Various spherical radiative transfer (SRT) computer codes that model ultraviolet (UV) sky intensities incident at the Earth's surface are compared for clear-sky (Rayleigh scattering and ozone absorption) and hazy-sky (stratospheric volcanic aerosol, Rayleigh scattering, and ozone absorption) atmospheric conditions. Calculated sky intensities using Dave vector (pseudospherical), Dave scalar (pseudospherical), Herman's vector (full spherical), and Herman's scalar (full spherical) codes are compared for various parameters including solar zenith angle, direction of view, and wavelength. The differences in the calculated downward UV intensities are believed to arise from the differences in the code geometries and the neglect of polarization effects. The difference between downward UV intensities is within ¿15% for clear-sky conditions and between -30 and 18% for hazy-sky conditions. The results of comparisons suggest that an experiment should be conducted to measure UV sky intensities for clear-sky conditions, with minimal aerosol, to test the quality of the radiative transfer codes with actual observations. The Dave scalar radiative transfer code has been used in the past to calculate aerosol error corrections to Umkehr measurements. To evaluate the accuracy of these calculations, we performed a set of comparisons with results of various spherical radiative transfer codes. Ground level sky intensities were calculated for various solar zenith angle directions, for a vertically inhomogeneous Rayleigh atmosphere with ozone absorption, including and excluding stratospheric aerosols. For 0.11 stratospheric aerosol optical thickness, the method for calculating stratospheric aerosol errors to retrieved Umkehr ozone profile predicts either -22% or -32% error in layer 8 depending on whether vector or a scalar radiative transfer code had been used. The calculations were also used to study the effect of full spherical and pseudospherical forward model differences on Umkehr ozone profile retrievals. The difference in retrieved ozone profiles was found to be within ¿4% for clear-sky conditions and up to 13% in layer 8 for hazy-sky conditions. The results of these comparisons suggest that further improvements to the profile retrieval and stratospheric aerosol error calculations could be made using a fully spherical RT code that accounts for polarization. ¿ 2000 American Geophysical Union