Erythemally Weighted UV Irradiance

One of the data types most frequently requested from UVMRP is a compilation of erythemally weighted irradiance over a specified geographic region and time period. The UVMRP employs YES UVB-1 pyranometers to measure the Commission Internationale de l'Eclairage (CIE) weighted erythemal irradiances (McKinlay and Diffey, 1987) at each climatological site. Initially, the UVB-1 pyranometers were calibrated and characterized annually by the instrument manufacturer. Comparisons between collocated radiometers show an agreement of better than ±2.3% when SZA is less than 80°,while absolute errors might reach 10% when SZA is greater than 80° (Bigelow et al., 1998; McKenzie et al., 2006). Since 1997, the broadband pyranometers have been calibrated and characterized annually by CUCF. Characterizations include the tests of cosine response and spectral response. The scale factor for erythemal calibration is determined by matching the output of the pyranometer to the standard triad at CUCF. The absolute calibration factor of the broadband pyranometer is determined annually by comparing the output of the broadband sensor in sunlight with that of three collocated UVB-1 standard pyranometers (triad). The inter-comparison with the triad typically lasts from one to two months in order to obtain sufficient data for the calibration that includes corrections for different amounts of columnar ozone in each zenith angle regime. The standard triad is in turn frequently calibrated for erythemal action spectrum as a function of SZA and total column ozone in the field against a collocated precision spectral radiometer employing the principle described in the literature (Grainger et al., 1993; Lantz et al., 1999; Xu and Huang, 2000; Vijayaraghavan and Goswami, 2002; Xu and Huang, 2003). This practical calibration scheme is supported by the uniformity of the cosine responses of different broadband pyranometers in the UVMRP network. The uncertainty in this calibration is from two sources. One consists of the random and systematic uncertainties of the ozone and SZA dependent erythemal calibration factor. The other is the difference in the scale factor for each pyranometer.

Although the absolute calibration uncertainty of the pyranometers could reach as high as +10%, relative uncertainties among the pyranometers are more important in certain applications (McKenzie et al., 2006). Differences in spectral response and angular response are the major sources for relative uncertainties. These differences are inherent in the calibration procedure, which is a comparison of the solar signal from the pyranometer to the triad as measured at CUCF. In order to examine the relative accuracy of the instruments, experiments that compared the solar signal from the standard triad to the field site UVB-1 pyranometers were conducted as the instruments cycled through the calibration process. The tests were performed as a function of SZA over 110 two-week periods under various weather conditions. The results indicate the rather remarkable level of consistency amongst the pyranometers in that, on average, the pyranometers differ from the standard triad by 0.1% at an SZA of 20°, and by 2.8% at an SZA of 80° for all 51 tested instruments. Other collocated measurements of the UVB-1 pyranometers also show a good agreement with the broadband pyranometers from other manufacturers, as well as within the tested UVB-1 pyranometers (Seckmeyer et al., 2007). Since a generic factor is applied for ozone-dependent calibration, differences of spectral responses between the pyranometers may introduce uncertainties. In order to more fully characterize the individual pyranometers, their signals are corrected for cosine response, and for the purposes of determining erythemal irradiance, individual characterizations as a function of zenith angle and total column ozone are made as discussed below.

8.4.2.1 Angular (Cosine) Response of the UVB-1 Pyranometer

The erythemal irradiance is almost always supplied as a total horizontal quantity. Thus, the angular response of each UVB-1 instrument must be utilized to correct for imperfect cosine response. The angular response is tested annually by performing a north-south scan and an east-west scan with an interval of one degree zenith angle. The average of the two scans is considered as the cosine response. Figure 8.3 displays the relative angular responses of 10 UVB-1 pyranometers that were characterized by the CUCF in 2004. The measurements are normalized to 0° SZA. These results have the typical shapes of those reported by previous studies (Grainger et al., 1993; Mayer and Seckmeyer, 1996; Landelius and Josefsson, 2000). The differences of the angular responses from an ideal cosine response have been discussed by Bigelow et al. (1998). Although the departure from the ideal cosine response exceeds 10% beyond 60° SZAs (Bigelow et al., 1998), the variability between different pyranometers is rather small. Figure 8.3 actually displays the cosine response of ten instruments and shows that the agreement is within the width of the plotted line. Figure 8.4 shows the standard deviations of the cosine responses of 20 randomly selected pyranometers. The variances tend to increase with SZA, but the standard deviations are within 0.02% in the zenith angle range of 0° - 89°. This repeatability among the angular responses of the UVMRP broadband pyranometers suggests that a single generic cosine correction may be applied to all pyranometers in the network without introducing significant uncertainties.

