INTRODUCTION
Mangrove forests are unique saline woodlands and shrublands formed by trees and shrubs in tropical and subtropical coastal regions, providing vital ecosystem services such as firewood, fish, and construction materials, and contributing significantly to primary productivity and economic development They play a crucial role in bio-protection by preventing coastal erosion, safeguarding against tropical storms, tsunamis, and other natural hazards, while also acting as carbon sinks that help mitigate global warming However, over the past fifty years, mangrove ecosystems have faced a sharp decline, with a 35% reduction worldwide due to human activities like aquaculture, agriculture, logging, and urban development, as well as natural disasters such as tsunamis and tropical storms Future threats include sea-level rise and climate change, alongside increasing coastal population growth, which threaten to accelerate mangrove loss Therefore, it is imperative for governments to develop effective plans and strategies for monitoring and conserving these invaluable ecosystems.
Remote sensing is highly effective for monitoring and mapping threatened mangrove ecosystems, as demonstrated by numerous global studies (Claudia Kuenzer, 2011) This technology provides essential data for habitat inventories, change detection, ecosystem evaluation, and productivity assessment of mangrove forests Additionally, remote sensing aids in planning field surveys and ongoing conservation efforts, making it an invaluable tool for the sustainable management of these vital ecosystems.
Vietnam is actively working to conserve and protect its mangrove forests, recognizing their vital ecological and economic importance Historically, the country has faced significant mangrove deforestation caused by land use changes and inadequate management policies, leading to severe habitat loss (Thuy Dang Truong, 2018) In recent years, Vietnam has implemented advanced conservation strategies and sustainable land management practices to restore and preserve these crucial ecosystems.
2 remote sensing, various techniques are provided from many satellite systems to improve the efficiency in monitoring mangrove forests (LU & WENG, 2007)
There is no universal classification method or satellite data recommended for mapping mangrove forests, as noted by Congalton (2001), Hankui K Zhang (2018), and Heumann (2011) For mapping moderate-sized areas, high-resolution aerial photography combined with ground validation offers the best accuracy in delineating different mangrove communities or zones While remote sensing provides significant advantages, acquiring high-resolution satellite imagery can be costly; however, platforms like Landsat-8 OLI and Sentinel-2 MSI offer accessible, free, medium to high-resolution images for effective mangrove forest monitoring.
Coarser spatial resolution images are well-suited to traditional pixel-based classification methods that rely on statistical analysis, while high-resolution images generally improve classification accuracy but can complicate land cover discrimination due to increased sub-classes and spectral mixing As resolution becomes finer, distinguishing between land cover types becomes more challenging, requiring more advanced approaches Object-based classification methods offer a promising alternative by focusing on true spatial patterns and meaningful object features rather than individual pixels, enhancing land cover discrimination and classification effectiveness.
The accuracy of the classification process is significantly influenced by the characteristics of the study site (Young, 2017) Boundary differentiation can be challenging when distinguishing scattered mangroves or clusters of trees along coastal areas (Manson, 2001; Heumann B W., 2011) In Ninh Binh province, the sparse and small canopy mangrove populations, typical of northern Vietnamese provinces, are particularly difficult to classify due to the large seasonal temperature variations and lower annual precipitation (Phan Nguyen Hong T V., 1999).
Tidal regime plays a crucial role in reducing the accuracy of mangrove mapping using remote sensing techniques (Kerrylee Rogers, 2017) The fluctuating presence or absence of sea water beneath the mangrove canopy significantly alters reflectance values, making remote sensing data interpretation more complex Understanding tidal influences is essential for reliable mangrove detection and monitoring from satellite imagery.
Discrimination at a single tidal stage can limit the accuracy of mangrove classification (Manson, 2001; Kuenzer, 2011) Combining images from different tidal stages enhances the ability to accurately identify and characterize mangrove zones, as it captures temporal variations that single-scene approaches may miss This multi-tidal analysis improves classification precision compared to traditional methods that rely on individual satellite images (Kerrylee Rogers, 2017).
The aim of this study is to determine the suitable classification methods of mangrove forests in Ninh Binh province with free satellite imagery
LITERATURE REVIEW
GIS and Remote sensing
2.1.1 Concept of GIS and Remote sensing
The term Geographical Information System (GIS) is used in geographically oriented computer technology that integrate systems in substantive applications or generation of massive interest worldwide (Paul A Longley, 2005)
Remote sensing (RS) is the science and practice of obtaining information about objects without direct contact by analyzing reflected and emitted electromagnetic radiation from Earth's terrestrial, aquatic ecosystems, and atmosphere It plays a crucial role in monitoring and understanding environmental changes, making it an essential tool in Earth observation.
Remote sensing data captures spectral, spatial, temporal, and polarization signatures of sensors and targets, enabling effective discrimination of earth surface features across different wavelengths Its ability to provide a synoptic view with repetitive, calibrated coverage allows for accurate detection of changes over time, making it a superior alternative to traditional natural resource management methods Widely applied in key sectors such as forestry, agriculture, water resources, land use planning, geology, environmental monitoring, coastal zone management, marine surveillance, and infrastructure development, remote sensing plays a crucial role in diverse operational applications.
Remote sensing data plays a crucial role in correcting, updating, and maintaining cartographic databases and GIS, as demonstrated by numerous studies and projects over recent decades (Campbell, 1987) GIS and remote sensing are specialized spatial data technologies, each with unique methods for data representation and analysis, particularly in vegetation analysis and modeling (Goodchild, 1994) Advances in integrating GIS with remote sensing tools and the development of sophisticated software have created powerful capabilities to acquire, store, manipulate, analyze, and visualize spatial data according to user-specific needs.
Launched in February 2013, the Landsat-8 satellite is equipped with the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS), offering a 16-day revisit cycle over a 185 km swath It features nine reflective wavelength bands, including six land application bands with a 30-meter resolution, integrated into its imagery Developed through a collaboration between NASA and the U.S Geological Survey (USGS), NASA was responsible for the satellite's design, construction, and launch, ensuring advanced capabilities for Earth observation and land monitoring.
During the on-orbit calibration phases, the satellite was known as the Landsat Data Continuity Mission (LDCM) Following this period, the US Geological Survey (USGS) took over routine operations, and the satellite was renamed Landsat-8 USGS is responsible for post-launch calibration, satellite operations, data product generation, and data archiving at the Earth Resources Observation and Science (EROS) Center.
Landsat-8's Operational Land Imager (OLI) marks a significant advancement in spatial technology, offering enhanced imaging capabilities As a push-broom sensor with a four-mirror telescope and 12-bit quantization, OLI captures data across visible, near-infrared, and shortwave infrared spectral bands, along with a panchromatic band Notably, it introduces two new spectral bands—one specifically designed for detecting cirrus clouds and another optimized for coastal zone observations—thereby expanding the satellite's applications and improving data accuracy.
Table 2.1 Specification of Landsat-8 OLI/TIRS
Band 6—Short-wave infrared (SWIR 1) 1.566 - 1.651 30
Band 7—Short-wave infrared (SWIR 2) 2.107 - 2.294 30
The Sentinel-2 satellites were launched in June 2015 having a 10-day repeat cycle over
290 kilometers swath and, is equipped with the Multi Spectral Instrument (MSI) which allows us to acquire 13 reflective wavelength spectral bands including 4 visible and near-
6 infrared bands with 10 meters resolution, 6 red edge, near infrared and short wave infrared bands with 20 meters resolution and 3 other bands with 60 meters resolution (M Drusch,
The Copernicus Sentinel-2 mission consists of two polar-orbiting satellites in a sun-synchronous orbit, phased at 180° to ensure comprehensive coverage These twin satellites are designed to monitor land surface variability by providing high-resolution, multispectral images with a high revisit frequency Their primary mission is to deliver systematic global data useful for creating land-cover maps and detecting land changes, supporting various environmental monitoring and land management applications.
Table 2.2 Specification of Sentinel-2 MSI
Band 10—Shortwave Infrared (SWIR) - Cirrus 1.373 60
Image classification
Pixel-based classification is the traditional method for land use and land change (LULC) mapping, relying on statistical analysis of individual pixels (L Wang, 2004) This approach is popular because pixels are the fundamental spatial units of satellite images, making the method straightforward and easy to implement However, pixel-based classification primarily focuses on spectral properties of single pixels, often neglecting spatial context and relationships within the image.
