You are on page 1of 8

This article was downloaded by: [University of Chicago Library]

On: 16 November 2014, At: 20:12


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Sustainable


Energy
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gsol20

Fitting Weibull distribution to


ultraviolet solar radiation data
a b
M. E. Ghitany & N. F. El-Nashar
a
Department of Statistics and Operations Research, Faculty of
Science , P.O. Box 5969, Safat, 13060, Kuwait
b
Applied Sciences Department, College of Technological Studies ,
P.O. Box 42325, Shuwaikh, 70654, Kuwait
Published online: 25 Feb 2007.

To cite this article: M. E. Ghitany & N. F. El-Nashar (2005) Fitting Weibull distribution to
ultraviolet solar radiation data, International Journal of Sustainable Energy, 24:4, 167-173, DOI:
10.1080/14786450500096761

To link to this article: http://dx.doi.org/10.1080/14786450500096761

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
International Journal of Sustainable Energy
Vol. 24, No. 4, December 2005, 167–173

Fitting Weibull distribution to ultraviolet solar radiation data


M. E. GHITANY† and N. F. EL-NASHAR*‡
Downloaded by [University of Chicago Library] at 20:12 16 November 2014

†Department of Statistics and Operations Research, Faculty of Science, P.O. Box 5969,
Safat 13060, Kuwait
‡Applied Sciences Department, College of Technological Studies, P.O. Box 42325,
Shuwaikh 70654, Kuwait

The aim of this paper is to fit the Weibull probability distribution to recently published data on monthly
average daily ultraviolet (UV) radiation over Kuwait. The goodness-of-fit of the Weibull model is
demonstrated both graphically and via several statistical tests such as modified Anderson–Darling,
Cramer–von Mises, and Kolmogorov–Smirnov tests. As a result, we can estimate statistical measures
associated with monthly average daily UV radiation data such as the mean, median, mode, and standard
deviation of the monthly average daily UV radiation over Kuwait.

Keywords: Ultraviolet radiation; Weibull distribution; Goodness-of-fit tests; Kuwait

1. Introduction

Ultraviolet (UV) solar radiation is the part of the electromagnetic spectrum having wavelengths
200–400 nm. The biomedical literature divided it into UV-A wave band (320–400 nm), which
consists mostly of UV solar radiation received at earth surface, UV-B wave band (290–320 nm),
which is partially absorbed or scattered in the atmosphere, and UV-C wave band (200–290 nm),
which is completely absorbed by stratospheric ozone.
Small traces of UV-B manage to reach the earth surface and excessive exposure to it
can increase the incidence of slow skin aging, skin cancer, cataract, immune deficiencies,
degradation of materials, and also harm crops.
The UV-B reaches the ground depending on many factors such as solar zenith angle, cloud
cover, humidity, dust or sand due to sandstorms, chemical pollutants, and the atmospheric
ozone.
The knowledge of UV solar radiation data and its distribution is vital for our environ-
ment. There are many studies at ground level in many countries. The Gulf area is a prime
candidate for this research from the health point of view. The UV solar radiation monitored
and analysed in Saudi Arabia by El Hadidy et al. (1990) and Khogali and Al-Bar (1992), in

*Corresponding author. Email: nnashar@paaet.edu.kw

International Journal of Sustainable Energy


ISSN 1478-6451 print/ISSN 1478-646X online © 2005 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/14786450500096761
168 M. E. Ghitany and N. F. El-Nashar

Bahrain by Som (1992), and in Kuwait by Al-Aruri et al. (1988), Al-Aruri (1990), Al-Aruri
and Amer (1993), Kollias and Baqer (1984), Kollias et al. (1988), El-Nashar et al. (2001) and
Al-Awadhi and El-Nashar (2002).
The study done by Blumthaler and Ambach (1990) has shown an increase in UV-B radiation
flux about 1% per year at the Jungfraujoch Station (Swiss Alps) since 1981. The deple-
tion of stratospheric ozone increases the UV-B part of the spectrum as shown by Kerr and
McElory (1993).
The effect of the chemical pollutants (e.g., SO2 , NO2 ) was reported by Schwander
et al. (1997), where the estimated percentage increase of the global irradiance due to the
decrease in SO2 was about 3% for wavelength 290 nm and 5–6% for 300 nm. Dvorkin and
Steinberger (1999) also found an appreciable reduction in the relevant wavelengths due to the
Downloaded by [University of Chicago Library] at 20:12 16 November 2014

