Cloud flags can be used as identifiers to classify measured spectra of
UV irradiance as to the conditions of cloudiness during the time of t
he spectral measurement. That classification is helpful for quality ch
ecks of spectra as well as for their further analysis and comparisons
with model calculations. A method has been developed to derive cloud f
lags for UV spectra. It uses readily available ancillary measurements
such as direct (or global) and diffuse radiation, relative sunshine du
ration and hourly observations of cloud cover and visibility as input
data. Application of the cloud flagging algorithm to 16,217 UV spectra
measured with Brewer spectroradiometers MW and MKIII at the station P
otsdam in the years 1995 and 1996 classified 807 UV spectra (5%) as 'c
lear sky', with almost all of the remaining data, i.e., 15,384 (95%) b
eing classified as 'cloudy', and-due to missing part of corresponding
ancillary data-26 spectra classified as 'unknown' with referring to cl
oud conditions. Among the combinations of ancillary data tested, both
direct/diffuse ratios and sky brightness turned out to be the most use
ful parameters for cloud detection including Cirrus clouds. They can a
lso be used to select those spectra distorted by moving clouds. In add
ition to the simple 'clear sky vs, cloudy' classification, direct/diff
use ratios together with sky brigthness have been tentatively used as
potential classifiers to separate five different conditions of cloudin
ess as to their optical effect. The cloud flagging algorithm is neithe
r confined to UV spectra nor to meteorological data of a specific site
, but it can be applied to any meteorological data at any other site.
(C) 1998 Elsevier Science B.V. All rights reserved.