Presented here are the results of a short but intense measurement campaign at Lauder, New Zealand, in which spectral irradiance from instruments operated by the National Institute of Water and Atmospheric Research (NIWA) and Austria/Innsbruck (ATI) were traced to different irradiance standards and compared. The observed spectral differences for global irradiance were relatively small (<5%) and were consistent with those expected from observed differences in the radiation standards used by each group. Actinic fluxes measured by both groups were also intercompared and found to agree at the 10% level. The ATI instrument had the additional capability of measuring solar direct beam irradiance and sky radiances. These provided the first series of sky radiance measurements at this pristine Network for the Detection of Atmospheric Composition Change (NDACC) site. The polarization of sky radiance results were compared with estimates from a radiative transfer model without any aerosols and was found to be up to 25% smaller. Total ozone values derived from Total Ozone Mapping Spectrometer (TOMS), Dobson measurements by NIWA, spectral direct sun measurements by ATI, and spectral global irradiance measurements by NIWA agreed generally within 2%–3%.
The National Institute of Water and Atmospheric Research’s (NIWA’s) atmospheric research site at Lauder, New Zealand (45.04°S, 169.68°E, 370 m), is important because it is one of very few Southern Hemisphere sites with long-term quality-assured measurements of UV radiation. The site is the Southern Hemisphere charter site of the Network for the Detection of Atmospheric Composition Change [NDACC; formerly known as the Network for the Detection of Stratospheric Change (NDSC)], and as such, it is well equipped with state-of-the-art instruments to monitor the composition and transmission of the atmosphere. Data from the UV spectrometers at Lauder comply with the exacting requirements of the NDACC (McKenzie et al. 1997). Important results from this site include 1) the finding that UV irradiance there is approximately 40% larger than at corresponding latitudes in the Northern Hemisphere (McKenzie et al. 2006; Seckmeyer et al. 1995; Seckmeyer and McKenzie 1992); 2) an observed increase in UV over the 1980s and 1990s in response to ozone loss, and a subsequent leveling off or decline in UV since then (Bais et al. 2007; McKenzie et al. 1999); and 3) accurate agreement with satellite-derived UV, which is more accurate than at more polluted sites (McKenzie et al. 2001). Other measurements at Lauder and elsewhere in New Zealand have shown that despite its very low aerosol extinctions, there is evidence of reduction in global irradiance prior to the mid-1980s, henceforth followed by an increase [i.e., a period of “dimming,” followed by a period of “brightening”; Liley (2006)]. In view of these important findings, it is desirable to regularly evaluate the performance of the instrumentation at this site against other independent measurements. Although NIWA instruments have regularly been involved in intercomparisons and found to perform well (Bais et al. 2001; Wuttke et al. 2006), the last such campaign at the Lauder site was more than 10 yr ago, in the summer of 1993 (McKenzie et al. 1993). The purpose of this study is twofold. The first aim is to cross-correlate spectral UV measurements of the NIWA instruments with those from an independent instrument with calibrations traceable to a different standards laboratory. The second aim is to measure the distribution of sky radiance to verify model calculations under pristine conditions at a relatively low altitude and to better quantify the magnitude of possible errors in cosine corrections that arise from the simplistic assumption that the sky radiance is homogenous in the UV region. The measurements described below were undertaken between 22 February and 3 March 2004 when the spectrometer system belonging to ATI (this abbreviation was used for the instrument of Austria/Innsbruck in several previous campaigns) operated alongside NIWA’s UV spectrometers at Lauder. Unfortunately, the weather was rather unsettled through most of the period. However, sufficient data were collected to achieve the aims of the study.