presence of SO2 and NO2 . The conclusion of their work was that SO2 has strong absorption
peak at 300 nm, where there is a reduction of 13–14% in the flux. At 365 nm, near the NO2
absorption maximum, the flux reduced by 11%. The analysis of monthly average daily UV
solar radiation data of Kuwait from January 1996 to December 1998 by El-Nashar et al. (2001)
showed that there is a decrease in the amount of UV solar radiation from 270–240 Wh/m2
during the period of study. The explanation of this is accounted for in the increase in the mean
concentration of chemical pollutants SO2 (10%, 24%) and NO2 (30%, 28%). That is, the
decrease in the UV solar radiation as a result of increasing chemical pollutants is a dilemma.
The aim of this paper is to fit Weibull probability distribution to the monthly average daily
UV solar radiation data of Kuwait. Here the distribution’s parameters are unknown and are
estimated from the data. A natural approach is to use some method of estimating the parameters
and then apply formal goodness-of-fit tests such as modified Anderson–Darling, Cramer–von
Mises, and Kolmogorov–Smirnov tests.

2. Fitting monthly average daily UV data

Table 1 shows the monthly average daily UV solar radiation (kWh/m2 ) measurements of
Kuwait during 1996–1998, recently reported by El-Nashar et al. (2001).
It is natural to ask whether there is a parametric probability distribution to fit the population
from which the data arise and also, how well the sample data agrees with the fitted distribution
as its population.

Table 1. Monthly average daily UV solar


radiation (kWh/m2 ) (1996–1998).

Year
Month 1996 1997 1998

January 0.154 0.162 0.161


February 0.196 0.233 0.210
March 0.254 0.253 0.227
April 0.323 0.295 0.306
May 0.354 0.313 0.329
June 0.397 0.337 0.337
July 0.356 0.318 0.327
August 0.328 0.310 0.275
September 0.288 0.271 0.242
October 0.254 0.204 0.203
November 0.179 0.177 0.147
December 0.161 0.133 0.129
Weibull distribution to UV solar radiation data 169

In the following, we fit the Weibull distribution with scale parameter a and shape parameter
b and probability density function (p.d.f.)
   
b  x b−1 x b
f (x) = exp − , x > 0, a, b > 0 (1)
a a a

to the monthly average daily UV data.


As the values of the parameters a, b are unknown, they must be estimated from the given
data by means of some statistical procedure such as the method of moments or the method of
maximum likelihood. The latter method produces estimates having desirable properties, e.g.
Downloaded by [University of Chicago Library] at 20:12 16 November 2014

consistency and asymptotic normality. For our data, the maximum likelihood estimates of the
parameters a and b are given by â = 0.2812 and b̂ = 3.9844, respectively (see Appendix A).
In the following we explore graphical assessment and formal goodness-of-fit tests to
assess the goodness-of-fit of the fitted Weibull distribution to the monthly average daily UV
data.

2.1 Graphical assessment

Here we plot the estimated cumulative distribution function (c.d.f.) of the Weibull distribu-
tion, i.e.
   
x b̂
F̂ (x) = 1 − exp − , x > 0, â, b̂ > 0 (2)

against the empirical c.d.f

1
Fn (x) = (number of observations ≤ x) (3)
n

on the same graph.


Figure 1 shows that the fitted Weibull and empirical distributions are close enough to each
other to conclude that the fit is good.

