177
Radiation Protection Dosimetry
Vol. 91, Nos 1–3, pp. 177–180 (2000)
Nuclear Technology Publishing
A PRELIMINARY ESTIMATION OF THE DIRECT
ULTRAVIOLET SPECTRAL IRRADIANCE IN VALENCIA (SPAIN):
COMPARISON WITH MEASURED VALUES
F. Tena, R. Pedro
´
s, L. Herna
´
ndez, M. P. Utrillas and J. A. Martı
´
nez-Lozano
Departamento de Termodina
´
mica, Universitat de Vale
`
ncia
46006 Burjassot (Valencia), Spain
Abstract — UVSPEC and SMARTS2 models have been used to estimate the UV spectral irradiance values in Valencia, Spain,
and the results of these estimations have been compared with experimental measurements of direct irradiance at normal incidence
obtained using an Optronic OL 754 in the 300–400 nm range. The relative RMSD and MBD parameters have been used to
estimate the deviations of the modelled values with respect to the experimental values. The results indicate that the deviations
are smaller when the SMARTS2 model is used with the appropriate aerosol coefficients.
INTRODUCTION
The detection of the thinning of the ozone layer lead-
ing to the appearance of an ozone hole in the Antarctic
in 1985 and the progressive ozone reduction in the fol-
lowing years
(1,2)
stimulated the scientific community to
investigate the consequences of this phenomenon,
especially the increase in incident ultraviolet radiation
at ground level. Over the past few years major efforts
have been made to assess the trends in the increase of
the UV radiation and the influence that different atmos-
pheric parameters exert on the transmission of UV solar
radiation through the atmosphere.
In order to carry out such analyses spectroradiometry
can be a useful tool for studying the UV radiation
characteristics at ground level, especially in relation to
the biological effects of increased UV radiation which
are known to be strongly wavelength dependent.
Unfortunately most spectroradiometers are too expens-
ive or have poor accuracy in the UV range and conse-
quently experimental solar irradiance spectral data at
ground level are scarce. To remedy this a number of
atmospheric algorithms have been described in the
literature dating back to the early 1980s which try to
simulate and predict irradiance values
(3)
.Such
algorithms can be classified into two different groups:
sophisticated rigorous codes (or radiative transfer
codes, RTC) and simple transmittance parameterisa-
tions (parameterised models, PM). All these models are
subject to several uncertainties — usually divided into
numerical errors and uncertainties in the input data —
which must be quantified by comparison with experi-
mental data.
In this paper the estimated values produced by two of
these models are compared with the experimental results
determined from spectral irradiance measurements at
Valencia, a Spanish Mediterranean coastal site. The
experimental data have been obtained by means of an
Contact author E-mail: tenaf얀uv.es
Optronic OL 754 spectroradiometer with a high resol-
ution in the UV spectral range. The models employed
were the UVSPEC (RTC)
(4)
and the SMARTS2 (PM)
(5)
which are summarised below. To evaluate the deviation
of the values computed by the models from the experi-
mental data, the relative RMSD and MBD statistical
parameters have been used.
The work presented here should only be considered
as a preliminary analysis of the use of the two models
in the UV range, the work chiefly considers the influ-
ence of the choice of the aerosol model on their pre-
cision. It is not an exhaustive analysis of the models’
validity since this would require a larger spectral data
base covering a wide variety of sky conditions
(6)
.
THE UVSPEC AND SMARTS2 MODELS
The UVSPEC code was developed by Kylling in
the University of Alaska and is available on the
World Wide Web
(4)
. The UVSPEC code covers the
wavelength region 176–850 nm and evaluates both
direct and diffuse irradiances with 1 nm resolution. The
radiative transfer equation is numerically integrated with
DISORT
(7)
, SDISORT (Pseudo-spherical or Spherical
DISORT version)
(8)
or TWOSTR (TWO-Stream
code)
(9)
. Mayer et al
(6)
have proposed a modification of
the original UVSPEC, the TWODIS, which combines
the benefits of TWOSTR and DISORT. The UVSPEC
uses the standard atmosphere profile USSA and the
aerosol models of Shettle
(10)
. The UVSPEC allows the
surface albedo, the cloud type, the ozone concentration,
and other conditions to be varied. The different atmos-
phere types and the temperature, pressure and ozone
profiles can also be modified if the corresponding data
are available.
