Characterization of Thermal Effects in Pyranometers: A Data Correction Algorithm for Improved Measurement of Surface Insolation

Transcription

1 165 Characterization of Thermal Effects in Pyranometers: A Data Correction Algorithm for Improved Measurement of Surface Insolation BRETT C. BUSH, FRANCISCO P. J. VALERO, AND A. SABRINA SIMPSON Atmospheric Research Laboratory, Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California LIONEL BIGNONE École Polytechnique, Palaiseau, France (Manuscript received 29 September 1998, in final form 15 March 1999) ABSTRACT Pyranometers are reliable, economical radiometers commonly used to measure solar irradiances at the surface in a long-term, monitoring mode. This paper presents a discussion of the response of these instruments to varying environmental conditions, including the magnitude and variability of the irradiance being measured. It is found that different conditions, commonly occurring in field experiments, affect the thermal balance and temperature gradients within the instrument in a variety of ways. Such an effect results in variable offset systematic errors whose origin and magnitude are investigated in laboratory and field experiments. It is shown that these offset errors are proportional to the difference between the fourth power of the dome and detector temperatures, following closely the Stefan Boltzmann radiation law. Results of field experiments are presented for daytime and nighttime operation over a variety of atmospheric conditions ranging from clear to heavy overcast and rain. All measurements took place from May through October 1998 in La Jolla, California, at the Scripps Institution of Oceanography. Laboratory experiments are used to quantify the magnitude of the thermal offset errors under controlled conditions and to calibrate them as a function of thermal gradients between the dome and the detector. The quality of the data resulting from pyranometer measurements can be improved in a significant way by proper knowledge of the thermal parameters affecting the operation of the thermopile system. To that end, a data correction algorithm that requires an extensive thermal calibration procedure and a simple modification of the instrument is proposed. Such an algorithm needs to be applied to the power calibration procedure as well as to the retrieval of data acquired during normal field operations. The experimental results presented in this paper could potentially affect analyses based on surface insolation measurements performed using pyranometers, in particular those related to the measurement of diffuse radiation fields. 1. Introduction Recent studies have used surface and airborne radiative flux measurements in conjunction with theoretical model predictions to attempt to answer fundamental questions related to radiative transfer processes in the atmosphere. One of the key issues in these studies is the question of the presence or absence of an excess shortwave absorption (relative to model predictions) in cloudy and/or clear skies (Arking 1996; Cess et al. 1995, 1996; Charlock and Alberta 1996; Conant et al. 1997; Francis et al. 1997; Imre et al. 1996; Kato et al. 1997; Li et al. 1995; Li and Moreau 1996; Pilewskie and Val- Corresponding author address: Dr. Francisco P. Valero, Scripps Institution of Oceanography, University of California, San Diego, 9500 Gilman Dr., Dept. 0242, La Jolla, CA fvalero@ucsd.edu ero 1995; Ramanathan et al., 1995; Stephens 1996; Stephens and Tsay 1990; Valero et al. 1997a; Valero and Bush 1999; Waliser et al. 1996; Zender et al. 1997). Extensive field campaigns, such as the Atmospheric Radiation Measurements Enhanced Shortwave Experiment (ARESE) (Bush and Valero 1999; Valero et al. 1997b) and the Subsonic Aircraft: Contrail and Cloud Effects Special Study (SUCCESS) (Michalsky et al. 1997; Pilewskie et al. 1998; Valero and Bush 1999) have been completed with the intent of addressing, at least in part, the clear sky and cloud absorption questions. Kato et al. (1997) utilize surface observations made in October 1995 during ARESE. They use a careful analysis of direct and diffuse radiative fields to conclude that the clear sky absorbs more solar radiation than accounted for by models and that the missing absorption is taking place at visible wavelengths of the solar spectrum. Additionally, they postulate that the unaccounted for absorption could be the result of the presence of an 2000 American Meteorological Society

2 166 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 17 unknown gas in the atmosphere. The conclusions reached by Kato et al. (1997) are based on the disagreement observed between calculated and measured diffuse components of the total insolation. Halthore et al. (1998), follow the Kato et al. (1997) arguments and conclusions but using data obtained during SUCCESS in April However, Zender et al. (1997) and Valero and Bush (1999) find agreement within one standard deviation (models always being higher) between observations and models for the same ARESE and SUCCESS periods. There are, however, differences between the data sets used by each group. Kato et al. (1997) and Halthore et al. (1998) used data generated by the standard Cloud and Radiation Testbed (CART) site radiometric instrumentation, while Zender et al. (1997) and Valero and Bush (1999) used data produced by the Radiation Measuring System (RAMS) (Bush et al. 1999). The standard CART and RAMS radiometers were located in close proximity (a few meters apart) during the period of the observations. The apparent inconsistencies and the major scientific importance of the conclusions drawn from the diffuse field measurements (Kato et al. 1997) motivated us to study the properties of the instrumentation. First, a comparison exercise was organized between the RAMS and World Radiation Reference (WRR) standard radiometric systems, operated by the Climate Monitoring and Diagnostics Laboratory (CMDL) of the National Atmospheric and Oceanic Administration (NOAA) (Wiscombe et al. 1998). Second, we proceeded with the investigation described in this paper to assess the performance of the thermopile-based radiometers that provided the data used for the ARESE and SUCCESS analysis. This decision was based on studies by Morikofer (1939), Beur (1950), Drummond and Roche (1965), Robinson (1966, ), Rodskjer (1971), Gulbrandsen (1978), and more recently Cess et al. (1999). These studies indicate the potential for significant thermally generated errors that affect measurements