A Study of Morphological Changes in the Coastal Areas and Offshore Islands of Sudarban Coastline Using Remote Sensing

Abstract

The Sundarbans, located along the coastal areas of India and Bangladesh, is the largest remaining single block of mangrove forest in the world, covering approximately 1 million hectares (10,000 km2) of the Ganges-Brahmaputra delta. This unique ecosystem is under threat from major disturbances such as sea level rise and alterations in water flows from the Himalayan headwaters. There have been very few studies on the current status and dynamics of the Sundarban’s coastline. To address this knowledge gap, we conducted a study utilizing Landsat images spanning from 1975 to 2022. Our findings reveal that the rates and directions of erosion and accretion varied across the different periods. During the 1994–2005 interval, erosion reached its peak with a land loss rate of ~17 km2 per year. However, this rate substantially declined in subsequent periods to ~8 km2. Accretion, on the other hand, showed a rate of ~7 km2 per year between 1975 and 1988 but declined to approximately ~6 km2 per year between 1988 and 1994. While the accretion rate has declined in recent years, the erosion rate has remained relatively high.

1. Introduction

The Sundarban area of the South Asian Region is located between 21^o N to 26^o30’N and 88 ^o E to 92^o 30′E in the Bay of Bengal. It consists of a cluster of low-lying islands covering an area of ~10,000 km² that were formed during the last 11,000 years. It is the second-largest river delta built by the Ganga-Bramhaputra river system driven by southwest monsoon rains. The rivers carry large amounts of sediment loads, ~10 ⁹ t/year, from the Himalayas and upper parts of the Bengal delta. The mangrove forest of the Sundarbans is shared between India (38%) and Bangladesh (62%). The Royal Bengal Tiger, Ganga River Dolphin, and certain endangered species, such the River Terrapin, may all be found in the Subarnarekha Mangrove woods, which are rich in biodiversity. In both nations, the vital ecosystem is protected and listed as a UNESCO World Heritage Site.

The mangrove ecosystem holds immense ecological and economic importance. It is home to numerous organisms that possess substantial ecological and economic values. These ecosystems play a crucial role in supporting both terrestrial and aquatic food chains, thereby sustaining a diverse range of plant and animal species. One of the notable contributions of mangrove ecosystems is their ability to serve as natural barriers, protecting shoreline and island areas from various natural hazards such as cyclones, hurricanes, and tsunamis. By absorbing and dissipating the force of waves, they effectively mitigate coastal erosion. Moreover, mangroves act as biological filters, helping to maintain water quality by trapping sediment and nutrients, thus purifying polluted coastal waters.

Furthermore, mangroves play a vital role in maintaining the carbon balance in coastal areas. They sequester and store large amounts of carbon dioxide, contributing to climate change mitigation efforts. Additionally, these ecosystems hold significance for tourism and recreation purposes, attracting visitors who appreciate their unique beauty and ecological value [1]. Overall, the preservation and conservation of mangrove ecosystems have far-reaching ecological, economic, and societal benefits.

Mangroves demonstrate remarkable ecological stability in terms of persistence and resilience. However, they are highly sensitive to changes in hydrology. Therefore, it is essential to prioritize the protection and restoration of mangrove ecosystems. Since gaining independence, a significant number of homeless individuals have migrated and settled in the reclaimed Indian Sundarban region. Despite various efforts to safeguard mangrove resources, they face substantial anthropogenic pressure resulting from unsustainable exploitation for multiple purposes, such as wood harvesting, fodder collection, fuel extraction, and charcoal production [2]. Additionally, the conversion of forested areas into aquaculture and agricultural lands, as well as the construction of jetties and harbors to meet the demands of the growing population, further exacerbate the challenges faced by mangroves [2, 3, 4].

On an average, 5–6 cyclones annually hit the Sundarban area. Of these cyclones, two are of severe category. Among these, Amphan in 2020 had the highest impact with an estimated loss of 128 lives and > USD 13 million in damages. The diversity and extent of the Sudarban are constantly declining due to anthropogenic and natural causes. Regular monitoring of extent and quality is necessary for this area to control or even regain the loss if given constant effort for a longer period of time. Land dynamics of delta coastline are controlled by three major factors.

