Uncertainties in modeled and measured clear‐sky surface shortwave irradiances

A comparison of five independent measurements of the clear‐sky downward shortwave irradiance at the surface shows that they scatter within a 5% range depending on their calibration constants. When the measurements are corrected using data from two cavity radiometers, three of the five independent measurements agree within 3 W m −2 over three clear‐sky days, which is well within the estimated error limit of ±1.5%. A comparison of these three sets of irradiance measurements with the computed irradiance by a δ2‐stream model reveals that the model overestimates the irradiance by 5%. Detailed investigation of the approximations and uncertainties associated with the computations (including the measurement error in the water vapor and ozone amounts, neglecting the state of polarization and trace gas absorption, the 2‐stream approximation, the neglect of the spectral dependence of the surface albedo, and the uncertainties associated with aerosols) demonstrates that the discrepancy is not due to these approximations. Further analysis of the modeled and measured irradiance shows that the discrepancy is almost entirely due to the difference between modeled and measured diffuse field irradiances. An analysis of narrow‐band diffuse to total irradiance ratios shows that this discrepancy is the largest near 400 nm and decreases with wavelength. These results rely on the absolute calibrations of two cavity radiometers, two shaded pyranometers, and one unshaded pyranometer, as well as ratios of irradiances measured by a multifilter rotating shadow‐band radiometer. Therefore, in order for instrumental error to account for the diffuse field discrepancy, three independent measurements of the diffuse field irradiance must be biased low by at least 40%. For an aerosol to account for this discrepancy, it must be highly absorbing with a single‐scattering albedo as low as 0.3. The unlikelihood of instrumental errors of 40% and aerosol single‐scattering albedos of 0.3 suggests a third possibility: the neglect of some gaseous absorption process at visible wavelengths.