Future projections of storm surges and associated coastal inundation along the east coast of India

The tropical cyclones (TCs) that make landfall (hereafter LF) along the Indian coasts have immense socio-economic implications. A TC is a rotating storm system characterized by low atmospheric pressure, strong winds, and heavy rain. The occurrence of TCs over the Bay of Bengal (BoB) and Arabian Sea (AS) continues to be of great concern to India's coasts. According to the study by Sahoo & Bhaskaran (2015, 2016) and Knutson et al. (2021), more than a 100 cyclones have hit the Indian coasts during the last five decades. Coastal storm surge, an abnormal raise in sea level at the coast due to an approaching tropical cyclone, is the prime hazard associated with the landfalling cyclone (Srinivasa Kumar et al. 2015; Murty et al. 2016, 2017; Luettich 2018). Therefore, accurate prediction of the cyclone tracks and intensity is vital for the precise estimation of storm surge amplitudes (Bhaskaran et al. 2014; Srinivasa Kumar et al. 2015; Murty et al. 2017). Understanding the stochastic nature of extreme storm surges and their effects on coastal locations is crucial for efficiently designing coastal protection structures and planning for future coastal adaptations (Arns et al. 2013).

A comprehensive study of storm surge projections considering climate change impacts is crucial for disaster preparation and future coastal infrastructure development activities. As per the literature survey, Needham & Keim (2012) generated a storm surge database, SURGEDAT, for the Gulf Coast of the United States, employing over 3,000 government and public sources of information since 1880. Kirezci et al. (2020) analyzed the projections of global-scale extreme sea levels and associated coastal flooding over the 21st century. Sahoo & Bhaskaran (2018) studied storm surges and associated coastal inundation along the east coast of India (ECI), but this study is location-specific and concentrates on the Odisha coast, and the future cyclone tracks were generated considering the worst-case scenarios. Rao et al. (2020) studied the impact of climate change on storm surges and associated inland inundation, but the generated tracks are completely ideal and do not adhere to the actual variation of wind intensity at different time steps. These tracks represent the worst-case scenario and may lead to significant overestimation. To the best of the author's knowledge, a detailed analysis of the entire ECI is still lacking.

The present study attempts to fill that gap through a comprehensive analysis of storm surges and associated coastal inundation along the entire ECI using historical and synthetic cyclone tracks over BoB. Inundation is the phenomenon of flooding normally dry land portions due to coastal storm surges. The current study concentrates on analyzing future storm surge characteristics by generating synthetic cyclone tracks based on historical cyclonic tracks. These synthetic tracks represent the future projections of the cyclone tracks. Assessments of projected coastal storm surges and associated inundation are critical in forming policy directions. Such assessments can also identify local hot spots, which demand more detailed modeling.

The historical cyclone tracks can be obtained from a variety of sources. The India Meteorological Department (IMD), which is part of the Ministry of Earth Sciences (MoES), has created a comprehensive electronic Atlas (eAtlas) that contains the tracks of cyclones and depressions over the northern Indian Ocean, covering the period from 1891 to till date. In Indian Ocean cyclones, IMD and the Joint Typhoon Warning Center (JTWC) maintain historical track details, available at https://rsmcnewdelhi.imd.gov.in/report.php?internal_menu=MzM= and http://www.usno.navy.mil/JTWC/. The best track cyclone data from JTWC are available for the Indian Ocean since 1972. Additionally, the World Meteorological Organization (WMO) has recognized the International Best Track Archive for Climate Stewardship (IBTrACS), which is maintained by the National Oceanic and Atmospheric Administration (NOAA). The National Environmental Information Centers (https://www.ncdc.noaa.gov/ibtracs/) provide the best track of TC in a centralized location to understand cyclone distribution, frequency, and intensity. Regional Specialized Meteorological Centers (RSMCs) and other international centers partnered with the IBTrACS to develop a global best track data set that combines storm information from multiple centers into a single product and archives the data for public use (Knapp et al. 2010).