acquired with radiometers of the type discussed in this paper. An essential goal is to quantify the reaction of a thermally characterized thermopile detector to various atmospheric conditions. In particular, we characterize the baseline or detector thermal offset signal with respect to varying thermal conditions. The observed thermal offset errors vary with environmental conditions, in particular those that affect the temperature gradients within the pyranometer (e.g., air temperature, wind speed, direct-to-diffuse radiation ratios, magnitude and variability of the irradiance being measured, cloud cover, mounting surface temperature and optical properties, etc.). To properly account for the systematic thermal offsets it is critical to accurately measure the temperature of those specific components of the instrument that play a role in driving the output signal of the detector. For ideal broadband solar flux measurements, this signal should react only to the atmospherically varying radiation field and not to any thermally changing conditions. By quantifying the thermal effects, the accuracy of these measurements may be better understood and algorithms can be developed for correcting data products, thus improving data quality. Thermal dependencies of the absolute calibration have been discussed in previous works (Smith et al. 1988) and are not included in this study. They conclude that relative measurement errors of radiation up to 2% exist for a temperature range of 20 to 40 C. The experimental results that we present here were obtained using an Eppley PSP pyranometer and apply to instruments using the same principle of operation and similar design. Other designs, including the use of forced air circulation around the dome (ventilation) may reduce or increase the thermal effects discussed below, depending on the particular conditions (i.e., air temperature). However, the characterization of thermal properties and the application of the correction algorithm proposed will result in an improvement of data acquired with any thermopile-based pyranometer. For some designs and applications the corrections may be small, in others more significant, but the certain way to determine the magnitude of the thermally induced errors, with the accuracy required by present-day climate research, is to accurately account for such effects in all cases. 2. Experimental a. Pyranometer modification A modified Eppley precision pyranometer (Model PSP) capable of measuring solar irradiance was used in this study. Basically, the PSP uses a wire-wound thermopile as its radiation sensor. The thermopile is in thermal contact with a large base that acts as a heat sink and is covered with two concentric hemispherical domes transparent in the m spectral region. However, the thermopile detector itself is sensitive to far-ir radiation. As in any optical system that does not use cryogenic cooling or balanced operation, the transfer of infrared radiation between components affects the performance of pyranometers by generating an internal infrared signal that is superimposed to the output signal. The temperatures of the detector and of the outer dome are the main drivers of the temperature gradients that generate the internal, spurious signal. The inner dome acts as a heat shield ; it reduces the amount of infrared radiation being transferred between the detector and the outer dome (see, e.g., Özisik 1985, ). Figures 1, 2, and 3 illustrate examples of the outstanding correlation between the internally generated offset signal and the temperature difference between detector and outer dome. Such correlation (R , Fig. 3) demonstrates the appropriateness of using the outer dome and detector temperatures to characterize the ther-

3 167 FIG. 1. (a) PSP signal and (b) thermal variations during a 12-h period. In (b) the left axis represents the individual detector and dome temperatures. The right axis represents their difference. The periodic jumps correspond to times when the air conditioning was turned on and off. The calibrated PSP signal during the same period is presented in (a). The signal variations are completely correlated to temperature changes in the PSP detector and dome. mal properties of the optical system as a whole. An equivalent characterization using the inner dome temperature could also be used with corresponding results. The PSP was modified by adding four precision thermistors (YSI series epoxy encapsulated) to measure the temperature difference between the outer dome and the detector. Two thermistors were attached directly to the detector mounting plate (as close to the detector as possible) and two to the outer dome using a special thermally conductive epoxy. The dome detectors were located about 15 mm from the base of the dome to minimize interference with the radiative signal. The thermistors are specified as accurate to 0.1 C. The time constant of the thermistors is approximately 10 s in still air and only 1 s in a well-stirred oil bath. In the present application the time constant is estimated at about 10 s. The redundant measurements allow sensing of the variation of both the detector reference T Detector and the dome, T Dome, temperatures. Averages of the two readings on each component were used to characterize the overall temperature of that component. In order to minimize heating from any source other than the dome itself, dome thermistors were shaded from direct sunlight using small ( 4 mm 2 ) pieces of foam-backed reflective tape. FIG. 2. (a) Variation of the PSP signal, and (b) dome and detector temperatures after cooling the instrument base. A slow warming of the detector and dome restore the equilibrium conditions after over 5 h, resulting in a negligible PSP offset and differential temperature. b. Environmental conditions Tests were made in both indoor (laboratory) and outdoor conditions. External heat sources were applied in the laboratory during some of the calibration procedures. Outdoor operations occurred in La Jolla, California, during May, June, and July During these operations, the modified PSP was placed on a platform on the roof of a building in close proximity to the Pacific Ocean with few neighboring obstructions. The platform served both to raise the radiometer off the surface of FIG. 3. Correlation of the PSP offset signal with the dome and detector temperatures. The fit of the offset signal response data with respect to the difference of the fourth power of the dome and detector temperatures gives a % linear correlation.

