Portal de Periódicos da CAPES

A novel nonintrusive method to resolve the thermal dome effect of pyranometers: Radiometric calibration and implications

2011; American Geophysical Union; Volume: 116; Issue: D24 Linguagem: Inglês

10.1029/2011jd016466

2156-2202

Q. Ji, S. Tsay, Kei May Lau, Richard A. Hansell, James J. Butler, J. Cooper,

Atmospheric Ozone and Climate

Journal of Geophysical Research: AtmospheresVolume 116, Issue D24 Climate and DynamicsFree Access A novel nonintrusive method to resolve the thermal dome effect of pyranometers: Radiometric calibration and implications Q. Ji, Q. Ji qiang.ji-1@nasa.gov ESSIC, University of Maryland, College Park, Maryland, USA Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorS.-C. Tsay, S.-C. Tsay Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorK. M. Lau, K. M. Lau Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorR. A. Hansell, R. A. Hansell ESSIC, University of Maryland, College Park, Maryland, USA Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorJ. J. Butler, J. J. Butler Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorJ. W. Cooper, J. W. Cooper Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USA Sigma Space Corporation, Greenbelt, Maryland, USASearch for more papers by this author Q. Ji, Q. Ji qiang.ji-1@nasa.gov ESSIC, University of Maryland, College Park, Maryland, USA Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorS.-C. Tsay, S.-C. Tsay Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorK. M. Lau, K. M. Lau Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorR. A. Hansell, R. A. Hansell ESSIC, University of Maryland, College Park, Maryland, USA Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorJ. J. Butler, J. J. Butler Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USASearch for more papers by this authorJ. W. Cooper, J. W. Cooper Earth Sciences Division, NASA Goddard Space Flight Center, Greenbelt, Maryland, USA Sigma Space Corporation, Greenbelt, Maryland, USASearch for more papers by this author First published: 17 December 2011 https://doi.org/10.1029/2011JD016466Citations: 12AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Abstract [1] Traditionally the calibration equation for pyranometers assumes that the measured solar irradiance is solely proportional to the thermopile's output voltage; therefore, only a single calibration factor is derived. This causes additional measurement uncertainties because it does not capture sufficient information to correctly account for a pyranometer's thermal effect. In our updated calibration equation, temperatures from the pyranometer's dome and case are incorporated to describe the instrument's thermal behavior, and a new set of calibration constants are determined, thereby reducing measurement uncertainties. In this paper, we demonstrate why a pyranometer's uncertainty using the traditional calibration equation is always larger than a few percent, but with the new approach can become much less than 1% after the thermal issue is resolved. The highlighted calibration results are based on NIST traceable light sources under controlled laboratory conditions. The significance of the new approach lends itself to not only avoiding the uncertainty caused by a pyranometer's thermal effect but also the opportunity to better isolate and characterize other instrumental artifacts, such as angular response and nonlinearity of the thermopile, to further reduce additional uncertainties. We also discuss some of the implications, including an example of how the thermal issue can potentially impact climate studies by evaluating aerosol's direct radiative effect using field measurements with and without considering the pyranometer's thermal effect. The results of radiative transfer model simulation show that a pyranometer's thermal effect on solar irradiance measurements at the surface can be translated into a significant alteration of the calculated distribution of solar energy inside the column atmosphere. Key Points The uncertainty of a pyranometer is <1% if its thermal effect is accounted for Radiometric calibrations were performed using NIST-traceable light sources An improved pyranometer can play an important role in climate studies 1. Introduction [2] The measurements of solar irradiance over the globe play an important role in climate studies because solar radiation is the primary driving force of climate. Pyranometers are used worldwide to measure solar irradiance. Traditionally, the uncertainty of a high-quality pyranometer is greater than a few percent, which is not