Figure 2 shows the spectral irradiance of the TUV and EUV from the UVSPEC simulation in representative cases in January and July. The vertical line in the middle is the spectral border line between UV-A and UV-B. Because of the characteristics of the spectral erythema weighting function, most of the EUV irradiance is estimated in the UV-B region, and the maximum value is estimated within the spectral range of 300-310 nm. However, the EUV irradiance in the UV-A region is also estimated to carry significant values. For example, the irradiance is estimated to be 0.3-0.8 mW m-2 nm-1 at 330 nm, which is about 10% of the irradiance at 315 nm. This significant value is caused by the large intensities of TUV in the UV-A region, which is estimated to be larger than 100 mW m-2 nm-1. However, because the effect of UV-A on EUV is spectrally varied, it is impossible to directly estimate the impact of UV-A on EUV from broadband observation.
The impact of UV-A on the EUV estimation is also shown in the broadband observation. Figure. 3a shows the time series of EUV from the 501 UV-biometer compared to those from the UVSPEC simulation. In selecting the input parameters of the UVSPEC calculations, the daily total ozone amounts measured from a Dobson spectrophotometer (instrument number: 124, WMO/Global Atmosphere Watch (GAW) station No. 252) are used. Thus, the UVSEPC simulations shown in Fig. 3 indicate the standards of UV-index under clear-sky conditions; the AOD at 550 nm of 0.2 and SSA of 0.98 for the background aerosol condition, and cloud-free conditions are assumed. As observational data are contaminated by cloud and hazardous aerosol events, the overall observational data are lower than the simulation. Additionally, the limit of spectral coverage in broadband observation may be a factor influencing the underestimation of the observation. From the comparison result under the condition below daily cloudiness of 2.0, as shown in Fig. 3b, the slope between the UVSPEC simulation and observation from the 501 UV-biometer is 1.085, with a correlation coefficient of 0.849.
Table 1 shows the monthly EUV irradiance and UV-index from the UVSPEC simulation. Also shown is the ratio of the sum of EUV(B) and EUV(A) [EUV(A+B)] to the EUV(B) [EUV(A+B)/EUV(B)] from the simulation results. To define the irradiance of EUV(A) and EUV(B), the integrated spectral range is defined from 320 to 400 nm, and from 280 to 320 nm for EUV(A) and EUV(B), respectively. Compared to the EUV(B) and EUV(A+B), the difference of UV-index between EUV(A+B) and EUV(B) is estimated to be 0.6 to 1.2, which means that the UV-index has been underestimated due to lack of consideration of UV-A, and this effect can be significant to the broadband observation.
The EUV(A+B)/EUV(B) is estimated to be in the range from 1.2 to 1.8, and the ratio is strongly dependent on the representative monthly conditions. The annual variation of the ratio is caused by the SZA, aerosol, and total ozone amounts, which affect the scattering and absorption properties of UV radiation. The SZA causes change in the optical path length, and makes the optical depth change at the slant path. Furthermore, the extinctions for aerosol and total ozone are normally stronger in the UV-B region than in the UV-A region due to the spectrally strong dependence of the extinction -such as Rayleigh scattering and absorption. Therefore, the characteristics of extinction due to aerosol and total ozone also lead to the annual variation of the EUV(A+B)/EUV(B). Therefore, linear conversion from EUV(B) to EUV(A+B) is insufficient to determine the total amount of EUV irradiance.
To identify the EUV(A+B) from the broadband observation, the ratio of EUV(A)/TUV(A) is introduced in this study, defined as follows: \[ EUV(A)/TUV(A)=\intΛ1Λ2W(Λ)I(Λ)dΛ/\intΛ1Λ2I(Λ)dΛ , \] where W(Λ) is the erythema weighting function, as normalized by the weighting at 298 nm (McKinlay and Diffey, 1987), and I(Λ) is the spectral irradiance at wavelength Λ. To calculate EUV(A)/TUV(A), the spectral irradiance data from the simulation is necessary. Table 2 shows the estimated TUV(A), EUV(A) and the EUV(A)/TUV(A) in the monthly representative cases. Because the erythema weighting function in the UV-A region is smaller than 8.0× 10-3, the order of EUV is 1000 times smaller than that of TUV. The value of EUV(A)/TUV(A) ranges from 6.6× 10-4 to 7.0× 10-4, and EUV(A)/TUV(A) is larger in summer than in winter. For the period from April to September when EUV is strong, EUV(A)/TUV(A) is estimated to be 6.8× 10-4 to 7.0× 10-4, and the representative value is estimated to be 6.9× 10-4.