4 168 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 17 the roof and allow prevailing winds to flow naturally past the detector. All measurements were made with a datalogger recording average data in 5-s intervals during the daytime and 60-s intervals at night. 3. Laboratory calibrations a. Thermistors Since the accuracy of determining the differential temperature between the detector and dome is important in this radiometer characterization, all four thermistors used were calibrated relative to one another before mounting them on the PSP. Adjustment factors to the resistance versus temperature specifications were made to ensure good agreement in the approximate measurement temperature range (roughly C). The resulting adjustments were all less than 1% and the resulting relative precision was about 0.02 C. b. PSP offset Calibration of the systematic error, the offset signal that is inherent to the PSP, was completed in the laboratory under controlled conditions. An example of the reaction of the PSP to slightly varying environmental conditions in a dark, closed room is shown in Fig. 1. The periodic variations in the signal correspond to times when the air conditioning was turned on and off inside the isolated room. It is apparent that the thermal time constant of the dome is smaller than that of the detector in that it reacts more quickly to the changing ambient temperature. The difference in the dome and detector temperatures ranged from nearly 0 to about 0.4 C in this case with the dome always being cooler than the detector. Figure 1b demonstrates that the average room temperature changed very slowly, approximately 2.3 C in 12 h. The detector and dome temperatures do not equalize at any time during this period. Simultaneously, the PSP signal was monitored. The room was completely dark so that any signal or variation was not due to any external light source. The radiometer was left uncovered so that it was not insulated from the external environment. The calibrated PSP signal is shown in Fig. 1a over the same 12-h period presented in Fig. 1b. The PSP signal variations are completely correlated with the temperature difference of the dome and detector. The small magnitude of the average baseline PSP offset ( 0.5 W m 2 ) results from close detector and dome temperatures (about 0.1 C). Furthermore, the negative offset in the PSP signal corresponds to the detector being warmer than the dome, which is consistent with the thermal nonequilibrium present in the system: the radiant infrared flux energy emitted by the dome at T Dome is less than that of the detector at T Detector. The radiative energy transfer is from the detector to the dome, hence the signal is negative. The magnitude of the offset error was calibrated by placing the PSP in a black chamber thus, eliminating any visible radiation from reaching the detector. In this fashion, it was possible to establish the relationship between the offset signal and the difference between dome and detector temperatures. Naturally, as described later, there is also a dependence of the offset error on absolute mean temperature of the instrument. Temperature gradients between dome and detector were produced in the laboratory in two ways. First, the detector heat sink was warmed and allowed to cool down until a quasi-stationary regime is reached. At this point, the dome and detector temperatures evolve slowly and converge to the temperature inside the black box after several hours. During this calibration experiment, negative offset values are related to the temperature gradient between the dome and the detector. Second, the PSP was cooled in a refrigerator. Immediately after removing the instrument from the refrigerator, the dome warmed up much faster than the detector, establishing a gradient of opposite sign to that of the first experiment. Again, the radiometer was placed inside the black box and dome and detector temperature allowed to evolve and converge to the ambient temperature. This second experiment provided the positive offset values. Figure 2a shows the response of the PSP signal during this test and Fig. 2b shows the dome, detector, and differential temperatures of the instrument. As seen in these figures, it took over 5 h for the system to return to a steadystate equilibrium where the detector and dome temperatures were nearly equal and the PSP offset was negligible. The data acquired during the calibration experiments described above are plotted in Fig. 3. Each point represents an average of calibration data values within a 1 Wm 2 bin. A total of over calibration points, or approximately 70 h of measurements, is represented in Fig. 3. The measured offset values, F, are plotted versus the difference in the fourth power of the temperatures (D T 4 Dome TDetector 4 ). A linear fit, weighted by the number of calibration points in each 1 W m 2 bin, gives F ( T 4 T 4 ) Dome Detector [W m 2 ], (1) with a correlation coefficient of The Wm 2 offset for the fit reinforces the fact that when the dome and detector are in thermal equilibrium, the detector offset is negligible. Equation (1) points to the very high correlation (essentially 1) between the PSP thermal error and the difference in temperature between the outer dome and the detector. The linearity of the fit with respect to the difference in the fourth power of the temperatures is not surprising when considering the B T 4 blackbody power law, where B Wm 2 K 4 is the Stefan Boltzmann constant. Due to its extreme linearity, as reflected in the correlation coefficient of , Eq. (1) will be used

