Vegetation Cover Change Analysis during 1989-2020 of Coastal Barguna District, Bangladesh Using Remote Sensing and GIS Technology

Abstract

This paper represents an advanced assessment of a satellite imagery-based vegetation transition detection method for Bangladesh's southern district of Barguna. Bangladesh has been identified as the most vulnerable country to the effects of climate change due to deteriorating vegetative cover and urbanization pushing forces. This paper aims to use GIS and remote sensing techniques to estimate vegetation change and determine the conversion pattern in the coastal district of Barguna, Bangladesh, from 1989 to 2020. The major methodology used in this research is NDVI differencing and supervised image classification to validate the conversion over time. With a combination of satellite imagery bands, NDVI uses multi-spectral remote sensing techniques to discover vegetation index, water, empty spaces, and forests. To categorize the canopy into distinct kinds, we use NDVI threshold values of 0, 0.1, 0.15, and 0.2. Our findings reveal significant geographical variations in bare regions, sparse, moderate, and thick vegetation types. We discovered that over the last 31 years, a total of 412.90 km² (49.25 percent) of land has been deforested, with the transition rate being particularly high in dense vegetation. This research of vegetation cover change can help predict the recurrence of natural disasters, give humanitarian assistance, and enable innovative protection tactics.

Introduction

Forests help to lessen the risk of natural disasters including floods, landslides, cyclones, and other calamities, as well as air pollution (Nowak et al., 2006). According to the United Nations Sustainable Development Goals (SDGs), target 3 and goal 11 seek to provide communities with safer, healthier, and more prosperous lifestyles by allowing forests to reduce climate change migration through carbon absorption, oxygen, and humidity balance in the atmosphere (Turner‐Skoff & Cavender, 2019). Climate change impacts have intensified in most regions of the world, mostly in the last three decades, putting severe pressure on forest ecosystems, forcing people to leave their homes, combat food poverty, and suffer the negative effects of deforestation, as well as loss of livelihood (IPCC, 2007;World Bank, 2020). With regards, the climate has an unavoidable impact on the condition of vegetation and the growing environment too (Gao et al., 2014). Bangladesh ranks among the most vulnerable countries in the world in terms of climate change. Due to the increase in demand for fuel, agricultural land, timber, food remaining forests are at threat where agriculture is a dominant driving force for decreasing forest additionally, out of 3 billion people who live in the rural area almost 2.5 billion people are dependent on agriculture (Potapov et al., 2017). Also, this south Asian country is situated in the floodplains of the Padma, Jamuna, and Meghna river systems which is the biggest delta land as well as tropical moist deciduous and semi-evergreen forests, mangroves, and freshwater wetlands are characteristics (Olson et al., 2001). Major Roads, Rivers and Cities in Bangladesh indicating study area which is represented in Figure 1. It has only 11% of land covered by forest and it has the lowest forest-covered land per capita (0.009 ha person-1) for 156 million people among other countries (FAO, 2015).

Figure 1. Major Roads, Rivers and Cities in Bangladesh indicating Study Area

By adjusting the energy balance, temperature, hydrological and biological cycles due to this conversion, the vegetation index can be viewed as an important measure of environmental and anthropogenic impacts (Rousvel et al., 2013).

Recently many studies have assessed the impacts of climate change on vegetation (Pettorelli et al., 2005). A study observed that for the same native forest NDVI and Normalized Difference Water Index (NDWI) showed the highest correlation coefficient of 0.68, as well as Enhanced Vegetation Index (EVI) and Soil-Adjusted Vegetation Index (SAVI), were found correlated for the native forest (Silva et al., 2020). A study found the highest NDVI in Chittagong and Sylhet division eastern part of Bangladesh and lower NDVI in the southern region, the coastal areas towards the Bay of Bengal and some regions in north-western Rajshahi and Rangpur divisions including a small part in Sylhet division showed lower NDVI, this study also found an association between NDVI and rainfall rate (Islam & Mamun, 2015). A study on Khadimnagar National Park showed that from 1998 to 2010 amount of dense forest fallen from 526 hectors (67%) to 417 hectors (59%) whereas an improvement was observed in medium dense forest as it increased from 155 hectors (20%) to 317 hectors (40%) additionally, empty land significantly decreased from 104 hectors (13%) to 8 hectors (1%) (Redowan et al., 2014). Rahman (2013) found continuous deforestation occurring in the coastal district Patuakhali by comparing data from 1989 to 2010 which increases soil erosion, the loss of habitats and biodiversity and decreases in catchment area water-holding capacity. A study on Tengragiri under Barguna district spotted that Keora, Baen-Passur, and Sundri species were increased in Tengragiri forest where Keora plantation was more significant and increasing rate was 13.17 ha/year additionally, Sundri (27% of total area) always showed increasing signs where another species increases or decreases over this period (2000-2017). Whereas Gewa-Goran showed decreasing signs over the period (Islam et al., 2020). Traditional ground forest cover monitoring methods need a huge labor force, cost and plenty of time. Nowadays remote sensing data have become the staple data source for vegetation change detection applications due to cut the larger required time, larger synoptic view and digital format. Still, now the satellite data provide more accurate assessments than traditional methods (Franklin et al., 2000). Soil moisture is considered a crucial parameter for assessing vegetation status over a region. But it is troublesome to obtain data about soil moisture over large geographical areas, where remote sensing precipitation information can be a solution (Nightingale & Phinn, 2003). Cyclone SIDR (A tropical storm named as SIDR, that was an extremely severe cyclone, made landfall in Bangladesh's coastal region on November 15, 2007) hit badly on Khulna, Bagherhat, Barguna, Patuakhali and Pirojpur district which is a cause of vegetation reduction (0-30%) in most of the area where 62 hectares of agricultural land was devastated by SIDR in 2007 and massive damage occurred in forest, 343 hectares forest were damaged in the middle and northern part Patharghata (Islam & Mamun, 2015). By providing information on the spatiotemporal trend of land-use change in Barguna, the findings of this analysis will assist government agencies and the natural resources management authority. The main objectives of this study were to estimate vegetation cover and detect changes for 31 years of Barguna District and identify the factors affecting the increasing and decreasing of vegetation of the study area. The central research question of this paper is: How has the vegetation cover of Barguna district, Bangladesh changed between 1989 and 2020, and what are the factors driving these changes?

