Land-Use Regression Models of Particulate Matter in Temuco, Chile Due to Residential Wood Burning

Abstract

The spatial model development of particle distribution are necessary for studying the chronic health impacts of particulate air pollution due to residential wood- burning. The spatial variability of fine particulate (PM2.5) and ultrafine particles (UFP) due to wood-burning has not been studied in Chile. In this study, we used data from a mobile sampling campaign to develop a Land-Use Regression (LUR) model in the city of Temuco, Chile, which is a city heavily impacted by wood-burning. This campaign was performed for 28 days in the winter of 2016, in which PM2.5 and UFP measurements of 1 s-intervals were recorded along four routes during nighttime when wood burning emissions are the highest. LUR models were developed to yield a map with spatial variability of PM2.5 and UFP. Predictor variables were obtained from land use, transportation, demography, meteorology, and geography information in a GIS environment. A background normalization of PM2.5 and UFP mean values at a 100 m resolution was applied to minimize day-to-day urban air quality variations. The results show that the LUR models were capable of explaining 0.43 and 0.31 of the variation in PM2.5 and PUF, respectively. The number of households, length of local roads, and residential areas were positively correlated with PM2.5 concentrations, whereas the average density of woodstoves, proximity to local roads, elevation, temperature, and precipitation were negatively correlated with PM2.5. The significant predictor variables positively associated with UFP concentrations were total population, major roads length, traffic intensity, and temperature, while the proximity to major roads and precipitation were negatively correlated with UFP. These results provide some insights for the development of a more sophisticated sampling campaign and LUR to be used to study chronic effects of particle exposure.