5 169 FIG. 4. PSP offset signal vs mean, (T Dome T Detector )/2, temperature, and temperature differential ( T T Dome T Detector ). The offset increases for a constant T as the mean temperature increases. for inferring PSP offset values in all conditions in this study. The difference in the fourth power of the temperatures, D, may be written as D (TDome 3 TDome 2 TDetector TDomeTDetector 2 T Detector 3 ) T, (2) where T T Dome T Detector and can be further approximated by 3 D 4T Mean T, (3) and T Mean (T Dome T Detector )/2 is the mean PSP temperature. Equations (1) and (3) point out the very important fact that the magnitude of the offset signal per degree of temperature difference depends on the mean temperature of the apparatus. Figure 4 illustrates this dependence for a range of temperatures and temperature differences commonly encountered in field measurements. 4. Field measurements PSP detectors are typically operated in the field under two different conditions: unshaded and shaded. When unshaded, the radiometers measure the total (global) hemispherical downwelling solar radiation. The shaded PSPs are typically used in conjunction with normal incidence pyrheliometers or cavity radiometers whose field of view mimics that of the shadow disk occulting the direct solar beam. In this case, the PSP approximates the diffuse portion of the downwelling flux and the normal incidence instrument measures the direct component with the component sum being equal to the total flux. Recent studies conclude that when using pyranometers, this component sum method is a more accurate means of determining the global surface irradiance than with a single, unshaded PSP (Michalsky et al. 1997; Bush and Valero 1999). In this study, we will characterize the thermal effects for both the unshaded and shaded PSP conditions. FIG. 5. (a) Surface insolation and (b) temperature evolution on a predominantly clear day, 19 Jun a. Clear day unshaded PSP measurements A variety of environmental conditions were present during the unshaded PSP operations. These conditions ranged from clear sky to scattered clouds, broken clouds, overcast, and light to mild rain days. Winds were light and normally rotated from the east to the west, approximately tracking the sun during its daily progression. Operations occurred uninterrupted for 24 h a day for a 3-week period in June Additional measurements were made in July and October The field observations that we report here characterize the very mild nonextreme conditions typical of the weather at La Jolla. During this period, daytime temperatures typically ranged from 15 to 20 C, whereas nighttime values were C. This range of conditions included complete solar elevation angle variations that were also essential in assessing the PSP thermal behavior over a gamut of atmospheric states. Furthermore, the nighttime measurements also serve as direct measurements of the PSP offset values, again in different situations. An example of observations during a clear day (19 June 1998) is shown in Fig. 5. The top panel depicts the pyranometer signal, while the bottom panel illustrates the evolution of the temperatures during the day as well as their differences. The difference between the dome and detector temperatures is roughly constant during this period except for some slight variations. Figure 6 shows a time series of the PSP signal and temperature differences of an unshaded pyranometer operating under clear sky conditions (5 October 1998). At about 15.6 h the instrument was covered with a black box, and the evolution of the PSP signal and temperatures are depicted in the figure. Note the striking similarity in the shape of the PSP signal and temperature

6 170 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 17 FIG. 6. Time series of the PSP signal and temperature differences of an unshaded pyranometer operating under clear sky conditions. At about 15.6 h the instrument was covered with a black box and the evolution of the PSP signal and temperatures are shown. Note the striking similarity in the shape of the PSP signal and temperature difference curves. As soon as the instrument is covered the signal drops to a negative value representative of the offset ( 17Wm 2 ). The thermal offset value calculated using Eq. (1) and temperatures measured just before covering, is about 17Wm 2. Note the similarity of the curves representing the difference between temperatures and the thermal offset output signal of the pyranometer. difference curves. Furthermore, as soon as the instrument is covered the signal drops to a negative value representative of the offset ( 17Wm 2 ). As expected, if the thermal calibration Eq. (1) is applied using the temperature difference measured just before the PSP is covered, a calculated offset of about 17Wm 2 is found. The calculated value is practically identical to the offset measured immediately after covering. This is the magnitude of the offset affecting the instrument before and at the moment the covering took place. The PSP signal is deduced from a direct voltage measurement (no instrument electronics