17 Deb and Ferreira.Pdf

Home , 1991 Bangladesh cyclone, Cyclone Sidr

Journal of Hydro-environment Research xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Journal of Hydro-environment Research

journal homepage: www.elsevier.com/locate/JHER

Research papers Potential impacts of the Sunderban mangrove degradation on future coastal ﬂooding in Bangladesh ⇑ Mithun Deb a, , Celso M. Ferreira a a Department of Civil, Environmental & Infrastructure Engineering, George Mason University, Fairfax, VA 22030, USA article info abstract

Article history: The coastal areas of Bangladesh are recognized by the United Nations (UN) as the most vulnerable areas Received 30 June 2015 in the world to tropical cyclones and also the sixth most vulnerable country to floods around the world. Revised 27 July 2016 Cyclone Sidr (2007) was one of the most catastrophic natural disasters in Bangladesh causing nearly Accepted 14 November 2016 10,000 deaths and $1.7 billion damage. During cyclone Sidr, mangrove forests in coastal areas played a Available online xxxx crucial role in the mitigation of these deadly effects. Sunderban mangrove, the world’s largest mangrove ecosystem with 7900 sq. miles, forms the seaward frontier of the bay and is now facing significant degra- Keywords: dation. The Sunderban mangrove ecosystem is increasingly being degraded for a variety of purposes such Bangladesh as agriculture, fishing, farming and settlement. In this study, we evaluate the potential impacts from the Sunderban mangrove Cyclone degradation of the Sunderban mangrove on storm surge flooding. We evaluate two hypothetical and Storm surge extreme scenarios: 1) the conversion of the entire mangrove land cover to an estuarine forested wetland; Coastal hazard and 2) by considering a full degradation scenario where the entire mangrove is converted to grassland. To ADCIRC quantify the benefits of the mangrove forests to attenuate storm surge in this area, we applied a framework combining a spatial characterization of mangroves vegetation with numerical simulations. Storms surge was calculated using a hydrodynamic model (ADCIRC) coupled with wave model (SWAN) under a High Performance Computing environment. An unstructured numerical mesh with 200,000 nodes was developed and validated along with the coupled SWAN + ADCIRC model at six separate locations in the Bangladesh coast using Cyclone Sidr (2007) meteorological inputs. Twenty-seven model simulations were performed considering nine cyclones of different categories to quantify the effects of mangrove degradation on spatial flood inundation and storm surge magnitude. Simulation results showed that, on average, the mangroves degradation to grassland could raise the surge elevation as high as 57% and had a significant impact on increasing the velocity of the flood wave by up to 2730% for category 3 cyclones. In addition, the inundation inland penetration and total flooded area would increase almost 10 km and 18% respectively for low intensity cyclones. Furthermore, these hypothetical scenarios support the importance of the existing Sunderban mangrove in the reduction of surge elevation, velocity, inundation penetration and flooded area. More importantly, it also demonstrates how the continuous degradation of this important ecosystem has the potential to adversely impact the future cyclone induced hazards in the region. Ó 2016 International Association for Hydro-environment Engineering and Research, Asia Pacific Division. Published by Elsevier B.V. All rights reserved.

1. Introduction the low-lying heavily inhabited areas and a continental shelf with shallow bathymetry leaves the entire coastal region of Bangladesh Cyclone surges in Bangladesh have caused deaths of over very vulnerable to catastrophic inundations along the coast (Murty 700,000 people since 1960 (Chowdhury and Karim, 1996) and it et al., 1986; Dube et al., 1997; Madsen and Jakobsen, 2004). Almost is still considered as one of the most damaging meteorological phe- one sixth of tropical cyclones that developed in the Bay of Bengal nomena in the region. The exposure to storm surge combined with made landfall on the Bangladesh coast (Islam et al., 2011); however, the death toll observed to be 80% of the global record (Debsharma, 2007). Numerous historic super cyclones, such as ⇑ Corresponding author at: Center for Applied Coastal Research, University of the 1970 Bhola cyclone, the 1991 Bangladesh cyclone and the Sidr Delaware, Newark, DE 19716, USA. 2007 cyclone had catastrophic effects in the coastal areas of E-mail addresses: [email protected], [email protected] (M. Deb), [email protected] (C.M. Ferreira). Bangladesh (Islam et al., 2011; Madsen and Jakobsen, 2004). http://dx.doi.org/10.1016/j.jher.2016.11.005 1570-6443/Ó 2016 International Association for Hydro-environment Engineering and Research, Asia Paciﬁc Division. Published by Elsevier B.V. All rights reserved.

