The application of semicircular-buffer-based land use regression models incorporating wind direction in predicting quarterly NO2 and PM10 concentrations

Abstract

Land use regression (LUR) models have proven to be a robust technique for predicting spatial distribution of pollutants with high resolution. Wind direction is an important factor affecting atmospheric environment quality. However, conventional LUR models have difficulties taking wind direction into consideration. This study put forward a semicircular-buffer-based (SCBB) LUR model to overcome this challenge. To assess the impact of wind direction on model performance, we set up two different LUR models for nitrogen dioxide (NO2) and particulate matter (PM10) in the urban area of Changsha, China. A location-allocation approach was used to identify sampling sites. Integrated 14-day mean concentrations of NO2 and PM10 were measured at 80 sites and 40 sites, respectively. Measured mean concentrations ranged from 17.0 to 75.7 for NO2 and 34.7 to 118.7 μg/m3 for PM10. Random samples of 75% of monitoring sites were used to the develop model and the remaining 25% of sites were retained for evaluation. Predictor variables were created in a geographic information system (GIS) and LUR models were developed with the most significant variables. The results showed SCBB LUR models had significantly higher R2 values than traditional LUR models, supporting the feasibility of this new approach incorporating wind direction in the LUR model.