The Total Ozone Mapping Spectrometer (TOMS) has the longest time series of globally distributed estimates of UV irradiance at the Earth's surface. The proper interpretation of TOMS estimated irradiances relies on well-calibrated and well-maintained spectrometers at the Earth's surface. In this study, daily erythemal irradiances measured by a Brewer spectrophotometer at the island of Lampedusa (35.5¿N, 12.6¿E), in the Mediterranean are compared with TOMS observations in the period January 1998 to August 2003. The comparison, also because of the peculiar conditions at Lampedusa, a very good site for ground-based validation of satellite observations, allows us to recognize how the space-borne observations are influenced by the presence of atmospheric aerosols. Two TOMS data sets, derived applying different algorithms to retrieve ozone and UV irradiance from the backscattered radiance, are used in this study: Version 7 (V7) and the recently developed version 8 (V8), which uses new climatologies for ozone and temperature profiles and accounts for the attenuation by tropospheric aerosols through the aerosol index (AI). As shown in previous studies performed with V7 TOMS data, satellite-derived erythemal doses systematically overestimate ground-based measurements, mainly because of uncorrected absorption by aerosols in the troposphere. The bias between the TOMS and Brewer doses for all-sky conditions is (9.4 ¿ 19.8)% for V7 and (7.3 ¿ 20.0)% for V8 and decreases to (5.6 ¿ 8.0)% for V7 and (3.4 ¿ 8.4)% for V8 for the cloud-free cases. The large standard deviations for all-sky conditions are due to nonhomogeneity in the cloud cover within the sensor field of view, while those for cloud-free days are caused by the large aerosol variability occurring at Lampedusa. The biases for cloud-free days have been related to differences in the TOMS AI UV attenuation algorithm and to the aerosol optical depth (AOD) at 415.6 nm measured with a Sun photometer at Lampedusa since 2001. The mean bias between the V7 TOMS and Brewer doses progressively increases with AI and AOD at 415.6 nm, from ¿ 3% for low AI and AOD up to 21% for 1.5 ≤ AI < 2.5 and 0.5 ≤ AOD < 0.6. The bias calculated with V8 data set varies between +6% for 0 ≤ AI < 1 and about -8% for 4 ≤ AI < 5, well within the respective uncertainties of the Brewer and TOMS measurements. TOMS V8 data show a smaller dependency on the aerosol absorption, indicating that the implemented corrections produce more reliable estimated doses. For very low aerosol loading (AOD at 415.6 nm below 0.2), the TOMS-to-Brewer erythemal dose ratio, both for V7 and V8, is approximately 1, indicating that the radiometric calibration of the Brewer instrument is consistent with the TOMS estimated irradiances from derived ozone and Rayleigh scattering attenuation.