involved) and is not affected by any electronic overshoot. FIG. 7. Thermal response of the PSP dome and detector during a shaded measurement (same format as Fig. 1b). The shadow disk is introduced at 1225 LT. The dome immediately cools by approximately 4 C, whereas the detector temperature decreases by about 2 C. Thermal steady state is reestablished about 30 min after the shadowing begins. b. Clear day shaded PSP measurements The major thermal effect of shading the PSP is to change the temperature balance between the dome and the thermopile detector. The shaded PSP measurements in this study were made in predominantly clear or mostly clear sky conditions. When overcast conditions existed, the effect of shadowing the direct solar beam (which, by the nature of the conditions, did not exist) was negligible on the temperature balance of the system. During the shaded measurements, the PSP was initially left unshaded in its environment for a period of several hours long enough for conditions to equilibrate. After a sufficient time period, the PSP was shaded from the direct solar beam and conditions were allowed to reequilibrate with respect to the modified experimental conditions. The shadow disk followed the sun to maintain solar occultation. A typical instrument response during a single shaded PSP measurement on a clear day (27 July 1998) is presented in Fig. 7. A shadow disk blocked the direct solar beam from the PSP at 1225 LT. Before introducing the shadow disk the temperatures of the detector and dome increased at approximately the same rate with a constant average difference of approximately 2.5 C. After introduction of the disk the dome temperature dropped about 4 C, whereas the detector temperature decreased by only about 2 C. Steady state was reestablished after about 20 min. The long-term trend in both the dome and detector temperatures was a gradual decrease at approximately the same rate. These results indicate that the shaded PSP operates in a different thermal state than the unshaded instrument, at least during clear sky conditions. Figure 7 shows 3.5 h of data after shadowing to illustrate the fact that steady state has been reached and that the observed temperature difference is not a transient effect resulting from the sudden blocking of the direct solar beam. Figure 8 is similar to Fig. 6 except that it depicts the pyranometer signal and temperatures evolution after covering a shaded pyranometer, under clear sky conditions (6 October 1998). In this case a negative offset of about 18 W m 2 is measured shortly after covering the instrument. The calculated value, using Eq. (1) and temperatures measured just before covering, is also 18 Wm 2. Again, note the similarity of the curves representing the difference between temperatures and the off-

7 171 FIG. 8. Pyranometer signal and temperatures evolution after covering a shaded pyranometer. In this case a negative offset of about 18Wm 2 is measured shortly after covering the instrument. The offset value calculated using Eq. (1) and temperatures measured just before covering is also 18Wm 2. Again, note the similarity of the curves representing the difference between temperatures and the offset output signal of the pyranometer. Note that after about 30 min the dome and detector reach steady state. set output signal of the pyranometer. Note that after about 20 min the dome and detector reach steady state. FIG. 9. (a) Surface insolation and (b) temperature evolution on a broken cloud day, 10 Jun c. Cloudy days Figure 9 shows data for broken cloud conditions during 10 June Note the wide oscillations in dome and detector temperatures that produce highly variable temperature differences, which are clearly correlated to the variations in insolation. This is illustrated by the similarity of the curves representing the power measured by the PSP (Fig. 9, top panel) and the temperatures (Fig. 9, bottom panel) of the dome and detector. Figure 10 shows data for an overcast day, 16 June The correlation between radiometer power measurement and temperature variations is also evident in this example. However, the difference in temperatures between the dome and the detector is smaller than in other cases. Cloudy and clear skies affect the thermal conditions in pyranometers in very different ways. This applies to both shaded and unshaded modes of operation. Even if clear sky shaded operation and overcast operations involve the measurement of diffuse radiation fields, the fundamental thermal processes affecting the pyranometer performance are quite different. In the case of cloud cover a strong infrared radiation field heats the dome, which is nearly black in the infrared. The net result is a reduced temperature difference between dome and detector and a corresponding reduction of the negative thermal offset in the output signal. As an example, Fig. 11 depicts output signal and temperatures for an overcast day, 2 October In this case, the temperature difference between dome and detector is reduced, compared to clear sky, resulting in a diminished offset (about 8 W m 2 ). Under certain conditions, the temperature difference was observed to even change sign, the detector was cooler than the dome, which resulted in positive offsets. FIG. 10. (a) Surface insolation and (b) temperature evolution on an overcast day, 16 Jun 1998.

