Spatial variation of ambient PM2.5 and PM10 in the industrial city of Arak, Iran: A land-use regression

Abstract

Land-use regression models (LUR) were developed to estimate the spatial patterns and seasonal variability of particulate matter with aerodynamic diameter ≤2.5 mm and ≤10 mm (PM2.5 and PM10) in the industrial city of Arak, Iran. PM2.5 and PM10 data were obtained from the air monitoring stations belonging to the Department of Environment from 2015 to 2020. Spatial interpolation of particles by ordinary kriging method was conducted to estimate pollutant variability in unmeasured areas. The predicted particles concentrations (PM2.5 and PM10) of the three hundred points were subsequently collected from the kriging map and accepted as a response variable in final LUR models. All available Geographic Information System (GIS) predictors in vector format were applied as potentially independent variables. The leave-one-out cross-validation (LOOCV) method was used to evaluate the performance of the LUR models. The annual mean PM2.5 and PM10 concentrations were 22.92 and 49.78 μg/m3 across the city. LUR models explained 88.5% of the variances (R2) for PM2.5 and 54.5% for PM10. For the PM2.5 and PM10 models, the mean LOOCV R2 values were 0.83 and 0.45, respectively. There were some variations between both LUR models for PM2.5 and PM10, but the heating and non-heating season models were nearly similar to the annual LUR models. These LUR models provide a spatial distribution of particles at the home addresses and predict concentrations for no-monitored areas. High cross-validation of this method confirms the robustness of the suggested approach to forecasting particle concentrations in another place.