Land use regression study in Lanzhou, China: A pilot sampling and spatial characteristics of pilot sampling sites

Background: Land use regression (LUR) has been widely used to estimate air pollution exposure in recent epidemiology studies. However, few LUR studies were conducted in China, and even fewer used purposefully designed monitoring networks. The objectives of this study are to obtain preliminary understanding of fine-scale air pollution distributions, and to provide a foundation for a future extended study in Lanzhou, China, a major industrial city. Methods: A pilot monitoring network was designed using stratified-random sampling, and purposeful selection in gaps of spatial predictor distributions. Based on this network, NO₂ were measured using Palmes tubes for 2 weeks in summer 2015, which were used to develop a pilot LUR model considering spatial information of traffic and population densities, elevation, land cover, and land use. We developed linear regression, kriging models, including ordinary kriging, universal kriging, and compared them using AIC. Results: The sampling sites of the pilot monitoring network represented wide ranges of spatial predictors (N = 47). The pilot LUR model explained 71% of the variance in the measured NO₂ at the sampling sites. The spatial predictors in the model included road densities, elevation, and district indicator. Predicted NO₂ concentrations were higher in the east of the city, which is more developed and has dense road networks. Linear regression model performed better than the kriging models due to the lowest AIC. Conclusions: This study developed a pilot monitoring network that can effectively capture variability in spatial characteristics and developed a robust LUR model capturing small-scale spatial variations of air pollution in an understudied area. The predicted and measured NO₂ showed substantial spatial heterogeneity that was not captured by the limited government monitors. A future study with extended monitoring network and measurements from more seasons is needed to fully understand the distribution of air pollution in Lanzhou, China.