desired for studying long-term climate change. For example, the radiative forcing due to doubling the amount of CO2 in the atmosphere is ∼4 W m−2 [Ad Hoc Study Group on Carbon Dioxide and Climate, 1979], and the global average of radiative forcing for different agents and mechanisms is around a couple of W m−2 [Pachauri and Reisinger, 2007]. However, a ± 2% uncertainty (specifications of pyranometer) [e.g., World Meteorological Organization (WMO), 2008] on the ∼184 W m−2 (∼263 W m−2 if cloud free) global annual mean solar radiation at the Earth's surface [Trenberth et al., 2009] is over ± 3.6 W m−2, comparable to or larger than the expected ranges of the climate change signals. A reduced uncertainty in pyranometer measurements will have profound implications, such as gaining the ability to better discern the radiative forcing, or the global dimming and brightening [Wild, 2009]. Furthermore, a more accurate data set of solar irradiance will also help improve the evaluation of the effects of clouds and aerosols on climate studies. An example of the latter effect is explored later in the paper. [3] A pyranometer's larger than 2%∼3% uncertainty [WMO, 2008] is related to its thermal dome effect (TDE), which traditionally is not quantified. Shading a pyranometer can help reveal and infer its thermal effect [e.g., Gulbrandsen, 1978]; however, it does not provide a direct nonintrusive measurement of TDE. In addition, any object casting the shade also imposes its own contribution dynamically to the thermal environment. This situation has changed since the studies that explicitly describe the TDE [e.g., Bush et al., 2000; Ji and Tsay, 2000; Haeffelin et al., 2001; Dutton et al., 2001]. Continuing the effort to directly addressing the thermal issue, Ji and Tsay [2010] have introduced an innovative nonintrusive method that enables a reliable and routine monitoring of TDE; thus this thermal effect of pyranometer can be correctly accounted for in measurements of solar irradiance. [4] Historically, the measured solar irradiance, I, is regarded as being simply proportional to the output voltage of the pyranometer's thermopile, V: where ch is an empirical calibration factor. Regardless of how ch is determined, this traditional calibration equation carries a large uncertainty because it is incapable of fully capturing the dynamic thermal behavior of the pyranometer. To reduce the uncertainty, we have derived a new calibration equation: where Ts is the temperature of the receiving surface of the thermopile, and Td is the effective dome temperature. In addition, σ = 5.67 × 10−8 J s−1 m−2 K−4 is the Stefan–Boltzmann constant; and the intrinsic calibration constants c and f are stable physical properties [Ji and Tsay, 2000]. f is also called the "dome factor," a quantity related to the dome emissivity and the geometry of a pyranometer (discussed in this paper). Furthermore, because the thermopile's output is referenced to the temperature of the pyranometer's case (i.e., Tc), Ts can be determined by Ts = Tc + αV, where α is a thermopile parameter, and we regarded it as a constant in our previous study [Ji and Tsay, 2010]. A schematic of a pyranometer illustrating a couple of our nonintrusive methods for deriving Td is described in Appendix A. [5] The difference between the traditional and the new calibration equations is obvious. The former has one measurement (i.e., V) and a single empirical factor (i.e., ch), unable to properly represent the TDE; while the latter incorporates three measurements (i.e., V, Tc, and Td) and four true physical constants (i.e., c, f, σ, and α), able to quantify the TDE by the temperature-related term. A variety of methods for calibrating pyranometers are listed in WMO's Guide to Meteorological Instruments and Methods of Observations [WMO, 2008, chapter 7.3.1]. Since they all depend on equation (1), we refer this type of calibration as a V1C1 calibration; standing for one measured variable, and one constant to be determined. For example, whenever a single calibration constant is provided by a manufacture or a research laboratory such as Broadband Outdoor Radiometer Calibrations (BORCAL at the National Renewable Energy Laboratory, available at http://www.nrel.gov/aim/borcal.html), it is a V1C1 calibration. Under this convention a method using equation (2) becomes a V3C4 calibration. However, because σ is an universal constant, and following our previous studies we assume α is a know constant, we are effectively dealing with a V3C2 calibration in this study. Additionally, our new method is backward