DEPARTMENT OF COMMERCE
SINCLAIR WEEKS, Secretary
WEATHER BUREAU
F. W. REICHELDERPER, Chief
MONTHLY WEATHER REVIEW
Editor, J A M E S E. CASKEY, IR.
Volume 82 Number 8 AUGUST 1954 Cloned October 15, 1954 hued November 15,1954
PYRHELIOMETER CALIBRATION PROGRAM OF THE
U. S. WEATHERBUREAU
T. H. MACDONALD AND NORMAN B. FOSTER
U. S. Weather Bureau, Washington, D. C.
~cinuscript received May 25, 1954; revised August 26, 19.541
ABSTRACT
A new system developed for calibrating the horizontal incidence pyrheliometer is described. The pyrheliometers
to be calibrated are exposed simultaneously with a standard pyrheliometer in an integrating sphere. Calibrations
are made by comparing voltages developed by the instruments undergoing calibration with those of the standard
pyrheliometer. Calibration of the standard pyrheliometer is based on comparisons with the Smithsonian Institution
pyranometer, both out-of-doors on clear days and within the integrating sphere. Advantages of the new system in-
clude reproducibility of the calibration within less than one percent. This is due t o t h e reproducibility of the radi-
ation field in the integrating sphere, in which there are relatively small variations in ambient temperature. The
calibrations can be done much more rapidly and accurately than was formerly the case when the work was done
out-of-doors; clear skies and minimum atmospheric pollution were necessary conditions previously.
CONTENTS
Introduction__ _ _ ___ _ _____ _ __ _ - _ _ _ - _ _ - _ _ - _ _ __ __ _ _ _ ___ __ Old calibration system .................................. Linkage of calibration of field pyrheliometers to water-
flow pyrheliometers ______________ - ____ - __ ________ Error s o ~c e s __-_________________-_____-_____-_-__
Details of old method of calibrating Eppley standard and
field pyrheliometers"_ _ _ _ __ __ _ _ _ - __ - _ - - __ _ _ _ _ __ _ -
Development of new calibration procedure ________________ The pyranometer as a standard- _ ___________________
Calibration of the standard Eppley against the pyran-
ometer_________--_______-_________-___-________
Calibration with sun as source- _ ________________ Calibration in the integrating sphere _____________ Discussion of results"- ___ ______ - __ _ _ _ ____ - _ _ ______ Calibration of field pyrheliometers against the standard
The question of pyrheliometric scale .....................
Eppley-_---_---_---,-------_-,"_-_,"-_-_-------
Summary___ - _ _ _ _ _ _ - _ _ - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - - _ _ _ - _ _ _ _ _ _ - _ Appendix___ _ _ _ _ _ _ - _ _ _ _ - _ - _ _ - - _ - _ _ _ _ - - _ _ _ _ _ _ _ - - _ _ - - _ _ -
Acknowledgments- _ _ _ _ _ _ _ _ _ _ - _ _ - _ - _ _ _ _ _ _ _ _ _ _ _ - _ _ - _ _ _ _ _ References ____ _ _ - _ _ _ _ _ _ _ _ _ _ - _ _ - _ _ _ _ _ _ _ _ _ _ _ _ _ - _ _ _ _ _ _ _ _
INTRODUCTION
Page
219
219
219
220
221
222
222
222
223
224
224
224
225
225
225
226
227
The purpose of this paper is to describe a new system
of calibrating the horizontal surface pyrheliometer. In
order to show the historical continuity of the radiation
817006"M-1
activities of the Weather Bureau, a description and an+
sis of the previous calibration procedures are given.
For some time prior to 1952 the need for improvement
in the procedures for calibration of the horizontal surface
pyrheliometer was becoming acute as the radiation net-
work expanded. I n the summer of that year the Instru-
ment Division and Scientific Services Division of the
Bureau reviewed the calibration procedures then in use,
with the object of improving them beforebeginning a
recalibration program for all Weather Bureau horizontal
surface pyrheliometers.
The horizontal surface pyrheliometer used by the
Weather Bureau, originally designed by Kimball and
Hobbs [l] and now universally known as the Eppley
pyrheliometer after the manufacturer, has been calibrated
in the past by indirect reference to the Smithsonian water-
flow normal-incidence pyrheliometer [2].
OLD CALIBRATION SYSTEM
LINKAGE OF CALIBRATION OF FIELD PYRHELIOMETERS
TO WATER-FLOW PYRHELIOMETERS
The old calibration process involved a series of steps as
follows :
1. By the Smithsonian Institution: (a) Absolute cali-
219
220 MONTHLY WEATHER REVlEW AUQUET 1954
bration of the water-flow normal-incidence pyrheli-
ometer [2]. (b) Calibration of the Smithsonian
Institution silver-disk pyrheliometer by direct com-
parison with the water-flow pyrheliometer 13, 4, 51.
2. By the Smithsonian Institution and Weather Bureau
jointly: Calibration of a Weather Bureau silver-disk
pyrheliometer against the Smi thsonian Insti tu tion
silver-disk pyrheliometer.
3. By the Weather Bureau: Calibration (61 of a standard
Eppley horizontal surface pyrheliometer against the
Weather Bureau silver-disk, for use by the manu-
facturer of the Eppley in calibration of pyrheliometers.
4. By the manufacturer: Calibration of the instruments
used by the Weather Bureau in its field network, by
comparison with the standard horizontal surface
Eppley calibrated by the Weather Bureau for that
purpose.
HEROR SOURCES
In order to determine how best to improve the calibra-
tion process, the precision of the above steps was examined.
The essential facts are: The wa.ter-flow pyrheliometer of
the Smithsonian Institution has long been the fundamental
standard of pyrheliometry in the United States, and
aeems to be regarded generally as being at least as good
as any other existing instrument for the measurement of
flux density at normal incidence to the sun. In several
European countries the standard of pyrheliometry is the
lhgstrom pyrheliometer, an instrument based on physicaI
principles different from those of the water-flow.Com-
parisons have been made indirectly between the two
standards. The results indicate that the standards are
in close agreement. hgstr6m [7] indicates that the two
can be reconciled within 0.1 percent. No further exam-
ination of the water-flow will be made here. The question
of "pyrheliometric scale" is treated in a h a 1 paragraph.
Calibration of the Smithsonian silver-disk pyrheliometer
against the water-flow is reproducible with an accuracy of
the order of a tenth of 1 percent. Data are shown in
table 1, the source being page 7 of [5].
Calibration of the Weather Bureau silver-disk No. 1
against the Smithsonian silver-disk pyrheliometer (No.
A. P. 0. 8 bis) is of the same order of accuracy as that of
the Smithsonian silver-disk against the water-flow-about
one-tenth of 1 percent. Data kindly supplied by Mr.
Aldrich of the Smithsonian Institution are shown in
table 2.
TABLE 2.-Calibration of Weather Bureau silver-d!sk pyrheliometer No. 1 against Smithsonian Institution silver-drsk pyrheliometer A. P. 0. 8 bis. (Source: data supplied by Mr. Lyle B. AMrich of the Smifhsonian Institution.)
No. of values I Date { Meanconstant
0.3714 .3 m .3703 .3m .3717
.3m .3731
.3m .374a
I I
*In October 1940 new mercury was inserted in 8. I. 1 and the silver disk reblackened.
TABLE 3.-Extreme oariation about its mean calibration for each o a
set of standard Eppley pyrheliometers. (Source of baric dab: k . Hedley Greer of Eppley Laboratories.)
Differenm between extremecslibrsr
of the mean tions as a permit Standard Eppley pyrheliometer no.
-~
The standard Eppley pyrheliometer is calibrated against
the Weather Bureau silverdisk as described in [6]. Some
information on the accuracy of the results of this process
is obtainable indirectly from an analysis of the calibration
constants obtained over a period of years for the standard
pyrheliometers. Data for six instruments indicate an
average deviation about the mean of f 2 percent. Ex-
treme variations (highest value minus lowest value divided
by the mean of the constant of the individual instrument)
for each of six pyrheliometers are shown in table 3. Time
series graphs of the constants of five pyrheliometers are
shown in figure 1. (Data on which this figure is based were
kindly provided by Eppley Laboratories, Inc.)
It was originally assumed that variations in the constant
of a pyrheliometer with time arose from substantid
changes in the instrument. However, in view of the
apparent random nature of the variations shown in figure
1, it appears that such an assumption is questionable,
and that the changes may well have been associated with
circumstances of the calibration process, such as ambient
temperature, rather than with actual changes in the
pyrheliometers.
It is difficult to obtain data on the error involved in
calibration of the instruments actually used in the field
against the standard pyrheliometer. For example, data
showing calibration of one pyrheliometer against a par-
ticular standard over a series of years are not available.
We are forced to consider indirect evidence.
Calibration constants for a set of 12 pyrheliometers
called in from the field for recalibration and calibrated
originally by the manufacturer against the same standard
Eppley (the calibration constant of which had remained
unchanged during the period of calibration of the 12 field
dUQU8T 1954 MONTHLY WEATHER REVIEW 221
1 1 ' 1 1 1 "' 1 "' 1 ' "' ' ' "' 1 '
-
SUN'S ENERGY THRU TWO ATMOSPHERES
LEGEND I----\ TUNGSTEN AT 2800'K THRU 4 MM GLASS, 4CMWATER
100
). 80
m z
+ 60
W
2
w
c
-
40
a
a
-I
W
2 0
0
0.0' 0.3 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2 .o 2.2 2.4 2.6 2.8 3.0
WAVELENGTH (MICRONS)
FIOURE k.-Time series of calibration constants of flve standard pyrheliometem. (Data from Eppley hboratories. Inc.)
TABLE 4.-Comparison of original calibration constants with laboratory constants
precedii column. -0.043 is the mean of the fractional deviations. *The "random error" is here obtained by adding O.M3 to the individual values in the
TABLE 5.--Uncertainty in the steps of the old calibration system
Step Method of determining mate un- certainty degree of uncertainty
Approxi-
percent
Calibration of water-flow ....................
4 ___..do ________ __ - __ __________ Field instruments v8. standard Eppley ______
2 _____do ....................... Etandard Eppley vs. Weather Bureau
0.1 -___. do ..... -
~
- - __ - __ _. _. ____ - Weather Bureau silver-disk vs. Smithosnian
.055 Average deviation about Emithsonian silverdisk vs. water-flow"_- __
0.1 Indirectcomparison with European standard.
mean calibration.
silverdisk.
sllverdisk.
pyrheliometers) can be compared with the calibrations
recently obtained for the same set of instruments under
laboratory conditions to be described. Results are shown
in table 4. These data indicate that the original calibra-
tionsof the set were systematically about 4 percent low,
and that there was a random component of error amount-
ing to about 3 percent. (If the calibration constant is
too low, indicated radiation is too high.)
The information outlined above is summarized in table
5. No precise or elaborate statistical treatment is needed
to show the relative importance of the several error
sources in the calibration chain connecting the water-flow
standard to calibration of the Eppley pyrheliometers
used in our field network. The brief examination above
indicates clearly that improvement in the process should
begin with the calibration of the standard Eppley, and
should include calibration of the field pyrheliometers
against the standard Eppley.
DETAILS OF OLD METHOD OF CALIBRATING EF'PLEY STANDARD AND FIELD
PYRHELIOMETEFS
In the old calibration process, the standard Eppley
(which measures flux density Q on a horizontal surface
from the entire hemisphere of the sky) was calibrated
against the Weather Bureau silverdisk (which measures
flux density N from the sun, on a surface normal to the
sun's direction). We designate the e. m. f. generated by
the Eppley in measuring Q as e,. The Eppley can also
be used to measure the diffuse sky radiation flux density
D on a horizontal surface by shading the pyrheliometer
from the direct sun by means of a shade-ring or disk [SI.
The e. m. f. of the Eppley when measuring D will be
written eb.
Calibration of the Eppley against the silverdisk is
done by recording eg and eb alternately over a short time
interval during which N is also measured, during a day
having cloudlessskies. The calibration constant 0, in
millivolts/langleys* per minute, is computed from
(e,-eb)/cos i
N C=
where i is angle of incidence.
*One langley is 1 gram calorie per square centimeter.
222 MONTHLYWEATHERREVIEW AUQUST 1954
Calibration of the field pyrheliometer against the stand-
ard Eppley (the standard being one calibrated by the
Weather Bureau as just described) was done by exposing
both instruments on a clear day to sun and sky, and
obtaining simultaneous records of the outputs of each by
means of either a portable potentiometer or recording
potentiometer. This was done on a day or a set of days
selected for low atmospheric pollution as judged by
visibility, haziness, pyrheliometer recordings, etc.
In both of these steps (i. e., calibration of the Eppley
standard, and calibration of the field pyrheliometers) the
results depended to some extent on effects of angle of
incidence, ambient temperature, and atmospheric pollu-
tion. The first two of these effects are known, for ex-
ample, from [8]. That the degree of atmospheric pollution
existing at the time of calibration is a factor in the cali-
bration can be seen from the following: Whereas with no
pollution there is a minimum of solar radiation scattered
from the angular area immediately surrounding the sun,
in the presence of pollution there is considerable scatter-
ing. Further, the intensity of this scattered radiation
varies markedly with angular distance from the sun.
Therefore, under conditions of pollution the degree to
which the silver-disk "sees" the same flux as the horizontal-
surface pyrheliometer being calibrated depends on the
precision with which the angular area presented by the
shade ring or disk matches the angular area irradiating the
silver-disk pyrheliometer. Also, with pollution, the
geometry of the normal-incidence pyrheliometer becomes
pertinent [9, lo]. Further, in the presence of appreciable
atmospheric pollution the flux density is likely to vary
rapidly with time, in a random fashion; possible differences
in time constants of the two instruments gives rise to
additional uncertainty in the calibrations.
DEVELOPMENT OF NEW CALIBRATION PROCEDURE
It is apparent, then, that ambient temperature, angle
of incidence, and degree of pollution all exert some effect
on the calibration obtained for the standard Eppley, and
that in general to every set of these three elements there
corresponds a calibration constant. The magnitude of
the effects of angle of incidence and temperature are shown
in [8], but we have not measured the magnitude of the
pollution effect. A new calibration procedure should take
account of these elements in the calibration of the Eppley
standard and in calibration of the pyrheliometers used in
&he radiation network. The two phases of the process
are treated here as separate problems.