Literature Review

Environmental and ecological research have been interested in the issue of changing vegetation cover for many years. Monitoring and analysing changes in land use over time has been easier with the expansion of remote sensing and GIS technology. Mia et al. (2020) conducted a study on land-use change in the coastal area of Cox's Bazar, Bangladesh, using remote sensing and GIS technology. According to the study, the region has experienced major changes in land use and land cover over the last two decades, including a sharp rise in built-up areas and a sharp decline in vegetation cover. Overall, the study shows how satellite imagery may be used to monitor and manage coastal regions. Faruque et al. (2022) analysed the land-use changes in the Sundarbans, a mangrove forest in Bangladesh, using remote sensing and GIS technology. They discovered that the decrease in agricultural land and the increase in aquaculture areas lead to a transformation in the local population's occupation over time. Yet, both the ecosystem's human-induced activities and the biodiversity of mangrove forests have been negatively impacted by this transition. Abdullah et al. (2022) studied the Land Use/Land Cover Change on Present and Future Land Surface Temperature Chittagong, Bangladesh. They found that population growth, agricultural expansion, and industrialisation were the primary causes of vegetation cover change. The data show that between 1990 and 2020, there was an increase in built-up areas, waterbodies, and agricultural fields, while vegetation cover dropped and bare regions arose. Mamnun & Hossen (2020) analysed the spatio-temporal dynamics of forest cover change in the Chittagong Hill Tracts of Bangladesh using remote sensing and GIS technology. They discovered that shifting agriculture, logging, and the growth of communities were the main causes of deforestation and forest degradation in the region. The research emphasized the necessity for sustainable forest management techniques to protect the Chittagong Hill Tracts' biodiversity and ecosystem services. Huq et al. (2015) conducted a study on the impacts of climate change on agriculture in Bangladesh. They concluded that growing agricultural yields and altering precipitation patterns were both contributing to an increase in pests and diseases. In light of climate change, the report highlighted the need for adaption techniques and policies to help the agricultural sector. Akbar Hossain et al. (2022) studied the spatio-temporal dynamics of vegetation cover change in the Sundarbans mangrove forest using remote sensing and GIS. They discovered that during the last few decades, the Sundarbans' vegetation cover has dramatically diminished, mostly as a result of cyclonic storms and human activities like shrimp farming and wood harvesting.

The objective of the study of Islam et al. (2021) was to analyse changes in forest cover over time using the Normalized Difference Vegetation Index (NDVI) approach, and to determine the effect of co-management on such changes. The NDVI methodology was employed in the study, which is a commonly used method for assessing forest cover change throughout the world. Ma & Li (2022) studied at the Spatial-temporal aspects of urban growth in Shenyang, China, between 2001 and 2014. Shenyang's urbanization trend has significantly increased the amount of land available for construction, which has caused rapid changes in the morphology of the city. The study emphasizes the demand for efficient urban development and management to ensure Shenyang's sustainable urban growth. Chopra et al. (2022) found vegetation in urban settings is important in mitigating numerous environmental concerns brought on by development. Unsustainable human activity, on the other hand, have resulted in the degradation of urban forests. This research employed GIS and remote sensing technologies to identify changes in urban forests, giving baseline data on these changes and the effects of urbanization. This data may be used to develop strategies to ensure the long-term protection of urban ecosystems. Alwedyan (2023) showed the efficacy of GIS technology in obtaining accurate land use maps and comprehensive change statistics for the city of Irbid. The findings provide important data for formulating policies related to land use and urban expansion in the city, as well as making judgments and developing plans for the supply of adequate infrastructure. This technique will aid in addressing issues in urban planning in order to achieve sustainable urban management. The forest cover change analysis in the study of Redowan et al. (2014) involved the use of both supervised and NDVI classification approaches, both of which generated comparable results in identifying forest changes. The results showed that the increase in the area of medium dense forest cover was attributed to the establishment of short and long rotation plantations at various intervals, while the decline in dense forest cover was primarily caused by illegal felling, encroachment, and human settlement near forest areas. Thus, it is suggested that preventative measures be put in place to stop further forest degradation. By identifying changes in the spatial and temporal biodiversity patterns within the Sundarbans region, time-series remote sensing data analysis can be a useful tool to assist forest management methods.

Overall, the research indicates that Bangladesh is facing severe problems linked to changing vegetation cover and other environmental issues. In order to track and analyse land-use changes through time and get useful insights into their causes and effects, remote sensing and GIS technologies have been used extensively. In order to address these issues and advance environmental conservation and sustainable development, the literature emphasizes the necessity for sustainable management methods, strategies, and regulations.

Methodology

Study area

Barguna district is located in Barisal division at 21º48' and 22º29' north latitudes and between 89º52' and 90º22' east longitudes with 1,831.31 sq. km. (707.07 sq. miles) area of which 399.74 sq. km. is riverine and 97.18 sq. km. is under forest (BBS, 2013). Barguna is a coastal district situated between Khulna & Patuakhali which is the southern part of Bangladesh. Study location map showing digital elevation model is shown in Figure 2. Characteristics of the land of Barguna district are plain which is full of estuarine creeks and many rivers such as Bishkhali, Baleshwari, Burishwar, Payra, Andharmanik, Gajalia intersected this district and fallen into Bay of Bengal which is the southern boundary of this district (BBS, 2013). Jhalokathi, Barisal districts are on the north boundary, Pirojpur district and a part of Sundarbans under Bagerhat district is on the west boundary of Barguna District. The maximum temperature of Barguna district is 33.3°C and the minimum temperature is 12.1°C. The district's annual average rainfall is reported at 2506 mm (BBS, 2013).

