Answers are grounded in this paper and cite their sources.
Aquatic vegetation (AV) dynamics are key indicators of shallow-lake ecological conditions, but sequential mapping remains challenging under complex and uneven observation conditions. Single-date methods lack temporal continuity, whereas conventional time-series approaches often depend on complete phenological sequences and are therefore limited in supporting sequential map updating. To address this problem, we propose a Bayesian multisource temporal fusion (BMTF) framework, which propagates historical class information through recursive Bayesian updating, integrates current optical and SAR observations with adaptive weighting, and updates optical likelihood parameters online to maintain temporal consistency. BMTF was evaluated in Baiyangdian Lake using 2025 Sentinel-1/2 time series, with independent validation from field surveys and high-resolution image interpretation, and further assessed using independent 2026 field samples to examine cross-year robustness. The framework achieved an overall accuracy of 91.42% and a Macro-F1 of 0.87, improved spatial completeness relative to a single-date AV mapping method, maintained reliable performance in complex vegetation-water transition zones, and showed cross-year robustness in the 2026 independent validation. The 10-day sequential monitoring results revealed distinct growth trajectories and changing contributions of submerged AV and floating-leaved/emergent AV. Overall, BMTF bridges static AV classification and sequential mapping, providing a Bayesian dynamical framework for continuous AV monitoring in shallow lakes.
No takes yet. Share an insight, caveat, or question.
Qi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: