A method previously developed for reconstructing daily erythemal UV doses at Sodankyl¿, northern Finland, was adjusted to the local conditions at Davos, Switzerland, and used for estimating the erythemal UV doses there over the period 1926--2003. The method uses total ozone, sunshine duration, and snow depth as input, and is based on the empirical relationship between relative sunshine duration and relative UV doses. In order to examine how the method behaves in different environments, the relationships found for Davos and Sodankyl¿ were compared. This revealed that the surface albedo and the cloud climate have a comparable influence on the relationship found. Although the method is fairly simple, it accounts for the most important factors affecting the amount of UV radiation reaching the Earth's surface. A comparison between estimated UV doses and the corresponding observations with a broadband biometer at Davos demonstrated the good performance of the method. The correlation coefficient for daily values varies between 0.95 and 0.98 depending on time of year, and the corresponding root mean square error is typically of the order of 20%. The monthly mean values show considerably less scatter around the regression line with a root mean square error of 4%. The time series of estimated UV shows that the UV level at Davos has varied considerably throughout the period of this study, with high values in the middle of the 1940s, in the early 1960s, and in the 1990s. Variations in the estimated UV doses prior to 1980, e.g., a steady decrease from the early 1960s to the late 1970s, were found to be caused primarily by changes in sunshine duration. Since 1980, on the other hand, there has been a distinct increase in the UV level caused mainly by the diminution of total ozone. This increase is most clearly seen during winter and spring, while the decrease from the early 1960s to the late 1970s is most pronounced during summer.