Estimating seasonal and spatial variability of particulate matter and sulfur dioxide concentrations by land use regression in Tehran, Iran

摘要

Background: The air quality in Tehran, Iran is poor compared with the cities in North America and Europe where land use regression (LUR) has been most frequently applied. Aims: We aimed to estimate seasonal and spatial variations of PM10 and SO2 pollutants in Tehran, Iran using LUR. Methods: The response variable data (SO2 and PM10) were obtained from 21 governmental automatic air pollution monitoring stations distributed across the metropolitan area of Tehran. A group of 210 potentially predictive variables was created within a Geographic Information System, including several variables that have not been used in other studies, and new variable types. A novel step-by-step algorithm for LUR model building was developed to select variables based on (a) consistency with the a priori assumptions about the assumed direction of the effect for each variable, (b) improvements to the leave-one-out cross-validation (LOOCV) R2, and (c) a multicollinearity index called variance inflation factor. Results: The annual average concentrations of PM10 and SO2 across the stations were 100.8 μg/m3 and 38 ppb, respectively. The LOOCV R2 values for the LUR models ranged from 0.52 to 0.71 for PM10 and from 0.62 to 0.73 for SO2. Although there was limited similarity between the PM10 and SO2 predictive variables, measures of bridge presence and bridge proximity were consistent across both pollutants. Conclusions: Resulting models and maps show that patterns were consistent throughout the year for PM10, but there were clear differences between the warmer and cooler seasons for SO2.