Nitrogen dioxide prediction in Southern California using land use regression modeling: potential for environmental health analyses

We modeled the intraurban distribution of nitrogen dioxide (NO2), a marker for traffic pollution, with land use regression, a promising new exposure classification technique. We deployed diffusion tubes to measure NO2 levels at 39 locations in the fall of 2003 in San Diego County, CA, USA. At each sample location, we constructed circular buffers in a geographic information system and captured information on roads, traffic flow, land use, population and housing. Using multiple linear regression, we were able to predict 79% of the variation in NO2 levels with four variables: traffic density within 40-300m of the sampling location, traffic density within 300-1000m, length of road within 40m and distance to the Paci. c coast. Applying this model to validation samples showed that the model predicted NO2 levels within, on average, 2.1 p.p.b for 12 training sites initially excluded from t he model. Our evaluation of this land use regression model showed that this method had excellent prediction and robustness in a North American context. These models may be useful tools in evaluating health effects of long-term exposure to traffic-related pollution.