1 occurred in November 2011, τU,τV, and AP together are able to explain the SLR variability up to ∼ 68 %. In January 2012, the Vare is less than 20 % at all the sites in the AS and the BOB (Fig. 8a and Table 4). At Verem (Fig. 6b), the total Vare is highest among all the locations in AS, when τU,τV, and AP together are used to estimate the daily mean SLR as presented in Fig. 6b.2. During E1, the -U(V) will tend to decrease (increase) the sea level at Verem (Fig. 6b.1), still the estimated daily mean SLR is able to reproduce the comparable response with measured daily mean SLR, with a minor overshoot by ∼ 6.5 % (Table 5). At Karwar also the U(-V) will tend to increase (decrease) the sea level during E1 (Fig. 6c.1); however, the estimated SLR is able to peak only at up to half of the measured daily mean SLR (Fig. 6c.2 and Table 5).

In the BOB, the τU,τV and AP individually could explain an average 44 % SLR variability, when τU,τV, and AP together are used to regress the daily mean SLR, the total Vare is ∼ 57 % (Table 4). The monthly Vare is low < 20 % in October and January for all the three stations in the BOB as shown in Fig. 8b. In the BOB, the daily mean U(τU), V(τV) and estimated SLR is plotted in Fig. 7 for Gopalpur, Gangavaram and Kakinada. As stated earlier, along the east coast of India in the BOB, the positive (negative) U and V will decrease (increase) the sea level. Figure 7a.1 shows the U(-V) at Gopalpur, which are seaward (southward) during E2 favouring the sea-level fall (rise) due to the surface stress. However, by 1 January 2012 both the cross- and along-shore component of wind -U(V) turns landwards (northwards), imposing a sea-level rise (fall). Sea-level appears to be influenced more by alongshore wind, where the estimated SLR follows the forcing (τV) of V. Figure 7a.2 is plotted with the estimated and measured daily mean SLR at Gopalpur, during E2 the estimated SLR is comparable and is ∼ 9 % more than the measured daily mean SLR (Table 5). It is also observed that the measured SLR remains high (∼ 15 cm) till 5 January 2012, whereas the estimated SLR falls to zero by 31 December 2011. At Gangavaram, the U (τU) V(τV) winds (wind stress) are plotted in Fig. 7b.1, where the daily mean along-shore (V) winds are observed to dominate with a range of ∼ ±10 m s-1. During E2, the measured (estimated) daily mean SLR peaks at up to 13.8 (18.5) cm; the estimate overshoots by ∼ 34 % (Table 5). At Gangavaram, the rise in SLR residual is predominantly due to high along-shore wind (-V), as explained by Vare for December 2011 which is 64 %. Also the measured daily mean SLR remained high from 22 December 2011 to 9 January 2012, whereas the estimated SLR remained high from 25 December 2011 to 2 January 2012 (Fig. 7b.2). Similarly, at Kakinada also the V winds (Fig. 7c.1) appears to dominate with peaks at up to ∼ -10 m s-1. During E2, the measured (estimated) daily mean SLR peaked at up to 22.3 (27.1) cm; the estimate overshoots by 21 % (Table 5). At Kakinada, the rise in sea level residual is predominantly due to strong southward wind (V). The measured SLR started rising above zero on 23 December, reached highest level on 29 December and descended by 9 January 2012, whereas the estimated SLR started ascending on 25 December, reached the highest level on 29 December 2011 and then dropped to zero level by 4 January 2012 (Fig. 7c.2).

Harbour oscillations (coastal seiches) as explained by Rabinovich (2009) are specific type of seiche motion that occur in partially enclosed basins (bays, fjords, inlets and harbours) and are connected through one or more openings to the sea. They are mainly generated by the long waves entering through the open boundary (harbour entrance) from the open sea. An another important property of harbour oscillations is that even small vertical motions (sea level oscillations) may be accompanied by large horizontal motions (harbour currents), resulting in increased risk of damage of moored ships, breaking mooring lines as well as affecting various harbour procedures (Rabinovich, 2009). These waves are similar to a tsunami; however, the catastrophic effects are normally observed in specific bays and inlets. Some specific sites, that have favourable conditions for the resonant generation of extreme ocean waves regularly, have been listed by Monserrat et al. (2006), Rabinovich (2009) and Joseph (2011). Similar phenomena occurs on the southern west coast of India, mainly during pre-southwest monsoon during April or May (Kurian et al., 2009).

