Wang, Shih-YaShih-YaWangLin, Tzu-ChiTzu-ChiLinLi, Hsueh-HsunHsueh-HsunLiYang, Chao-HsuChao-HsuYangPEI-TE CHIUEH2026-04-212026-04-212026-06-0113522310https://www.scopus.com/record/display.uri?eid=2-s2.0-105033437962&origin=resultslisthttps://scholars.lib.ntu.edu.tw/handle/123456789/737368Accurate spatial estimation of fine particulate matter (PM2.5) remains a central challenge in urban air quality assessment, largely due to limited monitoring coverage and the dependence of commonly used modeling frameworks on static spatial representations of the built environment. This study examines whether satellite-derived land cover (LC) data can provide an effective spatial representation for PM2.5 estimation within a land use regression (LUR) framework, in comparison with conventional administratively defined land-use (LU) datasets. Using Landsat 8 imagery processed on Google Earth Engine, LC classification was generated with an overall accuracy of 82% across five surface categories: agricultural and green land, forest, water bodies, transportation, and urbanized areas. A stepwise LUR framework was implemented to systematically compare LC- and LU-based models, incorporating meteorological variables as well as remote sensing indicators, including aerosol optical depth (AOD) and boundary layer height (BLH). Principal component analysis (PCA) was applied to mitigate multicollinearity among predictors. The LC-based model achieved an adjusted R2 of 0.791, demonstrating functionally equivalent predictive performance to the LU-based model (adjusted R2 = 0.790). The results indicate that satellite-derived LC data capture urban surface characteristics relevant to PM2.5 spatial variability with performance equivalent to traditional LU datasets. By relying on observation-based surface information rather than static administrative classifications, the proposed framework offers a flexible and transferable approach for urban PM2.5 estimation, particularly in regions undergoing rapid land transformation or facing constraints in land-use data availability.falseGoogle earth engine (GEE)Land use regression (LUR)Remote sensingSatellite-derived land coverUrban air quality modelingjournal article10.1016/j.atmosenv.2026.1219582-s2.0-105033437962