Solar Energy, Vol. 18, pp, 259--263. Pergamon Press 1976.

Printed in Great Britain

TECHNICAL NOTE

The relationship between diffuse, total and extra terrestrial solar radiation

STANTON E. TULLER
Department of Geography, University of Victoria, British Columbia, Canada V8W 2Y2

(Received 16January 1975; in revised[otto 3 November 1975)

The number of stations recording the flux of total solar radiation summer solstice. Cloudiness is extensive in the summer with 75
on a horizontal surface K ~ has increased rapidly in recent years. per cent mean cloud cover in July and 67 per cent for the
Many practical problems, however, require a knowledge of the March-October period[5]. Completely clear days are rare[6].
magnitude of the diffuse component D as well as the total solar Goose has an east coast, continental climate. It is located on
intensity. If both the diffuse and total solar radiation are known, Melville Sound approximately 150 miles inland from the Atlantic
the equally important direct beam component S can be easily Ocean. Mean annual cloud cover is high (70 per cent). The
obtained via subtraction. maximum of 76 per cent occurs in June and the minimum (63 per
Despite its obvious utility, the magnitude of diffuse beam cent) in January and February [5].
radiation is measured at only a few locations. This has led to the Montreal and Toronto both possess a mid-latitude continental
development of various methods for estimating the diffuse beam climate and are representative of major urban centers. The Jean
when only the total is known. Many of these, however, consider Brebeuf station in Montreal is located in a predominantly
only clear sky conditions (e.g. [1]). Liu and Jordan [2] developed a residential area on the western flanks of Mount Royal which
simple method that is applicable for all degrees of cloud cover. separates it from the central business district and industrial areas
They equated the ratio of diffuse and total solar radiation D/K of the city. The large open areas on Mount Royal and the
(herein called the diffuse ratio) and the ratio of total and predominantly residential nature of the area mean that air
extraterrestrial radiation K $/Ko. Liu and Jordan referred to the pollution would not be as extensive as in many other areas of the
latter ratio as the cloudiness index since it is largely controlled by city. Annual cloud cover in Montreal is 62 per cent and reaches a
the amount of cloud cover. The value of the index varies inversely maximum in early winter. The general climate of Montreal is
with cloudiness, however, so the term clearness index seems more described by Longley[7] and Powe[8].
appropriate and will be used here. The extraterrestrial radiation Two Toronto stations that report both total and diffuse solar
can readily be computed or determined from published tables or radiation are included in Monthly Radiation Summary for
nomograms [3] for any station where the total is measured. This different time periods. Toronto Scarborough, located to the east of
allows the determination of the diffuse component from either a the central business district near the shore of Lake Ontario, was
graph or regression equation relating the two ratios when both included until September 1967. Since that time Toronto
K $ and Ko are known. Meteorological Research Station (TMRS) located in the north-
Liu and Jordan suggested their results would be applicable for western suburbs has been used.
practical use at stations with similar atmospheric dust content and The effect of the station change was tested by comparing five
surface albedo as those for which their relationships were complete years of mean monthly data from each station.
developed. Choudhury[4] tested their results, which were based Scarborough (1%2--67) had a lower average diffuse ratio and a
on mid-latitude data, at New Delhi and found general good higher average clearness index than TMRS (1968-73). Differences
agreement although his experimental data showed a higher diffuse in the means were significant at the 2.5 and 5 per cent levels,
ratio for a given clearness index (Figs. 2 and 3). Choudhury respectively. The results are counter to those of Yamashita[9]
attributed the difference to the greater amount of dust in the air who found a less polluted atmosphere and a higher influx of total
over New Delhi and to possible errors inherent in his small sample solar radiation at TMRS. The differences, therefore, may result
size. He suggested that until more experimental evidence became more from a temporal variation in cloud cover than to the actual
available the median between his curves and those of Liu and station change. The average annual number of bright sunshine
Jordan be used in tropical regions. hours recorded at the Bloor Street station in downtown Toronto
The purpose of the present note is to extend the study of this decreased from 2166 in the 1962-67 period to 2011 in 1968-73. The
method of approximating diffuse beam solar radiation to the upper former is 142 hr above the 1930-60 normal and the latter 13 hr
middle latitudes through the use of several years of mean monthly below normal. The difference in means is significant at the 2.5 per
data and one year of daily data recorded at four stations in Canada cent level.
which represent a variety of climatic regions as well as both highly The mean monthly diffuse ratios computed via regression
urbanized and remote localities. analysis for the two Toronto stations differ by only 2.0 per cent in
METHODOLOGY
Table 1. Stations
Hourly, daily and monthly mean daily solar radiation data for
several stations in Canada are published by the Atmospheric Station Lat~tudo Long£tude Elevatinn L a n a t h Of r e c o r d
Environment Service in its Monthly Radiation Summary. Four ueed for t h i s study
o ,~. o 'l. (m)
locations have the required data for both total and diffuse solar
radiation and are used in this study (Table 1).
The values of solar radiation are recorded to the nearest 0.1 ly Toronto 43 43 79 14 157 1961 - aul~ust 1967
Scarboroueh
and are rounded off to whole units for publication. The
International Pyrheliometric Scale of 1956 is used. Toronto 43 ~ 79 33 194 Sept. 1967 - 197}
M@teorolotical
The stations represent a range of atmospheric environments. eolearch
Station
Resolute is a high latitude station located on the southern shore of
Cornwallis Island in the Canadian Arctic Archipelago. It receives Montreal 4~ 30 73 57 ~32 196~ - 1973
Jaan Brebouf
no direct sunlight from early November to early February and
Gooae 53 19 60 2~ 44 1962 - ~973
continuous daylight from the beginning of May to mid-August.
Only the March-October period is considered here. The sun is Ruolute 74 43 94 ~9 64 1961 - 1973
always low in the sky reaching a maximum altitude of 38° on the

