Mục tiêu nghiên cứu - Nghiên cứu được đặc điểm quang phổ của rừng ngập mặn ở các tỉnh ven biển phía Bắc trên các tư liệu viễn thám; - Lựa chọn được tư liệu viễn thám và phương pháp phân loại phù hợp phù hợp để phân loại rừng ngập mặn các tỉnh ven biển khu vực phía Bắc; - Xác định được ngưỡng phân loại rừng ngập mặn một số tỉnh ven biển phía Bắc; - Đánh giá được biến động về diện tích rừng ngập mặn giai đoạn 2016-2020 ở một số tỉnh ven biển phía Bắc - Xây dựng được hướng dẫn kỹ thuật phân loại rừng ngập mặn cho khu vực ven biển. 6. Đối tượng và phạm vi nghiên cứu - Đối tượng nghiên cứu: Đối tượng nghiên cứu của luận án là toàn bộ rừng ngập mặn ven biển ở khu vực phía Bắc thuộc các tỉnh Quảng Ninh, Thái Bình và TP Hải Phòng, sử dụng các tư liệu viễn thám miễn phí có độ phân giải trung bình gồm tư liệu ảnh Radar Sentinel-1, tư liệu ảnh quang học Sentinel-2 và ảnh Landsat-8. - Phạm vi nghiên cứu: + Phạm vi cấp 1: Nghiên cứu điểm để xây dựng ngưỡng phân loại rừng ngập mặn tại các vùng đại diện gồm các huyện Tiên Yên và TX Quảng Yên, hai xã Bàng La (quận Đồ Sơn), Đại Hợp (huyện Kiến Thuỵ) thuộc Thành phố Hải Phòng. + Phạm vi cấp 2: Phân loại và đánh giá biến động tại 3 tỉnh thành gồm: tỉnh Quảng Ninh, TP Hải Phòng và tỉnh Thái Bình. + Thời gian thu thập số liệu trên các ô tiêu chuẩn được tiến hành vào năm 2019, thời gian thu thập điểm mẫu cuối năm 2020 và nghiên cứu toàn bộ các cảnh ảnh của năm 2016 và 2020 7. Phương pháp nghiên cứu: Luận án được áp dụng các phương pháp chủ yếu để triển khai các nội dung nghiên cứu như sau: (1) Phương pháp thu thập tài liệu thứ cấp và điều tra ngoại nghiệp tại các ô tiêu chuẩn; (2) Phương pháp nghiên cứu đặc điểm phân bố và sinh thái của rừng ngập mặn các tỉnh phía Bắc; (3) Phương pháp lựa chọn tư liệu ảnh vệ tinh thích hợp để phân loại rừng ngập mặn khu vực phía Bắc; (4) Phương pháp phân loại rừng ngập mặn các tỉnh ven biển phía Bắc (5) Phương pháp phân loại rừng ngập mặn khu vực nghiên cứu từ các tư liệu viễn thám; (6) Phương pháp đánh giá biến động rừng ngập mặn một số tỉnh ven biển phía bắc giai đoạn 2016 – 2020. 8. Kết luận của kết quả nghiên cứu: - Rừng ngập mặn ở khu vực phía Bắc có thành phần loài tương đối phong phú, trong đó khu vực Đông Bắc có nhiều đặc điểm khác biệt về địa hình, nước biển độ mặn cao nên chủ yếu xuất hiện các loài như Đước vòi, Vẹt dù, Bần chua, Trang, Mắm biển, Sú...., các loài có chiều cao trung bình thấp, các loài cây có chiều cao giao động từ 1,5-5,9 m, cây rừng có tán nhỏ, giao động trong khoản từ 1,2- 6,0 m. Đối với vùng ven biển đồng bằng sông Hồng, có nhiều cửa sông nên có lượng phù sa bồi tụ nhiều hàng năm, nước biển độ mặn thấp (từ 5-15 ppt) nên thành phần loài chủ yếu các loài là Bần chua, Trang, Sú, Giá… chiều cao của cây rừng giao động từ 5,6 -15,5 m. - Đặc điểm quang phổ của rừng ngập mặn trên các tư liệu viễn thám cho thấy: Các dấu hiệu quang phổ của rừng ngập mặn và các lớp thực vật khác gồm thảm thực vật trên cạn và thực vật dưới nước như bèo, cỏ ngập nước...tại các vùng bước sóng nhìn thấy gần như tương đồng nhau. Tuy nhiên, ở các bước sóng cận hồng ngoại (NIR) và sóng ngắn (SWIR) giá trị phản xạ của rừng ngập mặn thấp hơn đối tượng khác nhưng lại cao hơn khu vực có sự hiện diện của nước. - Đối với hoạt động giám sát rừng ngập mặn ven biển phía Bắc, kết hợp hai tư liệu viễn thám quang học Sentinel-2 và Radar Sentinel-1 với số lượng lớn các cảnh ảnh giúp hạn chế tối đa ảnh hưởng của thuỷ triều đến kết quả phân loại rừng ngập mặn và ứng dụng được công nghệ tiên tiến cho phép giám sát được rừng ngập mặn thời ở điểm hiện tại hoặc trong quá khứ với kết quả nhanh chóng và đảm bảo độ chính xác cho phép, giúp tối đa giảm chi phí điều tra thực địa rừng ngập mặn. - Có ba bộ ngưỡng phân loại rừng ngập mặn đối với các tỉnh ven biển khu vực phía Bắc trên hai tư liệu quang học Sentinel-2 và tư liệu Radar Sentinel-1. Với khu vực TP Hải Phòng và tỉnh Thái Bình, ngưỡng phân loại rừng ngập mặn từ chỉ số CMRI trên tư liệu ảnh Senitnel-2 lớn hơn 0.25, với tư liệu Radar Sentinel-1, ngưỡng phân loại từng ngập mặn từ giá trị tán xạ ngược VH hai khu vực này trong khoảng từ (-15.8) – (-10). Khu vực Quảng Ninh có hai ngưỡng, trong đó các huyện Tiên Yên, TP Cẩm Phả, Vân Đồn, Đầm Hà, Hải Hà, Cô Tô và TP Móng Cái có ngưỡng phân loại rừng ngập mặn với giá trị Chỉ số CMRI lớn hơn 0.1, ngưỡng phân loại rừng ngập mặn từ giá trị tán xạ ngược VH ở các địa phương này từ (-17.3) – (-10.0). Khu vực TX Quảng Yên, Uông Bí, Hoành Bồ, TP Hạ Long có ngưỡng phân loại rừng ngập mặn từ chỉ số CMRI lớn hơn 0.41 và ngưỡng phân loại rừng ngập mặn từ giá trị tán xạ ngược VH từ (-16,8 – (-10). - Sử dụng kết hợp tư liệu viễn thám quang học Sentinel-2 và ảnh Radar Sentinel-1 trên nềng tảng công nghệ Google Earth Engine có thể giám sát biến động diện tích rừng ngập mặn các tỉnh ven biển phía Bắc một cách nhanh chóng, có độ chính cho phép. Cụ thể, giai đoạn từ năm 2016 – 2020, diện tích rừng ngập mặn ở các tỉnh khu vực nghiên cứu có sự biến động khá lớn, trong đó tỉnh Quảng Ninh phát hiện thấy diện tích rừng ngập ngập mặn giảm lớn nhất với khoảng 479,4 ha. Thành phố Hải Phòng có diện tích tăng lên 233,3 ha và tỉnh Thái Bình có diện tích rừng ngập mặn tăng lớn nhất với 428 ha. - Luận án đã xây dựng hướng dẫn kỹ thuật các bước đánh giá biến động rừng ngập mặn ở các tỉnh phía Bắc gồm 11 bước, trong đó các bước thực hiện trên GEE cho phép tự động thu ảnh, tiền xử lý ảnh, tính toán các giá trị chỉ số, phân loại rừng tính diện tích rừng ngập mặn từ các tư liệu quang học và Radar, đồng thời kết hợp hai tự liệu này để đánh giá biến động có độ tin cậy cao.
