1. Introduction
Because of the important role of radiation in the climate system, the World Climate Research Programme (WMO 1990, 1991) proposed in 1990 the establishment of a worldwide Baseline Surface Radiation Network (BSRN) for continuous long-term measurements of the highest attainable quality of radiative fluxes at the earth’s surface. About 30 stations located throughout the world are planned (two-thirds are collecting data at present) to measure upward and downward shortwave and longwave broadband irradiance. The objective of the BSRN is to provide surface radiation budget data for validating estimates inferred from satellite measurements, to verify the accuracy of radiation codes in climate models, and to detect possible long-term trends in radiative fluxes at the surface.
The performance and accuracy of broadband infrared instruments, such as pyrgeometers and pyrradiometers, to measure down- and upwelling longwave radiation has been an important issue for BSRN scientists. A number of questions were raised in this respect, especially because of conflicting information found in the literature and dithering views circulated within the scientific community. Two field tests led by J. DeLuisi have been conducted, under BSRN auspices: the first was at Coffeyville, Kansas, during the First ISCCP (International Satellite Cloud Climatology Project) Regional Experiment (FIRE II) and the other was at Boulder, Colorado, to evaluate the performance of pyrgeometers and pyrradiometers. The objective was to determine the best method for measurement of broadband longwave irradiance for application in the BSRN network and to quantify the likely error in measurements. Various instruments used by participating BSRN groups were intercompared, and measurements were also compared with“official” estimates of the downwelling infrared irradiance provided by the Spectral Radiation Experiment, which also participated prominently in FIRE II as an adjunct experiment.
Results from the FIRE II intercomparisons (DeLuisi 1992) reinforced conclusions from earlier intercomparison results published by various authors (e.g., Weiss 1981; Alados-Arboledas et al. 1988; Field et al. 1992; Dehne et al. 1993), showing differences between downwelling longwave irradiance measurements from different instruments up to more than ±10 W m−2. Two instrument calibrations were used—one the manufacturer provided with the instrument and an ice dome calibration. For the two calibrations, systematic differences of mean results between instruments of up to 13 W m−2 were seen. This result particularly stressed the importance of a comparison of calibration procedures.
In this paper we report on the results of the round-robin calibration campaign, which was initiated by J. DeLuisi and A. Ohmura. The experiment was intended solely to test and compare presently available blackbody calibration devices. Problems related to pyrgeometer field measurements and to the absolute accuracy are not investigated. Five Eppley precision infrared radiometer (PIR) pyrgeometers and one pyrgeometer modified by Foot (1986) were supplied by different proprietors as test instruments. All PIRs had spent some time in the field and had proven to be reliable. Specialists and research laboratories of the radiation community from all over the world were invited to take part in the experiment, which involved calibration of the six pyrgeometers using their individual methods and apparatus. The round-robin experiment was conducted in a “blind” sense in that the participants had no knowledge of the results of others until the experiment ended. Calibration factors determined by the individual institutes were gathered at the World Radiation Center at Davos, Switzerland, and results from the analysis are presented on the following pages.
2. Pyrgeometers
PIR, a pyrgeometer developed by the Eppley Laboratory, is a development from their precision spectral pyranometer (PSP) but with a silicon rather than a glass dome. This instrument was first described by Drummond et al. (1970).
Tests conducted at the Meteorological Research Flight (MRF) of the U.K. Meteorological Office at Farnborough led Foot (1986) to conclude that the dome temperature cannot be measured satisfactorily at a single point and that within the body of the instrument there may be additional temperature gradients due to conduction and convection effects producing significant errors. MRF consequently adopted a modified design “the MRF pyrgeometer,” which minimizes the correction term by eliminating the influence of heat conduction and convection between dome and sensor surface. This is accomplished by arranging the gold cold junction and the black hot junction of the thermopile on the same surface so that both face the dome of the pyrgeometer and, therefore, are only sensitive to the radiation term. The correction factor k of the MRF pyrgeometer is thus much reduced compared to a normal PIR, and the dome temperature measurement may be neglected. Equation (1) can then be used for computing the incoming irradiance with the reference temperature TB; the remaining small dome correction term is ignored.
