You are on page 1of 11

Energy and Buildings, 6 (1984) 293 - 303 293

Analysis o f A t m o s p h e r i c T u r b i d i t y f o r D a y l i g h t C alculations
M. NAVVAB, M. KARAYEL, E. NE' EMAN and S. SELKOWITZ
Energy Efficient Buildings Program, Lawrence Berkeley Laboratory, University of California, Berkeley, CA
94720 (U.S.A.)
A large set of illuminance and irradiance
data has been collected for four years at 15-
mi nut e intervals in San Francisco. This data
set has been used to investigate the impact of
atmospheric turbidity on daylight calcula-
tions. Existing predictive formulae for Li nke
turbidity, TL, provide moderate agreement to
measured values of T L when using nominal
design values for the ~ngst rSm scattering co-
efficient, ~, and precipitable water vapor, w.
When average measured values for [J and w are
used, the agreement improves. We suggest the
use of an illuminance turbidity, T~, to cal-
culate direct normal illuminance directly. We
derive a simple approximate solution, T~ =
1 + 21.6 [3. Til appears to be a better param-
eter to describe atmospheric conditions since,
unlike TL, it is insensitive to air mass and thus
solar altitude or time of day. We present and
compare plots of T~ and TL vs. solar altitude,
time of day, and month. Finally, we examine
and compare several alternative pat hways to
derive direct normal illuminance from irradi-
ance and luminous efficacy (dependent on fl
and w), or directly from ft.
INTRODUCTION
Predicting the illuminance from di rect sun
and sky on clear days is essential t o any st udy
of daylighting strategies in buildings. During
t he past t wo decades, a number of researchers
have devel oped predictive models t hat
account f or at mospheri c t urbi di t y. These
models quant i f y t he influences of at mo-
spheric aerosols, gases, and wat er vapor on
di rect and diffuse radi at i on at t he Eart h' s sur-
face. The models enable researchers t o esti-
mat e irradiance or illuminance on a given sur-
face at a specific t i me, day, and l ocat i on.
Al t hough t heoret i cal l y derived algorithms
for illuminance dat a can be devel oped, at mo-
spheric scattering and absorpt i on processes
are sufficiently compl ex and microclimate-
dependent t hat a measured dat a base is essen-
tial. Researchers have r epor t ed results of avail-
ability studies in Australia, Japan, Sout h
Africa, and several European countries, but
t here is a pauci t y of dat a for t he United
States.
For most daylight calculations with clear
skies, it is essential t o know t he mean clarity
of t he at mosphere, which is commonl y ex-
pressed in t erms of t urbi di t y fact or. The cur-
rent l y used paramet ers are Linke' s t urbi di t y
fact or (TL) and the Angst rbm coeffi ci ent (fl).
In this paper we compare these calculation
met hods with our measurement s and intro-
duce a new concept , which we call illu-
mi nance t ur bi di t y (Tu).
DATA COLLECTION
Data were col l ect ed in San Francisco, Cali-
fornia, at 38 nor t h latitude, 123 longitude
within t he Pacific t i me zone. The station is
situated on a peninsula t hat separates t he
large San Francisco Bay from t he Pacific
Ocean. Sea fog and low cl oudy skies in t he
early mornings are characteristic of this area.
There is a great variation of local climates
within t he Bay Area [1]. The data were col-
lected on t op of t he Pacific Gas and Electric
(PG&E) building in t he city' s financial district
at 140 m (450 ft) above sea level. The
i nst rument at i on at t he station consists of
seven illuminance sensors, of which t wo
measure global and diffuse illuminance on a
hori zont al plane and f our measure vertical
illuminance at each cardinal ori ent at i on. The
0378-7788/84/$3.00 Elsevier Sequoia/Printed in The Netherlands
294
seventh is used as a luminance sensor for
zenith luminance measurements. Two pyrano-
meters measure global and diffuse irradiance
on a horizontal surface. For detailed informa-
tion about the station and the t ype of data
collection system, see refs. 2 and 3. Addi-
tional short-term measurements made at this
location are described later in this paper.
THEORETI CAL BACKGROUND
The solar flux in t he direct beam t hat
reaches the Earth' s surface as a funct i on of
wavelength, )~, is given by t he Bouguer-
Lambert law. Direct normal irradiance is cal-
culated by t he integral over the entire range
of wavelengths (~):
1
/ Ee o( ~) exp{--a(}Qrn} d~ (1)
E ~ s n - S
0
where E~o(~) is the solar const ant at the mean
sun-Eart h distance as a funct i on of }~, E~n is
the direct normal irradiance at t he Earth' s sur-
face, m is the absolute air mass, S is a param-
eter t hat normalizes for t he variations of E+o
within t he sun-Eart h distance as defined
later, and a(X) is the overall ext i nct i on coeffi-
cient per unit air mass as a funct i on of X. The
absolute air mass is a product of a pressure
correction fact or and a geometric fact or
called the relative air mass, rnr (e~):
m = (p/lOOO)m~(es) (2)
where p is the pressure in mb and e~ is the
angular distance of t he sun from the zenith.