Pixel-based classification relies on clearly defined and differentiated class characterizations, which are often challenging to achieve in real-world scenarios In land use and land cover (LULC) studies, classes such as water bodies, bare land, and vegetation provide consistency and stability; however, issues arise from various sources at a more detailed level A key limitation of pixel-based methods is that they do not consider the surrounding pixels, which could assist in more accurate identification of the target pixel Common pixel-based classification techniques include the Normalized Difference Vegetation Index (NDVI) and supervised maximum likelihood classification, both widely used for their effectiveness in remote sensing analysis.
2.2.2 Normalized difference vegetation index (NDVI) classification method
Plants exhibit low reflectance in the blue and red bands due to chlorophyll absorption, which causes they to appear green to the human eye; however, they strongly reflect near-infrared radiation, with this reflectance influenced by leaf cellular structure affected by environmental factors like soil moisture, nutrients, salinity, and leaf stage (Machado, 2002) The contrast between vegetation and soil peaks in the red and near-infrared wavelengths, enabling spectral reflectance data to be used for calculating vegetative indices that correlate with plant health, photosynthetic activity, and productivity (Adamsen, 1999) Vegetation indices, such as those derived from red and near-infrared bands, can predict photosynthetic activity by leveraging the spectral relationships affected by pigment absorption, significantly enhancing the signal related to vegetation health (M I El-Gammal, 2014) These indices provide reliable measures of vegetative activity, and advancements in satellite technology have made sensors with red and near-infrared capabilities common, facilitating widespread use of indices like NDVI in recent research (Brown, Pinzon, & C.J., 2006) NDVI is calculated from reflectance measurements in the red and near-infrared spectrum, serving as a key indicator in remote sensing studies of vegetation vigor.
2.2.3 Supervised maximum likelihood classification method
Maximum likelihood classification, or supervised classification, has demonstrated superior accuracy compared to unsupervised methods when appropriate training sites are provided, leveraging its powerful decision rule (Md Mijanur Rahman, 2013) This classification approach is commonly used for the quantitative analysis of remote sensing data, requiring users to supervise the pixel classification process by specifying spectral signatures associated with each class through selected training sites The process involves selecting representative sample areas of known cover types, allowing the computer to determine their spectral signatures, including statistical parameters like mean and variance (Xavier Ceamanos, 2017) The accuracy of classification heavily depends on the proper selection of training sites that capture the full variability within each class (Md Mijanur Rahman, 2013) The algorithm then uses these spectral signatures to classify the entire image, ideally resulting in minimal overlap between different classes for precise results.
The maximum likelihood classifier is a widely used supervised classification method in remote sensing, where each pixel is classified into the class with the highest probability This method incorporates prior information and class probabilities based on knowledge of the area, enhancing classification accuracy By ruling out unlikely class options for individual pixels, the maximum likelihood classifier significantly improves the reliability of remote sensing image analysis.
2.2.4 Object-based classification (OBC) method
Advancements in spatial technology have led to improved satellite product spatial resolution over time, enabling more precise and detailed analysis This progress has facilitated the adoption of object-based classification methods that focus on extracting texture, tone, and geometric features of objects, enhancing accuracy in land cover and land use mapping (Qiong Hu).
The approach builds upon decades of remote sensing image analysis techniques, including segmentation, edge detection, feature extraction, and classification (Shao, 2012; Yu, 2006; R Kettig, 1976) Recently, OBC applications have emphasized the identification and classification of target object features, particularly in high-resolution imagery with spatial details of one meter or less.
Image classification using traditional supervised and unsupervised spectral algorithms often strains computational resources, highlighting the need for more efficient methods (G.J Hay, 2008) Object-Based Classification (OBC) relies fundamentally on image segmentation to generate homogeneous regions that provide enhanced spatial information over single-pixel approaches The latest advancements in Object-Based Image Analysis (OBIA) focus on automating image processing to create time-efficient tools for natural resource monitoring Although the use of OBC has increased in recent years due to its numerous advantages, its effectiveness has not yet been quantitatively validated, as reported in recent studies (T Blaschke, 2010).
Overview of mangrove
Mangroves are a diverse group of primarily tropical trees and shrubs that thrive in the challenging intertidal zones faced with dynamic natural conditions (J.E Sterling, 2006) Coastal mangroves are typically located within the tropics and subtropics, spanning approximately [insert specific geographic range], providing vital ecological functions such as shoreline protection, habitat for wildlife, and carbon sequestration Recognizing their importance and distribution is essential for environmental conservation and sustainable coastal management.
308 N and 308 S latitude (Long, 2014) The term ‘mangrove’ is used to describe the ecosystem and the plant families that have developed with distinguished adaptations to survive at the coastal environment (Tomlinson, 1986)
Mangrove forests thrive in diverse environmental conditions and exhibit unique adaptive features, including salt-excreting leaves, exposed breathing root systems, and highly productive viviparous propagules (Duke, 1992) These forests often consist of dwarf, stunted trees in areas with high salinity or disturbed natural conditions, while in more favorable estuarine environments, mangroves can extend for kilometers inland, forming extensive coastal ecosystems (FAO, 2007).
Mangroves have been widely exploited throughout history in coastal countries worldwide, serving various valuable purposes Understanding their current and historical extent, condition, and uses is essential for effective forest management and informed decision-making These ecosystems provide critical functions and services at both local and national levels, supporting livelihoods and protecting communities Rural populations depend heavily on wood and non-wood resources from mangrove forests, which also serve as natural barriers against natural disasters Additionally, mangroves play a vital role in conserving biological diversity by offering habitats, nutrients, and nurseries for numerous species.
The total number of mangrove species remains under debate, but they can be classified into key plant families such as Rhizophoraceae, Avicenniaceae, and Combretaceae, which have developed unique physiological and structural adaptations to thrive in brackish water habitats Despite extensive global research on mangrove forests, focusing on their dynamics across various spatial and temporal scales, comprehensive data on the overall status and trends of mangrove extent remains limited.
2.3.1 Mangrove status in the world
The estimations of global mangrove forest area vary among different projects and reports
A groundbreaking study by the Food and Agriculture Organization (FAO) and the United Nations Environmental Program (UNEP) estimated that approximately 15.6 million hectares of mangrove forests existed worldwide in 1980 More recent research indicates that the global mangrove area ranges between 12 and 20 million hectares, highlighting changes and uncertainties in mangrove coverage over time.
Asia holds the largest share of mangrove forests, accounting for 42% of the global area, followed by Africa with 20% Despite its smaller land area, Asia boasts the highest mangrove species diversity, with an estimated 50 to 70 species worldwide Coastal mangrove forests are under significant pressure from human activities such as agriculture, aquaculture, and tourism, leading to rapid habitat loss Since 1980, Asia has experienced the most substantial decline, losing approximately 1.9 million hectares of mangrove forest, highlighting a critical need for conservation efforts.
The global rate of mangrove loss has decreased to approximately 0.5% per year in recent years, thanks to dedicated conservation efforts by various environmental organizations and projects (FAO, 2007) Sustainable practices such as plantation initiatives and natural regeneration programs are actively being implemented in many countries to restore and expand mangrove forests effectively.
Vietnam is home to 29 provinces and cities that feature vital mangrove ecosystems and coastal wetland habitats, primarily located along the northern, southern, eastern, and southwestern coasts Historically, Vietnam's mangrove forests covered over 250,000 hectares in 1943, but this area declined significantly to approximately 168,689 hectares by 2014 Notably, mangrove forests are most abundant and flourishing in the Mekong Delta, highlighting their ecological importance in this region.
Southwest Vietnam's provinces and cities account for nearly 70% of the country's mangrove forest area, with an estimated 89,837 hectares—making it the largest mangrove zone in Vietnam The South East region follows, covering approximately 42,000 hectares, while North East encompasses about 20,486 hectares The composition, distribution, and development of Vietnamese mangroves are heavily influenced by environmental factors such as salinity levels, climate conditions, tidal regimes, and site-specific factors, with tidal regime being a crucial determinant of forest structure (Phan Nguyen Hong T V., 1999) Key mangrove species found in Vietnam include Rhizophora apiculata and Rhizophora mucronata, which play vital roles in maintaining the health and biodiversity of these ecosystems.