MINISTRY OF EDUCATION AND TRAINING MINISTRY OF AGRICULTURE AND RURAL DEVELOPMENT VIETNAM NATIONAL UNIVERSITY OF FORESTRY NGUYEN TRONG CUONG USING REMOTELY SENSED OPTICAL AND RADAR DATA FOR MONITORING OF MANGROVE FORESTS IN NORTHERN COAST OF VIETNAM MAJOR: FOREST MANAGEMENT CODE: 62 02 11 SUMMARY OF THE DOCTORAL DISSERTATION HANOI, 2022 The thesis was completed at the Vietnam National University of Forestry Supervisor: Assoc Prof Dr Tran Quang Bao, Vietnam Administration of Vietnam Assoc Prof Dr Nguyen Hai Hoa - Vietnam National University of Forestry Reviewer 1: ……………………………… Reviewer 2: ……………………………… Reviewer 3: ……………………………… The dissertation will be defended at the Doctoral Evaluation Dissertation Council: - Date and time: - Place: Vietnam National University of Forestry The dissertation can be found at: - National Library of Vietnam - Library of Vietnam National University of Forestry INTRODUCTION Justification for the research Mangroves are tidal wetlands with a diverse assemblage of trees and shrubs and are located in tropical and subtropical regions between latitudes about 30 degrees N and 30 degrees South, Lee and Yeh (2009) ) [83] Mangroves are valuable resources in many ways, such as slowing the flow and spreading the tidal water, drastically reducing the height of waves during high tide, protecting sea dykes, limiting saltwater intrusion and protecting groundwater Besides Mangrove forests are also a place to protect biodiversity, protect animals during high tides and high waves… Besides, they provide a wide range of ecosystem services as well as habitats for many marine aquatic species , water purification, coastal stabilization, biodiversity, Abdul Aziz et al (2015) [19], Giri et al (2015) [53], Rahman et al (2013) [96] Mangroves provide a number of ecosystem services and coastal protection for tropical and subtropical coastlines around the world, Veettil et al (2018) [118] But tropical and subtropical coastal mangroves are among the most threatened and vulnerable ecosystems worldwide, Valiela et al (2001) [116] In our country, according to Phan Nguyen Hong and Hoang Thi San (1993) [137] Maurand's statistics in 1943 there were about 400,000 of mangroves and mainly in the South (250,000 ha) of which Rung Sat (40,000 ha) ha), Ca Mau (150,000 ha), Central and Northern regions (40,000 ha) and other places (20,000 ha) However, this number decreased to 286,400 in 1975, Phan Nguyen Hong and Hoang Thi San (1993) [137] and to only 156,608 (reduced to 39.1% compared to 1943) in 2000, Pham Thu Thuy et al (2019) [16] Faced with this situation, the Government has taken many solutions to protect and restore mangrove forests The protection and restoration of mangroves not only serves the goal of sustainable development, conservation and development of genetic resources, but especially response to climate change, which is increasingly affecting our country Therefore, the requirement for monitoring, restoration and protection of mangroves is very urgent Remote sensing has proven to be essential in monitoring and mapping highly threatened mangrove ecosystems Many studies on this topic have been carried out in the world Monitoring mangroves in situ is a challenging task, because these ecosystems are difficult to access, surveying can be expensive and time consuming, but in situ monitoring is still considered a source of information important news, Moritz Zimmermann et al (2001) [87] Today, more than 300 earth observation satellites from more than 15 countries are active, Cárdenas et al (2017) [33] Regarding remote sensing data sources, it can be said that the development of research works on mangrove monitoring is associated with the development history of remote sensing images Traditional coastal wetland methods for monitoring and mapping are often time-consuming, labor-intensive and costly, and they often fail to detect changes over large coastal regions Ghosh et al (2016) [50] In recent years, with the advancement of remote sensing technology, it has greatly supported the investigation and monitoring of forest resources in general and mangroves in particular In which, studies using optical remote sensing data in survey, monitoring and assessment of forest status are widely used However, because the characteristics of optical and near-infrared imaging systems depend on solar energy, optical images are affected by weather, often cloudy and foggy on images information quality in images This is one of the basic limitations of optical imaging, especially for countries located in the tropical monsoon region like Vietnam Another fundamental limitation of optical satellite images is that they only provide information about the reflection and absorption characteristics of objects on the surface in the visible wave region, so the image often lacks structural information structure and roughness of the study surface Unlike optical images, because they are taken in the region of micro waves with longer wavelengths (about cm), satellite images of Radar allow to provide information about roughness, material texture and structural characteristics of objects on the ground surface Moreover, radar waves have the ability to penetrate clouds, so radar photography is independent of the weather, can be taken both during the day and at night, so it is proactive and has a very high probability of success in capturing and capturing images image However, Radar images also have very basic disadvantages such as many geometric distortions including front shortening, overlapping, shadows, high noise levels, and the image of objects on the Radar image is often incomplete It is completely similar to common human perception Therefore, it is much more difficult to process geometry, noise and information on Radar images than in traditional optical images, which limits the application of Radar images, especially in Vietnam In order to take