Eppley's precision spectral pyranometer (PSP) is used in networks around th
e world to measure down-welling diffuse and global solar irradiance at the
surface of the Earth. In recent years several studies have shown significan
t discrepancy between irradiances measured by pyranometers and those comput
ed by atmospheric radiative transfer models. Pyranometer measurements have
been questioned because observed diffuse irradiances sometimes are below th
eoretical minimum values for a pure molecular atmosphere, and at night the
instruments of ten produce nonzero signals ranging between +5 and -10 W m(-
2). We install thermistor sondes in the body of a PSP as well as on its inn
er dome to monitor the temperature gradients within the instrument, and we
operate a pyrgeometer (PIR) instrument side by side with the PSP. We derive
a relationship between the PSP output and thermal radiative exchange by th
e dome and the detector and a relationship between the PSP output and the P
IR thermopile output (net-IR). We determine the true PSP offset by quickly
capping the instrument at set time intervals. For a ventilated and shaded P
SP, the thermal offset can reach -15 W m(-2) under clear skies, whereas it
remains close to zero for low overcast clouds. We estimate the PSP thermal
offset by two methods: (1) using the PSP temperatures and (2) using the PIR
net-IR signal. The offset computed from the PSP temperatures yields a reli
able estimate of the true offset (+/-1 W m(-2)). The offset computed from n
et-IR is consistent with the true offset at night and under overcast skies
but predicts only part of the true range under clear skies. (C) 2001 Optica
l Society of America.