1. Tectonic subsidence and compaction

2. Relative sea level change and wave action

3. River sediment supply

Studies suggest the subsidence rate of the Bengal delta area is in the range of 15–50 mm annually. The exploration of oil and gas from delta, trapping of sediments due construction of reservoirs upstream, and other anthropogenic activities are considered the main cause of the subsidence of Bengal Delta. Also, the estimated rise of sea level of Bay of Bengal is at >10 mm/year which is among the world’s highest (Based on global sea level data and modeling, Eriscson et al. [5] are noticeable this coastline.

The objective of this study is to observe the change of shoreline in Bay of Bengal through time and its impact on the extent of mangrove forest cover area. The rate of shoreline erosion or accretion and movement of coastline are calculated using the series of Landsat images available from 1975 to 2022. Conducting field surveys in the swampy mangrove forest can be extremely challenging due to their inaccessibility (Nandy et al. [6]). In such circumstances, remote sensing techniques have emerged as increasingly valuable tools for mapping and monitoring mangroves in a timely manner [7]. While remote sensing data cannot replace field surveys, it offers several advantages. These include synoptic coverage, the availability of free or low-cost satellite data, and repetitive coverage [8, 9].

Remote sensing allows for the collection of data over large areas, providing a broad overview of mangrove ecosystems that would be difficult to achieve solely through field surveys. Additionally, satellite data can be obtained at regular intervals, enabling the monitoring of changes and trends in mangrove extent and condition over time. The availability of free or affordable satellite data further enhances the accessibility and affordability of remote sensing technology for mangrove monitoring purposes. While remote sensing data is a valuable tool, it is important to note that it should be complemented with field surveys to validate and ground truth the remote sensing results. By integrating remote sensing with field-based data collection, a comprehensive and accurate understanding of mangrove ecosystems can be achieved.

Allison [10] mentioned the Bengal delta is in a net erosional state at a rate of ~1.9 km² per year and the coastline retreat of ~3–4 km in some areas of the western edge since 1792 (~21 m year⁻¹). Some studies mention the accretion rate as ~7 km² year⁻¹ along the river mouth regions [11].

2. Study area

The coastline along Sundarban area, unlike adjacent areas, is not restricted by embankments, making it a preferred study area to observe the impact of sea level rise and coastal erosion on the shoreline. Anthropogenic activities, such as fishing, hunting, and resource harvesting, occurs inside the inland river channel and their subsidiaries and inland forest area. Excessive humidity prevails throughout the year. The monsoon occurs in the months of June to September with 2500 to 3000 mm of rain fall annually. In the summer temperature ranges between 25 to 35°C, while winter has a temperature range of 12 to 24°C.

Tidal levels in the area exhibit seasonal variations, ranging from 4 to 6.5 meters. Additionally, the pH of the water fluctuates between 7.2 to 7.9 [12]. These environmental factors, including humidity, rainfall, temperature, tidal levels, and water pH, collectively contribute to the unique and dynamic ecosystem of the mangrove forest. Therefore, the study is focused on the coastline that is affected.

3. Material and method

The present study uses a collection 2 level 1 images from LANDSAT series of satellites available from USGS Earth Explorer website from 1975 to 2022. To get cloud-free clear images of the study area, images are quired in the month of December each year where possible, or in adjacent months. For the years 1975 and 1980, images were selected from the Landsat Multi-Spectral Scanner (MSS), and the rest images were from Landsat Thematic Mapper (TM, 4 and 5) and Enhanced TM (ETM+) (Figure 1). Two adjacent Landsat images of path 148 rows 45 and 147 rows 45 for Landsat MSS; path 138 rows 45 and 137 rows 45 for TM and ETM+ are needed for the Sundarbans coastline of Bangladesh and India (Table 1).

Figure 1.

Location of the study area: Sundarban coastline along bay of Bengal.