The Sundarban mangrove offers coastal protection to millions of characterized as mangroves to an estuarine forested wetland and people in Bangladesh and lies in a zone facing cyclonic storms and subsequently to grassland. These hypothetical scenarios are based tidal bores from the Bay of Bengal periodically. The population on current trends of deforestation and situations already observed density surrounding the Sunderban mangrove is a significant fac- elsewhere in the world, for example, Colombia’s Caribbean coast tor, which made it a unique location for vulnerability related mangroves were converted into agricultural lands and shrimp assessments. During cyclone Sidr, mangrove forests in coastal aquaculture ponds, and in Quintana Roo (México), deforestation areas played a crucial role in the mitigation of the deadly effects of mangroves changed the land cover to low-density human settle- from coastal flooding. It has been well demonstrated that the man- ments and roads (Blanco et al., 2012). Afterwards, we used cate- groves play an important role in coastal defense and disaster risk gory 5 cyclone (Sidr 2007) modifying from its original track, and reduction as they reduce storm surge water levels by slowing the hypothesized about two additional cyclone categories 3 and 4 to flow of water, and reducing surface waves and wind stress observe the influence of storm attributes on inundation over the (McIvor et al., 2012). Several studies in recent past were conducted altered land covers. Then, the simulation results such as maximum to document the role of mangrove forests in the dissipation of water level, velocity, inundation distance and flood area were com- storm surges or tsunami inundation (e.g. Zhang et al., 2012; pared at different locations. Hiraishi and Harada, 2003; Yanagisawa et al., 2010). McIvor et al. As the coastal regions of Bangladesh are low-lying areas, the (2012) suggested that storm surge reduction through mangroves impact from mangrove degradation would certainly trigger wide- range from 5 to 50 cm of water level reduction per kilometer of spread damages. Surge elevation, velocity and inundation area mangrove width. However, the surge reduction rate depends on would increase largely and affect the existing dense population many significant contributing factors such as storm surge eleva- around the coast. The main goal of the analysis is to demonstrate tion, topography, distance to the shoreline and size of mangrove and quantify the impact that such drastic land cover changes could forests (Liu et al., 2013). Krauss et al. (2009) recorded water level potentially have on amplifying surge related hazards in the Sun- data in the mangrove zone of the South Florida coast during Hurri- derban mangrove region. canes Charley (2004) and Wilma (2005) and reported that mangroves reduce storm surge heights approximately 9.4 cm/km. 2. Study area However, the world’s largest mangrove forest, Sunderban, is facing a significant degradation. It occupied approximately twice The Sunderban mangrove lies on the delta of Ganges, Brahma- its current extent at the beginning of the colonial era in India putra and Meghna Rivers on the Bay of Bengal (Fig. 1). It is inter- (Islam et al., 1997). Globally, the area of mangroves has reduced connected by a complex network of tidal waterways, mudflats by almost 20% between 1980 and 2005 (Barros and Albernaz, and small islands of mangrove forests and the floor varies from 2014) and continues to decrease at a rate of 1–2% annually (Duke 1.5 to 3.0 m from the mean sea level (e.g. Karim and Mimura, et al., 2007). More than 50% of the country’s forests were extin- 2008). The Sundarbans was formerly (200 years ago) measured to guished in the last 30 years (MOEF, 2008; Roy et al., 2012). Other be of about 16,700 km2, and currently has reduced to an area of studies showed that 50% of the total tree cover has vanished over 10000 km2. Sixty-two percent of the total mangrove falls within the past 20 years (e.g. Kabir and Hossain, 2008). Population density the territory of Bangladesh and western boundary borders with close to Sundarban mangroves is among the highest in the world the Indian Sunderban (Islam and Wahab, 2005). The forest meets (Giri et al., 2007). Nearly 10 million people of the coastal regions the Bay of Bengal on the south side; to the east it is surrounded are relying on the Sunderban mangrove directly or indirectly, for by Baleswar River and agricultural lands on the north side (e.g. a variety of purposes such as agriculture, fishing, farming, human Wahid et al., 2007). The Sundarban has a unique biodiversity with settlement, collection of housing materials and human foods and a wide range of flora and fauna and the dominant mangrove Heri- employment opportunities in forestry practices (e.g. Islam, 2005). tiera fomes, locally known as Sundri or Sundari (e.g. Bhowmik and Shrimp farming is one of the major and perhaps the most detri- Cabral, 2013). Average tree height and bole diameter of the Sun- mental activity, which is playing a critical role in rapid mangrove derban mangrove vegetation varies from 5.81 to 17.55 m and deforestation in Bangladesh for the last two decades (Islam, 2005). 25.70 to 54.35 cm respectively (e.g. Abdul, 2014). The Sunderban In addition, the mangrove ecosystems are largely sensitive to mangrove has been guarding the coastal community of Bangladesh sea-level rise (SLR) as they originate in the transition zone between for a long time and acting as a vegetative shield in the direction of land and ocean in tropical and subtropical coasts (e.g. Lin, 1997; storm surges that helps to minimize the adverse impacts. Approx- Ellison, 2012). SLR along the coast of Bangladesh is much higher imately 2500 sq. km or 25% of the total area of the Sundarban man- than the global rate of 1.0–2.0 mm/year (Karim and Mimura, grove forest was impaired by cyclone Sidr, 2007 (CEGIS 2007). 2008) and apparently is an alarming threat to the Sunderban man- Subsequently, the devastating effects from Sidr carried out grove ecosystem. Based on Intergovernmental Panel on Climate destruction of the agricultural crops, fisheries, and other household Change (IPCC) reports and available SLR studies, the National assets, which drastically reduced livelihood earnings over both the Adaptation Program of Action (NAPA) for Bangladesh suggested short and long-term in the Sunderban area (Shamsuddoha et al., SLRs of 14, 32 and 88 cm for the years 2030, 2050 and 2100, corre- 2013). Moreover, the geomorphological characteristics, such as spondingly (Karim and Mimura, 2008). Along with the human very shallow estuary and marginally higher elevation of the man- raised damaging activities, SLR also will contribute to the degrada- grove from mean sea level made it an extremely vulnerable place tion of the current mangrove ecosystem. for coastal flooding in the Bay of Bengal region. In this study, to evaluate the potential impacts of the Sunderban mangroves degradation to storm surge flooding in the region, we applied a framework combining a spatial characterization of man- 3. Methodology groves vegetation with numerical simulations of waves and storm surge hydrodynamics. While there are studies promoting the 3.1. Hydrodynamic modeling implementation of mangrove or wetlands as a natural shield against storm surges, this research distinctively projected the Numerical models have been developed previously to simulate potential impact from mangrove losses in amplifying storm surges storm surges induced by cyclonic storms on the coast of Bangladesh specifically for the Sunderban region. We developed two hypothet- (e.g., Flather and Khandoker, 1993; Flather, 1994; Roy, 1995; Henry ical degradation scenarios by converting the land cover currently et al., 1997). Nowadays, depth-averaged two-dimensional (2D)

Fig. 1. Sunderban mangrove ecosystem in the coast of Bangladesh. hydrodynamic models are commonly used to solve the shallow- near shore complex coastal hydrodynamics with mesh size smaller water continuity and momentum equations and analyze storm in the coastal land-water transition zones and estuarine areas, and surges in the Bay of Bengal (e.g., Madsen and Jakobsen, 2004; larger away from the coast (e.g. Bhaskaran et al., 2013). For this Dube et al., 1994; Bhaskaran et al., 2013, 2014; Murty et al., analysis, global topographic and bathymetric datasets were used 2014). In this study, we applied the 2D coupled hydrodynamic to generate the numerical mesh for SWAN + ADCIRC simulation and wave model ADCIRC + SWAN (Dietrich et al., 2011) for our and collected from the freely available General Bathymetric numeric modeling of cyclone induced surge simulations. ADCIRC Chart of the Oceans (GEBCO) database (Bhaskaran et al., 2013). is a finite-element hydrodynamic model that is widely used for The GEBCO consists of a 30 arc-second (900 m) resolution dataset storm surge modeling in the east coast of United States and Gulf generated by joining quality-controlled ship depth soundings of Mexico, which can generate water levels and current velocities (Henstock et al., 2006) and land data based on the 3 arc second (e.g., Westerink et al., 2008; Bunya et al., 2010; Dietrich et al., (90 m) resolution Shuttle Radar Topography Mission (SRTM30) 2011; Ferreira et al., 2014a). The wave model SWAN follows a fully (Jarvis et al., 2004)(Fig. 2). The earlier Bay of Bengal numerical implicit finite difference method that calculates surface waves in meshes had a domain restricted to the vicinity of the coast only the nearshore region solving wave action balance equation (e.g. (Flather, 1994; Salek, 1998), and were not extended to inland. Zijlema, 2010). All of our results have demonstrated combined Roy (1995) has defined a model domain for the Bay of Bengal of results, where ADCIRC supplies calculated wind speeds, water about 1.4 1.3 sq.km resolution for shallow water zones to con- levels and currents to SWAN at a given time step, and SWAN applies sider small islands of the estuary and 22.2 21 sq.km of coarse the given information of water levels and ambient currents to resolution of the outer ocean boundary with 10,612 mesh points. recalculate water depth and wave processes (wave propagation, Furthermore, in a recent study, Mashriqui et al. (2010) has estab- depth-induced breaking, etc.). Comprehensive information about lished an unstructured finite element mesh including more than the fundaments and complete set of equations related to ADCIRC 363,399 elements and 186,981 nodes for the entire Bay of Bengal, and SWAN can be found in Luettich and Westerink (2004) and with the finest and largest element size of 150 m and 4 km respec- Booij et al. (1999) respectively. The 2D ADCIRC + SWAN coupled tively. However, the mesh domain focused mainly on the Ganges model solves both wave and circulation interactions on the same delta and vicinity, and did not include the mainland. In contrast, unstructured mesh (e.g. Dietrich et al., 2011). the model domain introduced in the present analysis contains the entire coast of Bangladesh uniquely focusing on coastal main- 3.2. Mesh generation land, mangrove and low-lying Ganges delta with a fine resolution of 300 m (Fig. 2). This unstructured numerical mesh comprises of A numerical mesh with high resolution in shallow water zones almost 200,000 nodes and 400,000 triangular elements that can is the most critical factor in storm surge computations (Blain et al., represent the estuarine rivers and entire area of Sunderban 1994). An unstructured mesh can be very effective in solving the mangrove. Small narrow channels inside the mangrove zone were

Fig. 2. a. Numerical mesh domain b. Modiﬁed GEBCO topo-bathymetry of the lower Bangladesh coast and Sunderban mangrove area (in meters) c. Mesh resolution (in kilometers).

