Ultraviolet radiation in the spectral region between 280 and 315 nm (often referred to as UV-B) is harmful to living organisms. Satellite-based estimation of surface UV-B supplements the sparsely distributed ground-based UV-B monitoring networks. This study is concerned with validation of an inversion algorithm <Li et al., this issue> for retrieving spectrally integrated UV-B (no spectral weighting) and erythemal UV (EUV) (with spectral weighting) fluxes at the surface from satellite. The physical inversion algorithm contains a few analytical expressions and input parameters: the solar zenith angle, ozone amount, albedo at the top of the atmosphere (TOA), and aerosol variables. The algorithm is applied to satellite measurements of total ozone amount and 360 nm reflectance from Meteor 3/TOMS and visible reflectance from NOAA/AVHRR. The retrieved UV-B and EUV fluxes are compared with ground UV observations made at six Canadian UV observation stations with Brewer instruments from 1992 to 1994. Under all-sky conditions the comparisons showed very small mean differences and relatively large standard deviations (s.d.): 0.033 W/m2 (mean) and 0.287 W/m2 (s.d.) for total UV-B and 3.02 mW/m2 (mean) and 12.0 mW/m2 (s.d.) for EUV radiation. The large standard deviations are attributed to the inhomogeneity in sky condition and mobility of cloudy scenes, which renders an inaccurate match between satellite and surface measurements. The comparisons under clear-sky conditions showed very small mean and standard differences. By means of a running average over a period of time, satellite inversion can track the variation of surface-observed UV-B and EUV very well. ¿ 2000 American Geophysical Union