advantage of the advantages of the two types of data, and at the same time limit the weaknesses of the two open remote sensing documents to solve the problem of monitoring mangrove forests in the northern region, which has many characteristics, PhD student carried out the project "Using remotely sensed optical and Radar data for monitoring of mangrove forests in Northern coast of Vietnam" Goal and objectives - Studying the spectral characteristics of mangroves in the northern coastal provinces based on remote sensing data; - Efforts to select remote sensing data and suitable classification methods to classify mangrove forests in the northern coastal provinces; - Determine the threshold for classification of mangroves in some northern coastal provinces - Assess changes in mangrove area in the period 2016-2020 in some northern coastal provinces - Develop technical guidelines for classification of mangroves for coastal areas Subject: Research object of the thesis is the entire coastal mangrove forest in the northern region of Quang Ninh, Thai Binh and Hai Phong provinces, using medium-resolution free remote sensing data including Sentinel-1 Radar image data, Sentinel-2 optical image document and Landsat-8 image Scope: Level 1: Study site to build threshold for mangrove classification in representative areas including Tien Yen district and Quang Yen town, two communes of Bang La (Do Son district), Dai Hop (Kien Thuy district) of the province Hai Phong city Level 2: Classification and assessment of changes in provinces including: Quang Ninh province, Hai Phong city and Thai Binh province Time of data collection was conducted in 2019, time to collect sample points at the end of 2020 and study all the scenes of 2016 and 2020 Scientific and practical significance 5.1 Scientific significance: The results of the thesis have allowed to automatic updating and monitoring the dynamics of mangrove forests on the basis of new technology developed The application of the step-by-steps guide to automatic mangrove change assessment allows managers and scientists to quickly check the areas with changes in mangrove forests at present or in the past at different time intervals quickly, ensuring accuracy by combining both optical and radar materials 5.2 Practical significance: + This is a very detailed study on the spectral characteristics of mangrove forests using the medium remote sensing data (Landsat-8, Sentinel-1, and Sentinel-2) to monitor mangroves This study has also selected suitable remotely sensed data for classification of mangrove forests; developed classification methods of mangrove forests based on index thresholds; and combined optical and RADAR data for mangrove monitoring + As findings showed that it has confirmed the ability to use remotely sensed data (Sentinel-1 and Sentinel-2 data); and combine these two kinds of data (optical and RADAR) in determining mangrove fluctuations for some northern coastal provinces in particular and the whole country in general The thesis has also confirmed that the use of the combined mangrove index index (CMRI) and backscatter value VH to classify and monitor mangroves is reliable, applicable and accurate in Vietnam The thesis also provides a theoretical basis for the application of mangrove index thresholds on the basis of Google Earth Engine (GEE) platform to classify and monitor mangrove forests, which can be used as a reference for future studies Contributions (1) Combining the optical remote sensing Sentinel-2 and Radar Sentinel-1 with a large number of images has minimized the influence of tides on the results of mangrove classification In addtion, applying the advanced technology allows to monitor mangrove forests in the present or in the past with quick results and ensures accuracy, helping to reduce the cost of mangrove field surveys (2) Thresholds for classifying mangroves were defined for the northern provinces from the combined mangrove index CMRI for Sentinel-2 data and VH backscatter values for Sentinel-1 data for some coastal provinces, which applied to classify and monitor the changes in mangrove extent in Quang Ninh, Hai Phong and Thai Binh provinces during the period of 2016-2020 quickly, and ensured the accuracy and fully automatic classification in GEE platform (3) Developing a technical guideline of step-by-steps to combine Sentinel-1 and Sentinel-2 data on the GEE platform, which allows automatic classification and monitoring of mangrove changes for the northern provinces Thesis outline The thesis consists of 323 pages, of which the Contents section is 143 pages, the Appendix is 180 pages, The layout includes the introduction (4 pages) and Chapters, Chapter 1: Overview of the research problem (31 pages), Chapter 2: Characteristics of the research area (07 pages), Chapter 3: Objectives, objects, scope and research methods (15 pages), Chapter 4: Research results and discussion (69 pages); Conclusion, existence, recommendations (03 pages); References (09 pages) The thesis uses 17 references in Vietnamese and 122 documents in English The thesis has 23 tables and 27 images The thesis has published 06 articles, including international articles, domestic articles (3 articles in Vietnamese and 01 article in English) Chapter : OVERVIEW The dissertation has overviewed and summarized about main relevant issues to mangroves and the use of remote sensing in mangrove monitoring, specifically: (1) Research on the distribution of mangroves shows that in the world in general and in Vietnam in particular, there have been many elaborate studies on the distribution and status of mangroves Although recently, research on mangroves has been greatly developed through the support of scientific and technological