Particulate　matter　with　an　aerodynamic　diameter　of　2.5　μm　or　less　(PM2.5)　concentrations　exhibit　significant　spatiotemporal　variability　due　to　diverse　urban　scenarios,　making　accurate　PM2.5　simulation　challenging.　This　study　used　PM2.5　mobile　monitoring　data　and　considered　the　impact　of　urban　scenarios　on　PM2.5.　A　hybrid　PM2.5　simulation　model　was　developed　by　integrating　the　Multiscale　Geographically　Weighted　Regression　model　(MGWR)　and　Extreme　Gradient　Boosting　(XGBoost),　referred　to　as　MGWR-XGBoost.　Nested　buffer　analysis,　combined　with　feature　importance　and　partial　dependence　plots,　was　employed　to　quantify　the　scale-dependent　effects　of　urban　scenarios　on　PM2.5.　Additionally,　spatiotemporal　patterns　of　PM2.5　concentrations　across　different　urban　scenarios　were　analyzed.　The　results　showed　that　the　MGWR-XGBoost　model,　which　introduced　urban　scenario　variables,　achieved　a　mean　improvement　of　10.5　%　in　the　coefficient　of　determination　(R2)　and　a　mean　reduction　of　1.52　μg/m3　in　the　root　mean　square　error　(RMSE),　thereby　enabling　intra-city　PM2.5　simulations　at　a　spatial　resolution　of　100　m　×　100　m.　Quantitative　analysis　revealed　that　roads　and　industrial　areas　could　influence　PM2.5　concentrations　at　a　regional　scale　(1000–1500　m　buffers),　whereas　residential　areas,　parks,　sports　services,　and　educational　and　medical　units　primarily　exhibited　more　localized　impacts　(100–500　m　buffers).　Spatially,　PM2.5　concentrations　in　the　study　area　exhibited　a　southeast-high,　northwest-low　pattern,　with　higher　pollution　in　construction　sites,　roads,　and　heavy　and　light　industrial　areas.　Temporally,　PM2.5　pollution　levels　across　different　urban　scenarios　exhibited　a　rise–fall–rise　pattern.　The　findings　provide　support　for　fine-scale　PM2.5　monitoring,　urban　planning,　and　pollution　control.　©　2025　Elsevier　B.V.