THE PYRANOMETER AS ASTANDARD
On May 3, 1952, the authors outlined the project to
Mr. L. B. Aldrich and the late Mr. W. H. Hoover, of the
Smithsonian Institution, and requested their advice on
the problem of improving the calibration of the standard
Eppley. Their reply stressed the adverse effects of the
shade-rings. They suggested that a substantial im-
provement in precision of calibrating the standard Eppley
couldbe expected through the use of the Smithsonian
pyranometer [l 1, 121, and gave convincing argumentsl to
that effect.
The pyranometer is a compensation-type pyrheliometer
which can be exposed to measure normal-incidence flux
density when shielded from sky radiation by means of an
apertured tube such as is used on the silver-disk and
water-flow. The instrument can also be used to measure
total solar and sky radiation on a horizontal surface by
removing the tube. I t is equipped with a precision-ground
hemispherical quartz envelope of uniform thickness. The
tube is constructed with the object that the angular area
about the sun "seen" by the instrument when measuring
only direct solar radiation should match as closely m
possible that of the silver-disk against which it is calibrated.
An examination of the data on calibration of the pyre-
nometer against the Smithsonian silver-disk at Mount Wil-
son at normal incidence, kindly supplied by Mr. Aldrich,
showed a mean deviation from average calibration amount-
ing to about 0.2 percent of the mean calibration. Table 6
contains calibration data for the pyranometer.
TABLE 6.-Calibrations of pyranometer no. 1.6 against silver-disk A. P. 0. 8 bis. (Basic data source: Mr. Aldrich, Smithsonian Institution)
Date
1 Number of 1
Mean con-
1 Deviation comparisons stant from mean
-I I-
""
.030/18.460 equals about 18.460 '
It was decided then to accept the advice of Mr. Aldrich
and Mr. Hoover, and to take advantage of their kind offer
to cooperate in our program by furnishing and operating
the pyranometer, and a procedure for calibration of a
standard Eppley against the pyranometer was arranged.
CALIBRATION OF THESTANDARDEPPLEYAGAINST THE PYRANOMETER
The requirements that the radiation field and ambient
temperature be standardized and pollution effectssup-
pressed in the calibration process suggest that laboratory
processes are desirable. Early in the study of the prob-
lem, it appeared that the use of a radiation integrating
sphere would provide a solution since the great precision
in measurement of distances generally involved in optical-
bench setups is not necessary in the integrating sphere.
The uniform flux density constituting the radiation field
in the sphere is desirable since possible directional effects
AUQUST 1954 MONTHLY WEATHER REVIEW 223
due to small irregularities in the glassenvelope and de-
tector sensitive surface are integrated. Ambient tempera-
ture varies over but a small range since the sphere is in a
well-insulated building; and pollution effects are elimi-
nated. The authors decided to carry out a program of
calibration of a standard Eppley out-of-doors under
clear-sky conditions on days of minimum pollution and to
compare the results with calibrations of the same Eppley,
against the pyranometer, in an integrating sphere. An
excellent integrating sphere located at the National Bureau
of Standards, Washington, D. C., was found to be avail-
able.
Calibration with sun as source.-Through the good
officesof Dr. W. F. Shenton, head of the mathematics
department of American University, Washington, D. C.,
permission for use of the attic and roofof Hurst Hall on
the university campus was kindly granted by the president
of the university, Dr. H. R. Anderson. The exposure
there is one of the best available in the Washington area,
being almost completely free from obstructions to the
nearly flat horizon, and having relatively few important
local sources of atmospheric pollution. Six Eppleys were
installed on the roof and leads were run down a venti-
lating shaft to the attic in which the auxiliary measuring
apparatus was installed. Mr. Aldrich and Mr. Hoover,
who handled all details of the pyranometer work, installed
the pyranometer within about 2 feet of the center of the
cluster of pyrheliometers. Their auxiliary measuring
equipmentwasalso installed in the attic. Since one of
the pyrheliometers exposed (No. 1973) previously had
been subjected to numerous tests to determine its charac-
teristics, it was chosen to be the standard. The others
were exposed to provide a margin against damage to
No. 1973.
Operation of the pyranometer requires that the detector
unit be exposed to the sun for 20 seconds by the flipping
openof a solenoid-operated lid and that the immediate
deflectionof a ballistic-type galvanometer be read. The
lid is then closed and electric power is metered to a strip
of the detector in quantity sufficient to produce a galvan-
ometer deflection matching that observed when the unit
was exposed to sun and sky. The observer then records
the current, as read from a suitable meter. (Full details
of the pyranometer are given in [l l , 121.) To coincide
with an observation made with the pyranometer, a read-
ing of the pyrheliometer was made at the instant the lid
to the pyranometer was flipped open. A system of
buzzers made it possible to synchronize the observrttions
of pyranometer and pyrheliometers satisfactorily. Buzzer
and solenoidwere controlled automatically by timing
switches. Two observers operated thermocouple switches
to connect the pyrheliometers to portable potentiometers.