For e~ < 80 , mr can be approxi mat ed by
SEC(e, ), where SEC(es) is t he slant height
t hrough a plane-parallel atmosphere [4]. For
larger angles a more exact formul a should be
used (see eqn. (7)). The normal i zat i on param-
eter, S, is given as the square of t he ratio of
t he actual sun-Eart h distance, Ra, to t he
mean sun-Eart h distance, Rm [5, 6]:
R:
s - (3)
The overall ext i nct i on coefficient per unit air
mass, a(X), has a number of complex consti-
t uent terms. In a perfect (non-absorbing)
Rayleigh atmosphere, scattering occurs from
molecules of atmospheric gases t hat are much
smaller t han t he wavelength of light; this
produces a strong wavelength dependence
(1/~ 4) in the scattering coefficient. However,
it is well known t hat even in the clearest
atmosphere, there is significant additional
scattering and absorption due to atmospheric
aerosols and ot her natural and man-made
atmospheric component s. Accordingly,
several models have been developed to predict
quantitatively t he at t enuat i on of direct beam
radiation in a turbid atmosphere. We review
these in t he sections t hat follow, describe the
results of measurements and calculations us-
ing each model, and comment on the relative
merits of each.
MODELS OF TURBI DI TY
Li nke turbidity factor
The intensity of solar radiation reaching
t he Earth' s surface t hrough a turbid atmo-
sphere depends on several scattering and
absorption processes in which gases, water
vapor, dust, and aerosols are the major
variables in addition to solar altitude. Equa-
t i on (1) can be rewritten with ar, t he mean
ext i nct i on coefficient integrated over all
wavelengths, where ~r accounts for scattering
and absorption processes. Linke proposed to
express ~r in terms of the product of a Ray-
leigh scattering term for an ideal clear atmo-
sphere, (xr, and the Linke t urbi di t y factor, TL,
where T L can be interpreted as t he number of
equivalent Rayleigh atmospheres required to
produce t he same extinction. Thus,
Ee~n = Eeo exp(--drm) (4.1)
and
Ees n = Eeo exp( - ~r m TL) (4.2)
where T L > 1. T L can t hen be calculated from
eqn. (4.2):
T L = (ln Eeo -- In E+sn)(oLrm) -1 (5)
where E+o = 1370 W/m 2 and
a~ = (9.4 + 0.9m) -1 (from Kasten, ref. 7) (6)
--sin 7s + [sin27s - - 1 + (I. 001572)2] j~2
m =
and
Ys = 9 0 - - e s
0. 001572
(from Mahotkin, ref. 7) ( 7)
The integrated Rayleigh scattering coeffi-
cient, ~, depends primarily on m. The Linke
t urbi di t y factor, TL, depends on precipitable
water vapor and scattering aerosols as well as
on m. Al t hough TL is useful for comparing
atmospheric properties for a variety of clear
sky conditions, it has one drawback: measure-
ments suggest t hat , under const ant clear con-
ditions, T L varies with m (see, for example,
ref. 5). This shows t hat the absorption and
scattering processes based on aerosol and
water vapor cont ent are also wavelength-
dependent . Thus TL by itself is not a good
i ndi cat or of t he atmospheric aerosol and
water vapor concent rat i ons t hat influence
daylight availability.
The dependence of TL on m creates a
problem when analyzing long-term solar
irradiation data. In t he analysis of our data,
we used t he criterion, recommended by t he
World Meteorological Organization (WMO),
t hat a clear sky condi t i on exists when the
direct normal irradiance equals or exceeds
200 watts per square met er (W/m2). This
irradiance corresponds to t he mi ni mum inten-
sity required to scorch t he paper strip of a
Campbel l -St okes sunshine recorder. The
criterion limits the maxi mum value of T L for
clear conditions and makes it a funct i on of
solar altitude. Equations (5) and (6) can be
solved for TL with Ees n/> 200 W/m 2 as a func-
t i on of solar altitude (through m), resulting in
eqn. (8):
TL(max) = (ln Eeo -- In 200)
9. 4+0. 9m
rr/
( 8 )
These results are presented in Fig. 1. They
suggest t hat a comparison across seasons of
calculated values of TL based on measured
dat a may be misleading because t he maxi mum
value of TL obtainable at low altitudes is in
part an artifact of t he clear sky radiation
criterion. Changing the clear sky ' defi ni t i on'
to Ees . / >400 W/m 2, so as to analyze only
very clear days, shifts the maxi mum TL line
lower, as shown in Fig. 1, furt her constraining
t he ' allowable' TL at low solar altitudes.
Furt hermore, any threshold t ype definition
of a ' clear sky' will still allow some cirrus
cloud cont ami nat ed dat a into t he ' clear' con-
ditions. Direct visual observation or addi-
tional test criteria (e.g., diffuse/global ratio)
may help in furt her analysis. This suggests
295
t hat alternative clear sky definitions may be
desirable.
Dogniaux [ 8] and Valko [ 9] have bot h
suggested empirical fits of the Linke t urbi di t y
fact or (TL) as a funct i on of aerosol and water
vapor cont ent and relative air mass.
17, 5" . . . . . . . . . . . ] b~i Ckl l ~ , ~ i < ........... : .......... ' ......... ~ .......... ; .........
I ~ . . . . . . . . . . . i . . . . . . . . . . i . . . . . . . . i . . . . . . . . . . ~ : > ~ ~ . . . . . . . . . . i . . . . . . . . . . i . . . . . . . . .
5-
2. 5.
0
Fig. 1. Maximum turbidity as a function of solar alti-
tude. A, 200 W/m2; X, 400 W/m2.
Dogniaux' s formula [8], valid for 5 < % <
65 , is:
[ + 8 5 ) ]
TL = [(39.-~-e--~ + 77.4) + 0.1
+ (16 + 0.22w}~ (9)
Valko' s formul a [9], valid for 5 < % < 65 ,
is:
TL = (B + 0. 54)[1. 75 l og( w/ m + 0.1) + 14.5]
- - 5 . 4 ( 1 0 )
The variables in bot h equations are: TL =
Linke' s t urbi di t y factor, m = air mass, "Ys =
solar altitude (deg.), ~ = AngstrSm t urbi di t y
coefficient (see below), w = water vapor con-
t ent in the atmosphere (cm), and B = 1.07 fl
(by Schuepp) [ 9]. Dogniaux provides nominal
annual average values of fl and w t hat can be
used to estimate T L for sites where fl and w
have not been measured. San Francisco is an
urban t emperat e zone, which gives (from
Dogniaux) f l =0. 1 and 4>w ( c m) >2. We
calculate TL at 15-minute intervals from our
measured data, using eqn. (5) and values of
Ees, derived from measured global irradiance,
measured diffuse irradiance, and shadow band
296
cor r ect i ons based on ref. 11. Fi gur e 2 pl ot s
our mont hl y aver aged meas ur ed t ur bi di t i es
agai nst val ues cal cul at ed f r om Dogni aux' s and
Val ko' s f or mul ae wi t h/ 3 = 0.1 and w = 2. The
f or mul ae do not give cons t ant val ues over t he
year because t he di st r i but i on of cl ear sky
sol ar al t i t udes changes over t he year . As an
annual s umma r y of t he adequacy of t he for-
mul ae, we have cal cul at ed t he average
mont hl y r at i os TL ( cal cul at ed) / TL( measur ed)
and s t andar d devi at i on t o be 1. 19- + 0. 03
( Dogni aux) and 1. 18 + 0. 04 ( Val ko) . Fi gure 2
shows t hat single mont hl y r at i os will di f f er by
a l arger f act or .
! i : : i i
! ........ . . . . ! ...... ..... i
W . . . . . . i . . . . . . . . i . . . . . . . . i . . . . . . . ~. . . . . . . . . ! . . . . . . . . ~. . . . . . . . . ! . . . . . . . . . !- . . . . . - $ . . . . . . . . i . . . . . . .
"1 2 .......... Legend ..... ~ ......~ .................
~ o ~ 1 4
Fig. 2. Monthly average values of Linke turbidity as
measured vs predictions from eqn. (9) (x) Dogniaux
[8] and eqn. (10) (~) Valko [9] using 13 =0.1, w=
2.0. ~ = 200 W/m ~.
I f we r epeat t he cal cul at i on usi ng w = 4 cm,
t he f or mul ae give a less accur at e answer in
compar i s on t o our meas ur ed dat a. The
TL( cal cul at ed) / TL( meas ur ed) r at i o ~or
Dogni aux is 1. 27 and f or Val ko 1. 29.
To bet t er de t e r mi ne t he accur acy of t he
basic f or mul ae, we need addi t i onal dat a on
and w. In t he next sect i ons, we di scuss how
t hese par amet er s can be es t i mat ed f r om
meas ur ed dat a. We t hen agai n c ompa r e t he
f or mul ae t o our meas ur ement s .
f i n g s t r 6 m t ur bi di t y c oe f f i c i e nt , fl
The Angs t r Sm t ur bi di t y coef f i ci ent , fl,
pr ovi des a means of est i mat i ng t he scat t er i ng
due t o aer osol s. The overal l ext i nct i on coeffi -
ci ent , a(~.), can be wr i t t en as a wei ght ed sum
of t hr ee ext i nct i on coef f i ci ent s:
m r m r
a( k) = ~a ( k) + - - ~D( k) + - - ~W( ) ~) (11)
/ 7l / 72
wher e a R (~) is t he Rayl ei gh ext i nct i on coeffi -
ci ent per uni t m, absol ut e air mass; ~w( k) is
t he sel ect i ve abs or pt i on coef f i ci ent f or gases
wi t hi n t he at mos pher e ( 02, O3, H20, CO2,
et c. ); and ~D( k) is t he aer osol ext i nct i on coef-
fi ci ent . The aer osol ext i nct i on coef f i ci ent is
wei ght ed by mr / m because t he dust concen-
t r at i on depends on t he pat h l engt h but not
t he air pr essur e, p. Thi s rat i o is also used t o
wei ght t he abs or pt i on coef f i ci ent f or gases be-
cause of t he domi na nc e of wat er vapor ( whi ch
is i ndependent of p) in t hi s t er m.
Accor di ng t o Angs t r bm t he aer osol ext i nc-
t i on coef f i ci ent can be cal cul at ed by t he fol-
l owi ng f or mul a:
aD(~) = ~/ ()~) (12)
wher e ~ is t he Angs t r 6m t ur bi di t y coef f i ci ent ,
and t he wavel engt h e xpone nt a depends on
t he size of at mos pher i c part i cl es. The mos t
c ommonl y used val ue is a = 1. 3, whi ch was
r e c omme nde d by Angst r 5m. Gi ven a val ue f or
~, it is necessar y t o det er mi ne ~D in or der t o
fi nd/ 3.