Year	Month	Satellite	Path	Cloud cover	Resolution (M)
1975	December	Landsat MSS 1	148/45147/45	0	60
1988	December	Landsat MSS 4	148/45147/45	0	30
1994	December	Landsat MSS 4	148/45147/45	0	30
2005	December	Landsat 7 ETM+	148/45147/45	<7%	30
2015	December	Landsat 8 OLI	148/45147/45	0	30
2022	December	Landsat 8 OLI	148/45147/45	0	30

Table 1.

Landsat images used in the study.

Images downloaded from USGS Earth Explorer are first georeferenced with a final georeferenced image having <±0.5 pixel root mean squared error (RMSE). As the adjacent images fall in two different countries with different UTM Zones of 46 N and 45 N The images are reprojected to Lambert Azimuthal Equal area projection to preserve the area of individual polygon and a true sense of direction from the center. This projection is also preferred for statistical analysis of land change. To maintain the spectral integrity of the image nearest neighbor resampling was used. All images from Landsat MSS1 are resampled to 30 m resolution. Resampling the 60 m MSS pixels to 30 m does not impact the spatial resolution of the images, whereas resampling the 30 m TM pixels to 60 m MSS pixels would degrade spatial resolution of the images.

To enable accurate classification and change detection from multi-temporal satellite imagery, it is crucial to perform radiometric calibration. This process involves correcting for gain and bias variations in the satellite data. In the case of Landsat data, the scattering effect is particularly prominent [13]. Additionally, for vegetation cover identification, atmospheric correction is necessary to mitigate the impact of scattering.

In our study, we conducted atmospheric correction on the Landsat visible and near-infrared (VNIR) bands. This correction involved radiometric calibration, which transformed the digital number (DN) values of the bands into the top of the atmosphere radiance (LTOA) using a sensor calibration function (Eq. 1) proposed by Chander et al. [14]. Subsequently, we converted the radiance of the VNIR bands into accurate surface reflectance using an image-based atmospheric correction model developed by Chavez [15]. This model was chosen for its simplicity and because radio-sounding data was not readily available (Eq. 2).

$$L_{TOA}=\left(\frac{L{\max }_{\lambda }-L{\min }_{\lambda }}{QCAL\max -QCAL\min }\right)\times \left(DN-QCAL\min \right)+L{\min }_{\lambda }$$E1

Where Lmax_λ and Lmin_λ represent the maximum and minimum radiance (in W/m⁻² sr⁻¹ μm⁻¹), QCALmax and QCALmin represent the maximum and minimum DN value possible (255/1).

$$\rho =\frac{\left(L_{TOA}-L_p\right)\pi d^2}{ESU{N}_{\lambda }\cos {\theta }_z{T}_z}$$E2

Where ρ represents the surface reflectance. d denotes the Earth-sun distance, which is measured in Astronomical Units (AU). ESUNλ refers to the band-pass solar irradiance at the top of the atmosphere (TOA) for a specific wavelength (λ); Z represents the solar zenith angle, measured in degrees; TZ represents the atmospheric transmission between the ground and the TOA. For band 4, the value of TZ is assumed as 0.85, while for band 5, it is taken as 0.95 (Figure 2) [15]. Lp represents the radiance that results from the interaction of aerosols and atmospheric particles. Its estimation is based on the studies conducted by Song et al. [13], Chavez [15], and Sobrino et al. [16].

Figure 2.

Overall methodology of the study.

