Development of Land Use Regression (LUR) models for assessment of annual average PM10 and endotoxin exposure levels in ambient air in a livestock dense area

Abstract

Introduction There is concern about the influence of livestock farms on the health of Dutch inhabitants. Results from studies on health effects of livestock farming based on exposure proxies are inconsistent. Dust exposure measurements may enable more refined exposure-response analyses. We aimed to develop LUR models for particulate matter 10 (PM10) and endotoxin, known to be emitted from livestock farms and associated with adverse health effects. Methods Ambient PM10 was collected with Harvard Impactors at 61 sites (residential gardens) representing a variety of nearby livestock related characteristics. Three to four 2-week averaged PM10 samples were collected at each site over the course of 1.5 year. In addition a local reference site was set up and measured continuously to take into account temporal variation. Samples were analyzed for PM10 mass by weighing and endotoxin using the LAL assay. LUR models were developed using temporally adjusted annual PM10 and endotoxin exposure averages and livestock-related GIS variables (distances to and number of farms /animals by animal species). Results More spatial variation was observed for endotoxin compared to PM10. The model explained variance was higher for endotoxin than for PM10 (Adj. R2 0.56 and 0.12 respectively; number of predictor variables 7 and 2 respectively). Predictor variables included number of farms and type of animal species kept in the surroundings, and distance to the farms. Conclusions In conclusion, the effect of livestock-related sources on annual average exposure levels seems considerable for endotoxin and more limited for PM10. The LUR approach used, similar to air measurement studies focusing on other air pollutants, was found to be suitable to predict spatial variation in a livestock dense area. Thus far only livestock related predictor variables were explored. The developed LUR models should be further enriched and validated, before they can be applied to predict air pollution concentrations.