Inducing flooding index for vegetation mapping in water-land ecotone with Sentinel-1 & Sentinel-2 images: A case study in Dongting Lake, China

Water-land ecotone is a core area of wetlands which play an important role in ecological function. The infor-mation of vegetation type on the water-land ecotone is important for detecting the health of the wetland, and its protection. Seasonal flooding is a key factor in controlling the environment in water-land ecotone. Even pre-venting direct detection of the flooded vegetation with satellites, seasonal flooding has effects on the distribution of plants, which may provide potential information on vegetation classification with satellites. Therefore, the study induces flooding index in vegetation mapping in water-land ecotone with satellites. We applied Sentinel-1 and Sentinel-2 images to map vegetation in the Dongting Lake with random forest method. The Sentinel-1 SAR images were used to obtain the flooding index (i.e. inundation frequency), and Sentinel-2 images were applied to acquire the optical characteristic of vegetation. A comparative experiment based on whether the flooding index was added or not during vegetation mapping in the Dongting Lake was carried out to explore the role of flooding index in vegetation mapping. The results concluded that (1) seasonal flooding will hinder vegetation mapping using a single phase of optical information; (2) the classification accuracy of single phase scenario was signifi-cantly improved with the aid of flooding index especially in wet season; (3) the classification accuracy can be somewhat improved by adding flooding index for the multi-temporal scenario. The study implies that inundation frequency information can be used for vegetation mapping in water-land ecotone, especially for complex regions with insufficient cloud-free satellite images.

Automated summary

Water-land ecotone plays a crucial role in wetland ecological function. Vegetation information in this area is important for wetland health assessment and conservation. Seasonal flooding affects plant distribution and can provide potential information for satellite-based vegetation classification. This study focuses on mapping vegetation in the Dongting Lake using flooding index derived from Sentinel-1 SAR images and optical characteristics from Sentinel-2 images. Results indicate that adding flooding index significantly improves the accuracy of vegetation mapping, especially during the wet season.