Classification Of Salt Marsh Vegetation In Coastal Zone Of China Based On Temporal Optics And Radar Images

Salt marshes are highly productive ecosystems in the mid-high latitude coastal zone,and the ecological service functions provided by different types of salt marsh vegetation are significantly different.Under the combined effects of human activities and climate change,such as reclamation,invasion of Spartina alterniflora,sea-level rise,the structure and spatial distribution of salt marsh vegetation in the coastal zone of China has changed rapidly.According to the phenological characteristics of salt marsh vegetation,time series method can improve the classification accuracy.However,the time distribution of optical data in the coastal zone is sparse and uneven due to the influence of clouds and tides.SAR can obtain complete time series image data and is sensitive to vegetation structure and water information,can be used as auxiliary data of optical image.Based on the annual time series sentinel-1 and sentinel-2 images and field survey vegetation distribution samples,combined with the analysis of time series NDVI and backscatter characteristics,this paper compared the decision tree method based on expert knowledge with the random forest method,discussed the classification accuracy of the combination of pure radar image and radar optical image,put forward the salt marsh vegetation classification method based on the combination of optical and radar image.The salt marsh vegetation distribution data in China's coastal zone in 2018 was obtained with an overall accuracy 91.2%,and the distribution status and historical changes of salt marsh in China and other provinces were analyzed in combination with the literature.This method can accurately and quickly obtain the spatial distribution and interspecific composition information of salt marsh vegetation,which is of great significance for biodiversity protection,wetland ecosystem function improvement and coastal ecological environment management.The main conclusions are as follows:(1)In terms of classification method,the analysis process of classification method based on expert knowledge is simple and clear,and the classification accuracy meets the needs of large-scale salt marsh vegetation survey.The optimal feature can be obtained by histogram statistical analysis,or precisely calculated by SEaTH method;the optimal classification threshold can be adjusted based on SEaTH method and histogram.If the optimal features are known in areas with similar vegetation types and phenological characteristics,the application range can be extended to different spatial time span by fine-tuning threshold.Although the accuracy of random forest method is slightly improved,the calculation is large,and the classification process is not visible.When the research area or data time changes,the samples need to be collected again,which is more dependent on the samples.Therefore,the classification method based on expert knowledge is more suitable for the situation that it is difficult to obtain and slow to update large-scale salt marsh vegetation samples.(2)In the use of data sources,combining of radar and optical remote sensing data have the highest accuracy when used together,and can ensure the stability of data and the precision of classification results within the year.The data source of radar image is stable,only using the radar image of the whole year can complete the classification of salt marsh vegetation,and it can be used as a more stable classification data source in the area covered by many clouds.But the results of radar data classification are relatively rough in tidal channel and vegetation junction.The optical image is seriously affected by cloud and tide level,and the available image time is uncertain.If we can get the appropriate image,it is better than the radar image in describing the details of features such as tidal channel.Therefore,the best classification accuracy can be obtained by adding appropriate optical image on the basis of stable acquisition of radar image.(3)The Yangtze River estuary is a typical area in the middle latitude.According to the characteristics of radar backscatter and vegetation growth in the Yangtze River estuary,the annual and monthly average VV,VH polarization and NDVI was selected as the available classification features.According to the time series radar backscatter analysis,the key radar features of the salt marsh vegetation extraction in the Yangtze River estuary are annual VH polarization,annual VV polarization,VH polarization in April and VV polarization in November,which matches with the vegetation phenological features.According to the time series NDVI analysis,the months suitable for extracting water body and tidal flat,Scirpus mariqueter,Phragmites australis and Spartina alterniflora are from July to October,late May to July,late April to early May and October to November,respectively.(4)The Yellow River estuary is a typical area in high latitude area,where there has little rainfall and many optical images.However,the coverage of Suaeda glauca in large area is low,and the spectral reflectance is affected by the background soil.The difference between near-infrared and infrared bands is very small,so it is difficult to distinguish between Suaeda glauca and tidal flat by NDVI.Based on the sensitivity of radar signal to the surface roughness,the VH polarization backscattering intensity of Suaeda glauca and tidal flat is quite different in summer.Therefore,the threshold value of distinguishing Suaeda glauca and floodplain can be determined by comparing the backscatter histogram of winter and summer images.(5)Jiulongjiang River estuary is a typical area in low latitude area,with large tidal range,large amount of rainfall especially in summer and few images available in the whole year.The average VH polarization image in summer is used to distinguish vegetation and water floodplain instead of optical image in summer,and the coincidence of classification results is more than 98%.It is difficult to distinguish salt marsh vegetation and mangrove by radar backscatter,but there are obvious differences in phenology between them in optical image,which can be distinguished by optical image in winter.(6)In 2018,there were 106352.5 hm~2 of salt marsh vegetation in the coastal zone of China,mainly including five types of Spartina alterniflora,Phragmites australis,Suaeda glauca,Scirpus mariqueter and Tamarix chinensis,respectively accounting for61.4%,18.2%,11.3%,7.4%and 1.6%of the total salt marsh area in China.Salt marsh vegetation in Shanghai,Jiangsu,Shandong and Zhejiang Province is widely distributed and covers all vegetation types.In other provinces and cities,the distribution area of salt marsh vegetation is small and the vegetation type is relatively simple.The main reason is high-intensity reclamation and the invasion of Spartina alterniflora.