Figure 2. Study location map showing digital elevation model

Data and satellite imageries collection

Landsat data were used in this study to show the vegetation coverage and identify change detection of vegetation overtimes. Landsat thematic mapper (TM) of 1989, 2004 and Landsat 8 Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) of 2020 were downloaded from the United States Geological Survey (USGS) Earth Explorer website ( https://earthexplorer.usgs.gov/) which is open-source data. Image acquisition date and characteristics are shown in Table 1. Landsat TM images have 7 spectral bands and spatial resolution is 30 meters for Bands 1 to 5 and 7, whereas Band 6 is thermal infrared which has 120 meters spatial resolution. Landsat 8 OLI-TIRS consists of 9 spectral bands. It has a spatial resolution of 30 meters for bands 1 to 7 and 9. Band 8 is a panchromatic band with a 15meters resolution.

Table 1. Image acquisition date and characteristics

Image acquisition date	Satellite/Sensor	Data Type	Path/Row	Sun Elevation	Projection, Datum, UTM Zone	Spatial resolution (meter)	Cloud cover (%)
11.02.2020	Landsat 8	OLI-TIRS	137/45	45.305171	UTM, WGS84, 45N	30	0
29.11.2004	Landsat 5	TM	137/45	44.125530	UTM, WGS84, 45N	30	0
11.04.1989	Landsat 5	TM	137/45	43.640022	UTM, WGS84, 45N	30	0

We used bands 1,2,3,4,5 and 7 of Landsat TM for supervised classification and bands 3, 4 for calculating Normalized difference vegetation index (NDVI). On the other hand, bands 2, 3, 4 and 5 of Landsat 8 OLI-TIRS were used for supervised classification and bands 4,5 for NDVI calculations. The spectral range in micrometers of each band has given in Table 2. Ancillary data such as Bangladesh administrative boundary, study area shapefile, roads shapefile, digital elevation models (DEM), water body shapefiles were acquired from the ESRI database. Temperature, precipitation, and Population data were collected from the Bangladesh Metrological Department (BMD) and Bangladesh Bureau of Statistics (BBS) respectively.

Table 2. Spectral range in micrometers of each band

Landsat TM Bands	Wavelength (micrometer)	Landsat OLI-TIRS Bands	Wavelength (micrometer)
Bands 1 (Blue)	0.45-0.52	Bands 1(coastal/aerosol)	0.43 - 0.45
Bands 2 (Green)	0.52 - 0.60	Bands 2 (Blue)	0.450 - 0.51
Bands 3 (Red)	0.63 - 0.69	Bands 3 (Green)	0.53 - 0.59
Bands 4 (NIR)	0.76 - 0.90	Bands 4 (Red)	0.64 - 0.67
Bands 5 (mid IR)	1.55 - 1.75	Bands 5 (Near Infra-red)	0.85 - 0.88
Bands 6 (thermal)	10.40 - 12.50	Bands 6 (Short wavelength Infra-red)	1.57 - 1.65
Bands 7 (Mid-Infrared)	2.08 - 2.35	Bands 7 (Short wavelength Infra-red)	2.11 - 2.29

Image processing

Image processing such as image atmospheric correction, image enhancement of all images required for this research has been done using the Semi-automatic classification plugin (SCP) of QGIS 3.10 and ArcGIS 10.4 version. We have set 0 percent cloud coverage, winter season, daytime as the criteria during image acquisition to minimize atmospheric disturbance like cloud, haze, water vapor, unwanted shade. The Landsat image has several bands that need to be composed. Layer stacking of individual bands was done to get the multispectral image with all band combinations. Region of interest (ROI) or study area has been extracted from the country shapefile of Bangladesh. The Layer stacking, Mosaicking and sub-setting process of Images were done using ERDAS IMAGINE 15 (Hussain et al., 2020; Prakasam & Biswas, 2009; Qasim et al., 2016). Landsat 8 data were calibrated to Top of Atmosphere (TOA) radiance using reflectance rescaling coefficients. In this process, thermal infra-red digital numbers (DN) have been converted to TOA spectral radiance using the metadata (MTL) file of Landsat that contained the necessary information. The following equation has been used to estimate TOA radiance:

$L\lambda =ML\times Q_{\mathrm{\text{cal}}}+AL$ ……………………………………... (1)

where, Lλ is TOA spectral radiance, without correction for the solar angle, ML is band-specific multiplicative rescaling factor from the metadata file, Q_cal denotes Quantized and calibrated standard product pixel values (DN) and AL means band-specific additive rescaling factor from the metadata file. Also, TOA spectral radiance was corrected for the sun angle following the equation:

$L\lambda =\frac{{\mathrm{\text{Lλ}}}^1}{{Sin(\theta }_{\mathrm{\text{SE}}})}$ …………………………………………………... (2)

where Lλ is TOA planetary reflectance, Lλ¹ is spectral radiance without correction, θ_SE means sun elevation angle that is found in the metadata file of Landsat data. Sun elevation is 49.00315815 for each scene of Landsat 8 OLI (Operational Land Imager).

Normalized Difference Vegetation Index (NDVI)

The Normalized Difference Vegetation Index (NDVI) is usually used for monitoring the changes in vegetation cover condition in various temporal and spatial scales (Liu et al., 2015) and also identify the responses of vegetation to regional and global climate change and it has a lot of advantage, the algorithm is simple in this method and it can roughly differentiate vegetation cover areas from other types of land cover area (Leprieur et al., 1994; Piao et al., 2011). It functions by considering the contrast of the absorption in the red band due to vegetation chlorophyll pigments as well as the reflection in the infrared band occurred by the cellular structure of the leaves. So, the resembling pattern to process as well as perceiving controls on the spatial pattern formation and variations are primary obstacles in vegetation science (W. Heil & P. A. van Deursen, 1996). Rouse et al. (1973) first discovered and accredited and later (Kriegler et al., 1969) elaborated the prime notion of NDVI (Hossain & Easson, 2015). NDVI calculation mainly depends on the reflectance value of Red and near-infrared (NIR) of the electromagnetic spectrum. It is evident and proved that chlorophyll in plants reflects quite well compare with visible bands. NDVI ranges from +1 to -1. The positive value of NDVI corresponds to vegetation while the negative value corresponds to non-vegetation. The more values, the more intensity of vegetation. Weier & Herring, (2000) stated that 0.1 and below the value of NDVI denotes water body, built-up area, barren soil, sand, rock, or snow; 0.2-0.3 value corresponds grassland and shrubs, while 0.6 or higher value indicates dense vegetation or tropical rainforests. The NDVI was estimated for Vegetation indices following the formula (Huyen et al., 2017):

$NDVI=\frac{NIR-RED}{NIR+RED}$ …………………………………………… (3)

where NIR is near infra-red reflectance (Band 4 of Landsat 5 TM and Band 5 of Landsat 8 OLI-TIRS), RED is the reflectance of visible red (Band 3 of Landsat 5 TM and Band 4 of Landsat 8 OLI-TIRS).