2. Materials and methods
a. ATI spectral measurements
The spectral measurements of ATI were carried out with a commercial Bentham DTM300 double monochromator spectroradiometer, which was modified by custom-made extensions in several ways to allow the measurement of global irradiance, actinic flux, direct irradiance, sky radiance, and polarization of sky radiance. The instrument is equipped with two sets of holographic gratings, 1200 and 2400 lines per millimeter, which can be changed automatically under software control. The entrance and exit slits were set to a width of 0.42 mm, resulting in an almost triangular optical slit function with 0.48-nm full width at half maximum (FWHM), when used in combination with the grating with 2400 lines per millimeter. This was the case for measurements of direct sun, sky radiance, and sky polarization. For measurements of global irradiance and actinic flux, the grating with 1200 lines per millimeter was used, resulting in 0.96-nm FWHM. The monochromator’s wavelength setting was regularly checked against the Fraunhofer structure of the solar spectrum (Huber et al. 1993); therefore, the wavelength uncertainty of the spectral measurements can be quantified to be less than 0.03 nm.
At the exit slit of the spectroradiometer, a side-window photomultiplier is mounted, which is operated at about 600 V. At each wavelength setting, the measurement time (average of 1-MHz 14-bit samples) is automatically adjusted between 2 s at very low signal levels down to 20 ms at high signals. A five-stage decadal amplifier allows a large dynamic range. The whole spectroradiometer system is fitted in a box, which is temperature controlled to about 0.1°C.
The spectroradiometer has two entrance ports (with identical slits), which can be selected with a “swing-away mirror.” At each port, a quartz fiber is attached with a length of about 5 m. One fiber is Y shaped, in which one end is connected to the global input optics, and the other to the one for actinic flux. Both input optics are equipped with a shutter that opens and closes the entrance of the fiber. With this setup, either radiation quantity can be measured within a few seconds at each wavelength setting, thus minimizing the effect of changing clouds on the ratio between the two radiation fluxes. Unfortunately, during the campaign in New Zealand, one of the magnetic shutters stopped working properly after a few days, so that only a few synchronous measurements of global irradiance and actinic flux could be carried out. The angular response of the global entrance optics is optimized by a special shape, so that the resulting cosine error is less than about 2% for zenith angles smaller than 75° (Schreder et al. 1998). In addition, the entrance optics are heated, so that the temperature of the Teflon diffuser is always above about 20°, which was found to be a critical temperature for changes in transmission of Teflon (Ylianttila and Schreder 2005). The entrance optics for the actinic flux have a 2-pi angular response (produced by Metcon, Inc.; R. Schmitt 2004, personal communication; information online at http://www.metcon-us.com), which was tested in the laboratory to be close to ideal. Any deviations were estimated to give <3% uncertainty for the 2-pi actinic flux. Only the downwelling actinic flux was measured.
The second fiber (connected to the second entrance port) leads to a small telescope, which is mounted on a custom-made sun-tracking device. This allows positioning of the optics at the sun or at any point on the sky. The telescope, which has a 25-mm quartz lens of 100-mm focal length and with appropriate baffles, was used to limit the field of view to a well-defined aperture of about 1.5° FWHM. For direct sun measurements, a 1-mm-thick piece of Teflon is positioned in the focal plane to reduce the intensity. In front of the lens, a linear sheet polarizer can be mounted manually, which can be rotated by a stepping motor under control of the custom-made measurement software. The polarizer was especially selected to allow a good transmission in the UV-wavelength range as well.
From the high-resolution spectral measurements of direct solar irradiance, the actual values for total atmospheric ozone and spectral aerosol optical depth were determined (Huber et al. 1995).
Radiance measurements were carried out for 60 directions in the solar principal plane (defined as the vertical plane through the sun and through the zenith) and in the almucantar (defined as the horizontal plane at the elevation of the sun, scanned in the direction of the sun’s movement). At each position on the sky, four measurements were taken, with orientations of the polarizer at 0°, 45°, 90° and 135°, relative to the vertical, which needed about 15 min. The scans were made for the wavelengths 310, 350, and 450 nm, consecutively.
b. NIWA spectral measurements
The spectral measurements of NIWA were carried out with two modified double monochromators: one configured to measure irradiance, and the other configured to measure actinic flux. These instruments are part of the NDACC suite of measurements at Lauder. In normal operations, scans are taken at 5° steps in solar zenith angle (SZA) and at 15-min intervals for a 2-h period centered at local noon. For this campaign, scans were undertaken at 10-min intervals to match the frequency of scans of the ATI instruments. However, the scan rates were not adjusted to match the ATI and operated in their normal modes, as described below. This has the advantage of the system performance being assessed against independent instruments under normal operating conditions. The disadvantage is that it complicates the comparison under cloudy conditions when radiation fields are rapidly changing. Such conditions prevailed during most of the campaign, so it was necessary to interpolate data using the high-temporal-resolution broadband data, as described below.
