Description: | A comparison of BSRN and SIROS solar radiometers for measuring downwelling
irradiances at the SGP central facility was made with field standards and two
absolute cavity radiometers brought to the site or a two-week period in April
1996 by Mike Rubes (formerly of the National and Oceanic Atmospheric
Administration, Air Resources Laboratory, Surface Radiation Research Branch in
Boulder, CO). A description of this effort can currently be found on the
World Wide Web at http://www.srrb.noaa.gov/apr96iop/hagsie.html. Analyses of
the data from these comparisons have resulted in several observations on the
quality of data collected at the BSRN and SIROS platforms since October 13,
1995, which are probably valid to the present time, until these sensors are
replaced with more recently calibrated sensors. On Oct. 13, 1995, the two
BSRN pyranometers (PSPs) were replaced, so the observations do not apply to
the BSRN measurements of global and diffuse irradiation before that date.
Another source of information is inspection of the SIROS and BSRN equipment by
Joe Michalsky (Atmospheric Science Research Center, State University of New
York at Albany) at various times. The results of the findings are summarized
as recommendations in the following several paragraphs. Some explanation and
further comments are provided in the parenthetical remarks.
ANALYSIS WHEN THE DIRECT BEAM WAS NOT OBSCURED BY CLOUDS
Direct-beam solar irradiance measured with the BSRN pyrheliometer (NIP) are
too large by approximately 0.5% compared to the two absolute cavity
radiometers. (This small underestimate is within the expected level of
uncertainty.)
Direct-beam solar irradiance measured with the SIROS pyrheliometer are too
small by approximately 2.1% compared to the two absolute cavity radiometers.
(This large discrepancy is unexplained and will be explored during future
calibration activities at the SGP Radiation Calibration Facility.)
Possibly the best estimate of downwelling total hemispherical solar (global)
irradiance can be made by summing the SIROS pyrheliometer irradiance reading
multiplied by 1.021 (and by the cosine of the solar zenith angle) and the
average of the readings for diffuse irradiance from the shaded BSRN and SIROS
pyranometers. The direct-beam part can alternatively be computed as 0.995
times the BSRN pyrheliometer reading. For data collected in October before
the 13th, when the BSRN shaded pyrheliometer was replaced, the diffuse
component is probably best computed directly from the SIROS shaded sensor
alone.
Downwelling total hemispherical solar (global) irradiance measured by the BSRN
unshaded pyranometer is approximately 2% too small (which is within the
expected level of uncertainty for unshaded pyranometer measurements) compared
to the values computed from the measured direct-beam and diffuse components.
(Downwelling total hemispherical solar irradiances measured by the SIROS
unshaded pyranometer systematically underestimates the global irradiances
by excessive amounts, i.e., by greater than 3%.)
The analyses leading to these recommendations are described in an extended
abstract presented in early February (J. Michalsky et al., "Optimal
Measurements of Surface Shortwave Irradiance Using Current Instrumentation--
The ARM Experience," in Preprint Volume, Ninth Conference on Atmospheric
Radiation, Feb. 2-7, Long Beach, California, pp. J5-J9, American
Meteorological Society, Boston, MA). Further relevant analyses were conducted
by Kato et al., (Seiji Kato, Pennsylvania State University) and are described
in a manuscript submitted for publication ("Uncertainties in Modelled and
Measured Clear Sky Surface Shortwave Irradiances")."
UNSHADED PYRANOMETER PERFORMANCE WHEN THE DIRECT BEAM WAS NOT OBSCURED
The above recommendations are based mostly on analyses conducted for
cloudless, midday conditions. Because the data reported from the unshaded
pyranometer were not corrected for cosine response, slight overestimates of
global irradiance from unshaded pyranometers tend to occur in cloudless
conditions at solar zenith angles less than 45 deg and slight underestimates
tend to occur for zenith angles greater than 55 deg. The maximum deviations
occur at extreme solar zenith angles and are about 2%.
TRACKER-SHADING PERFORMANCE
The data user should note, as has been noted in data release statements, that
analyses of the direct, diffuse, and/or direct beam irradiances should be
preceded by a check of sensor performances by summing the direct and diffuse
components and comparing the result to the directly measured global component.
When this is done, problems with solar tracking are usually apparent.
