We developed LUR models for traffic-induced PM10, PM2.5, and PM1 in dense neighborhoods, a 10-block area (about 4 km2) in Xi’an, China. Two monitoring routes and 22 sampling sites along the routes were defined for data collection. We then conducted a mobile sampling at peak hours during weekdays. In the regressions, the dependent variables were the logarithms of 3 types of PM concentrations. Independent variables included traffic intensity, meteorological data, and land use types. The final PM10, PM2.5, and PM1 models accounted for 29.3%, 50.5%, and 35.5% of the variances, respectively. Green space within a 150-m radius, relative humidity, and open space within 100m, were indicated as the influential variables. Our results showed that in this dense urban area, the LUR model could perform well with an acceptable R2, and provided accurate references concerning spatial PM distribution. Along with mobile monitoring, the LUR method can be more effective and flexible, compared with other methods like dispersion models.

CICTP