Spectral distribution of UV-B irradiance derived by synthetic model compared with simulation results of TUV and ground measurements

Multifilter rotating shadowband radiometers are deployed in the United States, Canada, and New Zealand by the USDA (United States Department of Agriculture) UV-B (ultraviolet-B) Monitoring and Research Program to measure UV-B irradiances at seven discrete wavelengths. A synthetic model is used to construct the continuous spectral distribution, from which irradiance integrals can be performed for various purposes. The derived spectral data are posted for public use through a web accessible database. Although the synthetic model has been validated with a certain data set, few works have been seen to compare the results of the synthetic model with simulations of other widely accepted models such as TUV. Through this comparison the validation of the synthetic model can be further confirmed and alternative techniques for constructing spectral irradiances from discrete narrowband measurements can also be explored. In this study the data from the USDA UV-B Monitoring and Research Program are used to evaluate the synthetic model and to explore the capability of the TUV model for constructing continuous spectra from discrete measurements. Simulations of the TUV model are compared with discrete measurements, erythema-weighted broadband measurements, and the results of the synthetic model. Good agreements between derived results by using TUV model and the synthetic model with measurements in general further confirm the validation of the synthetic model. Generally, the spectral irradiances constructed by using synthetic model are lower than those by using the TUV model at very shorter wavelengths (<301 nm) and at the wavelengths of 315-342 nm, but are higher at other wavelengths. The ratio of erythemal doses derived by using the TUV simulation to broadband measurements varies between 0.87-1.02. Constructed erythemal doses by using the TUV simulation are closer to broadband measurements than those obtained by using the synthetic model. These results suggest that the TUV model may be a good alternative to accurately estimate continuous spectral distributions from discrete measurements.