Renewable and Sustainable Energy Reviews 53 (2016) 338–362

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Diffuse solar radiation on a horizontal surface: Reviewing

and categorizing the empirical models
Hossein Khorasanizadeh a,n, Kasra Mohammadi b
a
Faculty of Mechanical Engineering and the Energy Research Institute, University of Kashan, Kashan, Iran
b
Department of Mechanical and Industrial Engineering, University of Massachusetts, Amherst, Massachusetts 01003, USA

art ic l e i nf o a b s t r a c t

Article history: Accessibility to the accurate knowledge of diffuse solar radiation is a fundamental prerequisite for
Received 24 May 2015 designing and monitoring the solar energy technologies. Nevertheless, the measured diffuse solar
Received in revised form radiation is not available in most regions across the globe. During the past 60 years in order to estimate
10 August 2015
horizontal diffuse solar radiation on both daily and monthly mean daily basis, numerous empirical
Accepted 18 August 2015
models have been developed or established for many locations around the world. For this purpose
Available online 19 September 2015
different parameters have been utilized and various functional forms employed. This study aims at
Keywords: providing a review on the diffuse solar radiation models developed, established and proposed thus far.
Solar energy To put forward a proper appraisal, the models are primarily categorized on the basis of how to correlate:
Diffuse solar radiation
(1) the diffuse fraction or cloudiness index (i.e. the ratio of diffuse solar radiation to the global solar
Empirical models
radiation), and (2) the diffuse coefficient or diffuse transmittance index (i.e. the ratio of diffuse solar
Review
Category radiation to the extraterrestrial global solar radiation) with different inputs. Both categories, which in
terms of the utilized inputs as well as the functional forms constitute 56 different models are re-
classified into several main sub-categories and then presented in a chronological manner. This review
would be profitable for solar energy researchers in terms of identifying the parameters and the
functional forms widely utilized up till now as well as recognizing their significance. Accessibility to
either solar and/or meteorological data varies from one location to another; thus, the appropriateness of
the models to estimate the horizontal diffuse solar radiation depends on the availability of such inputs at
a desired location. Additionally, to attain proper accuracy and reliability of estimation in a desired
location the inputs should be selected based upon the fact that how much they can provide strong
correlation with diffuse solar radiation.
& 2015 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
2. The main categories of diffuse solar radiation models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
3. Diffuse fraction models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
3.1. Sunshine duration based models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
3.1.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
3.1.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
3.1.3. Group 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
3.1.4. Group 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
3.1.5. Group 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
3.2. Clearness index based models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
3.2.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
3.2.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
3.2.3. Group 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

n
Correspondence to: University of Kashan, Kashan 8731751167, Iran.
E-mail addresses: khorasan@kashanu.ac.ir (H. Khorasanizadeh), kasra278@yahoo.com (K. Mohammadi).

http://dx.doi.org/10.1016/j.rser.2015.08.037
1364-0321/& 2015 Elsevier Ltd. All rights reserved.
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 339

3.2.4. Group 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348

3.2.5. Group 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
3.2.6. Group 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
3.2.7. Group 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
3.3. Sunshine duration and clearness index based models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
3.3.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
3.3.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.3.3. Group 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.4. Cloud cover based models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
3.4.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
3.4.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
3.4.3. Group 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
3.4.4. Group 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
3.5. Other suggested multi parameters diffuse fraction models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
4. Diffuse coefficient models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
4.1. Sunshine duration based models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
4.1.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
4.1.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
4.1.3. Group 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
4.1.4. Group 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
4.1.5. Group 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
4.2. Clearness index based models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
4.2.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
4.2.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.2.3. Group 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.3. Sunshine duration and clearness index based models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.3.1. Group 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.3.2. Group 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
4.4. Other multi parameters diffuse coefficient models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
Acknowledgement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361

1. Introduction
According to status quo, solar energy is being particularly
appealing as an appropriate energy source. Solar energy is clean,
environmentally friendly and inexhaustible that can be effectively
utilized for many applications [1–3]. Harnessing the solar energy
may be profitable to deal with the existing barriers in the field of
energy production. As a matter of fact, solar energy technologies
and solar power plants can be regarded as suitable substitutes for
traditional energy systems for the sake of providing and develop-
ing sustainability around the world [4–6]. Nevertheless, in spite of
considerable efforts to exploit solar energy via various technolo-
gies by scientists, governments as well as investors so far, its
potential is fundamentally unexploited yet. About the significance
of solar energy it is worthwhile to state that in case of converting
only 0.1% of the solar energy arrived at the earth surface to
electricity with the efficiency of 10%, the output power would be
17,300 GW, which is 7 times higher than the global average
momentary electricity consumption in 2012 [7,8].
Owing to geographic dependency as well as variability of solar
radiation, high-quality and reliable information on solar radiation
data is unquestionably essential prior to conducting any effort to
design and install solar energy systems. In fact, having favorable
knowledge of solar radiation is the first and primary step in
designing, establishing, optimizing and assessing the output
energy of solar energy systems [9–15].
Solar radiation passing through the atmosphere is subjected to
variation due to two phenomena: (1) atmospheric scattering by air
molecules, water and dust and (2) atmospheric absorption by O3,
H2O and CO2. Part of the scattered radiation eventually arrives at the
earth surface in the form of diffuse radiation. The measure of
radiation scattering depends on the number of particles on its path
and the size of the particles relative to the radiation wavelength. The
path length of the radiation through the atmosphere is described by
the air mass. The number of water and dust particles encountered by
the radiation is not only dependent on the air mass but also on the
quantities of dust and moisture present in the atmosphere, which are
both themselves time and location dependent.
For any particular location, alongside the global solar radiation, the
knowledge of its diffuse fraction is an essential element for designing
and monitoring the active and passive solar energy systems; from
which one application is building illumination [16–20]. Besides, as
most of the solar energy applications involve inclined solar surfaces, it
is necessary to transform the horizontal diffuse solar radiation to that
for inclined surfaces; as an example, the performance of flat plate solar
surfaces can be appraised if their tilt is adjusted properly.
Generally, the superior way of knowing the amount of solar
radiation at any specific location is to set up the high precision
measuring equipment. Solar radiation is measured using different
instruments such as Pyranometer and Pyrheliometer. The recorded
data may be available for various time scales such as hourly, daily
and monthly mean daily. However, due to a series of obstacles
such as paucity of required instruments as well as fiscal issues the
solar radiation data are not easily accessible, particularly for
stations in developing countries and isolated regions. In fact,
despite worldwide continuous effort to establish further solar
radiation measurement stations in recent years, the number of
stations recording the solar radiation, especially the diffuse com-
ponent, is still limited. As a consequence, it is of indispensible
significance to determine the amount of diffuse solar radiation
alternatively based upon estimating methods.
340 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362
Nomenclature
Gsc solar constant (equal to 1367 W/m2)
H daily or monthly averaged daily global solar radiation
on horizontal surface (MJ/m2)
Hd daily or monthly averaged daily diffuse solar radiation
on horizontal surface (MJ/m2)
Ho daily or monthly averaged daily extraterrestrial solar
radiation on horizontal surface (MJ/m2)
Kdc ¼(Hd/H) daily or monthly averaged daily diffuse fraction or
cloudiness index (dimensionless)
Kdf ¼(Hd/Ho) daily or monthly averaged daily diffuse coefficient
or diffuse transmittance index (dimensionless)
KT ¼(H/Ho) daily or monthly averaged daily clearness index
(dimensionless)
n daily or monthly averaged daily sunshine hours (hr)
nday day number
N daily or monthly averaged daily maximum possible
sunshine hours (hr)
N' the maximum sunshine duration for which the
Campbell-Stokes sunshine recorder is active
Global solar radiation has been considered as the most signifi-
cant parameter in estimating the diffuse fraction [21]. Clearness
index, the ratio of horizontal global terrestrial radiation in a region
to the extraterrestrial radiation, is an element highly correlated
with diffuse radiation. The first model suggested in this regard,
developed by Liu and Jordan [22], aimed at correlating the diffuse
radiation with clearness index. Their investigation has been adapted
by many researchers as a basis in the realm of diffuse solar radiation
models for further studies. In this regard, during the recent decades
a considerable number of methods, models and correlations have
been proposed by many researchers to estimate horizontal diffuse
solar radiation on both daily and monthly mean daily scales in some
considered locations. In the developed models, in addition to
clearness index, different other parameters have been used as input
elements to estimate diffuse solar radiation. Besides, to attain
further preciseness in estimation, various mathematical functional
forms have been utilized even by setting up complex equations.
Owing to a remarkable number of correlations which have
been proposed, applied and tested, identifying the various types of
models as well as commonly utilized input parameters is a difficult
and time consuming task on some occasions. On the other hand,
accessibility to the required data varies from one location to
another; thus, the appropriateness of the methods to estimate
the horizontal diffuse solar radiation depends on the availability of
inputs at a desired location. Additionally, the inputs should be
selected based upon the fact that how much they can provide
strong correlations with diffuse solar radiation to offer further
accuracy and reliability of estimations in the location under
consideration.
The chief goal of this study is to present a review on the
empirical models reported in the literature to estimate horizontal
diffuse solar radiation on both daily and monthly mean daily basis
for various locations distributed across the globe. To fulfill this
objective properly, the models are primarily classified in two major
categories on the basis of how to correlate: (1) the diffuse fraction
or cloudiness index (i.e. the ratio of diffuse solar radiation to the
global solar radiation) and (2) the diffuse coefficient or diffuse
transmittance index (i.e. the ratio of diffuse solar radiation to the
extraterrestrial global solar radiation) with different inputs. After-
wards each category is reclassified into different sub-categories
based upon the utilized input parameters. From each category, the
Ne daily or monthly averaged daily cloud cover rank
Rh daily or monthly averaged daily relative humidity (%)
Tavg daily or monthly averaged daily average ambient
temperature (1C)
Tmax daily or monthly averaged daily maximum ambient
temperature (1C)
Tmin daily or monthly averaged daily minimum ambient
temperature (1C)
W daily or monthly averaged daily atmospheric water
vapor content per unit volume of air (gm/m3)
Greek letters
α solar altitude angle (deg.)
δ solar declination angle (deg.)
φ latitude of the location (deg.)
ω daily or monthly averaged daily atmospheric total
precipitable water column (cm)
ωs sunset hour angle (deg.)
most significant existing empirical models that principally rely
upon weather conditions, seasonal variations, geographical location
and solar radiation characteristics of the sites are presented and
reviewed in a chronological order.
This review would be helpful to determine various inputs that
have been widely utilized for estimation of horizontal diffuse solar
radiation and also to recognize their significance.
2. The main categories of diffuse solar radiation models
As stated in the introduction, many correlations have been
reported in the literature to estimate the diffuse solar radiation on
a horizontal surface (Hd) either on a daily basis (Hereafter DB)
or monthly mean daily basis (Hereafter MB). In this study, the
proposed diffuse solar radiation models are classified in two
primary categories.
(1) In the first major category, the diffuse fraction or cloudiness
index (Kdf) (i.e. the ratio of horizontal diffuse solar radiation,
Hd, to the global solar radiation, H) is correlated with different
input elements. Accordingly, on the basis of parameters
utilized as inputs, this category is classified into five different
sub-categories. These five sub-categories are:


H n
K df ¼ d  f ð1Þ
H N
n refers to the daily or monthly averaged daily sunshine hours
which is usually measured in weather stations and N refers to
the day length or the monthly averaged day length (the
maximum possible sunshine hours).



H H
K df ¼ d  f ð2Þ
H Ho
Ho refers to the daily or monthly averaged daily extraterrestrial
solar radiation on horizontal surface at the same geographical
location as that of H. The ratio of H/Ho is called clearness index,
KT, which indicates the measure of solar attenuation at a
particular location. Its value close to one indicates the negligible
effect of the earth atmosphere on solar radiation, whereas its
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 341

Table 1
The general form of the existing DB or MB diffuse fraction (cloudiness index) models.

No. Functional form Sub-category Number of input parameters Input parameters

   
1 H d =H ¼ aþ b n=N 1 1 n=N
2      2 1 1 n=N
H d =H ¼ aþ b n=N þ c n=N
3      2  3 1 1 n=N
H d =H ¼ aþ b n=N þ c n=N þ d n=N
   
4 H d =H ¼ 1=a þ b n=N 1 1 n=N
5    b 1 1 n=N
H d =H ¼ a n=N c
   
6 H d =H ¼ aþ b H=H o 2 1 H=Ho
    
7 H d =H ¼ aþ b 1= H=H o 2 1 H=Ho
    
8 H d =H ¼ aþ bexp 1= H=H o 2 1 H=Ho
   
9 H d =H ¼ aþ exp bH=H o 2 1 H=Ho
   
10 H d =H ¼ aexp bH=Ho 2 1 H=Ho
    
11 H d =H ¼ ln a þ b H=Ho 2 1 H=Ho
   
12 H d =H ¼ aþ b log H=H o 2 1 H=Ho
13      2 2 1 H=Ho
H d =H ¼ aþ b H=H o þ c H=Ho
14      2  3 2 1 H=Ho
H d =H ¼ aþ b H=H o þ c H=Ho þ d H=H o
15      2  3 2 1 H=Ho
H d =H ¼ aþ b H=H o þ c H=Ho þ d H=H o
 4
þ e H=Ho

     
16 H d =H ¼ aþ b H=H o þ c n=N 3 2 H=Ho ; n=N
17      2 3 2 H=Ho ;n=N
H d =H ¼ aþ b H=H o þ c H=Ho
   2
þ d n=N þ e n=N

18      2  3 3 2 H=Ho ; n=N
H d =H ¼ aþ b H=H o þ c H=Ho þ d H=H o
   2  3
þ d n=N þ e n=N þ f n=N

 
19 H d =H ¼ aþ bðN e Þ 4 1 Ne
   
20 H d =H ¼ aþ b N e =8 4 1 Ne
21      2 4 1 Ne
H d =H ¼ aþ b N e =8 þ c N e =8
22      2  3 4 1 Ne
H d =H ¼ aþ b N e =8 þ c N e =8 þ d N e =8
     
23 H d =H ¼ aþ b n=N þ c N e =8 5 2 n=N; N e
     
24 H d =H ¼ aþ b H=H o þ c N e =8 5 2 H=Ho ; N e
   
25 H d =H ¼ aþ b H=H o þ cRh 5 2 H=Ho ; Rh
   
26 H d =H ¼ aþ b H=H o þ cδ 5 2 H=Ho ; δ
   
H d =H H=Ho ; ωS
 ¼ aþ
 b π=180 ðωs  901Þ
27 5 2
   
 c þ d π=180 ðωs  901Þ  cos 2 H=H o  0:9

    
28 H d =H ¼ aþ b 1=H=H o þ c 1=ðωs  901Þ 5 2 H=Ho ; ωS
     
29 H d =H ¼ aþ b H=H o þ c n=N þ d cos δ 5 3 H=Ho ; n=N; δ
     
30 H d =H ¼ aþ b H=H o þ c n=N þ dδ 5 3 H=Ho ; n=N; δ
     
31 H d =H ¼ aþ b H=H o þ c n=N þ d cos δ þ e cos φ 5 3 H=Ho ; n=N; δ; φ
   
32 H d =H ¼ aþ b H=H o þ dT avg þ eRh 5 3 H=Ho ; T avg ; Rh
   
33 H d =H ¼ aþ b H=H o þ dT avg þ eRh þ e sin α 5 4 H=Ho ;T avg ; Rh ; α
     
34 H d =H ¼ aþ b H=H o þ c n=N þ dT avg þ eRh 5 4 H=Ho ; n=N; T avg ;Rh
       
35 H d =H ¼ aþ b H=H o þ c n=N þ d T max =T min 5 4 H=Ho ; n=N;T max =T min ; Rh
þ eRh

value close to zero states strong solar attenuation. (2) In the second main category, the diffuse coefficient or diffuse


 transmittance index (Kdc) (i.e. the ratio of horizontal diffuse
H n H
K df ¼ d  f ; ð3Þ solar radiation, Hd, to the extraterrestrial solar radiation, Ho) is
H N Ho
correlated with different input elements. This major category

 is classified into four different sub-categories based upon the
H
K df ¼ d  f ðN e Þ ð4Þ type of input parameters used. These four sub-categories are:
H

 n
H
K dc ¼ d  f ð6Þ
Ne is the daily or monthly averaged daily cloud cover rank. In Ho N
Section 3.4 necessary information about cloud cover rank is