difficult to recognize with the available bathymetric data and thus tion parameters from the GLC2000 (Fig. 3). Land cover classes from neglected. Bathymetry of the some near shore water bodies, such GLC2000 such as, mangroves and irrigated agricultures were con- as rivers and channels where the width is greater than 300 m, sidered as woody wetlands and cultivated crops correspondingly has been extracted from local riverine studies because of the lower to match the NLCD 2001 classification (Table 1). resolution in GEBCO data (e.g., Ali et al., 2007; Lewis and Bates, 2013). Consequently, the GEBCO database provided unrealistic ele- 3.3.1.2. Scenario 1: degradation to estuarine forested wetland. Because vation information about the mangrove zone, where it varied of the alarming mangroves degradation in the Sunderban area for between 5.0–10.0 m. SRTM datasets are not efficient in forested last few decades due to deforestation, shrimp farming and sea level areas (Lewis and Bates, 2013), hence, we assumed the elevation rise, we anticipate a significant change in the existing land cover in for Sunderban mangrove to be about 1.5–3.0 m from the Mean future. Initially, we considered a gradual degradation of the existing Sea Level (MSL) (e.g. Karim and Mimura, 2008). mangroves by assuming that the entire mangrove area is trans- formed into an estuarine forested wetland. The conversion is carried 3.3. Land cover scenarios and parametrization out by reclassifying all the existing mangrove pixels to estuarine forested wetland. Thereafter, the newly generated land cover is uti- 3.3.1. Land cover scenarios lized to produce a different set of friction parameters. Land cover plays a significant role in the forcing and dissipation mechanisms of storm surges and it has been demonstrated that 3.3.1.3. Scenario 2: degradation to grassland. To account for an land cover such as wetlands and mangroves can effectively impact extreme degradation scenario, in scenario 2 we converted the hurricane storm surges responses (e.g., Zhang et al., 2012; Liu et al., entire mangroves land cover into grasslands, speculating an ulti- 2013; Barbier et al., 2013; Ferreira et al., 2014b). In this study, we mate scenario of the current land cover degradation. Similarly, to are simulating the degradation of the present Sunderban man- scenario 1, the conversion is carried out by reclassifying all the groves in the future by altering the existing land cover to estuarine existing mangrove pixels to grasslands. forested wetland and grassland. Changes in surge behavior with respect to the land cover alteration will illustrate the role of the 3.3.2. Incorporating land cover in the ADCIRC model existing Sunderban mangrove in surge attenuation for Bangladesh. Detailed representation of vegetation (i.e. mangroves or wetlands) in hydrodynamic models could increase the precision of 3.3.1.1. Current land cover condition. Land cover dataset of 1 km res- inundation extent and duration assessments (McIvor et al., olution from the Global Land Cover 2000 Project (GLC2000) was 2012). It can lessen the surge amplitude and flow velocity influenc- collected to look at the present land cover condition of the Sunder- ing the bottom friction force and wind stress (Liu et al., 2013). The ban mangrove area. Land cover maps were generated using daily forcing mechanisms impacted by land cover are represented in the data collected from the vegetation sensor on-board of earth obser- SWAN + ADCIRC model by computing its effect on the bottom fric- vation satellite SPOT 4, and land cover categories recognized were tion and changes in the transfer of momentum transmitted to the then generalized into GLC2000 (Bartholomé and Belward, 2005). water column by the wind (Westerink et al., 2008; Wamsley et al., For the current analysis, we have reclassified the GLC2000 dataset 2009; Bunya et al., 2010). These factors are spatially quantified by according to National Land Cover Database 2001 (NLCD 2001), as ADCIRC in three georeferenced parameters: 1) frictional drag at the there is no developed methodology to extract ADCIRC specific fric- sea bottom (Manning’s n); 2) the blocking effect of wind momen-

Fig. 3. GLC2000 Land cover data reclassiﬁed to NLCD 2001 for the Sunderban mangrove area.

Table 1 GLC2000 reclassiﬁcation to NLCD 2001 and corresponding Mannings’s n values.

GLC2000 GLC2000 legend NLCD 2001 NLCD 2001 legend Manning’s n 0 Sea 11 Open water 0.02 1 Tropical Evergreen 42 Evergreen forest 0.11 5 Tropical Semi evergreen 41 Deciduous forest 0.1 6 Temperate Conifer 8 Tropical Moist Deciduous 9 Tropical Dry Deciduous 11 Mangroves 90 Woody wetlands 0.2 12 Degraded forest 41 Deciduous forest 0.1 14 Thorn Forest/Scrub (Northern) 52 Shrub/scrub 0.05 15 Thorn Forest/Scrub (Southern) 17 Abandoned Jhum 18 Sparse woods 42 Evergreen forest 0.11 19 Bush 52 Shrub/scrub 0.05 23 Slope Grasslands 52 Shrub/scrub 0.045 25 Alpine Meadow 71 Grassland/herbaceous 0.034 26 Alpine Grasslands 32 Irrigated Intensive Agriculture 82 Cultivated crops 0.037 33 Irrigated Agriculture 35 Rainfed Agriculture 39 Water Bodies 11 Open water 0.02 40 Snow 12 Perennial ice/snow 0.01 41 Barren 31 Barren land (rock/sand/clay) 0.09 45 Settlement 24 Developed high intensity 0.13 46 No Data 95 Open water 0.02

tum transfer to the water column by vegetation (Surface canopy); 3.4. Meteorological parameters and 3) the roughness of the land due to vegetation that can impede wind flow (Land roughness length) (e.g., Ferreira et al., 2014b). The wind and pressure fields during a cyclone are the most ADCIRC related spatial attributes (i.e., Manning’s n, Land roughness important forcing mechanisms for storm surge modeling length and Surface canopies) were generated from the existing and (Holland, 1980). For this study, meteorological forces for Sidr hypothesized land covers (scenario 1 and 2) to represent the Sun- 2007 cyclone (category 5) which hit the coast of Bangladesh were derban mangrove region accurately in our model (Fig. 4). Methods collected from the U.S. Navy Joint Typhoon Warning Center to extract friction parameters such as, Manning’s N, Land roughness (JTWC) (http://www.usno.navy.mil/JTWC/). Storm attributes such length and Surface canopies from the NLCD 2001 database are as, storm track, central pressure and wind speed were used to explained in Atkinson et al. (2011) and Bunya et al. (2010) with drive ADCIRC’s Holland wind model (e.g. Bhaskaran et al., 2014), detailed information. Generally, Manning’s N for woody wetlands with which ADCIRC calculates wind field and pressure data at has been kept to 0.15 in the previous literatures, yet, we assumed each node internally and contribute towards information on the coefficient to be about 0.20 for our study area because of the sea-level pressure distribution and wind gradient. A maximum largely dense canopies (e.g. McIvor et al., 2012; Lewis and Bates, wind radius (Rmax) of 25 km was used to generate the symmetric 2013)(Table 1). wind field for Holland model from Dube and Murty (2009), due to

Fig. 4. a1. b1. c1 Representing the land cover types of mangroves, estuarine forested wetland and grassland respectively, and a2. b2. c2 showing the friction parameter (Manning’s n) generated from them.

Table 2 Cyclone characteristics.