The SMARTS2 model (Simple Model for the Atmos-
pheric Radiative Transfer of Sunshine) was proposed by
Gueymard in 1995
(5)
. It is based on an extensive
revision of the algorithms used to calculate direct beam
radiation with the spectral model SMARTS
(11)
and con-
sists of a separate parameterisation of the different
Article in press
F. TENA, R. PEDRO
´
S, L. HERNA
´
NDEZ, M. P. UTRILLAS and J. A. MARTI
´
NEZ-LOZANO
178
extinction processes ocurring in the atmosphere. To
make this parameterisation, accurate transmittance
functions for all atmospheric extinction processes are
introduced and the effects of temperature and
humidity are considered. The model also includes
very accurate absorption coefficients derived from
spectroscopic data. The version used in this work was
kindly provided by its author and corresponds to the
latest revised version (Gueymard, 1997, personal
communication). SMARTS2 code allows the introduc-
tion of ground meteorological data or the choice of ten
different reference atmospheres if ground data are not
available. It also allows the user to choose from nine
predefined aerosol models or to introduce their own
aerosol coefficients. The aerosol models available in the
present version are: (a) four proposed by Shettle and
Fenn
(12)
; (b) two proposed by Braslau and Dave
(13)
; and
(c) three corresponding to the Standard Radiation
Atmosphere (SRA)
(14)
.
Both models allow the introduction of the spectrora-
diometer slit function in order to compare the esti-
mations with experimental measurements.
INSTRUMENTATION AND METHODOLOGY
Spectral solar irradiance measurements were obtained
using two commercial spectroradiometers, an Optronic
754 and a Licor Li-1800. The spectral band of the OL
754 ranges from 250 nm to 800 nm, with a 2 nm band-
width. The measurements can be made with minimum
pass bands of 0.05 nm. The OL 754 is a double mono-
chromator with holographic gratings of 1200 lines.
mm
⫺1
. The input optics is an integration sphere covered
by PTFE and the detector is a silicon photodiode with
a photomultiplier with temperature stabilisation. The
spectral band of the Li-1800 ranges from 300 nm to
1100 nm, with a 6 nm bandwidth. The characteristics of
this instrument are detailed elsewhere
(15)
. The spectro-
radiometers are absolutely calibrated every three months
in our laboratory by means of 200 W reference lamps,
the Optronic with a 754 OL 752-10E (Optronic), and
the Li-1800 with a 1800/ORL815 (Licor).
The OL 754 has been used to obtain direct spectral
irradiance measurements at normal incidence in the UV
range, 280–400 nm. The Li-1800 has been used to
determine the aerosol characteristics by means of direct
spectral irradiance measurements in the visible (400–
670 nm). These aerosol characteristics were then intro-
duced into the models in order better to approximate the
actual atmospheric conditions. For the direct sun
measurements the spectroradiometers were oriented
manually by means of a three-axis ball and socket joint
and alignment system. For the direct irradiance
measurements, radiance limiting tubes (collimators)
with a field of view of 4.7° were used. Simultaneously
with the spectral irradiance measurements, the total
atmospheric content of ozone and water vapour were
also measured with a Microtops II. Other parameters
such as temperature and pressure were routinely
recorded.
All the measurements corresponded to clear days and
were taken on the terrace (40 m above sea level) of the
Faculty of Physics, Valencia University, in the Burjassot
Campus, located in the outskirts of the city of Valencia
(Spain). The latitude is 39.5° N and the longitude is 0°.
The obstructions above the horizon were less than 4°,
except in a small zone in the Northwest. A previous
paper describes site obstructions and the measuring set-
up in detail
(16)
.
From the spectral measurements at normal incidence
in the 400–670 nm range obtained with the Licor, values
of the total optical thickness for all the different atmos-
pheric extinction processes,
T
were deduced by using
the Beer law. Once the total atmospheric optical thick-
ness had been determined, the value of the aerosol
optical thickness,
a
, was obtained by removing the
contributions due to Rayleigh scattering and to absorp-
tion by the other atmospheric components from the total
transmitance
(17)
. Applying the power law relationship
proposed by Angstrom
(18,19)
a
= 
−␣
␣ and  were evaluated and used as inputs in the spec-
tral irradiance models
(20)
.
In order to analyse the deviation of the values esti-
mated by both models from the experimental values,
two different statistical estimators which are commonly
used for evaluating the accuracy of models have been
used: RMSD (root mean square deviation) and MBD
(mean bias deviation). RMSD always gives positive
values whilst MBD, defined as modelled minus
observed, may be positive or negative, with the positive
values corresponding to overestimation by the model.