Each observer began a series of readings on the three
pyrheliometers assigned to him when the buzzer signaled
that the pyranometer lid had opened. The sequence of
readingof the pyrheliometer voltages was reversed by
each observer on each successive calibration to avoid hav-
ing a systematic time displacement between any of the
pyrheliometer readings and those of the pyranometer.
A series of readings was taken every 2% minutes. I t took
an observer about 40 seconds to read the 3 pyrheliometers.
Readings were taken both in the forenoon and afternoon.
Readings of ambient temperature were obtained by
means of a resistance thermometer, the sensing element
of which was shielded from the sun and mounted among
the pyrheliometers. A Wheatstone bridge unit, connected
to the thermometer element, was mounted near one of
the portable potentiometers in the attic.
When these arrangements had been completed, a day
of clear weather, with a level of atmospheric pollution as
low as might be expected in any reasonable length of time,
was awaited. Such a day occurred on August 25, 1952,
and the first measurements were made then. On that
date 97 calibration sequences were taken, spanning a
range of solar elevations from 62' to 24'. The following
day was also suitable and 106 sequences were taken span-
ning solar elevations from 61' to 24'. Other observations
were obtained on January 26 and February 9, 1953, but
the elevation angles were so small that the data were of
only secondary importance; e. g., testing temperature
effects. Processing of data consisted of the following
steps for each sequence:
1. Tabulation of Eastern Standard Time of the
2. Computation and tabulation of the corresponding
3. Computation and tabulation of solar altitude.
4. Tabulation of the voltage output for each of the
pyrheliometers.
5 . Tabulation of the corresponding flux density
measured by the pyranometer.
6. Determination of the calibration constants by
dividing millivolts output by flux density to
obtain millivolts/langleys per minute.
beginning of the sequence.
apparent solar time.
7. Tabulation of ambient temperature.
The above provided the desired fundamental data.
Various studies were made attempting to reconcile
laboratory data on cosine response and ambient tempera-
ture with similar data obtained in the calibrations at
American University-"sun calibrations". These were
not entirely successful, and indicated the possibility of a
cosine-responseeffect in the pyranometer and suggested
the form and amount. Subsequent tests on the pyranom-
eter at the Smithsonian Institution did not bear out the
details of the suggested cosine curve, but did indicate
that there was a small cosineeffect in the pyranometer
and that the best results in calibrating the Eppley against
the pyranometer using the sun as source could be expected
a t low values of angle of incidence, corresponding to those
at which the pyranometer had been calibrated against
the silvor-disk. Accordingly, to represent the sun calibra-
tion of Eppley standard No. 1973, an average calibration
over the angle-of-incidence interval 28' to 60' (28' being
the smallest angle of incidence at which data were ob-
MONTHLY W E A T H E R R E V I E W AUQTJST 1954
tained, and the span 28' to 60' being an interval of nearly
constant small slope in the cosine curve) was computed
as representative of the constant for the midpoint of that
interval-i.e., angle of incidence 44'. As mentioned,
these data were obtained on August 25 and 26, 1952.
Temperature conditions during the two days were quite
similar. Average ambient temperature was about 75' F.
(While [SI shows that the temperature effect is not quite
h e a r , departure from linearity introduces no more than
a trivial difference from the mean temperature obtained
b y a n arithmetic averaging.) The calibration figure
obtained is :
Pyrheliometer No- - - __________________ - - 1973.
Calibration Constant (millivolts/langleys per
minute) - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 2.40.
Angle of incidence ____ ___________________ 44.
Ambient temperature .................... 75" F.
Calibration in the integrating sphere.--While the cali-
brations were being carried out at American University,
arrangements were made for carrying out calibration work
at the National Bureau of Standards integrating sphere,
through the generous cooperation of Mr. Ray Teele, of
the Radiometry Section of the Division of Optics, in the
Materials Testing Building of the National Bureau of
Standards in Washington, D. C.
The integrating sphere (see [13] for theory) is about 15
feet in diameter. It is arranged with hinges and casters
to open along a vertical diametrical plane, giving ready
access to the interior. The walls of the interior are painted
with a special highly reflective magnesium oxide paint.
The radiation source is a tungsten-in-glass lamp of 2,500
watts, the voltage to which was controlled during our
operations by means of a variable transformer, spanned
by an accurate a. c. voltmeter. Temperature of the
flament was estimated by Mr. Teele as about 2,800' K.
On two separate occasions (March 23,1953, and April 1,
1953) the pyranometer and pyrheliometer No. 1973 were
set up in the integrating sphere. Calibrations obtained
were 2.343 and 2.334, respectively, a t ambient tempera-
tures 85'-90' F. The mean of thess calibrations is 2.34
to the nearest hundredth. This figure can be corrected
to ambient temperature 75' F. by reference to data in
181 ; the resulting constant is 2.36. Summarizing]
Out of doors at American University, calibration minute
at ambient temperature 75" F_- - __ - - - _____ - __ 2.40
Integrating sphere:calibration]reduced to 75" F" 2.36
rnillivoltallangley~
These values disagree by about 1.7 percent.