In order to understand the harbour oscillations induced by tropical cyclones at various locations, the SLRs are high-pass filtered (time period ≤ 2 h) using a 5th order Butterworth filter (Fig. 9). The amplitude of high frequency SLR (hf-SLR) oscillations in response to E1 at Ratnagiri is ∼ ±10 cm (Fig. 9a), less at Verem and Karwar (Fig. 9b and c). The Karwar station is located in open ocean and therefore does not have the resonance features of a harbour. However, the Verem station is located in Mandovi estuary and Ratnagiri station is located in a cove and may experience resonance with meteorological disturbances. In a similar study at Verem, Mehra et al. (2012) reported the hf-SLR oscillations of ∼ ± 15 (10) cm in response to the Cyclone Yemyin (Phyan) which occurred in the BOB (the AS) during 23–25 June 2007 (9–12 November 2009). The hf-SLR oscillations at Tuticorin (Mandapam) are up to 10 (5) cm during E1 (Fig. 9d and e). Mandapam sea level gauge is located on the common boundary of Palk Strait and the Gulf of Mannar, whereas the Tuticorin sea-level gauge is located in the Gulf of Mannar (Fig. 1). The hf-SLR oscillations at the stations located in the BOB are also shown in Fig. 9. The hf-SLR amplitude due to E2 at Tuticorin, Mandapam is not observable, and at Gopalpur a brief amplitude of 10 cm is observed (Fig. 9d, e and f). At Gangavaram (Kakinada) the hf-SLR variations are ∼ ±10 (5) cm as both the gauges are located in the harbour (Fig. 9g and h), and ∼ ±4 cm at Port Blair (Fig. 9i).

Event E1 and background SLR spectra estimated at Karwar, Verem and Ratnagiri are indicated in Fig. 10. The spectrum of SLR data is obtained using “pwelch” function from MATLAB with Hamming window of 256 data points and 50 % overlap. Rabinovich (1997) proposed an approach to separate the influence of source and topography in observed tsunami spectra. The method assumed a linear tide gauge response to external forcing and is based on comparative analysis of tsunami and background spectra. This method will be used to understand the resonant influence of local topography and spectral characteristics of SLR during an event at a particular location. The data duration for estimating the spectrum of the SLR during E1 (background) is from 26 November to 1 December (1 September to 20 November) 2011. Similarly, the data duration during the event E2 (background) is 25–31 December (1 September–10 December) 2011, respectively.

The background spectra of different sites have significant differences at high frequencies as seen in Fig. 10, indicating the influence of local topography. The event spectrum at Ratnagiri is high in energy, well lifted above that of background with major peaks at 127, 80, 47, 30, 26 and 14 min during E1 as shown in Fig. 10a. At Verem the event spectra is intertwined with background spectra with peaks at 182, 91, 40 and 20 min as shown in Fig. 10b. However, the event spectrum at Verem was energetic during the cyclone Yemyin (2007), September Sumatra Tsunami (2007), and the cyclone Phyan (2009), where a distinct peak was observed at ∼ 43 min (Mehra et al., 2012). Similarly, the event spectrum during E1 (Fig. 10c) at Karwar is similar to the background with some detectable peaks at 106, 67, 44 and 21 min, and further higher frequencies are merged with the background spectra, indicating open-ocean behaviour (lack of harbour resonance). The influence of E1 is also visible at Tuticorin with dominant spectral peaks at 106, 53, 44, 24 and 18 min (Fig. 10d). However, at Mandapam (Fig. 10e) the event spectra are intertwined with background spectra with peaks at 116, 80, 42 and 26 min.

Event E2 and background SLR spectra estimated at Gopalpur, Gangavaram, Kakinada, and Port Blair during E2 are shown in Fig. 11. The event spectrum during E2 (Fig. 11a) at Gopalpur is intertwined with the background spectra with some detectable peaks at 106, 80, 60, 45, 36, 21 and 12 min. The spectral peak at 45 and 21 min are also present in the background signal. The event spectral energy at Gangavaram (Fig. 8b) is higher compared to the background SLR spectra with peaks at 213, 98, 67, 41, 25 and 18 min; however, the background spectra shows peaks at ∼ 128, 98 and 17 min. The phenomena of harbour resonance is clearly visible at Gangavaram station (Fig. 5b.1), where it is not the surge but high frequency oscillations triggered by the long waves arriving from the open ocean. At Kakinada (Fig. 11c) the spectra for lower frequencies (time period > 41 min) are similar, but the energy is enhanced for higher-frequency (shorter time-period) oscillations with time periods 41, 37, 25 and 12 min, suggesting resonance occurring in the harbour. At Port Blair, the SLR spectra of both the event and background show similar variability with event peaks at 160, 85, 47, 41, 26 and 17 min (Fig. 11d). The background SLR spectrum has noticeable peaks at 41 and 26 min. Event spectra at shorter time periods are marginally above the background spectra, which indicates the absence of harbour resonance as Port Blair station provides open-ocean conditions.