3. The round-robin unit
S. Sandberg and J. Wendell prepared the round-robin instrumentation, which consisted of five standard PIR pyrgeometers (one dome thermistor at the rim) and one MRF pyrgeometer (no dome thermistor). They added a laptop computer, a Campbell Scientific data acquisition system for logging the pyrgeometer signals, and a nonintrudable computer program. This same autonomous measurement system was used by all participants for homogeneity, who otherwise used their own apparatus, calibration method, and equipment.
Climate Monitoring and Diagnostics Laboratory (CMDL) and the Applied Research Laboratory (ARL), both of the National Oceanic and Atmospheric Administration, organized the shipping of the equipment to the participating laboratories, which had agreed to calibrate the six pyrgeometers, and ARL periodically checked the entire unit. The first calibration was made at CMDL in 1993, and the same institution and ARL recalibrated the six instruments at the end of the round robin in 1996. The only problem encountered with the suite of equipment during the experiment was a malfunctioning instrument due to a ground pin that had been disconnected from the case by its owner. Eppley Laboratory, third in the round-robin sequence, discovered that the instrument (PIR 26181) had been modified and reconnected the ground pin. However, this did not appear to have significantly affected the results.
4. Participating laboratories and calibration devices
Eleven laboratories from seven countries participated in the round-robin calibration of the six pyrgeometers. Table 1 lists addresses of the laboratories in alphabetical order and the names and e-mail addresses of the responsible scientists. To learn more about the different calibration methods and devices used by the different institutes, and to get more insight into the parameter range in which the instruments were calibrated, a questionnaire was sent to the individual participants at the end of the experiment in 1995. Table 2 describes the blackbody radiation sources and gives a general description of the calibration apparatus. All the 11 blackbody radiation sources used are of different construction and shape. The emitted hemispherical radiation is computed from Planck’s law using the mean temperature of the blackbody Tbb and the values of total emittance of the cavities, which range from 0.995 to 1.
Table 3 summarizes general characteristics of the calibration procedure and devices employed by the individual laboratories. Different pyrgeometer exposures were used during calibration, most of which were made with the pyrgeometer body temperature at around 20°C with extremes from −1° to 28°C. Blackbody temperatures varied in a wide range from −50° to +127°C. The fifth column of Table 3 indicates that for a temperature difference of 25°C between pyrgeometer body and blackbody radiation source, maximum differences between the dome and the body temperature vary from 0° to 2.5°C for the different calibrations. The number of calibration points varied within a factor of 10. Only three laboratories used ventilation during the calibration.
5. Results
a. Calibration evaluation
As pointed out in section 3, round-robin participants were free to adopt their own method of calibration. Five participants not only determined the responsivity C but also an individual dome correction factor k for the five PIRs (see the two last columns of Tables 3 and 6). In these cases, the pyrgeometer equation (2) was used in the evaluation. At three laboratories, a fixed dome correction factor of k = 4 was taken for all the instruments together with Eq. (2) for the evaluation. The remaining three laboratories ignored the dome correction by setting k = 0 and employed Eq. (1) with the body temperature TB as a reference for the pyrgeometer. The MRF pyrgeometer was calibrated by seven participants and Eq. (1) was used to calculate C. The responsivity C is given in μV/W m−2 and the individual correction factors k range between about 2 and 5.
b. PIR responsivities C
The responsivities of the five PIR pyrgeometers found by the 11 laboratories are shown in Table 4. The original responsivities given by the manufacturer and the value found by CMDL and ARL at the end of the round-robin experiment are also included at the end of the table. No distinction was made whether C was determined with or without the dome correction factor k. The median of the 11 values was determined for each instrument and the minimum, maximum, and the absolute deviation of C are given as percentages with respect to the median. (The median, rather than the arithmetic average, is likely to be nearer the average value of the non-Gaussian distribution expected from the 11 very different calibrations.) The difference ΔC, between the individual C values and the median of the respective instrument, is given as a percentage. Moreover, the median of the ΔC values of the five instruments was determined for each laboratory, and an absolute deviation of the five ΔC values from the median of the ΔC’s is shown in the last column of Table 4.