Because slight misalignments in the tracking and shading devices can be
difficult to detect, small deviations of the component sum from expected
behavior are sometimes difficult to explain. If such deviations tend to recur
for specific time intervals for several days, one might suspect a tracking or
shading problem. For the time period addressed here, the modern tracking-
shading assembly used with the SIROS sensors appeared to work well. For the
BSRN sensors until January 1996, an older tracking-shading system was used
that was not as reliable as the modern assembly used with the SIROS sensors;
problems with this BSRN tracking and shading system, were usually evident when
they occurred. A modern tracker-shader was installed for the BSRN sensors in
January 1996. The tracker was not aligned as well as it could be. Efforts are
underway to improve tracker alignment checks and procedures at all SIROS sites
and the BSRN site.
PARTLY CLOUD CONDITIONS
An analysis by Chuck Long (formerly at the Pennsylvania State University and
now with the University of Colorado and the National Oceanic and Atmospheric
Administration) indicated that data users who are investigating partly cloudy
sky conditions will usually find that the BSRN outputs are more reliable for
short periods of time, say less than 30 min, than are the SIROS outputs. This
tends to occur because the SIROS data are recorded only every 20 s while the
BSRN data represent one-minute averages computed on the basis of sampling once
per second. Under partly cloudy conditions, sampling only once every 20 s
tends to provide inadequate statistical representation of downwelling
irradiances.
ESTIMATES FOR CLOUDY CONDITIONS
The component sum technique is not applicable for overcast conditions. For
the time period addressed here, the SIROS shaded sensor appears most reliable
before October 13, 1995. Thereafter, an average of data from the SIROS shaded
pyranometer, the shaded BSRN sensor, and the shaded BSRN sensor multiplied by
1.02 might be the best estimate of global irradiance for cloudy conditions.
However, a rigorous analysis on the results of this procedure has not been
carried out, so the data user should approach this technique with caution.
SOME ADDITIONAL INFORMATION
The excessively large deviations noted above for the pyranometers result in
part from a mixture of different sources of calibration procedures. The
following table lists the sources of calibration:
Sensor Coefficient used Calibration Installation
to process data date date
BSRN PSP DS BORCAL Sept. 1995 Oct. 13, 1995
BSRN PSP DD Eppley June 1995 Oct. 13, 1995
BSRN NIP BORCAL July 1993 March 17, 1994
SIROS PSP DS Eppley June 1995 July 25, 1995
SIROS PSP DD Eppley June 1995 July 25, 1995
SIROS NIP BORCAL Sept. 1994 July 25, 1995
DS = downwelling solar or global
DD = downwelling diffuse
PSP = precision spectral pyranometer
NIP = normal incidence pyrheliometer for direct-beam solar
BORCAL = broadband outdoor radiometer calibration, conducted by the National
Renewable Energy Laboratory (NREL)
Eppley = denotes calibrations in an integrating sphere by the manufacturer,
Eppley Laboratory, Inc.
The BORCAL calibrations result in estimates of solar irradiances that are
typically 1.5% larger than Eppley calibrations, a situation which is under
investigation by Tom Stoffel at NREL and John Hickey at Eppley. They are
working together to document this difference. This difference helps to
explain the larger estimates of global irradiance measurement with the BSRN
sensor than with the SIROS sensor.
A greater source of concern than over differences between the NREL versus the
Eppley calibrations at this time is the insufficiently frequent recalibrations
of sensors in operation at the SGP site. Although the NIPs are expected to
hold their calibrations for rather long periods of time, the pyranometers
typically should be recalibrated at least once every 12 months. Change out
with freshly calibrated pyranometers and pyrheliometers at the SGP site will
begin in 1997, with the goal of routinely replacing every pyranometer and
pyrheliometer with freshly calibrated sensors once every year.
Data users can inspect metrics provided on the World Wide Web by the SGP site
scientist team on data quality at the following address:
http://manatee.gcn.uoknor.edu/metrics/METRICS.html
Other observations/measurements impacted by this problem:
Any derived estimates of downwelling solar radiation components using data
from central facility SIROS or BSRN sensors (for downwelling solar radiation)
for the time period indicated.
Suggested Corrections of the Problem: (e.g. change calibration factor and
recompute, flag data with this comment, etc.)
Use of these recommendations by data users. Ideally, the component sum
technique would be applied in a value-added product (VAP) implemented at the
Experiment Center, but this has not been done yet. In the meantime, users of
recent data can inspect plots of component sum technique on the World Wide Web
site noted above. |