Traditional environmental assessments often treat urbanized bays as homogeneous units, failing to detect localized degradation hotspots. Consequently, semi-enclosed bays remain the frontline of anthropogenic and natural conflicts where cumulative stressors often outpace resilience, a dynamic often hidden from aggregate metrics. To address these complexities, this study established a high-resolution “Drivers-Pressure-State-Impact-Response” (DPSIR) framework comprising 81 comprehensive indicators. A spatially explicit assessment was conducted across five distinct functional zones (e.g., mariculture, shipping, and wetland conservation areas) in Jiaozhou Bay, China, from 2013 to 2022. Quantitative analysis reveals a distinct recovery trajectory: the composite Environmental Quality Index (EQI) escalated by 135% from 0.31 (“Poor”, Class IV) in 2013 to 0.73 (“Moderate/Good”, Class II – III) in 2022, with 2017 identified as the critical inflection point. After 2017, a “Governance–Pressure Inversion” mechanism emerged, where governance intensity (R) structurally surpassed pollution accumulation (P). This phase transition effectively uncoupled economic growth from environmental degradation. However, recovery was not spatially uniform. We identified a distinct “Asynchronous Recovery” phenomenon: hydrodynamic “dead zones” lagged 2–3 years behind well-flushed areas, constrained by sediment memory effects. These findings challenge “one size fits all” governance models. We propose a “Zone-based Adaptive Management” strategy that prioritizes sediment remediation in the inner bay while focusing on risk prevention in port areas, offering a blueprint for sustainable governance in semi-enclosed coastal ecosystems globally. • Spatially explicit model assesses urban bay recovery (2013−2022). • Applied a spatially explicit model with 81 indicators over a decade. • Governance intensity structurally surpassed pollution pressure in 2017. • Inner bay recovery lagged years behind the flushed bay mouth. • Zone—based adaptive management is proposed for heterogeneous bays.
Han et al. (2026) studied this question.