These historical track data give a good picture of past climatology. For future TC projections, the best option is to generate synthetic TC track. The historical tracks can be utilized to generate synthetic cyclone tracks. Based on theory and high-resolution dynamical models, future predictions show that global warming will lead the mean strength of TCs to shift toward greater storms, with an intensity increase of 2–11% by 2100 (Duan et al. 2018). While developing future TC tracks, synthetic track models will have the ability to incorporate the influence of climate change. Two approaches are available in the literature to produce synthetic TCs: the simple track model (STM) and the empirical track model (ETM). The first one developed to create synthetic cyclones was the STM (Vickery & Twisdale 1995). The central concept of the STM (Vickery & Twisdale 1995) is to obtain certain observable TC features and utilize them to build probability density functions. These characteristics are then sampled from their distributions using Monte-Carlo simulations and passed along the track that does not vary, ensuring that TC properties remain constant along the track. The disadvantage of this approach is that it is highly site-specific. ETM is, in theory, is the progression of STM (Vickery et al. 2000). It uses a similar approach of the congregation of statistics and then sampling them using Monte-Carlo simulations. Rather than sampling all parameters once, the variables’ properties are varied at each step along the track.

Numerous synthetic TC databases and methods have lately been published in the literature. Vickery et al. (2000) produced a considerable number of synthetic storms in the North Atlantic (NA) basin using statistical features of historical tracks and intensities. Changes in TC attributes over 6 h were modeled as linear functions of preceding values of those variables, position, and sea surface temperature (SST). James & Mason (2005) applied a similar but relatively more superficial and less data-intensive method. Arthur (2021) has used a procedure like that of James & Mason (2005) but engrossed in the whole continent of Australia. Vickery et al. (2009) used the second stage in the TC generation process by comprising thermodynamic and atmospheric environmental factors. Nederhoff et al. (2021) used the Markov chain method-based ETM. Based on a given historical data source, it can simulate synthetic TC tracks and wind fields in each ocean basin for the user-specified number of years, with and without considering climate change.

In the present work, past five decades, TC tracks are collected and using these TC tracks, synthetic tracks are generated for the future 100 years considering the climate change scenario. Numerical simulations are conducted for past and future TC tracks to estimate storm surge and inundation in the past and future, and a comprehensive analysis is carried out. The subsequent sections provide more details on the data and methodology, modeling system, results, discussion, and conclusions.

Two main aspects dealt in the present study are (1) Obtaining historical cyclone tracks and generating synthetic cyclone tracks considering the impact of climate change on TC intensity. (2) Numerical simulations of storm surges, generation of storm surge, and coastal inundation maps for comprehensive analysis.

Historic cyclone track data are collected from the JTWC Best Track database, IMD best track data, etc. Multiple track data are collected, compared, and checked with each other; a database of historical tracks to the maximum possible continuous duration is obtained by combining multiple source data for the BoB. The past five decades, i.e., from 1970 to 2020, historical tracks are considered in the current study. The track data have geographical coordinates of cyclone eye location (latitude and longitude), maximum wind speed; central pressure; the radius of maximum winds, etc., at every 6-h interval. If the data are missing at any timestamp, interpolation techniques were used to get the track to the required 6-h interval.

Various synthetic track generation models (Bloemendaal et al. 2020; Nederhoff et al. 2021) are available in the literature. In the current work, we have chosen a model described by Nederhoff et al. (2021). Their group has developed the Tropical Cyclone Wind Statistical Estimation (TCWiSE) toolbox and made it publicly free and open source. TCWiSE is adaptable based on the user's requirements and preferences. A thorough description of the TCWiSE is available at (Nederhoff et al. 2021). TCWiSE gives the option of considering the impact of climate change on the synthetic track and intensity. The factors that control the impact of climate change are defined based on the expert assessment of TC climate predictions (Knutson et al. 2015). The current study utilized TCWiSE to generate synthetic tracks over BoB for the next 100 years, considering the impact of climate change with the past five decades of historical cyclone tracks as input.