achievements, but these works still have great scientific value, providing the basis for theory and practice for mangrove research (2) Research on the structural characteristics of mangroves shows that: The results of the overview study show that mangroves have many characteristic features such as lung roots, tuberous roots, salt leaves , mangroves have the ability to Highly adaptable to harsh coastal environments but very sensitive to changes, especially in the hydrological environment (e.g changes in water quality) In our country, the number of mangrove tree species is quite rich with many ecological and biological characteristics due to the difference in natural conditions in geographical areas Therefore, in different areas, the characteristics of mangroves on remote sensing data are not completely the same (3) Research on the spectral characteristics of mangroves based on remote sensing data shows that: In general, it is very difficult to separate mangroves from other vegetation using only greenness and strength forest health The key difference between mangroves and other vegetation and forest species is that mangroves have a higher leaf water content than other plants because the shortwave region is largely affected by water absorption in the leaves Therefore, the use of image channels (Green, NIR, SWIR) has the reflectivity of mangroves different from other plants (on optical materials and VH values on Sentinel-1 Radar images to highlight Enable mangrove forests This is the scientific basis for separating mangroves from terrestrial vegetation (4) Research on the use of documents used to establish mangroves shows that about the monitoring ability of remote sensing data, the support of the GEE platform shows the potential of satellite data high resolution to produce highly accurate maps of changes in mangroves Studies also show that medium-resolution image document classification techniques are well suited for mapping ecosystems (except at the species level), monitoring of large-scale changes, and analysis regional environmental relationships and assessment of mangrove status (quality, density, ) Global variability of mangroves is readily discernible from mid-resolution data analysis jar (5) Regarding methods and indicators of mangrove classification, research shows that, for mangrove detection indicators, each index has different accuracy and advantages relative to the rest of the index The CMRI index gives good mangrove classification results, but requires large input data The MVI is an index that takes into account the greenness and moisture content of mangrove vegetation on Sentinel-2 and Landsat-8 images, but has not been verified in many areas (6) Regarding the combination of optical and Radar image materials in the monitoring, it shows that: The combination of optical and Radar remote sensing materials shows many advantages when taking advantage of the advantages and limiting the disadvantages The scores of both are photographic, however, they have not been studied much for mangroves and studies rarely use average images over a period (month, quarter, year) because of the complexity and difficulties when using software to support processing and calculating digital images and few studies on using new technology platform GEE in mapping mangrove forests Some of comments: - The results confirm the difference in reflectance characteristics on photographic records of mangroves with terrestrial vegetation, in which, the difference is in the nearinfrared and short-wave regions, in addition to the blueness is the basis for classifying mangroves on optical remote sensing images - Multi-temporal remote sensing data better represents the unique nearshore coastal wetland ecosystems and habitats compared with single data Furthermore, multi-temporal remote sensing data improved the accuracy because the number of sample images obtained in a given period of time would have better statistical significance - Rapid and accurate mapping techniques are required for effective monitoring and management of mangrove resources Field surveys typically take a lot of time and effort Field surveys are also difficult to determine the distribution and abundance of mangroves due to the inaccessibility of mangrove communities Studies on mangroves often use a supervised classification method, the commonly used technique is the vegetation index However, each usually only applies to a specific area, with a large area that has not been studied much - Optical images with many cloudy and foggy days and complex land cover patterns in coastal wetlands, Synthetic Aperture Radar (SAR) Satellite Imagery not limited by climatic conditions and meteorology However, due to the imaging mechanism of SAR, it causes speckle, shadow and overlap noise in the SAR image Combining SAR and optical imaging can improve the accuracy of mangrove extraction, both taking advantage of the advantages and limiting the disadvantages of both photographic materials, however, they has not been studied much for mangroves and studies often rarely use average images of a period (month, quarter, year) because of the complexity and difficulty of using software to support processing and computational digital images and few studies on using new technology platform GEE in mapping mangrove forests It is necessary to have research to solve the above problems, which are: (1) using the vegetation index specifically for mangroves in the northern region, (2) incorporating both optical remote sensing data Science and Radar, (3) using a large number of images to limit the influence of tides and (4) applying advanced and modern technology for quick and accurate research results allowable Chapter 2: CHARACTERISTICS