UV Index from ERA5 reanalysis

Exposure to ultraviolet (UV) radiation significantly impacts human health. Consequently, comprehensive UV climatological databases are of great interest. UV exposure is evaluated by weighting UV spectra with spectral functions that describe physiological responses at each wavelength. The most widely used function is the erythemal weighting function, which is used to compute the UV Index (UVI) to assess the health risk associated with UV overexposure. The ERA5 datasets, produced by the Copernicus Climate Change Service (CDS), offer hourly ground-level UV radiation ([Formula: see text]), but do not include UVI. This study proposes a model to compute hourly UVI using exclusively ERA5 data, enabling direct access through the CDS to derive UVI statistics for locations of interest and potentially supporting the integration of a dedicated UVI product into ERA5. The model was developed using UV spectra simulated under clear-sky conditions with the uvspec radiative transfer model, accounting for atmosphere type, solar zenith angle, visibility, altitude, albedo, total ozone, and aerosol type. For these parameters, representative values typical of tropical, mid-latitude, and subarctic regions were used, effectively excluding Arctic conditions, and considering UVI values ≤ 12. The resulting formulation expresses UVI as a function of [Formula: see text], sun elevation, and total ozone. The model was validated using ground-based UVI measurements from six stations (over 17000 cases) and, in addition, compared with UVI derived from Copernicus Atmospheric Monitoring Service (CAMS) products (over 6000 cases) taken as a reference. Performance was assessed through the statistics of the differences between measured/modelled values and CAMS data under three scenarios: clear-sky conditions, varying cloud cover, and all-sky conditions. Under clear-sky conditions, the model uncertainties showed a small positive bias (≤ 0.5), with the absolute difference (AD) < 1 in 73% of the cases for ground measurements and in 86% of the cases for CAMS. The root mean squared difference (RMS) and the mean absolute deviation (MAD) were 0.9 and 0.7, respectively, for ground measurements, and 0.7 and 0.6 for CAMS. Under cloudy-sky conditions, model performance worsens significantly for CC > 0.4, with RMS and MAD reaching values of about 1.5. However, when considering relative uncertainties (percentage ratio between RMS and reference values of UVI), up to CC < 0.7 the RMS% remains below 15% for Very High-to-Extreme WHO/ICNIRP exposure categories and below 20% for Moderate-to-Extreme categories. A comparison between CAMS UVI and ground measurements was also performed, yielding results consistent with those described above. As an example, Appendix A illustrates how the model can be applied to generate daily and monthly UVI statistics over large geographical areas using only ERA5 data accessed through the CDS web portal.