259
SE Vol. I8,No. 3--G
260 Technical Note

the middle range of the clearness index where the bulk of the Table 2. Regression equations and correlation of diffuse radiation
observations are located. This is well within the accuracy of the transmission coefficient (D/Ko) as a function of clearness index
present method of determining diffuse radiation. The results from (K ~/Ko) on clear days
the two Toronto stations, therefore, are pooled to give a single
overall relationship covering the period 1961-73. A discussion of S o l a r Constant
Station ( l y min -1 ) • b r S.E.
the climate of Toronto is given by Smith[10].
The extraterrestrial solar radiation on a horizonal surface was
calculated for every day of the study period using a computer Hu=p Mountain, N.C. 2.00 -.42 .38
program [11]. Monthly means were then computed from the daily Montreal 2.00 -.79 .68 -.79 .O28
values. Two solar constants were used in the calculations. In the 1.95 --79 .69 "'78 .030
first instance a value of 2.001y min -~ was used to allow
Toronto 2.00 --55 .55 -.70 .042
comparison with the results of Liu and Jordan who utilized this
1.93 --55 .56 -.70 .Oh5
value and adjustments were made to account for differences in
Pyrheliometric Scales. It is suggested for actual working Oooee 2.00 -.41 .47 -.60 .045
relationships that the results using the more currently accepted 1.93 -.41 .48 -.61 .044
solar constant of 1.95 ly min-' be employed. R, e o l u t e 2.00 .27 .08 +.20 .038
Values of the diffuse ratio D/K ~ and clearness index K $/Ko 1.95 .2? .O8 +.20 .O60
were calculated and graphed in order to determine the diffuse ratio
£11 i t • • i o n 8 combined 2.00 -.49 .51 -.52 .0.~
as a function of the clearness index. Inspection of the plotted data
indicated that a linear relationship would provide an adequate fit, 1.93 -.49 .52 -.52 .054