Hd H
presented. K dc ¼ f ð7Þ
Ho Ho





H n H Hd n H
K df ¼ d  f ; ; T avg ; Rh ; … ð5Þ K dc ¼ f ; ð8Þ
H N Ho Ho N Ho
342 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

 n 
H
K dc ¼ d  f ; T avg ; Rh ; … ð9Þ 3.1.1. Group 1
Ho N Models from this group have a simple linear form such that the
diffuse fraction is linearly related to relative sunshine duration as
All of the parameters introduced in Eqs. (1)–(9) have been Hd n
¼ aþb ð15Þ
defined in the nomenclature. H N
Daily extraterrestrial solar sradiation on a horizontal surface, Iqbal [24] based upon the measured data of 3 properly dis-
Ho, is calculated via [21,23]: tributed Canadian locations of Montreal, Toronto and Goose Bay

 derived the following MB model:
24  3600 360nday
Ho ¼ Gsc 1 þ 0:033 cos n
π 365 Hd
  ¼ 0:791  0:635 ð16Þ
πωs H N
 cos φ cos δ sin ωs þ sin φ sin δ ð10Þ
180 Barbaro et al. [25] obtained the following MB models for three
where φ is the latitude of the location and nday is the day number of Italian stations of Palermo, Macerata and Genova.
the year, counted from the first of January. For the 1st of January nday For Palermo:
is 1 and for 31st of December it is 365. Also, δ and ωs are the daily Hd n
¼ 0:6603  0:5272 ð17aÞ
solar declination and sunset hour angle, respectively, given as [21,23] H N

   For Macerata:
nday þ284 360
δ ¼ 23:45 sin ð11Þ n
365 Hd
¼ 0:6603  0:5717 ð17bÞ
H N
 
ωs ¼ cos 1
 tan φ tan δ ð12Þ For Genova:
Hd n
The daily maximum possible sunshine duration is [21,23] ¼ 0:5866  0:4264 ð17cÞ
H N
2  
N¼ cos  1  tan φ tan δ ð13Þ Khogal et al. [26] established the following MB model for
15
Sana'a in Yemen as
It is important to note that in the case of monthly mean daily n
Hd
estimation, the average day of each month is used in Eqs. (10)–(13) [21]. ¼ 0:920  0:830 ð18Þ
The aforementioned two categories and all of the sub- H N
categories along with the empirical models which belong to them Lewis [27] suggested the following MB models for Salisbury
are defined and introduced in the following two sections. It should and Bulawayo in Zimbabwe.
be noted that for the DB models either the daily diffuse fraction or For Salisbury:
diffuse coefficient is related to daily values of the input para- Hd n
meters, whereas for the MB models either the monthly averaged ¼ 0:779  0:689 ð19aÞ
H N
daily diffuse fraction or coefficient is related to the monthly
For Bulawayo:
averaged daily values of the input parameters; however, a similar
Hd n
notation has been used hereafter for both types of correlations.
¼ 0:754  0:654 ð19bÞ
H N
Benson et al. [28] derived the following two MB models for
3. Diffuse fraction models winter and summer seasons for Atlanta in USA.
For winter:
In the first main category the diffuse fraction or cloudiness index Hd n n
is correlated with different input parameters. Table 1 summarizes ¼ 0:66  0:61 0:2 r r 0:7 ð20aÞ
H N N
the general functional forms and input parameters of the existing
For summer:
models of this category. As can been from Table 1, various kinds of
Hd n n
functional forms and ten different inputs parameters have been
¼ 0:88  0:80 0:2 r r 0:7 ð20bÞ
utilized so far. In the following sub-sections the DB or MB models H N N
related to each sub-category are introduced. Nagaraja Rao et al. [29] reported the following DB model for
Corvallis, Oregon in USA as
Hd n n
3.1. Sunshine duration based models
¼ 1:0070  0:9316 0:05 r ð21Þ
H N N
The relative sunshine duration has been widely recognized as Ibrahim [30] established the following MB model for Cairo in
an influential and determinant parameter for modeling the hor- Egypt as
izontal diffuse solar radiation. In the models from this sub- n
Hd
category the diffuse fraction, or in other words the cloudiness ¼ 0:79  0:59 ð22Þ
H N
index, is only function of relative sunshine duration as

 n Al-Hamdani et al. [31] developed the following models for
H
K df ¼ d  f ð14Þ different periods of the year for Fudhaliyah, Iraq.
H N For the whole year:
Since sunshine duration is measured at a vast number of Hd n
¼ 0:650  0:494 ð23aÞ
meteorological stations around the world, so many types of H N
empirical models with various functional forms have been sug- For winter:
gested so far. According to the functional form, these models are
Hd n
classified in different groups. The existing models in the literature
¼ 0:782  0:655 ð23bÞ
relating to each group are introduced in the following, accordingly. H N
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 343

For spring: MB model as

Hd n Hd n
¼ 0:734  0:687 ð23cÞ ¼ 0:8009  0:7247 ð30Þ
H N H N
For summer: Trabea [38] obtained the following MB model for AL-Arish, AL-
Hd n Tahrir, Marsa Matroh, Cairo, AL-Kharga and Aswan located in
¼ 0:897  0:790 ð23dÞ Egypt as
H N
Hd n
For autumn: ¼ 0:896  0:688 ð31Þ
H N
Hd n
¼ 0:771  0:646 ð23eÞ Bashahu [39] derived the following MB model for Dakar in
H N
Senegal as
Jain [32] suggested the following MB models for Macerata in n
Hd  
Italy as well as Salisbury and Bulawayo situated in Zimbabwe. ¼ 1:059  0:984 0:6 r n=N r 0:8 ð32Þ
For Macerata: H N

H  n Aras et al. [40] developed the following MB model for 12

¼ 0:824 þ 0:385 1= ð24aÞ provinces in the Central Anatolia Region of Turkey as
H  Hd N
Hd n
Hd  n ¼ 0:6630  0:4883 ð33Þ
¼ 0:177 þ 0:385 1= ð24bÞ H N
H  Hd N
Mubiru and Banda [41] obtained the following MB model for
For Salisbury: Kampala in Uganda as
H  n n n
Hd
¼ 0:516 þ 0:628 1= ð24cÞ ¼ 0:837  0:723 0:31 o o 0:62 ð34Þ
H  Hd N H N N

Hd  n Ulgen and Hepbasli [17] obtained the following MB model for

¼ 0:484 þ 0:628 1= ð24dÞ three Turkey's big cities of Ankara, Istanbul and Izmir as
H  Hd N
Hd n
For Bulawayo: ¼ 0:5456  0:2242 ð35Þ
H N
H  n
¼ 0:590 þ 0:573 1= ð24eÞ Jiang [42] obtained the following DB model for Beijing, China as
H  Hd N
Hd n
 n ¼ 1:0247  0:8119 ð36Þ
Hd H N
¼ 0:410 þ 0:573 1= ð24f Þ
H  Hd N
El-Sebaii et al. [43] presented the following MB model for
Tasdemiroglu and Sever [33] utilized the data of five cities of Jeddah in Saudi Arabia as
Turkey, with different climatic conditions, and derived the follow- Hd n
ing MB model as ¼ 5:488  5:672 ð37Þ
H N
Hd n n
¼ 0:622  0:350 0:2 r r 0:94 ð25Þ
H N N
3.1.2. Group 2
Ahmad et al. [34] developed an MB model for Karachi, Pakistan,
In this group, diffuse fraction, also called cloudiness index, is
as
parameterized as the second-order polynomial function of the
Hd n
relative sunshine duration. The general functional form of this
¼ 0:584  0:378 ð26Þ
H N group is
Nfaoui and Buret [35] developed the following DB and MB Hd  n   n 2
models, respectively, for Rabat in Morocco as ¼ aþb þc ð38Þ
H N N
Hd n
¼ 1:00  0:77 ð27aÞ Barbaro et al. [25] obtained the following MB model for three
H N Italian stations of Palermo, Macerata and Genova.
n n For Palermo:
Hd
¼ 0:89  0:66 0:4 r r 0:8 ð27bÞ n  n 2
H N N Hd
¼ 0:7434  0:8203 þ 0:2454 ð39aÞ
H N N
Lewis [36] derived the following MB model using measured
data of Memphis, Tennessee in USA as For Macerata:
Hd n Hd n  n 2
¼ 0:730  0:440 ð28Þ ¼ 1:0297  2:1096 þ 1:5193 ð39bÞ
H N H N N
Gopinathan and Soler [16] suggested the following MB model For Genova:
for 40 locations widely distributed in the latitude range of 361S to n  n 2
Hd
601N: ¼ 0:8159  1:3289 þ 0:8668 ð39cÞ
H N N
Hd n
¼ 0:79819  0:69930 ð29Þ Lewis [36] utilized the measured data of Memphis of Tennessee
H N
in USA and obtained the following MB model as
In another study, Gopinathan and Soler [37] used measured
Hd n  n 2
data of 17 European locations in France, Belgium, the United ¼ 1:480  2:890 þ1:970 ð40Þ
Kingdom, Portugal, Italy and Greece to establish the following H N N
344 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

Said et al. [44] suggested the following MB correlation for Also, Pandey and Katiyar [47], by utilizing the measured data of
Tripoli, Libya, as Jodhpur, Calcutta, Bombay and Pune derived the following MB
Hd n  n 2 model for the all of these locations as
¼ 1:625  3:421 þ 2:185 ð41Þ n  n 2  n 3
H N N Hd
¼ 0:09781 þ4:7630  11:3200 þ 7:1670 ð52Þ
Trabea [38] obtained the following MB model for AL-Arish, AL- H N N N
Tahrir, Marsa Matroh, Cairo, AL-Kharga and Aswan located in Ulgen and Hepbasli [17] obtained the following MB model for
Egypt as Turkey's big cities of Ankara, Istanbul and Izmir as
n  n 2 Hd n  n 2  n 3
Hd
¼ 0:839  0:537  0:098 ð42Þ ¼ 0:6595  0:7841 þ 0:7461  0:2579 ð53Þ
H N N H N N N
El-Sebaii and Trabea [45] by utilizing the measured insolation Jiang [42] suggested the following DB model for Beijing, China
data of Matruh, Al-Arish, Rafah and Aswan obtained the following as
MB model for the whole Egypt as Hd n  n 2  n 3
n  n 2 ¼ 0:976  0:685 þ0:653  0:892 ð54Þ
Hd H N N N
¼  0:209 þ 2:183  1:785 ð43Þ
H N N Li et al. [48] derived the following MB model for Guangzhou,
Aras et al. [40] established the following MB model for 12 China, as
provinces over Central Anatolia Region of Turkey as Hd n  n 2  n 3
¼ 1:4069  6:3866 þ 18:4206  18:3755 ð55Þ
Hd n  n 2 H N N N
¼ 0:6492  0:4323  0:0512 ð44Þ
H N N Sabzpooshani and Mohammadi [49] established the following
Mubiru and Banda [41] obtained the following MB model for MB model for Isfahan in Iran as
Kampala in Uganda as Hd n  n 2  n 3
¼  0:59276 þ 4:60382  6:85670 þ 3:06795
Hd n  n 2 n H N N N
¼ 0:352 þ 1:386  2:235 0:31 o o0:62 ð45Þ ð56Þ
H N N N
Jiang [42] obtained the following DB model for Beijing, China as
n  n 2 3.1.4. Group 4
Hd In this group, the diffuse fraction, or cloudiness index, is
¼ 0:9612  0:2275  0:6278 ð46Þ
H N N correlated with the rational form of relative sunshine duration as
Hd 1
¼   ð57Þ
3.1.3. Group 3 H a þ b Nn
Some researchers have correlated the diffuse fraction or clou-
diness index, with a third order polynomial function of relative Jain [32] developed the following MB models for Macerata in
sunshine duration as Italy as well as Salisbury and Bulawayo situated in Zimbabwe.
For Macerata:
Hd  n   n 2  n 3
¼ aþb þc þd ð47Þ Hd 1
H N N N ¼   ð58aÞ
H 0:631 þ 4:220 Nn
Tiris et al. [46] obtained the following MB model for Gebze in
Turkey as For Salisbury:

Hd n  n 2  n 3 Hd 1
¼ 0:4177  0:07702  1:9069  1:1900 ð48Þ ¼   ð58bÞ
H N N N H  1:305 þ7:128 Nn
Aras et al. [40] developed a MB model for 12 provinces over For Bulawayo:
Central Anatolia Region of Turkey as Hd 1
n  n 2  n 3 ¼   ð58cÞ
Hd H  1:453 þ7:273 Nn
¼ 0:5562 þ 0:1536  1:2027 þ 0:7122 ð49Þ
H N N N
Mubiru and Banda [41] suggested the following MB model for 3.1.5. Group 5
Kampala in Uganda as In this group, the diffuse fraction, or cloudiness index, is
Hd n  n 2  n 3 n correlated with the exponential form of relative sunshine duration
¼ 1:517  6:434 þ 14:830  12:130 0:31 o o 0:62
H N N N N as
ð50Þ n b
Hd
¼a c ð59Þ
Pandey and Katiyar [47] suggested the following MB models, H N
for Jodhpur, Calcutta, Bombay and Pune stations in India, respec- Pandey and Katiyar [47] using the data of Jodhpur, Calcutta,
tively, as Bombay and Pune in India derived the following MB model as
Hd n  n 2  n 3  n   0:4014
¼  2:887 þ 17:95  29:4 þ 14:92 ð51aÞ Hd
H N N N ¼ 0:9891  0:7839 ð60Þ
H N
Hd n  n 2  n 3
¼ 3:419  15:03 þ 25:18  13:86 ð51bÞ 3.2. Clearness index based models
H N N N

Hd n  n 2  n 3
¼ 0:8384  0:2841  0:8208 þ 0:4315 ð51cÞ In the models from this sub-category, the diffuse fraction or
H N N N cloudiness index is only function of clearness index as



Hd n  n 2  n 3 H H
¼ 1:033  0:9107 0:1288 þ 0:0972 ð51dÞ K df ¼ d  f ð61Þ
H N N N H Ho
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 345

Based on the experiences, clearness index is closely correlated they have correlated the daily averaged diffuse fraction with daily
to diffuse solar radiation, thus has been considered as a determi- averaged clearness index in that particular period.
nant element for estimation of diffuse radiation. One of the For time interval of 5 days:
greatest features of the models from this group is their convenient