8 172 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 17 TABLE 1. Average and standard deviation PSP offsets for unshaded daytime measurements. The before and after noontime conditions are considered independently. Cloud cover Frequency (10-min events) Before noon (W m 2 ) Avg Std dev After noon (W m 2 ) Avg Std dev Clear Partly cloudy Mostly cloudy Overcast FIG. 11. Output signal and temperatures for an overcast day. In this case, the temperature difference between dome and detector is reduced, compared to clear sky, resulting in a diminished offset (about 8Wm 2 ). Equation (1) again gives the same offset. Under certain conditions, the temperature difference was observed to even change sign, detector cooler than the dome, which resulted in positive offsets. 5. Analysis a. Unshaded For PSP operations in the unshaded mode, data were averaged for 10-min time intervals throughout the day and night. For each averaged dome and detector temperature pair, the corresponding PSP offset was determined via the calibration Eq. (1) and the D value. Only in the nighttime measurements was this offset value directly measured. The resulting averages provided about 900 points (150 h) of data during a variety of daytime measurements and about 500 points (80 h) during nighttime periods. 1) DAYTIME MEASUREMENTS Each 10-min period with the sun above the horizon was classified into one of four categories describing the atmospheric conditions: clear, partly cloudy, mostly cloudy, and overcast. Subjective criteria were used to separate the data into the different degrees of cloud cover existing at the time of measurement. Further discrimination was provided for periods before and after local solar noon. The results of the unshaded daytime analysis are presented in Fig. 12. Data are binned into 0.5Wm 2 intervals of the derived PSP offset parameter. Table 1 gives the mean and standard deviations of the PSP offset for each of the conditions shown. The negative derived offsets correspond to the dome temperature being less than the detector temperature. In only a few cases (mostly A.M. overcast conditions), the offsets were positive. For all conditions, the offsets after noon were greater in magnitude (more negative) than the before-noon values. Furthermore, the clear and mixed clear/cloud conditions all showed offsets significantly larger than the overcast case average. The overcast conditions, however, have the tightest distribution, as reflected by the magnitude of the standard deviations in Table 1. The broadness of the other three distributions indicate the varying conditions remaining in the cloud fraction classifications. FIG. 12. Histograms of the unshaded PSP data during daytime measurements. The four panels indicate the frequency of events for varying cloud cover conditions binned with respect to the PSP offset value. The portions before (dashed) and after (dotted) local solar noon are separated in the analysis. 2) NIGHTTIME MEASUREMENTS A marine stratus layer, better known in La Jolla as the June gloom, typically characterized overnight

9 173 FIG. 13. Directly measured PSP signal vs the derived PSP offset for nighttime measurements. The dashed line is the 1:1 reference line. The linear correlation coefficient between the measured and inferred offsets is 94.6%. FIG. 14. Histogram of the PSP offset before (solid) and after (dotted) the offset correction factor is applied. The corrected data have a much narrower distribution than the uncorrected data and is also centered about 0.0 W m 2. conditions during the tests completed in this study. As with the daytime measurements, data were averaged into 10-min intervals. Nighttime conditions were defined as the period approximately 30 min after sunset to 30 min before sunrise. The calibrated PSP signal is plotted versus the derived PSP offset in Fig. 13. The linear correlation between the two parameters is 94.6%, which implies excellent agreement between the measured radiometric signal and the corresponding offset, as determined from the D of the dome and detector. Furthermore, when subtracting this offset from the measurement, the actual dark signal offset results. Figure 14 represents the PSP offset both before and after the correction term is applied. The peak of the corrected distribution now occurs in the 0.0 Wm 2 bin and the corresponding average gives a value of W m 2. b. Shaded During five predominantly cloud-free days, the pyranometer was operated in the diffuse irradiance mode for portions of each day. Differences in the thermal equilibrium established before and after the shadow disk was in place are examined to determine its effect on the PSP measurements. Both dome and detector temperatures are averaged during time intervals when thermal quasi-stationary state is reestablished after either mounting or removing the shadow disk. The corresponding offsets determined from these temperatures and Eq. (1) are giving in Table 2. The unshaded offset value range from approximately 6 to 17 W m 2 and the corresponding shaded values are lower and range from approximately 16 to 23 Wm 2. The average change in the offset when introducing the shadow disk into the system is approximately 8.2Wm 2 with the shaded value being smaller (more negative) than the unshaded one. This change verifies the fact that the effect of shadowing the pyranometer changes the thermal environment of the measurement and consequently the associated offset value. 6. Algorithm proposed to improve performance and data quality It is proposed that an equation of the general form F a( T 4 Dome T 4 Detector ) b (4) be utilized to correct for systematic offset errors in the pyranometer, as deduced in section 3. The application of Eq. (4) to correct the data requires the modification of the pyranometer by adding temperature monitoring TABLE 2. Instrument offsets using Eq. (1) and measured dome and detector temperatures during unshaded and shaded clear sky conditions along with the corresponding differences. Date 17 Jun Jun Jul Jul Jul 1998 PSP offset (W m 2 ) Unshaded Shaded Difference