Storm Landfall location Category Central pressure (mb) Maximum wind speed (knots) Forward speed (m/s) 1 Barguna, Bangladesh 5 918 140 9.26 2 Sunderban 3 955 106 7.72 3 Sunderban 3 955 106 9.26 4 Sunderban 3 955 106 11.32 5 Sunderban 4 933 123 7.72 6 Sunderban 4 933 123 9.26 7 Sunderban 4 933 123 11.32 8 Sunderban 5 918 140 7.72 9 Sunderban 5 918 140 9.26 10 Sunderban 5 918 140 11.32

the absence of the Rmax value in the collected dataset. The wind storm characteristics on land cover scenarios (Table 2). Storm file with original track, speed and central pressure, hereafter 8–10 in Table 2 represents the original Sidr 2007 cyclone attri- referred as storm 1, has been applied to the current analysis for butes, and storm 2–7 were hypothesized based on historic storm model validation purposes (Table 2). Then, to look at the maxi- surge reports (e.g. Khan, 1995). Ultimately, we developed three mum impact of storm surges for land cover changes, we modified different intensities (category 3, 4 & 5) for Sidr 2007 cyclone, the Sidr 2007 cyclone track from its original track and changed based on the Saffir-Simpson hurricane wind scale. Then, three the landfall location to Sunderban mangrove (Fig. 5). Subse- types of forward speed during landfall 7.72 m/s, 9.26 m/s & quently, with the modified track we developed nine synthetic 11.32 m/s, hereafter referred as slow, normal & fast were assigned storms (storm 2–10) by adjusting the Sidr 2007 cyclone central to the different categories. Table 2 shows the setting of the matrix pressures and forward speeds to evaluate the sensitivity from consisting cyclone category and forward speed.

other in geography and topography. We picked the first site, zone 1, to show maximum water surface elevation and velocity at cyclone landfall location, and zone 4 perpendicular to the maximum surge propagation direction. Meanwhile, zone 2 has been selected from the land–water interference zone, and zone 3 from slightly inward mangrove forest exposed to the riverine flooding. Elevation at each zone varies from each other, for example, the elevation at zone 2 is approximately 1.5 m, while elevation varies from 2.0–3.0 m at zone 4. Seven mesh nodes close to each other were chosen from every site to calculate the average maximum water elevation and velocity data to account for the topographic variation. To evaluate the increase in surge elevation and velocity due land cover changes from mangroves to estuarine forested wetland or grassland, the following metrics were used: 1 Xn f ¼ f ð1Þ Z n i i¼1 f f D ¼ Z;m Z;wg % ð Þ Z 100 2 fZ;m Fig. 5. Cyclone Sidr modified track with four zones and three cross-sections used f for the model analysis in the Sunderban mangrove area. where i is water surface elevation (WSE) above mean sea level at a given node or the velocity, fZ is the average result at individual zones, D is the percent change, Z is the sampling zone, n is the num- 3.5. Tidal constituents ber of nodes and m, w, g represents the land cover type of mangroves, estuarine forested wetland and grassland respectively. Astronomic tides in the lower western coast of Bangladesh close Inundated areas from the twenty-seven simulations were to Sunderban mangrove system ranges from 1.0 m to 1.5 m from extracted with Surface-water Modeling System (SMS) and the mean sea level (e.g. Salek, 1998). Tides originating from the Indian increase in the flooded area for land cover degradation scenarios Ocean enters the Bay of Bengal through two submarine canyons, were computed using Eq. (2). Calculation of the flood area extent the ‘Swatch of No Ground’ and the ‘Burma Trench’, where M2 and due to mangroves degradation aimed to provide a spatial view S2 are the principal constituents with oscillation period of 12 h on future inundation scenarios. In addition, we have looked into 25 min and 12 h respectively (Mondal, 2001). In this study, tidal the surge attenuation rate by existing Sunderban mangrove and elevation forcing constituents (M2,N2,S2,O1,P1,Q1,K1 and K2) compared it to estuarine forested wetland and grassland for nine from the ‘LeProvost’ global tidal database (Le Provost et al., 1998) cyclones (storm 2–9), according to the land cover degradation sce- were applied at the open ocean boundaries in a geographic coordi- narios discussed earlier at Section 3.3.1. Three cross-sections were nate system. In recent similar studies, Bhaskaran et al. (2014) and picked from the existing mesh nodes, spacing 5 km apart at three Murty et al. (2014) have validated model results using 6 (K1,M2, different locations. Cross-section 1 is located in the maximum N2,O1,P1, and S2) and 13 (K1,M1,N2,O1,P1,S2,K2,L2,2N2,MU2, surge propagation area adjacent to the open ocean, while cross- NU2,Q1 and T2) constituents respectively from ‘Leprovost’ tidal section 2 is in the landfall location and cross-section 3 exposed database in the Bay of Bengal. This database was generated from to the estuarine river. We have measured the maximum water sur- a global tidal model and observed to perform well in deep ocean face elevation at the first and last node of the cross-sections and waters (Mashriqui et al., 2010). In addition, Mashriqui et al. divided the difference by inundation distance to evaluate the surge (2010) had applied five tidal constituents (M2,S2,K1,O1,N2) from attenuation rate. the ‘Leprovost’ database in the Bay of Bengal and noticed a good agreement with the Colorado University and the University of 4. Results and discussion Texas global tidal models. 4.1. Model validation 3.6. Simulation plan and evaluation metrics The model validation was performed including tidal forcing The first phase of the analysis was the model validation, where from the open ocean boundaries for the entire cyclone period of we ran the coupled model with Sidr 2007 cyclone (storm 1) for 5 days from its generation at November 11th, 2007:0000 UTC to 5 days, specifying November 11th, 2007:0000 UTC as the model its dissipation on November 16th, 2007:0000 UTC. Fig. 6b shows cold start time. In addition to the meteorological forcing, eight tidal the resulting maximum water elevation during Sidr 2007 cyclone constituents were also used from the open ocean boundaries. Then, in the lower coastal districts of Bangladesh. We carefully assessed to evaluate the impact of land covers changes on storm surges in the University of Hawaii Sea Level Center (UHSLC) database, the the Sunderban area based on scenario 1 and 2 (section 3.3.1), nine only resource for publicly available water level data in the region synthetic storms (storm 2–10) were applied on 3 different land and observed that it misrepresents the surge elevation. Afterwards, covers. Storms uniquely different from each other based on the we looked for the historic reports published on Sidr 2007 cyclone intensity and speed were used separately for the model run with to validate our model results and continue further analysis. mangroves, estuarine forested wetland and grassland, which con- Additional information related to data collection and model tributed toward a total of 27 simulations. Results generated from validation can be found in a separate study by Deb and Ferreira three land covers for each storm type were compared using maxi- (2016). For Sidr 2007 cyclone, model results were validated at six mum water level, maximum velocity, inundation with respect to different locations along the Bangladesh coast against the storm distance and the inundated area. To compare water level and surge contours from the Japan Society of Civil Engineers (JSCE) velocity, we had selected four zones uniquely different from each model results and the field investigation report of JSCE (2008)

Fig. 6. a. Locations used for model validation b. maximum water surface elevation during cyclone Sidr 2007 c. comparison of storm surges during Sidr 2007 super cyclone with historic studies at Khepupara.

Table 3 Comparison of Model outcomes with high water levels (HWL) extracted from existing studies on storm surge modeling in Bangladesh coast.

Locations HWL during Sidr 2007 cyclone (m) R squared RMSE Chittagong Cox’s Bazar Sandwip Khepupara Galachipa Hiron point Model 3.0 2.0 4.5 5.9 6.5 2.5 Lewis and Bates (2013) 3.2 2.0 3.4 5.5 6.1 2.5 0.95 0.512 JSCE (2008) 3.0 1.7 4.2 6.0 6.0 2.2 0.98 0.297

Fig. 7. Comparison of maximum water surface elevation (in meters) at four different zones for the combination of three land cover types and nine cyclones.

Fig. 8. Comparison of percent increase in maximum water surface elevation at four different zones.

(Fig. 6a). Additionally, high water marks were also extracted from between 5.0–7.0 m along the coastal districts and estuaries where another similar study by Lewis and Bates (2013) for validation Sidr 2007 made landfall, illustrating a good agreement with Lewis purposes. Fig. 6c shows that the computed storm surge varied and Bates (2013) and JSCE (2008). The correlation coefﬁcient

Fig. 9. Comparison of maximum velocity (in m/s) at four different zones for the combination of three land cover types and nine cyclones.

Fig. 10. Comparison of percent increase in maximum velocity at four different zones.