DISCUSSION OF RESULTS
The difference of 1.7 percent in the calibrations obtained
in the integrating sphere and out of doors requires com-
ment. Conditions in the integrating sphere differ in two
essential respects from those existing at American Uni-
versity during the work there. They are (1) the spectral
distribution of radiation from the radiation source (sun
and sky a t American University vs. the 2,800° K. incan-
descent tungsten-in-glass lamp in the integrating sphere),
and (2) radiation "field" (uniform diffuse flux density in
the sphere vs. largely unidirectional radiation from an
angle of incidence varying from 22O to 65' at American
University).
If the pyrheliometer and pyranometer differ with
respect to spectral response over the intervals of the
spect,rum involved in the two tests, the results might be
reflected in a discrepancy between the calibrations of the
pyrheliometer obtained in those tests. Anexperiment
performed in the laboratory (described in the appendix)
indicated that such an effect would influence the calibra-
tion by not more than about one-tenth of one percent.
Evidently the observed discrepancy of 1.7 percent cannot
be ascribed to spectral considerations.
Differences in the radiation field can influence the
calibration of the pyrheliometer due to the differing cosine
response characteristics of the two instruments. If suffi-
ciently comprehensive data were available for both
instruments, it mould be possible to compute the magni-
tude of the effect and so verify the hypothesis that the
discrepancy arises from cosine characteristics. This
would require data for angles of incidence and azimuth
corresponding to the solar paths "seen" by each instru-
ment during the tests since, especially in the pyrhelio-
meter, there is a small random variation in calibration
with azimuth as well as angle of incidence. It hardly
seems feasible to do this at present. The cosine response
data available for the pyrheliometer suggest that the
integrator calibration should be slightly lower than the
out-of-doors calibration and it is believed likely that a
large part of the 1.7-percent discrepancy arises from the
pyrheliometer cosine characteristic and cannot be sup-
pressed without substantial (and probably expensive)
improvements in the pyrheliometer.
The question now arises as to which calibration to
assign the instrument. In the aggregate, it seems
likely that more recordwillbe taken with the sun
obscured by clouds than with the sky clear and the
sun near 45' (or any other specified) angle of incidence,
The former condition corresponds very roughly to the
sphere calibration conditions, the latter to the American
University calibrations. On this basis, calibration under
the diffuse radiation of the integrating sphere calibration
seems preferable and is taken as definitive. Ananalysis
of the cosine response of pyreheliometer No. 1973 together
with the Moon-Spencer 1141 cloudy-sky radiation distri-
bution also indicates the sphere calibration is preferable
for cloudy-sky conditions.
CALIBRATION OF FIELDPYRHELIOMETERSAGAINST THE STANDARD EePLEY
Use of the integrating sphere makes it possible for two
men to easily calibrate a dozen pyrheliometers in a day
against the standard Eppley. Experience to date on
about 50 pyrheliometers has shown that without exception
the calibration is reproducible to better than one percent
accuracy.
AUQU8T 1954 MONTHLY WEATHER REVIEW 225
THE QUESTION OF PYRHELIOMETRIC SCALE
The concept of “pyrheliometric scale” arises in the
following way. The water-flow pyrheliometer [2] was
established in the period 1910 to 1913 as the standard
pyrheliometer by the Smithsonian Institution. From
time to time since then, improvements have been made in
thewater-flow pyrheliometer, each of whichled to an
estimation of the systematic error involved in previous
measurements with the water-flow and hence of instruments
calibrated against it, directly or indirectly. However, in
order that measurements with field instruments should
be kept comparable over the years, in general, this sys-
tematic error, while recognized, was not corrected for in
thefield measurements. This has proved to be a sound
policy, in view of the several changes in “scale” which
would otherwise have greatly complicated comparisons of
observations over a period of years.
The Weather Bureau has observed such a policy in its
field measurements [6, p. 4181, and while there are several
arguments pro and con on the problem of changing to a
basis eliminating all known systematic error insofar as
possible, the policy of the Weather Bureau has not been
changed and for the time being at least the basis of
calibration remains the “Smithsonian scale of 1913.”
Present opinion [5, p. 71 is that the 1913 scale is 2.5 percent
too high. This indicates that field measurements cali-
brated to the 1913 scale give data systematicallg2.5percent
too high.
SUMMARY
A system for calibrating the standard Eppley against
the pyranometer has beendevised and placed in effect.
A new system for calibration of t.he pyrheliometers used
in the network has also been devised and placed in effect.
The latter system, which involves comparison calibrations
against a standard Eppley in an integrating sphere,
makes it possible for two men to calibrate a dozen pyrhelio-
meters a day with reproducibility accurate to better than
one percent. Previously, twomencould calibrate only
about 3 pyrheliometers a day, and then only during the
rareoccasionswhen skies were clear and atmospheric
pollution a t a minimum, and with repeatability of about
3 percent.