especially in the middle range of the clearness index where the

bulk of the data points were concentrated. The best fit linear
regression line was calculated by the method of least squares. clear days in the Arctic Archipelago is very small during the polar
Two time scales were investigated. The monthly mean day and the scatter in the data is quite pronounced. Solar angles
incorporated all months with adequate data from 1961 to 1973. are low at the latitude of Resolute. It may be that, in the range of
Individual daily data, for both clear days and all weather types the clearness index considered here, as the transmission of total
combined, from twelve complete months spanning 1972 and 1973 solar radiation increases so does that of the diffuse beam that
were also used. A clear day was defined as one with 80 per cent or makes up an important proportion of the total. The diffuse beam
more of the possible bright sunshine. made up 39 per cent of the total solar radiation at Resolute on the
clear days investigated in this study. This compares with 28 per
RESULTS cent at both Toronto and Goose and 22 per cent at Montreal.
Daily data--clear days
Figure 1 and Table 2 show the relationship between the Daily data--all weather types
transmission of diffuse D/Ko and total K $/Ko solar radiation for The major variable which determines the distribution of total
clear days. The closer the line lies to the x-axis the lower the and diffuse solar radiation is the amount and type of cloud cover.
amount of diffuse beam and the greater the direct beam radiation The advantage of a method such as that tested here for
that make up the total. approximating diffuse radiation is that it considers the effect of
The regression lines for the Canadian stations are arranged in a clouds. The diffuse ratio will increase and the clearness index
general latitudinal sequence, illustrating the effect of solar angle or decrease as cloud cover becomes more extensive. It is expected,
optical air mass. The lower diffuse radiation at Montreal could be therefore, that a suitable relationship will exist between the two
a result of the cleaner atmosphere at the Jean Brebeuf station ratios at a given locality[2].
compared with Toronto, and to a slightly lower humidity. Liu and The linear regression lines and equations computed for the four
Jordan's relationship was derived from data taken at Hump Canadian stations are given in Fig. 2 and Table 3. The correlation
Mountain, North Carolina. The lower latitude (36° 08') and higher coefficients are significant at the 0.1 per cent level. The curve for
elevation (1463 m) would be expected to reduce the proportion of Blue Hill, Massachusetts, determined by Liu and Jordan and that
diffuse radiation to a value below that found at the Canadian for New Delhi by Choudhury are also included. These show a
stations used in this study. nearly linear trend in the middle range of the clearness index and
The positive slope of the line for Resolute is difficult to explain were theoretically extended so that the diffuse ratio would
and may result from the very small sample size. The number of approach 1.0 when the clearness index fell to near 0.0 under heavy
overcast.
The regression line for Montreal is very close to that of Liu and
o, I Jordan. The other Canadian regression lines indicate a higher
.......................Liu and Jordan
--- Resolute proportion of diffuse radiation for a given clearness index as
........... Goose
.... Toronfo latitude or atmospheric pollution increase.
.... Man•rear
All combined

Table 3. Regression equations and correlation of daily diffuse

ratio as a function of clearness index--all weather types
Solar Cons~lnt
Station (1~ = i n " ) • b r S.E.
0.2 " x ~

..... X Montreul 2.00 -1.30 1.14 --95 .098

1.95 -1.26 1.14 --95 .098
Toronto 2.00 -1.32 1.20 -.95 .O96

""'"'"'"'.. I 1.95 -1.28 1.20 -.95 .0~

GOOSe 2.00 -1.33 1,23 -.92 .114

xx 1*95 -1.29 1.23 -.92 .114

0 Resolute 2.00 -1,23 1.36 -.82 .133
0 55 060 0.65 070 0.75 080 0.85 0.90 0.95 1.00
1.95 -1.23 1.38 -.82 .133
Clearness, K t / K ' o &ll atationa combined 2.00 -1.23 1.20 -.88 .158

1.95 -1.20 1.20 -.88 .139

Fig. 1. Relationship between the transmission of diffuse and total
solar radiation on clear days.
 Technical Note 261

.9

........... e o o ~ '~,,,-,,,,~'%'t...,.~.
o.2 ~ ~ ~~ , ~ _ _ .......7.~.........
~,, ~ ' ~ --~'~...x~

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 J.O

Clearness index, /(~/K o

Fig. 2. Relationship between daily diffuse ratio and clearness index, all weather types.