For our dayl i ght i ng appl i cat i ons, t he ef f ect
of aw(h) on t he t ot al ext i nct i on coef f i ci ent
can be l argel y el i mi nat ed by not i ng t hat
aw(}~) is domi na t e d by abs or pt i on of r adi at i on
by wat er vapor , whi ch occur s pr i mar i l y in t he
far red and i nf r ar ed. Thus t he Rayl ei gh coef-
fi ci ent and aer osol ext i nct i on coef f i ci ent can
be es t i mat ed f r om me a s ur e me nt s of t he sol ar
i r r adi ance in t he visible regi on of t he spec-
t r um. To ma ke t hese meas ur ement s , we used
a Schot t RG2 fi l t er havi ng a nomi nal c ut of f at
630 nm. The fi l t er was mount e d on a nor mal -
i nci dence pyr he l i ome t e r wi t h a sun- t r acki ng
syst em.
Coul son pr ovi des a fi gure t o det er mi ne
( f or a = 1. 3) as a f unct i on of meas ur ed i rradi -
ance ( f or h < 630 nm) and air mass, rn (Fig.
3. 5 in ref. 6). To f aci l i t at e comput er - as s i s t ed
anal ysi s of our meas ur ed dat a, we used
Coul son' s dat a t o fi t t he aer osol coef f i ci ent ,
a D , as a f unct i on of/3 and m:
c~ D - fla~R (13)
(Co + d c m)
Using this equat i on for aD, we can subst i t ut e
eqn. (11) i nt o eqn. (4.1) and derive t he fol-
lowing expression for/~:
fl = [ l n ( E / E~ ) --~Rcm] [C + dc m] (14)
OLRcD2
w h e r e E is t h e extraterrestrial s o l a r c o n s t a n t
f o r )~ ~< 6 3 0 n m , E ~ n is t h e m e a s u r e d flux,
~ R c is t h e R a y l e i g h c o e f f i c i e n t for )~ ~< 6 3 0
n m ( 0 . 1 8 6 5 f r o m ref. 6), a n d C ~ a n d d c are
t h e coefficients w e d e r i v e d t o fit t h e air m a s s
d e p e n d e n c e o f t h e m e a n aerosol e x t i n c t i o n
coefficient.
B e c a u s e t h e actual c u t o f f w a v e l e n g t h , ) ~ ( ~
at w h i c h T = 5 0 % ) , w i l l v a r y f r o m t h e
n o m i n a l v a l u e (~ = 6 3 0 n m f o r t h e R G 2
filter), e a c h o f t h e w a v e l e n g t h - d e p e n d e n t
p a r a m e t e r s in e q n . ( 1 4 ) m u s t b e a d j u s t e d t o
reflect t h i s A)~ = ~ ~ 6 3 0 n m . U s i n g C o u l -
son's t a b u l a r d a t a f o r t h i s A ~ e f f e c t , w e
d e v e l o p e d a series o f e q u a t i o n s t o correct t h e
n o m i n a l values in e q n . (14):
E c = E 6 3 0 + A E A ) ~ = 4 9 9 + 1.6A)~ ( W / m 2)
(14.1)
0LRc -~ 0LR630 "{" A0~RA)~ -- 0. 1865 -- 0.00062A)~
(14.2)
C = C6a 0 + ACAh = 0. 0696 -- 0.00008A?~
(14. 3)
d= d6a 0 + AdA~ = 0. 00272 -- 0. 0000340Ah
(14. 4)
We used a spect r ophot omet er t o measure t he
spectral t ransmi t t ance of our filter and deter-
mi ned t hat A~ = 0.5 rim. Thus our correct i on
t erms were very small.
The resul t ant values of fl est i mat ed by
appl yi ng eqn. (14) t o hourl y data col l ect ed
during a 3-mont h peri od (December 1982-
Febr uar y 1983) show no obvious dependence
on air mass. Figure 3 shows t he f r equency
di st ri but i on of our results, whi ch have a mean
of 0. 088, a st andard deviation of t he distribu-
t i on, a, of 0. 03 and a st andard deviation of
t he mean of 0. 002 (o/x/-~). This compares t o
t he nomi nal value of 0.1 suggested by
Dogni aux f or urban t emper at e climates.
At mo s p h e r i c wat er vapor
Water vapor cont ent , w, for this area was
calculated from mean daffy dew-poi nt t emper-
297
atures, td, t aken from ref. 1 and are based on
measurement s from 7 am t o 4 pm. These
values were calculated using Reitan' s regres-
sion equat i on [ 10] :
In w (cm) = 0.981 + 0. 0341td (F). (15)
Water vapor can also be calculated f r om
at mospheri c pressure and dewpoi nt using an
ASHRAE algorithm (see ref. 4, chapt er 5).
40-
35-
30-
25-
20-
;3
~ - 15-
10-
5-
0.00 0.05 0.10
i
0.15 0.20
Me a s u r e d Be t a
Fig. 3. Frequency di st ri but i on of/ ~ngst rSm t ur bi di t y
coefficient (~).
2. 4-
2. 2-
1. 8-
1. 6-
1.4~
M O N T H
Fig. 4. Mont hl y average values of water vapor cont ent
of t he at mosphere for San Francisco.
This approach di ffered by no more t han 3%
from t he results in eqn. (15). Mont hl y average
values of w f r om t hr oughout t he four-year
measurement peri od are shown in Fig. 4. The
298
distribution w = 1.58 cm through 2.2 cm has
a somewhat lower average t han t he range re-
commended by Dogniaux for a typical tem-
perature climate (2 to 4 cm).