The mangrove ecosystem, which includes species such as Bruguiera gymnorrhiza, Avicennia alba, Avicennia marina, Sonneratia alba, and Nypa fruticans, has experienced significant damage due to human activities Overexploitation of mangrove resources, land use changes for agriculture and tourism, and increased economic development and population growth have led to widespread habitat loss Consequences of this degradation include land deterioration, water pollution, spread of plant diseases, abandonment of bare land, coastal erosion, and elevated salinity levels, posing serious challenges for government conservation efforts.
2.3.3 Remote sensing application on mangrove forest management
Remote sensing technology plays a crucial role in the effective management and conservation of mangrove forests by enabling comprehensive, large-scale data collection essential for decision-making Its synoptic, repetitive, and multi-spectral capabilities make it particularly suitable for monitoring the dynamic characteristics of mangrove ecosystems over time and space Satellite data provides valuable insights into various components of the coastal environment, including wetland conditions, shoreline changes, brackish water areas, suspended sediments, and mangrove density, demonstrating its importance for mapping, monitoring, and planning efforts As the threat of mangrove area loss increases, leveraging remote sensing is increasingly vital for sustainable coastal ecosystem management at the national level.
For decades, remotely sensed data has been essential in monitoring the condition and extent of mangrove forests using various satellite imagery, from low to very high spatial resolution, and diverse classification methods (Wang et al., 2004; Spalding, 1997; Vaiphasa, 2005) Despite no universal classification method or satellite data type being established for mangrove mapping, research has focused on developing complex algorithms that combine indices like NDVI with traditional classification techniques such as maximum likelihood and minimum-distance methods (Congalton, 2001; Hankui Zhang, 2018; Heumann, 2011; Bahuguna, 2001) Pixel-based approaches have enabled discrimination of mangrove vegetation from other land uses but often yield moderate to poor results when leveraging multispectral data (Neukermans, 2008; Vaiphasa, 2006) To overcome these limitations, object-based classification methods were introduced, utilizing more spatial information to improve accuracy in mangrove detection.
Commercial satellite imagery has been essential for mangrove forest monitoring for decades; however, high-resolution images like IKONOS or QuickBird can be costly, posing financial challenges for researchers Medium-resolution imagery, such as Landsat-8 and Sentinel-2, offers a cost-effective alternative by providing multispectral data suitable for regional-scale mangrove mapping Studies have demonstrated that while high-resolution images improve discrimination among mangrove and other vegetation types, medium-resolution images have proven effective for mapping and classifying mangrove and non-mangrove areas at broader scales (Aschbacher, 1995; Rasolofoharinoro, 1998; Selvam, 2003; Gao, 1998; Brown, Pinzon, & C.J., 2006; Hankui K Zhanga, 2018).
An intensive field campaign is crucial for accurately distinguishing mangroves from neighboring land covers, as emphasized by Claudia Kuenzer (2011) Gaining a deep understanding of local knowledge and conducting comprehensive fieldwork are essential for calibrating and validating classification results However, the often inaccessible or complex nature of mangrove ecosystems presents significant challenges to field research efforts, underscoring the need for advanced methods to improve classification accuracy.
Remote sensing technology has been utilized for decades to effectively manage mangrove forests, owing to its capabilities in capturing spatial and temporal data Various studies have demonstrated the potential of remote sensing for mapping and classifying land use and land cover (LULC) across multiple scales, employing different classification approaches and satellite imagery The availability of freely accessible satellites like Landsat-8 and Sentinel-2 provides significant advantages for monitoring mangrove ecosystems, enabling researchers to track changes and delineate mangrove extents over time Despite these advancements, there is a notable lack of remote sensing studies focused on mangrove forests in Ninh Binh province, which this study aims to address by utilizing free satellite imagery to accurately map and assess mangrove cover in the region.
GOAL, OBJECTIVES AND SCOPE
Goal
Providing scientific basis for remote sensing application for better management of mangrove forest in Ninh Binh province.
Objectives
To achieve the study’s goal, essential steps must be undertaken, with clearly defined specific objectives to guide the process These objectives aim to establish a scientific foundation for analyzing mangrove forests in Ninh Binh province using remote sensing technology By delineating precise actions, the research ensures accurate assessment and effective monitoring of the mangrove ecosystems, supporting conservation efforts and sustainable management in the region.
- Constructing thematic map of current mangrove forest in Ninh Binh province and estimating the mangrove forest extent area
- Assessing and comparing the classification accuracy of pixel-based and object-based approaches with Landsat-8 and Sentinel-2 satellite images
- Constructing dynamic map of mangrove forest in Ninh Binh province within the period of 2013 to 2019 using multi-temporal remote sensing data
- Quantifying the changes of mangrove forest extent in Ninh Binh province from 2013 to
- Proposing appropriate remote sensing approaches and further study for the mapping of mangrove forest area in Ninh Binh.
Study scope
Ninh Binh province is located in northern Vietnam with the population of 169,000 and total area of 215 km 2 The province is bordered by four provinces, including: Thanh Hoa province,
Ha Nam, Nam Dinh, and Hoa Binh provinces were separated from Ninh Binh province by natural barriers including the Tam Diep mountain range, which extends from northwest to southeast, marking the boundary between Ninh Binh and Thanh Hoa province The Day River acts as a natural border, dividing Ninh Binh from Nam Dinh and Ha Nam provinces The region's economy is predominantly driven by the industrial and service sectors, with an impressive average annual GDP growth rate of 15.35% (Cuc, 2011).
The province is traversed by two major rivers, the Red River and the Ma River, which play a vital role in the region's ecology and economy Its tropical monsoon climate features a hot season from April to October and a cooler season from November to March, creating diverse weather patterns The area maintains an average annual temperature of approximately 24°C, contributing to its lush landscapes and agricultural productivity.
The region experiences an annual rainfall of approximately 1,760 mm, contributing to its lush and diverse ecosystem (Nguyen Khanh Van, 2000) The tidal cycle lasts about 23 hours, with a mean tidal amplitude ranging from 150 to 180 cm, leading to significant high tidal inundation during the cold season from December to February (Phan Nguyen Hong V., 2004) This combination of substantial rainfall and tidal variations plays a crucial role in shaping the area's coastal environment and hydrological patterns.
Ninh Binh province boasts the largest forest area in the Red River Delta, covering over 19,000 hectares of both natural and plantation forests Among these, mangrove forests are prominent, particularly along the coastal line where wetland habitats are abundant.
Ninh Binh's topography consists of three primary terrains: delta, mountainous, and coastal areas The delta region, covering over 100,000 hectares, is the most populous, housing nearly 90% of the province’s residents and primarily supporting agriculture The mountainous terrain, situated along the Tam Diep range, spans approximately 35,000 hectares, representing about 24% of the province and encompassing parts of five districts The smallest area is the coastal zone, which extends over 6,000 hectares along 15 kilometers of shoreline within five communes of Kim Son district; this coastal land area is rapidly expanding due to sediment accumulation in the Red River Delta, averaging 28 meters of land growth annually The coastal mangrove forests, primarily located on the east coast, cover a limited area due to historical planting initiatives like the Japanese Red Cross project in the 1990s The 15 km coastline supports over 500 plant and animal species, including 50 types of cultivated and natural mangrove plants, vital for the region’s biodiversity.
A specific study site for classification processing was identified based on reference material from Google Earth Pro The mangrove forest in Ninh Binh province is delineated by rivers on both sides, serving as clear administrative boundaries that separate it from neighboring mangrove areas in Nam Dinh and Thanh Hoa provinces The study site boundary was precisely created using the “add polygon” tool in Google Earth Pro, then exported as a KML file and converted to KMZ format for further analysis.
Using the “Conversion” tool in ArcMap 10.4 software, various classification methods were then applied to the designated study site These techniques aimed to reduce redundant land cover information, thereby enhancing the accuracy of distinguishing mangrove forests from other land types.