Fig. 11. a1. b1. c1 Represents the increase in velocity (in m/s) due to the land cover change from mangrove to wetland, and a2. b2. c2 represents the increase due to change from mangrove to grassland for category 3, 4 & 5 slow moving cyclones respectively. values (R-square) for model results were 0.95 and 0.98 with exist- 1, 3 and 4 respectively. While, at zone 2 which is located near to ing studies of Lewis and Bates (2013) and JSCE (2008) correspond- the land–ocean interface area compared to other zones, percent ingly and minimum RMSE found to be about 0.30 m and 0.5 m increase in surge amplitude found to be lower for faster and slower (Table 3). tracks than the normal one (Fig. 8). Subsequently, for category 4 & 5 cyclones, maximum increase in surge elevation occured with fast 4.2. Impact of mangrove degradation on maximum water level moving tracks at all the zones. However, when we altered it to grassland, increase in surge elevation varied depending on the In this section, we have examined and compared the percent locations. At zone 2 & 3, surge increase was dominated by the increase in storm surge elevation generated by the two land cover slower tracks, while faster tracks made the maximum increase in degradation scenarios with different cyclone characteristics (storm surge elevation at zone 1 & 4. Zone 1 & 4 are quite distant from 2–10). Appendix A summarizes the results from land cover scenar- the land-sea interface zone compare to other zones, which could ios simulated with different cyclone categories and forward speed have inﬂuenced the surge water levels during fast moving at four different zones (see Supplementary Appendix A) in the Sun- cyclones. Maximum increase in surge elevation at zone 1 & 4 derban mangrove forest along the Bangladesh coast. For both observed to be of 31.92% & 57.35%, 28.18% & 56.72% and 24.30% degradation scenarios, surge elevation was higher for slow moving & 49.09% for category 3, 4 and 5 cyclones. While, at zone 2 & 3, cyclones at all the zones compared to increased forward speed of surge increased up to 26.96% & 20.49%, 21.70% & 19.04% and each categories 3, 4 and 5 (Fig. 7). Also, the water surface elevation 16.18% & 17.91% for category 3, 4 and 5 cyclones (Fig. 8). In general, observed to be greater at zone 4, where the land elevation ranges difference in surge amplitudes for different land covers were min- from 1.0 to 1.5 m and located in the track of maximum surge prop- imum for slower tracks, as cyclones created higher surges for every agation. First, when we changed the land cover from mangroves to land cover type. In a similar study, Zhang et al. (2012) compared estuarine forested wetland, maximum increase in surge occured the role of mangrove width in reducing surge height and found for category 3 slower track up to 14.27%, 8.80% and 16.49% at zone that a 1-km mangrove zone can lessen the surge height up to

Fig. 12. Storm surge elevation along cross-section 01 (in meters) for the combination of three land cover types and nine cyclones.

Fig. 13. Storm surge elevation along cross-section 02 (in meters) for the combination of three land cover types and nine cyclones.

30%, where a 10-km mangrove zone capable of 86% drop. This is in 4.3. Impact of mangrove degradation on surge propagation agreement with the current analysis as we also have observed almost 60% increase in surge elevation because of the mangroves Here, we explored the maximum velocity during cyclones at degradation to grassland. individual zones to compare the effectiveness of mangroves in

Fig. 14. Storm surge elevation along cross-section 03 (in meters) for the combination of three land cover types and nine cyclones. slowing down the surge propagation with estuarine forested wet- illustration of the response from surge propagation throughout land and grassland (see Supplementary Appendix B). Mazda et al. the entire mangrove area. In a recent study, Teo (2008) had com- (1997) had assessed the influence of Mangrove species composi- pared tsunami surge velocities in an area considering scenarios tion in reducing ebb flow velocity and stated that, species with with and without mangrove, and observed an increase in velocity dense roots holds water more effectively than species without up to 500%. In our study, surge velocity during storm events roots. In our study, land cover alteration from mangroves to estu- peaked for category 5 cyclones at all of the selected zones, while arine forested wetland has increased surge velocities up to 246%, the percent increase in surge velocity was highest for category 3 while a significant increase of 2730% occurred for a change from storms. A large portion of mangrove forests got inundated for cat- mangroves to grassland (Fig. 10). At first, when we changed the egory 4 and 5 storms, which lead to less change in surge velocities land cover to estuarine forested wetland, maximum increase in with respect to land cover alterations. Therefore, the existing the surge velocity, calculated using Eq. (2) for category 3 cyclones, Sunderban mangrove plays a larger role in decreasing surge prop- varied from 181% to 226%, 176% to 231% and 186–246% for slow, agation velocities for category 3 cyclones rather than category 4 normal and fast moving cyclones respectively at our four different and 5. Moreover, a major increase in velocity was observed only zones (Fig. 10). Similarly, for category 4 and 5 cyclones, velocity at the mangrove areas where we altered the land covers subse- increase for slow, normal and fast moving tracks varied between quently to estuarine forested wetland and grassland. Disparity in 71% to 159.8%, 87.8–146.7% and 108–145% respectively. Simula- velocity at other locations (e.g., channels and rivers) found as triv- tion results show that the Category 3 cyclones produced maximum ial, where it varied from 2 to 8% for a specific storm. increase in velocity. This is because most of the mangrove areas were dry during Category 3 cyclones and became inundated after 4.4. Inundation with respect to distance changing it to estuarine forested wetlands. In contrast, most of the mangrove areas were already wet with category 4 & 5 cyclones, To investigate the importance of mangroves in attenuating and alteration of the land cover did not produce any major changes storm surges with respect to its propagating distance, results are in the percent increase of surge velocity (Fig. 9). Secondly, when illustrated here using three different cross-sections. First at we changed the land cover to grassland, surge velocity increased cross-section 1, we observed a significant increase in the inundation on a greater scale at all the locations. For category 3 cyclones, distance when we changed the existing mangroves to estuarine increase in surge velocity range from 1084% to 2633%, 996% to forested wetland and then to grassland (see Supplementary 2109% and 1702–2730% for slow, normal and fast moving cyclones Appendix C). In the land–ocean interface, surge water piled up (Fig. 10). While, for category 4 and 5, changes in surge velocity before getting into the land areas of mangroves and generated a declined compared to category 3, and the velocity at all the zones greater surge compared to estuarine forested wetland and for slow, normal and fast moving tracks found to vary between grassland. In a similar study, Zhang et al. (2012) observed a 398.8–1250%, 571.5–1137.5% and 623.4–1090.5% correspondingly 10–30% rise in water levels infront of the mangrove zone, compared (Fig. 10). The simulation results displayed analogous scenarios in to simulations without mangroves. For category 3 cyclones, man- velocity increase to the degradation to estuarine forested wetland, groves and estuarine forested wetlands were observed to have a however with greater magnitude. The spatially varying velocity similar impact for any types of forward speed, and the surge ampli- increase due to the land cover degradation is shown in Fig. 11 for tude dropped significantly at a rate of 0.4 m/km (Fig. 12). In higher category 3, 4 & 5 slow moving cyclones, which provides a better categories, mangroves were found to be very efficient for storm

Fig. 15. a. b. c. Flooded area for the land cover types: grassland, estuarine forested wetland and mangroves respectively, during a category 3 fast moving cyclone.

