<![CDATA[Reinforced concrete bridges in coastal environments are highly vulnerable to chloride-induced corrosion, which accelerates structural deterioration and increases maintenance demands. Conventional prediction models often fail to capture spatial heterogeneity in environmental exposure and temporal uncertainty in deterioration processes. This study aims to develop an uncertainty-aware spatio-temporal framework for predicting corrosion deterioration and supporting sustainable bridge maintenance decision-making. A hybrid Kriging–Hidden Markov Model is proposed to integrate spatial and temporal uncertainties within a unified probabilistic framework. The Kriging module reconstructs spatial chloride concentration fields from sparse environmental data, while the Hidden Markov Model captures stochastic transitions among latent deterioration states based on corrosion observations, with spatial exposure explicitly incorporated as a probabilistic driver. The framework is applied to three reinforced concrete bridges in Indonesia with varying exposure conditions. The results indicate distinct deterioration trajectories, with nearshore bridges reaching critical damage states at approximately 30 years, compared to about 58 years for inland structures. Model validation against inspection data demonstrates robust predictive performance. The probabilistic outputs provide actionable indicators, including time-to-failure distributions and critical deterioration thresholds, supporting risk-informed and sustainable maintenance strategies. The key novelty of this study lies in integrating spatial environmental variability with probabilistic temporal deterioration modeling within a unified framework to enable adaptive and lifecycle-oriented bridge maintenance decisions.]]>