OF 2.1 General characteristics of the northern coastal area 2.1.1 Area 1: Northeastern coast The area along the Northeast Sea is an area with complex climate, hydrology and topographical characteristics This is a tropical climate with cold winters The Northeast coast has complex geomorphological, hydrological and climatic features; This area has a relatively rich mangrove flora, including highly salt-tolerant species 2.1.2 Area 2: Coast of the Northern Delta Although the coastal area of the Northern Delta is a tropical monsoon climate with cold winters, the temperature here is higher than that of region I Mangrove tree communities include brackish-loving species, in which the most dominant species is coriander distributed in estuaries Under the canopy of cork are black tiger and holly, forming a layer of shrubs; in some places black tiger and holly develop in clusters 2.2 Natural and socio-economic characteristics of the study area 2.2.1 Basic natural and socio-economic characteristics of Quang Ninh province 2.2.2 Basic natural and socio-economic characteristics of Thai Binh province 2.2.3 Basic natural and socio-economic characteristics of Hai Phong city Chapter 3: RESEARCH OBJECTIVES, SUBJECTS, SCOPE, CONTENTS AND METHODS 3.1 Objectives of the study - Studying the spectral characteristics of mangroves in the northern coastal provinces based on remote sensing data; - Selected remote sensing data and suitable classification method to classify mangrove forests in the northern coastal provinces ; - Determine the threshold for classification of mangroves in some northern coastal provinces - Assess changes in mangrove area in the period 2016-2020 in some northern coastal provinces - Developed technical guidelines for mangrove classification for some northern coastal provinces of Vietnam 3.2 Research subjects The research object of the thesis is the entire coastal mangrove forest in the Northern region of Quang Ninh, Thai Binh and Hai Phong provinces , using free remote sensing data with medium resolution including Whether Sentinel-1 Radar images , Sentinel-2 optical documents and Landsat-8 images 3.3 Research scope 3.3.1 Research area Level 1: Case study in representative areas including Tien Yen district and Quang Yen town (Quang Ninh province), two communes of Bang La (Do Son district), Dai Hop (Kien Thuy district) of Hai Phong city Level 2: Classification and assessment of changes in provinces including: Quang Ninh province, Hai Phong city and Thai Binh province 3.3.2 Research time The time to collect data on the standard plots of image key code (MKA) was conducted in 2019, the time to collect research sample points in the end of 2020 sample, using all photos of 2020 (January 01, 2020) 2020-December 31, 2020), to monitor changes in forest area at two times, the thesis collects photos of the last months of the year (from October 01 to December 31) 2016 and 2020 3.4 Research content 1) Distribution and ecology characteristics of mangroves in the Northern region 2) Selecting appropriate remote sensing data to classify coastal mangrove forests in the Northern provinces 3) Selection of methods to classify mangroves in the northern coastal provinces 4) Results of classification of mangroves in the study area from remote sensing data 5) Assessment of changes in mangrove forests in some northern coastal provinces in the 2016-2020 period 6) Development of technical guidelines for coastal mangrove classification for the northern provinces of Vietnam 3.5 Methods 3.5.1 Methodological basis 3.5.1.1 Theoretical basis For images with very short wavelengths (Opticals), the reflectance values of plants in general and mangroves at visible wavelengths are lower than in other states because most of the energy is absorbed by the leaves, a small part is absorbed through the remaining leaf is reflected For Radar images, in principle, depending on the wavelength, the image will mainly carry information about different objects, for mangroves the difference lies in the sensitivity to the environment different humidity (with water and without water) 3.2.1.2 Practical basis Using satellite images to monitor mangroves has been studied by many scientists and has achieved positive results Studies often use different vegetation indices, in which common indexes such as NDVI, SAVI, EVI are mainly used, some studies use scattering values of Radar images In addition, scientists have also developed many indicators specific to mangroves, which have been applied and proven to be relevant to many regions of the world 3.5.2 Methods of studying distributional and ecological characteristics of mangrove forests in the Northern provinces 2.1.1 Collect secondary documents: - Documents on natural, socio-economic conditions of Quang Ninh, Thai Binh and Hai Phong cities Map of planning for three types of forests in 2015; Map of results of provincial forest inventory in 2015, Map of updating forest changes to 2019 of the provinces of Quang Ninh, Thai Binh, and Hai Phong city 2 Mangrove field survey In the area of Quang Ninh and Hai Phong conducted to establish 15 standard mangrove forest plots representing a circle with an area of 1,000 m In the standard plots, the total number of trees was measured, including the species name, tree quality, and growth parameters of D 00 , and D 1.3 (for trees with diameter at chest height) >6cm) height (Hvn , Hdc), canopy diameter 3.5.3 Method of selecting suitable satellite image data to classify mangroves in the Northern region 3.1.1 Preprocessing of remote sensing images on Google Earth Engine a Landsat-8 datas Landsat-8 images are mined from the Landsat-8 collection in GEE and have been processed at T1 level Remove clouds on images by automatic