We summarise the response of sea level of the two events in the AS and the BOB. The SLR rise (fall) in the AS due to E1 reflects the winds as is also seen in the estimated SLR. The estimated SLR peak value at Ratnagiri and Verem is comparable to the measured SLR during E1 but at Karwar it is short by ∼ 50 %. The Vare accounted by local surface meteorological parameters is ∼ 69 %. Vare is small in January at all the locations of the present study (Table 4). In a similar study by Mehra et al. (2010), multi-linear regression analysis (each with a 2-month duration) was also used to resolve the dependence of sea level on various forcing parameters for 2007 and 2008 at Verem, Goa. During the summer monsoon (May–September), the sea level variability attributable to wind was up to 47 and 75 %, respectively, for 2007 and 2008; however, it decreased to < 20 % during the winter monsoon (November–February). A significant part of the variability observed in sea level remains unaccounted for and was attributed to remote forcing. In the BOB, the SLR response to E2 is of a plateau shape with rising peaks and prolonged falls during E2, which the estimated SLR is not able to capture. This persistence of high daily mean SLR state may be attributed to intensity, direction and duration of the event, the distance from the source etc. The distance of Thane (E2) track is ∼ 570 km from Gangavaram as shown in Fig. 1. The slope of the continental shelf will also affect the level of surge in a particular area. For example areas with shallow slopes of the continental shelf (as in AS) will allow a greater storm surge, and areas with deep water just offshore experience large waves, but little storm surge (SLOSH, 2003). At Gangavaram, the daily mean SLR is low (∼ 13.8 cm) compared to the other sites on the east coast of India (Table 5), and this could be attributed to distinct harbour oscillations at this location (Fig. 5b.1). Also note that when Thane crossed the Tamil Nadu coast just south of Cuddalore between 01:00 and 02:00 UTC of 30 December 2011, no distinct surge is observed at Mandapam and Tuticorin, even though Mandapam (Tuticorin) is in close proximity ∼ 237 (360) km to Thane track. The highest surges usually occur to the right of the storm track (travelling with the storm) at approximately the radius of maximum wind whereas Mandapam and Tuticorin were to the left of the track.

Storm surge is generated partly by the atmospheric pressure variations, but the main contributing factor is wind acting over the shallow water and it is an air-sea interaction problem. The basic mechanism involved in the generation of coastal surges is the influence of a long-shore wind stress, driving an Ekman transport towards the coast, causing the piling-up of water within a Rossby radius of deformation. Only the long-shore wind stress and atmospheric pressure variations associated with a cyclone generate a surge (Thomson, 1970). The surge travels along the coast as a Kelvin wave, away from the finite area of forcing, at a speed c=√(gh), where g is the acceleration due to gravity and h is the depth. Such coastally trapped motions (with the coast to the right (left) in the Northern (Southern) Hemisphere) are called forced Kelvin waves (e.g. LeBlond and Mysak, 1978; Gill, 1982). In the spectrum, the storm surges are centred about 10-4 Hz, which corresponds to a period of about 3 h (Platzman, 1971). A few parameter estimates of the E1 and E2 are listed in Table 6. To estimate the surge propagation speed, we have also added a few more locations, where hourly sea level data are available from www.gloss-sealevel.org (marked as red star in Fig. 1). Figure 12 shows the sea level response from Colombo, Sri Lanka to Jask, Iran in the Indian Ocean. Some relevant parameters of the storm surge are listed in Table 7. The average surge propagation speed is estimated to be ∼ 6.5 m s-1. The E1 moved northwards with an average along-shore speed of ∼ 6.2 m s-1, with the track almost parallel to the west coast of India. The match of the surge propagation speed of 6.5 m s-1 with that of E1 alongshore speed is evidence of a forced wave. The residual surge lagged the storm by 3, 4, 6.5 and 8.5 h (Fig. 1 and Table 7) to its nearest proximity at Colombo, Kochi, Karwar and Verem, respectively, with constant peak amplitude of ∼ 34.6 cm (Fig. 12). However, at Kochi, the development of secondary peak is clearly visible with time difference of ∼ 14 h between the two peaks of ∼ 13.5 cm. At Ratnagiri and Karachi the surge peak leads the storm by ∼ 1 h with the constant peak amplitude of ∼ 33.5 cm.