The study focuses on the coastal area of Ninh Binh province in Vietnam, as illustrated in Figure 3.1 The map provides an overview of Vietnam and highlights Ninh Binh's location, while Sentinel-2 satellite images depict both the broader coastal region and the specific study site within Ninh Binh province These visualizations are essential for understanding the geographic context and environmental characteristics of the research area.
METHODOLOGY
Materials
Mangrove forests are primarily found in unique wetland areas that are periodically submerged by high tides, which can significantly affect detection and mapping accuracy The contrasting spectral signatures of submerged versus land-covered mangroves often lead to underestimating their true area (Xuehong Zhang, 2017) Various advanced techniques and technologies can help mitigate the impact of tidal regimes on mangrove mapping, but they tend to be costly and require multiple sensor types to accurately identify submerged forest areas Studying the tidal characteristics of the Gulf of Tonkin, where the Ninh Binh province mangrove coastal area is located, provides valuable references for establishing consistent and reliable mangrove boundary delineations.
Neap tides and spring tides are critical stages of the tidal cycle, characterized by significant fluctuations in sea water levels, especially along coastlines The timing and frequency of tidal inundation, or hydroperiod, play a central role in determining the distribution of mangrove communities (Kuenzer, 2011) Developing an automated classification algorithm for mangrove mapping that works consistently across different tidal stages is challenging due to these dynamic water level changes According to Kerrylee Rogers (2017), integrating satellite imagery captured at various tidal stages improves classification accuracy, as combining data from different dates—based on automatic tidal estimation algorithms and tidal regime studies—enhances the discrimination of mangrove forests across varying hydroperiod conditions.
Ninh Binh province, located along the shoreline of the Red River delta, experiences diurnal tides with a 14 to 16-day neap-spring tide cycle (Minh Luu, 2014) Using online archives such as https://earthexplorer.usgs.gov/, we accessed Sentinel-2 and Landsat-8 satellite imagery, which offer 5- and 16-day temporal resolutions, respectively By selectively selecting cloud-free images from these archives that align with low and high tide periods, we captured spectral reflectance of mangrove forests at different tidal stages Specifically, four images from each satellite were chosen, spaced 14 to 16 days apart, to ensure coverage across tide cycles Landsat-8 multi-temporal data were then used to analyze changes in mangrove forest extent and to create an updated thematic map of the mangrove ecosystems in Ninh Binh province.
The online archive at https://earthexplorer.usgs.gov/ offers a comprehensive, free, and long-term collection of Landsat-8 and Sentinel-2 satellite imagery, enabling the multi-temporal analysis of mangrove forest extent changes Since 2013, these high-resolution images have been instrumental in monitoring and delineating shifts in mangrove ecosystems in Ninh Binh province Accessing these datasets supports researchers and environmentalists in assessing habitat dynamics, identifying deforestation patterns, and informing conservation strategies for sustainable mangrove management.
Since 2015, Sentinel-2 satellite images have been selectively assessed and downloaded from the online archive, focusing on cloud-free adjacent scenes of the study site to ensure high-quality data for analysis.
Table 4.1 Details of remote satellite images selection
Image type Product ID Date of
LC08_L1TP_126046_20131008_20170429_01_T1 08/08/2013 LC08_L1TP_126046_20190603_20190618_01_T1 03/06/2019 LC08_L1TP_126046_20190619_20190703_01_T1 19/06/2019 LC08_L1TP_126046_20190705_20190719_01_T1 05/07/2019 LC08_L1TP_126046_20190721_20190801_01_T1 21/07/2019
Methodology
This study's methodology involves selecting satellite images and performing pre-processing on Landsat-8 data with Top of Atmosphere (TOA) correction to ensure accurate analysis Various classification methods are applied to distinguish mangrove areas effectively, followed by the composition of pixel layers for detailed mapping The accuracy of the classification results is thoroughly assessed to ensure reliability A mangrove dynamic map is constructed to visualize spatial-temporal changes, enabling precise quantification of mangrove extent variations over the defined period The entire methodology is systematically illustrated through a flowchart in Figure 5.1, providing a clear framework for understanding the research process.
In March 2019, two field visits were conducted to the mangrove forest area in Ninh Binh province to gather data for accuracy assessment, perform visual inspections, and collect local knowledge regarding the tidal regime During these visits, location data on land use and cover classes were recorded using the Garmin 78s GPS device, ensuring precise spatial information for effective mangrove ecosystem analysis.
A simple random sampling method was used to assess NDVI, supervised maximum likelihood, and OBC classification methods independently, with a minimum of 50 samples per class to ensure accurate and optimized classification results (Congalton, 2001) A total of 100 sample points—split between mangrove and non-mangrove cover types—were generated for the three classification techniques using the “random sampling” tool in ArcMap 10.4 software.
Figure 4.2 Sampling points for field data collection
Landsat data require preprocessing to minimize the effects of sensor, solar, atmospheric, and topographic variations across images for accurate analysis The recent Landsat 8 satellite enhances data quality by providing processed products using the Operational Land Imager (OLI) and Thermal Infrared Sensors (TIRS), supporting advanced remote sensing applications (Young, 2017).
This study utilizes Landsat 8 OLI/TIRS Level-1 Collection 1 products processed at the L1TP (Precision Terrain Level) for accurate terrain analysis To enable effective vegetation classification, the digital numbers (DN) of key Landsat 8 bands were converted to top of atmosphere (TOA) reflectance using scaling factors and solar elevation data derived from the product metadata, following Hankui K Zhang's methodology (2018) The conversion was performed with the Raster Calculator tool in ArcMap, applying the formal provided in the metadata MTL file to ensure precise radiometric correction.
TOA reflectance value, denoted as 𝜌𝜆 ′, is derived using the bands' calibration coefficients Specifically, 𝑀𝜌𝑄𝑐𝑎𝑙 represents the REFLECTANCE_MULT_BAND_x, and 𝐴𝜌 corresponds to the REFLECTANCE_ADD_BAND_x, where x indicates the specific band number These values can be located within the metadata file of a Landsat 8 image, enabling accurate surface reflectance calculations for remote sensing analyses.
The TOA reflectance value after solar correction, denoted as 𝜌𝜆, provides an accurate measurement of surface reflectance by adjusting for solar illumination conditions Prior to correction, the TOA reflectance is represented as 𝜌𝜆 ′, which needs to be adjusted based on the solar zenith angle, 𝜃𝑆𝐸, to account for the position of the sun in the sky Each spectral band used in the study is individually converted to TOA reflectance, ensuring consistent data quality These corrected reflectance values are then combined to create a comprehensive image, where each pixel's reflectance accurately represents its surface properties, facilitating precise analysis and interpretation.
Supervised maximum likelihood classification method:
Maximum likelihood classification was processed on ArcMap 10.4 software The combination of band 2 ,3, 4, 5, 6, 7 was obtained from Landsat-8 data and band 2, 3, 4, 8,
Sentinel-2 data provides the red, blue, and green bands crucial for creating RGB images To generate the RGB band combination, the selected bands from both satellites were composited using the “Composite” tool in ArcMap 10.4 software This process ensures a high-quality visualization by merging the spectral information into a single, cohesive image.
To maximize classification accuracy, it is essential to carefully select training sites for each class by comparing Google Earth images with reference materials such as local knowledge, previous maps, and related documents Ensuring these training sites have sufficient quantity, appropriate shape, variety, homogeneity, and proper distribution is crucial for reliable land classification results (Md Mijanur Rahman).
2013) The training sites were created using “Draw polygon” tool and managed using
The "Training Sample Manager" tool was used to reclassify training site results into mangrove and non-mangrove cover classes using the "Reclassify" tool The number of training sites, selected pixels, and total area for each class are summarized in Tables 4.2 and 4.3, providing essential data for accurate land cover classification.
