The precise classification of coastal zone is of great significance for the protection and restoration of coastal wetlands. However, due to the mutual interference of sea and land background noise in nearshore environments, periodic tidal inundation, and frequent cloud and rain coverage, traditional coastal zone classification methods based on instantaneous remote sensing images have limited accuracy. A hierarchical classification method is proposed to address this issue, which combines land region information extraction based on the optimal image of object-oriented random forest and intertidal zone information extraction based on dense time series composite images. The method was applied to the coastal zone of Jiangsu Province, and its spatiotemporal evolution characteristics over the past 20 years were quantitatively described. The results show that the method has good application effectiveness, with an overall accuracy of over 87% and kappa coefficients exceeding 0.86 for the classification of coastal wetlands in Jiangsu Province at different periods. And in the past 20 years, the transformation of coastal wetlands in Jiangsu Province has shown a trend of natural wetlands, artificial wetlands, and non-wetlands. Land based natural and artificial wetlands are severely affected by human interference, with reduced area and intensified landscape fragmentation.

coastal wetlands; time series images; Google Earth Engine; classification; long-term changes