Comparison of calculated and measured TL
The calculated mont hl y average water
vapor cont ent of the atmosphere, w, and the
mean annual average/3 value derived from our
measured data and t he % for which E~n>
200 W/m 2, were used in Dogniaux' s and
Valko' s equations to calculate t he mont hl y
variation in turbidities. Figure 5 shows the
ratio of the calculated values to our measured
values. Some of t he error we found when us-
ing design values for w and ~ (see Fig. 2) is
eliminated when using more appropriate
values for w and ft. A summary measure of the
degree of the fit is again expressed by the
average ratio and standard deviation of
TL(calculated)/TL(measured) = 1.12 -+ 0.03
(Dogniaux) and 1.13 -+ 0.03 (Valko).
*~o ~ I [quotlo~ (9) ~etos.08715 wmmonthly ovg
Equation (10) b~om,O~'75 wmmo~thly avg
1.5-
i
&
I---
"
M O N T H
Fig, 5. Monthly ratio of calculated Linke's turbidity
using eqn. (9) (z~) Dogniaux [8] and eqn. (10) (m)
Valko [9] to measured turbidity, fl = 0.09, w from
Fig. 4 is monthly average.
Both formulae provide a good rough esti-
mate of measured TL. We need additional
dat a and analysis to det ermi ne if eqns. (9) and
(10) will accurately predict T L for the San
Francisco climate. There are several potential
experimental and analytical sources which
could reduce accuracy and account for the
discrepancies between measured and cal-
culated values. These include: (1) errors in
our estimates of w and fl; (2) the use of
average ~/s rather t han average air mass, m;
(3) t he shadow band correction techniques
[11]; and (4) error i nt roduced by use of the
200 W/m 2 criterion for a clear day. Changing
this criterion might add more low-altitude
points, which would lower t he calculated
mean turbidities. The effect on measured
values is less clear, so it is not certain t hat it
would improve t he formulae.
IUuminance turbidity
The concept of t urbi di t y factor has been
developed for irradiance measurements. In
the daylighting field, Linke' s t urbi di t y factor
(TL) is used to calculate irradiances t hat are
t hen converted into illuminances, using t he
luminous efficacy of radiation, K. Both T L
and K depend on m,/3, and w, which increases
the opport uni t y for error and t he compl exi t y
of t he calculation. The calculation of an
illuminance t urbi di t y, TH, provides a way
around these problems. The concept is
analogous to t hat of irradiance t urbi di t y.
Direct normal illuminances would be cal-
culated by eqn. (4.1) with the irradiance
parameters Eesn, Eeo , ~r, and T L replaced by
their illuminance counterparts, E~n, E~o, au,
and T H .
Because water vapor absorption occurs
predomi nant l y in the infrared, T~ should be
insensitive to w. In addition, because illumi-
nance strongly weights a narrow band of
wavelengths, the ext i nct i on coefficient should
be relatively insensitive t o m. We now derive
simple approxi mat i ons for the illuminance-
weighted Rayleigh coefficient, a~, and the
illuminance t urbi di t y, TH.
Because water vapor absorption is not im-
port ant for illuminances, we approxi mat e the
ext i nct i on coefficient, di~, in terms of ~H
and an aerosol term:
a H = ~ H + # l ( X ) ~ ( 1 6 )
where X = 0.5527 pm is t he mean wavelength
weighted by t he V(X) curve [6]. Solving eqns.
(4.1) and (4.2) for T~ gives:
T~ = d~,/~ (17)
Substitution of eqn. (17) into eqn. (16) gives
T ~ = 1 + ~ / a H ( X ) ~ ( 1 8 )
The r e c omme nde d val ue of t he e xpone nt a is
1. 3. An expr essi on f or an can be der i ved f r om
eqns. (1) and (4):
1 I ]
a~ = - - In (19)
m f~E1o(h) exp(--~(k)m} dX
where E,o(k ) =Eeo(h)V(h), and V(h) is the
CIE photoptic sensitivity curve. The integral
in eqn. (19) can be estimated from the best
fitting Gaussian (the method of steepest des-
cents). This gives a reciprocal expression for
ai, (see also eqns. (6) and ( 13) ) :
an = 0. 1/ ( 1 + 0. 0045m) (20)
Equat i on (20) suggest s t hat an can be as s umed
t o be 0.1 t o about 5% accur acy f or 3'~ > 5 ,
(m < 10). Subst i t ut i ng t he val ues of k, a, and
an i nt o eqn. (18) yi el ds t he f ol l owi ng si mpl e
expr essi on:
Tn = 1 + 21.6/3 (21)
Now l et us r econsi der t he pr obl e m of com-
put i ng t he di r ect nor mal i l l umi nance, E~.
Dogni aux has l i st ed desi gn val ues of fl and w
f r om whi ch desi gn t ur bi di t i es and effi caci es
are c omput e d as a f unct i on of ai r mass. Usi ng
t he i l l umi nance t ur bi di t y concept , we can re-
pl ace t hi s pr ocedur e wi t h t hr ee desi gn i l l umi -
nance ext i nct i on coef f i ci ent s: dn = 0. 21, 0. 32,
and 0. 53 f or rural , ur ban, and i ndust r i al areas,
r espect i vel y. ( Coul son quot es an uncer t ai nt y
299
on c~ of +0. 2, whi ch t r ansl at es i nt o about a
10% uncer t ai nt y on t hese desi gn val ues. ) Cal-
cul at i on of Els n usi ng desi gn val ues f or ~,
f ol l ows di r ect l y f r om t he i l l umi nance anal-
ogue of eqn. {4).