Table 4.2 Training sites description of each class for Sentinel-2 images
Land Use/Land Cover class Mangrove Non-mangrove
Table 4.3 Training sites description of each class for Landsat-8 images
Land Use/Land Cover class Mangrove Non-mangrove
Finally, training sites assigned with given classes were automatically classified using
“Maximum likelihood classification” tool to the final map and ready for accuracy assessment The process of maximum likelihood classification is presented in the Figure 4.3
Figure 4.3 Flowchart of supervised maximum likelihood classification process
Normalized Difference Vegetation Index (NDVI) classification method:
Utilizing multi-spectral data from Landsat 8 OLI/TIRS and Sentinel-2 MSI enhances vegetation cover detection in the study area Both satellite images provide red (R) and near-infrared (NIR) bands—Sentinel-2 offers NIR band 8 and R band 4, while Landsat-8 provides NIR band 5 and R band 4—accessed via metadata from the USGS Earth Explorer platform Classification techniques focus on these bands to emphasize plant density, and the processing is conducted using ArcMap 10.4 software Vegetation indices are calculated using specific equations to improve the accuracy of vegetation detection.
NDVI values range from -1 to 1, with values between 0.6 and 0.8 indicating temperate and tropical rainforests (Gross, 2005) For healthy mangrove ecosystems, the recommended NDVI value range is from 0.3 to 1 (Umroha, 2016) In Ninh Binh, mangrove forests typically exhibit lower heights compared to the more developed mangrove forests found in southern Vietnam.
Based on high-resolution imagery from Google Earth Pro, numerous saplings and newly planted mangroves were identified through visual inspection Pixels were classified according to their vegetation index values using the “Reclassify” tool, enabling accurate mapping of the mangrove regeneration area.
Table 4.4 Recommended NDVI values for different LULC types
Mangrove - Dense and healthy mangrove area
- Saplings and newly planted mangrove plants
After the being reclassified into given classes, the results were ready for accuracy assessment The process of NDVI classification method is presented at the flowchart in Figure 4.4
4.2.3.2 Object-based classification (OBC) method
Advancements in spatial technology have led to increased resolution of satellite imagery, challenging traditional pixel-based classification methods due to their inability to effectively utilize texture, tone, and geometric features of objects As a result, statistical separability among different land cover classes diminishes, reducing classification accuracy (Qiong Hu, 2013; Shao, 2012; Yu Q., 2006) To address this, our study capitalized on the advantages of True Color Image (TCI) band combinations from Landsat-8 and Sentinel-2, which were created using the “Composite” tool in ArcMap 10.4 before processing in E-Cognition Developer 9.0 software.
Figure 4.4 Flowchart of NDVI classification method process
Object-based image analysis was conducted using E-Cognition Developer 9.0, following a systematic process that included multiresolution segmentation, selecting relevant features or classification rules, applying the nearest neighbor algorithm, and performing the classification This method enhances accuracy in remote sensing data analysis by effectively identifying and categorizing land cover types.
RESULTS AND DISCUSSIONS
Results
5.1.1 Image classification methods’ results and thematic maps
The current status of mangrove forests was mapped using Landsat-8 and Sentinel-2 satellite imagery combined with three classification methods: NDVI, supervised maximum likelihood classification, and OBC Employing multiple satellite datasets and classification techniques enhanced the accuracy and detection of diverse mangrove forest areas This integrated approach provides a comprehensive understanding of mangrove distribution and health.
Table 5.1 Mangrove forest area detected by different classification on Landsat-8 and
The OBC method applied to Landsat-8 satellite images identified the largest mangrove forest area of 392.58 hectares, compared to 322.53 hectares detected using Sentinel-2 In contrast, NDVI-based analysis showed slightly different results, with Landsat-8 identifying 331.11 hectares and Sentinel-2 detecting 334.92 hectares When employing NDVI for classification, Landsat-8 mapped 361 hectares, while Sentinel-2 classified 378 hectares of mangrove forest The thematic maps illustrating the current extent of mangrove forests are displayed in Figures 5.3 to 5.8.
Satellite Image Type Classification Method
Detected Mangrove Forest Area (Hectare)
Figure 5.1 NDVI classification method with Landsat-8 images
Figure 5.2 Supervised maximum likelihood classification method with
Figure 5.4 OBC with Landsat-8 images Figure 5.3 NDVI classification method with Sentinel-2 images
Figure 5.5 Supervised maximum likelihood classification method with Sentinel-2 images
Figure 5.6 OBC method with Sentinel-2 images
5.1.2 Image classification methods’ accuracy assessment
The accuracy assessment results, based on simple random sampling, are presented from Tables 5.2 to 5.7, evaluating various classification methods and satellite images This assessment includes key metrics such as producer’s accuracy, user’s accuracy, overall accuracy, and Cohen’s Kappa coefficient Producer’s and user’s accuracy evaluate the detection accuracy for each class, while Cohen’s Kappa coefficient measures the overall agreement and reliability of the classification results (Congalton, 2001).
Table 5.2 Error Matrix of NDVI classification with Landsat-8 images
NDVI classification method with Landsat-8 Accuracy (%)
Class Mangrove Non-mangrove Reference
Table 5.3 Error matrix of supervised classification with Landsat-8 images
Supervised maximum likelihood classification method with
Class Mangrove Non-mangrove Reference
Table 5.4 Error matrix of OBC with Landsat-8 images
OBC method with Landsat-8 Accuracy (%)
Class Mangrove Non-mangrove Reference
Table 5.5 Error matrix of NDVI classification with Sentinel-2 images
NDVI classification method with Sentinel-2 Accuracy (%)
Class Mangrove Non-mangrove Reference
Table 5.6 Error matrix of supervised maximum likelihood with Sentinel-2 images
Supervised maximum likelihood with Sentinel-2 Accuracy (%)
Class Mangrove Non-mangrove Reference
Table 5.7 Error matrix of OBC with Sentinel-2 images
OBC method with Sentinel-2 Accuracy (%)
Class Mangrove Non-mangrove Reference
The highest overall accuracy was achieved using Sentinel-2 images, with the NDVI classification method reaching 87% Landsat-8 images demonstrated overall accuracies of 86% with NDVI, 83% with supervised maximum likelihood, and 82% with the OBC method In comparison, Sentinel-2 images outperformed Landsat-8 across all classification methods, indicating the superior performance of Sentinel-2 for thematic mapping The NDVI classification method on Sentinel-2 satellite imagery provided the most accurate results, emphasizing its effectiveness for land cover analysis This study highlights the importance of selecting appropriate satellite data and classification techniques to optimize the accuracy of thematic maps.
The Kappa statistics for Landsat-8 classification were 0.71 with NDVI, 0.66 with supervised maximum likelihood, and 0.62 with the OBC method, indicating moderate to strong agreement In comparison, Sentinel-2 achieved higher accuracy, with Kappa values of 0.73 using NDVI, 0.70 with supervised maximum likelihood, and 0.60 with the OBC method, demonstrating generally superior classification performance Notably, applying NDVI to Sentinel-2 imagery resulted in the highest Kappa statistic of 0.73, highlighting its effectiveness for land cover classification.
5.1.4 Mangrove dynamic map and quantifying mangrove forest changes from 2013 to
Landsat-8 satellite images from 2013 and 2019 were used to construct a comprehensive mangrove dynamic map, highlighting changes over the years The NDVI classification method was selected for assessing these changes due to its superior accuracy, achieving an overall accuracy of 86%, making it the most reliable approach compared to other methods tested on Landsat data This approach provides valuable insights into mangrove ecosystem dynamics and supports effective environmental management.
8 Thematic mangrove forest maps of 2013 and 2019 classified 2 main LULC of mangrove and non-mangrove The changes of forest extent included the conversion of non-mangrove to mangrove cover, mangrove to non-mangrove cover and the stable area of mangrove forest and non-mangrove area
Figure 5.7 Mangrove dynamic map of Ninh Binh province from 2013 to 2019 using Landsat-
Over a six-year period, the dynamic map reveals four notable changes in both mangrove and non-mangrove areas In 2013, the mangrove forest in Ninh Binh Province covered 262.08 hectares as determined by NDVI classification, increasing to 361.35 hectares by 2019 Despite experiencing some loss, the overall mangrove forest in the region showed a significant net gain, reflecting ongoing changes in coastal vegetation.
The total mangrove area spans approximately 100 hectares, with significant expansion observed along the dike constructed along the Kim Son province coastline, as evidenced by recent field surveys and drone imagery While most of the mangrove at the study site results from ongoing plantation projects by organizations such as the Japanese Red Cross, Ninh Binh Red Cross, and the Ministry of Natural Resources and Environment, some small areas of less than 4 hectares have experienced losses at the edges of dense mangrove zones Since 2000, rapid development of mangrove plantations has been a key strategy for coastal protection, helping to strengthen dikes and prevent coastal soil erosion.
