Information Extraction And Ecological Quality Assessment Of Yancheng Coastal Wetland Based On Remote Sensing

The narrow coastal zone in the north of Jiangsu Province has given birth to a large number of coastal wetlands,which are complex systems formed by the interaction of sea and land,providing a series of important ecosystem service functions such as carbon sink,purification,climate regulation,maintenance of biodiversity,resource supply and tourism culture,which are closely related to human life,so its habitat complexity determines the universality of non-zonal distribution rules,providing rich habitats and breeding sites for many organisms.It is the type with the highest value of ecological services per unit area of all ecosystem types.In order to carry out more comprehensive monitoring and protection of coastal wetland environment,the use of remote sensing technology to extract information for wetland environmental assessment provides great help,the current remote sensing technology can provide qualitative and quantitative analysis,has the advantages of both temporal resolution and spatial resolution,with the improvement of satellite image quality,gradually realize real-time monitoring and management of changes in wetland resources..Yancheng Coastal Wetland is one of the most typical silty tidal coastal wetlands in China and even the world,and was included in the World Natural Heritage List in 2019,becoming the first World Natural Heritage Site in Jiangsu.In this paper,combined with remote sensing technology and machine learning methods,Sentinel-2 images with representative four seasonal phases in 2020 were selected,and spectral features such as texture features,red edge index,vegetation and water body index were extracted for multi-temporal phase combination,and two machine learning methods such as K-nearest proximity method and random forest were used to establish a vegetation information extraction model in the study area,and 14 groups of vegetation information extraction schemes,including single-season phase,multi-seasonal phase combination experiment and preferred feature combination experiment,were designed.Taking the core area of Yancheng Rare Bird Reserve as the study area,96 characteristics such as spectral characteristics,vegetation and water index characteristics and texture characteristics of Sentinel-2 images were extracted in each season.In this paper,the method of information extraction of Yancheng coastal wetland is improved,and the ecological quality of the study area is graded and evaluated by combination with the remote sensing ecological index,and the following conclusions are mainly drawn:(1)In this paper,the object-oriented multi-scale image segmentation method is selected for the information extraction of the coastal wetland,and the multi-scale segmentation parameters that are most suitable for various features in Yancheng coastal wetland are compared: the segmentation scale is 131,the shape factor is 0.2,the compactness factor is 0.5,the weight of the red edge band is assigned to 2,and the weight of the remaining bands is 1.An object-oriented coastal wetland information extraction model was constructed,which had better applicability at the temporal and spatial scales.(2)The random forest method was better than the K nearest neighbor method in single-season phase experiments;According to the vegetation growth law,the four seasonal phases were divided into vegetation growth period and vegetation dormancy period for wetland information extraction of multi-seasonal facies,and the preferred feature combination after ranking the feature importance of the images by random forest model had a good dimensionality reduction effect: the overall classification accuracy of vegetation growth period reached 98.83%,KAPPA coefficient was 0.986,the overall accuracy of vegetation dormancy period was 97.79%,and the KAPPA coefficient was0.978,which verified the effective correlation between vegetation growth law and information extraction results.The technical scheme implemented in this paper improves the difficulties of scattered distribution and similar shape and characteristics of vegetation information extracted by remote sensing technology,and reduces the redundant information generated by feature combination,and successfully extracted the feature information of Yancheng coastal wetland.(3)The average of the remote sensing ecological index of 4 Landsat 8 images that can be obtained from Yancheng Coastal Wetland in 2015,2017,2019 and 2020 and the composite images in spring and summer each year was calculated through the GEE platform,and the ecological quality grades were divided and compared,and they were divided into 5 grades according to the difference of 0.2,namely: poor(0～0.2),poor(0.2～0.4),medium(0.4～0.6),good(0.6～ 0.8)and excellent(0.8～1).The average RSEI of the study area increased steadily from 2015 to 2020,and the RSEI in spring and summer was higher than that of the annual image due to high vegetation coverage,indicating that the ecological quality of Yancheng coastal wetland showed a trend towards better development overall,indicating that the ecological redline planning and the implementation of wetland protection plan in Jiangsu Province achieved remarkable results.A good wetland environment can provide a good habitat for waterbirds such as red-crowned cranes and protect biodiversity in coastal areas,which indicates that the establishment of a hierarchical wetland protection system,the improvement of wetland protection system,the implementation of wetland ecological redline system,the implementation of multiple measures for degraded wetland restoration projects,and the increase of wetland area need to be continuously implemented to provide more guarantees for coastal wetlands.