Satellite NO2 data improve national land use regression models for ambient NO2 in a small densely populated country

Abstract Land use regression (LUR) modelling has increasingly been applied to model fine scale spatial variation of outdoor air pollutants including nitrogen dioxide (NO 2 ). Satellite observations of tropospheric NO 2 improved LUR model in very large study areas, including Canada, United States and Australia. The aim of our study was to assess the value of satellite observations of NO 2 in modelling the spatial variation of annual average NO 2 concentrations in a small densely populated country. We used surface level annual average NO 2 concentration and geographic information system data from 144 monitoring sites spread over the Netherlands: 26 regional background, 78 urban background and 40 traffic sites for developing land use regression models. For the 144 monitoring sites we obtained the annual average tropospheric NO 2 concentration for 2007 from the Ozone Monitoring Instrument (OMI) satellite sensor. These OMI data reflect a spatial scale of about 10 × 10 km. We calculated the correlation between satellite and surface level NO 2 concentrations for all sites and for background sites only. We next evaluated whether adding satellite observations improved land use regression models. Annual average satellite observations of tropospheric NO 2 correlated well spatially with annual average urban plus regional background (R = 0.74, n = 104 sites) and especially regional background NO 2 concentrations (R = 0.88, n = 26). The correlation was moderate for all sites, including traffic locations (R = 0.51, n = 144). A LUR model including satellite NO 2 observations performed better (overall R 2  = 0.84) than LUR models including geographical coordinates or indicator variables (overall R 2 65–74%) in modeling concentrations at the 104 background sites across the Netherlands. Satellite NO 2 observations agreed well with measured surface concentrations at background locations and improved land use regression models, even in a small densely populated country.