The high amount of diffuse radiation at Resolute is controlled by Table 4. Regression equations and correlation of monihly mean
two factors. The first is the low sun angle which will produce a daily diffuse ratio as a function of clearness index
higher proportion of diffuse radiation on clear or lightly overcast
days. The second is the thin vertical depth and lower liquid water Btatte~
Solar Const~t
( 1 5 m i n -1 ) • b r 8.E.
content of clouds in the Arctic region[12, 13]. The average
transmission of diffuse radiation D/Ko is 0.34 at Resolute
compared to 0.20 at Toronto and 0.23 for the combined value of all Montreal E,OO --93 .91 -.77 .O51

four Canadian stations. It appears that for general usage, where 1.95 -.91 .91 -.97 .o51
extreme accuracy is not essential, that the median between the Toronto 2,00 -.¢// .95 -.82 .0.5~
New Delhi and Liu and Jordan curves would provide the broadest 1.95 -.94 .95 -.82 .o~
range of application.
Oooee 2 .CO -.66 .87 -.60 .065
The linear trend in the data used to compute the regression
1.95 -.65 .86 -.60 ,o6~
equations in Table 3 is most apparent in the middle range of the
clearness index from about 0.30-0.70. Below this interval the Resolute 2,00 -.68 .98 -.66 .090
scatter in the data points increases as the diffuse ratio equals 1.0 1,95 -.66 .98 -.66 .090
for a wide range of the clearness index. This decrease in accuracy All stations ¢ ~ b i n e d 2~00 -.6% .8~ -.52 .09Z
at the lower values of the clearness index should be kept in mind 1.95 -.62 .8q -.52 .092
in practical applications. The equation to calculate diffuse
radiation in actual usage would take the form:

D = a (K ~ 21Ko)+ bK ~ (1) the middle range of the clearness index. Ninety-seven per cent of
or the values are between 0.30 and 0.70 and 69 per cent between 0.40
D = K J, (aK ~ IKo + b). (la) and 0.60. The extreme range is 0.26-0.71. The lines, therefore, are
only applicable in this interval and extension beyond is not valid,
It is recommended that for practical application the equations or indeed, seldom needed.
using a solar constant of 1.95 ly min t be used. The correlation coefficients are lower for the monthly period
than for the daily period, but the standard errors of estimate are
Monthly data also smaller. The smaller variation of the data points about the
The monthly mean of diffuse radiation is often sufficient for regression line is offset by the smaller range of clearness index.
general planning and in purely climatological studies. Regression Given the sample size the correlation coefficients are still
lines and equations for the relationship between the diffuse ratio significant at the 0.1 per cent level.
and clearness index on a mean monthly basis are given in Fig. 3 The seasonal differences in diffuse and total radiation caused by
and Table 4. variations in solar angle, atmospheric dust and water vapor and
The majority of the data points for the Canadian stations lie in cloud type can produce a range of diffuse ratio for a given

.o-
"6

0.4 ~ u r Y "'~'~ -~tk4~. "~..

~ o r d o n ~",,~ -'t~b.. ~.,...
r • R e s o l u t e - , . , . , , , ~ _ ~'~d.... I
...........Goose .....~ . ~ ~ .
o.a I .... ~ ~ . , ~
.... Montreal "~, " n ~ _ J
All combined ~ i

0L . . . . . . . . . "...,'~.~
0.1 0.2 0.5 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Clearness index, /(t #(o