The use of annual average desi gn val ues of
a~ (or any r el at ed pa r a me t e r such as 3) ma y
pr ovi de adequat e average est i mat es f or i l l umi -
nance but will nor mal l y not be a good
i ndi cat or of di f f er ences due t o at mos pher i c
ef f ect s on a s hor t er t i me scale (e.g., hour l y,
dai l y, mont hl y) . For ma ny dayl i ght i ng cal-
cul at i ons, val ues f or i r r adi ance (Eesn) ma y be
avai l abl e so t hat i l l umi nance can be cal cul at ed
i f K, t he l umi nous ef f i cacy, is known. K, how-
ever, is a c ompl e x f unct i on of sol ar al t i t ude,
%, wat er vapor cont ent , w, and at mos pher i c
scat t er i ng ef f ect s, 3. In Tabl es 1 and 2 we
c ompa r e hour l y/ mont hl y val ues of be a m
l umi nous ef f i cacy obt ai ned in t wo di f f er ent
ways.
In Tabl e 1, we use our meas ur ed annual
aver age val ue f or 3 ( 0. 088) and mont hl y
average da yt i me val ues of wat er vapor con-
t ent t o cal cul at e T L based on eqn. (9). TL is
t hen used in eqn. (4. 2) t o cal cul at e Eesn. For
t he i l l umi nance, we use 3 t o cal cul at e Tu
(eqn. (21)), and t hen Tu is used t o cal cul at e
E~n usi ng t he i l l umi nance anal og of eqn. (4. 2).
The val ues in Tabl e 1 s how K = El s n/ Ee s n on
an average hour l y basis f or each mont h. I n
Tabl e 2, we pr esent val ues of K obt ai ned by
di vi di ng me a s ur e d beam i l l umi nance by
TABLE1
Lumi n~useffi ca~y~fdi re~t n~rma~radi at i ~nasafun~t i ~n~fs~l aral t i t udeusi ngaverage~=~. ~88
Alt. Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.
5 25 25 25 26 26 26 27 27 27
10 67 67 68 68 69 69 70 70 70
15 85 86 86 86 87 88 89 89 89
20 94 95 95 95 96 96 97 97 97
25 99 99 99 99 100 101 101 102 102
30 101 102 102 102 103 103 104 104 104
35 103 103 104 104 104 105 105 106 106
40 105 105 105 105 106 107 107 107
45 106 106 106 106 107 107 107 108
50 106 107 107 107 108 108 108
55 107 107 108 108 109 109 109
60 108 108 108 109 109 109 110
65 108 109 109 110 110 110
70 109 110 110 111 111
75 110 110 111 111 111
27 26 25
70 69 68
88 87 86
97 96 95
101 100 99
104 103 102
105 104 104
106 106
107
108
3OO
TABLE 2
Luminous efficacy of direct normal radiation as a function of solar altitude based on measured Eem and Elm
Alt. Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.
5 37 30 36 28 28 40 21 34 39
10 65 59 52 53 57 61 58 70 67
15 87 80 78 80 82 86 80 90 89
20 99 94 90 90 91 97 93 105 100
25 104 102 95 98 95 104 97 109 106
30 107 105 101 102 99 110 107 108 110
35 110 106 103 103 99 110 108 112 111
40 105 102 102 97 110 106 110 109
45 101 102 102 98 106 105 109 108
50 102 102 98 108 105 108 109
55 103 102 97 106 105 110 108
60 106 104 98 106 106 109 104
65 101 98 106 106 111 106
70 102 98 106 107 110
75 102 98 106 105 104
37 34 32
65 64 6O
89 87 88
103 100 101
109 107 108
114 112 111
116 112 114
114 112
112
113
meas ur ed be a m i r r adi ance. We not e in Tabl e 1
t hat t he use of a single annual val ue f or 3 ( and
mont hl y w) resul t s in val ues f or K whi ch
show a st r ong dependence on sol ar al t i t ude
but ot her wi s e l i t t l e mont h t o mont h varia-
t i on. The r api dl y decr easi ng beam ef f i cacy at
l ow sol ar al t i t udes is e xpe c t e d and resul t s
f r om t he wavel engt h- dependent scat t er i ng
pr oper t i es of t he at mos pher e.
Compar i ng t he meas ur ed dat a in Tabl e 2 t o
t hose of Tabl e 1 i l l ust r at es t he di f f er ences t o
be e xpe c t e d when compar i ng average mea-
sur ed hour l y/ mont hl y dat a t o annual resul t s.
May and Oc t obe r s how l ower and hi gher
val ues r espect i vel y t ha n we pr edi ct ed f or
Tabl e 1. We are t r yi ng t o det er mi ne t he
sour ce of t he at mos pher i c ef f ect s t hat caused
t hose di f f er ences.
We concl ude t hat t he t r adi t i onal appr oach
of usi ng desi gn val ues of fl and w t o cal cul at e
E~n f r om Ees n usi ng a l umi nous ef f i cacy wor ks
well as l ong as desi gn val ues are accur at el y
known. However , if 3 can be meas ur ed or
accur at el y es t i mat ed, t hen eqn. (21) can be
used di r ect l y t o cal cul at e E~n, bypassi ng t he
r equi r ement t o conver t i r r adi ance t o illu-
mi nance.
The f act t hat T~ der i ved f r om meas ur ed
i l l umi nance is essent i al l y i ndependent of m
and t hus ~'s is an advant age i n eval uat i ng
at mos pher i c condi t i ons. As me nt i one d earl i er,
a pl ot of T L agai nst t i me of day or sol ar alti-
t ude is di f f i cul t t o i nt er pr et because TL is
de pe nde nt on ~'s and at mos pher i c condi t i ons.