Figure 4 shows the EUV(A)/TUV(A) in clear cases as a function of SZA, total ozone, and the AOD at 550 nm. For the calculation shown in the figure, the following conditions are assumed: SSA = 0.90; day of year = 75; altitude = 84 m (the altitude of the Yonsei University site). Furthermore, reference conditions of SZA = 30°, total ozone = 340 DU, and AOD at 550 nm = 0.4, are assumed. As shown in Fig. 4a, the EUV(A)/TUV(A) is estimated to be 6.1× 10-4 to 7.0× 10-4, and the ratio rapidly decreases due to the large SZA. The extinction of spectral irradiance is relatively high at short wavelengths due to the spectral characteristics of Rayleigh scattering and absorption by ozone. For this reason, the decreased intensity of spectral irradiance is large at short wavelengths as the path length increases. Furthermore, the erythema weighting function has a large portion in short wavelengths, especially those shorter than 328 nm. A relatively large portion of the weighting function and strong extinction at short wavelengths affects the decrease in EUV(A)/TUV(A) as the SZA increases. Therefore, EUV(A)/TUV(A) is relatively sensitive to the change at short wavelengths, and optical path length, which affects the air-mass factor, the secant function of the SZA.
As suggested by the results shown in Fig. 4b, EUV(A)/ TUV(A) is significantly influenced by the total ozone value. Although the strong absorption by ozone is mostly found in the UV-B region, the absorption band of ozone in the UV-A region below 350 nm is also important, referred to as the Huggins band (e.g., Inn and Tanaka, 1953; Griggs, 1968; Le Qu\'er\'e and Leforestier, 1992). Based on (Molina and Molina, 1986), the absorption cross section of ozone is estimated to be 0.377× 10-20 cm2 molecule-1 at 330.0 nm and 0.167× 10-20 cm2 molecule-1 at 340.0 nm, if the temperature is assumed to be 263 K. Therefore, the irradiance in the UV-A region is partially affected by the total ozone amount, which causes the decrease of EUV(A)/TUV(A) as the total ozone amount increases. From the simulation in the reference case, the decreasing rate of EUV(A)/TUV(A) in the total ozone is estimated to be -1.0× 10-5 (100 DU)-1. For example, the maximum and minimum value of total ozone at Seoul was 499 and 225 DU from 1985 to 2009, respectively (Park et al., 2011). Therefore, the variance of EUV(A)/TUV(A) is 2.7× 10-5, which is equal to 3.9% of the relative variance. Figure. 4c shows the change of EUV(A)/TUV(A) due to the AOD change at 550 nm. Similar to the results shown in Fig. 4b, the EUV(A)/TUV(A) decreases with increasing AOD, and the slope of EUV(A)/TUV(A) with respect to AOD is -1.5× 10-5. From observation at Seoul, the Angstr\"om exponent of aerosol is estimated to be in the range from 0.905 to 1.193 (Koo, 2008). For this reason, the spectral change of AOD normally decreases as the wavelength increases, producing the decrease of EUV(A)/TUV(A). (Kim et al., 2007) revealed that the monthly mean of AOD over East Asia to be lower than 2.0. When this value is considered as the maximum value, the relative variance of EUV(A)/TUV(A) due to the AOD is estimated to be about 4.4%.
To calculate the EUV(A+B) from the broadband observation, the EUV(A)/TUV(A) is adopted to convert from the observed TUV(A) to the EUV(A). Figure 5 shows a scatter plot of the EUV irradiance between the Brewer spectrophotometer and the UV-biometers. To adjust the observation time, the time difference between the Brewer spectrophotometer and the UV-biometers is established to be less than 30 seconds for the comparison. Furthermore, the EUV data from the Brewer spectrophotometer that are larger than 10 mW m-2 are used for the validation, to neglect the effect of large-scale clouds (number of data = 560). The EUV(A)/TUV(A) value uses the fixed value of 6.9× 10-4, as described above. From Fig. 5a, the EUV(B) from the UV-biometer 501 is partially underestimated, as compared to the EUV from the Brewer spectrophotometer. Although the correlation coefficient (R2) is 0.820, which is a statistically significant value, the slope is estimated to be 0.951. This statistical result means that the observation of the UV-biometer 501 in the UV-B region only explains 95.1% of the total irradiance of EUV from the Brewer spectrophotometer. To adopt the EUV(A) value by using the TUV data from the UV-biometer 501A in the UV-A region, the slope is calculated to be 1.103 and the R2 is 0.818, as shown in Fig. 5b. Compared to Fig. 5a, the R2 value is almost the same but the slope is significantly improved; the intensity of enhancement is about 0.152. From this result, the EUV in the UV-A region is calculated to take about 15% of the EUV irradiance, which is comparable to previous findings (van Geffen et al., 2004).
To consider the different conditions for aerosol and total ozone amount, a look-up table (LUT) for EUV(A)/TUV(A), which adopts the observed TUV(A) data, is developed. The LUT considers the change of SZA, AOD, and total ozone amount. To cover most of the observed conditions, the EUV(A)/TUV(A) value is estimated in the LUT for cases in which the SZA ranges from 0.0° to 80.0°, with a 10.0° degree interval; the AOD ranges from 0.0 to 3.0, with an interval of 0.2; and the total ozone ranges from 250 to 550 DU, with a 30 DU interval. From the LUT simulation, the EUV(A)/TUV(A) is calculated to be 6.60× 10-4 0.35× 10-4 (1-σ level).
