Detecting and quantifying long-term oceanic chlorophyll-a (Chla) trends while mitigating associated uncertainties in multi-sensor merged satellite ocean color products poses persistent challenges. These challenges are intensified by the relatively short length of satellite observational time series and regionally low signal-to-noise ratios. Bayesian hierarchical space-time models improve long-term Chla trend estimation by accounting for spatial and temporal dependence in ocean biogeochemical signals, but most existing implementations rely on static, fixed ocean boundaries. Such rigid ocean partitioning approaches cannot reflect evolving bio-optical conditions across the global ocean, introducing additional uncertainty into long-term trend estimates. Here, we develop a dynamic Bayesian space-time framework that incorporates time-varying optical classes to represent shifting ocean optical structure across space and time. We applied this framework to global merged satellite Chla records spanning 1997 to 2025 to assess and quantify long-term trends and constrain product uncertainties. Our framework improves model fit and effectively reduces uncertainty in trend estimates, with the largest improvements occurring in optically heterogeneous and transition regions. The estimated global trends show widespread Chla declines across oligotrophic open oceans and localized increases in coastal and high-latitude waters, with similar spatial patterns in two independent merged satellite products. Our results show that dynamic optical variability provides a stronger basis to quantify long-term ocean biogeochemical changes and constrain uncertainties when applying multi-decadal merged ocean color CDRs for climate and ecosystem assessment.

暂无