Based on a 4-year REKLIP data-set of global solar radiationG, shortwave surface albedo a, air temperature Ta and net radiative flux Rn, four types of regression models for the estimation of net radiative flux for three sites at different altitudes, located in the southern Upper Rhine valley have been proposed. In order to make for the limitation associated with the basic regression model (BRM) which relates net radiative flux over a surface to only incoming shortwave radiation, a longwave exchange coefficient has been introduced thus giving rise to the modified regression model (MRM). During daytime, the longwave exchange coefficient is observed to be negative for all three sites averaging about -0.20. The suitability of MRM over BRM becomes particularly obvious with respect to the mountainous site of Feldberg where the mean absolute error between measured and simulated Rn using MRM amounts to just half of that observed using BRM. Furthermore the role of clearness index and air temperature in the estimation of the net radiative flux have each been examined. The incorporation of the former is to make up for the effect of cloudiness on the net radiative flux budget, while the latter is an independent variable arising from the effective terrestrial radiation which thus allow for the estimation of the net radiative flux during all hours of the day. The regression models been proposed here have each been validated and their efficiency in reproducing actual measurements have been reported.
