Diffuse irradiance, G(d), is an important variable in solar resource a
ssessment. The diffuse irradiance can be worked out from global, G, an
d direct, G(b), irradiance measurements, but this method involves the
use of relatively expensive tracking mechanisms. Alternatively, a wide
ly accepted technique uses a pyranometer with a shadowband. Because th
e shadowband screens the sensor from part of the diffuse radiation com
ing in from the sky, a correction must be made to the measurements. Ho
wever, because of the anisotropy of diffuse radiation it is difficult
to compute an exact theoretical correction. In this study we use two d
ata sets registered in two locations in Spain. The first one consists
in coincident hourly values of global, direct, and diffuse irradiance;
the latter by means of shadowband. The other data set includes the sa
me variables but as 5-minute values. Our goal is to study the necessar
y correction factor for diffuse irradiance measurements obtained by me
ans of shadowband. After testing several well-known correction methods
, we have developed two different correction models, using two-thirds
of the hourly data set, while the remaining one-third and the whole 5-
minute data set have been used for validation purposes. The last valid
ation test suggests that our anisotropic models provide reliable corre
ctions for conditions different than the ones where they have been dev
eloped. The results obtained by the developed models show a negligible
mean bias deviation. Approximately 55% of cases present deviations lo
wer than 5% over the mean value of diffuse irradiance.