Fragmentation Typologies, Trajectories, and Path Dependence in Urban Mangrove Landscapes of Southeast Asia

Context In rapidly urbanizing coastal regions, mangrove ecosystems are increasingly fragmented by cumulative human and natural disturbances. Although fragmentation is widely recognized as a dynamic process shaped by historical legacies, mangrove fragmentation is still most often assessed using snapshot-based measurements that obscure the spatio-temporal pathways through which fragmentation develops. Objectives This study aims to (1) characterize urban mangrove fragmentation as a set of distinct, process-based spatio-temporal transformations using patch-level typologies, and (2) evaluate whether fragmentation trajectories exhibit path dependence, such that patch fate is conditioned by prior fragmentation states across diverse urban landscapes. Methods Mangrove fragmentation was analyzed across 21 Southeast Asian urban areas from 1996 to 2020 using patch-level landscape metrics derived from Global Mangrove Watch data. Fragmentation typologies were identified using a Random Forest machine-learning approach and linked across time periods using a maximum-overlap child–parent tracking framework to reconstruct multi-step patch trajectories. Path dependence was quantified within a closed cohort of patches traceable from 1996 using an information-theoretic framework that compared history-conditioned models with memory-less baselines. Results Across urban areas, mangrove patches consistently became smaller, more edge-exposed, geometrically simpler, and more isolated, indicating widespread intensification of fragmentation. However, similar trends in aggregate fragmentation metrics arose from fundamentally different spatial processes. These differences were captured by four dominant fragmentation typologies—Nibbling, Clearing, Shattering, and Displacing—which varied substantially in both relative frequency and extent among urban areas. Fragmentation trajectories were highly uneven, with a small number of recurring pathways accounting for most observed sequences. Trajectories dominated by Shattering and Clearing were strongly associated with subsequent patch loss. History-conditioned models consistently outperformed memory-less baselines, providing robust evidence of path dependence along persistent patch lineages. Although entropy-based effect sizes were modest, they were stable across robustness checks and increased with longer fragmentation histories, indicating partial but meaningful historical constraint on fragmentation dynamics. Conclusions These findings demonstrate that urban mangrove fragmentation unfolds through historically structured sequences of spatial transformations. By shifting emphasis from snapshot metrics to fragmentation trajectories, this study reveals lineage-conditioned constraints on landscape change that are not detected by conventional assessments. The proposed typology- and trajectory-based framework provides a transferable approach for diagnosing fragmentation dynamics and identifying high-risk pathways in urban and other human-modified landscapes.