Abstract: [Purposes] The large-scale outbreak of water hyacinth can exert significant adverse impacts on flood control,water supply security,and aquatic ecological health of rivers and lakes.To address the problems that water hyacinth is prone to cause misjudgments in cyanobacteria monitoring of Taihu Lake and interfere with cyanobacteria salvage and treatment during cyanobacteria bloom periods,a targeted water hyacinth monitoring method is proposed.[Methods] Based on the spectral response differences of water hyacinth,cyanobacteria,and other aquatic vegetation in specific bands of Sentinel-2 images,this study constructs corresponding feature indices,and extracts water hyacinth using threshold segmentation and decision tree classification methods.[Findings] According to the experimental results,the accurate extraction of water hyacinth in Taihu Lake can be achieved under the conditions of VREI >0.05,FLVI>0.57,and WVI>0.065.[Conclusions] Field verification indicates that this method can be applied to water hyacinth extraction and can distinguish water hyacinth from cyanobacteria,nymphoides peltatum,and other aquatic vegetation in Taihu Lake.The application of this method to Ge Lake further verifies its applicability．

Key words: water hyacinth, Sentinel-2, spectral features, decision tree, threshold segmentation