We used Landsat 5 TM data (Band 4 and 3) for 1989 and 2004 for NDVI calculation where the Spectral range of Band 4 and Band 3 is 0.77-0.90 µm and 0.63-0.69 µm respectively. Landsat 8 OLI-TIRS data have been also used for the year 2020 to show the coverage and change detection of vegetation of Barguna district. Band 5 (spectral range 0.85-0.88) and Band 4 (spectral range 0.64-0.67) generally uses for NDVI calculation for Landsat 8 OLI-TIRS Images. For Landsat 5 TM and Landsat 8 OLI-TIRS, NDVI was estimated following the equation:

${\mathrm{\text{NDVI}}}_{Landsat5\_TM}=\frac{Band4-Band3}{Band4+Band3}$ …………………………… (4)

${\mathrm{\text{NDVI}}}_{Landsat8\_OLI}=\frac{Band5-Band4}{Band5+Band4}$ …………………………… (5)

In this study, NDVI was calculated followed by Arc Toolbox-Spatial Analyst Tools-Map Algebra-Raster calculator of ArcGIS 10.4 version. Vegetation classification threshold values are given in Table 3. NDVI values have been set as ranged from 0 to 0.2, where the value lower than 0 indicates no vegetation in a specific area and there is the possibility of the presence of water bodies or wetland like ponds, canals, etc. A value greater than 0 to 0.1 represents there are soil, settlement, roads, char lands. And the value greater than 0.1 to less than 0.2 implies an area with moderate vegetation like a shrub, grassland, cropland/agricultural land. Greater than 0.2 value denotes a dense vegetation area like forest land.

Table 3. Vegetation classification threshold value in this study

Change classes	Value intensity
No Vegetation	Less than 0
Sparse Vegetation	0 to 0.1
Moderate Vegetation	Greater than 0.1 to less than 0.2
Dense Vegetation	>0.2

Supervised image classification

The land cover classification was performed for 1988, 2004 and 2020 imageries. We applied the maximum likelihood algorithm of the supervised classification method. The false-color tone of images was used to prepare training samples in training samples Manager of Image classification tools of ArcGIS 10.4. Author’s prior knowledge regarding the study area also helped in this regard. A field survey was conducted to verify ground truth data and to get better accuracy in image classification. Seventy-four points of Barguna district have been selected in this regard. The latitude and longitude of these places were recorded using a Global positioning system (GPS) named Garmin eTrex 30 GPS. To validate and crosscheck the accuracy of NDVI maps, supervised classification has been performed. Later, the maps produced from the supervised classification technique were checked by user accuracy, produce accuracy, overall accuracy, and kappa coefficient method. We classified NDVI maps into four categories (Table 3), similarly supervised classification maps have been done using reclassify tools of ArcGIS 10.4.

Change detection and transition analysis

We have considered there specific years of 1989, 2004 and 2020 for conducting this vegetation change detection study on a certain time interval. The main reason to select these three years is to show the status of vegetation and changes over the years due to several factors such as natural disasters, urbanization and population migration. Firstly, three different years (1989, 2004 and 2020) of imageries were classified using NDVI techniques to identify vegetation coverage of the study area. Later, supervised classification of these imageries has been done for the same years accordingly. All classified images were converted from Raster to vector data using Raster to Polygon tools of ArcGIS software. Then each area coverage based on vegetation categories has been calculated through Microsoft Excel 2016. Image differencing techniques have been followed to visualize and calculate the vegetation change detection. Vegetation coverage change detection maps from 1989-2000, 2000-2020 and 1989-2020 were calculated for both NDVI and supervised land use classified images. Area coverage and change detection were quantified in square kilometer (km²) units. A transition map was also prepared for the year 1989 to 2000, 2000-2020 and 1989 to 2020 to show the interconversion of vegetation coverage using Geoprocessing (intersect) tools of ArcGIS. In this process, from (before image) and to (after image) images transformation were visualized in maps and calculated through calculate geometry function of ArcGIS software. All data generated from the GIS and Remote sensing software have been calculated and visualized by using Microsoft Excel 2016 version and IBM SPSS version 25.

Accuracy assessment

Accuracy assessment is vital in terms of validation of classified images to justify the acceptability of the output. We performed an accuracy assessment for the supervised land use classification images of 1989, 2004 and 2020. For this purpose, we generated 45, 50 and 54 points randomly as user values according to the supervised land use classified images of 1989, 2004 and 2020 respectively (Table 4). Every point is then organized according to the identification (ID) pixel numbers such as 1, 2, 3, and so on based on vegetation categories. To crosscheck and find out producer values comparing with the user values, we used false color tone images of these years along with the surveyed ground truth data, google earth image, and literature review and author's prior knowledge regarding the study location. We clicked one by one user values and checked point accuracy (producer value). Though this process is time-consuming, but it brings better results in terms of image accuracy assessment. We used the following formula (6, 7, 8, 9) to calculate User accuracy, produce accuracy, overall accuracy and kappa coefficient for this study are as follows:

$UsersAccuracy=\frac{\mathrm{\text{CCPC}}}{\mathrm{\text{TCPC}}}\times 100$ …………………. (6)

$ProducerAccuracy=\frac{\mathrm{\text{CCPC}}}{\mathrm{\text{TRPC}}}\times 100$ …………….…. (7)

$OverallAccuracy=\frac{\mathrm{\text{TCCP}}}{\mathrm{\text{TRP}}}\times 100$ ………..………...…… (8)

$KappaCoefficient\left(T\right)=\frac{(TS\times TCS)-{\sum }_{}^{}(CT\times RT)}{{\mathrm{\text{TS}}}^2-{\sum }_{}^{}(CT\times RT)}\times 100$ ………...…… (9)

where CCPC is the number of correctly classified pixels in each category; TCPC is the total number of classified pixels in that category (the row total); TRPC is the total number of reference pixels in that category (the column total); TCCP is the total number of correctly classified pixels (diagonal); TRP is the total number of reference pixels; TS is the total sample; TCS is the total corrected sample; CT is the column total; RT is the row total.