Hd H
usage, since for utilizing them only measured global solar radia- ¼ 1:155  1:388 ð69aÞ
H Ho
tion data is required. Various functional forms have been proposed
for clearness index based empirical models in the literature. For time interval of 10 days:
These models, classified in different groups, are presented in the

Hd H
following. ¼ 1:118  1:341 ð69bÞ
H Ho

3.2.1. Group 1 For time interval of 15 days:


The models from this group enjoy a simple form, in which the Hd H
¼ 1:104  1:325 ð69cÞ
diffuse fraction is a linear function of clearness index; as H Ho


Hd H For time interval of 30 days:
¼ a þb ð62Þ
H Ho

Hd H
¼ 1:093  1:313 ð69dÞ
Page [50], using the data from 10 widely-spread sites in the H Ho
401N to 401S latitude belt, derived the following MB relation as

 Nagaraja Rao et al. [29] suggested the following DB model for
Hd H Corvallis, Oregon in USA as
¼ 1:00  1:13 ð63Þ
H Ho 8  
>
> 1:00 H=H o r 0:20
>
>    
H d < 1:130  0:667 H
Iqbal [51] established the following MB model for two Cana-
dian locations of Toronto and Montreal as ¼ Ho 0:20 o H=H o r 0:26 ð70Þ


 H > >    
>
>
Hd H H : 1:403  1:725 HHo 0:26 o H=H o r 0:75
¼ 0:958  0:982 0:3 o o0:6 ð64aÞ
H Ho Ho
Spitters et al. [54] developed the following DB models for de
Also, using data of three Canadian locations of Toronto, Mon-
Bilt in Netherlands and stated that these models are valid for a
treal and Goose Bay, Iqbal [51] obtained the following MB model as


 wide range of climate conditions:
Hd H H 8  
¼ 0:914  0:847 0:3 o o0:6 ð64bÞ >
> 1:00 H=H o o 0:07
H Ho Ho >
>  
>
> 2  
>
H d < 1:203 Ho  0:07 0:07 r H=H o o 0:35
H
Barbaro et al. [25] reported the following MB models for three
¼     ð71Þ
Italian stations of Palermo, Macerata and Genova. H > >
> 1:33  1:46 Ho 0:35 r H=H o o 0:75
H
>
For Palermo: >
>  

 >
: 0:23 0:75 r H=H o
Hd H
¼ 1:0492  1:3246 ð65aÞ
H Ho Lalas et al. [55] derived the following DB and MB models for
For Macerata: three locations of Athens, Rodos and Kythnos in Greece, respec-

 tively, as
Hd H


¼ 0:9918  1:0336 ð65bÞ Hd H H
H Ho ¼ 1:360  1:600 0:22 o o 0:80 ð72aÞ
H Ho Ho
For Genova:




Hd H Hd H H
¼ 0:7273  0:7770 ð65cÞ ¼ 1:270  1:450 0:22 o o 0:80 ð72bÞ
H Ho H Ho Ho

Lewis [27] derived the following MB models for Salisbury and Also for H/Ho o 0.22, they offered Hd/H¼ 0.98.
Bulawayo in Zimbabwe. Al-Hamdani et al. [31] reported the following MB models for
For Salisbury: different periods of the year for Fudhaliyah, Iraq.

 For the whole year:
Hd H

¼ 1:260  1:501 ð66aÞ
H Ho Hd H
¼ 1:102  1:299 ð73aÞ
H Ho
For Bulawayo:

 For winter:
Hd H
¼ 1:049  1:128 ð66bÞ

H Ho Hd H
¼ 1:199  1:494 ð73bÞ
H Ho
Khogal et al. [26] obtained the following MB model for Sana'a,
Yemen, as For spring:



Hd H Hd H
¼ 1:380  1:800 ð67Þ ¼ 1:325  1:634 ð73cÞ
H Ho H Ho
Hawas and Muneer [52], utilizing the data of 13 stations For summer:
distributed across India, suggested the following MB correlation as


 Hd H
Hd H ¼ 1:202  1:480 ð73dÞ
¼ 1:3500  1:6075 ð68Þ H Ho
H Ho
For autumn:
Vignola and McDaniels [53] developed the following models,

Hd H
for six locations distributed in Oregon and Idaho, USA, to give a ¼ 1:127  1:390 ð73eÞ
H Ho
daily average diffuse fraction for a specified time interval. In fact,
346 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

Newland [56] obtained the following MB model for Macau, For winter:
China, as

Hd H

 ¼ 1:31  1:53 ð83bÞ
Hd H H Ho
¼ 1:020  1:157 ð74Þ
H Ho
For spring:
Tasdemiroglu and Sever [33], using data of five cities with

Hd H
different climates in Turkey, derived the following MB model as ¼ 1:31  1:45 ð83cÞ
H Ho



Hd H H For summer:
¼ 0:791  0:775 0:2 r r 0:7 ð75Þ
H Ho Ho

Hd H
Veeran and Kumar [57] established the following MB model for ¼ 0:96  1:06 ð83dÞ
H Ho
Madras in India:

 For autumn:
Hd H

¼ 1:238  1:456 ð76Þ Hd H
H Ho ¼ 1:26  1:59 ð83eÞ
H Ho
Nfaoui and Buret [35] obtained the following DB model for
Moreover, Jacovides et al. [60] suggested the following DB
Rabat in Morocco as


 model for Athalassa, Cyprus for different sky conditions:
Hd H H 8    
¼ 1:33 1:37 0:25 o ð77aÞ > 0:992  0:068 H H=H o r 0:16
H Ho Ho Hd < Ho
¼     ð83fÞ
H > : 1:26  1:425 HHo 0:16 o H=H o r 0:71
Also, Nfaoui and Buret [35] obtained the following MB model
for Rabat in Morocco as


 Lam and Li [61] developed the following DB correlation for
Hd H H Hong Kong as
¼ 0:96 0:83 0:40 r r0:7 ð77bÞ
H Ho Ho 8  
>
> 0:974 H=H o r 0:15
Martinez-Lozano et al. [58] established the following MB model >
<    
Hd
¼ 1:192  1:349 Ho 0:15 o H=H o r 0:7
H
for Valencia in Spain as ð84Þ
H > >  

 >
:
Hd H 0:259 H=H o 4 0:7
¼ 0:17 þ0:81 ð78aÞ
H Ho
Janjai et al. [62], using data of four stations in different parts of
Also, Martinez-Lozano et al. [58] suggested the following DB Bangkok in Thailand, suggested the following MB model as
model for Valencia in Spain as

Hd H
8   ¼ 1:09  1:22 ð85Þ
H=H o r 0:2 H Ho
Hd <
0:99
¼     ð78bÞ
H : 1:36  1:65 HHo 0:2 o H=H o o 0:8 Hove and Göttsche [63] established the following MB model for
Bulawayo in Zimbabwe as
8    
Tiris et al. [59] suggested the following MB model for Gebze,
H d < 1:0294  1:1440 Ho 0:47 o H=H o o 0:75
H
Turkey, as ¼   ð86Þ

 H : 0:1750 0:75 o H=H o
Hd H
¼ 0:6875  0:4981 ð79Þ
H Ho Trabea [38] derived the following MB correlation for AL-Arish,
AL-Tahrir, Marsa Matroh, Cairo, AL-Kharga and Aswan in Egypt as
Lewis [36], using measured data of Memphis of Tennessee in


USA reported the following MB model as Hd H
¼ 0:924  0:894 ð87Þ

 H Ho
Hd H
¼ 0:870  0:840 ð80Þ
H Ho Kaygusuz and Ayhan [64] established the following MB model
for Trabzon, Turkey, as
Gopinathan and Soler [16] obtained the following MB correla-


tion for 40 widely distributed locations in the latitude range of Hd H H
¼ 0:789  0:869 0:252 o o0:610 ð88Þ
361S to 601N as H Ho Ho


Hd H Oliveira et al. [65] derived the following MB model for São
¼ 0:91138  0:96225 ð81Þ
H Ho Paulo in Brazil as


Gopinathan and Soler [37], using measured data of 17 European Hd H
¼ 1:20  1:70 ð89Þ
locations in France, Belgium, the United Kingdom, Portugal, Italy H Ho
and Greece, established the following MB model as El-Sebaii and Trabea [45] using data of Matruh, Al-Arish, Rafah

 and Aswan reported the following MB model for the whole Egypt
Hd H
¼ 0:9851  1:0680 ð82Þ as
H Ho

Hd H
Jacovides et al. [60] reported the following MB model for ¼ 1:242  1:337 ð90Þ
Athalassa, Cyprus, as H Ho

 Bashahu [39] established the following MB model for Dakar,
Hd H
¼ 1:03 1:17 ð83aÞ Senegal, as
H Ho



Hd H H
Also, Jacovides et al. [60] reported the following DB models for ¼ 1:13  1:29 0:51 r r 0:68 ð91Þ
H Ho Ho
Athalassa, Cyprus for 0.16o H/Ho o0.71.
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 347

Paliatsos et al. [66] derived the following DB models for 0.1o H/ 3.2.2. Group 2
Ho o0.8 for Athens, Greece. The models from this group are simple rational functions of
For the whole year: clearness index; such that:



Hd H Hd H
¼ 1:771  1:4285 ð92aÞ ¼ a þ b 1= ð100Þ
H Ho H Ho
For warm period: Jain [32] proposed the following MB models for Macerata in

 Italy as well as Salisbury and Bulawayo in Zimbabwe.
Hd H
¼ 1:1582  1:3705 ð92bÞ For Macerata:
H Ho


For cold period: Hd H
¼ 0:193 þ 0:343 1= ð101aÞ

 H Ho
Hd H
¼ 1:2124  1:5088 ð92cÞ For Salisbury:
H Ho


For spring: Hd H
¼ 0:602 þ 0:571 1= ð101bÞ

 H Ho
Hd H
¼ 1:1783  1:3741 ð92dÞ
H Ho For Bulawayo:


For summer: Hd H

 ¼ 0:492 þ 0:506 1= ð101cÞ
H Ho
Hd H
¼ 1:0522  1:2314 ð92eÞ
H Ho Tiris et al. [59] reported the following MB model for Gebze,
For autumn: Turkey, as



Hd H Hd H
¼ 1:2165  1:5380 ð92f Þ ¼ 0:2599 þ 0:0880 1= ð102Þ
H Ho H Ho

For winter:


Hd H 3.2.3. Group 3
¼ 1:2347  1:5573 ð92gÞ
H Ho In this group the diffuse fraction is dependent upon the
exponential functional form of clearness index. Thus far, to provide
Jin et al. [67], using data of 78 stations across China, derived the
exponential relations between diffuse fraction and clearness index,
following DB model as
8 the three following functional forms have been proposed as
 
>
> 0:987 H=H o o 0:2

>   Hd H
Hd <   ¼ a þ bexp 1= ð103Þ
¼ 1:292  1:447 Ho 0:2 r H=H o r 0:75
H
ð93Þ H Ho
H > >  
>
: 0:209

H=H o 4 0:75
Hd H
¼ a þ exp b ð104Þ
Aras et al. [40] obtained the following MB correlation for 12 H Ho
provinces over Central Anatolia Region of Turkey as


 Hd H
Hd H ¼ aexp b ð105Þ
¼ 1:0212  1:1672 ð94Þ H Ho
H Ho
Tiris et al. [59] developed the following MB model for Gebze,
Mubiru and Banda [41] obtained the following MB model for Turkey, as
Kampala in Uganda as



 Hd H
Hd H H ¼ 0:3917 þ 0:0065exp 1= ð106Þ
¼ 0:980  1:046 0:32 o o 0:59 ð95Þ H Ho
H Ho Ho
Lealea and Tchinda [69] proposed the following MB models for
Jiang [68] utilized the data of 8 typical stations distributed in
four stations situated in North and far North of Cameroon.
different parts of China and derived the following MB model as
For Kousseri:


Hd H

¼ 1:0120  1:1440 ð96Þ Hd H
H Ho ¼ 3:97 þ exp  3:96 ð107aÞ
H Ho
Ulgen and Hepbasli [17] obtained the following MB model for
For Maroua:
Turkey's big cities of Ankara, Istanbul and Izmir as



Hd H Hd H
¼ 0:6772  0:4841 ð97Þ ¼ 3:57 þ exp  3:80 ð107bÞ
H Ho H Ho

Jiang [42] obtained the following DB model for Beijing, China, For Garoua:


as Hd H

 ¼ 3:93 þ exp  3:99 ð107cÞ
Hd H H Ho
¼ 1:236  1:422 ð98Þ
H Ho For Touboro:
El-Sebaii et al. [43] derived the following MB model for Jeddah

Hd H
in Saudi Arabia as ¼ 3:64 þ exp  3:83 ð107dÞ
H Ho


Hd H
¼ 4:618  6:269 ð99Þ Boukelia et al. [20] established the following MB models for six
H Ho
stations in Algeria.
348 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

For Algiers: For Ghardaia:



Hd H Hd H
¼ 1:832exp  2:900 ð108aÞ ¼ 0:002  0:870 log ð112cÞ
H Ho H Ho

For Constantine: For Bechar:



Hd H Hd H
¼ 1:093exp  2:090 ð108bÞ ¼ 0:033  0:560 log ð112dÞ
H Ho H Ho

For Ghardaia: For Adrar:



Hd H Hd H
¼ 5:188exp  4:010 ð108cÞ ¼ 0:095  0:380 log ð112eÞ
H Ho H Ho

For Bechar: For Tamanrasset:



Hd H Hd H
¼ 2:414exp  3:340 ð108dÞ ¼  0:185 1:240 log ð112fÞ
H Ho H Ho

For Adrar:

 3.2.5. Group 5
Hd H In this group, the diffuse fraction is correlated by a second
¼ 1:097exp  2:230 ð108eÞ
H Ho order polynomial function of the clearness index as
For Tamanrasset:

2
Hd H H

 ¼ aþb þc ð113Þ
Hd H H Ho Ho
¼ 21:880exp  6:380 ð108f Þ
H Ho
Barbaro et al. [25] obtained the following MB model for three
Italian stations of Palermo, Macerata and Genova.
For Palermo:
3.2.4. Group 4


2
In the models from this group, the diffuse fraction is correlated Hd H H
¼ 1:0896  1:4797 þ 0:1471 ð114aÞ
with the first-order of clearness index with the logarithmic H Ho Ho
functional form. The first type of models from this group was
For Macerata:
proposed by Lealea and Tchinda [69] as




2
Hd H Hd H H
¼ ln a þb ð109Þ ¼ 2:6845  6:6848 þ 4:6701 ð114bÞ
H Ho H Ho Ho
For Genova:
Lealea and Tchinda [69] proposed the following MB models for


2
four stations in North and far North of Cameroon: Hd H H
For Kousseri: ¼ 0:6153  0:2738  0:5561 ð114cÞ
H Ho Ho



Hd H Lalas et al. [55] suggested the following DB model for three
¼ ln 2:72  2:08 ð110aÞ
H Ho locations of Athens, Rodos and Kythnos in Greece as
For Maroua:

2

Hd H H H


 ¼ 1:240  1:080  0:510 0:22 o o 0:80
Hd H H Ho Ho Ho
¼ ln 2:73  2:09 ð110bÞ
H Ho ð115Þ

For Garoua: And for H/Ho o0.22, they proposed Hd/H ¼0.98.