surge attenuation for faster tracks, where they reduced water level the land-ocean interface for category 3 cyclones (Fig. 14). Then, by 0.5 m/km and 0.4 m/km for category 4 and 5 cyclones respec- we observed a reduction in surge elevation by 0.62 m/km with man- tively. Recently, Zhang et al. (2012) observed the decay rate to vary groves for every combination of category 3 cyclones. Moreover, between 0.2 and 0.5 m/km across mangroves in the Gulf Coast of surge amplitude lessened very slowly for higher categories after South Florida which shows good agreement with our model results. travelling 15 km inland and the mangrove areas had a difference At cross-section 2, for category 3 cyclones, mangrove forests have in inundation distance of approximately 5 km with the estuarine produced a maximum attenuation of 0.5 m/km for faster and slower forested wetland. Reduction in surge elevation occurred signiﬁ- tracks and the surge has travelled only 5 km inland compared to cantly at an approximate rate of 0.60 m/km for category 4 and 5 10 km for estuarine forested wetland (Fig. 13). In addition, for cyclones after propagating 15 km inland (see Supplementary higher storm categories, the mangrove forests have attenuated Appendix C). In all of our assessed cross-sections, we observed max- surge at a rate of 0.5 m/km for faster tracks, while, the inundation imum efﬁciency in storm surge reduction by mangroves for low distance increased up to 10 km with estuarine forested wetland intensity cyclones. In a similar study, Liu et al. (2013) also showed and grassland. Finally, at cross-section 3, surge amplitude was that, mangroves reduces storm surge quicker for low intensity hur- almost the same for each land cover type up to 10 km inland from ricanes than high intensity hurricanes. In most of the cases, longer

Please cite this article in press as: Deb, M., Ferreira, C.M. Potential impacts of the Sunderban mangrove degradation on future coastal flooding in Bangla- desh. J Hydro-environ Res (2016), http://dx.doi.org/10.1016/j.jher.2016.11.005 M. Deb, C.M. Ferreira / Journal of Hydro-environment Research xxx (2016) xxx–xxx 15 distances were required for surge dissipation for slower moving of the zone to a river or stream, topography etc. We considered cyclones than the faster ones, which showed good agreement with Zone-4 (Fig. 5) as an ideal place to provide an overall summary of the studies by Liu et al. (2013) and Zhang et al. (2012). On the other the impacts of Suderban mangrove loses to the surge response as hand, estuarine forested wetland and the grassland seemed to it is perpendicular to the surge direction and also the topography increase the inundation distances up to 5–10 km from the same is almost flat. Surge elevations were found to be significant for slow storm. moving cyclones for any land covers, however, percent increase in the amplitudes of surge for land cover changes from mangroves 4.5. Changes in the flood area to estuarine forested wetland or grassland, was highest for fast moving cyclones. Similarly, maximum velocities were found to In this section, we present the impacts of the different scenarios peak during slow moving cyclones for mangroves and estuarine on the overall inundated area. Flooded area has increased signifi- forested wetland, while the percent increase in velocity topped cantly up to 682 sq.km for category 4 and 1834 sq.km for category for faster ones to almost 1703%. Also, we found mangroves to be 3 cyclones, when we changed the land covers from existing man- most effective in decreasing surge propagating distances and groves to estuarine forested wetland, and grassland respectively flooded areas for weaker and low intensity cyclones. The scenario (see Supplementary Appendix D). Zhang et al. (2012) suggested analysis showed that the existing Sunderban mangrove can play a that the mangroves has a large effect on reducing the inundation substantial role in the reduction of the inundation distance as they extent from storm surge, where model results showed inundation decreased storm surge at a rate of 0.62 m/km for category 3 areas of 4220 sq.km without mangroves and 2450 sq.km with cyclones. Likewise, the inundated area would increase significantly mangroves. In our analysis, the surge affected area was largely up to 18% for category 3 cyclones if the land cover changes to reduced by the mangroves demostrating maximum efficiency for grassland. the category 3 cyclones (Fig. 15). Liu et al. (2013) had found similar It should be noted that, we encountered some limitations during results in the Gulf coasts of South Florida, where maximum flood our analysis because of the available datasets. The low resolution reduction occurred for low intensity hurricanes compared with topo-bathymetric dataset was modified at the mangrove areas higher intensities. In our study, the maximum increase in the because of the inaccurate GEBCO data. Then, the land cover dataset flooded area of 5.3% and 17.6% were observed by the normal mov- GLC2000 was reclassified to NLCD 2001 based on a fixed set of ing cyclones when we altered the land cover from mangroves to assumptions. A high resolution, precise topo-bathymetric dataset estuarine forested wetland, and grassland (see Supplementary and proper land cover classification could have contributed Appendix D). For higher categories, maximum efficiency in the significantly to the current analysis. Hypothetical storms used in reduction of inundated area by mangroves was observed for faster the analysis were generated in an ad-hoc manner based on Saffir- tracks, which shows good agreement with Rego and Li (2009) and Simpson scale due to the absence of potential attributes in the Peng et al. (2006). For category 4 and 5 cyclones, the overall inun- recorded historic storm datasets. A further analysis in future is dated area increased with intensity, however, the percent increase required to create synthetic storms based on measured data in in the flooded area reduced gradually. Such as, category 4 faster the Bay of Bengal area to get better prediction of the surge proper- moving cyclone had inundated 13342 sq.km areas when we ties. In addition, inclusion of the potential impacts of climate altered the land cover to grassland, where the inundation area change to storm intensities, tracks or sizes would strengthen the was already 11823 sq.km for mangroves, reducing the percent prospective research works in constructing hypothetical storm increase in flooding area to 12.9%. Furthermore, the alteration of surge scenarios in the region and undoubtedly improve the reliabil- the mangroves to estuarine forested wetland has produced a ity. Sensitivity numerical experiments showed that the potential 5.8% increase in the inundation area. During fast moving category role of mangroves in reducing storm surge and flooding depends 5 cyclone, the total amount of flooded area peaked to 12556 sq.km, on storm characteristics, such as moving forward speed and 13075 sq.km and 13900 sq.km for mangroves, estuarine forested intensity. Our analysis evaluated the importance of the existing wetland and grassland correspondingly, increasing the flooded Sunderban mangrove in attenuating cyclone-induced surges and area nearly 4.1% for estuarine forested wetland and 10.7% for grass- provided a better understanding of the potential effects of land land (see Supplementary Appendix D). cover changes in coastal flooding in this region. Finally, the research has demonstrated the importance of the Sunderban mangrove to 5. Summary and conclusion protect the people living in the lower western coast of Bangladesh and how its degradation would have hazardous consequences. It In this study, we have assessed the potential impacts of Sunder- also supports the importance of implementing preventive ban mangrove losses to the surge response by simulating degrada- measures against the degradation for this ecosystem. tion scenarios. A Global Land Cover dataset (GLC2000) was adapted to represent the landcover effects to the storm surge propagation in Acknowledgements the area. The coupled wave + hydrodynamic model (SWAN + ADCIRC) was validated for the Bangladesh coast using Sidr 2007 This work used the Extreme Science and Engineering Discovery cyclone and historical data. Afterwards, existing land cover Environment (XSEDE), which is supported by National Science conditions were modified assuming two degradation scenarios: 1) Foundation grant number ACI-1053575. The authors acknowledge mangroves to estuarine forested wetland and 2) mangroves to the Texas Advanced Computing Center (TACC) at The University of grassland. Simulation results showed how surge amplitude, Texas at Austin for providing HPC resources (Stampede) that have velocity and flooding area would increase due to the mangroves contributed to the research results reported within this paper. URL: degradations. Responses of mangrove areas in reducing storm surge http://www.tacc.utexas.edu. during different categories of cyclones were assessed at 4 different zones and 3 cross-sections along the Bangladesh coast, and results for maximum surge elevation, maximum velocity, inundation dis- Appendix A. Supplementary data tance and flood area were compared with two land cover degradation scenarios. Variation in the results from one zone to another Supplementary data associated with this article can be found, in was affected by several geographic features, such as the distance the online version, at http://dx.doi.org/10.1016/j.jher.2016.11.005.