cloud detection algorithm built-in GEE system based on image channel "'BQA'" After removing clouds, all channels of all images in the time series are merged using the me an () function on GEE to create a cloud-free image b Sentinel-2 datas Sentinel-2 images are mined from the collection: Sentinel-2: on GEE, level 1C, using the built-in cloud detection algorithm on GEE based on the "QA60" image channel After removing clouds, all channels of all images in the time series are merged using the mean () function on GEE to create a cloud-free image c Sentinel-1 datas Sentinel-1 images are mined from GEE's free data source, processing level 1-GRD T All Senitnel-1 images follow time series are merged according to the function me an () on GEE to produce an averaged image, less affected by image acquisition conditions and more likely to reduce speckle noise d SRTM DEM elevation digital model data In this thesis , DEM data was obtained from NASA's system integrated on GEE platform 12 3.5.6 Methods of assessing changes in mangrove forests in some northern coastal provinces in the 2016-2020 period 3.5 6.1 Locate forest fluctuations From the automatically classified mangrove distribution map from GEE, overlay the map at two assessment points including the 2016 and 2020 Mangrove Forest Map on QGIS 3.18, then use the tool Vector overlay/Difference to identify areas of forest variation between the latter and the former The thesis uses the updated map of forest changes in 2019 of Quang Ninh, Thai Binh and Hai Phong provinces to update information on forest plots from the Update Colum tool in the Table Module using the Mapinfo Pro15.0 software After having the information of the forest plot changes, use the area function ($area) in the Field Calculator toolbox on QGIS 3.18 software to calculate the specific area for each forest plot 3.5.6.2 Check changes in forest area in the field Synthesized to determine the change in forest area between times for a specific location in three cases as follows: Case 1: (1) On optical image, there is a fluctuation; (2) On the image Radar determines there is a change, this location has a plant change with 100% confidence Case 2: One of the two documents has fluctuations, then this location has plant fluctuations with 50% confidence Case 3: The remaining cases are determined to have no fluctuations The thesis uses 95 points identified with fluctuations in Thai Thuy district, Thai Binh province to evaluate the accuracy of the results of fluctuations Volatility lots (increase or decrease) will be pre-selected on the computer, then checked through high-resolution Google Earth images at the time of December 2016 and December 2020 for each specific location Brief description of the research process: CHAPTER 4: RESEARCH RESULTS AND DISCUSSION 4.1 Current status of mangroves in the Northern region 13 1.1.1 Distribution of mangrove forests in the northern coastal provinces According to administrative units, mangrove forests in Vietnam are distributed in 28 provinces and cities directly under the Central Government, of which there are provinces in the North region including: Quang Ninh, Hai Phong, Thai Binh, Nam Dinh and Ninh Binh (Figure 4.1) Figure 1: Distribution of mangroves in the northern coastal provinces The area of mangrove forests in the study area according to the statistical results in 2020 by region and sub-region is shown in Table 4.1 Table Area of mangrove forest in the period of 2020 according to statistics TT first Area Northeast Red river delta Total Conscious Quang Ninh Hai Phong peaceful Nam Dinh Ninh Binh Area (ha) 19,601 2.536 3,727 2,699 614 29.177 Percentage (%) 67.2 8.7 12.8 9.3 2.1 100 Source: MARD, 2021 [5] In terms of species composition, the results of field investigations show that mangroves in the North have a relatively rich species composition, a relatively diverse species of mangrove species, mainly species such as Mangroves Proboscis, Parakeet, Coriander, Trang, Mam Sea, Su , species with low average height, tree species with height ranging from 1.55.9 m, For coastal areas In the Red River Delta, there are many estuaries, so there is a lot of alluvium accretion every year, the sea water is low in salinity, so the main species composition of the species is Ban Chua, Trang, Su, Gia the height of the forest trees range from 5.6 15.5 m, about 10.1 ± 5.4m 4.1.2 Ecological characteristics of mangrove forests in the northern coastal provinces a Area 1: Northeastern coast This area has relatively rich mangrove flora, including highly salt-tolerant species, without typical brackish-loving species, except for swamps located inland such as Yen Lap and a part south of the Bach River Due to the strong influence of the flow 14 b Area 2: Coastal plains Rad river The coastal area of the Red River Delta has more distinctive features, with mangroves distributed at the coastal edge , most of which are delimited from agricultural land and residential areas by roads combined with sea dykes Selecting appropriate remote sensing data to classify coastal mangrove forests in the Northern provinces 4.2.1 _ Reflective characteristics of mangroves and cover states of the study area on remote sensing data Reflection characteristics of mangroves in Sentinel-2 image documents: Calculation results of reflection ratio of image channel overlay states on Sentinel-2 image are shown in the following figure Figure 4: Reflectance graph of overlay states on Sentinel-2 image in 2020 There are two reflection peaks of mangroves and terrestrial forests in the range with central wavelengths of 665 nm and 865 nm, corresponding to bands and 8A on the Sentinel2 image For terrestrial vegetation (including terrestrial forests and agricultural