The absolute deviation of ΔC to the median of the ΔC’s of the five instruments best quantifies the calibrations of the individual laboratories. The small scatter demonstrates the reproducibility of the calibration result. Whether the median of the ΔC’s is close to zero or not is less important, since a bias is related to a systematic error that could be corrected. Six out of the 11 laboratories determined the responsivity C of the five PIRs within about 2% of the median, with an absolute deviation of ΔC less than 1%. Figure 1 represents the ΔC’s from all the five instruments and the 11 laboratories and clearly shows the difference between the small scatter of the calibrations of 6 laboratories compared to the significantly larger scatter of the others. An error of 1% in C corresponds to an error in the measured flux of about 1 W m−2. Hence, the large errors found between individual pyrgeometer measurements of 10 or more W m−2 are not explained with errors on the calibration contant. It is also worthwhile noting that the values of the original calibrations are within 2% of the median except for the oldest instrument (13678). Furthermore, the two CMDL calibrations made in 1993 and 1996 are almost identical.
c. MRF responsivity C
The results from the MRF pyrgeometer calibrations are shown in Table 5. The responsivity found by the seven participants who calibrated the instrument are within 5% of the median, which is overall a better result than for the PIRs.
d. PIR correction factor k
Individual dome correction factors were determined by only five laboratories and are shown in Table 6. Generally speaking, the k values were not as precisely determined as the C values. Nevertheless, four out of the five participants found the correction factor k of the five PIRs within 20% of the median. These four calibration laboratories are among the group of six with responsivity values C close to the median.
6. Discussion of results
A first glance at the results of the round-robin calibration experiment shows notable differences of up to 20% from the median of the responsivities of the five PIR pyrgeometers. The maximum difference from the median of the MRF pyrgeometer is only about 5%. Unfortunately, only one instrument of this kind was included in the experiment and it was calibrated only by 7 of the 11 participants. Hence, it is difficult to compare the results obtained with the MRF pyrgeometer directly with the PIR calibrations.
However, looking more closely at the results, it becomes apparent that six laboratories determined the PIR responsivities C with remarkably lower scatter around the median values than the other five participants (see also Fig. 1). The fact that among these six laboratories the absolute deviation of ΔC to the median of the ΔC’s is less than 1% is strikingly good. At the outset, this provides a strong indication of the stability of PIR pyrgeometers, which is further underlined by the two calibrations made by CMDL (Dutton 1993), of the five PIRs, one at the beginning in 1993 and one at the end of the experiment in 1996, which are both almost identical and are within the 1% limit. Certainly, the instruments had very little exposure to the environment during the 3-yr time period. However, they had traveled to many different and distant locations.
A second important result is the good agreement found among results from very different calibration methods and apparatus. Among the six laboratories that achieved the good “median” results, pyrgeometer positioning during calibrations were upward, downward, and sideward. The body temperature was set between 0° and 25°C. Blackbody temperatures from −50° to +70°C were used. One participant used ventilation during calibration. For the evaluation, four laboratories determined C and k using Eq. (2), whereas two participants simply neglected the dome correction term and used Eq. (1). Despite of all these differences the good agreement between the responsivities suggests that blackbody radiation sources, although of very diverse construction, are capable of producing consistent results. Thus, the calibration procedure itself and, in particular, interchanging blackbody radiation sources, does not seem to be a matter of serious consequence.
