Modelling the decay of concentrations of nitrogenous compounds with distance from roads

Abstract

Passive sampler concentration measurements of nitrogen dioxide (NO 2) and ammonia (NH 3) were performed in the framework of a transect study to investigate the impact of vehicle emissions on ecosystems dissected by highways. The concentrations of both gases decrease markedly with distance from highway to regional background pollution values. Modelling the functional form of the decay is of interest for reducing the amount of measurements, for exposure assessment, and for predicting background concentrations. Three modelling approaches are compared: the exponential decay function, the shifted power-law function, and the linear-logarithmic function. The models were fitted to four compound- and year-specific data sets from one transect, applying mixed-effects models for repeated-measurement designs. The goodness-of-fit did not differ consistently between the model classes. Combined data from four transects with different characteristics were analysed with the exponential decay model, allowing for transect-specific random coefficients. From the empirical point of view, none of the three model classes is consistently superior to the others. But for prediction beyond the observed distance range it is essential to consider a model with meaningful parameters. The final choice of a model depends on the amount of data and on the characteristics to be represented by the model.