Routine uniform spectral UV-B measurements with Brewer spectrophotometers in the Canadian network began in 1989. This relatively short duration of UV measurements militates against reliable detection of long-term changes in UV. A statistical model has been developed to extend the record of UV back to the early 1960s. It estimates UV values (at individual wavelengths and spectrally integrated) from global solar radiation, total ozone, dew point temperature, and snow cover. The model results are demonstrated to be in good agreement with the measurements. For example, the standard deviation of the difference between monthly values of measured and derived erythemally weighted UV irradiation is 3.3% for summer months. The major source of error in the model estimates is probably linked to rare occurrences of absorbing aerosols in the atmosphere. Long records of reliable measurements of total ozone, global solar radiation, and other parameters made it possible to derive UV-B values at three Canadian stations from the mid-1960s. Trends in derived erythemally weighted UV at two stations (Toronto and Edmonton) are similar to those expected from total ozone trends although the estimated error of the UV trends is more than 2 times larger. However, the increase in annual UV at Chruchill (59 ¿N) in 1979--1997 was found to be more than twice that expected from the ozone decline. This is a result of long-term changes in snow cover and clouds. ¿ 2001 American Geophysical Union