The entrance optic for the irradiance measurements is a shaped polytetrafluoroethylene (PTFE) diffuser that was designed in-house at NIWA. The cosine error is less than about 3% for SZA < 85°. In addition, its temperature is monitored and corrections are applied to account for the recently documented changes in transmission as a function of temperature for these diffusers (McKenzie et al. 2005). The diffuser is coupled via a 2-m fiber optic bundle to an Acton Spectra Pro 275 double monochromator (NIWA designation UV4), equipped with 3600 g mm−1 holographic gratings. The entrance and exit slits are 1 mm, giving a near-triangular slit function of 0.6-nm FWHM. In this system, each spectrum consists of the average of a reverse scan plus a forward scan over the wavelength range 285–450 nm. This order of a downward scan followed immediately by an upward scan minimizes any small effects of nonlinear changes of the solar spectrum during the scanning time, especially at the shortest wavelengths for larger SZA. The scan speed is adjusted so that more time is spent at shorter wavelengths where signal strengths are smaller. The total scan time is approximately 5 min, and the time stamp that is logged with the data corresponds to the midpoint turnaround at 285 nm. Dark currents measured before and after each scan are subtracted. The signal is measured by a side-windowed Hamamatsu R1527 photomultiplier (bialkali photocathode) with a voltage setting of approximately 555 V, and data are logged with a 24-bit A/D converter. Apart from the monochromator itself, this system is the same as NIWA’s current Bentham DTM300 UV instruments (http://www.niwa.co.nz/rc/fac/instruments/lauder/uvspec).
The entrance optic for the actinic flux measurements was supplied by R. Schmitt (Meteor, Inc., 2000, personal communication). Its 2-pi angular response was tested and found to be within 5% of the ideal response for all SZA < 75°, and these deviations were estimated to give <5% uncertainty for the 2-pi actinic flux. Only the downwelling actinic flux was measured. The diffuser is coupled via a 2-m fiber optic bundle to a Bentham DM300 (NIWA designation UVM), equipped with 1200 g mm−1 holographic gratings and with entrance and exit slits of 1 mm, to give a near-triangular slit function of 0.90-nm FWHM. This system logs the UV data at a constant rate over the spectral range of 290–450 nm in steps of 0.2 nm. Each scan takes approximately 3 min to complete, and the time stamp for the data is for the midpoint wavelength of 370 nm. Dark currents measured before and after each scan are subtracted. In this older system, the signal is measured by a front-windowed EMI 9804 photomultiplier (bialkali photocathode) and logged with a 12-bit A/D converter. The high dynamic range necessary is achieved by adjusting the voltage applied to the photomultiplier tube (PMT), and therefore the gain, as discussed previously (McKenzie et al. 1992).
The software to control the instrument and to analyze the results was developed in-house. The photomultipliers are used in analog current mode, with a filter time constant of approximately 27 ms. The wavelengths are sampled in steps of 0.2 nm, which is significantly less than the 0.6-nm FWHM. A consequence of this oversampling is that very accurate wavelength alignment can be achieved by correlation with the Fraunhofer structure of the solar spectrum. By these means, small nonlinearities in the wavelength drive are also corrected. The resulting wavelength uncertainty of the spectral measurements is less than 0.03 nm. Both systems are temperature controlled and stabilized to within ±0.5°C, and they include diode detectors that are sensitive to UVA wavelengths at the entrance slit, to monitor changes in intensity during the scans.
