A coastal wetlands mapping approach of Yellow River Delta with a hierarchical classification and optimal feature selection framework

Wetlands play an important role in ecological health and sustainable development, their spatial distribution and explicit thematic information are crucial for developing management and conservation measures. The Yellow River Delta is an important coastal wetland reserve in China, its wetland types are complex and diverse, natural and artificial wetlands are easily confused, making refined classification more difficult. To address this challenge, we proposed a new wetland mapping approach by combing hierarchical classification framework (HCF) and optimal feature selection. First, inheritance-based multiscale segmentation was carried out to obtain object-oriented images, and decision tree classification was used for preliminarily identify wetland and non-wetland. Second, recursive feature elimination and cross-validation (RFECV) was used to select optimal features, which was then utilized for wetland refinement extraction by using random forest (RF) algorithm. The experiments were performed based on Sentinel-1, Sentinel-2 and NASADEM datasets. The results show that effective wetland classification features can be selected by using RFECV. The feature scores are as follows, red edge index > spectral features > vegetation/water body index > backscatter coefficient > topographic features > texture features > location feature > urban index > geometric feature. The overall accuracy and Kappa coefficient of the method in this paper are 92.36 % and 0.915, which are 14.62 % and 6.68 % higher than using only HCF or only RFECV. Compared with the GlobeLand30 and CAS_Wetlands datasets, the refinement of wetland mapping in this paper is higher. This study provides a new idea in methodological selection for wetland information extraction, and the resulting coastal wetland map can be used for sustainable management, ecological assessment and conservation of the Yellow River Delta.