Although it is not appropriate to judge individual laboratories and their calibration procedures and apparatus, we still feel that it is necessary to consider the reasons for the good “median” results of six participants and those for possible shortcomings of the other laboratories. Three of the participants [the Atmospheric Environment Service (AES), the Bureau of Meteorology (BoM), and Geographisches Institut ETH (GI-ETHZ)], which have a large scatter and are not among the six“median” results have only one calibration point and, thus, there is no indication of the internal consistency. AES and BoM calibrated with a blackbody temperature of 70° and 127°C, which is probably not the optimal temperature for pyrgeometers that rather measure thermal radiation corresponding to a temperature around or below 0°C. The three participants who evaluated their results using a fixed dome correction factor of k = 4 [BoM, GI-ETHZ, and National Aeronautics and Space Administration Applied Research Center (NASA/ARC)] are not among the six “median” results. The NASA/ARC calibration apparatus is designed to calibrate flight instruments that are not sensitive to dome temperature effects. During this calibration the dome temperature is not constant and drops to near the blackbody temperature, while the radiometer body remains at near room temperature. Dome correction factors determined by MRF have the largest scatter among the five laboratories that did determine the individual k values, otherwise there is no obvious reason why MRF is not among the six.
Dome correction factors are more difficult to determine and the almost 20% difference between the four participants whose results are nearest the “median” is large. The main reason for the large scatter is probably due to the dome temperature measurement, which seems to be influenced differently by the different calibration devices and, hence, that the dome thermistor at the rim does not indicate a representative dome temperature.
7. Conclusions
Six of the 11 participants of the round-robin pyrgeometer calibration experiment found very close responsivities C around the median for the five pyrgeometers. This result shows that pyrgeometers are stable and that blackbody calibrations are reproducible. In this context the term stability is used to mean that the pyrgeometers consistently reproduce their calibration constants over long periods of time and after exposure to many environmental factors.
Participants were entirely free to use their own methods and calibration devices in this experiment and yet good results were achieved. However, there is no doubt within the BSRN community that a certain standardization of calibration procedures would improve the results even further. Therefore, for the BSRN, recommendations have been made with regard to the temperature range in which pyrgeometers are calibrated and the equation used for the evaluation of pyrgeometer calibrations. Although, for the thermopile, temperature compensation circuits are used (not in the MRF instrument), a certain enhancement of pyrgeometer performance might be achieved, if the body temperature during calibration could be set close to the annual mean of the ambient temperature of the site on which the pyrgeometer was to be deployed. Also, to simulate the downwelling atmospheric radiation as realistically as possible during calibration, the blackbody source temperature should be about 10°–25°C below the pyrgeometer’s body temperature. It must be said, however, that calibrations made at more “extreme,” in particular lower temperatures, are not necessarily poorer performers. But the analysis clearly shows that a certain number of calibration points are necessary.
The good “median” results of the Deutscher Wetterdienst Meteorologisches Observatrium Potsdam, and the Eppley Laboratory, that did not determine and, therefore, did not use the correction factor k, carries an implication that the use of the dome correction term for the calibration may not be essential. However, this requires, that during calibration, the temperature difference between body and dome is very small. It is well known and accepted that the dome correction term and Eq. (2) must be used for accurate pyrgeometer measurements in the field. Hence, the dome correction factor k has to be known accurately and should be determined during calibration. The unsatisfactory results with regard to the estimation of k in this experiment is most likely connected to the dome temperature measurement at the rim, which does not provide a representative dome temperature. The new dome temperature measurement system described by Philipona et al. (1995), which uses three thermistors separated by 120° and glued at 45° elevation, to assess more exactly the representative dome temperature might be a solution to this problem.
The investigations presented in this paper are limited to calibration of pyrgeometers using a blackbody source and are therefore clearly separated from problems related to pyrgeometer field measurements. Results from the field test in Boulder, Colorado, using the round-robin instruments will be published in a separate paper. Encouraging results have been found with regard to the stability of pyrgeometers, the interchangability of blackbody radiation sources, and to the determination of the responsivity C of PIRs. However, further investigations are needed, in particular of the determination of the dome correction factor k, and a certain standardization of the pyrgeometer calibration procedures is inevitable for the future.
Acknowledgments
The authors would like to thank Jim Wendell CMDL, Scott Sandberg ARL, Tony Vergamini ARL, Mike Rubes ARL, Tezz Johnson ARI, and David Theisen ARL of NOAA at Boulder, Colorado, for assistance and their fine work. The Boulder group is grateful to Steve Cox for the loan of the CSU calibration device.
