The US Department of Agriculture's (USDA), UVB Radiation Monitoring and Research Program1 makes routine measurements of ultraviolet radiation at over 30 sites in the United States, Canada and New Zealand. UV measurements of total, direct and diffuse horizontal irradiances, in seven spectral channels at two nm nominal bandwidths are made with a Yankee Scientific Inc., Multiple Filter Rotating Shadow band Radiometer (UV MFRSR). A similar instrument takes measurements in the visible region with 10 nm bandwidths. The UVB group has provided, upon request, a high resolution UV product referred to as "synthetic spectra," based on application of a non-linear estimation method described in Min and Harrison2 (1998) to UV MFRSR data. This presentation examines typical problems encountered when the synthetic spectra algorithm is applied to data collected at large solar zenith angles and when the application is extended to spectral regions beyond 368 nm, the center of the longest wavelength UV MFRSR channel. In particular, the effects on derived products such as the Caldwell or Flint3 action spectra are discussed. The useful spectral region of the algorithm has been expanded by including one or more of the datum from the visible MFRSR. This extension properly constrains the derived spectrum beyond 368 nm providing especially improved Flint action values, and can be used to estimate a PAR value if extended to include the 862 nm measurement. The extent of disagreement between measurements from LICOR PAR sensors and 'synthetic PAR' values will be presented. Planning for the next version of the synthetic spectra algorithm on the new USDA UVB web site is discussed.