compatible, because the V1C1 calibration remains intact and available. [6] A characteristic listed in the specification of a pyranometer [WMO, 2008] is its "zero offset," but how it affects the uncertainty of a pyranometer is not specified. In order to correct the offset errors in measurements of diffuse solar irradiance, Dutton et al. [2001] developed a data correction procedure where the correction factor is derived from the output of a collocated pyrgeometer that measures atmospheric infrared irradiance. This is an indirect approach involving multiple instruments. These results are useful if the diffuse component is required when calibrating total irradiance radiometers by reference to a standard pyrheliometer and a shaded reference pyranometer [International Organization for Standards, 1993]. The method is expensive if used in the routine measurement of total irradiance because it requires a pyrheliometer for the direct solar component, a shaded pyranometer for the diffuse sky component, and a shaded pyrgeometer for the thermal correction of the shaded pyranometer. In addition, all three instruments must be mounted on a solar tracker; and the pyranometer and pyrgeometer need to be ventilated. Another complication is related to the pyrheliometer being subject to thermal effects (discussed in this paper). In contrast, our new method eliminates the "zero offset" directly from a pyranometer [Ji and Tsay, 2010]. [7] Different design and make of pyranometers are available commercially. Their response to thermal effect are different too [e.g., Michalsky et al., 2003]. We focused on Epply Precision Spectral Pyranometer (PSP) (available at http://www.eppleylab.com), because its design has remained relatively unchanged for decades; therefore, there are relatively longer records of consistent measurement, which is useful for climate change studies. Nevertheless, our new method and discussions are not limited to PSP, we expect them to be also valuable in improving other instruments and measurements. [8] This paper is structured as follows. In section 2we present our method for radiometric calibration of a pyranometer against a calibrated NIST-traceable light source. Insection 3 we show the measurement results illustrating the weakness of the V1C1 calibration and the advantage of the V3C2 one. In section 4 we discuss implications of this new approach, including some of the knowledge gained, followed by an example of model simulation demonstrating the potential impact of TDE on climate studies. Finally, the conclusion and future work are presented in section 5. 2. Radiometric Calibration [9] In this study, integrating spheres maintained in a clean room of the Radiometric Calibration Facility at NASA Goddard Space Flight Center (available at http://spectral.gsfc.nasa.gov) are used as references to examine both the traditional and new calibration equations. Figure 1aillustrates a PSP mounted in front of an integrating sphere. This light source is calibrated by comparison with NIST-calibrated standard irradiance lamps using monochromators [Walker et al., 1991]. Its spectral radiance is shown in Figure 1b. Figure 1Open in figure viewerPowerPoint (a) Grande is a 1 m Teflon integrating sphere with an aperture of 25.40 cm in diameter. It has 9 independently controlled lamps for producing different levels of radiance. A PSP can be mounted with the thermopile facing the aperture. (b) The irradiance from Grande is determined by a combination of its spectral radiance, viewing geometry, and spectral cut-off of PSP's dome. The gray curve illustrates a transmittance for WG295 glass. The thick curve highlights the currently NIST-traceable part of spectral radiance (i.e., between 0.4 and 2.4μm). [10] The calibration procedure is straightforward: for V1C1, a calibration factor is determined by ch = {I/V} for equation (1); for V3C2, c and f can be derived as the intercept and slope of a linear fit from {I/V} = c + f{σ(Ts4 − Td4)/V} for equation (2). In addition, although Td is not directly measured, it can be verified under special conditions from Tc by maintaining a pyranometer in the dark until a thermal equilibrium state is achieved (i.e., V = 0 and Td = Tc). [11] Note that the parameter space over which the overall calibration uncertainty of a pyranometer can be adequately quantified is a much bigger question to tackle. To properly address this question will require the collective efforts of the scientific community as a whole whereby the current methodology can be fully tested in different environments under different situations. Here we regard a light source as an absolute radiometric reference for showing how V1C1 and V3C2 track a reference, thereby demonstrating why measurements can be improved. 