Discussion
5.2.1 Suitable satellite image classification methods
This study employs both pixel-based and object-based classification methods to analyze mangrove ecosystems The pixel-based approach includes NDVI and supervised maximum likelihood classification, while the object-based classification (OBC) offers an alternative method However, both techniques often encounter challenges such as mixed zonation within forest areas, misclassification of fringe zones, and difficulties in accurately identifying sparse mangrove regions in typical mangrove forests (Trang Nguyen, 2016).
NDVI and maximum likelihood supervised classification revealed a significant presence of scattered pixels from both classes, resulting in a “salt and pepper” effect, which is commonly regarded as a limitation of pixel-based land cover mapping methods (Yuehong Chen, 2018) Despite this, accurately classified “salt and pepper” pixels contributed to increased overall accuracy in mapping sparsely distributed mangroves Additionally, distinguishing the boundary between mangrove and non-mangrove classes proved challenging, as fringe mangrove pixels often appeared unclear and were frequently mixed with non-mangrove pixels, complicating precise boundary delineation.
Previous studies have shown that applying image segmentation and specific-rule classification methods on Object-Based Classification (OBC) often struggles to accurately link training sample areas with natural landscape features, as visual interpretation can be vague compared to human assessment (Heumann B W., 2011; Trang Nguyen, 2016; Hu, 2013) While OBC can improve the separation between mangrove and non-mangrove classes by reducing scatter misclassification, it may also lead to coarser land use maps, with reduced mapping accuracy due to "salt and pepper" pixel effects The standard field sampling protocol, which focuses on single pixel reflectance rather than entire heterogeneous objects, can be time-consuming and limits the effectiveness of OBC, highlighting the need for further field analysis and the development of object classification rules to improve spatial information Although multi-segmentation and class-specific rules enhance spatial pattern recognition beyond spectral data, integrating low to medium spatial resolution remote sensing data with OBC yields only moderate accuracy in mangrove classification, as evidenced by similar coastal mangrove studies.
38 environment in Vietnam (Ruiz-Luna, 1999) (Trang Nguyen, 2016) (Berlanga-Robles, 2002) (F Blasco, 1998) (Horning, 2008)
Pixel-based classification methods are widely recommended for mangrove forest mapping using low to medium spatial resolution satellite imagery such as Landsat-8 and Sentinel-2, achieving high accuracy (Giri C P., 2007; Hernández Cornejo, 2005) Studies like Seto and Fragkias (2007) successfully employed supervised classification and vegetation indices with Landsat TM scenes to map mangrove extents in Vietnam's Red River Delta, demonstrating its effectiveness in systematic monitoring Similarly, other research, including Aschbacher (1995) in Thailand and Thu and Populus (2007) in Tra Vinh Province, Vietnam, confirms the success of pixel-based approaches in similar environments Literature indicates that supervised classification, which integrates spectral information from multiple bands, is considered the most effective method for mapping extensive mangrove forests at medium resolutions (Thu, 2007; Aschbacher, 1995; Giri M J., 2008; Heumann B W., 2011; Rasolofoharinoro, 1998) This technique enhances overall mapping accuracy by leveraging the spectral characteristics of different land cover types.
This study demonstrated that the NDVI classification method using Sentinel-2 images achieved the highest accuracy in distinguishing mangrove cover from other land use/land cover types Supervised classification methods, often employed to differentiate mangroves from infrastructures and various vegetation types, benefit from NDVI’s high efficiency in separating mangrove areas from non-vegetation classes, as supported by studies from Tong (2004), Thu (2007), and Green (1998) Additionally, a strong correlation between NDVI data and mangrove canopy density across different satellite platforms has been established (Ramsey and Jensen, 1996) Monitoring gaps and closures in mangrove forests through NDVI provides valuable insights into the extent, health, and structure of mangroves, thereby enhancing mapping accuracy and supporting conservation efforts.
39 the conservation and monitoring of mangrove forest by modeling of the ecological process including evapotranspiration, photosynthesis, primary production and so on
Supervised classification techniques provide higher accuracy when mapping mangrove forests in mixed environments by effectively distinguishing multiple land use and land cover (LULC) types, including both vegetation and non-vegetation classes Conversely, the NDVI method is more effective for differentiating mangrove areas from non-vegetation and non-mangrove classes, making it particularly suitable for mapping the extent of mangrove forests in Ninh Binh province.
5.2.2 Mitigating tidal regime impact on remote sensing processing
Previous studies often overlooked the impact of tidal variations on classification results, focusing mainly on combining images from multiple tidal stages to address cloud cover issues (Kerrylee Rogers, 2017; Kirui, 2013) However, as interest in mangrove ecosystems and their spectral characteristics has grown, the influence of tidal regimes has gained recognition in remote sensing applications for mangrove mapping Utilizing scenes from multiple tidal stages can reduce pre-processing efforts and enhance classification accuracy compared to single-scene approaches.
Kerrylee Rogers (2017) demonstrated that analyzing water presence and absence during different tidal regimes—such as low tide, neap tide, and spring tide—within submerged mangrove forests enhances the discrimination of different mangrove zones By creating a combined multi-tidal image, this approach improves the accuracy of mangrove classification compared to traditional methods that focus solely on single tidal scenes Utilizing tidal variation data is essential for more precise mapping and monitoring of mangrove ecosystems in Australia.
Accurately modeling tidal regime activity requires long-term field data collection and correlation analysis to account for inherent variability in tidal cycles Research indicates that a 30-year record of in situ tidal data is sufficient to reliably estimate the differences between neap and spring tides within a tidal cycle.
Due to time and financial constraints faced by most researchers (Kovacs, 2001; Kirui, 2013; Kerrylee Rogers, 2017), this study focused on the tidal regime of the north coastal area of Vietnam and Ninh Binh province by selecting satellite images representing multiple tidal scenes for overlay analysis Future research will explore the use of various software tools recommended by Kerrylee to enhance tidal studies.
(2017) can be utilized to model tidal activity and put to use along with selection of remote
Satellite sensing data over a 40-day period provides a suitable calibration window for accurate analysis Acquiring compatible remote sensing data that aligns with tidal cycles and cloud-free conditions remains challenging due to weather variability To improve classification accuracy and deepen understanding of the mangrove ecosystem, long-term recording of tidal regimes at different stages is essential.
CONCLUSIONS
General conclusion
This study utilized freely available satellite images from Landsat-8 and Sentinel-2 to accurately map current mangrove forest extents and analyze their dynamic changes in Ninh Binh Province between 2013 and 2018 Various pixel-based classification techniques, including NDVI, maximum likelihood supervised, and OBC, were tested to enhance classification accuracy Selecting satellite image dates aligned with tidal cycles, based on previous research, and employing image overlaying techniques helped delineate submerged mangrove areas and improve discrimination results Computer processing and fieldwork confirmed the effectiveness of these methods, providing comprehensive insights into the spatial and temporal dynamics of mangrove ecosystems.
The study found that the overall accuracy for Landsat-8 images was 86%, 83%, and 82% using NDVI, supervised maximum likelihood, and OBC methods, respectively, while Sentinel-2 achieved higher accuracies of 87%, 85%, and 80% with the same classification approaches Notably, the highest overall accuracy of 87% was obtained with Sentinel-2 data using the NDVI classification method Additionally, the Kappa statistics for Landsat-8 were 0.71, 0.66, and 0.62 for NDVI, supervised maximum likelihood, and OBC methods, respectively, whereas Sentinel-2 demonstrated slightly higher Kappa values of 0.73, 0.70, and 0.60 with similar classification techniques Overall, Sentinel-2 images provided more accurate and reliable thematic maps across the evaluated classification methods.
This study confirms that pixel-based applications, particularly NDVI, generally achieve higher accuracy in detecting mangrove forests, consistent with previous research The most effective approach involved using NDVI derived from Sentinel-2 MSI imagery, which successfully identified 378.56 hectares of mangrove forest The classification achieved an overall accuracy of 87%, with a producer’s accuracy of 91.38% and a user’s accuracy of 86.89%, corresponding to a Cohen’s Kappa value of 0.73, demonstrating reliable detection performance in Ninh Binh province.