RESULTS AND DISCUSSION
Our analysis of the results was limited to establishing
general comparisons between the spectral values of the
experimental direct irradiance and those obtained by the
models. The UVSPEC and SMARTS2 models were
applied to evaluate the direct irradiance at normal inci-
dence on four clear days in July 1999. Forty two spec-
tral series, corresponding to solar zenith angles varying
between 20° and 60° have been compared. For each of
these measurements the ␣ and  Angstrom coefficients
have been evaluated from the irradiance values in the
visible range, following the methodology described
above. The spectral irradiance values have been com-
pared in the range 300–400 nm with a 1 nm pass band.
For the estimation by the SMARTS2 model, the
Angstrom turbidity coefficient  and three aerosol
models, rural (SFR), urban (SFU) of Shettle and
Fenn
(12)
and urban (SRAU) of SRA
(14)
have been used.
These models were selected because they produced the
best results in a previous work carried out in
Valencia
(17)
. As an example, Figure 1 shows the experi-
ESTIMATION OF UV SPECTRAL IRRADIANCE IN VALENCIA
179
mental and the estimated values obtained by employing
these three aerosol models for a solar zenith angle of
20° on 15 July. The representative curves show that the
model that best fits with the experimental values was
the aerosol urban model. The MBD values of the three
curves were 3.9% (SFU), 9.5% (SFR) and 12.7%
(SRAU). The corresponding values of the RSMD were
8.7% (SFU), 11.8 (SFR) and 14.2% (SRAU).
The UVSPEC model takes as input the Angstrom ␣
and  coefficients and an aerosol model. The Shettle
urban, rural and marine models have been used.
Figure 2 presents the results obtained for the same day
and for the same solar zenith angle considered for the
SMARTS2. As can be observed, the curves correspond-
ing to the values deduced by employing the different
aerosol models practically coincide and the MBD
(6.3%) and the RSMD (15.5%) were practically the
same for the three models considered, varying only in
the second decimal digit between one aerosol model and
another. The same happened with all the spectral
measurements analysed. Indeed, it appeared that, for the
version used (libRadtran 0.14), the UVSPEC was almost
completely insensitive to the choice of aerosol model
when the ␣ and  Angstrom coefficients were used as
input parameters. This observation, however, should be
more carefully analysed in the future based on a higher
number of experimental measurements.
For the whole set of available measurements the aver-
1.00
0.80
0.60
0.40
0.20
0
Experimental
SFR
SFU
SRA
400380360340320300
l (nm)
Irradiance (W.m
2
.nm
1
)
Figure 1. Comparison between the experimental data and the
values estimated by the SMARTS2 code for a solar zenith
angle of 20° on 15 July 1999.
REFERENCES
1. Bojkov, R. D. and Fioletov, V. E. Estimating the Global Ozone Characteristics during the Last 30 Years J. Geophys. Res.
100, 16537–16551 (1995).
2. Stolarsky, A. S., Bojkov, R. D., Bishop, L., Zerefos, C., Stachelin, J. and Zawodny, J. Measured Trends in Stratospheric
Ozone. Science 256, 343–349 (1992).
3. Koepke, P. and 23 others. Comparison of Models used for UV Index Calculations. Photochem. Photobiol. 67, 657–662 (1998).
4. Kylling, A. UVSPEC. A Program for Calculation of Diffuse and Direct UV and Visible Fluxes and Intensities at Any Altitude.
Available on ⬍ftp://kaja.gi.alaska.edu/pub/arve⬎ (1995).
5. Gueymard, C. SMARTS2, a Simple Model of the Atmospheric Radiative Transfer of Sunshine: Algorithms and Performance
Assessment. FSEC-PF-270–95 (Florida Solar Energy Center, Cocoa, FL) (1995).
age values of the deviations were the following: (a) for
the UVSPEC the value of the MBD was 7.4 ⫾ 0.1%,
depending on the aerosol model used, and the RSMD
was 15.7%, independent of the aerosol model; (b) the
SMARTS2 was more sensitive to the aerosol model,
with values of 5.6% (MBD) and 9.4% (RMSD) obtained
for the SFU model, 11.5% (MBD) and 13.3% (RMSD)
for the SFR model and 14.8% (MBD) and 15.9%
(RMSD) for the SRAU model.
As was noted at the beginning of the paper, the results
presented here must be interpreted with caution because
they are only a first approximation to the use of the
models and not an analysis of their validity. Neverthe-
less it is possible to point out three qualitative character-
istics of the models used: (a) SMARTS2, with the
appropriate aerosol model, produces estimates nearer to
the experimental measurements than UVSPEC; (b) the
version of the UVSPEC used, with the Angstrom coef-
ficients as input parameters, seems to be insensitive to
the choice of aerosol model; and (c) both SMARTS2
and UVSPEC systematically underestimate the direct
irradiance values.
