Different aspects of UV radiation were investigated. The analyses incorporate measurements of surface UV irradiance and UV actinic flux and various calculation methods for these quantities. Measurements of erythemally weighted broadband UV irradiance were performed with UV Biometers. More elaborate measurements of spectral irradiance and spectral actinic flux between 290 and 500 nm were performed with Bentham spectroradiometers in three measurement compaigns. The measured data were used to evaluate existing state of the art calculation methods, improve new calculation methods (3-D-radiative transfer model GRIMALDI) and to reconstruct erythemal UV-doses for the past with no available measurement data. It was found, that 1-D radiative transfer models like DISORT are very accurate and able to reproduce the ground-based measurements even in mountainous terrain with an uncertainty close to the measurement uncertainty. However, some typical effects of radiation in montainous terrain (shading effects, multiple reflection on mountain slopes) are not covered by these models. Calculations with the basically modified 3-D-radiative transfer model GRIMALDI reproduced such 3-D effects. In addition to 1-D and 3-D radiative transfer models, UV irradiance retrieved from space are compared with measurements. Surface UV irradiance retrieved from measurements of the ozone monitoring instrument (OMI) onboard the NASA EOS Aura Spacecraft is difficult to compare with ground-based measurements. A crucial issue is how representative a satellite pixel is. Therefore the variability of ground UV irradiance was estimated within the area of one satellite pixel. It was found that in relatively flat terrain and under clear-sky conditions, the deviations of erythemal UV irradiance inside one satellite pixel are within the measurements uncertainty. Thus, satellites can in principle determine ground UV irradiance in such situations, but so far OMI retrieved values are significantly lower than the measurements due to problems with cloud correction. Apart from this issue (not cloud corrected values are also available), the accuracy is only slightly worse in comparison with 1-D and 3-D simulations. The main source of uncertainty is related to the coarse resolution of OMI, it is not possible to resolve very inhomogeneous terrain in the Austrian Alps.