2.2 Formal goodness-of-fit tests

We apply three well-known tests for goodness-of-fit for the Weibull distribution when the
parameters are unknown and must be estimated from the data. These are modified Anderson–
Darling, Cramer–von Mises, and Kolmogorov–Smirnov tests (see Appendix B).
Table 2 shows the test statistics and the percentage points of the considered tests.
From table 2, since the test statistics are smaller than their corresponding upper tail
percentage points, we conclude that the fitted Weibull distribution is accepted (at α = 0.05
significance level) as a suitable model for the monthly average daily UV data.
Once the fitted Weibull distribution passes both graphical and formal goodness-of-fit tests,
it can be used to estimate certain statistical measures for the monthly average UV data.
Table 3 gives the estimated values of the mean, median, mode, and standard deviation of
the monthly average UV data.
The observed sample values of the mean, median, mode, and standard deviation of the given
data are 0.254, 0.254, 0.251, and 0.074 respectively. These sample values are very close to the
estimated population ones given in table 3.
170 M. E. Ghitany and N. F. El-Nashar

1.0

.8
Probability Distribution

.6
Downloaded by [University of Chicago Library] at 20:12 16 November 2014

.4

.2

0.0
0.0 .1 .2 .3 .4 .5
Monthly Average Ultraviolet Radiation (Kw.h/m2)

Figure 1. Comparison of the empirical and fitted Weibull distribution for the average monthly UV data.

Table 2. Test statistics and upper tail percentage points for goodness-of-fit
tests.

Upper tail percentage point


Modified test Test statistic (α = 0.05)

Anderson–Darling 0.622864 0.757


Cramer–von Mises 0.097011 0.124
Kolmogorov–Smirnov 0.68849 0.843 for n = 20, 0.856 for n = 50

Table 3. Estimated mean, median, mode, and standard deviation of the fitted Weibull
distribution.

Statistical measure Formula Estimated value (kwh/m2 )


 
1
Mean â +1 0.254824

Median â(ln 2)1/b̂ 0.256487


  

 1 1/b̂
Mode â 1 − , b̂ > 1 0.261527

0

b̂ ≤ 1
    1/2
2 1
Standard deviation â  + 1 − 2 +1 0.0717425
b̂ b̂

(c) = 0 t c−1 e−t dt, c > 0 is the gamma function.
Weibull distribution to UV solar radiation data 171

3. Conclusion

It is shown that the Weibull probability distribution is suitable for modelling average daily UV
radiation data of Kuwait. The maximum likelihood method is used to estimate the parameters
of the distribution from the given data. Graphical and formal goodness-of-fit tests are used
to confirm the goodness of the fitted Weibull distribution. This, in turn, will enable us to
estimate certain statistical measures associated with average daily UV radiation data such as
the mean, median, mode (measures of central tendency), and the standard deviation (measure
of variability).
Downloaded by [University of Chicago Library] at 20:12 16 November 2014

References
Al-Aruri, S.D. Rasas, M., Al-Jamal, K. and Shaban, N., An assessment of global ultraviolet solar radiation in the
range (0.290–0.385 µm) in Kuwait. Sol. Energy, 1988, 41(2), 159–162.
Al-Aruri, S.D., The empirical relationship between global radiation and global radiation ultraviolet (0.290–0.385) µm
solar radiation components. Sol. Energy, 1990, 45(2), 61–64.
Al-Aruri, S.D. and Amer, M.F., Empirical regression models for weather data measured in Kuwait during the years
1985, 1986 and 1987. Sol. Energy, 1993, 50(3), 229–233.
Al-Awadhi, S. and El-Nashar, N.F., Stochastic modelling of global solar radiation measured in the state of Kuwait.
Environmetrics, 2002, 13, 751–758.
Blumthaler, M. and Ambach, W., Indication of increasing solar ultraviolet-B radiation flux in Alpine regions. Science,
1990, 248, 206–208.
Chandra, M., Singapurwalla, N.D. and Stephens, M.A., Kolmogorov statistics for tests of fit for the extreme-value
and Weibull distributions. J. Am. Stat. Assoc., 1981, 76, 729–731.
Dvorkin, A.Y. and Steinberger, E.H., Modeling the altitude effect on solar UV radiation. Sol. Energy, 1999, 65(3),
181–187.
El Hadidy, M.A., Abdel-Nabi, D.Y. and Krauss, P.D., Ultraviolet radiation at Dhahran, Saudi Arabia. Sol. Energy,
1990, 44(6), 315–319.
El-Nashar, N.F., Abdullah, A.H. and Al-Zenki, M., Solar global and ultraviolet radiation measurements over Kuwait.
Int. J. Sol. Energy, 2001, 21(4), 281–291.
Kerr, J.B. and McElory, C.T., Evidence for large upward trends of ultraviolet-B radiation linked to ozone depletion.
Science, 1993, 262, 1032–1034.
Khogali, A. and Al-Bar, O.F., A study of solar ultraviolet radiation at Makkah solar station. Sol. Energy, 1992, 48(2),
79–87.
Kollias, N. and Baqer, A., Measurements of solar middle ultraviolet radiation in Kuwait. Sol. Wind Technol., 1984,
1(1), 59–62.
Kollias, N., Baqer, A. and Sadiq, I., Measurements of solar middle ultraviolet radiation in desert environment.
Photochem. Photobiol., 1988, 47(4), 565–569.
Schwander, H., Koepke, P. and Ruggabar, A., Uncertainties in modeled UV irradiances due to limited accuracy and
availability of input data. J. Geophys. Res., 1997, 102(D8), 9419–9429.
Som, A.K., Solar UV-B radiation measurement over Bahrain. Renew. Energy, 1992, 2(1), 93–98.
Stephens, M.A., Goodness-of-fit for the extreme value distribution. Biometrika, 1977, 64, 583–588.