Figure 6 shows a similar result as Fig. 5, but the ratio of EUV(A)/TUV(A) is used from the LUT simulation, as changed by the conditions of AOD and total ozone. The AOD and total ozone values are daily averaged data from the Brewer spectrophotometer. Because the reference wavelength of the AOD measured by the Brewer spectrophotometer is 320.1 nm, the Angstr\"om exponent is assumed to be 1.0, which is the mean value (340-1020 nm) in 2007 at Seoul (Koo, 2008), to estimate the AOD at 550 nm. From the scatter plot, the slope and R2 value are estimated to be 1.107 and 0.819, respectively. Compared to Fig. 5b, the slope and R2 value show insignificant differences. The difference of the statistical result between Fig. 5b and Fig. 6 is related to the variation of EUV(A)/TUV(A) as a consideration of total ozone and aerosol, and the proportion of EUV(A) to total EUV. The proportion of EUV(A) to total EUV is about 15% from the previous result, and the EUV(A) difference between the fixed and LUT method for EUV(A)/TUV(A) is 2.1% 2.3%. Therefore, the effect of AOD and total ozone is insignificant during the comparison period.
From Figs. 5b and 6, the EUV(A+B) from the UV-biometers is partially overestimated compared to that from the Brewer spectrophotometer. The reason for the overestimation is caused by the sensitivity of UV-A for the 501 UV-biometer in the UV-B observation. (Kudish et al., 2005) reported that the normalized spectral response is ∼0.001 at 330 nm, which is about 10% of the sensitivity at 320 nm. Furthermore, (Singh et al., 2005) showed that the UV-B biometer actually measures the spectral band of 280-340 nm, by using the broadband filter characteristics. Although the absolute value of the spectral response is small, this sensitivity significantly affects the EUV calculation because the erythema weighting function is also small at this wavelength. For this reason, the effect of spectral overlap has to be subtracted for accurately estimating the EUV irradiance from the broadband observations. To correct the spectral overlap between 320 and 340 nm, the ratio of the spectrally integrated EUV (280-320 nm) to EUV (280-340 nm) is calculated as a function of SZA, AOD, and total ozone. Thus, an LUT of the ratio with the same dimension, identical to the LUT of EUV(A)/TUV(A), is considered. In addition, another LUT for EUV, spectrally integrated from 363 to 400 nm, is developed to supplement the spectral coverage of the Brewer spectrophotometer. Because of the limited spectral coverage of the Brewer spectrophotometer up to 363 nm, it potentially affects the comparison of spectral EUV integration between the UV-biometers and the Brewer spectrophotometer.
Figure 7 shows a scatter plot between the EUV from the Brewer spectrophotometer and the EUV(A+B) from the two different UV-biometers, after correcting the spectral coverage of the instruments using the two methods. Firstly, Fig. 7a compares the two EUVs; the EUV(B)+EUV(A) from the two UV-biometers minus the overlap of 320-340 nm and EUV(B + partial A) from the Brewer UV scans are examined using each LUT of the EUV(320/340) and EUV(A)/TUV(A) ratios. A close look at the statistics reveals a slope of 1.022 with standard deviation of 0.019 and the R2 of 0.835. Compared to the previous results in Figs. 5b and 6, the slope between two values is decreased and the correlation coefficient is slightly better. Thus, the former overestimation of the EUV(A+B) calculated from the UV-biometers is to some extent corrected by removing the overlap effect of EUV(320/340). Secondly, Fig. 7b compares the Brewer UV scan extended from 363 nm to 400 nm. This result indicates a slope of 0.992 with standard deviation of 0.019, and an R2 similar to that in Fig. 7a. While there are no significant differences for the two comparisons in the slope and R2 values, the slope is more precisely fitted on a 1:1 line (i.e., slope of 1.0) by considering the expansion of the Brewer UV scan.
From the statistical analysis, the broadband observation of the EUV(B) accounts for about 85% of total EUV from the Brewer spectrophotometer, and its underestimation is caused by neglecting the contribution of UV-A. By adding the broadband UV-A data, the total EUV irradiance from broadband observation is increased by about 15%. It can be further suggested that the enhancement of the EUV value by considering UV-A is larger than the inherent uncertainty of the instrument. Although the effect of spectral overlap within 320-340 nm causes overestimation by 10%, the correction factor of the EUV(320/340) can reduce the overlap effect to about 8% of total EUV, and the spectral limitation of the Brewer spectrophotometer explains the remainder of the overestimation. Through several correction methods, the correlation coefficient is increased from 0.818 to 0.836. The calibration of the instrument is one of main effects of the discrepancies between the two instruments' data. Furthermore, the temporal variation of clouds and aerosol types also affects the disagreement.