 Bindi et al. [70], based upon the data of 7 stations (2 stations in
Hd H
¼ ln 2:78  2:18 ð110cÞ Netherlands and 5 stations in Italy), suggested the following DB
H Ho
model, from which one is the second order polynomial function of
For Touboro: the clearness index as


 8  
Hd H > 1:00 H=H o o 0:07
¼ ln 2:75  2:11 ð110dÞ >
>
H Ho >
>  2  
>
>
Hd < 1:00  2:30 H
H o
 0:07 0:07 r H=H o o 0:35
Another type of logarithmic functional form was proposed by ¼   ð116Þ
H >  
Boukelia et al. [20] as >
> 1:33  1:46 H
0:35 r H=H o o 0:75
>
> Ho

 >
>  
Hd H : 0:23 H=H o Z 0:75
¼ a þ b log ð111Þ
H Ho Nfaoui and Buret [35] suggested the following DB model for
Boukelia et al. [20] proposed the following MB models for six different sky conditions for Rabat in Morocco:
stations in Algeria. 8  
0:98 H=H o o 0:10
For Algiers: Hd <    2
¼   ð117Þ

 H : 0:98 þ 0:15 HH  1:48 HH 0:10 r H=H o
Hd H o o
¼ 0:029  0:570 log ð112aÞ
H Ho
Martinez-Lozano et al. [58] established the following DB model
For Constantine: for Valencia in Spain as



2

Hd H Hd H H H
¼ 0:113  0:390 log ð112bÞ ¼ 1:26  1:19  0:48 0:2 o o 0:8 ð118aÞ
H Ho H Ho Ho Ho
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 349

Also, Martinez-Lozano et al. [58] obtained the following MB Jiang [68] derived the following MB model, utilizing the data of
model for Valencia in Spain as 8 typical stations distributed in China as


2

2
Hd H H Hd H H
¼  0:86 þ 4:59  3:46 ð118bÞ ¼ 1:0180  1:1670 þ 0:0240 ð125Þ
H Ho Ho H Ho Ho

Said et al. [44] suggested the following MB correlation for Jiang [42] reported the following DB model for Beijing, China,
Tripoli in Libya as as


2


2 Hd H H
Hd H H ¼ 1:0286  0:1225  1:5784 ð126Þ
¼ 2:944  8:474 þ 6:767 ð119Þ H Ho Ho
H Ho Ho

Trabea [38] established the following MB model for AL-Arish,

3.2.6. Group 6
AL-Tahrir, Marsa Matroh, Cairo, AL-Kharga and Aswan in Egypt as
In this group, the diffuse fraction is correlated by a third order


2
Hd H H polynomial function of the clearness index such that:
¼ 0:534 þ 0:384  1:036 ð120Þ

2
3
H Ho Ho
Hd H H H
¼ aþb þc þd ð127Þ
Paliatsos et al. [66] derived the following DB models for 0.1 oH/ H Ho Ho Ho
Ho o 0.8 for different periods of the year for Athens, Greece. Klein [72] using data of [22] developed the following MB model
For the whole year: for USA:


2

2
3

Hd H H Hd H H H H
¼ 1:0827  0:8670  0:7405 ð121aÞ ¼ 1:390  4:027 þ 5:531  3:108 0:3 o o 0:7
H Ho Ho H Ho Ho Ho Ho
ð128Þ
For warm period:


2 Barbaro et al. [25] obtained the following MB model for three
Hd H H Italian stations of Palermo, Macerata and Genova.
¼ 1:0704  1:0266  0:3359 ð121bÞ
H Ho Ho ForPalermo:


2
3
For cold period: Hd H H H
¼ 13:9375  76:2760 þ 144:3846  92:1480


2 H Ho Ho Ho
Hd H H
¼ 1:0801  0:7617  0:8835 ð121cÞ ð129aÞ
H Ho Ho
For Macerata:
For spring:

2
3


2 Hd H H H
Hd H H ¼ 1:2634 þ 0:3801  6:9645 þ 6:3493 129bÞ
¼ 1:0490  0:7548  0:6455 ð121dÞ H Ho Ho Ho
H Ho Ho
For Genova:
For summer:

2
3
Hd H H H


2 ¼ 1:1845  4:2489 þ 8:5833  6:9198
Hd H H H Ho Ho Ho
¼ 0:7585  0:1496  0:9614 ð121eÞ
H Ho Ho ð129cÞ

For autumn: Erbs et al. [73], by utilizing data of four locations in USA,
established the following MB model as


2
Hd H H

2
¼ 1:0827  0:8670  0:7405 ð121f Þ Hd H H
H Ho Ho ¼ 1:3170  3:0230 þ 3:3720
H Ho Ho
For winter:
3

H H
 1:7690 0:3 r r 0:8 ð130Þ


2 Ho Ho
Hd H H
¼ 1:0810  0:5446  1:1038 ð121gÞ
H Ho Ho Also, Erbs et al. [73] proposed the following two seasonal MB
correlations, using data of four locations in USA as
Tarhan and Sari [71] established the following MB model for

five provinces in Central Black Sea Region of Turkey as H
for ωs r 81:41 and 0:3 r r0:8


2 Ho
Hd H H
¼ 0:9885  1:4276 þ 0:5679 ð122Þ

2
3
H Ho Ho Hd H H H
¼ 1:391  3:560 þ 4:189  2:137 ð131aÞ
H Ho Ho Ho
Aras et al. [40] reported the following MB model for 12
provinces over Central Anatolia Region of Turkey as

H


2 for ωs 4 81:41 and 0:3 r r0:8
Ho
Hd H H
¼ 1:1244  1:5582 þ 0:3635 ð123Þ
H Ho Ho

2
3
Hd H H H
¼ 1:311  3:022 þ 3:427  1:821 ð131bÞ
Mubiru and Banda [41] obtained the following MB model for H Ho Ho Ho
Kampala in Uganda as
Vignola and McDaniels [53] suggested the following DB model


2
 for six locations distributed in Oregon and Idaho in USA as
Hd H H H
¼ 1:868  4:952 þ 4:183 0:32 o o 0:59

2
3
H Ho Ho Ho Hd H H H
¼ 0:921 þ 1:218  5:510 þ 3:241 ð132Þ
ð124Þ H Ho Ho Ho
350 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

Muneer and Hawas [74] reported the following DB model, Miguel et al. [75] derived the following DB correlation, using
based upon measured data of 13 stations located in the wide measured data from 11 locations in South France, Greece, Italy,
latitude range of 081290 to 281350 in Nover India as Portugal and Spain as
8   8  
>
> 0:98 H=H o o 0:20 >
> 0:952 H=H o r 0:13
> >
Hd <    2  3   Hd <    2  3  
¼ 1:024 þ 0:470 HHo  3:622 HHo þ 2:0 HHo 0:20r H=H o r 0:77 ¼ 0:868 þ 1:335 HHo  5:782 HHo þ 3:721 HHo 0:13 o H=H o r 0:8
H >
>   H >>  
>
: >
:
0:16 0:77o H=H o 0:141 H=H o 4 0:8
ð133Þ ð139Þ
Ibrahim [30] established the following MB model for Cairo in Tarhan and Sari [71] reported the following MB model for five
Egypt as provinces in Central Black Sea Region of Turkey as


2
3

2
3
Hd H H H Hd H H H
¼ 0:636  0:279  0:194  0:383 ð134Þ ¼ 1:0270  1:6582 þ 1:1018  0:4019 ð140Þ
H Ho Ho Ho H Ho Ho Ho
Tiris et al. [59] obtained the following MB model for Gebze in Aras et al. [40] obtained the following MB model for 12
Turkey as provinces over Central Anatolia Region of Turkey as


2
3

2
3
Hd H H H Hd H H H
¼ 0:5830 þ 0:9985 5:2400 þ5:3220 ð135Þ ¼ 1:7111  4:9062 þ 6:6711  3:9235 ð141Þ
H Ho Ho Ho H Ho Ho Ho
Martinez-Lozano et al. [58] established the following MB model Mubiru and Banda [41] obtained the following MB model for
for Valencia in Spain, as Kampala in Uganda as


2
3

2
Hd H H H Hd H H
¼  8:38 þ 48:9  85:0 þ 49:6 ð136aÞ ¼ 2:333  8:137 þ 11:320
H Ho Ho Ho H Ho Ho

3

Also, Martinez-Lozano et al. [58] reported the following DB H H
model for Valencia in Spain as  5:233 0:32 o o 0:59 ð142Þ
Ho Ho


2
3

Hd H H H H Ulgen and Hepbasli [17] derived the following MB model for
¼ 0:72 þ 2:73  9:12 þ 5:97 0:2 o o 0:8
H Ho Ho Ho Ho
Turkey's big cities, named Ankara, Istanbul and Izmir as
ð136bÞ

2
3
Hd H H H
Jacovides et al. [60] obtained the following DB models for ¼ 0:981  1:9028 þ1:9319  0:6809 ð143Þ
H Ho Ho Ho
different periods of the year for Athalassa, Cyprus.
For the whole year Jiang [68] reported the following MB model for 8 stations
8     distributed in different parts of China as
>
> 0:992  0:0486 Ho H=H o r 0:10
H
>
>
<

2
3
Hd    2  3   Hd H H H
¼ 0:954 þ 0:734 H  3:806 H þ 1:703 H
> 0:10 o H=H o r 0:71 ¼ 0:8220 þ 0:1930  2:9280 þ 2:017 ð144Þ
H > Ho Ho Ho H Ho Ho Ho
>
>  
: 0:165 H=H o 40:71
Jiang [42] reported the following DB model for Beijing, China as
ð137aÞ

2
3
Hd H H H
For dry months ¼ 0:920 þ 1:346  5:989 þ 3:684 ð145Þ
 
H Ho Ho Ho
8  
>
> 0:987  0:12 Ho H=H o r 0:13
H
>
>
< Li et al. [48] derived the following MB model for Guangzhou in
Hd    2  3  
¼ 0:919 þ 0:996 H  4:553 H þ 2:340 H 0:13 o H=H o r 0:71 China as
H > > Ho Ho Ho
>
>  

2
3
: 0:163 H=H o 40:71 Hd H H H
¼ 1:2433  3:7455 þ 9:3735  10:8486
ð137bÞ H Ho Ho Ho
ð146Þ
For wet months
8     Bortolini et al. [76] derived the following DB models based
>
> 1:00  0:0676 Ho H=H o r 0:10
H
>
> upon the measured data of 44 stations located in 11 European
Hd <    2  3  
¼ 0:987 þ 0:946 H 4:135 H þ 1:831 H 0:10 o H=H o r 0:715 countries.
H >> Ho Ho Ho
>
>   For the whole year:
: 0:169 H=H o 4 0:715


2
3
ð137cÞ Hd H H H
¼ 0:9888 þ 0:3950  3:7003 þ 2:2905
H Ho Ho Ho
Bashahu [39] derived the following MB model for Dakar,
Senegal, as ð147aÞ


2 For summer:
Hd H H
¼ 37:4368  61:3179 þ 0:2856

2
3
H Ho Ho Hd H H H

3
 ¼ 1:0172 þ 0:0158  2:7036 þ 1:5729
H H H Ho Ho Ho
 0:8451 0:51 r r 0:68 ð138Þ
Ho Ho ð147bÞ
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 351

For winter: Martinez-Lozano et al. [58] reported the following DB model


2
3 for Valencia in Spain as
Hd H H H
¼ 0:9403 þ 0:9887  5:2499 þ3:4586

2
3
H Ho Ho Ho Hd H H H
¼ 2:02  10:05 þ 34:7 þ57:3
ð147cÞ H Ho Ho Ho

4

H H
þ 32:7 0:2 o o0:8 ð155Þ
Ho Ho
3.2.7. Group 7
In this group, the diffuse fraction is correlated with the fourth Oliveiraet al. [65] recommended the following DB model for
order polynomial function of the clearness index as São Paulo in Brazil as


2
3
4

2
3
4
Hd H H H H Hd H H H H
¼ a þb þc þd þe ð148Þ ¼ 1:00 þ 0:27  2:50  2:60 þ 4:30
H Ho Ho Ho Ho H Ho Ho Ho Ho
ð156Þ
Collares-Pereira and Rabl [77], using the data of 5 locations in
USA, proposed the following DB model as
8  
> 0:99 H=H o r 0:17 3.3. Sunshine duration and clearness index based models
>
>
>
>      
H d < 1:188  2:272 H þ 9:473 H
2
¼ Ho Ho 0:17 o H=H o o 0:8 In some types of models, both influential and determinant
H > >  3  4 elements of clearness index and relative sunshine duration have
>
>
>
:  21:856 HHo þ 14:648 HHo been utilized together. In another word, the diffuse fraction is not
only function of clearness index but is also function of relative
ð149Þ
sunshine duration; such that:
Erbs et al. [73] proposed some seasonal DB correlations using


H n H
data of four locations in USA. for ωs o 81.41 K df ¼ d  f ; ð157Þ
H N Ho
   2  
Hd
H ¼ 1:00  0:2727 H o þ 2:4495 H o H=H o o0:715
H H

 3  4
 11:9514 HHo þ 9:3879 HHo ð150aÞ
  3.3.1. Group 1
Hd
H ¼ 0:143 H=H o Z0:715 In this group the diffuse fraction is correlated with the first
order of both clearness index and relative sunshine duration as
And for ωs Z81.41
 n
   2  3 Hd H
Hd   ¼ aþb þc ð158Þ
H ¼ 1:00 þ 0:2832 H
Ho  2:5557 HHo þ 0:8448 HHo H=H o o 0:722 H Ho N
Hd
 
H ¼ 0:175 H=H o Z 0:722 Al-Hamdani et al. [31] established the following MB correla-
ð150bÞ tions for different specific intervals of the year for Fudhaliyah
in Iraq.
Nagaraja Rao et al. [29] developed the following DB model for For the whole year
Corvallis, Oregon in USA as
 n
Hd H


2
3
4 ¼ 0:962  0:758  0:277 ð159aÞ
Hd H H H H H Ho N
¼ 0:9493 þ 1:1314  5:7688 þ 4:5503  1:2457
H Ho Ho Ho Ho
For winter:
ð151Þ

Hd H n
Newland [56] suggested the following DB correlation for ¼ 0:923  0:429  0:497 ð159bÞ
H Ho N
Macau, China, as

8      3  2  
H d < 0:9713 þ 0:5614 HHo  3:3534 HHo þ 1:0339 HHo þ 0:5136 HHo 0:10 o H=H o r0:71
¼   ð152Þ
H : 0:180 0:71 o H=H o

Tasdemiroglu and Sever [33] obtained the following MB model For spring:
for five cities with different climatic conditions in Turkey as

Hd H n


2
3
4 ¼ 0:951  0:565  0:471 ð159cÞ
Hd H H H H H Ho N
¼ 1:6932 8:2262 þ25:5532  37:8070 þ19:8178
H Ho Ho Ho Ho
ð153Þ For summer:

 n
Bindi et al. [70] suggested the following DB model, based upon Hd H
¼ 1:093  0:507 þ 0:629 ð159dÞ
the data of 7 stations, from which 2 were in Netherlands and H Ho N
5 were in Italy, as

8  
>
> 0:99 H=H o r0:17
>
Hd <    2  3  4  
¼ 1:188  2:272 HHo þ 9:473 HHo  21:856 HHo þ 14:648 HHo 0:17 o H=H o r 0:80 ð154Þ
H >>  
>
: 0:25 0:80 o H=H o
352 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

