Development Of A Land Use Regression Model For Tree Pollen Exposure

Background: Little is known about how tree pollen varies within cities or whether spatial variation in pollen contributes to the development or exacerbation of allergic disease. We developed a land use regression (LUR) model for tree pollen that can be used to estimate exposure information for epidemiologic studies. Methods: We measured pollen at 45 sites within a U.S. city using modified Tauber traps placed 8 feet off the ground. We collected integrated samples for the entire pollen season of 2013 (March through October). We evaluated four types of variables characterized at five radial buffer sizes (50, 100, 250, 500, and 1,000 meters) for inclusion in an LUR model for log-transformed total tree pollen: tree canopy, building density, elevation, and traffic density. We also evaluated the distance to water for each monitoring site. For each variable that was significantly correlated with tree pollen, we passed the most highly correlated buffer size on to the LUR. We evaluated model performance using leave...