The analysis includes corrections for stray light, dark current, nonlinearities in the wavelength drive, and departures from the ideal angular response. Outputs, which are traceable to the National Institute of Standards and Technology (NIST), include spectral irradiance or actinic flux, biologically weighted irradiance (such as erythemally weighted irradiance), or chemical photolysis rates [such as J(NO2) and J(O3)]. Additionally, ozone retrievals can be made by using the algorithm described by Stamnes et al. (1991). Software has also been developed to estimate irradiance from the actinic flux measurements and actinic flux from the irradiance measurements. These algorithms use model calculations to apportion diffuse and direct radiation. The following two assumptions are made in these algorithms: first, when the UVA transmission is less than 0.5, it is assumed that all radiation is diffuse; second, the diffuse radiation field is assumed to be isotropic.
c. Broadband data
Broadband UV irradiance and photolysis rate data were available at 1-min intervals from NIWA and ATI sensors. Here, we used data from the ATI instruments to minimize cloud effects. These comprised pyranometers for total solar radiation (300–3000 nm), for UVA radiation (315–400 nm), and for erythemally weighted UV irradiance. The data from these detectors were used for quality control of the measurement of global spectral irradiance. The corresponding spectral integrals agreed with the broadband data during the whole campaign within a few percent, without any significant long-term drift. These broadband data were also used for interpolating the spectral measurements between the two instruments of NIWA and ATI. Because during each scan of the ATI instrument the broadband detectors were sampled every second, the variability due to changing conditions of cloudiness could be quantified and applied to the interpolated data.
a. Comparison between lamps
Absolute radiometric calibration of the two spectroradiometers is achieved by reference to 1000-W FEL lamps. In the past there have been large divergences in the irradiance scales maintained in different countries. These differences sometimes exceed 4% in the UV region (Walker et al. 1991) and can be the limiting factor for comparison between different groups. In this study, the measurements by the NIWA group are traceable to the irradiance scale maintained by NIST in the United States (lamp F726 and F666), whereas the measurements from the ATI group are traceable to the irradiance scale maintained by Physikalisch-Technische Bundesananstalt (PTB) in Germany (lamp F1353). The irradiance scales were assessed by measuring the irradiance from all three lamps with the NIWA spectrometer and calculating the resulting instrument calibration factors using the lamp irradiance provided by the respective calibration laboratories. The results comparing the measured irradiance ratios between lamps F1353/F666 and F726/F666 measured with the NIWA spectrometer and the calibration factors for the UV4 instrument, derived from all three lamps, are shown in Fig. 1. The ratios of these k factors vary between 0.99 and 1.01 with some spectral structure. The largest discrepancies (∼1%) are due to slightly larger emissions from aluminum imperfections in the ATI lamp near 308 and 390 nm. The differences between the lamps from the two calibration laboratories are on the same order as the differences from two lamps from the same calibration laboratory. Differences between the NIWA lamp (F666, to NIST) and the ATI lamp (F1353, to PTB) calibration could explain only ∼1% of the difference in erythemally weighted UV. The smaller calibration factor from the ATI lamp would result in slightly larger irradiances.
From this we conclude that the laboratory calibration of the standard lamps agree within the stated uncertainty limits of the calibration laboratories. These small differences are consistent with the expected variability when using FEL lamps as transfer standards (Harrison et al. 2000) and with differences between the NIST and PTB scales that have been reported recently (Woolliams et al. 2006).
b. Comparison of spectral irradiance
NIWA data were increased by up to 2% to correct for the temperature dependence of the Teflon diffuser, taking into account that calibrations are performed at a different temperature than solar measurements. No temperature correction was necessary for the ATI diffuser because it is heated to about 20°C. The ATI analysis also includes a correction for the shape of the diffuser, which affects the measurement of the distance between the lamp and the diffuser during calibrations (Gröbner and Blumthaler 2007). This resulted in a 1.5% decrease of data. Although it is likely that a similar correction is necessary for the NIWA data, it has not yet been applied because of difficulties in determining the position of the entrance aperture with sufficient precision, but this work is in progress.