Table 4. Overall classification accuracy and Kappa coefficient of supervised classification of 1989, 1997 and 2017 imageries of Barguna District.

Year	Overall classification accuracy	Overall Kappa statistics
1989	86.79	82.29
2004	92.16	89.52
2020	94.44	92.62

Results and Discussion

The NDVI analysis of Barguna district from 1989 to 2020 revealed that the area with no vegetation had the highest coverage of 226.65 km² (14.94%) in 2004, compared to 222.04 km² (14.64%) in 1989. However, the coverage decreased to the lowest of 215.72 km² (14.22%) in 2020. Figure 3 shows the image classification of vegetation coverage using NDVI and supervised image classification. The coverage of sparse vegetation increased to 257.28 km² (16.96%) in 2004, which was two times higher than the 134.44 km² (8.86%) coverage in 1989. It increased to 461.94 km² (30.46%) in 2020, which was two times higher than the coverage in 2004. The maximum increase of 541.01 km² (35.67%) in moderate vegetation occurred in 2004, and dense vegetation decreased since 1989, with a coverage of 423.99 km² (27.96%) in 2020.

Moreover, the supervised classification result of vegetation status in Barguna District from 1989 to 2020 showed that the no vegetation coverage area increased from 210.66 km² (13.89%) in 1989 to 219.18 km2 (14.45%) in 2004 but decreased to 217.22 km² (14.32%) in 2020. The coverage of sparse vegetation was the highest in 2020, at 467.92 km² (30.84%). The coverage of moderate vegetation increased to 529.21 km² (34.89%) in 2004, compared to 483.24 km² (31.85%) in 1989 and 485.01 km² (31.96%) in 2020. Dense forest cover continuously decreased to 347.21 km² (22.88%) in 2020 from 710.03 km² (46.8%).

Both NDVI and supervised classification images showed a similar trend in vegetation cover change from 1989 to 2020. The maximum coverage of no vegetation was observed in 2004, at 14.94% and 14.45% in NDVI and supervised classification images, respectively. The NDVI image showed a continuous decrease in dense forest cover from 32.42% in 2004 to 27.96% of the total land cover in 2020.

Figure 3. (a), (b), (c) Normalized Difference Vegetation Index; (d), (e), (f) Supervised Classification Vegetation Coverage of Barguna District from 1989 to 2020

Table 5 presents the output results of the study. The results show that there has been a slight increase in the no vegetation areas, with an increase of 0.67% from 1989 to 2004 and 2.67% from 2004 to 2020. However, the sparse vegetation cover has increased significantly. In 1989, the total area covered by sparse vegetation was 122.84 km², which increased to 327.50 km² in 2020. The increase rate of moderate vegetation cover has decreased from 31.54% to 10.94% from 1989 to 2020. In the same period, it was observed that 126.04 km² (30.79%) of moderate vegetation cover had decreased. The dense vegetation cover has declined drastically. Although the rate of change in forest areas was about 50% from 1989 to 2020, the area degraded was 67.68 km2 (16.54%) from 2004 to 2020 compared to other following years.

Table 5. Change Detection of Vegetation Status of Barguna District from 1989 to 2020 (NDVI Result)

NDVI Categories	Change Detection (1989-2004)		Change Detection (2004-2020)		Change Detection (1989-2020)
	Area (km2)	Area (%)	Area (km2)	Area (%)	Area (km2)	Area (%)
No vegetation	+4.61	+0.67	-10.93	-2.67	-6.32	-0.75
Sparse Vegetation	+122.84	+17.79	+204.66	+50.00	+327.50	+39.06
Moderate Vegetation	+217.77	+31.54	-126.04	-30.79	+91.73	+10.94
Dense Vegetation	-345.22	-50.00	-67.68	-16.54	-412.90	-49.25

Spatial techniques and indices were used to analyze the vegetation cover change from 1989 to 2020 in the study area, which covers a coastal region affected by significant environmental changes. The analysis involved NDVI and supervised classification, which helped to recalculate the study area to consider the massive environmental change. To ensure accuracy, multiple maps were compared, and inter-conversion results were analyzed. The displacement of a huge amount of vegetation areas was mainly caused by environmental change and population growth, as reflected in the different images and summarized in Table 6. All IV classes were visualized the different colors according to their magnitude and conversion of vegetation cover from less to dense categories. In this fact, the present environmental conditions of the study area are not properly developed and preserved of vegetation cover. As a result, the areas of sparse, moderate and forests are going to regression and decline (Lech-hab et al., 2015). Inter-conversion map of studied years is given in Figure 4. With the growth of the population, the demand for necessities is also getting higher. As the vegetation cover is relatively high and the public amenities are flawed at the beginning phase of population growth, the vegetation cover diminishes quickly when the consuming-through obliteration impact is a lot more grounded than the planting development impact. So, to meet these demands, vast areas with high vegetation cover will be destroyed for the construction of homes, highways, industrial plants, and stores, and numerous vegetation assets will be razed, resulting in a reduction in vegetation cover (Li et al., 2013). Forests are being quickly transformed into arable land as a means of meeting the growing need for food. Chemical fertilizer use would destroy the ecosystem of the forest by interacting with groundwater and moisture at plant roots. Yet another factor of climate change-induced geomorphological change affects the vegetation area. Coastal district Barguna is faced Sea Level change impact. Due to the rising of mean sea levels, erosion and deposition as well as the shoreline change are common in context (Hasan et al., 2021). Therefore, the coastal mangrove vegetation covers are also hampered by a lowering in rate.