References Henry, R.F., Duncalf, D.S., Walters, R.A., Osborne, M.J., Murty, T.S., 1997. A study of tides and storm surges in offshore waters of the Meghna estuary using a finite element model. Mausam 48 (4), 519–530. Abdul, A.M., 2014. Analysis of vegetation structure causing top-dying in mangrove Henstock, T.J., McNeill, L.C., Tappin, D.R., 2006. Seafloor morphology of the forest trees in the Sundarbans in Bangladesh. Am. J. BioSci. 2 (4), 135–146. Sumatran subduction zone: surface rupture during mega thrust earthquakes? Ali, A., Mynett, A.E., Azam, M.H., 2007. Sediment dynamics in the Meghna Estuary Geology 34, 485–488. Bangladesh: a model study. J. Waterway Port Coast. Ocean Eng. 133, 255–263. Hiraishi, T., Harada, K., 2003. Greenbelt tsunami prevention in South-Pacific region. Atkinson, J., Roberts, H., Hagen, S.C., Zhou, S., Bacopoulos, P., Medeiros, S., Rep. Port Airport Res. Inst. 42 (2), 1–23. Weishampel, J., Cobell, Z., 2011. Deriving frictional parameters and Holland, G.J., 1980. An analytic model of the wind and pressure profiles in performing historical validation for an ADCIRC storm surge model of the hurricanes. Mon. Wea. Rev. 108, 1212–1218. Florida gulf coast. Florida Watershed J. 4 (2), 22–27. Investigation Team of Japan Society of Civil Engineering (JSCE), 2008. Investigation Barros, D.F., Albernaz, A.L.M., 2014. Possible impacts of climate change on wetlands report on the storm surge disaster by cyclone Sidr 2007, Bangladesh. and its biota in the Brazilian Amazon. Braz. J. Biol. 74 (4), 810–820. Islam, M.J., Alam, M.S., Elahi, K.M., 1997. Remote sensing for change detection in the Barbier, E.B., Georgiou, I.Y., Enchelmeyer, B., Reed, D.J., 2013. The value of wetlands Sundarbans, Bangladesh. Geocarto Int. 12, 91–100. in protecting southeast Louisiana from hurricane storm surges. PLoS ONE 8 (3), Islam, N., 2005. Environmental issues in Bangladesh: an overview. J. Soc. Sci. 3, 671– e58715. 679. Bartholomé, E., Belward, A.S., 2005. GLC2000: a new approach to global land cover Islam, A.S., Bala, S.K., Hussain, M.A., Hossain, M.A., Rahman, M.M., 2011. mapping from Earth observation data. Int. J. Remote Sens. 26 (9), 1959–1977. Performance of coastal structures during cyclone Sidr. Nat. Hazards Rev. 12 Bhaskaran, P.K., Nayak, S., Bonthu, S.R., Murty, P.L.N., Sen, D., 2013. Performance and (3), 111–116. validation of a coupled parallel ADCIRC–SWAN model for THANE cyclone in the Islam, M.S., Wahab, M.A., 2005. A review on the present status and management of Bay of Bengal. Environ. Fluid Mech. 13 (6), 601–623. mangrove wetland habitat resources in Bangladesh with emphasis on Bhaskaran, P.K., Gayathri, R., Murty, P.L.N., Reddy, B., Sen, D., 2014. A numerical mangrove fisheries and aquaculture. Hydrobiologia 542, 165–190. study of coastal inundation and its validation for Thane cyclone in the Bay of Jarvis, A., Rubiano, J., Nelson, A., Farrow, A., Mulligan, M., 2004. Practical use of Bengal. Coast. Eng. 83, 108–118. SRTM data in the tropics: Comparisons with digital elevation models generated Bhowmik, A.K., Cabral, P., 2013. Cyclone Sidr impacts on the Sundarbans floristic from cartographic data Working Document no. 198, 32. International Centre for diversity. Earth Sci. Res. 2 (2). Tropical Agriculture (CIAT). Blanco, J.F., Estrada, E.A., Ortiz, L.F., Urrego, L.E., 2012. Ecosystem-wide impacts of Kabir, D.M.H., Hossain, J., 2008. Resuscitating the Sundarbans: Customary Use of deforestation in mangroves: The Urabá Gulf (Colombian Caribbean) case study. Biodiversity and Traditional Cultural Practices in Bangladesh. Unnayan ISRN Ecol. 2012, 1–14. Onneshan-The Innovators, Dhaka, p. 98. Blain, C.A., Westerink, J.J., Luettich, R.A., 1994. Domain and grid sensitivity studies Karim, M.F., Mimura, N., 2008. Impacts of climate change and sea-level rise on for hurricane storm surge predictions, Computational Methods in Water cyclonic storm surge floods in Bangladesh, Global Environ. Change 18, Resources X, A. Peters et al. (eds), Heidelberg. 490–500. Booij, N., Ris, R.C., Holthuijsen, L.H., 1999. A third-generation wave model for coastal Khan, S.R., 1995. Geomorphic Characterization of Cyclone hazards along the Coast of regions: 1. Model description and validation. J. Geophys. Res. 104 (C4), 7649– Bangladesh M.Sc. Thesis. ITC, Enschede, the Netherlands. 7666. http://dx.doi.org/10.1029/98JC02622. Krauss, K., Doyle, T., Doyle, T., Swarzenski, C., From, A., Day, R., Conner, W., 2009. Bunya, S., Dietrich, J.C., Westerink, J.J., Ebersole, B.A., Smith, J.M., Atkinson, J.H., Water 666 level observations in mangrove swamps during two hurricanes in Jensen, R.E., Resio, D.T., Luettich Jr., R.A., Dawson, C.N., Cardone, V.J., Cox, A.T., Florida. Wetlands 29, 142–149. Powell, M.D., Westerink, H.J., Roberts, H.J., 2010. A high resolution coupled Le Provost, C., Lyard, F., Molines, J.M., Genco, M.L., Rabilloud, F., 1998. A riverine flow, tide, wind, wind wave and storm surge model for Southern hydrodynamic ocean tide model improved by assimilating a satellite Louisiana and Mississippi: Part I—Model development and validation. Mon. altimeter-derived data set. J. Geophys. Res. 103 (C3), 5513–5529. http://dx. Weather Rev. 138 (2), 345–377. doi.org/10.1029/97JC01733. Center for Environmental and Geographic Information Services Bangladesh (CEGIS), Lewis, M., Bates, P., 2013. A storm surge inundation model of the northern 2007. Effect of cyclone Sidr on the Sundarbans. A Preliminary Assessment. Bay of Bengal using publicly available data. Q. J. R. Meteorol. Soc. 139, Chowdhury, J.U., Karim, M.F., 1996. A risk-based zoning of storm surge prone area of 358–369. the Ganges Tidal Plain. J. Civ. Eng. 24, 221–233. Lin, P., 1997. Mangrove ecosystems in China. Science Press, Beijing, pp. 11–250. (in Deb, M., Ferreira, C.M., 2016. Simulation of cyclone-induced storm surges in the Chinese). low-lying delta of Bangladesh using coupled hydrodynamic and wave model Liu, H., Zhang, K., Li, Y., Xie, L., 2013. Numerical study of the sensitivity of mangroves (SWAN + ADCIRC). J. Flood Risk Manag. http://dx.doi.org/10.1111/jfr3.12254. in reducing storm surge and flooding to hurricane characteristics in southern Debsharma, S.K., 2007. Numerical Simulations of Storm Surges in the Bay of Bengal. Florida. Cont. Shelf Res. 64, 51–65. JCOMM Scientific and Technical Symposium, Korea. Luettich, R.A., Westerink, J.J., 2004. Formulation and numerical implementation of Dietrich, J.C., Zijlema, M., Westerink, J.J., Holthuijsen, L.H., Dawson, C., Luettich, R.A., the 2D/3D ADCIRC Finite Element Model Version 44.XX . waves and storm surge using integrally-coupled scalable computations. Coast. Madsen, H., Jakobsen, F., 2004. Cyclone induced storm surge and flood forecasting in Eng. 58, 45–65. the northern Bay of Bengal. Coast. Eng. 58, 277–296. Dube, S.K., Sinha, P.C., Rao, A.D., Chittibabu, P., 1994. A real time storm surge Mashriqui, H., Kemp, G., Heerden, I., Binselam, A., Yang, Y., Streva, K., Westerink, J., prediction system: an application to east coast of India. Proc. Indian Natl. Sci. Ropers-Huilman, B., 2010. Experimental storm surge forecasting in the Bay of Acad. 60, 157–170. Bengal. Wind Storm Storm Surge Mitigation, 15–25. Dube, S.K., Rao, A.D., Sinha, P.C., Murty, T.S., Bahulayan, N., 1997. Storm surge in the Mazda, Y., Magi, M., Kogo, M., Hong, P.N., 1997. Mangroves as a coastal protection Bay of Bengal and Arabian Sea: the problem and its prediction. Mausam 48, from waves in the Tong King delta, Vietnam. Mangroves Salt Marshes 1 (2), 283–304. 127–135. Dube, S.K., Murty, T.S., 2009. Storm surge modelling for the Bay of Bengal and McIvor, A.L., Spencer, T., Möller, I., Spalding, M., 2012. Storm surge reduction by Arabian Sea. Nat. Hazards 51, 3–27. mangroves. Natural Coastal Protection Series: Report 2. Cambridge Coastal Duke, N.C., Meynecke, J.O., Dittmann, S., Ellison, A.M., Anger, K., Berger, U., Cannicci, Research Unit Working Paper 41. The Nature Conservancy and Wetlands D., Diele, K., Ewel, K.C., Field, C.D., Koedam, N., Lee, S.Y., Marchand, C., Nordhaus, International, p. 35. ISSN 2050-7941. I., DahdouhGuebas, F., 2007. A world without mangroves? Science 317 (5834), Ministry of Environment and Forests (MOEF), 2008. Nishorgo: Bangladesh’s 41–42. Protected Area Management Program. Ministry of Environment and Forests, Ellison, J., 2012. Climate change vulnerability assessment and adaptation planning Bangladesh. viewed 06/02/2010, http://www.nishorgo.org/N. for mangrove systems. World Wildlife Fund (WWF), Washington DC, pp. 3–85. Mondal, A.M., 2001. Sea Level Rise Along the Coast of Bangladesh. Bangladesh Ferreira, C.M., Irish, J.L., Olivera, F., 2014a. Uncertainty in hurricane surge simulation Inland Water Transport Authority, Ministry of Shipping. http://www.gloss- due to land cover specification. J. Geophys. Res. Oceans 119, 1812–1827. http:// sealevel.org/publications/documents/bangladesh_2001.pdf. dx.doi.org/10.1002/2013JC009604. Murty, T.S., Flather, R.A., Henry, R.F., 1986. The storm surge problem in the Bay of Ferreira, C.M., Irish, J.L., Olivera, F., 2014b. Quantifying the potential impact of land Bengal. Prog. Oceanogr. 16, 195–233. cover changes due to sea-level rise on storm surge on lower Texas coast bays. Murty, P.L.N., Sandhya, K.G., Bhaskaran, P.K., Jose, F., Gayathri, R., Nair, T.M.B., Coast. Eng. 94, 102. Kumar, T.S., Shenoi, S.S.C., 2014. A coupled hydrodynamic modeling system for Flather, R.A., Khandoker, H., 1993. The storm surge problem and possible effects of PHAILIN cyclone in the Bay of Bengal. Coast. Eng. 93, 71–81. sea level changes on coastal flooding in the Bay of Bengal. In: Warrick, Barrow, Peng, M., Xie, L., Pietrafesa, L.J., 2006. A numerical study on hurricane-induced E.M., Wingley, T.M.L. (Eds.), Climate and Sea Level Change: Observations, storm surge and inundation in Charleston Harbor, South Carolina. J. Geophys. Projections and Implications. Cambridge University Press, Cambridge, pp. 229– Res. 111, C08017. http://dx.doi.org/10.1029/2004JC002755. 245. Rego, J.L., Li, C., 2009. On the importance of the forward speed of hurricanes in storm Flather, R.A., 1994. A storm surge prediction model for the Northern Bay of Bengal surge forecasting: a numerical study. Geophys. Res. Lett. 36, L07609. http://dx. with application to the cyclone disaster in April 1991. J. Phys. Oceanogr. 24, doi.org/10.1029/2008GL036953. 172–190. Roy, G.D., 1995. Estimation of expected maximum possible water level along the Giri, C., Pengra, B., Zhu, Z., Singh, A., Tieszen, L.L., 2007. Monitoring mangrove forest Meghna estuary using a tide surge interaction model. Environ. Int. 25 (1), 671– dynamics of the Sundarbans in Bangladesh and India using multi-temporal 677. satellite data from 1973 to 2000. Estuar. Coast. Shelf Sci. 73, 91–100.