Using Landsat-8 imagery and NDVI classification, a dynamic map of mangrove extent was developed to assess changes from 2013 to 2019 Over this six-year period, the mangrove forest in Ninh Binh province expanded by nearly 100 hectares, increasing from 262.08 hectares in 2013 This study demonstrates significant growth in local mangrove ecosystems, highlighting the effectiveness of remote sensing and NDVI analysis for monitoring coastal habitat changes.
Recommendation for further study and limitation
This study offers valuable insights into the scientific foundation of remote sensing applications for mangrove forests in Ninh Binh province However, it also encountered several limitations and challenges during the research process that may affect the accuracy and effectiveness of the findings Despite these obstacles, the research contributes important knowledge to the field of environmental monitoring and conservation of mangrove ecosystems.
Due to the absence of tidal regime recording stations along the Kim Son district coastline, this study primarily relied on theoretical references and secondary data from previous research Incorporating in situ coastal sea water level measurements simultaneously with remote sensing data and fieldwork would enhance tidal activity modeling and improve image overlay accuracy The use of multi-tidal stage composition scenes showed only a minor improvement, achieving just a 1-3% increase in overall accuracy compared to single-scene approaches Additionally, the Object-Based Classification (OBC) method provided moderate classification accuracy, as it is better suited for higher spatial resolution remote sensing data and is more compatible with fieldwork guidelines than pixel-based classification techniques.
Future research should focus on integrating tidal studies with comprehensive field data collection to better understand current and historical tidal regimes Enhancing the application of object-based analysis techniques can be achieved through the development of more detailed and standardized fieldwork guidelines, leading to more accurate and reliable results.
1 Adamsen, F P (1999) Measuring wheat senescence with a digital camera Crop sci,
2 Aschbacher, J O (1995) An integrated comparative approach to mangrove vegetation mapping using advanced remote sensing and GIS technologies: Preliminary results
3 Aschbacher, J O (1995) An integrated comparative approach to mangrove vegetation mapping using advanced remote sensing and GIS technologies: Preliminary results
4 Bahuguna, S a (2001) Application of remote sensing data to monitor mangroves and other coastal vegetation of India Indian Journals of Marine Science, 30, 195-213
5 Béland, M G (2006) Assessment of land-cover changes related to shrimp aquaculture using remote sensing data: A case in the Giao Thury District, Vietnam Int J Remote Sens,
6 Berlanga-Robles, C R.-L (2002) Land use mapping and change detection in the coastal zone of northwest Mexico using remote sensing techniques J Coast Res, 18, 514-522
7 Blaschke, T S (2001) What’s wrong with pixels? Some recent developments interfacing remote sensing and GIS GeoBIT/GIS, 6, 12–17
8 Brown, M E., Pinzon, J D., & C.J (2006) Evaluation of the consistency of long-term NDVI time series derived from AVHRR, SPOT-Vegetation, SeaWiFS, MODIS, and Landsat ETM+ sensors Geoscience and Remote Sensing, IEEE Transactions on 44(7),
9 Camile Sothe, C M (2017) Evaluating Sentinel-2 and Landsat-8 Data to Map Sucessional Forest Stages in a Subtropical Forest in Southern Brazil Remote Sens., 9, 838
10 Campbell, J B (1987) Introduction to Remote Sensing New York, US: The Guildford Press
11 Claudia Kuenzer, A B (2011) Remote Sensing of Mangrove Ecosystems: A Review
12 Congalton, R G (2001) Accuracy assessment and validation of remotely sensed and other spatial information International Journal of Wildland Fire, 10
13 Duke, N (1992) Mangrove floristics and biogeography In A a Robertson, Tropical Mangrove Ecosystems, Coastal Estuarine Studies (pp 63–100) Washington, D.C:
14 F Blasco, T G (1998) Recent advances in mangrove studies using remote sensing data Marine and Freshwater Research, 49: 287–96
15 FAO (2007) The World’s Mangroves 1980-2005 Rome, Italy: Food and Agriculture Organization of the United Nations
16 Field, C., Osborn, J., Hoffmann, L., Polsenberg, J., Ackerly, D., Berry, J., Matson, P a
(1998) Mangrove biodiversity and ecosystem function Global Ecology and Biogeography
17 G.J Hay, G C (2008) Geographic Object-Based Image Analysis (GEOBIA): A new name for a new discipline In S L T Blaschke, Object Based Image Analysis (pp 93-112) Heidelberg, Berlin, New York: Springer
18 Gao, J (1998) Hybrid method toward accurate mapping of mangroves in a marginal habitat from SPOT Multispectral data Int J Remote Sens, 19, 1887-1899
19 Gilman, L., Ellison, J., & Duke, N a (2008) Threats to mangroves from climate change and adaptation options: a review Aquatic Botany, 89(2), 237–250
20 Giri, C K ( 1996) Assessing Land Use/Land Cover Dynamics in Two Identified ―Hot Spot‖ Areas: Oudomxay Province of Lao P.D.R and Mekong Delta of Vietnam Proceeding of The 17th Asian Conference on Remote Sensing, (pp 4–8 ) Colombo, Sri Lanka
21 Giri, C P (2007) Monitoring Mangrove forest dynamics of the Sundarbans in Bangladesh and India using multi-temporal satellite data from 1973 to 2000 Estuar Coast
22 Giri, M J (2008) Mangrove forest distribution and dynamics in Madagascar (1975–
23 Goodchild, M F (1994) Integrating GIS and remote sensing for vegetation analysis and modeling: methodological issues Journal of Vegetation Science, 615-626
24 Green, E C (1998) Remote sensing techniques for mangrove mapping Int J Remote
25 Gross, D (2005) Monitoring Agricultural Biomass Using NDVI Time Series Food and
Agriculture Organization of the United Nations (FAO), Rome, Italy
26 H.Strahler, A (1980) The use of prior probabilities in maximum likelihood classification of remotely sensed data Remote Sensing of Environment, 10: 135-163
27 Hankui K Zhanga, D.C (2018) Characterization of Sentinel-2A and Landsat-8 top of atmosphere, surface,and nadir BRDF adjusted reflectance and NDVI differences Remote Sensing of Environment 215, 482 - 494
28 Hernández Cornejo, R K.G (2005) Remote sensing and ethnobotanical assessment of the mangrove forest changes in the Navachiste-San Ignacio-Macapule Lagoon Complex, Sinaloa, Mexico Ecol Soc , 10, art 16
29 Heumann, B W (2011) An Object-Based Classification of Mangroves Using a Hybrid Decision Tree—Support Vector Machine Approach Remote Sensing, 3, 2440-2460
30 Heumann, B W (2011) Satellite remote sensing of mangrove forests: recent advances and future opportunities Progress in Physical Geography, 35, 87–108
31 Horning, N (2008) Remote Sensing In S Jorgensen, & B Fath, Encyclopedia of Ecology (p 2986) London: Elsevier Science
32 Hu, Q (2013) Exploring the Use of Google Earth Imagery and Object-Based Methods in Land Use/Cover Mapping Remote Sensing, 5, 6026-6042
33 International Federation of Red Cross, R C (2009) Mangrove plantation in Viet Nam: measuring impact and cost benefit Hanoi: ifrc
34 Irons, J D (2012) The next Landsat satellite: the Landsat data continuity mission
35 J.E Sterling, M H (2006) Vietnam: a natural history Yale Univ Press, 1-21 and 91-
36 Kerrylee Rogers, L L (2017) Mapping of mangrove extent and zonation using high and low tide composites of Landsat data Hydrobiologia
37 Kirui, K B (2013) Mapping of mangrove forest land cover change along the Kenya coastline using Landsat imagery Ocean & Coastal Management, 83, 19–24
38 Knudby, A L (2014) Using multiple Landsat scenes in an ensemble classifier reduces classificlassification error in a stable nearshore environment International Journal of Applied Earth Observation and Geoinformation, 28, 90–101