REFERENCES
Alados-Arboledas, L., J. Vida, and J. I. Jimenez, 1988: Effects of solar radiation on the performance of pyrgeometers with silicon domes. J. Atmos. Oceanic Technol.,5, 666–670.
Albrecht, B., and S. K. Cox, 1977: Procedures for improving pyrgeometer performance. J. Appl. Meteor.,16, 188–197.
——, M. Peollet, and S. K. Cox, 1974: Pyrgeometer measurements from aircraft. Rev. Sci. Instrum.,45, 33–38.
Dehne, K., U. Bergholter, and F. Kasten, 1993: IEA comparison of longwave radiometers in Hamburg, 1989/90. Rep. IEA-SHCP-9F-3. [Available from Deutscher Wetterdienst, Meteorologisches Observatorium Hamburg, Postfach 65 01 50, D-22361 Hamburg, Germany.].
DeLuisi, J., K. Dehne, R. Vogt, T. Konzelmann, and A. Ohmura, 1992: First results of the Baseline Surface Radiation Network (BSRN) broadband infrared radiometer intercomparison at FIRE II. Proc. Conf. Int. Radiation Symp. 92, Tallinn, Estonia, 559–564.
Drummond, A. J., W. J. Scholes, and J. H. Brown, 1970: A new approach to the measurement of terrestrial long-wave radiation. WMO Tech. Note 104, 383–387.
Dutton, E. G., 1993: An extended comparison between LOWTRAN7 computed and observed broadband thermal irradiances: Global extreme and intermediate surface conditions. J. Atmos. Oceanic Technol.,10, 326–336.
Enz, J. W., J. C. Klink, and D. G. Baker, 1975: Solar radiation effects on pyrgeometer performance. J. Appl. Meteor.,14, 1297–1302.
Field, R. T., L. J. Fritschen, E. T. Kanemasu, E. A. Smith, J. B. Stewart, S. B. Verma, and W. P. Kustas, 1992: Calibration, comparison, and correction of net radiation instruments used during FIFE. J. Geophys. Res.,97, 18 681–18 695.
Foot, J. S., 1986: A new pyrgeometer. J. Atmos. Oceanic Technol.,3, 363–370.
Miskolczi, F., and R. Guzzi, 1993: Effect of nonuniform spectral dome transmittance on the accuracy of infrared radiation measurements using shielded pyrradiometers and pyrgeometers. Appl. Opt.,32, 3257–3265.
Philipona, R., C. Fröhlich, and C. Betz, 1995: Characterization of pyrgeometers and the accuracy of atmospheric longwave radiation measurements. Appl. Opt.,34, 1598–1605.
Shiobara, M., and S. Asano, 1992: The dome emission effect on the performance of pyrgeometers with silicon domes. Pap. Meteor. Geophys.,43, 17–31.
Weiss, A., 1981: On the performance of pyrgeometers with silicon domes. J. Appl. Meteor.,20, 962–965.
WMO, 1990: World Climate Research Programme: Workshop on implementation of the baseline surface radiation network, Washington, USA. WMO/TD-No. 406.
——, 1991: World Climate Research Programme: Second workshop on implementation of the baseline surface radiation network, Davos, Switzerland. WMO/TD-No. 453.
Difference ΔC from the median in percent for the 5 PIRs of the 11 participants. Six out of the 11 groups determined C within about 2% of the median, whereas the other groups have considerably larger deviations.
Citation: Journal of Atmospheric and Oceanic Technology 15, 3; 10.1175/1520-0426(1998)015<0687:TBSRNP>2.0.CO;2
Round-robin participating laboratories and responsible scientists.
Description of blackbody and calibration apparatuses.
(Continued).
Temperature range and general characteristics of calibration procedures and apparatus.
Responsivity C and deviation to median ΔC (%) of PIR pyrgeometers.
Responsivity C and deviation to median ΔC (%) of MRF pyrgeometer.
PIR dome correction factors k and deviations to median Δk (%).