ADvanced CIRCulation (ADCIRC) is used for storm surge simulations. Storm surge simulations are performed using historical and synthetic tracks. These tracks are used to generate the cyclonic wind and pressure fields which are the primary forcing to the model. The moored buoy recorded wind data are used to validate the modeled wind speed. The peak storm surge height over the entire simulation at each model grid point is obtained for each simulation, and this peak value is called Maximum Envelopes of Water (MEOW) (Rygel et al. 2006; Kleinosky et al. 2007; Frazier et al. 2010; Maloney & Preston 2014). The MEOWs for each simulation form a composite dataset called the Maximum of MEOWS (MOMs) (Rygel et al. 2006; Kleinosky et al. 2007; Frazier et al. 2010). The MOMs give the composite picture of potential storm surge elevation and associated coastal inundation spatial distribution due to TCs. The MOMs are generated separately for the past using historical tracks and for the future using synthetic tracks. Tide gauge data along the ECI are collected to validate the computed storm surge heights at the tide gauge locations. Further brief details on the numerical model are given in the subsequent section.

Storm surge modeling mainly requires surface wind fields and atmospheric pressure as forcing at each time step in the simulations. Thus, the most critical challenge for accurate storm surge computations is obtaining errorless meteorological forcing (Fleming et al. 2008). The global meteorological models cannot provide accurate wind fields and underestimate the TC peak intensity (Fleming et al. 2008). In contrast, the parametric wind models produce better wind fields and are simpler to use for storm surge computations (Houston et al. 1999; Mattocks et al. 2006). The parametric wind models need cyclone eye location, cyclone maximum wind speed, and regional domain to generate the wind fields at a given timestamp. In this context, in the current study, the recent version of the Holland wind model (Holland et al. 2010) is used to generate wind fields as atmospheric forcing to the storm surge model.

Modeled storm surge heights are verified using the available observations from tide gauges along the ECI. The tide gauge network used in the study is shown in Figure 4. The tidal contribution is removed from the tide gauge records, i.e., the observations are de-tided. The total water elevation is primarily controlled by the amplitude and phase of the tide at the time of LF, mainly where the tidal ranges are high. However, tides are not considered in the current work because the time of LF of the future cyclone projections is not known as a priori (in contrast with historic tracks). De-tiding the surge heights gives the information of water level that is merely due to the cyclone's wind contribution, i.e., residual storm surges. The short wave contribution is also not included in the simulations because it requires coupled ADCIRC + SWAN model, which is computationally expensive.

Furthermore, the present model mesh is very fine, with over one million grid points, and we had to simulate more than 200 simulations. Hence, we have chosen ADCIRC alone for the model simulations with the minimum available computational resources. The simulated storm surges and inundation extent using ADCIRC alone reached the maximum value for the next ten decades. Hence, superposing the small value (compared to the surge height) of the wave setup will not significantly differ as we are providing the composite analysis. Model simulations were performed in parallel mode on the Mihir high-performance computing (HPC) system at National Center for Medium-Range Weather Forecasting (NCMRWF), India. Mihir HPCS is a Cray-XC40 Liquid Cooled System with 2,320 nodes running with a peak performance of 2.8 PF and a total system memory of 290 TB (Mamgain et al. 2018). Each model simulation took around 19 min using 360 processors for an average simulation length of about 120 h.

The main cause of destruction due to a coastal storm surge is its associated inland flooding which is potentially hazardous and can substantially impact any coastal area. Coastal flooding triggered by storm surges is as devastating as the wind and poses a significant risk to life and property along the coast. Hence, to minimize storm surge damage, a realistic inland inundation estimation is as important as the height of the storm surge (Bhaskaran et al. 2014). The coastal inundation parameter is crucial for improving cyclone preparedness and optimizing evacuation scenarios. In the current study, we have also simulated the coastal inundation along the ECI for all the historical and synthetic tracks. To accurately estimate the threat posed by coastal inundation, the inundation depth must be calculated (in addition to the horizontal inundation extent) at each inland grid point over the model domain. Inundation depth is calculated by subtracting the local grid point topography value from the run-up height at the respective points. Run-up height is the storm surge height above mean sea level (MSL) at its farthest point inland. Run-up height and topography values are with reference to the MSL, whereas the inundation depth value is with reference to the local topography value. For example, if the run-up height at some location is 9 m (w.r.t. MSL) and the topography value at the respective location is 5 m (w.r.t. MSL), then the inundation depth at the same location will be 9–5 m = 4 m. Hence, the inundation depth will determine the possible local hazard severity due to a storm surge. Horizontal inundation extent at any coastal location is the distance (normal) from that coastal location to the last wet cell.