10 174 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 17 devices at appropriate locations inside the instrument, very much in a similar way as described in section 2a or as presently done with pyrgeometers. Additionally, a new calibration process, within the lines of that described above in section 3, is necessary to determine the coefficients, a and b, in Eq. (4) for each radiometer unit. Naturally, application of the correction algorithm to the pyranometer signal leads to the fact that the overall power calibration of the pyranometer will also need to take into account the correction for thermal offset errors. 7. Discussion It has been shown that the performance of the pyranometer is dependent on the environment in which it is operated and on the magnitude and variability of the radiant power being measured. Furthermore, the thermal balance between the components of the instrument must be accounted for to correct for errors resulting from offsets generated by the transfer of infrared radiation between the detector and the surrounding instrument components. One of the major consequences of the PSP offset is that it affects the determination of the absolute calibration constant of the pyranometer. As shown above, the uncertainties in the fluxes due to the thermally generated offsets associated with both the shaded and unshaded instruments will enter into the total uncertainty in the determination of the calibration constant. Accurate calibration of pyranometers requires that the thermal effects discussed above be accounted for during the calibration procedure by applying the appropriate corrections to the calibration and to the standard data. Many factors can affect the thermal environment during the PSP measurements. This study comprises a range of daytime sky conditions from extremely clear to heavy overcast during a 3-month summer period in a very mild marine environment. Typical to the specific measurement period, the nighttime conditions were predominantly overcast resulting in a fairly good thermal balance between the PSP dome and detector. The result of this near equilibrium is that the average nighttime offset is only about 1 Wm 2 (see Fig. 13). The excellent agreement between the inferred offset values and the directly measured nighttime signals supports the transfer of the laboratory calibrations to the PSP operating environment. Since the offsets depend so greatly on measurement conditions and on particular instrument designs and operating modes, it is impossible to generalize the behavior and thermal offset characteristics for all instrument designs, locations, and environments. For instance, clear sky Arctic conditions would affect the thermal balance differently than clear sky desert, tropical, or marine environments, or the use of different materials for the inner dome will affect differently the transfer of IR radiation between the detector and the outer dome. Using the temperature differential, which may be monitored coincidentally with the PSP power signal and in the exact configuration of the measurement, is an efficient means of correcting for the systematic PSP offset error. 8. Conclusions As discussed in this paper, it is possible to correct for thermal offset errors in pyranometers. An algorithm and thermal offset calibration procedure is proposed to improve the accuracy of the data. The implementation of this procedure requires 1) accurate temperature sensors installed in appropriate locations, 2) thermal calibration to determine the thermal calibration constants [Eq. (4)], 3) a change of the power calibration methods to include the thermal offset correction algorithm, and 4) application of the thermal offset correction to the standard observational data. It should also be noted that particular designs and environmental circumstances might result in smaller or larger corrections than those found in this study for the PSP instrument. To maximize the quality of surface insolation data acquired with pyranometers, it is suggested that the inherent thermal offset errors be evaluated in every case and that the proposed algorithm be applied when necessary. Acknowledgments. The authors wish to thank Jürgen M. Lobert and Jens Meywerk of the Center for Clouds, Chemistry, and Climate, Center for Atmospheric Sciences, at Scripps Institution of Oceanography, University of California, San Diego for facilitating the use of a pyranometer and datalogger for this study. This research was partially funded by NASA Grant NAG to the Atmospheric Research Laboratory, Scripps Institution of Oceanography, University of California, San Diego. REFERENCES Arking, A., 1996: Absorption of solar energy in the atmosphere: Discrepancy between models and observations. Science, 273, Bener, P., 1950: Untersuchung uber die Wirkungsweise des Solarigraphen Moll-Gorczynski. Beitrage zur Strahlungsmethodik III. Arch. Meteor. Geophys. Bioklimatol., B2, Bush, B. C., and F. P. J. Valero, 1999: Comparison of ARESE clear sky surface radiation measurements. J. Quant. Spect. Rad. Trans., 61, , S. K. Pope, A. Bucholtz, F. P. J. Valero, and A. Strawa, 1999: Surface radiation measurements during the ARESE campaign. J. Quant. Spect. Rad. Trans., 61, Cess, R. D., and Coauthors, 1995: Absorption of solar radiation by clouds: Observations versus models. Science, 267, ,, Y. Zhou, X. Jing, and V. Dvortsov, 1996: Absorption of solar radiation by clouds: Interpretations of satellite, surface, and aircraft measurements. J. Geophys. Res., 101,