The new calibration linkage fromwater-flow normal-
incidence pyrheliometer to field pyrheliometer is now the
following :
1. By the Smithsonian Institution: (a) Absolute calibra-
tion of the water-flow normal-incidence pyrheliometer.
(b) Calibration of the Smithsonian 1nstitut.ion silver-
disk pyrheliometer by direct comparision with the
water-flow pyrheliometer. (c) Calibration of the
pyranometer by direct comparison with the Smith-
sonian silver-disk pyrheliometer.
2. By the Smithsonian Institution and Weather Bureau
jointly: Calibration of the standard Eppley against
the pyranometer.
3. By the Weather Bureau: Calibration of pyrheliometers
used in the radiation network in the integrating
sphere against the standard Eppley.
The calibrations are based on the 1913 scale, as heretofore.
APPENDIX
In the case of the calibration of the Eppley standard
against the pyranometer, the method consists essentially
of a determination of the flux density Q by means of the
pyranometer and the millivolts e , simultaneously generated
by the Eppley, both instruments being exposed to the sun’
and sky. The calibration constant Cis then
C=e,lQ
A s a c i e n t condition that a calibration under a radiation
source other than sun and sky be accurate when measuring
sun-and-sky radiation is that e,/& as determined under the
calibration source should be the same as e,/& measured
under sun-and-sky radiation source. It is desired then to
determine whether e,/& given with the tungsten lamp as a
radiation source is the same as e,/Q with the sun as the
source.
It is difficult to make a direct comparison, by attempting
to obtain a calibration from the two sources at identical
angles of incidence. I t is very much simpler to obtain
ratios by using the unfiltered tungsten lamp as a source
first, and then by using the same source with a water filter,
as is described in the following experiment:
The pyranometer and pyrheliometer were mounted on
opposite sides of a precision turntable, above whichwas
mounted a tungsten lamp similar to the one used in the
integrating sphere and with the same voltage across its
terminals. The turntable was equipped with a vernier
for precisely positioning the detectors.
With the lamp turned on (power was fed through a volt-
age stabilizer and regulator), the pyranometer and pyrheli-
ometer were successively placed beneath the lamp and
readings taken. A water-cell filter, consisting of about
one-eighth inch of glass and 4 cm. of water, was introduced
into the radiation beam and readings were repeated. The
calibration constants computed under the two radiation
fields agreed within one-tenth of 1 percent. This indicates
that the relatively large amount of radiation coming in at
wavelengths greater than 1 micron from the unfiltered
tungsten lamp introduces no error greater than 0.1 percent
into the calibration of the pyrheliometer.
Figure 2 shows spectral distribution of radiation from
the tungsten lamp, with and without the water filter,
together with solar radiation spectral distribution [15].
The first two were computed from spectral distribution of
radiation data for tungsten at 2,800° I(. (estimated by
Ray Teele as being the temperature of the tungsten lamps
used during the calibration in the integrating sphere) to-
gether with liquid water transmission data [16] and trans-
226 MONTHLY WEATHER REVIEW AUGUST 1954
+6
o\o
I I I I I I I I I I 1
- NO. 245 50 JCT. -
c5 1 1 I I L E G E N D I I I I I
Y
z
l-
----
9 9 362 IO JCT
--_ -_
9 9 389 50 JCT 0 +4
a
-
a
m +2 - .-
a
a 0
5 -
- .......... ' 9 395 I O JCT
-
A
0
z
W
0 0
0
-
z 0 -2 a -
LL
W - -
a 3 -4 I- -
a
a
-
WEATHER BUREAU * a - -
-6 I I I I I I I I I I I I I I I I I I
1931 1933 1935 1941 1943 1945 1947 I949 1951
YEAR,
FIGURE 2.-Normalized transmission curves of solar radiation and tungsten-lamp radiation with and without the water filter.
mission data for glass.* The water cell reduces the pro-
portion of radiation from wavelengths greater than 1
micron from 87 percent to 33 percent-i.e., from more
than solar radiation amount to less than that for solar
radiation.
Further examination of figure 2 shows that the intro-
duction of the water-glass filter into the radiation beam
from the tungsten filament lamp does not produce radia-
tion matching the solar radiation, but that the tungsten is
displaced somewhat to the long-wave side of the solar
distribution curve. The question arises as to the possible
differences in ratio of response of the pyranometer and
pyrheliometer under these two slightly different spectral
distributions.
The coverings of the two instruments are different.