Roy, A.K.D., Alam, K., Gow, J., 2012. A review of the role of property rights and forest Dr. Celso Ferreira is an Assistant Professor of Water policies in the management of the Sundarbans Mangrove Forest in Bangladesh. Resources Engineering in the Civil, Infrastructure and Forest Policy Econ. 15, 46–53. Environmental Engineering Department of George Salek, J.A., 1998. Coastal trapping and funneling effect on storm surges in the Mason University. He is also an Associate Researcher at Meghna estuary in relation with the cyclones hitting Noakhali-Cox’s Bazar coast the USGS National Research Program. He has a PhD from of Bangladesh. J. Phys. Oceanogr. 28, 227–249. Texas A&M University in Civil Engineering with a focus Shamsuddoha, M., Islam, M., Haque, M. A., Rahman, M. F., Roberts, E., Hasemann, A. on Water Resources Engineering. and Roddick, S., 2013. Local Perspective on Loss and Damage in the Context of Extreme Events. Teo, F.Y., 2008. Attenuation of tsunami currents in an estuary with mangroves, . Wahid, S.M., Babel, M.S., Bhuiyan, A.R., 2007. Hydrologic monitoring and analysis in the Sundarbans mangrove ecosystem, Bangladesh. J. Hydrol. 332, 381–395. Wamsley, T., Cialone, M., Smith, J.M., Atkinson, J.H., Rosati, J.D., 2009. Inﬂuence of landscape restoration and degradation on storm surge and waves in southern Louisiana. Nat. Hazards 51, 207–224. Westerink, J.J., Luettich Jr., R.A., Feyen, J.C., Atkinson, J.H., Dawson, C.N., Roberts, H.J., Mithun Deb is currently a PhD student in the Civil and Powell, M.D., Dunion, J.P., Kubatko, E.J., Pourtaheri, H., 2008. A basin to channel Coastal Engineering program of the University of Dela- scale unstructured grid hurricane storm surge model applied to Southern ware and working as a Graduate Research Assistant in Louisiana. Mon. Weather Rev. 136 (3), 833–864. the Center for Applied Coastal Research. Recently, he Yanagisawa, H., Koshimura, S., Miyagi, T., Imamura, F., 2010. Tsunami damage graduated with a Master of Science degree in Civil reduction performance of a mangrove forest in Banda Aceh, Indonesia inferred Engineering from George Mason University focusing on from ﬁeld data and a numerical model. J. Geophys. Res. 115, C06032. http://dx. Environmental and Water Resources Engineering. doi.org/10.1029/2009JC005587. Zhang, K.Q., Liu, H., Li, Y., Hongzhou, X., Jian, S., Rhome, J., Smith III, T.J., 2012. The role of mangroves in attenuating storm surges. Estuar. Coast. Shelf Sci. 102, 11–23. Zijlema, M., 2010. Computation of wind-wave spectra in coastal waters with SWAN on unstructured grids. Coast. Eng. 57, 267–277.