For autumn: Jiang [42] obtained the following DB model for Beijing, China,

 n as
Hd H
¼ 0:856 þ 0:275  0:538 ð159eÞ

2 n  n 2
H Ho N Hd H H
¼ 0:971 þ 0:268  1:241  0:258  0:157
H Ho Ho N N
Gopinathan and Soler [16] suggested the following MB model
ð170Þ
for 40 locations, widely distributed in the latitude range of 361S to
601N, as Li et al. [48] reported the following MB model for Guangzhou,

 n China, as
Hd H

2
¼ 0:87813  0:33280  0:53039 ð160Þ Hd H H n  n 2
H Ho N ¼ 0:7463 þ 1:2922  3:7966  0:7285 þ 1:0592
H Ho Ho N N
Lewis [36] derived the following MB model, using measured ð171Þ
data of Memphis, Tennessee in USA as

 n
Hd H 3.3.3. Group 3
¼ 0:880  0:910 þ 0:040 ð161Þ
H Ho N In the models classified in this group, the diffuse fraction is
correlated with the third order polynomial function of not only the
Gopinathan and Soler [37] suggested the following MB model,
clearness index but also the relative sunshine duration; such that:
utilizing the data of 17 European locations in France, Belgium, the


2
3  n   n 2  n 3
United Kingdom, Portugal, Italy and Greece as Hd H H H
¼ aþb þc þd þe þf þg

 n H Ho Ho Ho N N N
Hd H
¼ 0:9599  0:6945 0:3381 ð162Þ ð172Þ
H Ho N
Jiang [68], by utilizing the data of 8 locations in different parts
Trabea [38] obtained the following MB model for AL-Arish, AL-
of China, obtained the following MB model as
Tahrir, Marsa Matroh, Cairo, AL-Kharga and Aswan in Egypt as


2
3

 n Hd H H H
Hd H ¼ 0:7470 þ 1:5020  4:9560 þ 3:3210
¼ 0:9270  0:1640 0:5950 ð163Þ H Ho Ho Ho
H Ho N
n  n 2  n 3
Mubiru and Banda [41] established the following MB model for  1:0040 þ 1:7470 1:2260 ð173Þ
Kampala in Uganda as N N N

 n
 Jiang [42] obtained the following DB model for Beijing, China,
Hd H H as
¼ 1:010  0:959  0:149 0:32 o o0:59;
H Ho N Ho

2
3
n Hd H H H
¼ 0:945 þ 0:734  2:876 þ 1:421
0:31 o o 0:62 ð164Þ H Ho Ho Ho
N n  n 2  n 3
Jiang [68], using the measured data of 8 typical stations  0:392 þ 0:378  0:404 ð174Þ
N N N
distributed in China, established the following MB model as
Khorasanizadeh et al. [78] established the following MB model

 n
Hd H for Tabass in Iran as
¼ 1:00 0:8580  0:2350 ð165Þ
H Ho N

2
3
Hd H H H
¼ 0:95930 0:87130 þ 0:29191 0:09790
Jiang [42] obtained the following DB model for Beijing, China, H Ho Ho Ho
n  n 2  n 3
as
 0:28419 þ0:02653  0:02083 ð175Þ

 n N N N
Hd H
¼ 1:144  0:707  0:433 ð166Þ Bakirci [79], by utilizing the measured data for 8 cities in
H Ho N
Turkey, suggested the following MB model as
El-Sebaiiet al. [43] reported the following MB model for Jeddah

2
3
Hd H H H
in Saudi Arabia as ¼ 0:8130  0:2041  0:8108 þ 0:5217
H Ho Ho Ho

 n n  n 2  n 3
Hd H
¼ 4:609  6:318 þ 0:047 ð167Þ  0:0491  0:5646 0:3961 ð176Þ
H Ho N N N N

3.4. Cloud cover based models

3.3.2. Group 2
In this group, the diffuse fraction is correlated with the second It is a well-known fact that diffuse solar radiation is highly
order polynomial function of not only clearness index but also the contingent upon the presence of clouds, as it results in scattering
relative sunshine duration as the solar radiation remarkably. Cloud cover, Ne, is ranked between


2 0 and 8, where Ne ¼0 denotes clear sky and Ne ¼ 8 denotes overcast
Hd H H n  n 2
¼ aþb þc þd þe ð168Þ sky. When the sky is totally covered by clouds, the relative cloud
H Ho Ho N N cover, Ne/8 is equal to one; thus the relative sunshine duration, n/
N, is zero and all of the global radiation has diffuse nature. In this
Jiang [68] obtained the following MB model using data of
sub-category, the diffuse fraction is only function of cloud cover;
8 typical locations distributed in China as
such that:


2 n  n 2

Hd H H H
¼ 0:9450 0:6750  0:1660  0:1730  0:0790 K df ¼ d  f ðN e Þ ð177Þ
H Ho Ho N N H
ð169Þ
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 353

3.4.1. Group 1 3.5. Other suggested multi parameters diffuse fraction models
In this group, the diffuse fraction is correlated with the first
order of cloud cover as In some established diffuse solar radiation empirical models,
Hd along with clearness index and/or relative sunshine duration some
¼ a þb ðN e Þ ð178Þ other parameters such as air temperatures, relative humidity,
H
cloud cover, sunset angle and solar declination angle have been
Lewis [27] proposed the following MB models for Salisbury and utilized, such that:
Bulawayo in Zimbabwe.


For Salisbury: H n H
K df ¼ d  f ; ; T avg ; Rh ; … ð187Þ
H N Ho
Hd
¼ 0:111 þ 0:059 Ne ð179aÞ Due to diversity of extra parameters utilized, all of the estab-
H
lished multi parameter models from this category are introduced
For Bulawayo:
as they have been reported in the literature without placing them
Hd in different sub-groups.
¼ 0:123 þ 0:058 Ne ð179bÞ
H Considering the influence of seasonal variation of day length as
a variable, Collares-Pereira and Rabl [77] suggested the following
functional forms using a combination of clearness index and
3.4.2. Group 2
sunset hour angle as
In the models classified in this group, the diffuse fraction is       
correlated with the first order of relative cloud cover as H d =H ¼ a þb π =180 ðωs  901Þ  c þ d π =180 ðωs  901Þ

   
Hd Ne  cos 2 H=H o  0:9 ð188Þ
¼ a þb ð180Þ
H 8
Collares-Pereira and Rabl [77] using the data of 5 locations in
Bashahu [39] proposed the following MB correlations for Dakar USA proposed the following MB model as
in Senegal as  
H ¼ 0:775 þ 0:347 π =180 ðωs  901Þ 
Hd


Hd Ne  


¼ 0:12 þ 0:52 0:28 r N e =8 r0:65 ð181aÞ    H
H 8 0:505 þ 0:261 π =180 ðωs 901Þ  cos 2  0:9 ð189Þ
Ho


Hd Ne   where ωs is the sunset hour angle for a horizontal surface in
¼ 0:67  0:59 0:35 r 1  N e =8 r 0:72 ð181bÞ
H 8 degrees and 2(H/Ho–0.9) is in radians. Also, for the considered
Mubiru and Banda [41] obtained the following MB model for locations in their study, during Feb–Apr and Aug–Oct ωs was close
Kampala in Uganda as to 901, and during May–July and also during November–January

 ωs was close to 1001.
Hd Ne  
¼  0:472 þ 1:285 0:66 r N e =8 r0:80 ð182Þ In another study, Tiris et al. [59] proposed the following
H 8 functional forms based upon the combination of clearness index
and sunset hour angle as
    
3.4.3. Group 3 H d =H ¼ a þb 1=H=H o þ c 1=ðωs  901Þ ð190Þ
In this group, the diffuse fraction is correlated with the second
Tiris et al. [59] proposed the following MB correlation for
order of relative cloud cover as
Gebze, Turkey, as


2

Hd Ne Ne Hd H 
¼ a þb þc ð183Þ ¼ 0:2751 þ 0:8210 1=  0:0251 1=ðωs  901Þ ð191Þ
H 8 8 H Ho
Mubiru and Banda [41] proposed the following MB correlation Gopinathan and Soler [16] suggested the use of the solar
for Kampala in Uganda as declination angle in a combination with clearness index and


2 relative sunshine duration in the following form:
Hd Ne Ne  
¼ 2:796  7:676 þ 6:126 0:66 r N e =8 r0:80
 n
H 8 8 Hd H
¼ aþb þc þd cos δ ð192Þ
ð184Þ H Ho N
Gopinathan and Soler [16], based upon the data of 40 locations
3.4.4. Group 4 widely distributed in the latitude range of 361S to 601N, proposed
In this group, the diffuse fraction is correlated by a third order the following MB model as

 n
polynomial function of the relative cloud cover; such that: Hd H
¼ 1:01833  0:33029  0:53110  0:14696 cos δ


2
3 H Ho N
Hd Ne Ne Ne
¼ a þb þc þd ð185Þ ð193Þ
H 8 8 8
Also, Gopinathan and Soler [16] introduced both solar declina-
Mubiru and Banda [41] suggested the following MB model for
tion angle and latitude angle to the combination of clearness index
Kampala in Uganda as
and relative sunshine duration and proposed the following form:


2
Hd Ne Ne
 n
¼ 87:650  356:400 þ 482:900 Hd H
H 8 8 ¼ aþb þc þd cos δ þe cos φ ð194Þ

3 H Ho N
Ne  
 216:900 0:66 r N e =8 r 0:80 ð186Þ Gopinathan and Soler [16] proposed the following MB model
8
for 40 locations widely distributed in the latitude range of 361S to
354 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

601N as locations in Nigeria, all having the following structure as

 n
  

Hd H Hd H n T max
¼ 1:01267  0:33058  0:53656  0:14698 cos δ ¼ aþb þc þd þ e Rh ð206Þ
H Ho N H Ho N T min
þ 0:011427 cos φ ð195Þ For Sokoto:

 n

In another study, Gopinathan and Soler [37], using the combi- Hd H T max
¼ 1:055  0:815  0:0353 þ 0:142 þ0:00078 Rh
nation of solar declination angle, clearness index and relative H Ho N T min
sunshine duration, suggested the following functional form as ð207aÞ

 n
Hd H For Maiduguri:
¼ 0:9843  0:7233 0:3610 þ 0:000748 δ ð196Þ


H Ho N Hd H n T max
¼ 0:7795  0:830  0:0364 þ 0:0152 þ 0:00023 Rh
H Ho N T min
Gopinathan and Soler [37] based upon the data of 17 European
locations in France, Belgium, the United Kingdom, Portugal, Italy ð207bÞ
and Greece proposed the following MB model as For Port Harcourt:

 n
 n

Hd H Hd H T max
¼ 0:9843  0:7233 0:3610 þ 0:000748 δ ð197Þ ¼ 0:684  0:735  0:095 þ 0:0248 þ0:00065Rh
H Ho N H Ho N T min
Janjai et al. [62] using the combination of average air tempera- ð207cÞ
ture, relative humidity and clearness index introduced the follow- For Owerri:
ing structure:
 n

Hd H T max

 ¼ 0:775  0:954  0:056  0:000042 þ 0:0762 Rh
Hd H H Ho N T min
¼ aþb þ c T avg þ d Rh ð198Þ
H Ho ð207dÞ
Janjai et al. [62] proposed the following MB model for four For Enugu:
stations located at different parts of Bangkok in Thailand as
 n

Hd H T max

 ¼ 0:642  0:851  0:016 þ 0:087 þ 0:00057 Rh
Hd H H Ho N T min
¼ 0:913  1:46 þ 0:0118 T avg 0:014 Rh ð199Þ
H Ho ð207eÞ
Li et al. [48] obtained the following correlation on MB for For Abeokuta:
Guangzhou, China, as
 n

Hd H T max

 ¼ 0:6796  1:099 þ0:044 þ 0:079 þ0:0014 Rh
Hd H H Ho N T min
¼ 1:3686  1:6036 þ0:0080 T avg  0:4912 Rh ð200Þ
H Ho ð207fÞ

Li et al. [48] examined the influence of adding average air For Yola:

 n

temperature and relative humidity to the linear sunshine duration Hd H T max
¼ 1:1007  0:9844 þ 0:0214  0:1117  0:0001 Rh
and clearness index based model by suggesting the following H Ho N T min
structure: ð207gÞ

  
Hd H n For Jos:
¼ aþb þc þ d T avg þ e Rh ð201Þ
H Ho N
 n

Hd H T max
¼ 1:028  1:0081 0:0229  0:031  0:0011 Rh
Li et al. [48] proposed the following multiple parameters MB H Ho N T min
model for Guangzhou, China, as ð207hÞ

 n
Hd H Mubiru and Banda [41], by considering four different combina-
¼ 0:4461 þ 0:4187 0:8972 þ 0:0049 T avg þ 0:3231 Rh
H Ho N tions of two diverse input parameters, proposed the following
ð202Þ structures as
Hd  n 
N 
Li et al. [80], using the data of 17 stations across China, derived ¼ aþb þc
e
ð208Þ
the following MB model as H N 8

 n

Hd H Hd H
¼ 1:1937  0:6821 0:4658  0:0008 T avg  0:1987 Rh ¼ aþb þ c Rh ð209Þ
H Ho N H Ho
ð203Þ

Hd H
Li et al. [48] considered the influence of solar elevation angle by ¼ aþb þc δ ð210Þ
H Ho
suggesting the following multi parameters structure as




Hd H Hd H Ne
¼ aþb þ c T avg þ d Rh þ e sin α ð204Þ ¼ aþb þc ð211Þ
H Ho H Ho 8
Mubiru and Banda [41] proposed the following four MB
Li et al. [48] developed the following MB model for Guangzhou,
correlations for Kampala in Uganda as
China, as


 Hd n Ne  
Hd H ¼ 0:245  0:593 þ0:707 0:66 o Ne =8 o 0:80;
¼ 1:4668  1:7335 þ0:0121 T avg  0:3707 Rh  0:2683 sin α H N 8
H Ho  
0:31 o n=N o 0:62 ð212aÞ
ð205Þ


Falayi et al. [81] used a new combination of meteorological Hd H  
¼ 0:759  1:014 þ 0:294Rh 0:32 o H=H o o 0:59;
parameters to propose eight MB models for some nominated H Ho
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 355

Table 2
The general form of the existing DB or MB diffuse coefficient models.