The pyranometer is covered with a quartz envelope,
whereas the Eppley cover is glass. The receiving surface
of the pyranometer is finely divided carbon as is the black
receiving element in the Eppley. There is no surface in
the pyranometer corresponding to the white receiving
annular ring of the Eppley, under which the cold junctions
of the thermopile are located. It appears from this that
if there are differences in relative response of the pyranom-
eter and pyrheliometer with the change in spectral dis-
tribution in question, they must arise from differences in
'Glass transmission data by Corning Glass Works, National Bureau of Btandards, and
General Electric Go.
relative transmission of the two covers, or from variations
in the spectral reflectance of the white element in the
Eppley, or from both.
An examination of the data indicates no measurable
change in the relative transmissions of glass and quartz from
0.4 to 1 micron. Data for spectral reflection of MgO by
Middleton [17] show no considerable change in spectral
reflectivity of MgO in the region. It appears that the
spectral effects on the calibration do not exceed about 0.1
percent.
ACKNOWLEDGMENTS
We wish to thank Mr. Ray Teele, of the National Bureau
of Standards, for making available for our work the inte-
grating sphere and auxiliary equipment; and Dr. W. F.
Shenton and Dr. H. R. Anderson for providing facilities
at American University. We extend thanks also to Eppley
Laboratories and particularly Mr. Hedley Greer for pro-
viding needed data; to Dr. Sigmund Fritz for reading the
manuscript and suggesting the inclusion of a paragraph on
the scale of pyrheliometry; Mr. C.L. Mateer, of the
Canadian Meteorological Service, for contributing to the
editing; and Mr. Benjamin LeBlanc for extensive compu-
tational work. The project would have beenimpossible
without the participation of Mr. Lyle B. Aldrich and the
late Mr. W. H. Hoover of the Smithsonian Institution.
Auaum 1954 MONTHLY WEATHER REVIEW 227
REFERENCES
1. H. H. Kimball and H. E. Hobbs, “A New Form of
Thermoelectric Recording Pyrheliometer,” Monthly
WeatherReview, vol. 51, No. 5, May 1923, pp.
239-242.
2. C. G. Abbot and L. B. Aldrich, “An Improved
Water-flow Pyrheliometer and the Standard of
Radiation,” SmithsonianMiscellaneousCollections,
vol. 87, No. 15, Nov. 11, 1932.
3. C. G. Abbot, “The Silver-Disk Pyrheliometer,” Smith-
sonianMiscellaneousCollections, vol. 56, No. 19,
1911.
4. L. B. Aldrich, “The Abbot Silver-Disk Pyrheli-
ometer,” SmithsonianMiscellaneous Collectims,
vol. 111, No. 14, 1949.
5. W. H. Hoover and A. G. Froiland, “Silver-Disk
Pyrheliometry,” SmithsonianMiscellaneousCol-
lections, vol. 122, No. 5, 1953.
6. I. F. Hand, “Review of United States Weather Bureau
Solar Radiation Investigations,” MonthlyWeather
Review, vol. 65, No. 12, Dec. 1937, pp. 415-441.
(See p. 424.)
7. Aaders hgstrom “Actinometric Measurements,”
Compendium of Meteorology, American Meteoro-
logical Society, 1951, pp. 50-57. (Seep. 53.)
8. T. H. MacDonald, “Some Characteristics of the
Eppley Pyrheliometer,” MonthlyWeatherReview,
POI. 79, NO. 8, Aug. 1951, pp. 153-159.
9. L. Bossy et R. Pastiels, “fitude des propri6tes fonds-
mentales des actinom&tres,” Memoirs de Imtitut
RoyaleMdtdorologique de Belgique, vol. 29, 1948.
10. M. Nicolet, “Quelques problames foundamenfaux
pour les etudes de la radiation,” Jouml Scientijqw
de la Mdte’orologie, Janvier-Maxs 1949.
11. C. G. Abbot and L. B. Aldrich, “On the Use of the
Pyranometer,” Smi.&+onian Miscellaneous Collec-
tions, vol. 66, No. 11, Nov. 1916.
12. C. G. Abbot and L. B. Aldrich, “The Pyranometer:
An Instrument for Measuring Sky Radiation,”
SmithsonianMiscellaneousCollections, vol. 66,
No. 7, May 1916.
13. John W. T. Walsh, Photometrg, Constable and Co.,
Ltd., London, 1926, pp. 205-235.
14. R. G. Hopkinson, “Measurements of Sky Luminance
Distribution at Stockholm,” Journal of theOptical
Society of America, vol. 44, No. 6, June 1954, pp.
455459.
15. Parry Moon, “Proposed Standard Soh-Radiation
Curves for Engineering Use,” Journal of the Franklin
16. R. J. List, “Smithsonian Meteorological Tables,”
6th Revised Edition, Smithsonian Miscellaneozcs
17. W. E. K. Middleton, “The Absolute Spectral Diffuse
Reflectance of Magnesium Oxide,” Journd of th6
Optical Society of America, vol. 41, No. 6, June 1951,
pp. 419-423.
Instit&, VO~. 230, NO. 5, NOV. 1940, pp. 583-618.
. Collection, vol. 114, 1951.