Visibility: an investigation based on a multivariate adaptive regression spline technique

Visibility in the Po Valley in northern Italy is a meteorological parameter of extreme interest. In autumn and winter radiational cooling of the Valley bottom coupled with cold air drainage from surrounding mountain slopes leads to conditions favouring the formation of fog with a highly persistent character. This regime may be temporarily interrupted when vigorous northwest flow over the Alps produces descending dry currents in the lee of the range (foehn) or when an intense disturbance affects the area.A preliminary application of a statistical methodology called MARS (Multivariate Adaptive Regression Splines) to a visibility time series of Milan-Linate Airport is carried out with the primary intent of investigating the efficiency of the technique in detecting the non-linear dynamics of the series and the physical factors involved. Although the development of a statistical prediction scheme requires much more data in order to guarantee the stability of the model equation and reliable tests, results are interesting and indicate, beyond any doubt, the skill of MARS as a tool for making causal inferences from bservables.