Search

Land Use Regression (LUR) models have been used to describe and model spatial variability of annual mean concentrations of traffic related pollutants such as nitrogen dioxide (NO2), nitrogen oxides (NOx) and particulate matter (PM). No models have yet been published of elemental composition. As part of the ESCAPE project, we measured the elemental composition in both the PM10 and PM2.5 fraction sizes at 20 sites in each of 20 study areas across Europe. LUR models for eight a priori selected elements (copper (Cu), iron (Fe), potassium (K), nickel (Ni), sulfur (S), silicon (Si), vanadium (V), and zinc (Zn)) were developed. Good models were developed for Cu, Fe, and Zn in both fractions (PM10 and PM2.5) explaining on average between 67 and 79% of the concentration variance (R(2)) with a large variability between areas. Traffic variables were the dominant predictors, reflecting nontailpipe emissions. Models for V and S in the PM10 and PM2.5 fractions and Si, Ni, and K in the PM10 fraction performed moderately with R(2) ranging from 50 to 61%. Si, NI, and K models for PM2.5 performed poorest with R(2) under 50%. The LUR models are used to estimate exposures to elemental composition in the health studies involved in ESCAPE.

Land use regression models (LUR) frequently use leave-one-out-cross-validation (LOOCV) to assess model fit, but recent studies suggested that this may overestimate predictive ability in independent data sets. Our aim was to evaluate LUR models for nitrogen dioxide (NO2) and particulate matter (PM) components exploiting the high correlation between concentrations of PM metrics and NO2. LUR models have been developed for NO2, PM2.5 absorbance, and copper (Cu) in PM10 based on 20 sites in each of the 20 study areas of the ESCAPE project. Models were evaluated with LOOCV and "hold-out evaluation (HEV)" using the correlation of predicted NO2 or PM concentrations with measured NO2 concentrations at the 20 additional NO2 sites in each area. For NO2, PM2.5 absorbance and PM10 Cu, the median LOOCV R(2)s were 0.83, 0.81, and 0.76 whereas the median HEV R(2) were 0.52, 0.44, and 0.40. There was a positive association between the LOOCV R(2) and HEV R(2) for PM2.5 absorbance and PM10 Cu. Our results confirm that the predictive ability of LUR models based on relatively small training sets is overestimated by the LOOCV R(2)s. Nevertheless, in most areas LUR models still explained a substantial fraction of the variation of concentrations measured at independent sites.