11 175 Charlock, T. P., and T. L. Alberta, 1996: The CERES/ARM/GEWEX experiment (CAGEX) for the retrieval of radiative fluxes with satellite data. Bull. Amer. Meteor. Soc., 77, Conant, W. C., V. Ramanathan, F. P. J. Valero, and J. Meywerk, 1997: An examination of the clear-sky solar absorption over the central equatorial Pacific: Observations versus models. J. Climate, 10, Drummond, K. L., and J. J. Roche, 1965: Corrections to be applied to measurements made with Eppley (and other) spectral radiometers when used with Schott colored glass filters. J. Appl. Meteor., 4, Francis, P. N., J. P. Taylor, P. Hignett, and A. Slingo, 1997: On the question of enhanced absorption of solar radiation by clouds. Quart. J. Roy. Meteor. Soc., 123, Gulbrandsen, A., 1978: On the use of pyranometers in the study of spectral solar radiation and atmospheric aerosols. J. Appl. Meteor., 17, Halthore, R., S. Nemesure, S. Schwartz, D. Imre, A. Berk, E. Dutton, and M. Bergin, 1998: Models overestimate diffuse clear-sky surface irradiance: A case for excess atmospheric absorption. Geophys. Res. Lett., 25, Imre, D. G., E. H. Abramson, and P. H. Daum, 1996: Quantifying cloud-induced shortwave absorption: An examination of uncertainties and of recent arguments for large excess absorption. J. Appl. Meteor., 35, Kato, S., T. P. Ackerman, E. E. Clothiaux, J. H. Mather, G. G. Mace, M. L. Wesely, F. Murcray, and J. Michalsky, 1997: Uncertainties in modeled and measured clear-sky surface shortwave irradiances. J. Geophys. Res., 102, Li, Z., and L. Moreau, 1996: Alteration of atmospheric solar radiation by clouds: Simulation and observation. J. Appl. Meteor., 35, , H. W. Barker, and L. Moreau, 1995: The variable effect of clouds on atmospheric absorption of solar radiation. Nature, 376, Michalsky, J., M., Rubes, T. Stoffel, M. Wesely, M. Splitt, and J. DeLuisi, 1997: Optimal measurement of surface shortwave irradiance using current instrumentation The ARM experience. Proc. Ninth Conf. on Atmospheric Radiation, Long Beach, CA, Amer. Meteor. Soc., J5 J9. Morikofer, W., 1939: Meteorologische Strahlunsmessmethoden. Handbuch der Biologischen Arbeitsmethoden, E. Abderhalden, Ed., Urban & Schwarzenberg, Özisik, M. N., 1985: Heat Transfer: A Basic Approach. McGraw- Hill, 780 pp. Pilewskie, P., and F. P. J. Valero, 1995: Direct observations of excess solar absorption by clouds. Science, 267, , and Coauthors, 1998: Observations of the spectral distribution of solar irradiance at the ground during SUCCESS. Geophys. Res. Lett., 25, Ramanathan, V., B. Subasilar, G. J. Zhang, W. Conant, R. D. Cess, J. T. Kiehl, K. Grassl, and L. Shi, 1995: Warm pool heat budget and shortwave cloud forcing: A missing physics? Science, 267, Robinson, N., 1966: Solar Radiation. Elsevier, 347 pp. Rodskjer, N., 1971: A pyranometer with dome of RG8 for use in plant communities. Arch. Meteor. Geophys. Bioclimatol., B19, Smith, W. L., S. K. Cox, and V. Glover, 1988: A thermopile temperature sensitivity calibration procedure for Eppley broadband radiometers. NCAR Tech. Note NCAR/TN-320 STR, 20 pp. [Available from NCAR, Atmospheric Technology Division, P. O. Box 3000, 1850 Table Mesa Drive, Boulder, CO ] Stephens, G. L., 1996: How much solar radiation do clouds absorb? Science, 271, , and S. C. Tsay, 1990: On the cloud absorption anomaly. Quart. J. Roy. Meteor. Soc., 116, Valero, F. P. J., and B. C. Bush, 1999: Measured and calculated clear sky solar radiative fluxes during SUCCESS: A sensitivity study. J. Geophys. Res., in press., R. D. Cess, M. Zhang, S. K. Pope, A. Bucholtz, B. C. Bush, and J. Vitko, 1997a: Absorption of solar radiation by the cloudy atmosphere: Interpretations of collocated aircraft measurements. J. Geophys. Res., 102, , A. Bucholtz, B. C. Bush, S. K. Pope, W. D. Collins, P. Flatau, A. Strawa, and W. J. Y. Gore, 1997b: Atmospheric Radiation Measurements Enhanced Shortwave Experiment (ARESE): Experimental and data details. J. Geophys. Res., 102, Waliser, D. E., W. D. Collins, and S. P. Anderson, 1996: An estimate of the surface shortwave cloud forcing over the western Pacific during TOGA COARE. Geophys. Res. Lett., 23, Wiscombe, W., A. Marshak, E. G. Dutton, D. Nelson, B. C. Bush, and F. P. J. Valero, 1998: Summer 1997 comparison of surface shortwave downfluxes from Valero RAMS and world standard cavity shaded pyranometer combinations. Preprints. 8th Annual Atmospheric Radiation Measurement Science Team Meeting, Tucson, AZ, Atmospheric Radiation Measurements, Zender, C. S., B. C. Bush, S. K. Pope, A. Bucholtz, W. D. Collins, J. T. Kiehl, and F. P. J. Valero, 1997: Atmospheric absorption during the Atmospheric Radiation Measurement (ARM) Enhanced Shortwave Experiment (ARESE). J. Geophys. Res., 102,