-60 -30 0 30 Solar 2enith angle (degree)

Figure 8.3 Normalized cosine responses of ten UVB-1 pyranometers. Numbers in the legend indicate the serial numbers of the sensors. The 10 curves are overlaid over each other. It is difficult to visibly separate them, which exhibits uniform cosine responses

Figure 8.4 Standard deviations of cosine responses of 20 UVB-1 pyranometers as a function of SZA. The results show that the responses are within 0.02% of the mean cosine response and generally within the digitization error of the data logger

Figure 8.4 Standard deviations of cosine responses of 20 UVB-1 pyranometers as a function of SZA. The results show that the responses are within 0.02% of the mean cosine response and generally within the digitization error of the data logger

8.4.2.2 UVB-1 Spectral Response and Influence of Columnar Ozone

The erythemally weighted irradiance is determined by convolving the CIE action spectrum (McKinlay and Diffey, 1987) shown in Fig. 8.5 with the measured spectral irradiance. Measuring the erythemal UV irradiances requires that the spectral response of the pyranometer is identical with the CIE action spectrum because the spectral distribution of the solar radiation varies with many factors, such as ozone content, SZA, clouds, and other components in the atmosphere. However, the spectral response of the UVB-1 broadband pyranometer does not perfectly simulate the CIE action spectrum (Fig. 8.5), especially in the regions of shorter and longer wavelengths. Therefore, all factors that affect the wavelength distribution of solar irradiance in the erythemal wavelength band will impose potential impacts on the measurements of erythemal UV irradiance. The most significant variables are SZA, total ozone content and aerosol optical properties (Bodhaine et al., 1998; Lantz et al., 1999). Considering the effects of ozone content and SZA, the erythemal calibration factors for the UVB-1 broadband pyranometers are determined by the CUCF as a function of SZA and total ozone, or as a function of SZA with the assumption that the overhead total column ozone is 300 DU. The latter calibration factors are more conveniently applied when ozone measurements are not available at a site. The UVMRP data set provides erythemal UV irradiance data with the calibration factors of SZA dependence that assume overhead column ozone is 300 DU. Due to variable ozone contents and SZA, this assumption may result in an error of up to 25% (Fig. 8.6) for the ozone range of 200 DU - 500 DU for different SZAs. For a more accurate erythemal irradiance data set, however, ozone correction will be necessary. Lantz et al. (1999) detailed

280 320 360 400

Wavelength (nm)

Figure 8.5 CIE action spectrum and spectral responses of three UVB-1 pyranometers. The numbers in the legend indicate the serial numbers of the tested sensors

280 320 360 400

Wavelength (nm)

Figure 8.5 CIE action spectrum and spectral responses of three UVB-1 pyranometers. The numbers in the legend indicate the serial numbers of the tested sensors

200 300 400 500

Tola] column ozone (DU) Figure 8.6 Erythemal calibration factors for the UVB-1 as a function of total column ozone for three SZAs normalized by their respective values for total column ozone of 300 DU

an approach to derive erythemal UV irradiance on clear days. This method is also recommended by the World Meteorological Organization/Global Atmosphere Watch (Seckmeyer et al., 2007) for erythemal calibration of broadband pyranometers. In the proposed scheme, the dependence of the calibration factor on ozone is curve-fitted with a third order polynomial for different SZAs from 5° to 80° using a step interval of 5 ".Factors between the step intervals can be interpolated using any efficient interpolating technique. In cases where the total column ozone is known to differ significantly from 300 DU, the user may contact UVMRP to obtain erythemally weighted irradiances calibrated at other ozone levels.

200 300 400 500

Tola] column ozone (DU) Figure 8.6 Erythemal calibration factors for the UVB-1 as a function of total column ozone for three SZAs normalized by their respective values for total column ozone of 300 DU

+1 0

Post a comment