Variation-based spatiotemporal fusion of WorldView-2 and Planet imagery for intra-annual monitoring of S. alterniflora invasion in mangroves

Mangrove ecosystems are increasingly threatened by the invasion of Spartina alterniflora (S. alterniflora), which poses serious challenges to their biodiversity and sustainable management. Two significant research gaps remain unresolved: First, what is an efficient method to capture the high spatiotemporal dynamics of invaded mangrove forests? Second, how do invaded mangrove forests change within a year? To address these questions, this study proposes a novel framework for intra-annual monitoring of S. alterniflora invasion using synthetic imagery that retains fine spatial details and a dense time series. Specifically, high-resolution WorldView-2 imagery and high-temporal-resolution Planet data were fused using two spatiotemporal data fusion methods: the Flexible Spatiotemporal Data Fusion (FSDAF) method and the Variation-based Spatiotemporal Data Fusion (VSDF) method. Subsequently, Support Vector Machine (SVM) classification was applied to the synthetic images to map the distribution of mangroves and S. alterniflora across monthly intervals at 2-metre spatial resolution. The results indicate that VSDF outperformed FSDAF by more effectively preserving fine spatial features, such as vegetation boundaries. The classification map reveals the intra-annual dynamics of S. alterniflora and the stable condition of mangrove forests, achieving an overall accuracy of 92.5%. Classification performance was notably higher during the senescence period of S. alterniflora. Furthermore, the intra-annual dynamics of S. alterniflora, primarily driven by the transition from the growth period to the senescence period, resulted in a 1.4 ha reduction in its coverage. This framework provides a valuable approach for advancing mangrove ecosystem monitoring and guiding targeted management strategies.