No. Functional form Sub-category Number of parameters Input parameters

   
1 H d =H o ¼ a þ b n=N 1 1 n=N
2      2 1 1 n=N
H d =H o ¼ a þ b n=N þ c n=N
3      2  3 1 1 n=N
H d =H o ¼ a þ b n=N þ c n=N þ d n=N
   
4 H d =H o ¼ a þ b log n=N 1 1 n=N
    
5 H d =H o ¼ a exp b n=N 1 1 n=N
   
6 H d =H o ¼ a þ b H=H o 2 1 H=Ho
7    h  i
2 1 H=Ho
H d =H o ¼ H=H o 1  exp a  H=H ab
o

8      2 2 1 H=Ho
H d =H o ¼ a þ b H=H o þ c H=H o
9      2  3 2 1 H=Ho
H d =H o ¼ a þ b H=H o þ c H=H o þ d H=H o
     
10 H d =H o ¼ a þ b H=H o þ c n=N 3 2 H=Ho ; n=N
11      2 3 2 H=Ho ; n=N
H d =H o ¼ a þ b H=H o þ c H=H o
   2
þ d n=N þ e n=N

   
12 H d =H o ¼ a þ b n=N 0 þ cW 4 2 n=N ' ; W
 
13 Hd =Ho ¼ aþ b n=N þ c ω 4 2 n=N; ω
   
14 H d =H o ¼ a þ b n=N þ cT avg 4 2 n=N; T avg
   
15 H d =H o ¼ a þ b n=N þ cRh 4 2 n=N; Rh
 
16 H d =H o ¼ a þ bT avg þ cRh 4 2 T avg ; Rh
   
17 H d =H o ¼ a þ b n=N þ cT avg þ dRh 4 3 n=N; T avg ; Rh
18      2 4 2 n=N; T avg
H d =H o ¼ a þ b n=N þ c n=N þ dT avg
19      2 4 2 n=N; Rh
H d =H o ¼ a þ b n=N þ c n=N þ dRh
20      2 4 3 n=N; T avg ; Rh
H d =H o ¼ a þ b n=N þ c n=N þ dT avg þ eRh
     
21 H d =H o ¼ a þ b n=N þ c log n=N þ dT avg þ eRh 4 3 n=N; T avg ; Rh

0:63 o Rh o0:73 ð212bÞ

index) is function of the relative sunshine duration only; such that:


 n
Hd H H
¼ 0:9572  0:9975 þ 0:0005 δ ð212cÞ K dc ¼ d  f ð213Þ
H Ho Ho N



Hd H Ne  
¼ 0:512  0:968 þ 0:577 0:32 o H=H o o 0:59;
H Ho 8 4.1.1. Group 1
  In this group, the diffuse coefficient or diffuse transmittance
0:66 o Ne =8 o 0:80 ð212dÞ
index is parameterized as a first-order function of the relative
sunshine duration as
Hd n
4. Diffuse coefficient models ¼ aþb ð214Þ
Ho N

The diffuse solar radiation models are not restricted to those Barbaro et al. [25] reported the following MB models for three
from the first main category, which were presented in Section 3. Italian stations of Palermo, Macerata and Genova.
Some researchers have aimed at modeling the diffuse solar For Palermo:
radiation utilizing the diffuse coefficient (also called diffuse Hd n
transmittance index), which is defined as the ratio of horizontal ¼ 0:2626  0:1391 ð215aÞ
Ho N
diffuse solar radiation to the extraterrestrial horizontal solar
radiation. The models from this category are called the diffuse For Macerata:
coefficient or diffuse transmittance index models, accordingly. In Hd n
Table 2, the functional forms and the input parameters of the ¼ 0:2989  0:1577 ð215bÞ
Ho N
existing models of this category in the literature have been
presented. As seen from Table 2, the functional forms as well as For Genova:
the number of parameters used as inputs for this category are Hd n
fewer compared to those of the first main category. In the ¼ 0:1532 þ 0:0283 ð215cÞ
Ho N
following sub-sections, the models from each group of this
category are introduced. Jain [32] proposed the following MB models for Macerata in
Italy as well as Salisbury and Bulawayo situated in Zimbabwe.
For Macerata:
4.1. Sunshine duration based models n
Hd
¼ 0:293  0:135 ð216aÞ
Ho N
As stated earlier, sunshine duration has a direct influence on
different components of solar radiation, particularly on the diffuse For Salisbury:
solar radiation. Therefore, it has been vastly considered as an n
Hd
important element to estimate the horizontal diffuse solar radia- ¼ 0:374  0:271 ð216bÞ
Ho N
tion. In some models the diffuse coefficient (diffuse transmittance
356 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

For Bulawayo: 4.1.3. Group 3

Hd n In the models of this group, the diffuse coefficient is correlated
¼ 0:364  0:253 ð216cÞ with the exponential of the first-order relative sunshine duration
Ho N
as
Aras et al. [40] established the following MB model for 12   n 
Hd
provinces over Central Anatolia Region of Turkey as ¼ a exp b ð225Þ
Ho N
Hd n
¼ 0:2595  0:0978 ð217Þ Boukelia et al. [20] proposed the following MB models for six
Ho N
Algerian locations.
Jiang [68], by utilizing the data of 8 locations distributed in For Algiers:
China, obtained the following MB model as   n 
n Hd
Hd ¼ 0:227 exp  0:170 ð226aÞ
¼ 0:2420  0:0650 ð218Þ Ho N
Ho N
For Constantine:
Ulgen and Hepbasli [17] obtained the following MB model for
Hd   n 
three big cities of Ankara, Istanbul and Izmir in Turkey as ¼ 0:207 exp  0:140 ð226bÞ
n Ho N
Hd
¼ 0:1677 þ 0:0926 ð219Þ
Ho N For Ghardaia:
Hd   n 
El-Sebaii et al. [43] established the following MB model for
¼ 0:667 exp  1:460 ð226cÞ
Jeddah in Saudi Arabia as Ho N
Hd n For Bechar:
¼ 3:542  3:664 ð220Þ
Ho N
Hd   n 
¼ 0:045 exp 1:617 ð226dÞ
Tiris et al. [46] derived the following MB correlation for Gebze Ho N
in Turkey as
For Adrar:
Hd n
¼ 0:1426  0:1190 ð221Þ Hd   n 
Ho N ¼ 0:119 exp 0:394 ð226eÞ
Ho N
Mubiru and Banda [41] obtained the following MB model for
Kampala in Uganda as For Tamanrasset:
n n Hd   n 
Hd ¼ 0:904 exp  2:100 ð226fÞ
¼ 0:257  0:070 0:31 o o0:62 ð222Þ Ho N
Ho N N

4.1.2. Group 2 4.1.4. Group 4

In the models from this group, the diffuse coefficient is In this group, the diffuse coefficient is correlated with the
correlated with the logarithmic of first-order relative sunshine second order polynomial function of the relative sunshine dura-
duration as tion as
Hd n Hd  n   n 2
¼ a þ b log ð223Þ ¼ aþb þc ð227Þ
Ho N Ho N N
Boukelia et al. [20] developed the following MB models for six Iqbal [24] obtained the following MB model, using the mea-
stations in Algeria. sured data of 3 widely spread Canadian locations of Montreal,
For Algiers: Toronto and Goose Bay as
Hd n n  n 2
¼ 0:194  0:020 log ð224aÞ Hd
Ho N ¼ 0:1633 þ 0:4778  0:6555 ð228Þ
Ho N N
For Constantine: Barbaro et al. [25] obtained the following MB models for three
Hd n Italian stations of Palermo, Macerata and Genova.
¼ 0:182  0:010 log ð224bÞ For Palermo:
Ho N
Hd n  n 2
For Ghardaia:
¼ 0:2205 þ 0:0126  0:1292 ð229aÞ
n Ho N N
Hd
¼ 0:164  0:200 log ð224cÞ
Ho N For Macerata:
For Bechar: Hd n  n 2
¼ 0:3627  0:4259 þ 0:2678 ð229bÞ
Hd n Ho N N
¼ 0:214 þ 0:218 log ð224dÞ
Ho N For Genova:
For Adrar: Hd n  n 2
n ¼ 0:1717  0:0461 þ 0:0725 ð229cÞ
Hd Ho N N
¼ 0:178 þ 0:061 log ð224eÞ
Ho N Ibrahim [30] reported the following MB model for Cairo in
For Tamanrasset: Egypt as
Hd n Hd n  n 2
¼ 0:106  0:280 log ð224f Þ ¼ 0:2520  0:0001  0:0830 ð230Þ
Ho N Ho N N
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 357

Massaquoi [82] using measured data of four locations in Ulgen and Hepbasli [17] reported the following MB model for
Canada, Italy and Zimbabwe obtained the following MB model as three big cities of Ankara, Istanbul and Izmir in Turkey as
Hd n  n 2 Hd n  n 2  n 3
¼ 0:143 þ 0:368  0:434 ð231Þ ¼ 0:1437 þ 0:2151  0:1748 þ 0:0697 ð243Þ
Ho N N Ho N N N
Ahmad et al. [34] derived the following MB model for Karachi, Li et al. [48] reported the following MB model for Guangzhou in
Pakistan as China as
Hd n  n 2 n  n 2  n 3
Hd
¼ 0:181 þ 0:145  0:179 ð232Þ ¼ 0:1637  0:1875 þ 1:7268  2:2804 ð244Þ
Ho N N Ho N N N
Trabea [38] obtained the following MB model for six Egyptian
locations of AL-Arish, AL-Tahrir, Marsa Matroh, Cairo, AL-Kharga
4.2. Clearness index based models
and Aswan as
Hd n  n 2 As mentioned in Section 3, clearness index is one of the
¼ 0:101 þ 1:092  0:854 ð233Þ
Ho N N elements that highly influence the diffuse solar radiation so that
it is considered as a principal input parameter to estimate diffuse
El-Sebaii and Trabea [45], using data of Matruh, Al-Arish, Rafah
radiation. There are numerous established diffuse coefficient
and Aswan established the following MB model for the whole
models, in which the diffuse transmittance index is function of
Egypt as
clearness index only; such that:
n  n 2
Hd


¼  0:113 þ 1:217  0:954 ð234Þ H H
Ho N N K dc ¼ d  f ð245Þ
Ho Ho
Bashahu [39] obtained the following MB model for Dakar in
Senegal as These models are presented in the following sub-sections.

Hd n  n 2 n
¼  7:0744 þ 31:4386  15:5906 0:6 r r0:8 4.2.1. Group 1
Ho N N N
The models classified in this group are simple linear functions
ð235Þ
of clearness index; such that:
Aras et al. [40] established the following MB model for 12

Hd H
provinces located in Central Anatolia Region of Turkey as ¼ aþb ð246Þ
Ho Ho
Hd n  n 2
¼ 0:2142 þ 0:0863  0:1684 ð236Þ Jain [32] proposed the following MB models for three locations
Ho N N
of Macerata in Italy as well as Salisbury and Bulawayo in
Mubiru and Banda [41] obtained the following MB model for Zimbabwe.
Kampala in Uganda as For Macerata:
Hd n  n 2 n

¼ 0:219 þ 0:097  0:177 0:31 o o0:62 ð237Þ Hd H
Ho N N N ¼ 0:341  0:189 ð247aÞ
Ho Ho
Jiang [68] obtained the following MB model for 8 stations
For Salisbury:
distributed in different parts of China as

n  n 2 Hd H
Hd ¼ 0:570  0:600 ð247bÞ
¼ 0:1050 þ 0:5500  0:6190 ð238Þ Ho Ho
Ho N N
For Bulawayo:


4.1.5. Group 5 Hd H
¼ 0:514  0:505 ð247cÞ
In this group, the diffuse coefficient is correlated with the third Ho Ho
order polynomial function of the relative sunshine duration; such Aras et al. [40] reported the following MB model for 12
that: provinces over Central Anatolia Region of Turkey as
Hd n  n 2  n 3

¼ a þb þc þd ð239Þ Hd H
Ho N N N ¼ 0:3310  0:2330 ð248Þ
Ho Ho
Aras et al. [40] established the following MB model for 12
Ulgen and Hepbasli [17] obtained the following MB model for
provinces over Central Anatolia Region of Turkey as
Turkey's three big cities of Ankara, Istanbul and Izmir as
Hd n  n 2  n 3

¼ 0:2427  0:0933 þ 0:1846  0:2184 ð240Þ Hd H
Ho N N N ¼ 0:1155 þ 0:1958 ð249Þ
Ho Ho
Mubiru and Banda [41] obtained the following MB model for
El-Sebaii et al. [43] established the following MB model for
Kampala in Uganda as
Jeddah in Saudi Arabia as
Hd n  n 2  n 3

¼ 1:459  8:232 þ 18:000 12:920 Hd H
Ho N N N ¼ 2:973  4:037 ð250Þ
n Ho Ho
0:31 o o 0:62 ð241Þ
N Moreover, Al-Hamdani et al. [31] proposed the following DB
Jiang [68] reported the following MB model based upon the model for Fudhaliyah, Iraq as
data of 8 typical stations distributed in different parts of China as



Hd H ab
n  n 2  n 3 ¼ 1  exp a  ð251aÞ
Hd Ho Ho H=H o
¼ 0:1610 þ 0:1320 þ 0:3030  0:6190 ð242Þ
Ho N N N
where a and b are empirical coefficients. Also, a is presented in
358 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362

terms of b as El-Sebaiiet al. [43] suggested the following MB model for

Jeddah in Saudi Arabia as
a ¼ 0:41=ðb  0:45Þ ð251bÞ

 n
Hd H
By substituting the Eq. (230b) into Eq. (230a) the following ¼ 3:002  3:882  0:150 ð260Þ
Ho Ho N
equation was attained:



 
Hd H 0:41 b
¼ 1  exp 1 ð251cÞ
Ho Ho b 0:45 H=H o 4.3.2. Group 2
In this group, the diffuse coefficient is correlated with the
where b is the maximum clear-sky transmissivity, which is 0.74 for
second order polynomial function of not only clearness index but
the considered location.
also relative sunshine duration as


2 n  n 2
Hd H H
4.2.2. Group 2 ¼ aþb þc þd þe ð261Þ
Ho Ho Ho N N
In this group, the diffuse coefficient is correlated by a second
order polynomial function of the clearness index as Li et al. [48] developed the following MB model for Guangzhou


2 located in China as
Hd H H
¼ aþb þc ð252Þ

2 n
Ho Ho Ho Hd H H
¼  0:0493 þ 1:4146  1:9500  0:0306
Ho Ho Ho N
Aras et al. [40] suggested a MB model for 12 provinces over
 n 2
Central Anatolia Region of Turkey as þ 0:1269 ð262Þ


2 N
Hd H H
¼ 0:0511 þ 0:8267 0:9854 ð253Þ
Ho Ho Ho
4.4. Other multi parameters diffuse coefficient models

4.2.3. Group 3 In a series of proposed diffuse solar radiation models, along

In this group, the diffuse coefficient is correlated with the third with relative sunshine duration some other supplementary para-
order polynomial function of the clearness index as meters such as average daily temperature, relative humidity and


2
3 water vapor content have been utilized; such that:
Hd H H H
 n 
¼ aþb þc þd ð254Þ H
Ho Ho Ho Ho K df ¼ d  f ; T avg ; Rh ; … ð263Þ
Ho N
Aras et al. [40] reported the following MB model for 12
Due to diversity of extra parameters utilized, all of the estab-
provinces over Central Anatolia Region of Turkey as
lished multi parameter models from this group, which have been


2
3
Hd H H H reported in the literature, are introduced in the following, but
¼ 0:3276  0:7515 þ1:9883 1:8497 ð255Þ
Ho Ho Ho Ho without placing them in different sub-groups.
Hussain [83] proposed a new combination of input parameters
Ulgen and Hepbasli [17] obtained the following MB model for with the following structure:
Turkey's big cities of Ankara, Istanbul and Izmir as
Hd n


2
3 ¼ aþb 0 þc W ð264aÞ
Hd H H H Ho N
¼ 0:0273 þ 0:7270 1:0411 þ0:6659 ð256Þ
Ho Ho Ho Ho
where N0 is the maximum sunshine duration for which the
Li et al. [48] reported the following MB model for Guangzhou, Campbell  Stokes sunshine recorder was active. Also, W (in gm/
China, as m3) is the monthly mean atmospheric water vapor content per


2
3 unit volume of air, calculated based on the relative humidity, Rh,
Hd H H H
¼ 0:1366  0:4425 þ3:8771 5:6212 ð257Þ and the average ambient temperature, Tavg (in 1C), as [83]:
Ho Ho Ho Ho  
W ¼ Rh 4:7923 þ 0:3647 T avg þ0:0055 T 2avg þ 0:0003 T 3avg ð264bÞ