As noted earlier, spectral scans were not synchronized because we wanted to test the performance under normal operating conditions. However, because of the unsettled weather conditions, there were few periods when the conditions were sufficiently stable to allow a meaningful spectral comparison between the instruments. Figure 2 shows one example of irradiance and their ratios with respect to the ATI measurements. In this example, the spectra were taken on day 56 (day of the year) near noon [1215 New Zealand standard time (NZST); SZA = 36.6°]. The plot includes ratios of irradiance calculated with the Tropospheric Ultraviolet–Visible (TUV) radiative transfer model (Madronich and Flocke 1998) for the input parameters as discussed below. It also includes the ratios of irradiance deduced from the measurements of actinic flux with the UVM spectrometer system described above. Much of the scatter is due to mismatches between the spectral resolution of the spectrometers and the model. The scatter is slightly larger for the UV4 instrument because its slit function is the narrowest (0.6-nm FWHM). The mean ratios with respect to ATI are 1.006 ± 0.200, 0.993 ± 0.124, and 1.027 ± 0.092 for the TUV model, the UV4 spectrometer, and the UVM actinic flux spectrometer, respectively. The similar agreement between the UV4 and UVM instruments shows that for these conditions, at least (11% cloud cover mainly near the horizons, with the sun not obscured by clouds), the conversion from spectral actinic flux to spectral irradiance is accurate.
The retrieved ozone amount was 287, 286, and 287 DU from the UV4, UVM, and ATI spectrometers, respectively. The corresponding erythemally weighted irradiance was 19.5, 19.2, and 19.3 μW cm−2.
c. Comparison of erythemally weighted irradiance
The results of the comparison of erythemally weighted irradiance are shown in Fig. 3. Here, we express the irradiance in terms of the UV index (UVI = 0.4 UVEry measured in μW cm−2). The criterion for acceptance here was that the variance of the pyranometer signal was less than 5% of the mean for measurements over the period of interpolation. For the 62 data points that satisfied our acceptance criterion, the mean ratio between the two instruments was 1.00, with a standard deviation of 0.027. For the calculation of these ratios, a specific interpolation was applied, because data were not measured synchronously and weather conditions were generally variable with rapidly changing cloud conditions: the interpolation is based on the temporal changes, as measured with broadband detectors (total irradiance and erythemally weighted UV irradiance). The mean and standard deviations of these data were available over 1-min intervals. If the variation was too high during the measurement times of both spectroradiometers, this data point was not used for comparison; otherwise, the relative variability of the erythemal detector was applied to the spectral integrals. Thus, the scatter of the interpolated points was reduced significantly, while a significant number of data points were still included. There is no systematic deviation within the dataset. The ratios are independent of both the SZA and the absolute value of the signal. Only the random scatter increases somewhat for very small signals (below 1 UVI), because the weather conditions were the worst when the UVI was low, resulting in increased interpolation uncertainty.
d. Comparisons with model calculations and satellite-derived values
Retrievals of total column ozone amounts are available from both instruments. A summary of these as well as corresponding peak-UVI retrievals are given in Table 1. Results derived from satellite are also provided in Table 1.
Generally, the ozone retrievals from the spectrometers in this study showed good agreement with the independent measurements by the NIWA Dobson instrument (Bodeker et al. 2001) and from the Total Ozone Mapping Spectrometer (TOMS). On day 62, the value retrieved from the NIWA spectrometer was 6% less than that from the Dobson or overpass data, but otherwise the agreement was well within the expected 2%–3% absolute uncertainties of the reference Dobson instruments (Basher 1982).
The peak UVI is taken from the TOMS UV data product and from the NIWA UV4 spectrometer.
The only cloudless day of the campaign was day 57. The peak-UVI values on this day, calculated with the TUV radiative transfer model (Madronich and Flocke 1998), are shown in Table 2, for various estimates of ozone and aerosol conditions.