FCC (false-color composite) images are commonly used in remote sensing and satellite imaging to enhance the interpretation of land cover and vegetation. These images are created by combining different bands of electromagnetic radiation, typically in the green, red, and near-infrared regions of the spectrum. The interpretation of FCC images relies on the fact that different materials reflect and absorb different wavelengths of light. In the case of distinguishing between land and water, the choice of bands is important. By using the green, red, and near-infrared bands, it becomes possible to differentiate between various land cover types. In an FCC image, vegetated areas appear in shades of red. This is because healthy vegetation strongly reflects near-infrared light while absorbing more of the red light. As a result, when the near-infrared band is assigned to the red channel in the composite image, it gives vegetation a distinct red color. Bare soils, on the other hand, appear in tones of brown. Since bare soils have little vegetation cover, they reflect both green and red light, giving them a brownish appearance in the composite image. Mudflats or sandy beaches generally appear as shades of white in an FCC image. These areas have a high reflectance in both the green and red bands, resulting in a bright appearance. Water bodies, such as lakes or oceans, appear blue or black in the FCC image. This is because water absorbs near-infrared light, and in the absence of strong vegetation reflectance, it reflects more of the blue and green light. Therefore, water bodies tend to appear darker compared to other land cover types. In addition to FCC images, the normalized difference vegetation index (NDVI) can be calculated using the red and near-infrared bands. The NDVI is a quantitative measure of vegetation health and density. It is computed by taking the difference between the near-infrared and red reflectance values and dividing it by their sum. The resulting NDVI values can help confirm the land-water boundary. Water bodies typically have a negative NDVI value, indicating the absence of vegetation, while dry land usually has a positive NDVI value, reflecting the presence of vegetation.

Overall, FCC images and the NDVI provide valuable information for land cover classification, vegetation monitoring, and understanding the distribution of different land and water features in remote sensing applications.

4. Error estimation

The two main independent sources of uncertainty mentioned in your statement are the uncertainty caused by the georeferencing process and the uncertainty caused by the digitizing process. These two independent sources of uncertainty, namely georeferencing and digitizing, contribute to the overall uncertainty in the spatial data analysis and should be considered when interpreting or using the data. The georeferencing process involves aligning spatial data to a known coordinate system or reference imagery. The error associated with this process is assumed to be normally distributed with a mean of 0 and a standard deviation equal to the Root Mean Square Error (RMSE) resulting from the georeferencing procedure. In this case, the RMSE is ±0.5 pixels. The assumption of normal distribution implies that the errors are symmetrically distributed around the mean, and the RMSE provides an estimate of the typical magnitude of the georeferencing errors. Digitizing refers to the process of converting analogue or physical data into digital form. In this context, it involves delineating the boundary of an area on a map or image. The error associated with this process is assumed to be uniformly distributed, ranging between 0 square meters and 900 square meters. This assumption implies that the digitizing errors have an equal likelihood of occurring within this range. A mixed pixel occurs when a pixel represents a mixture of different land cover types, in this case, soil and water along the coastline. This can be a result of the spatial resolution of the data, where a single pixel covers an area that includes both land and water. The presence of mixed pixels along the coastline can introduce additional uncertainty in the analysis, as the classification or interpretation of such pixels becomes more challenging (Figure 3).

Figure 3.

Changes in coastline from 1975 to 2022.

5. Results and discussion

Accretion and erosion were not the same throughout the different segments of time considered for the study. For example, erosion was dominating in the 1975–1988 interval with a land loss of ~12 km²/year, while erosion declined in the following periods to a rate of ~8 km² per year. If we look at results as a whole erosion wins over accretion and the coast registers a net land loss of ~280 km² in this study period. Accretions in the 1994–2005 and 2005–2015 periods were ~ 10 km² and ~ 5 km² per year respectively. In this segment, a total gain of ~210 km² of land can be observed.

According to the present study, the erosion and accretion rates in the Subarnarekha Delta have been highly dynamic over the past 47 years. It is observed that the delta has experienced both erosion and accretion at different times during this period. The total land change estimated over the last 47 years in the Subarnarekha Delta is a loss of 250 km2 of land. This suggests that the delta experienced a net loss of land during this time period. The study also indicates that accretion has been dominant in the east direction, while erosion has been more pronounced in the south-to-west direction. This spatial variation in erosion and accretion patterns suggests that different parts of the delta have been subjected to varying degrees of land loss or gain. These findings highlight the dynamic nature of the Subarnarekha Delta, with ongoing processes of erosion and accretion shaping its coastline over the years. Understanding the spatial patterns and rates of erosion and accretion is crucial for managing and mitigating the impacts of coastal changes in the delta region.