Table 6. Inter-conversion of NDVI Classes from 1989 to 2020

1989	2020
		No Vegetation (km2)	Sparse (km2)	Moderate (km2)	Dense (km2)	Total (km2)
	No Vegetation (km2)	197.43	9.34	6.56	7.32	220.65
	Sparse (km2)	9.16	36.13	47.91	38.37	131.56
	Moderate (km2)	5.78	94.35	112.35	107.35	319.83
	Dense (km2)	7.01	322.95	243.48	271.53	844.96
	Total (km2)	219.38	462.77	410.30	424.56	1517

Figure 4. (a) Conversion map of Vegetation Cover from 1989 to 2004; (b) Conversion map of Vegetation Cover from 2004 to 2020; (c) Conversion map of Vegetation Cover from 1989 to 2020

Interconversion of NDVI classes (Table 6) from 1989 to 2020 revealed that most significant changes had been occurred dense vegetation to no vegetation (7.01 km²), moderate vegetation to sparse vegetation (94.35 km²), dense vegetation to sparse vegetation (322.95 km²) and dense vegetation to moderate vegetation (243.48 km²). It also noted that total 844.96 km² Dense vegetation had been interchanged out of 1517 km² area in the last 31 years. The primary causes of the decrease in vegetation cover in the Barguna District are climate change, natural catastrophes like floods and cyclones, population expansion, and rising demand for agricultural land, which results in deforestation and altered land use. Urbanization and industrialization cause land use changes and habitat fragmentation, and socioeconomic issues like poverty and lack of access to education can have an impact on how land is used and managed.

Conclusion

In this study, we compared the vegetation coverage of Barguna District in Bangladesh using the Normalized Difference Vegetation Index (NDVI), supervised classification, and transition analysis. We determined the quality of change detection by conducting an accuracy assessment. Such measures are crucial when satellite imagery is used for decision-making. As there is a practical need for precise studies using high-resolution data, it is expected that the demand for automated systems to extract surface reflectance data will increase. Therefore, change detection maps were created while considering socio-economic research and environmental repercussions. The change in the studied region from 1989 to 2020 was dominated by deforestation and loss of vegetation due to agricultural operations, urbanization, and particularly intense weather occurrences. Dense vegetated areas such as coastal mangrove forest areas were the most affected by the change. This study revealed that approximately 49.25 percent of the vegetation areas in Barguna District have undergone alterations over the past 31 years. However, Yet, this study might add to the current literature on vegetation cover change and its causes in Bangladesh's coastal regions. It can give useful insights on how to employ remote sensing and GIS technologies to monitor and analyze land-use changes over time. The findings of the study can potentially be used as a reference for future studies on vegetation cover change in similar areas. Practitioners concerned in land-use planning and management, conservation, and natural resource management may find the paper valuable. This analysis of vegetation cover change can assist in identifying areas in need of special attention for conservation and restoration activities. Remote sensing and GIS technologies can also aid conservation practitioners in monitoring and analyzing the effectiveness of their activities. Moreover, the research findings can aid decision-makers in formulating conservation, natural resource management, and land-use planning strategies that support sustainable development and safeguard the environment. Policymakers may also track and analyze land-use changes in other parts of Bangladesh by using the methods and tools given in the research.

Author Contributions

Conceptualization, M.S.I. and L.S; methodology, M.S.I; software, M.S.I., T.Y; investigation, M.S.I; data curation, M.S.I, T.Y.; writing—original draft preparation, M.S.I., T.Y; writing—review and editing, T.Y., M.S.I., S.K., M.S.H and L.S.; supervision, L.S. All authors have read and agreed to the published version of the manuscript.

Ethics Declaration

The authors declare that they have no conflicts of interest regarding the publication of the paper.

Funding

This research was funded by the National Key Research and Development Program of China (2018YFC0704703), the National Natural ScienceFoundation of China (71874174) and CAS Belt and Road MastersFellowship Program for international students at University of Chinese Academy of Sciences (UCAS), China.

Acknowledgments

We would like to thanks to the respondents and relevant officials of Barguna District of Bangladesh to collect necessary data for the study. Authors are much grateful to the Institute of Urban Environment (IUE), Chinese Academy of Sciences (CAS) and University of Chinese Academy of Sciences (UCAS), China. The first author would like to express gratitude to EQMS Consulting Limited for giving study leave opportunity for higher study. Authors also acknowledged USGS Earth Explorer for making Landsat data available online. In addition, Authors would like to show their gratitude to anonymous reviewers for their valuable comments and suggestions to improve the quality of the article.

