Development of a land use regression model for characterizing intra-urban spatial variation of fine particles in an Andean valley

Abstract

Background: Air pollution is an environmental and public health problem of special interest due to its impacts on human health. In recent years, PM2.5 levels in the city of Medellin and the Aburrá Valley (Colombia), mainly linked to mobile emission sources, have exceeded the national threshold and WHO guidelines. Although there is an active air quality monitoring network, an important limitation related to spatial distribution is associated. Methods: To characterize the spatial distribution of fine particles in the city of Medellin, Land Use Regression (LUR) models were developed based on PM2.5 monthly average for august 2018, available for 13 monitoring stations. The models were built with meteorological, demographic, traffic and urban land uses data, with buffers of 50 to 500 m. Based on the specification criteria of the Ordinary Minimum Squares (OLS) method, the best model for the month of august was selected and a predictive map was developed for the entire study area. Results: Sampler height above ground surface, average monthly wind speed, distance to residential land use from the monitoring station, and vehicle flow within a buffer of 300 m, explained 79% of variability of PM2.5. Based on the prediction map, the most contaminated areas were found in the southern region of the study area, with levels between 18 and 30 µg/m3. In addition, the least contaminated areas were found in the northwest region, with levels below 16 µg/m3. Conclusions: The selected model included a variable that represented emission sources and pollutant dispersion, showing the influence of vehicle fleet and meteorology on PM2.5 levels. The LUR methodology is a simple and replicable alternative to estimate exposure to particulate matter; however, this method is susceptible to limited measurements sites.