With decreasing regional-transported PM2.5 levels, health risk assessment has become insufficient to reflect local source heterogeneity's contribution to exposure differences. This study combined ultra-high-resolution PM2.5 concentration with population distribution to estimate personal daily PM2.5 internal dose, considering indoor/outdoor exposure differences. A 30-m PM2.5 assimilating method was developed, fusing multiple predictors, achieving higher accuracy (R²=0.78-0.82) than chemical transport model outputs (R² = 0.31-0.64). Weekly differences were identified from hourly mobile signaling data. Population-weighted ambient PM2.5 concentrations varied among districts but didn't fully reflect exposure differences. The average indoor PM2.5 concentration was 26.5 µg/m³. The internal dose showed high exposure diversity among subgroups, and attributed mortality increased by 24.0% compared to coarser models.
