There have been relatively few studies that have attempted to separate the effects of the different factors affecting the short- and long-term changes of UV radiation. The effect of ozone is fairly easily represented, but that of other factors (surface albedo, aerosols, clouds) is more complex. In this paper, we present a methodology that can account for the other effects as well. We have used this methodology to study what part of the short- and long-term variability of measured spectral UV data from Sodankyl¿, Finland (67¿N), and Thessaloniki, Greece (40¿N), is explained by each factor. It was found that the effect of ozone on the short-term variability of monthly mean irradiance can be almost as high as 100%, whereas on average it is about 35%. The corresponding impacts by clouds are typically smaller, 40% and 12%, respectively. During May the albedo-related effect is strongest at Sodankyl¿, being 21% at its maximum and 7% on average at monthly levels. The amplitude of the variability caused by ozone is much stronger at Sodankyl¿ than at Thessaloniki, so a longer time series at the former place is needed to detect any possible long-term trend. In the Thessaloniki time series there is no significant ozone-related increase. In the summer data from Thessaloniki, however, there is a long-term increase, which is mainly caused by cloudiness. It was found that it is crucial to check the spectral data for any possible wavelength shift, if the long-term variability at a single wavelength is studied. Moreover, it was demonstrated that if the irradiance measurements of a given solar zenith angle (SZA) range are selected, the use of even the rather narrow band of 2¿ can introduce an effect that hampers trend detection. In other words, it is important to remove the effect of SZA and normalize all the measurements to some constant SZA value.