3. Results [12] Eight rounds of calibration are depicted in Figure 2 as an example to illustrate the effects of TDE. The measured temperatures are shown in Figure 2a, where Ts and Tc are plotted as the thin curves at the top and on the bottom, respectively. In the middle is Td shown as the bold curve, which is proportional to the pressure of the air sealed between the domes as plotted in Figure 2b. Figure 2Open in figure viewerPowerPoint Example of eight rounds of calibration which took about 3 h. (a) The light is turned on at point A, making all the measured temperatures in a PSP to increase (see section 3). The light is blocked at point B about 15 min later, causing the temperatures to decrease and approach each other. The thermal equilibrium is reached after several minutes. The light is turned on again at point C, followed by a reduction of the light intensity at point D. The similar pattern is repeated in the later rounds. (b) The pressure in the sealed volume between domes (compare Appendix A). (c) The varying output voltage of thermopile. See section 3 for details. [13] This example reveals that Ts responds immediately when a PSP is exposed to light, while Td changes slowly and Tc lags further behind. Naturally, in response to the "solar heating," the larger the thermal mass is, the smaller the rate of change will be. When it is blocked from light, Ts approaches Tc promptly while all the temperatures decrease, and the thermal gradient diminishes after several minutes. Once exposed to light the second time, all temperatures start to rise again, but begin with a higher value due to the heating in the previous round, and will reach higher levels. Notice that if we just reduce the intensity of light instead of fully blocking it to return the PSP to dark conditions, then Td will decrease, but other temperatures may still be increasing with smaller rates under the reduced heating than in the previous round; however, all temperatures start to decrease if the irradiance is further reduced to a point that the heating became insufficient. In this example the irradiance was set to 879.6 W m−2 in rounds 1 and 2, then reduced to 668.3 W m−2 in rounds 3 to 5, and further reduced to 455.2 W m−2 in rounds 6 and 7, and finally to 244.3 W m−2 in round 8. Generally the system warms up in the early rounds when irradiance is larger and cools down later when irradiance becomes smaller, which to a certain extent represents morning and afternoon conditions, respectively. [14] It is important to point out that in each round of calibration the output voltage of the thermopile varies noticeably and does not reach a stable value for a prolonged period of time while the variation of the light source is negligible (less than 0.1%), as illustrated in Figure 2c. The changing output voltage is cause by the TDE. In this example the transitional period lasted about 10 min in the first round when the temperature swing was relatively large. The thermal gradient inside the PSP diminishes faster in the later rounds when the irradiance is smaller. 3.1. V1C1 Calibration [15] The limitation of a V1C1 calibration is demonstrated in Figure 3, where the curve shows the calibration factor (i.e., I/V, or ch). Evidently, ch neither starts at a fixed value nor reaches a fixed one in each round of calibration. This is unwanted for equation (1), because it may make a PSP's thermopile seemingly nonlinear against temperature. However, it is expected according to equation (2), which predicts that if I remains stable, then the larger the TDE, the smaller the V, and vice versa. Figure 3Open in figure viewerPowerPoint The I/V is the calibration factor that should be a constant according to the traditional calibration equation. For example, it should remain at 133.95 W m−2 mV−1 according to a BORCAL result, marked by the solid line. In reality I/V varies with temperature across a wide range as shown in the thick curve. See section 3.1 for details. [16] Also marked in Figure 3 is a V1C1 calibration from BORCAL (i.e., ch = 133.95 W m−2 mV−1, +2.84%, −4.43%, for PSP#33109F3). Notice its large uncertainty range, which is typical in a V1C1 method related to the lack of an accurate interpretation of the thermal effect. To correct the errors in the V1C1 calibrated measurements, a correction method such as the one developed by Dutton et al. [2001] is needed. 