The aerosol conditions [i.e., the Angstrom coefficients (α, β) and the single scattering albedo ωss] for each model run are defined as follows:
UVI(1): α = 1.3, β = 0.015, ωss = 0.9 (pristine);
UVI(2): α = 1.3, β = 0.05, ωss = 0.98 (clean);
UVI(3): α = 1.3, β = 0.05, ωss = 0.90 (near clean); and
UVI(4): α = 1.3, β = 0.1, ωss = 0.9 (polluted).
The calculated UVI value that is most consistent with the known input parameters (ozone = 269 DU, α = 1.2, β = 0.02; see discussion below) is UVI = 8.5. This result is consistent with the satellite-derived values (UVI = 8.8). The measured values from the two NIWA instruments was UVI = 8.0. This is about 5% less than the model-calculated value, but would be in better agreement with the model if the ozone had been at the upper limit of its uncertainty range or if the single scattering albedo of the aerosols were smaller (i.e., aerosols more absorbing) than the assumed value of 0.98. Unfortunately, the ATI instrument was operating in a different mode, making sky radiance measurements (see below) during this period. However, the close agreement between the NIWA and ATI instruments at other times gives confidence that it would also be in agreement with these model results and satellite retrievals.
4. Comparison of measured actinic flux
The spectral actinic flux measured by each system was used to calculate the photolysis rates of ozone and nitrogen dioxide according to the following reactions:
Here, the photolysis rate (J) is defined as
where F(λ) is the actinic flux and σ(λ) and ϕ(λ) are the absorption cross sections and quantum yields for photodissociation, respectively (Molina and Molina 1986; Harder et al. 1997; DeMore et al. 1997).
In each case, the photolysis rates for ozone J[O(1D)] and for nitrogen dioxide J(NO2) were calculated using the same cross sections and quantum yields. These two integrals succinctly summarize the performance of the spectrometers, because J[O(1D)] is dominated by UV-B radiation, whereas J(NO2) is dominated by UV-A radiation. The results are shown in Fig. 4 (open circles). For both quantities, the photolysis rates measured by the ATI system were higher than those measured by the NIWA system. For J[O(1D)], the ratio ATI/NIWA is about 1.11 ± 0.04; and in the case of J(NO2), the ratio ATI/NIWA is about 1.08 ± 0.05. Here, the differences are significantly higher than for the irradiances discussed previously, but the results are still promising because these measurements are relatively new and fewer intercomparisons of these quantities are available. Possible reasons for the differences include differences in angular response, which have not been accounted for, and uncertainties in the precise location of the entrance aperture.
5. Comparison of measured and calculated actinic flux
Photolysis rates were calculated from spectral global irradiance measurements of ATI using an algorithm developed by Schallhart et al. (2004). The calculated rates were compared with photolysis rates directly measured by ATI: ratios of calculated and measured photolysis rates are 0.94 ± 0.04 for J[O(1D)] and 0.95 ± 0.05 for J(NO2). This level of agreement is somewhat worse than at other measurement sites (Schallhart et al. 2004). However, it may be affected by the relative small number of data points.
Ratios of calculated photolysis rates by ATI and measured photolysis rates by NIWA are 1.03 ± 0.05 for J[O(1D)] and 1.01 ± 0.05 for J(NO2). The ratios of photolysis rates measured by each instrument (ATI/NIWA; Fig. 4, circles) are closer to unity than the ratios for measured/model comparison (Fig. 4, crosses). This could possibly indicate that the ATI measurements are slightly high. But, on the other hand, a comparison of the calibration of the actinic entrance optics of ATI in the laboratory in Innsbruck before and after the campaign in New Zealand showed agreement on the 2% level. Thus, it is not clear how this difference for the ratio “calculated by ATI/measured by ATI” on the order of about 5% can be explained.