Appendix A

A.1 Maximum likelihood estimation of the parameters of the Weibull distribution

Given a random sample x1 , x2 , . . . , xn from a Weibull distribution with scale parameter a and
shape parameter b having p.d.f. (1), the maximum likelihood estimates â, b̂ of a, b, respectively,
are given by

1/b̂
1 b̂
n
â = x
n i=1 i
172 M. E. Ghitany and N. F. El-Nashar

where b̂ is the solution of the equation


n
1
n
xib̂ ln(xib̂ )
1+ ln(xib̂ ) − i=1
n =0
n i=1 b̂
i=1 xi

Appendix B

B.1 Goodness-of-fit tests for Weibull distribution with unknown parameters


Downloaded by [University of Chicago Library] at 20:12 16 November 2014

B.1.1 Null and alternative hypotheses


H0 : the given random sample comes from Weibull distribution with estimated scale and shape
parameters
H1 : the given random sample does not come from Weibull distribution with estimated scale
and shape parameters

B.1.2 Test statistics


1. Modified Anderson–Darling statistic
  
1
n
0.2
A0 = 1 + √
2
−n − (2i − 1) ln[z(i) (1 − z(i) )]
n n i=1

2. Modified Cramer–von Mises statistic


  n 
2 
0.2 1 2i − 1
W02 = 1 + √ + z(i) −
n 12n i=1 2n

where n is the sample size, zi = F̂ (xi ), i = 1, 2, . . . , n, with F̂ (x) being the estimated
Weibull distribution function given by function (2), and z(i) , i = 1, 2, . . . , n, is the ith
order statistic of z1 , z2 , . . . , zn .
3. Modified Kolomogorv–Smirnov statistic

D0 = n Supx |Fn (x) − F̂ (x)|

where Fn (x) is the empirical distribution function (3).

B.1.3 Upper tail percentage points. Upper tail percentage points for the modified test
statistics A20 and W02 for the Weibull distribution with unknown parameters are given in the
following table; see Stephen (1977).

Significance level α
Modified statistic 0.25 0.10 0.05 0.025 0.01

A20 0.474 0.637 0.757 0.877 1.038


W02 0.073 0.102 0.124 0.146 0.175
Weibull distribution to UV solar radiation data 173

Upper tail percentage points for the modified test statistics D0 for the Weibull distribution
with unknown parameters are given in the following table; see Chandra et al. (1981).

Significance level α
n 0.10 0.05 0.025 0.01

10 0.760 0.819 0.880 0.944


20 0.779 0.843 0.907 0.973
50 0.790 0.856 0.922 0.988
∞ 0.803 0.874 0.939 1.007
Downloaded by [University of Chicago Library] at 20:12 16 November 2014

B.1.4 Decision rule. Reject H0 on α significance level if computed test statistic is larger
than the upper tail percentage point for that statistic, otherwise we fail to reject H0 .

You might also like