land) the reflectance peak is obtained at 865 nm, but at this wavelength, the reflectance value of agricultural land is lower than that of flooded forest salty The reflectivity of mangroves in the SWIR range (wavelengths 160 and 2190) is lower than that of other objects Reflective characteristics of mangroves in the Landsat-8 image document The result of calculating the reflectance value and the graph of reflectance rates of overlay states on the Landsat-8 image document is shown in the following figure Figure : 5: Reflectance graph of overlay states on Landsat-8 image 15 considerable similarities in the spectral signatures of mangroves and other vegetation classes ( terrestrial and aquatic vegetation such as duckweed, submerged grass, etc ) in the visible wavelength regions However, the canopy reflectance value of mangroves is usually lower than that of terrestrial vegetation in the NIR and SWIR bands Reflection characteristics of mangroves in Sentinel-1 image documents From the points to verify the overlays, the thesis calculates the VV and VH values of each state, the results are shown in the following figure: Figure 6: Channel reflectance values VV and VH of states on Sentinel-1 image Mangroves have the highest average values of VV and VH at -8.25 and -13.88 , respectively, followed by other objects with VV and VH values of -9.70 and -17.56, respectively the lowest observed in the area of frequent water presence with VV values of 20.46 and VH values of -27.78 VV and VH values are directly proportional to each other, however, because VH has characteristics of signal transmission and reception, and VH values are more sensitive to plants, VH values have many advantages for classification more mangroves 4.2 Evaluation of the monitoring capacity of mangrove forests of photographic materials theoretical conditions , the repetition period of each type of image Landsat-8, Sentinel2 , Sentinel-1 are: 16 days, days and days, respectively From there, determine the number of images obtained, the repetition period of the image at a location when using each type of image and combine them together according to the synthesis theory in the following table: Table : 2: Number of images, image repetition cycle at a location when using Landsat8, Sentinel-2, Sentinel-1 images and combining them together under theoretical conditions TT Photo material Landsat-8 Sentinel-2 Sentinel-1 Landsat-8+Sentinel-2 Landsat-8+Sentinel-1 Sentinel-1+Sentinel-2 Landsat-8 + Sentinel-1+ Sentinel-2 Cycle repeats 16 2 Year 23 73 122 96 145 195 218 Number of images Season Month 18 31 10 24 36 12 49 16 55 18 Week 2 4 16 Experimentally, the thesis has determined for Sentinel-1 documents: only 50% of the images of each type of shooting can be used for monitoring in a specific area For two types of documents Landsat-8 and Sentinel-2, the project processes clouds and cloud shadows of all 20-20 image scenes of these two documents, and then determines the average time of image repetition after treatment (no clouds and clouds left) for provinces with mangrove forests The results of determining the average repetition time of images at a location on the ground in the provinces with mangroves are shown in Table 4.3 Table : 3: Number of images, image repetition cycle at a location when using image materials and combining these documents together in real conditions TT Photo material Landsat-8 Sentinel-2 Sentinel-1 Landsat-8+Sentinel-2 Landsat-8+Sentinel-1 Sentinel-1+Sentinel-2 Landsat-8+Sentinel-1+ Sentinel-2 Cycle repeats 74 32 12 22 10 Number of images Year Season Month 11 30 16 35 42 11 47 12 Monitoring frequency Year Season Month Season Month Month Month Considering at one location, according to actual conditions 4.2.3 Discuss and select suitable remote sensing image materials to classify mangroves for the Northern coastal provinces Coastal mangroves are often narrow bands, so to ensure that these types of photographic materials are applied correctly, the size of an inland mangrove area should be at least one pixel at 30 meters for the Landsat image 7, 15m for the Landsat-8 image and 10m for the Sentinel-2 image Calculation results in Figures 4.4 and 4.5 also show that the reflectance rates of mangroves and other plants on the Landsat-8 and Sentinel-2 documents are similar In addition, the number of images obtained in years is not too big of a difference Compared with the Landsat-8 image (30 m, 8/16 days), Sentinel-2 has a clear improvement in spatial and temporal resolution Sentinel-2 images with better spatial resolution, more spectral channels provided, will have many advantages in building mangrove maps In addition, the research results of several authors have demonstrated that Sentinel images with a resolution of 10m have the ability to accurately detect mangroves on a national or regional scale For small patches of mangrove forest (< 01 ha), it is difficult to exploit with Landsat image with 30m resolution In order to both ensure the accuracy of the research results, save time and take advantage of the advantages of different types of documents, the study uses two main materials for research, which are Sentinel image documents -1 and Sentinel-2 Results of selecting methods to classify mangroves in the northern coastal provinces 17 4.3.1 Selecting a composite image method to classify mangroves 4.3.1.1 Variation of index value between single image scenes over time The results of calculating index values on image documents of MKA locations are as follows: At the same location there is almost no change in the forest in fact, considering the same index between two scenes, there can be very large variation (3 % between Sentinel-2 images and 9% - 12% between two Sentinel-1 images) 4.3.1.2 