A notable pattern observed in the land dynamics of the Sundarbans coastline is the declining rate of accretion in successive periods. This trend could be attributed to the overall sediment deprivation of the delta caused by human activities such as dam construction and other anthropogenic disturbances upstream (Table 2) [17].

Period	Accretion (Km^2 year^−1)	Total accretion (km^2)	Erosion (km^2 year^−1)	Total erosion (km^2)	Difference (km^2)
1975–1988	7.3	92.7 ± 0.8	12.9	167.8 ± 0.8	−75.1
1988–1994	6.58	32.9 ± 0.5	10.94	54.7 ± 0.3	−21.8
1994–2005	10.22	112.5 ± 0.4	17.27	190.2 ± 0.3	−77.7
2005–2015	6.1	61.5 ± 1.3	11.5	115.5 ± 1.6	−54
2015–2022	4.18	29.3 ± 0.5	8.51	59.6 ± 0.9	−30.3

Table 2.

Accretion and erosion rates in the Sundarbans coastline estimated.

Erosion is observed in all directions except for the landward directions (N, NE, NW), suggesting the potential influence of sea level rise (SLR) impacts in the Bay of Bengal. The variation in erosion rates in different azimuthal directions may be attributed to a combination of surface wave and tidal actions. Surface waves primarily originate from the southwest direction in the Bay of Bengal, while tidal actions predominantly occur from the south [18]. Additionally, the East India Coastal Current (EICC) flows northward during the rainy seasons (March–September) and reverses its direction during the dry seasons (October–January). As a result, the combined forces of waves and tides are stronger in the south azimuthal direction compared to other directions.

These factors contribute to the complex interplay of erosion and accretion along the Sundarbans coastline, with variations in sediment dynamics influenced by SLR, wave action, tidal forces, and seasonal currents (Figures 4 and 5).

Figure 4.

Erosion and accretion over the past 47 years in suburban coastal area.

Figure 5.

Accretion and erosion of the coastline between 1975 and 2022.

Since there is no dike or other construction to safeguard the shoreline, the coastline retreat in the Sundarban area was expected due to sea level rise. By storing sediments upland and reducing their availability at the shore, sediments from dam construction have also had a substantial impact. While tidal action occurs from the south, surface waves in Sundarbans coastline region are primarily from the southwest. The East India Coastal Current (EICC) flows southward during the dry seasons of October to January and northward during the rainy seasons of March to September. The southern region is more affected by the combined impact of waves and tides than other regions. During the final 47 years of the research period, the Subarnarekha coast lost 270 km2 in total, or 5.7 km2 per year. This number is higher than what preceding researchers had predicted. There has been a loss of land, but not equally. Despite the fact that some new islands have appeared along the coast, the overall picture shows a loss of land mass there.

6. Conclusions

Our research emphasizes the intricate nature of the spatiotemporal dynamics observed along the retreating Sundarbans coastline in the Bengal delta. To delineate the coastline, we utilized cloud-free Landsat images and developed algorithms that consistently derived distances and areas of land dynamics for the entire coastline at regular intervals. While coastal retreat is a natural global phenomenon associated with sea-level rise (SLR), our study delved into the specific impacts of SLR, as well as reduced discharge and sediment flow from the contributing river, on the coastline.

The formation of the Bengal delta was primarily driven by the discharge of the Ganges-Brahmaputra (GB) river, resulting in accretion dominating the region for thousands of years. However, while some previous sampling studies suggested ongoing accretion in the Bengal delta [11], recent modeling studies indicate that sediment compaction has caused the delta to sink [17]. Our study presents the first evidence that the entire non-diked portion of the Sundarbans coastline in the Bengal delta is currently experiencing a net erosional state.

By analyzing a time series of satellite images, we were able to characterize the spatial and temporal aspects of the retreat. This approach reduced uncertainties inherent in modeling and sampling studies of continuous spatial processes such as coastal dynamics. The spatiotemporal analysis conducted in our study may facilitate future research in understanding the local and global factors contributing to the reported spatial variations in erosion and accretion.

We anticipate that the findings of our study will have practical implications for the management planning of the Sundarbans—the world’s largest remaining patch of mangrove forests.