Climate research studies of solar radiation instrumentation, such as those made by the Baseline Surface Radiation Network (BSRN) participants, have characterized thermal offsets in thermopile pyranometers with all-black sensors measuring diffuse or global radiation (see http://www. gewex. org/bsrn. html). Thermal offsets produce negative data at night, and lower clear-sky diffuse or global irradiances during daytime.
This systematic negative bias explains in great part the discrepancy found between measurements and predictions from sophisticated radiative transfer models (e. g., Arking 1996; Philipona 2002). Similarly, thermal offsets explain why diffuse irradiance under very clean conditions has been reported lower than what pure Rayleigh-scattering theory (with no additional atmospheric constituents) predicts (Kato 1999; Cess et al. 2000). Other investigations confirmed the importance of thermal offsets, and offered correction methods (e. g., Dutton et al. 2001) as well as improved techniques for optimal pyranometry (Michalsky et al. 1999), which are summarized in Sect. 7. Thermal offsets produce absolute errors of typically -5 to -20Wm-2 in clear-sky diffuse or global irradiance with all-black thermopile pyranometers, and are dependent on instrument installation (e. g., use of ventilators or heaters, etc.), design, deployment site, and atmospheric conditions. Current calibration methods cannot compensate directly for these errors.
For black-and-white sensors, the reference and absorbing thermopile junctions are in a similar thermal environment. These radiometers have lower (typically ±0-2Wm~2) offsets and normally produce more accurate diffuse sky measurements than all-black sensors without appropriate post-measurement corrections, or special considerations in their construction, such as compensating thermopiles.
Other Spectral Effects
Diffuse sky radiation has little energy in the shortwave near-infrared region 1000-2800 nm, while the direct beam has significant energy in that region. Therefore, nothing affecting the direct beam total irradiance between 1000 and 2800 nm, such as atmospheric watervapor, affects a shaded pyranometer signal. Consequently, for several different water vapor concentrations, and direct normal irradiances, the same shaded signal is possible from the pyranometer. By varying total precipitable water vapor from 0.5 to 3.5 cm, this “spectral mismatch” effect can be shown to result in differences of about 0.5% in Rs (Myers et al. 2004).