References

•  Abdullah, S.,  Barua, D.,  Abdullah, Sk.  Md, A.,&  Rabby Y. W., (2022). Investigating the Impact of Land Use/Land Cover Change on Present and Future Land Surface Temperature (LST) of Chittagong, Bangladesh. Earth Systems and Environment, 6(1), 221–235. doi: https://doi.org/10.1007/s41748-021-00291-w.
•  Akbar Hossain,  K.,  Masiero,  M., &  Pirotti,  F. (2022). Land cover change across 45 years in the world’s largest mangrove forest (Sundarbans): The contribution of remote sensing in forest monitoring. European Journal of Remote Sensing, 1–17. doi: https://doi.org/10.1080/22797254.2022.2097450.
• Alwedyan, S. (2023). Monitoring Urban Growth and Land Use Change Detection with Gis Techniques in Irbid City, Jordan. International Review for Spatial Planning and Sustainable Development, 11(1), 253–275. doi: http://dx.doi.org/10.14246/irspsd.11.1_253.
•  BBS . (2013). District Statistics 2011, Barguna. http://203.112.218.65:8008/WebTestApplication/userfiles/Image/District%20Statistics/Barguna.pdf.
•  Chopra,  B.,  Singh,  B., &  Khuman , Y. S. C. (2022). Spatio-temporal Analysis of Land Use/Land Cover Changes in Urban Forest Ecosystem: A Case Study of Delhi Ridge Forest. International Review for Spatial Planning and Sustainable Development, 10(3), 128–147. doi: http://dx.doi.org/10.14246/irspsd.10.3_128.
•  FAO . (2015). Global Forest Resources Assessment 2015. Desk reference. Food and Agriculture Organization of the United Nations, Rome, 2015.
•  Faruque , Md. J.,  Vekerdy,  Z.,  Hasan , Md. Y.,  Islam , K. Z.,  Young,  B.,  Ahmed , M. T.,  Monir , M. U.,  Shovon , S. M.,  Kakon , J. F., &  Kundu,  P. (2022). Monitoring of land use and land cover changes by using remote sensing and GIS techniques at human-induced mangrove forests areas in Bangladesh. Remote Sensing Applications: Society and Environment, 25, 100699. doi: https://doi.org/10.1016/j.rsase.2022.100699.
•  Franklin,  J.,  Woodcock,  C., &  Warbington,  R. (2000). Multi-Attribute Vegetation Maps of Forest Service Lands in California Supporting Resource Management Decisions. Undefined. https://www.semanticscholar.org/paper/Multi-Attribute-Vegetation-Maps-of-Forest-Service-Franklin Woodcock/95e5a0fdb5e86855b03f1612e26ff0753b272dfb.
•  Gao , J.G.,  Zhang , Y.L.,  Liu , L.S., &  Wang , Z.F. (2014). Climate change as the major driver of alpine grasslands expansion and contraction: A case study in the Mt. Qomolangma (Everest) National Nature Preserve, southern Tibetan Plateau. Quaternary International, 336, 108–116. doi: https://doi.org/10.1016/j.quaint.2013.09.035.
•  Hasan , Md. M.,  Karmaker,  S., &  Rahman , Md. R. (2021). Geomorphological change assessment of south western coastal region: A case study of Mongla Upazila, Bagerhat, Bangladesh. Journal of Applied Water Engineering and Research, 9(1), 45–51. doi: https://doi.org/10.1080/23249676.2020.1831977.
•  Hossain , A. K. M. A., &  Easson,  G. (2015). Potential Impacts of the Growth of a Mega City in Southeast Asia, a Case Study on the City of Dhaka, Bangladesh. In W.-Y. Chen, T. Suzuki, & M. Lackner (Eds.), Handbook of Climate Change Mitigation and Adaptation (pp. 1–24). Springer New York. doi: https://doi.org/10.1007/978-1-4614-6431-0_68-1.
•  Huq,  N.,  Hugé,  J.,  Boon,  E., &  Gain,  A. (2015). Climate Change Impacts in Agricultural Communities in Rural Areas of Coastal Bangladesh: A Tale of Many Stories. Sustainability, 7(7), 8437–8460. doi: https://doi.org/10.3390/su7078437.
• Hussain, S., Mubeen, M., Akram, W., Ahmad, A., Habib-ur-Rahman, M., Ghaffar, A., Amin, A., Awais, M., Farid, H. U., Farooq, A., & Nasim, W. (2020). Study of land cover/land use changes using RS and GIS: A case study of Multan district, Pakistan. Environmental Monitoring and Assessment, 192(1), 2. doi: https://doi.org/10.1007/s10661-019-7959-1.
•  Huyen , N. T.,  Tu , L. H.,  Liem , N. D.,  Tram , V. N. Q.,  Minh , D. N., &  Loi , N. K. (2017). Assessing Impacts of Land Use and Climate Change on Soil and Water Resources in the Srepok Watershed, Central Highland of Vietnam (No. 2017–2; Agriculture and Development Discussion Paper Series). Southeast Asian Regional Center for Graduate Study and Research in Agriculture (SEARCA). doi: https://ideas.repec.org/p/sag/seadps/2017342.html.
• IPCC (Ed.). (2007). Climate change 2007: The physical science basis: contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press.
•  Islam , M. M., &  Mamun , M. M. I. (2015). Variations of NDVI and Its Association with Rainfall and Evapotranspiration over Bangladesh. Rajshahi University Journal of Science and Engineering, 43, 21–28. doi: https://doi.org/10.3329/rujse.v43i0.26160.
•  Islam , Md. A.,  Jimmy , A. N.,  Alam , Md. S., &  Khan , N. A. (2021). The use of multi-temporal Landsat normalized difference vegetation index (NDVI) data for assessing forest cover change of Lawarchara National Park. Environment, Development and Sustainability, 23(12), 17702–17722. doi: https://doi.org/10.1007/s10668-021-01408-x.
•  Islam, Md. M., Chowdhury, R. M. , Mostafa Zaman, A. K. M., Saifur Rahman, Md., Nazrul Islam, Md., Rudra, A. K., &  Salim Azad,  Md. (2020). Spatiotemporal mapping mangroves of Tengragiri wildlife sanctuary under Barguna district of Bangladesh using freely available satellite imagery. Modeling Earth Systems and Environment, 6(2), 917–927. doi: https://doi.org/10.1007/s40808-020-00728-7.
•  Kriegler , F. J.,  Malila , W. A.,  Nalepka , R. F., &  Richardson,  W. (1969). Preprocessing Transformations and Their Effects on Multispectral Recognition. 97.
•  Lech-hab , K. B. H.,  Issa , L. K.,  Raissouni,  A.,  El Arrim,  A.,  Tribak , A. A., &  Moussadek,  R. (2015). Effects of Vegetation Cover and Land Use Changes on Soil Erosion in Kalaya Watershed (North Western Morocco). International Journal of Geosciences, 06(12), 1353–1366. doi: https://doi.org/10.4236/ijg.2015.612107.