A pl ot of Til agai nst t i me of day or sol ar alti-
t ude gives mor e i nf or mat i on about t he varia-
t i on in at mos pher i c condi t i ons.
Fi gur e 6 pl ot s T~ der i ved f r om meas ur ed
i l l umi nance as a f unct i on of sol ar al t i t ude.
The r eader shoul d r e me mbe r t hat t he val ues
at l ow al t i t udes are bi ased due t o t he 200 W/
m 2 cr i t er i on (see Fig. 1).
Til shoul d be cl osel y r el at ed t o par t i cul at e
concent r at i ons over t he cour se of a day,
wher eas T L will be mor e de pe nde nt on chang-
ing ai rmass. Fi gur e 7 pl ot s T~ and T L as a
f unct i on of t i me of day. The resul t s i ndi cat e
t hat t her e are subst ant i al var i at i ons in T L as a
f unct i on of t i me of day due in par t t o its
dependence on air mass, whi ch vari es wi t h t he
mont h of t he year . T~ is rel at i vel y st abl e
dur i ng t he cent r al por t i on of t he day but we
also not i ced what ma y be a l ocal peak in T~
caused by af t er noon r us h- hour t r af f i c or t he
fog t hat of t en covers t he Bay in t he earl y
eveni ng. Thi s suggest s t hat i l l umi nance
t ur bi di t y ma y i ndeed be mor e sensi t i ve t o
at mos pher i c condi t i ons t han is i r r adi ance
t ur bi di t y.
The hour l y average dat a in Fig. 7 do not
s how mont hl y cl i mat i c t r ends, so in Fig. 8 we
pl ot i l l umi nance and i r r adi ance t ur bi di t i es as a
f unct i on of t i me of year . The average val ue
over t he year and st andar d devi at i on f or
i l l umi nance t ur bi di t y is 2. 6 0. 4, and f or
i r r adi ance t ur bi di t y is 3. 4 + 0. 5. The at mo-
301
.-i=
I-
El
r -
E
i i : : i ! : i i : i : : i i : i i i i i i : i : i : : ! i : i i i
i i i ! ! i i ! i ! ! i i ! i i i i i i i i i i i i i ! i i i i ! i
4 ....
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : ? ? ? : i ? ~ ~ : ? ~ ~ : : ? ? :
2~ .+.--i.--i...-i...i---i.-.i..-~-..i-..i....i..-~-...i-..~...~...i...~ .... i-.-F.i-.-.i--.i---+--.i---+--.i-.-~-.i.-.~....i.--i.--~...i...~-.,
i i l : i i l : i i i i i i i i i
1 1 : : : : 1 : ; : : : : : 1 1 :
i i i i i i i i i i i i i i i i i
1 "-'i'"'i'--!'-"i"-i'-"~" ' ' ' i'"'i---!-'-'i"'i---'i"'i'-'-:"
; : : : : : : : ; : : : : : : : :
!!!!: !!!!!!:!!!!
S o l a r A l t i t u d e ( d e g )
Fi g. 6. I l l umi na nc e t ur bi di t y, Ti], as f unc t i on of s ol ar a l t i t ude f or hour l y da t a f or all da t a me e t i ng t he ' cl ear da y'
cr i t er i on.
5 . s ............ : ........ i f i ........ ; ...... ~ < ~ ......... : .........
/ : 7 ~ : ,
~ . .
/ ~ I " ' " ~ ' ' ~ ' ~ ~ i
l s - [ i [ _ ~ ' ~ " ~ , ~ _ : . .
/
, f i i i i i
4 6 e 10 ~2 ~. 1~ ~e 20
Time of Day
Fig. 7. Average calculated illuminance t u rb idit y (~)
a nd i r r a di a nc e t ur bi di t y (X) as a f unc t i on of t i me of
da y, f or hour l y da t a f or all da t a me e t i ng t he ' cl ear
da y' c r i t e r i on.
a - '
L e g e n d
. L ~ l l l u m l n o n o T ~ r l ~ d l ~
X : l , ~ " o d l a r ~ o e T u t b l d l ' l y
M O N T H
Fi g. 8. Mont hl y aver age val ue of i l l umi na nc e Til (A)
a nd i r r a di a nc e T L ( ) t ur bi di t i e s .
spheric conditions t hat result in a maxi mum
average mont hl y value for TL in May cause an
even sharper peak in T~ for May.
As a check on the illuminance turbidities,
we can compare t he value of fl derived from
our illuminance measurement s and from
illuminance t urbi di t y, eqn. (21), to t he value
of fi we calculate from our irradiance turbidi-
ties, which are based on measured irradiance,
mont hl y average w, and either Dogniaux' s or
Valko' s fits (eqns. (9) and (10)). The
irradiance-derived fls consistently under-
predict t he illuminance fis with the average
annual ratios, r = 0.87 + 0.03 (Dogniaux) and
r = 0.82-+ 0.04 (Valko). Figure 9 shows the
mont h by mont h values and Fig. 10 shows
mont hl y ratios t hat indicate the magni t ude of
t he mont hl y differences. Furt her measure-
302
0, 14
i i i l .4" , ~ I x _ _ '~',_ _ ~,..~._ ~o,,C IQ _ .L_ I
, ) i : I
o . I o - . . . . . . . . ! . . . . . . . . I . . . . . . . . ' . ' n o ~, ~, . . , . . , _ _ _ . _ . _ . _ . _ . ~. . . . , , _ ~ . . . . . . .
