多源遥感协同下红树林的尺度效应OACHSSCD

Vegetation cover classification is a critical step in Earth system science and landscape ecology research,and the spatial resolution of multi-source remote sensing images exerts significant impact on the accuracy of vegetation in-formation extraction.The Dongzhai Port Mangrove Nature Reserve in Hainan and the Yingluo Mangrove Nature Reserve in Beihai,Guangxi,were selected as the study areas.Based on multi-source remote sensing synergy,the scale effects of mangrove extraction under different spatial resolutions were investigated,and the applicability and differential characteris-tics of multi-source remote sensing images in mangrove classification were quantitatively evaluated.Four sets of remote sensing data with varying spatial resolutions were selected:Jilin-1(0.75 m),Sentinel-2(10 m),GF-6(16 m),and Landsat-8(30 m).The Random Forest(RF)and Support Vector Machine(SVM)classification algorithms on the ENVI platform were employed to realize the classification and identification of mangroves and surrounding ground objects.In ad-dition,mixed pixel decomposition and spatial heterogeneity analysis were introduced to explore the influence mechanism of resolution variation on the classification accuracy of mangroves.The results showed that:1)Sentinel-2 and GF-6 im-ages achieve higher classification accuracy in mangrove extraction,and the optimal spatial resolution range for mangrove remote sensing classification is 10～16 m;2)The effectiveness of the linear decomposition method is significantly re-stricted by the spatial structural characteristics of the study areas;3)A significant correlation exists between mangrove extraction accuracy and regional spatial heterogeneity,i.e.,the higher the surface complexity of a region,the higher the requirement for the spatial resolution of remote sensing images for mangrove extraction;4)Resampled images exhibit weak sensitivity to scale changes,which will increase the uncertainty of mangrove classification results.This study pro-vides a reference for the remote sensing classification of mangrove vegetation,the selection of optimal resolution,and the uncertainty analysis of scale effect.