3.2. V3C2 Calibration [17] The potential of using the V3C2 calibration equation is demonstrated in Figure 4a, where the calibration constants (i.e., mean value ± standard deviation) are based on four rounds of calibration; two each at I = 879.6 and 668.3 W m−2. Notice that the linearity of the curve is an indication on how well equation (2) can capture the reality. To show repeatability, Figure 4b overlays twelve more rounds that include I = 455.2, 244.3, also 960.3, 729.7, 497.0, and 266.7 W m−2; realized by turning on different lamps in the integrating sphere, and by altering the distance between the PSP and the integrating sphere. In this particular example, the I/V versus σ(Ts4 − Td4)/V curve shifted slightly during those extra rounds when irradiances were smaller; however, the slope (i.e., f) remains relatively unchanged after a shift, which is consistent with f being a stable factor independent of the uncertainty in I and c. We found that the shift does not diminish when the PSP's "thermopile temperature compensation circuit" is disabled, indicating that it is not caused by the slight drift of temperature during these particular rounds. The shift can be caused by several reasons, such as biases in I, nonlinearity in V, or error in Td. Future work is necessary to resolve the issue to further reduce the uncertainty. Figure 4Open in figure viewerPowerPoint Example of the new calibration. (a) A consistent result from selected rounds. (b) Result including additional rounds. (c) Same as Figure 4a but without modification to the PSP (compare Appendix A). See section 3.2 for details. [18] In order to determine Td, we sealed the space between the inner and outer domes of a PSP to create a constant volume gas thermometer [Ji and Tsay, 2010]. This method is nonintrusive because it does not block the field of view of thermopile; however, it modifies a PSP slightly. To evaluate whether this modification impacts the performance of the PSP, we tested one of our alternative methods that does not require such modification (see Appendix A). Although this alternative method is less straightforward in determining Td, and produces slightly more noise in the calibration results as shown in Figure 4c, it indicates that a PSP's performance remains consistent regardless of modification. 3.3. Contrast Between the Two Results [19] An example of comparing the V1C1 and the V3C2 calibrations is given in Figure 5. When exposed to a known irradiance of 893.7 W m−2, the V1C1 result reached 870 W m−2 initially, then drifted to 901 W m−2 in about 10 min. Once repeated in the next round, it approached a couple of W m−2 lower, corresponding to about 1°C rise in temperature. In contrast, the result from the V3C2 calibration started to track the known irradiance within a few seconds every time. Figure 5b highlights a situation when light was blocked. After an initial drop, it took over 10 min for the V1C1 calibrated result to decrease from about 30 to 0 W m−2, while the V3C2 calibrated result promptly jumped to 0 W m−2. The high precision of the new calibration is evident. Figure 5Open in figure viewerPowerPoint The traditional (thick curve) versus the new (thin curve) calibration results. The sampling rate is about every 5 s. The PSP is exposed to light at point A, then blocked from light at point B, and exposed to light again at point C. The arrows highlight that the new calibration tracks the correct irradiance promptly. (a) Notice that the traditional result is slightly different in period C to D from the result in period A to B. (b) The prolonged nonzero results of the traditional calibration in dark conditions. See section 3.3 for details. [20] According to the WMO specification, the response time of a high-quality pyranometer is less than 15 s. This is based on using a V1C1 calibration for the measurement to reach 95% of final value. In contrast, it will reach 99.9% of final value within 15 s in a V3C2 calibration. More importantly, the results fromequation (2) will remain consistent when the calibration is repeated, which reflects that because c and f represent stable physical properties their values can be more accurately determined over time with better statistics. This is unachievable in a V1C1 calibration, because the empirical factor in equation (1) depends on environmental conditions and does not converge to a constant. 