Fig. 3. Relationship between monthly mean daily diffuse ratio and clearness index.
262 Technical Note

clearness index. Certain months at a given station tend to occupy the year as a whole the variation in atmospheric parameters
preferred positions on the graphs (Fig. 4). At Toronto the winter assumes the greater importance.
months have a higher diffuse ratio and lower clearness index
reflecting the winter maximum of cloud cover and low solar RELIABILITY

angles. Autumn shows a lower proportion of diffuse radiation than The use of relationships between diffuse ratio and clearness
winter for a similar clearness index. The clearness index values index provides a rather simple method of determining the relative
for October and January are 0.4481 and 0.4497 respectively, but proportions of direct and diffuse solar radiation at a station where
the diffuse ratios are 0.4783 and 0.5832. Although the diffuse ratio only the total is available either from measurement or via
during the spring (April) and summer (July) are nearly the same, calculation. Once the proper relationship has been determined for
the summer has a higher clearness index. a region the only other piece of input data which is needed is the
The scatter in the data is greater for Goose and Resolute than extraterrestrial solar radiation. The present study has shown,
for the two lower latitude stations, and the proportion of diffuse however, that no single form of the equation is applicable to all
radiation is higher. The slopes of the lines at Goose and Resolute atmospheric environments.
are very similar indicating an analogous response of the diffuse The reliability of the method must be evaluated in terms of both
ratio to a change in the clearness index. Toronto and Montreal the scatter of individual data around the regression line and the
have a steeper slope indicating that the diffuse radiation at these error introduced when the relationship established at one station is
stations is more sensitive to the degree of cloudiness. This is what applied at another. The former is inherent for any station. Since
would be expected at a station with a higher solar angle and a the use of the method is only needed at stations where no diffuse
greater cloud liquid water content. radiation is measured, the latter source of error will also be
involved wherever the method is used in practical application.
The effect o[ surface albedo This error is greater the more conditions at the reference station
The amount of diffuse radiation reaching the surface under a differ from the point of interest.
given atmospheric state is directly related to the surface albedo. Assuming a normal distribution of diffuse ratios for each
Some of the radiation reflected by the surface will be backscat- clearness index and with the large sample size used in this study,
tered downward to add to the total of measured diffuse radiation. approximately 95 per cent of the deviations will be within two
The importance of variations in surface albedo in determining standard errors of estimate from the regression line[14].
the amount of diffuse radiation was tested by correlating the Ninety-five per cent of the individual data points, therefore,
diffuse radiation transmission coefficient D/Ko and albedo at should be within 0.10-0.18 of the value predicted from the
Toronto, Goose and Resolute (Table 5). Montreal does not record regression line on a monthly basis and within 0.20-0.27 on a daily
reflected solar radiation. basis. Extrapolating the findings from the four Canadian stations
The positive correlation coefficients indicate the expected direct would indicate that stations in clearer climates and lower latitudes,
relationship between surface albedo and diffuse radiation. The with a lower proportion of diffuse radiation, would have less
annual peak of albedo occurs during the presence of snow cover in scatter in their data. Although the range of deviation of individual
the winter. This coincides with the season of maximum cloud data points around the regression line is large and prediction of
cover, lowest solar angles and maximum pollution, however, so diffuse radiation for an individual day or month may differ as
that no direct cause and effect relationship can be proved. In much as 40-50 per cent from the actual value, the results over the
general the correaltion coefficients are not large and only 27 per long term would be much more consistent. Design requirements
cent of the variance in diffuse radiation transmission coefficient is and climatological studies are often more concerned with the
explained by variations in albedo for monthly mean data at all long-term mean for a particular month or day. It is recommended
stations combined. It is safe to conclude that surface albedo does that the method presented here is best used to ascertain these
have some effect on the amount of diffuse radiation, but through long-term expected values. If used to predict a particular day or
month because of the lack of other data, the degree of possible
variation should be kept in mind.
An indication of the degree of error introduced in transferring
x Jon~ory i
0.8 * Ap~,l the relationship established at one station to another where the
~- • Jury
"~ ~
o
•
Oct0t~
Norther diffuse component is not measured is indicated by an inspection of
07 Figs. 2 and 3.
, • ee°q • The regression lines for the mid-latitude Canadian stations
OS (Montreal, Toronto and Goose) differ by 0.08 on a daily basis and
o .~ xo ~-c5- between 0.04 and 0.12 on a monthly basis in the middle range of
00 • --o *
05 the clearness index (0.3-0.6). For daily values adopting
Choudhury's suggestion of using a median position between his
O.4 curve and that of Liu and Jordan (very close to the Toronto line)
would give a diffuse ratio within 0.07 of the extremes on either
I side and should be practical for most situations in populated areas
0"3i
from the tropics to the upper middle latitudes. Maximum error
02 0.3 04 0.5 0.6 0.7
would be between 9 and 23 per cent depending on the clearness
Clearness index, K¢/i~"o
index. The Toronto line would provide a better median for the
monthly period, the error would range from about 8 per cent at a
Fig. 4. Distribution of data points of the relationship between clearness index of 0.3 to 22 per cent at 0.6.
monthly mean daily diffuse ratio and clearness index for selected In summary, the use of the relationships between the diffuse
months at Toronto. ratio and clearness index is one method of approximating diffuse
radiation when only the total is known. Its utility stems from the
Table 5. Coefficients of correlation between diffuse solar radia- fact it is useful over a wide range of cloud conditions. It is hoped
tion transmission coefficient and surface albedo that the present study will allow the limitations and applications of
this method to be better appreciated and will stimulate its further
Station Clear days All days H o n t h l y s e a Ds testing in an even wider variety of climatic regimes.

