Effect of Opaque and Thin Clouds on Broadband UVB Irradiance Using
One-Minute Multi-Sensor and Objective Cloud Cover Measurements
C. A. Gueymard
Solar Consulting Services, Colebrook, New Hampshire, USA
Introduction
Global irradiance in the UVB is a strong function of zenith
angle, total ozone amount, and cloudiness. The latter is
highly variable in time, space, and optical characteristics.
At nearly all ground sites, the available cloud data are subjectively estimated by observers only once every hour or
three hours, typically. In this contribution, objective
ground measurements of both thick (opaque) and thin
(translucent) cloudiness are used, resulting in detailed, high
spatial resolution (1%) and high-frequency (1 min) data.
Broadband UV radiometers are very useful because of
their lower cost and complexity than spectroradiometers.
However, they have problems of their own, particularly
regarding spectral responsivity, ozone-dependent calibration, and cosine response. In this study, five collocated
UVB instruments of different characteristics are used to
compare their response to variable cloudiness.
with tracking shade for diffuse irradiance (UVB-1 type).
Only global UV data are used in this report. The instruments under scrutiny are shown in Figs. 1 and 2, in company of some of the numerous other radiometers deployed
at SRRL. The spectral response of the five UVB sensors
varies widely, as evidenced in Fig. 3, which compares generic (and coarse) manufacturer’s data. All broadband UV
radiometers are regularly calibrated the same way, i.e.,
outdoors near solar noon against an Optronics OL-754.
Figure 2. Various radiometers at SRRL, including the SL-501A
and UVB-1 (first row), and CUVB and MS-210W (top row).
1
UVB Instruments
Normalized Response
Abstract The response of five UVB instruments to variable cloudiness, ozone and zenith angle is investigated. It
is found that the differing instruments’ optical characteristics explain most of the observed output differences. On
average, their response under thick cloud is similar, but
diverge slightly under thin cloud cover greater than 20%.
0.8
MS-210W
UVS B-T
CUVB
SL-501A
UVB-1
0.6
0.4
0.2
0
280
290
300
310
320
Wavelength (nm)
330
340
Figure 3. Spectral response of the five instruments used here.
Figure 1. UVS A-T and UVS B-T instruments at SRRL. Note
the photochemical brown cloud over Denver, to the East.
Experimental Setup
This study uses data from the Solar Radiation Research
Lab (SRRL) of the National Renewable Energy Laboratory
(NREL) in Golden, Colorado. This site is located on a
mesa (1829 m) overlooking the agglomeration of Denver
(Fig. 1). This sunny climate is dry with low turbidity, but
the abundant UV frequently triggers photochemical smog
(“brown cloud”) episodes over Denver (Fig. 1).
The UV irradiances reported here, like all other radiometric or meteorological variables at SRRL, are acquired at
1-min intervals from 4-sec samples. The UVB instrumentation consists of five sensors measuring global irradiance
(Eko’s MS-210W; Kipp & Zonen’s UVS-B-T and CUVB;
Solar Light’s 501A; and Yankee’s UVB-1), one tracking
sensor for direct irradiance (CUVB type), and one sensor
Total ozone is not monitored on site, but is available from
a Dobson instrument located at Boulder, 37 km north of
SRRL. Cloudiness is measured on site with a Yankee Total
Sky Imager (TSI), which provides total, opaque and thin
cloud cover on a scale of 100. A typical combination of
thick and thin clouds, as sensed by the TSI, is illustrated in
Fig. 4. The dataset assembled here covers the 6-month
period January–June 2006.
Results and Discussion
After quality control and elimination of low-sun (< 7°) and
snow-on-ground conditions, the resulting experimental
dataset consists of 107,867 data points. Owing to this large
number, the clear-sky UV irradiance was not evaluated
through radiative transfer modeling. Rather, the function
relating irradiance to zenith angle was first determined
empirically from all cloudless data points. A more detailed
function, incorporating ozone and aerosol optical depth
(AOD), is desirable and will be investigated in a subse-
quent study. The combined effect of cloud cover and zenith angle on the relative outputs of the five UVB instruments has been studied for a high-ozone day (405 DU, Fig.
5) and a low-ozone day (232 DU, Fig. 6). It appears that
most of the differences in relative output can be explained
by the differing optical characteristics of the instruments
(spectral response and cosine response, mainly). This is
particularly evident for the CUVB, which has a very narrow spectral response (Fig. 3).
Figure 4. Processed image from the TSI for 03 June 2006, 1030
LST. The opaque and thin cloud covers were 26% and 38%, resp.
instrument, using polynomial fits over the whole dataset. It
is obvious that the effects of thin and thick clouds (or their
combination) are widely different. The average transmittance under dense overcast is !45% at this site. Remarkably, all instruments behave similarly, with only a diverging
trend for thin cloud fractions above !20%, at which point
the mean cloud transmittance also increases with increasing cloud cover. A more detailed analysis reveals that, on
average, thin cloud fractions in the range 15–25% are associated with significantly larger thick cloud fractions
(!50%) than either below 15% or above 25%.
Over the whole dataset with cloud fraction >10%, the irradiance ratios relative to the SL-510A are 1.078, 1.134,
0.966, and 1.012 for the MS-210W, UVS B-T, CUVB, and
UVB-1 instruments, respectively. Comparatively, the same
ratios are 1.104, 1.127, 0.997, and 1.017 under clear-sky
conditions, indicating a smaller relative effect from
cloudiness than from the sensors’ optical characteristics.
Figure 5. Irradiance ratios relative to the SL-501A instrument
for a high-ozone morning with low cloud cover.
Figure 6. Irradiance ratios relative to the SL-501A instrument for
a low-ozone morning with high cloud cover.
For each data point and each instrument, the measured
irradiance normalized by the estimated clear-sky irradiance
represents a proxy for the spatially-integrated cloud-field
transmittance. A part of the observed scatter in this ratio is
caused by unaccounted-for variations in ozone and AOD,
but these tend to cancel out on the average due to the large
number of data points from various days.
Figure 7 shows the effect of total, opaque and thin cloud
cover on the normalized irradiance. Thick lines represent
the average cloud-field transmittance as sensed by each
Figure 7. Mean variation in UVB irradiance sensed by five
radiometers, as a function of total cloud cover (top plot), opaque
cover (middle plot) and thin cover (bottom plot).
Acknowledgments The information and data provided by Daryl
Myers and Tom Stoffel of NREL were essential to this study.