....................... i ........ i ? i . . . . . . . . . i . . . . . . . i . . . . . . . . i
Fig. 9. Monthly average of the irradianee-derived
using eqn. (9) (El) Dogniaux [8], eqn. (10) () Volko
[9] and eqn. (21) (A) LBL.
1. 2
~ . t , o . ( 9 )
I " ( ' ) I
E
" ~ 0 . 8 -
u
e'-
.~ o.e-
, i r 1 , , , , , , 1
MONTH
Fig. 10. Ratio of irradiance-derived f~ to illuminance-
derived ~ using eqn. (9) (:'-~) Dogniaux [8], eqn. (10)
( m) Volko [9] with eqn. (21).
( appr oxi mat el y 12% di f f er ence) when t he
cor r ect local values of fi and w are used in t he
f or mul ae. The use of nomi nal design values
f or fi and w gives cor r espondi ngl y r ougher
agr eement ( 18 - 28% di f f er ence) . The
Campbel l - St okes cl ear sky cr i t er i on of 200 W/
m 2 affect s t he ma xi mum cal cul at ed t ur bi di t y
at ver y l ow sol ar al t i t udes. The resul t s at t hese
al t i t udes are t her ef or e not compar abl e t o
resul t s at hi gher al t i t udes. Addi t i onal cri t eri a
t o unambi guousl y def i ne cl ear sky condi t i ons
woul d be desi rabl e.
We pr esent a si mpl e equat i on f or an illumi-
nance t ur bi di t y T~ ( anal ogous t o TL) , and
suggest t hat t he use of T, may si mpl i fy some
dayl i ght i ng cal cul at i ons. Thi s i l l umi nance
t ur bi di t y also has t he advant age t hat it bet t er
i ndi cat es at mospher i c condi t i ons t han does
Li nke t ur bi di t y. The i l l umi nance t ur bi di t y has
an heuri st i c advant age in t hat an appr oxi mat e
anal yt i cal expr essi on f or it can be der i ved
di r ect l y f r om t he Angst r Sm f or mul a.
ACKNOWLEDGEMENTS
This wor k was s uppor t ed by t he Assi st ant
Secr et ar y f or Conser vat i on and Renewabl e
Ener gy, Of f i ce of Bui l di ng Ener gy Resear ch
and Devel opment , Bui l di ng Syst ems Di vi si on
of t he U.S. Depar t ment of Ener gy under
Cont r act No. DE- AC03- 76SF00098.
Special t hanks are due t o Rober t Clear of
t he Li ght i ng Syst ems Resear ch Gr oup at
Lawr ence Ber kel ey Labor at or y f or his many
hel pf ul suggest i ons and t o Moya Mel ody f or
her edi t ori al assistance.
ment and analysis are r equi r ed t o det er mi ne
how much of t he di scr epancy is due t o t he
use of mont hl y average w, t o i naccuraci es in
Val ko' s and Dogni aux' s f or mul ae, or t o mea-
s ur ement error.
CONCLUSIONS
A compar i s on of cal cul at ed T L (usi_qg ex-
pressi ons der i ved by Dogni aux and Val ko)
and measur ed T L shows good agr eement
REF ERENCES
1 National Oceanic and Atmospheric Administra-
tion, National Climatological Data, U.S. Dept. of
Commerce, 1982.
2 M. Navvab, M. Karayel, E. Ne'eman and S. Selko-
witz, Daylight availability data for San Francisco,
Energy and Buildings, 6 (3) (1984) 273 - 281.
3 M. Karayel, M. Navvab, E. Ne'eman and S. Selko-
witz, Zenith luminance and sky luminance distri-
butions for daylighting calculations, Energy and
Buildings, 6 (3) (1984) 283 - 291.
4 ASHRAE Handbook of Fundamentals, American
Society of Heating, Refrigerating and Air-Condi-
tioning Engineers, Inc., New York, 1981, Chapter
5.
5 N. Robi ns on, Sol ar Radi at i on, El sevi er Publ i shi ng
Co. , New Yor k, 1966, pp. 113 - 131.
6 K. L. Coul son, Sol ar and Terrestrial Radi at i on,
Academi c Press, New Yor k, 1975, pp. 40 - 50.
7 F. Kast en, A si mpl e par amet er i zat i on of t wo pyr-
hel i omet r i e f or mul ae for det er mi ni ng t he Li nke
t ur bi di t y f act or , Met eorol . Rdsch. , 33 ( August )
( 1980) 124 - 127.
8 R. Dogni aux, Avai l abi l i t y o f Daylight, CIE Tech-
ni cal Commi t t e e 4. 2, 1975.
303
9 S. Aydi nl i , The Avai l abi l i t y o f Sol ar Radi at i on
and Daylight, Draft Techni cal Repor t t o CIE
Techni cal Commi t t e e 4. 2, Oct ober 1981.
10 C. H. Rei t an, Sur f ace dewpoi nt and wat er vapor
al of t , J. Appl . Met eorol . , 2 ( December ) ( 1963)
776 - 779.
11 B. A. LeBar on, W. A. Pat er s on and I. Di r mhi r n,
Cor r ect i on for di ffuse i r r adi ance meas ur ed wi t h
s hadowbands , Solar Energy, 25 ( 1983) 1 - 13.

You might also like