It should be noted that these gentle slope regions are also covered by major river deltaic regions along the ECI, through which water gets inundated through these channels to a more considerable distance. During historical and future projections, the inundation extent reaches greater distances over common portions with gentle slopes from 0 to 8 m land contour (central TN, southern and central AP, northern OD, and almost the entire WB). However, the same is further large in future projections, which clearly warns that these parts of the ECI might be under significant threat due to future storm surge projections by 2100 (MoWR RD&GR CWC GEF ADB-TA8652 IND 2019). At some locations, the inundation extent reaches beyond 30–40 km. These greater inundation extents could be due to the high storm surges that penetrated river channels and inundated the portions along the riverbanks. Along the OD and WB coasts, the greater inundation extents are seen in historical and future storm surge maps, whereas the inland penetration of seawater is increased by two-fold to three-fold along the TN and AP coasts due to future storm surges. Hence, TN and AP coasts might be significantly affected due to future storm surges in changing climate scenarios. Hence, in the face of changing climate, the future storm surge hazard will likely worsen (Khan et al. 2022; Leijnse et al. 2022). From Figure 14, minimal or no inundation is observed along the northern parts of AP and southern OD coasts during both historical and future projections, and the leading cause is the high steepness along these parts. Hence, north AP and south OD coasts are under low or no threat due to future storm surge induced coastal inundation by 2100.

The coastal belt encompassing the ECI experiences storm surges and inland flooding associated with landfalling cyclones almost every year. Prior studies for this region are very location-specific, and comprehensive research on TC-induced storm surges for the entire ECI is still lacking. Storm surge characteristics along the ECI are analyzed using the historical and synthetic cyclone tracks. Using the past five decades of historical TC tracks as input, the synthetic tracks are generated using the TCWiSE tool for the next 100 years while considering the impact of climate change. A thorough analysis of TC tracks reveals that most of the coastal locations along the ECI will experience high LF intensities (wind speeds >44 m/s). OD and AP will experience more cyclones in the future. The inclusion of climate change impact results in about 12% increase in TC peak intensity in the projections of future TC tracks. The LF time intensities of future tracks exhibit a comparable pattern, leading to increased storm surge heights throughout the ECI. Composite map analysis reveals a significant rise in peak storm surge heights and subsequent coastal inundation along the ECI. The increased storm surge heights and associated inundation extent are due to the increased TC frequency and intensity in synthetic tracks due to climate change. It can also be observed that the coastal locations falling in low- and moderate-risk zones are decreased, and high-risk zones are significantly increased from historical to future projections. Hence, the above observation reveals that most parts of the ECI are highly vulnerable due to future cyclones.

The current study benefits academic, scientific, and operational users and disaster management officials. This study can also assess the risks of vital facilities and infrastructure to assess potential economic and insured damages. The authors hope the current study will educate the readers on the risk of storm surges and helps in mitigating the loss of life and property in future cyclone events. The present study has vast potential and practical applications in coastal zone management, risk evaluation, and preparedness planning for the ECI. The outcome of the study will allow new and novel analysis for the study region that could not have been done before.

The authors take this opportunity to thank the Ministry of Earth Sciences (MoES), Government of India, for the extended support in carrying out the work. Thanks also to the ADCIRC developing team for sharing the code.

All relevant data are available from an online repository or repositories. GEBCO bathymetry topography used for meshing can be downloaded from: https://www.gebco.net/data_and_products/gridded_bathymetry_data/; IMD best track data can be found at: https://rsmcnewdelhi.imd.gov.in/report.php?internal_menu=MzM=; JTWC best track data can be found at: https://www.metoc.navy.mil/jtwc/jtwc.html?north-indian-ocean; TCWISE toolbox can be obtained from: https://download.deltares.nl/en/download/tcwise/

The authors declare there is no conflict.