Toronto -.02 +.19 +.~ Acknowledgements--The author wishes to thank Dr. Jen-hu
Goose +.2~ +.62
Chang of the University of Hawaii, who originally proposed this
line of research and provided useful encouragement during its
Resolute +.6~ +.35 +0~
progress.
&11 stations oomblned +.28 +.40 +°22 REFERENCES

1. G.V. Parmelee, Irradiation of vertical and horizontal surfaces

 Technical Note 263

by diffuse solar radiation from cloudless skies. Heating, 8. N. N. Powe, The Climate of Montreal. Queen's Printer,
Piping and Air Conditioning 26, 129 (1954). Ottawa (1969).
2. B. Liu and R. C. Jordan, The interrelationship and characteris- 9. S. Yamashita, Air pollution study from measurements of solar
tic distribution of direct, diffuse and total solar radiation. radiation. Arch. Meteor. Geophys. Bioklim. 21B, 243 (1973).
Solar Energy 4, 1 (1960). 10. R. J. Smith, The Climate o[ Toronto. Canada Dept. of the
3. R. List, Smithsonian Meteorological Tables, 6th Rev. Edn., Environment, Toronto (1972).
pp. 418-19. Smithsonian Institution, Washington, D.C. (1968). 11. G. M. Furnival et al., Solar climate calculator, Proc. 3rd Forest
4. N. K. D. Choudhury, Solar radiation at New Delhi. Solar Microclimate Symposium, pp. 181-187. Canadian Forest
Energy 7, 44 (1963). Service, Calgary, Alberta (1970).
5. Canada, Dept. of Transport, Climatic Normals, Vol. 3, 12. E. Vowinckel, Cloud Amount and Type over the Arctic.
Sunshine, Cloud, Pressure and Thunderstorms. Canada, Dept. AFCLR-62-663, Air Force Cambridge Research Laboratories,
of Transport, Toronto (1968). Bedford, Mass. (1962).
6. R. W. Rae, Climate of the Canadian Arctic Archipelago. 13. S. I. Sivkov, Computation of Solar Radiation Characteristics.
Canada Dept. of Transport, Toronto (1951). Israel Program for Scientific Translations, Jerusalem (1971).
7. R. W. Longley, The Climate o[ Montreal. Queen's Printer, 14. P. G. Hoel, Elementary Statistics, p. 149. Wiley, New York
Ottawa (1954). (1960).