As a result of sky radiance anisotropy, significant errors spoil diffu
se solar irradiance data obtained by means of noncosine-true instrumen
ts, which happens in spite of knowledge of the relative cosine respons
e. To reduce these errors we recommend a method that involves an addit
ional measurement of diffuse irradiance from a horizon band. The effic
iency of the so-called double-measurement method is found to be optimu
m for a 25 degrees horizon band elevation. We confirm the feasibility
of the method with relevance tests that use diffuse solar spectral mea
surements carried out on cloudless skies by means of an UV spectroradi
ometer. (C) 1996 Optical Society of America