Refined mangrove mapping in a highly-invaded area with multi-temporal Planet-Landsat-8 data fusion

Abstract

The mangrove ecosystems are significantly affected by the invasion of the exotic species such as Spartina alterniflora, challenging their biodiversity and sustainable development. In such heterogeneous landscapes composed of both mangroves and invaded Spartina alterniflora, the task of mapping mangroves become very challenging, owing to the fact that few remote sensing images maintain both high spatial and temporal resolutions as needed. In this study, we proposed a multi-temporal classification method based on spatiotemporal fusion of Planet and Landsat data to generate high-spatial resolution image time series. It attempts to achieve higher classification accuracy by inputting 3m resolution Planet data and utilizes the Flexible Spatiotemporal Data Fusion (FSDAF) method to synthesize Planet sample images from different phenological periods. The results show that the overall classification accuracy has increased by X% compared to the results of Planet-Landsat-8 data fusion and WorldView-Sentinel-2 data fusion. This indicates that enhancing the spatial resolution of input images is one way to improve classification accuracy. Given the ongoing advancement of commercial satellite systems, which could enable the use of a singular data type for classification and monitoring in the future, the exploration of more precise classification methods maybe a promising direction for future research.