4.3. Sunshine duration and clearness index based models Hussain [83] suggested the following MB correlation for some
Indian locations as
As both clearness index and relative sunshine duration are Hd n
influential and determinant elements in estimating the diffuse ¼ 0:30  0:165 0 þ 0:0025W ð265Þ
Ho N
solar radiation, at times they have been used together as inputs for
modeling the diffuse coefficient; such that: Bashahu [39] proposed the following structure based upon the


 relative sunshine duration and total precipitable water column as
H n H
K dc ¼ d  f ; ð258Þ Hd n
Ho N Ho ¼ aþb þc ω ð266aÞ
Ho N
whereω (in cm) is computed by the following expression [39]:

4.3.1. Group 1 ω ¼ 0:227ρω ð266bÞ

In this group, the diffuse coefficient is correlated with the first ρω is the water vapor density and is expressed in terms of
order of not only the clearness index but also the relative sunshine water vapor pressure (ev) and the average ambient temperature
duration as Tavg (in 1K) as [39]:

  
Hd H n eV
¼ aþb þc ð259Þ ρω ¼ 217 ð266cÞ
Ho Ho N T avg
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362 359

Bashahu [39] suggested the following MB correlation for Dakar MB models developed are introduced by Eqs. (272)–(278) as
in Senegal as Hd n
n ¼ aþb þ c T avg ð272aÞ
Hd   Ho N
¼ 0:2271  0:1302 þ 0:0198 ω 0:6 r n=N r 0:8;
Ho N n
2:91 r ω r 4:96 ð267Þ Hd
¼ 0:4028  0:2410 0:0003 T avg ð272bÞ
Ho N
Li et al. [48] proposed a new structure for diffuse radiation
Hd n
estimation by correlating the diffuse coefficient with relative ¼ aþb þ cRh ð273aÞ
sunshine duration, average daily temperature and relative humid- Ho N
ity as n
Hd
Hd n n ¼ 0:3231  0:1930 þ0:0006Rh ð273bÞ
¼ a þb þc log þd T avg þ e Rh ð268Þ Ho N
Ho N N
Hd
Li et al. [48] developed the following MB model for Guangzhou ¼ a þ bT avg þ cRh ð274aÞ
Ho
in China as
Hd n n Hd
¼ 0:5686  0:3724 þ 0:2991 log þ 0:0031 T avg  0:2035 Rh ¼ 0:0051 þ 0:0033T avg þ 0:0020Rh ð274bÞ
Ho N N Ho
ð269Þ n
Hd
Li et al. [80], employing the measured data of 17 locations in ¼ aþb þ c T avg þ d Rh ð275aÞ
Ho N
China, reported the following MB model, as
Hd n
Hd n n ¼ 0:2008  0:1223 þ0:0014 T avg þ0:0011 Rh ð275bÞ
¼ 0:7537  0:5832 þ 0:4954 log  0:0005 T avg  0:1123 Rh Ho N
Ho N N
ð270Þ Hd  n   n 2
¼ aþb þc þd T avg ð276aÞ
Li et al. [19] obtained the following MB models for 7 locations Ho N N
in China. n  n 2
Hd
For Kunming: ¼ 0:3867  0:1831 0:0454  0:0004 T avg ð276bÞ
Ho N N
Hd n n
¼ 0:1902  0:0651 þ 0:0317 log þ 0:0040 T avg  0:0035 Rh  n   n 2
Ho N N Hd
¼ aþb þc þd Rh ð277aÞ
ð271aÞ Ho N N
For Wuhan: Hd n  n 2
n n ¼ 0:2872  0:0879 0:0789 þ 0:0006 Rh ð277bÞ
Hd Ho N N
¼ 0:0230 þ 0:1916  0:1870 log þ 0:0008 T avg þ 0:0187Rh
Ho N N  n   n 2
Hd
ð271bÞ ¼ aþb þc þd T avg þ e Rh ð278aÞ
Ho N N
For Beijing:
Hd n  n 2
n n
Hd ¼ 0:2601  0:4171 þ0:2440 þ 0:0021 T avg þ 0:0012 Rh
¼ 0:6264  0:4045 þ 0:2636 log þ 0:0004 T avg  0:1691 Rh Ho N N
Ho N N
ð271cÞ ð278bÞ
Karakoti et al. [84] tested the performance of all of the seven
For Harbin:
established models for another 5 Indian locations. They found that
Hd n n
¼ 1:1072  1:0391 þ 1:1722 log  0:0010 T avg  0:0086 Rh all of the proposed models enjoy good generalization capabilities
Ho N N and favorable precision for other Indian locations; nevertheless,
ð271dÞ the correlation (252b), in which the relative sunshine duration,
For Lanzhou: relative humidity and average air temperature have been utilized,
n n showed more exactness.
Hd
¼ 1:6504  1:4101 þ 1:8113 log  0:0018 T avg  0:2370 Rh
Ho N N
ð271eÞ 5. Discussion
For Lasa:
Reviewing the existing empirical models in the literature
Hd n n
¼ 1:3688  1:3269 þ 1:4478 log  0:0011 T avg  0:0317 Rh revealed that up to now the linear, second order, third order and
Ho N N forth order polynomials as well as the logarithmic and exponential
ð271f Þ functions are the most typical mathematical functions utilized to
For Wulumuqi: correlate the diffuse radiation with different input parameters. The
n n elements frequently utilized as inputs are:
Hd
¼ 1:1750  0:9171 þ 0:9899 log  0:0038 T avg  0:2851 Rh
Ho N N
(1) Meteorological parameters: extraterrestrial solar radiation,
ð271gÞ
horizontal global solar radiation, sunshine duration, maximum
Karakoti et al. [84] proposed seven new structures to correlate the possible sunshine duration, maximum, minimum and average
diffuse coefficient with different input parameters. Using the data of ambient temperatures, relative humidity, cloud cover, atmo-
18 stations in India, they developed the new MB regression coeffi- spheric water vapor content and atmospheric total precipita-
cients for the proposed structures. The proposed structures and the ble water column.
360 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362
(2) Astronomical parameters: solar declination and solar hour
angles.
(3) Geographical parameters: latitude.
(4) Geometrical parameters: solar altitude angle.
Applying a piece-wise regression test has been a typical way
undertaken either to develop or to establish clearness index-based
diffuse solar radiation models. By dividing the sky conditions into
three cases including overcast, partly-cloudy and clear sky the
more appropriate fits between diffuse solar radiation and clear-
ness index have been developed. Mathematical representation for
the overcast, partly-cloudy or clear sky conditions may be differ-
ent; nevertheless, for the overcast skies the correlation between
diffuse radiation and clearness index has been generally presented
in a linear form, while for the partly cloudy skies a polynomial
functional form has been specified. Finally, for the clear sky
condition a constant value has been regarded as a broadly
accepted choice. It is worthwhile to state that the mentioned
mathematical representations are not strict, but variable depend-
ing upon the climatic condition and solar radiation characteristics
at the nominated site.
Reviewing the literature revealed that it is not possible to
introduce a set of input parameters with particular functional
forms for optimal estimation of horizontal diffuse solar radiation
in general for the whole globe. In fact, the possibility of improving
the accuracy of estimation by considering some certain inputs
with a specific functional form is site dependent. This is due to the
specific nature of solar radiation, which is totally dependent upon
the geographical location as well as the details of climatic condi-
tions of every region. To restate this, a brief review of the previous
efforts to improve the accuracy of estimation of diffuse solar
radiation is presented in the following.
El-Sebaii et al. [43] for Jeddah (211420 N) in Saudi Arabia and by
utilizing the clearness index instead of sunshine duration in both
diffuse fraction and diffuse coefficient models, achieved more
accuracy. Also, the models based on both sunshine duration and
clearness index showed further precision over either sole clear-
ness index based or sole sunshine duration based models. The
results of proposing and establishing ten different empirical MB
models using different inputs by Li et al. [48] for Guangzhou
(221500 N) in China showed that combination of clearness index,
relative sunshine duration, average ambient temperature and
relative humidity with linear functional form is the most proper
choice. Jiang [68], by performing an investigation for eight cities
distributed in China, found that the model with the second order
polynomial function of both clearness index and relative sunshine
duration show superiority over the models which are only clear-
ness index based or sunshine duration based. Janjai et al. [62] in an
investigation to estimate the diffuse solar radiation enhanced the
accuracy of MB linear correlation between diffuse fraction and
clearness index by introducing average temperature and relative
humidity. Gopinathan and Soler [16] using measured data of 40
widely distributed locations in the latitude range of 361S to 601N
showed that higher accuracy of MB models is achieved when both
clearness index and relative sunshine hours are inputs compared
to the models which are functions of only clearness index or
relative sunshine duration. Furthermore, they concluded that
estimation precision can be improved further by addition of solar
declination angle as a third input. Also, adding the latitude of the
locations as fourth input did not bring about a noticeable
enhancement in estimation. Trabea [38], in an investigation to
estimate MB diffuse solar radiation over Egypt, proved that
considering both clearness index and relative sunshine duration
as inputs provide further exactness compared to either only
clearness index based or sunshine duration based models. Li
et al. [80], based upon an investigation undertaken for estimating
the MB diffuse solar radiation in 17 locations of China, demon-
strated the significance of average ambient temperature and
relative humidity in enhancing the accuracy of the models. They
also showed that considering average ambient temperature and
relative humidity as extra inputs offer more accuracy compared
with the models which are functions of only relative sunshine
hours or functions of both relative sunshine hours and clearness
index. Karakoti et al. [84], by proposing 7 models, which correlated
the diffuse coefficient with different inputs, for various Indian
locations, found that higher precision is attained if combination of
linear form of relative sunshine hours, relative humidity and
average ambient temperature are inputs. Khorasanizadeh et al.
[78], using the diffuse solar radiation estimations of 20 existing
models for city of Tabass (331360 N) in Iran, showed that combina-
tion of clearness index and relative sunshine hours provide more
precision compared to the case in which only clearness index or
relative sunshine hours was the inputs. Sabzpooshani and
Mohammadi [49], on the basis of adapted diffuse solar radiation
data of 16 existing models for city of Isfahan (30143/ N) in Iran,
demonstrated that relative sunshine duration is a more dominant
parameter compared to the clearness index. Jiang [42] demon-
strated that for DB diffuse solar radiation estimation in Beijing
(391500 N) in China considering both clearness index and relative
sunshine hours provides superiority over the cases for which
either clearness index or relative sunshine hours are predictors.
Gopinathan and Soler [37], on the basis of the results of an
investigation for 17 European locations in France, Belgium, the
United Kingdom, Portugal, Italy and Greece, found that combina-
tion of only two parameters of clearness index and relative
sunshine duration leads to enhancing the accuracy of the models.
Also, they showed that when the solar declination angle was
considered as the third predictor the accuracy improved merely
slightly. The results achieved by Bashahu [39] showed that using
relative sunshine duration as a predictor provided more precision
compared with the model in which the clearness index was
utilized. Also, Bashahu [39] confirmed that the performance of
cloud cover based models was close to those of both sunshine
duration based and clearness index based models.
As it is clear from the above brief appraisal, introducing an
appropriate set of inputs for diffuse radiation estimation in all
locations is not a viable task. As a consequence, it would be highly
important to select proper input elements for diffuse radiation
prediction for achieving more precision and less complexity.
Generally speaking, the inclusion of numerous input variables for
developing a new model will certainly bring a series of drawbacks
including its complexity because of employing high number of
required inputs, inaccuracies associated with irrelevant para-
meters, the difficulty in explaining the model and time consuming
task for collecting the required data. These factors consequently
weaken the generalization capacity of the model.
Over the previous years, the artificial intelligence and computa-
tional intelligence techniques have been broadly applied in several
scientific fields for modeling, prediction, optimization etc. Solar
radiation prediction is also among the fields of application of these
techniques. In recent years, different algorithms and techniques
such as the automatic relevance determination (ARD) methodology
[85], the niching genetic algorithms [86] and the adaptive neuro-
fuzzy inference system (ANFIS) technique [87,88] have been utilized
to determine the most relevant input variables out of various
possibly linked parameters. Therefore, an important task for future
studies should be utilization of these algorithms and techniques to
identify the most relevant parameters for diffuse solar prediction.
Moreover, the least significant parameters should also be identified.
In fact, providing comparisons between the predictions by using the
most and least relevant variables can be highly useful to highlight
the importance of appropriate selection of input parameters.
 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362
6. Conclusion
Basically, along with global solar radiation, the availability of
precise and reliable diffuse solar radiation information is undoubt-
edly requisite for designing, simulating, optimizing and monitoring
different solar energy technologies. Nonetheless, the horizontal
diffuse solar radiation is scarcely measured owing to the technical
and fiscal obstacles. As a result, a substantial number of empirical
models, by considering various mathematical functional forms and
engaging various parameters as input elements, have been pro-
posed, to estimate diffuse solar radiation on a horizontal surface in
many locations around the world.
In this study, a comprehensive review was provided on the
existing empirical models reported in the literature to estimate
horizontal diffuse solar radiation on both daily and monthly mean
daily scales for numerous locations distributed around the world.
To achieve this, all of the utilized input parameters up to now,
were recognized and the models, which have been put forwarded
yet, were categorized in two major categories based on how:
(1) the diffuse fraction or cloudiness index, and (2) the diffuse
coefficient or diffuse transmittance index have been correlated
with different parameters. Then each category was classified into
some main sub-categories based upon the employed input para-
meters for diffuse solar radiation estimation and the functional
forms employed.
Revision of the existing empirical models indicated that
empirical coefficients (constants) of the models vary remarkably
for any specific location and from one period to another one. This
convincingly illustrates that the values of these constants are
dependent upon several factors such as climatic conditions,
geographical position, monthly and seasonal variations of solar
radiation as well as characteristic of diffuse solar radiation. This
accentuate on the fact that the empirical coefficients attained for a
diffuse solar radiation model, which has been developed using
measured data of a location, can be utilized for other regions as
long as the details of their weather conditions are similar.
Appraising the literature revealed that, without due care,
employing a set of some definite input parameters with certain