39 Kovacs, J M.-C (2001) Mapping disturbances in a mangrove forest using multi-date Landsat TM imagery Environmental Management, 27, 763–776
40 Kuenzer, C A (2011) Remote sensing of mangrove ecosystems: a review Remote Sensing 3, 878
41 Kuenzer, C A (2011) Remote sensing of mangrove ecosystems: a review Remote Sensing 3, 878
42 L Wang, W P (2004) imagery, Integration of object-based and pixel-based classification for mapping mangroves with IKONOS International Journal of Remote Sensing, 25:24, 5655-5668
43 L X Tuan, Y M (2003) Environmental Management in Mangrove Environmental Informatics Archives, 1, 38-52
44 Long, J N (2014) A mapping and monitoring assessment of the Philippines’ mangrove forests from 1990 to 2010 Journal of Coastal Research, 30(2), 260–271
45 Lu, D., Weng, Q (2007) Survey of image classification methods and techniques for improving International Journal of Remote Sensing, 28, No 5, 823–870
46 M Drusch, U D (2012) Sentinel-2: ESA's optical high-resolution mission for GMES operational services Remote Sens Environ., 120, 25 - 36
47 M I El-Gammal, R R (2014) NDVI Threshold Classification for Detecting Vegetation Cover in Damietta Governorate, Egypt American Science 10(8), 108-113
48 Machado, S E (2002) Spatial temporal variability of sorghum grain yield: Influence of soil, water, pests, and diseases relationship Precise Agriculture, 3: 389-406
49 Manson, F J (2001) Assessing techniques for estimating the extent of mangroves: topographic maps, aerial photographs and Landsat TM images Marine and Freshwater Research 52, 787–792
50 Mavis, B (2001) A socio-economic analysis of mangrove degradation in Samoa
51 Md Mijanur Rahman (2013) Comparison of Landsat image classification methods for detecting mangrove forests in Sundarban International Journal of Remote Sensing, 34:4,
52 Md Mijanur Rahmana, M R (2013) Comparison of Landsat image classification methods for detecting mangrove forests in Sundarban International Journal of Remote Sensing, 34, 4, 1041–1056
53 Minh Luu, D O (2014) Hydrological regime of a tidal system in the Red River Delta, northern Vietnam Hydrology in a Changing World: Environmental and Human Dimensions, 363
54 Neukermans, G D.-G (2008) Mangrove species and stand mapping in GAzi bay (Kenya) using Quickbird satellite imagery J Spat Sc, 53, 75-86
55 Nguyen Khanh Van, N T (2000) Bioclimatic charts of Vietnam Vietnam National University Press
56 Paul A Longley, M F (2005) Geographic information systems and science John Wiley
57 Phan Minh, J P (2007) Status and changes of mangrove forest in Mekong Delta
Estuarine, Coastal and Shelf Science 71, 98-109
58 Phan Nguyen Hong, T V (1999) Mangrove Forests of Vietnam Hanoi: Agriculture
59 Phan Nguyen Hong, V T (2004) Some Natural feature of the mud flats in the coastal estuarine mangrove area of Thai Binh and Nam Dinh Provinces in Mangrove Ecosystem in the Red River Coastal Zone Agriculture Publishing House, 1 - 15
60 Primavera, J 2 (2000) Development and conservation of Philippine mangroves: institutional issues Ecological Economics, 35, 91–106
61 Qiong Hu, W W (2013) Exploring the Use of Google Earth Imagery and Object-Based Methods in Land Use/Cover Mapping Remote Sensing, 5, 6026-6042
62 R Kettig, D L (1976) Classification of multispectral image data by extraction and classification of homogeneous objects IEEE Transactions on Geoscience Electronics, 19-
63 Ramsey, E I., & Jensen, J (1996) Remote sensing of mangrove wetlands: Relating canopy spectral to site-specific data Photogramm Eng Remote Sensing, 62, 939-948
64 Rasolofoharinoro, M B (1998) A remote sensing based methodology for mangrove studies in Madagascar Int J Remote Sens, 19, 1873-1886
65 Rasolofoharinoro, M B (1998) A remote sensing based methodology for mangrove studies in Madagascar Int J Remote Sens, 19, 1873-1886
66 Robert C Weih, N D (2010) Object-based classification vs Pixel-based classification: Comparitive importance of multi-resolution imagery
67 Ruiz-Luna, A B.-R (1999) Modifications in coverage patterns and land use around the Huizache-Caimanero lagoon system, Sinaloa, Mexico: A multi-temporal analysis using Landsat images Estuar Coast Shelf Sci, 49, 37-44
68 Selvam, V R (2003) Assessment of community-based restoration of Pichavaram mangrove wetland using remote sensing data Curr Sci, 85, 794-798
69 Seto K.C., F M (2007) Mangrove conversion and aquaculture development in Vietnam: A remote sensing-based approach for evaluating the Ramsar Convention on Wetlands Glob Environ Change, 17, 486-500
70 Shaban, M A (2001) Improvement of classification in urban areas by the use of textural features: the case study of Lucknow City, Uttar Pradesh International Journal of Remote Sensing, 4, 565–593
71 Shao, Y L (2012) Comparison of support vector machine, neural network, and CART algorithms for the land-cover classification using limited training data points ISPRS J.Photogramm Remote Sens, 70, 78–87
72 Spalding, M (1997) The global distribution and status of mangrove ecosystems Int NewsLett Coast., 1, 20-21
73 Story M, C R (1986) Accuracy assessment: A user's perspective Photogrammetric Engineering and Remote Sensing, 52, 397-299
74 T.Blaschke (2010) Object based image analysis for remote sensing ISPRS Journal of
75 Thomas N, L R (2017) Distribution and drivers of global mangrove forest change, 1996–2010 PLoS ONE, 12, 6
76 Thu, P P (2007) Status and changes of mangrove forest in Mekong Delta: Case study in Tra Vinh, Vietnam Estuar Coast Shelf Sci, 71, 98-109
77 Thuy Dang Truong, L H (2018) Mangrove forests and aquaculture in the Mekong river delta Land Use Policy, 20-28
78 Tomlinson, P B (1986) The botany of mangroves Cambridge, England: Cambridge University Press
79 Tong, P A (2004) Assessment from space of mangroves evolution in the Mekong Delta; in relation to extensive shrimp farming Int J Remote Sens , 25, 4795-4812
80 Trang Nguyen, L Q (2016) Object-Based vs Pixel-Based Classification of Mangrove Forest Mapping in Vien An Dong Commune, Ngoc Hien District, Ca Mau Province Using VNREDSat-1 Images Advances in Remote Sensing, 5, 284-295
81 Umroha, W A (2016) Detection of Mangrove Distribution in Pongok Island Procedia
82 Vaiphasa, C O (2005) Tropical mangrove species discrimination using hyperspectral data: A laboratory study Estuar Coast Shelf Sci, 65, 371-379
83 Vaiphasa, C S (2006) post-classifier for mangrove mapping using ecological data
84 Wang, L., Sousa, W., Gong, P., & Biging, G (2004) Comparison of IKONOS and QuickBird imagery Remove Sensing Environment, 91: 432-440
85 Wolanski, E M (2015) Coastal Wetlands: An Integrated Ecosystem Approach The Netherlands
86 Xavier Ceamanos, S V (2017) Processing Hyperspectral Images In M Z Nicolas Baghdadi, Optical Remote Sensing of Land Surface (p 121) ISTE Press - Elsevier
87 Xuehong Zhanga, P M (2017) Mapping mangrove forests using multi-tidal remotely- sensed data and a decision-tree-based procedure Int J Appl Earth Obs Geoinformation 67,
88 Young, N E (2017) A survival guide to Landsat preprocessing Ecology 98(4), 920 -
89 Yu Q., G P (2006) Object-based detailed vegetation classification with airborne high spatial resolution remote sensing imagery Photogramm Eng Remote Sens, 72, 799–811
90 Yuehong Chen, Y Z (2018) Enhancing Land Cover Mapping through Integration of Pixel-Based and Object-Based Classifications from Remotely Sensed Imagery Remote Sensing, 10, 77
ID Coordinates Land Use/ Land Cover
Mangrove forest in Ninh Binh province during relatively spring tide (04/07/2019)
Using GPS device to assess classification accuracy during field work (05/07/2019)