4. Implications [21] There are many important implications in realizing that a pyranometer can have a much smaller uncertainty once its thermal effect is accounted for. For example, we used to assign each PSP to a specific ventilator hoping the set would maintain a consistent thermal characteristic. We also tried to use reversed ventilation to reduce the thermal effect. With the new calibration, ventilation is rendered noncritical concerning the thermal effect. This makes it easy for us to deploy a solar-powered network of PSPs and to generate a more consistent data set, in light of that a consistent global data set is of importance for climate studies. A few other things we have learned are briefly highlighted in the following. 4.1. Different Domes and Dome Factors [22] We tested several "clear domes" (WG295 glass, available at http://schottglass.com) on the same PSP, and found that they yield a consistent result using the new calibration. Other than WG295, we also use "color domes," such as GG395, for selecting spectral bands from 0.4 to 3 μm and RG695 for 0.7 to 3 μm. Traditionally a PSP is calibrated with a clear dome; and when a color dome is needed for quantifying the energy partitioning in solar irradiance, an empirical scale factor is applied. It is difficult to determine and to justify a scale factor without knowing the TDE. [23] With equation (2), we treat all types of domes equally except when considering the corresponding spectral transmittance in the determination of the irradiance from the light source. We found that an RG695's TDE changes faster and varies over a larger range but can balance at a smaller value than a WG295's. It is because an RG695 dome absorbs more solar radiation than a WG295 dome does; therefore, its Td becomes closer to Ts than with a WG295 dome. The resulting distinct thermal behavior indicates that a scale factor is not appropriate for an RG695 in terms of a highly accurate and consistent measurement. [24] As listed in Table 1, an RG695 dome has a slightly larger dome factor than a WG295 dome, which in turn has a slightly larger dome factor than a quartz dome (i.e., 1.8 versus 1.5 versus 1.3 for PSP#33109F3). In theory, for an idealized pyranometer whose thermopile's receiving surface occupies the whole area underneath the dome, its dome factor is the emissivity of dome which is smaller than 1.0 [Ji and Tsay, 2000]. In reality, the receiving surface in a PSP only covers a small fraction of the area, allowing its surroundings to contribute noticeably to TDE, leading to a larger "effective value" of f. It may be improved in the future with a better calibration equation. Table 1. Sample Calibration Results of a PSP Using Different Types of Domes Type of Dome f c (W m−2 mV−1) WG295 #1 1.5 130 WG295 #2aa Same inner dome as in WG295 #1. 1.5 130 GG395 (yellow dome)aa Same inner dome as in WG295 #1. 1.5 124 RG695 (red dome)aa Same inner dome as in WG295 #1. 1.8 128 WG295 #3bb A different set of inner and outer domes. 1.5 130 Quartzcc The transmittance is assumed to be 100% between 0.2 and 3.5 μm, 0% elsewhere. 1.3 129 WG295 #4dd Yet another set of inner and outer domes; mounted on a brass collar for testing. 1.6 132 WG295 #4ee Same as above except sealed. 1.7 133 WG295 #5ff One more set of sealed inner and outer domes. 2.0 130.3 WG295 #5gg Same as above, except unsealed. 1.9 130 a Same inner dome as in WG295 #1. b A different set of inner and outer domes. c The transmittance is assumed to be 100% between 0.2 and 3.5 μm, 0% elsewhere. d Yet another set of inner and outer domes; mounted on a brass collar for testing. e Same as above except sealed. f One more set of sealed inner and outer domes. g Same as above, except unsealed. 4.2. Essence of Thermal Effect [25] To truly understand TDE it is important to realize that the thermal effect is not limited to a PSP-like pyranometer. Generally speaking, there will be "TDE" as long as the detector senses any blockages thermally in its field of view. A blockage can be a dome or a collimator and so on. An Eppley Normal Incidence Pyrheliometer (NIP) is an example: without a large thermal mass, a NIP can approach thermal equilibrium more easily than a PSP does, therefore displays a smaller "nighttime offset"; however, the sunlight or a sudden change of air temperature can cause a large thermal gradient along the cylinder of a NIP, introducing a significant thermal effect. [26] Figure 6ashows the output of a NIP during four rounds of tests when it is alternatively exposed and blocked from the NIST-traceable light source. Similar to a PSP, after an initial quick response to the change of light, the output of the NIP starts to drift slowly depending on temperature and its gradient in the instrument. As long as the temperature changes, the NIP's output will drift and balance to a different value. In