Similar response as above was observed by Fandry et al. (1984), when Cyclone Glynis moved slowly and almost parallel to the western coast of Australia in February 1970. In this event, a strong coastal peak travelled down the coast well ahead of the cyclone. In this example μ>UmfC and Vm<c (wheref is Coriolis parameter, Um (Vm) is eastward (northward) velocity of cyclone and μ is decay rate, and theory predicts a coastal peak of constant amplitude moving ahead of the cyclone. In their study, they characterised the sea level response to tropical cyclone as follows: a.

μ<UmfC: coastal peak of increasing in magnitude propagating along the coast with speed Vm. The maximum peak occurs at the edge y=Vmt which is the leading (trailing) edge if Vm>c(Vm<c). Where y is alongshore axis and positive towards north, t is time.

In the present study, our observations indicate that the surge peak lagged E1 up to Verem. At Ratnagiri to Karachi the surge peak leads the E1 with an almost constant amplitude (Table 7). Also note that the surge amplitude is almost constant (∼ 34 cm) at all the four locations from Karwar to Karachi. The impact of E1 in the BOB at Port Blair (Fig. 3i) and in the northern parts of east coast of India at Kakinada, Gopalpur and Gangavaram (Fig. 3f–h) is not observable. The response of sea level due to E1 at Masirah (Fig. 12k) is also not observable as the location is on the right hand side of the event track. Similarly, the sea level variations at the island locations of Minicoy and Hanimaadhoo are negligible due to E1 (Fig. 13), even though the track of E1 is only ∼ 170 and 280 km away, respectively. However, the absence of closed boundary at these Island locations and their locations on the left side of E1 track imply that no surges are predicted. The impact of the E1 is observed only in the AS at the coastal boundary located towards the right side of the event track.

It is being realised increasingly that a near real-time network of sea level and surface meteorological measurements along the coastal and Island locations of India such as ICON (http://inet.nio.org/) established by CSIR-NIO could play an important role in improving the operational (routine) predictions of coastal flooding and enable to understand the fundamental dynamics of these events. Presently, there are only a few mesoscale weather and sea level networks in some coastal segments of the Indian and eastern Atlantic oceans to observe such events. It is also expected that this kind of relatively inexpensive and simple networks, similar to the one in-house developed and established by CSIR-NIO will be affordable to limited-budget institutions in their natural hazard mitigation efforts.

This study attempts to investigate the meteorologically induced surges and water level oscillations along the select locations in response to the passage of the storm (E1) in the Arabian Sea and “Thane (E2)” in Bay of Bengal. The high frequency water level oscillations observed, such as at Gangavaram during the events are found to be due to the result of harbour resonance. Each location has a typical event spectrum related to the local topography. For example, Rabinovich (1997) observed two prominent peaks in Dimitrova Bay. The first peak with a period of 50 min related to the amplification of incoming waves over the shelf of the southern Kuril Islands (i.e to the shelf resonance). The second peak with a period of 18.5 min is caused by the standing oscillation of the bay itself. Some of the locations in the present study showed SLR oscillations around periods of ∼ 92, 43 and 23 (100, 42 and 24) min in the AS (the BOB). However, influence of local topography is clearly noticeable at shorter time periods at different stations. During E1, even though the local winds are small in magnitude (4–10 m s-1), the surges are 39–47 cm at Ratnagiri, Verem and Karwar with similar peaks in SLR. In this case, the cyclone track is parallel or at an oblique angle to the coastline and hence the surge occurs on longer stretches of the shoreline. Whereas, during E2 (Thane) the cyclone makes a perpendicular approach to the coast at landfall, and therefore the surges are low at Mandapam and Tuticorin, as these stations are on the left side of the Thane track, as compared to the stations on the right side (Gopalpur, Gangavaram and Kakinada). SLR at Gopalpur, Gangavaram and Kakinada remained high for ∼ 9 days even after the land fall of Thane. The residual sea levels from tide gauge stations along the west coast of India and the coast of Pakistan showed surge peak of constant amplitude propagating northwards with a speed of ∼ 6.5 m s-1 during E1. The propagating surges along the western coast have been identified as forced Kelvin waves with almost constant amplitude at Karwar, Verem, Ratnagiri and Karachi. The multi-linear regression using local daily mean winds (cross- and along-shore) in association with the the atmospheric pressure is able to account for up to 69 % of daily mean sea level residual (SLR) at Ratnagiri, Verem and Karwar during E1. However, in the BOB up to 57 % of daily mean SLR is accounted for at Gopalpur, Gangavaram and Kakinada.