Development of Land Use Regression models for particulate matter and associated components in a low air pollutant concentration airshed

Perth, Western Australia represents an area where pollutant concentrations are considered low compared with international locations. Land Use Regression (LUR) models for PM₁₀, PM_2.5 and PM_2.5 Absorbance (PM_2.5Abs) along with their elemental components: Fe, K, Mn, V, S, Zn and Si were developed for the Perth Metropolitan area in order to estimate air pollutant concentrations across Perth. The most important predictor for PM₁₀ was green spaces. Heavy vehicle traffic load was found to be the strongest predictor for PM_2.5Abs. Traffic variables were observed to be the important contributors for PM₁₀ and PM_2.5 elements in Perth, except for PM_2.5 V which had distance to coast as the predominant predictor. Open green spaces explained more of the variability in the PM₁₀ elements than for PM_2.5 elements, and population density was more important for PM_2.5 elements than for PM₁₀ elements. The PM_2.5 and PM_2.5Abs LUR models explained 67% and 82% of the variance, respectively, but the PM₁₀ model only explained 35% of the variance. The PM_2.5 models for Mn, V, and Zn explained between 70% and 90% of the variability in concentrations. PM₁₀ V, Si, K, S and Fe models explained between 53% and 71% of the variability in respective concentrations. Testing the models using leave one-out cross validation, hold out validation and cross-hold out validation supported the validity of LUR models for PM₁₀, PM_2.5 and PM_2.5Abs and their corresponding elements in Metropolitan Perth despite the relatively low concentrations.