mathematical functional forms to provide accurate estimations of
horizontal diffuse solar radiation for either the whole globe or a
large region (like a whole country) is not a feasible task. Further,
enhancing the estimation precision using an optimum set of inputs
and proper mathematical functional forms has to be performed
exclusively for a specifically nominated site, for which meteorolo-
gical and/or solar data are prerequisites. This owes to geographical
position, details of climatic conditions and particular features of
solar radiation which are distinct for every region.
It is worthwhile to mention that the availability of the required
input data varies from one location to another, thus the estimation
of the horizontal diffuse solar radiation should be performed
based on the available inputs at a desired location. Furthermore,
the accuracy and reliability of diffuse solar estimations in a
particular location would be dependent on both the utilized inputs
and the dependency measure of diffuse solar radiation to such
inputs. Therefore, utilization of the artificial intelligence and
computational intelligence techniques may be proper implements
to identify the most relevant parameters.
Acknowledgement
The authors would like to express their deep appreciation to
the Energy Research Institute of the University of Kashan for
supporting this study (Grant no. 1394/463851/1).
361
References
[1] Mohammadi K, Shamshirband S, Tong CW, Alam KA, Petkovic D. Potential of
adaptive neuro-fuzzy system for prediction of daily global solar radiation by
day of the year. Energy Convers Manag 2015;93:406–13.
[2] Karthikeyan S, Ravikumar Solomon G, Kumaresan V, Velraj R. Parametric
studies on packed bed storage unit filled with PCM encapsulated spherical
containers for low temperature solar air heating applications. Energy Convers
Manag 2014;78:74–80.
[3] Benmouiza K, Cheknane A. Forecasting hourly global solar radiation using
hybrid k-means and nonlinear autoregressive neural network models. Energy
Convers Manag 2013;75:561–9.
[4] Dincer I. Renewable energy and sustainable development: a crucial review.
Renew Sustain Energy Rev 2000;4(2):157–75.
[5] Kaygusuz K. Renewable and sustainable energy use in Turkey: a review.
Renew Sustain Energy Rev 2002;6(4):339–66.
[6] Bakirci K. General models for optimum tilt angles of solar panels: Turkey case
study. Renew Sustain Energy Rev 2012;16:6149–59.
[7] Thirugnanasambandam M, Iniyan S, Goic R. A review of solar thermal
technologies. Renew Sustain Energy Rev 2010;14:312–22.
[8] Ullah KR, Saidur R, Ping HW, Akikur RK, Shuvo NH. A review of solar thermal
refrigeration and cooling methods. Renew Sustain Energy Rev 2013;24:499–513.
[9] Besharat F, Dehghan AA, Faghih AR. Empirical models for estimating global solar
radiation: a review and case study. Renew Sustain Energy Rev 2013;21:798–821.
[10] Khalil SA, Shaffie AM. A comparative study of total, direct and diffuse solar
irradiance by using different models on horizontal and inclined surfaces for
Cairo, Egypt. Renew Sustain Energy Rev 2013;27:853–63.
[11] Bortolini M, Gamberi M, Graziani A, Manzini R, Mora C. Multi-location model
for the estimation of the horizontal daily diffuse fraction of solar radiation in
Europe. Energy Convers Manag 2013;67:208–16.
[12] Boland J, Huang J, Ridley B. Decomposing global solar radiation into its direct
and diffuse components. Renew Sustain Energy Rev 2013;28:749–56.
[13] Diagne M, David M, Lauret P, Boland J, Schmutz N. Review of solar irradiance
forecasting methods and a proposition for small-scale insular grids. Renew
Sustain Energy Rev 2013;27:65–76.
[14] Yao W, Li Z, Wang Y, Jiang F, Hu L. Evaluation of global solar radiation models
for Shanghai, China. Energy Convers Manag 2014;84:597–612.
[15] Mohammadi K, Khorasanizadeh H. A review of solar radiation on vertically
mounted solar surfaces and proper azimuth angles in six Iranian major cities.
Renew Sustain Energy Rev 2015;47:504–18.
[16] Gopinathan KK, Soler A. Diffuse Radiation models and monthly-average, daily
diffuse data for a wide latitude range. Energy 1995;20(7):65–7.
[17] Ulgen K, Hepbasli A. Diffuse solar radiation estimation models for Turkey’s big
cities. Energy Convers Manag 2009;50:149–56.
[18] Karakoti I, Pande B, Pandey K. Evaluation of different diffuse radiation models
for Indian stations and predicting the best fit model. Renew Sustain Energy
Rev 2011;15:2378–84.
[19] Li H, Bu X, Long Z, Zhao L, Ma W. Calculating the diffuse solar radiation in
regions without solar radiation measurements. Energy 2012;44:611–5.
[20] Boukelia TE, Mecibah M-S, Meriche IE. General models for estimation of the
monthly mean daily diffuse solar radiation (Case study: Algeria). Energy
Convers Manag 2014;81:211–9.
[21] Duffie JA, Beckman WA. Solar engineering of thermal processes. 3rd ed. New
York: John Wiley & Son;; 2006.
[22] Liu BYH, Jordan RC. The interrelationship and characteristics distribution of
direct, diffuse and total radiation. Solar Energy 1960;4(3):1–19.
[23] Kalogirou SA. Solar energy engineering: processes and systems. 1st ed.Elsevier
Inc; 2009.
[24] Iqbal M. Correlation of average diffuse and beam radiation with hours of
Bright sunshine. Sol Energy 1979;23:169–73.
[25] Barbaro S, Cannata G, Coppolono S, Leone C, Sinagra E. Diffuse solar radiation
statistics for Italy. Sol Energy 1981;26:429–35.
[26] Khogali A, Ramadan MRI, Ali ZEH, Fattah YA. Global and diffuse irradiance in
Yemen (Y.A.R). Sol Energy 1983;31:55–62.
[27] Lewis G. Diffuse irradiation over Zimbabwe. Sol Energy 1983;31:125–8.
[28] Benson RB, Paris MV, Sherry JE, Justus CG. Estimation of daily and monthly
direct, diffuse and global solar radiation from sunshine duration measure-
ment. Sol Energy 1984;32:523–35.
[29] Nagaraja Rao CR, Bradley WA, Lee TY. The diffuse component of the daily
global solar irradiation at Corvallis, Oregon (USA). Sol Energy 1984;32:637–41.
[30] Ibrahim SMA. Diffuse solar radiation in Cairo, Egypt. Energy Convers Manag
1985;25:69–72.
[31] Al-Hamdani N, Al-Riahi M, Tahir K. Estimation of the diffuse fraction of daily
and monthly average global radiation for Fudhaliyah, Baghdad (Iraq). Sol
Energy 1989;42:81–5.
[32] Jain PC. A model for diffuse and global irradiation on horizontal surfaces. Sol
Energy 1990;45:301–8.
[33] Tasdemiroglu E, Sever R. Estimation of monthly average daily, horizontal
diffuse radiation in Turkey. Energy 1991;16(4):787–90.
[34] Ahmad F, Aqil Burney SM, Husain SA. Monthly average daily global beam and
diffuse solar radiation and its correlation with hours of bright sunshine for
Karachi, Pakistan. Renew Energy 1991;1:115–8.
362 H. Khorasanizadeh, K. Mohammadi / Renewable and Sustainable Energy Reviews 53 (2016) 338–362
[35] Nfaoui H, Buret J. Estimation of daily and monthly direct, diffuse and global
solar radiation in Rabat (Morocco). Renew Energy 1993;3:923–30.
[36] Lewis G. Estimates of monthly mean daily diffuse irradiation in the south-
eastern United States. Renew Energy 1995;6:983–8.
[37] Gopinathan KK, Soler A. Effect of sunshine and solar declination on the
computation of monthly mean daily diffuse solar radiation. Renew Energy
1996;7:89–93.
[38] Trabea A. A multiple linear correlation for diffuse radiation from global solar
radiation and sunshine data over Egypt. Renew Energy 1999;17:411–20.
[39] Bashahu M. Statistical comparison of models for estimating the month-
lyaverage daily diffuse radiation at a subtropical African site. Sol Energy
2003;75:43–51.
[40] Haydar A, Balli O, Hepbasli A. Estimating the horizontal diffuse solar radiation
over the Central Anatolia region of Turkey. Energy Convers Manag
2006;47:2240–9.
[41] Mubiru J, Banda EJKB. Performance of empirical correlations for predicting
monthly mean daily diffuse solar radiation values at Kampala, Uganda. Theor
Appl Climatol 2007;88:127–31.
[42] Jiang Y. Correlation for diffuse radiation from global solar radiation and
sunshine data at Beijing, China. ASCE. J Energy Eng© 2009:107.
[43] El-Sebaii AA, Al-Hazmi FS, Al-Ghamdi AA, Yaghmour SJ. Global, direct and
diffuse solar radiation on horizontal and tilted surfaces in Jeddah, Saudi
Arabia. Appl Energy 2010;87:568–76.
[44] Said R, Mansor M, Abuain T. Estimation of global and diffuse radiation at
Tripoli. Renew Energy 1998;14:121–7.
[45] El-Sebaii AA, Trabea AA. Estimation of horizontal diffuse solar radiation in
Egypt. Energy Convers Manag 2003;44:2471–82.
[46] Tiris M, Tiris Ç, Türe İE. Correlations of monthly-average daily global, diffuse
and beam radiations with hours of bright sunshine in Gebze, Turkey. Energy
Convers Manag 1996;37(9):1417–21.
[47] Pandey CK, Katiyar AK. A comparative study to estimate daily diffuse solar
radiation over India. Energy 2009;34:1792–6.
[48] Li H, Ma W, Wang X, Lian YW. Estimating monthly average daily diffuse solar
radiation with multiple predictors: a case study. Renew Energy 2011;36:1944–8.
[49] Sabzpooshani M, Mohammadi K. Establishing new empirical models for
predicting monthly mean horizontal diffuse solar radiation in city of Isfahan,
Iran. Energy 2014;69:571–7.
[50] Page JK. The estimation of monthly mean values of daily total short-wave radiation
on vertical and inclined surfaces from sunshine records for latitudes40_N–40_S.
In: Proceedings of UN conference on new sources of energy, vol. 4, August 21–31,
Rome, Italy, United Nations, Paper no. 598; 1961: p. 378–90.
[51] Iqbal M. A study of Canadian diffuse and total solar radiation data-I. Monthly
average daily horizontal radiation. Sol Energy 1979;22:81–6.
[52] Hawas MM, Muneer T. Study of diffuse and global radiation characteristics in
India. Energy Convers Manag 1984;24:143–9.
[53] Vignola F, Mc Daniels DK. Correlations between diffuse and global insolation
for the pacific northwest. Sol Energy 1984;32:161–8.
[54] Spitters CJT, Toussaint HAJM, Goudriaan J. Separating the diffuse and direct
component of global and its implications for modeling canopy photosynthesis.
Part I. Components of incoming radiation. Agric Forest Meteorol 1986;38:217–29.
[55] Lalas DP, Petrakis M, Papadopoulos C. Correlations for the estimation of the
diffuse radiation component in Greece. Sol Energy 1987;39:455–8.
[56] Newland FJ. A study of solar radiation models for the coastal region of south
China. Sol Energy 1989;43:221–35.
[57] Veeran PK, Kumar S. Diffuse radiation on a horizontal surface at Madras.
Renew Energy 1993;3:931–4.
[58] Martinez-Lozano JA, Utrillas MP, Gomez V. Estimation of the diffuse irradia-
tion from global solar irradiation. Daily and monthly average daily values.
Renew Energy 1994;4:95–100.
[59] Tiris M, Tiris Ç, Varol HS, Türe İE. Determination of solar energy potential of
Gebze (in Turkish). Project Report (16.2.001), TÜBİTAK-MRC, Kocaeli; 1994.
[60] Jacovides CP, Hadjioannou L, Pashiardis S, Stefanou L. On the diffuse fraction of
daily and monthly global radiation for the island of Cyprus. Sol Energy
1996;56:565–72.
[61] Lam JC, Li DHW. Correlation between global solar radiation and its direct and
diffuse components. Build Environ 1996;31:527–35.
[62] Janjai S, Praditwong P, Moonin C. A new model for computing monthly
average daily diffuse radiation for Bangkok. Renew Energy 1996;9:1283–6.
[63] Hove T, Göttsche J. Mapping global, diffuse and beam solar radiation over
Zimbabwe. Renew Energy 1999;18:535–56.
[64] Kaygusuz K, Ayhan T. Analysis of solar radiation data for Trabzon, Turkey.
Energy Convers Manag 1999;40:545–56.
[65] Oliveira Amauri P, Escobedo João F, Machado Antonio J. Jacyra Soares.
Correlation models of diffuse solar-radiation applied to the city of São Paulo,
Brazil. Appl Energy 2002;71:59–73.
[66] Paliatsos AG, Kambezidis HD, Antoniou A. Diffuse solar irradiation at a
location in the Balkan Peninsula. Renew Energy 2003;28:2147–56.
[67] Jin Z, Yezheng W, Gang Y. Estimation of daily diffuse solar radiation in China.
Renew Energy 2004;29:1537–48.
[68] Jiang Y. Estimation of monthly mean daily diffuse radiation in China. Appl
Energy 2009;86:1458–64.
[69] LealeaT Tchinda R. Estimation of diffuse solar radiation in the north and far
north of Camerron. Eur Sci J 2013;9(18).
[70] Bindi M, Miglietta F. zipoli G. Different methods for separating diffuse and
direct components of solar radiation and their application in crop growth
models. Clim Res 1992;2:47–54.
[71] Tarhan S, Sari A. Model selection for global and diffuse radiation over the Central
Black Sea (CBS) region of Turkey. Energy Convers Manag 2005;46:605–13.
[72] Klein SA. Calculation of monthly average insolation on titled surfaces. Sol
Energy 1977;19(4):325–9.
[73] Erbs DG, Klein SA, Duffie JA. Estimation of the diffuse radiation fraction for hourly,
daily and monthly average global radiation. Sol Energy 1982;28:293–302.
[74] Muneer T, Hawas MM. Correlation between daily diffuse and global radiation
for India. Energy Convers Manag 1984;24:151–4.
[75] deMiguel A, Bilbao J, Aguiar R, Kambezidis H, Negro E. Diffuse solar irradiation
model evaluation in the North Mediterranean Belt area. Sol Energy
2001;70:143–53.
[76] Bortolini M, Gamberi M, Graziani A, Manzini R, Mora C. Multi-location model
for the estimation of the horizontal daily diffuse fraction of solar radiation in
Europe. Energy Convers Manag 2013;67:208–16.
[77] Collares-Pereira M, Rabl A. The average distribution of solarradiation—correla-
tions between diffuse and hemispherical and between daily and hourly
insolation values. Sol Energy 1979;22(2):155–64.
[78] Khorasanizadeh H, Mohammadi K, Mostafaeipour A. Establishing a diffuse
solar radiation model for determining the optimum tilt angle of solar surfaces
in Tabass, Iran. Energy Convers Manag 2014;78:805–14.
[79] Bakirci K. Models for the estimation of diffuse solar radiation for typical cities
in Turkey. Energy 2015;82:827–38.
[80] Li H, Bu X, Lian YW, Zhao L, Ma W. Further investigation of empirically derived
models with multiple predictors in estimating monthly average daily diffuse
solar radiation over China. Renew Energy 2012;44:469–73.
[81] Falayi EO, Rabiu AB, Teliat RO. Correlations to estimate monthly mean of daily
diffuse solar radiation in some selected cities in Nigeria. Adv Appl Sci Res
2011;2(4):480–90.
[82] Massaquoi JGM. Predicting diffuse radiation where only data on sunshine
duration is available. Sol Wind Technol 1987;4:205–10.
[83] Hussain M. Estimation of global and diffuse irradiation from sunshine
duration and atmospheric water vapor content. Sol Energy 1984;33:217–20.
[84] Karakoti I, Das PK, Singh SK. Predicting monthly mean daily diffuse radiation
for India. Appl Energy 2012;91:412–25.
[85] Bosch JL, López G, Batlles FJ. Daily solar irradiation estimation over a mountai-
nous area using artificial neural networks. Renew Energy 2008;33:1622–8.
[86] Will A, Bustos J, Bocco M, Gotaya J, Lamelas C. On the use of niching genetic
algorithms for variable selection in solar radiation estimation. Renew Energy
2013;50:168–76.
[87] Petkovic D, Gocic M, Trajkovic S, Shamshirband S, Motamedi S, Hashim R, et al.
Determination of the most influential weather parameters on reference
evapotranspiration by adaptive neuro-fuzzy methodology. Comput Electron
Agric 2015;114:277–84.
[88] Petkovic D, Cojbašic Z, Lukic S. Adaptive neuro fuzzy selection of heart rate
variability parameters affected by autonomic nervous system. Expert Syst
Appl 2013;40:4490–5.