•  Leprieur,  C.,  Verstraete , M. M., &  Pinty,  B. (1994). Evaluation of the performance of various vegetation indices to retrieve vegetation cover from AVHRR data. Remote Sensing Reviews, 10(4), 265–284. doi: https://doi.org/10.1080/02757259409532250.
• Li, C., Kuang, Y., Huang, N., & Zhang, C. (2013). The Long-Term Relationship between Population Growth and Vegetation Cover: An Empirical Analysis Based on the Panel Data of 21 Cities in Guangdong Province, China. International Journal of Environmental Research and Public Health, 10(2), 660–677. doi: https://doi.org/10.3390/ijerph10020660.
•  Liu, X.,  Zhu, X.,  Li, S.,  Liu, Y., &  Pan,  Y. (2015). Changes in Growing Season Vegetation and Their Associated Driving Forces in China during 2001–2012. Remote Sensing, 7(11), 15517–15535. doi: https://doi.org/10.3390/rs71115517.
• Mamnun, M., & Hossen, S. (2020). Spatio-temporal analysis of land cover changes in the evergreen and semi-evergreen rainforests: A case study in Chittagong Hill Tracts, Bangladesh. International Journal of Forestry, Ecology and Environment, 2(2), 87–99. doi: https://doi.org/10.18801/ijfee.020220.10.
•  Ma, F., &  Li,  M. (2022). Spatial-temporal Characteristics of Urban Expansion in Shenyang from 2001 to 2014. International Review for Spatial Planning and Sustainable Development, 10(4), 161–173. doi: http://dx.doi.org/10.14246/irspsd.10.4_161.
• Mia, M. B., Hasan, T., & Akhter, S. H. (2020). Change Detection of Landuse-landcover in and around Cox’s Bazar-Teknaf Coastal Area of Bangladesh Using Satellite Images. The Dhaka University Journal of Earth and Environmental Sciences, 8(1), 1–9. doi: https://doi.org/10.3329/dujees.v8i1.50754.
•  Nightingale , J. M., &  Phinn , S. R. (2003). Assessment of Relationships Between Precipitation and Satellite Derived Vegetation Condition Within South Australia. Australian Geographical Studies, 41(2), 180–195. doi: https://doi.org/10.1111/1467-8470.00204.
•  Nowak , D. J.,  Crane , D. E., &  Stevens , J. C. (2006). Air pollution removal by urban trees and shrubs in the United States. Urban Forestry & Urban Greening, 4(3–4), 115–123. doi: https://doi.org/10.1016/j.ufug.2006.01.007.
•  Olson , D. M.,  Dinerstein,  E.,  Wikramanayake , E. D.,  Burgess , N. D.,  Powell , G. V.,  Underwood , E. C.,  ... &  Kassem, K. R. (2001). Terrestrial Ecoregions of the World: A New Map of Life on EarthA new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience, 51(11), 933-938. doi: https://doi.org/10.1641/0006-3568(2001)051[0933:TEOTWA]2.0.CO;2.
•  Pettorelli,  N.,  Vik , J. O.,  Mysterud,  A.,  Gaillard,  J.-M.,  Tucker , C. J., &  Stenseth , N. C. (2005). Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution, 20(9), 503–510. doi: https://doi.org/10.1016/j.tree.2005.05.011.
•  Piao,  S.,  Wang,  X.,  Ciais,  P.,  Zhu,  B.,  Wang,  T., &  Liu,  J. (2011). Changes in satellite-derived vegetation growth trend in temperate and boreal Eurasia from 1982 to 2006: CHANGE IN VEGETATION GROWTH OVER EURASIA. Global Change Biology, 17(10), 3228–3239. doi: https://doi.org/10.1111/j.1365-2486.2011.02419.x.
•  Potapov,  P.,  Hansen , M. C.,  Laestadius,  L.,  Turubanova,  S.,  Yaroshenko,  A.,  Thies,  C.,  Smith,  W.,  Zhuravleva,  I.,  Komarova,  A.,  Minnemeyer,  S., &  Esipova,  E. (2017). The last frontiers of wilderness: Tracking loss of intact forest landscapes from 2000 to 2013. Science Advances, 3(1), e1600821. doi: https://doi.org/10.1126/sciadv.1600821.
•  Prakasam,  C., &  Biswas,  B. (2009). Land use, land cover change study in Ausgram I&II blocks, Burdwan District, West Bengal using remote sensing and GIS. Indian Journal Landscape Systems and Ecology Studies, 32(2), 8–10.
•  Qasim,  H.,  Luqman,  M., & Ullah  Khan,  S. (2016). A Study of Forest Land Cover Changes Using Satellite Remote Sensing in Thatta District Pakistan. Science International, 28(3).
•  Rahman , Md. M. (2013). Temporal Change Detection of Vegetation Coverage in Patuakhali Coastal Area of Bangladesh Using GIS & Remotely Sensed Data. International Journal of Geomatics and Geosciences, 4, 36–46.
•  Redowan,  M.,  Akter,  S., &  Islam,  N. (2014). Analysis of forest cover change at Khadimnagar National Park, Sylhet, Bangladesh, using Landsat TM and GIS data. Journal of Forestry Research, 25(2), 393–400. doi: https://doi.org/10.1007/s11676-014-0467-9.
•  Rouse,  J.,  Haas , R. H.,  Schell , J. A., &  Deering,  D. (1973). Monitoring vegetation systems in the great plains with ERTS. Undefined. https://www.semanticscholar.org/paper/Monitoring-vegetation-systems-in-the-great-plains-Rouse-Haas/fb2f60fe0fe2874e5cbf927a2556d719c32eac29.
•  Rousvel,  S.,  Armand,  N.,  Andre,  L.,  Tengeleng,  S.,  Alain,  T., &  Armel,  K. (2013). Comparison between Vegetation and Rainfall of Bioclimatic Ecoregions in Central Africa. Atmosphere, 4(4), 411–427. doi: https://doi.org/10.3390/atmos4040411.
• Silva, M. V. da, Pandorfi, H., Lopes, P. M. O., Silva, J. L. B. da, de Almeida, G. L. P., Silva, D. A. de O., Santos, A. dos, Rodrigues, J. A. de M., Batista, P. H. D., & Jardim, A. M. da R. F. (2020). Pilot monitoring of caatinga spatial-temporal dynamics through the action of agriculture and livestock in the brazilian semiarid. Remote Sensing Applications: Society and Environment, 19, 100353. doi: https://doi.org/10.1016/j.rsase.2020.100353.
• Turner‐Skoff, J. B., & Cavender, N. (2019). The benefits of trees for livable and sustainable communities. PLANTS, PEOPLE, PLANET, 1(4), 323–335. doi: https://doi.org/10.1002/ppp3.39.
• W.  Heil, G., & P. A. van Deursen,  W. (1996). Searching for patterns and processes: Modelling of vegetation dynamics with Geographical Information Systems and Remote Sensing. Acta Botanica Neerlandica, 45(4), 543–556.
•  Weier,  J., &  Herring,  D. (2000). Measuring vegetation (NDVI & EVI). Earth observatory
•  World  Bank. (2020). Climate change. https://www.worldbank.org/en/topic/climatechange/overview.