Constructing and Evaluating Models for Predicting Visibility for Data-Void Locations in Norway Using Weighted Least Squares.

Abstract : The Weibull distribution was used to model visibility at 51 Norway stations. Weighted least squares was used to investigate which variables, when incorporated into the coefficients of alpha and beta, would give a better fit. The model demonstrated that, as in Germany, incorporation of elevation and relative elevation into the variables model gives dramatically lower RMS over the constants model. Inclusion of latitude and longitude did not seem to improve the model significantly. The prediction ability of the model at data-void locations was estimated using sample re-use and indicated the model will be reasonably accurate. (Author)