Modelling the effects of meteorological variables on ozone concentration—a quantile regression approach

Abstract

Abstract This paper proposes the use of the conditional quantile regression approach for the interpretation of the nonlinear relationships between daily maximum 1-h ozone concentrations and both meteorological and persistence information. When applied to eight years (1992–1999) of data from four monitoring sites in Athens, quantile regression results show that the contributions of the explanatory variables to the conditional distribution of the ozone concentrations vary significantly at different ozone regimes. This evidence of heterogeneity in the ozone values is hidden in an ordinary least-square regression that is confined to providing a single central tendency measure. Furthermore, the utilization of an ‘amalgated’ quantile regression model leads to a significantly improved goodness of fit at all sites. Finally, computation of conditional ozone densities through a simple quantile regression model allows the estimation of complete density distributions that can be used for forecasting next day's ozone concentrations under an uncertainty framework.