ACKNOWLEDGEMENTS
This work was supported by the Ministerio de Educa-
cio
´
n (CLI97–0345–C05) and the Generalitat Valenciana
(GV97–RN–14–18).
1.00
0.80
0.60
0.40
0.20
0
Irradiance (W.m
2
.nm
1
)
400380360340320300
l (nm)
Experimental
Rural
Figure 2. Comparison between the experimental data and the
estimated values by the UVSPEC code for a solar zenith angle
of 20° on 15 July 1999.
F. TENA, R. PEDRO
´
S, L. HERNA
´
NDEZ, M. P. UTRILLAS and J. A. MARTI
´
NEZ-LOZANO
180
6. Mayer, B., Seckmeyer, G. and Kylling, A. Systematic Long-term Comparison of Spectral UV Measurements and UVSPEC
Modeling Results. J. Geophys. Res. 102, 8755–8767 (1997).
7. Stamnes, K., Tsay, S., Wiscombe, W. and Jayaweera, K. Numerically Stable Algorithm for Discrete-ordinate-method Radiat-
ive Transfer in Multiple Scattering and Emiting Layered Media. Appl. Opt. 27, 2502–2509 (1988).
8. Dahlback, A. and Stamnes, K. A New Spherical Model for Computing the Radiation Field Available for Photolysis and
Heating at Twilight. Planet. Space Sci. 39, 671–683 (1991).
9. Kylling, A., Seckmeyer, G. and Mayer, B. A Reliable and Efficient Two-stream Algorithm for Spherical Radiative Transfer:
Documentation of Accuracy in Realistic Layered Media. J. Atmos. Chem. 21, 115–150 (1995).
10. Shettle, E. P. Models of Aerosol, Clouds and Precipitation for Atmospheric Propagation Studies. In: Atmospheric Propagation
in the UV, Visible, IR and mm-region and Related System Aspects. AGARD Conf. Proc. 15, 1–15, 13 (1989).
11. Gueymard, C. Development and Performance Assessment of a Clear Sky Spectral Radiation Model. In: Proc. Solar ‘93–
22nd ASES Conf., Washington D.C. (1993).
12. Shettle, E. P. and Fenn, R. W. Models for the Aerosol of the Lower Atmosphere and the Effects of Humidity Variations on
their Optical Properties. AFGL-TR-79–0214 Environmental Res. Paper no 676. Hanscom, AFL (Mass) (1979).
13. Braslau, N. and Dave, J.V. Effect of Aerosols on the Transfer of Solar Energy through Realistic Model Atmospheres.J.
Appl. Meteorol. 30, 601–619 (1973).
14. WMO. A Preliminary Cloudless Standard Atmosphere for Radiation Computation. World Climate Programme, WCP-112.
WMO/TD-No 24 (Geneva, Switzerland: World Meteorological Organization, (1986).
15. Cachorro, V. E., Utrillas, M. P., Martinez-Lozano, J. A. and De Frutos, A. A Preliminary Assessment of a Detailed Two
Stream Short-wave Narrow-band Model using Spectral Radiation Measurements Solar Energy 60, 265–273 (1997).
16. Utrillas, M.P., Martinez-Lozano, J.A. and Casanovas, A.J. Evaluation of models for estimating solar irradiation on vertical
surfaces at Valencia, Spain Solar Energy 47, 223–229 (1991).
17. Utrillas, M. P., Bosca
´
, J. V., Martı
´
nez-Lozano, J. A., Can
˜
ada, J., Tena, F. and Pinazo, J.M. A comparative study of SPEC-
TRAL2 and SMARTS2 parametrized models based on spectral irradiance measurements at Valencia (Spain) Sol. Energ. 63,
161–171 (1998).
18. Angstrom, A. On the Atmospheric Transmission of Sun Radiation and on the Dust in the Air Geografis. Annal. 11, 156–
166 (1929).
19. Angstrom, A. On the Atmospheric Transmission of Sun Radiation. Geografis. Annal. 12, 130–159 (1930).
20. Martı
´
nez-Lozano, J.A., Utrillas, M.P., Tena, F. and Cachorro, V. The Parametrization of the Atmospheric Aerosol Optical
Depth using the Angstrom Power Law. Sol. Energ. 63, 303–311 (1998).