Index value fluctuations between average scenes a Calculation results for Sentinel-2 images: at the same MKA, the NDVI values between scenes of Sentinel-2 images, on average, have smaller fluctuations than between single images The average coefficient of variation S NDVI between years is only 8% b Calculation results for Sentinel-1 images: The results of determining the fluctuations in the VH index value between the quarterly average Sentinel-1 images for the 3rd year of 2018 - 2020 show that this coefficient for the whole year of 2018 is 3%, 4% in 2019 and 5% in 2020 The average coefficient of variation is 4% MKA the VH values between scenes of Sentinel - images on average fluctuate less than between single images The average coefficient of variation S NDVI between years is only 3%, lower than the quarterly and monthly averages research results on the variation of selected index values at each MKA location between single-image scenes and between time-averaged images in the gathering area in Table 4.7 Table 7: Coefficient of variation between single time image and average image over each time interval TT Photo material Sentinel-2 Sentinel-1 Coefficient of Variation (%) between scenes Mean Mean Mean by By Day by year Season Monthly 32 27 9-12 Monitoring frequency Year, Season Month, Season, year 4.3.2 Choosing a method to classify mangroves 3.2.1 Sentinel-2 image For accurate results when using the CMRI index, a large amount of training data is required to create the output map, but the Northern provinces have many characteristics to be able to test these indicators 3.2.2 Sentinel-1 image The thesis uses Sentinel-1 images used to remove year-round wetlands but not mangroves , in addition, the VH value on Sentinel-1 images can help to classify quite well among mangroves and non-mangrove plants and water surfaces 4.3.3 Combination of optical and radar materials for mangrove classification Both documents are used in conjunction on the basis of GEE through the function (.and) when the mangrove areas are within the threshold limits of classification of vegetation indices on Sentinel-2 optical images and value VH value of Sentinel-1 radar image Using the DEM 18 elevation map to separate terrestrial forest from the mangrove map, areas with elevations above 10 meters will be excluded 4.4.4 P classification of mangroves in the study area from remote sensing data 4.4.1 Check distribution of index values with each overlay state Density 1.5 0 Density 10 2.5 The test results of the normal distribution of CMRI, MFI, MVI index values from Senitnel-2 image and VH from Sentinel- image show that the value of deviation (Skewness) of the CMRI, MFI, and VH indexes of all states are close to zero, this result shows that the distributions of CMRI, MFI, VH all have normal distribution a For mangrove status: 0.40 0.60 0.80 CMRI S2 1.00 0.05 1.20 0.10 0.20 0.25 (b) Density 0 01 Density 02 03 (a) 0.15 MFI S2 -400 -200 MVI S2 (c) 200 400 -18.00 -16.00 -14.00 VH S11 -12.00 -10.00 (d) Figure 4.8: Results of testing the distribution of indicators for mangroves 4.4.2 Calculate index values for each state on image documents 4.2.1 CMRI For mangroves, there is a huge difference between the CMRI values across the two documents and varies from area to area CMRI TB mangroves all The area is larger than the CMRI TB of other plants, but the value of CMRI and CMRI max For other plants, the CMRI value of Quang Ninh area is less than in both documents, while the CMRI value of Hai Phong area is greater than 4.2.2 MFI index The results show that, although the average MFI value of mangroves in all areas is lower than that of Others, the minimum MFI values on Sentinel-2 images of mangroves and other plants are the same (MFI = 0) so, if in this case, using the minimum threshold to classify mangroves and other vegetation as MFI on Sentinel-2 would be very difficult 4.2.3 MVI The results of calculating the value of the MVI index have many anomalous values, especially for mangroves and other objects, this is reflected in the maximum and minimum values for the status of mangroves as well as other vegetation There are differences on the two photographic materials and in the two regions This calculation result is not suitable and it is difficult to classify mangroves in the study area 19 4.2.4 VH channel on photo Sentinel-1 The VH max and VH values of the water surface in all three areas are smaller than the VH values of mangroves, but there is an overlap between the VH max of the water surface and the VH of other plants in Quang Yen and Quang Yen areas Tien Yen The results of VH of the states in Hai Phong area are higher than in Quang Ninh, while the VH max values of the two regions are not significantly different 4.4.4 Establishing thresholds for classification of mangroves from vegetation indices Sentinel-1 Sentinel-2 4.4.3.1 Setting thresholds for classification From the above results, the thesis uses CMRI, MFI indexes for optical images and VH values to develop thresholds according to the method of statistical interval estimation and post-construction accuracy assessment The threshold for classification is as follows: Table 12: Thresholds for classification of states of study areas Image Data/Index C MRI Bang La - Dai Hop Quang Yen Tien Yen MFI Bang La - Dai Hop Quang Yen Tien Yen VH Bang La - Dai Hop Quang Yen Tien Yen Mangroves Others 0.10 - 0.41 0.25 0.41 0.10 >0.09 - 0.1 0.09 0.10 0.09 ( - 17.3) - 10.0 ) 0.0 - 0.41 0.0 - 0.25 0.0 - 0.41 0.0 - 0.10 0.0 - 0.1 0.0 - 0.09 0.0 - 0.10 0.0 - 0.09 ( - 26.0) - ( - 3) ( - 24.0) - ( - 15 8) ( - 26.0) - ( - 16 8) ( - 26.0) - ( - 17 3) ( - 15.8) - ( 10.0) ( - 16.8) - ( 10.0) ( - 17.3) - ( 10.0) Water surface < 0.01 < 0.01 _ < 0.01 _ < 0.01 _