Introduction: Metro station entrances are among the most acoustically exposed urban public spaces, where transport, commercial activities, and pedestrian flows converge. Yet large-scale 24-hour noise evidence for these sites remains scarce, and existing studies seldom compare activity, built-environment, and source-composition drivers within the same model or translate findings into station-level intervention priorities. This study investigates spatiotemporal noise patterns, their multi-domain drivers, and governance prioritisation at metro station entrances in Shenyang, China.

Methods: Taking Shenyang, China, as a case study, a one-day 24-hour monitoring programme was conducted at 111 station entrance sites (one entrance per station) during April–September 2025, yielding 109 valid sites. Hourly A-weighted equivalent sound levels (LAeq) were integrated with a population-heat activity proxy, built-environment indicators, and sound-source composition derived from audio event recognition. Five noise outcomes (LAeq_24h, Lday, Lnight, LPeak, LInter-peak) were evaluated using robust OLS with cross-period stability comparison. Governance outputs combined a Baseline–Surplus typology with bootstrap-tagged Top-10 prioritisation.

Results: A clear diurnal rhythm was identified (peak-to-trough difference: 7.61 dB), alongside significant spatial clustering (Moran's I = 0.196–0.246, all p = 0.001). Source composition was the most stable driver domain: traffic share was consistently positive and natural share consistently negative across all five outcomes. The Baseline–Surplus framework differentiated structurally high-burden stations from those with excess noise, and 9 of the Top 10 priority stations showed high bootstrap stability (topk_prob ≥ 0.80).

Conclusions: This study provides city-scale screening evidence for noise exposure at metro station entrances. Source composition, particularly traffic and natural sound shares, offers the most stable diagnostic signal for governance. The Baseline–Surplus typology and bootstrap stability tagging can help managers identify priority stations and allocate resources under uncertainty, rather than relying on uniform exposure ranking alone.