The corresponding measured/calculated ratios for the photolysis rates from the NIWA algorithm generally give values close to one (Fig. 5), but the scatter is higher (10%–20%) and there are groups of data with a ratio of about 1.3. These groupings are probably due to the bimodal way in which clouds are treated in this algorithm. If the calculated transmission is less than a certain threshold (τ < 0.5), the algorithm assumes that the sun is obscured and that the radiation field is completely diffuse and isotropic. If the transmission is above the threshold, the diffuse and direct components are treated separately, with their ratio being estimated using a radiative transfer model.
6. Direct sun measurements
With the ATI spectroradiometer, spectral scans of the direct solar irradiance were carried out during time periods when the sun was not obscured by clouds. During the whole campaign of 8 full-measurement days this was possible only on 3 days (days 56, 57, and 61). Then, from the direct sun spectra, the total ozone amount and the aerosol optical depth as a function of wavelength were calculated. Total ozone varied between 287 and 271 DU, which are about 5 DU higher than those measured with the Dobson instrument operated at the measurement site. This corresponds to the known uncertainty of the method of derivation of ozone from absolute direct sun spectra and is a consequence of unknown details as temperature profile, vertical distribution, and absorption cross sections.
The derived aerosol optical depth and its wavelength dependence on 26 February 2004 (day 57) are shown in Fig. 6. On this day, the wavelength dependence of the aerosol extinction (α) was 1.42. The optical depth at 350 nm was 0.054, so the deduced optical depth at 1 μm (β) is about 0.01. Throughout the campaign the aerosol optical depth at 350 nm varied between 0.04 and 0.09 and the Angstrom exponent derived for the wavelength range 320 to 450 nm varied between 1.0 and 1.5. The corresponding optical depths at 1 μm (β) therefore ranged from less than 0.01 to about 0.02. This means that during the campaign the amount of aerosols was as low, as is typical for this site (Liley 2006). Independent measurements with narrowband sun photometers, which are permanently located at the site as part of the Baseline Surface Radiation Network (BSRN), were also available and were provided by B. Forgan (Australian Bureau of Meteorology, 2004, personal communication). Results from these instruments were in good agreement with the spectral measurements (i.e., 0.04 ± 0.004 at 412 nm with narrowband filter instrument, corresponding to 0.044 with spectroradiometer). The uncertainty (1σ) of aerosol optical depth (AOD) derived from direct sun spectra is in the order of ±0.01, the uncertainty of the Angstrom exponent is about ±0.1.
7. Sky radiance measurements
During the campaign the sky was completely cloudless only for 1 day (day 57). On this day, spectral measurements of the diffuse radiance distribution were carried out with a telescope with a field of view of about 1.5°. The sky was scanned in two planes: a vertical one through the sun and the zenith and a horizontal one through the sun. Measurements were taken for the wavelengths 450, 350, and 310 nm. In total, five such scans could be carried out, each taking about 15 min.
The results of the sky radiance measurements are summarized in Fig. 7a, in which the sky radiance is normalized to the intensity of the direct sun at the different wavelengths. As the wavelength decreases from 450 to 310 nm, the diffuse fraction increases markedly. At 450 nm, the skylight is less isotropic (changing by a factor of about 3.5 between 50° and 120°), and there is significant horizon brightening. By contrast, at the UV wavelengths, the skylight is relatively isotropic (changing by a factor of 1.7 between 50° and 120° at 310 nm) with a suggestion of lower radiances near the horizon.
Figure 7b compares these sky radiances at Lauder with corresponding radiances measured at Ispra (45.814°N, 8.627°E, alt 214 m), which is a relatively polluted site in Italy. The aerosol optical depth during the sequence at Ispra was 0.58 at 320 nm. The solar zenith angle was about 50° (53°) for the Lauder (Ispra) data. Note also that while the two shorter wavelengths were the same at each site, the longest wavelength scanned was 450 nm at Lauder, but 400 nm at Ispra. For this comparison, the measured data of the sky radiance in the principal plane were normalized to the zenith radiance (elevation 90°). The general shapes of the curves were similar at the two sites. However, the circumsolar contribution was larger at Ispra, reflecting its greater turbidity, evidenced by the smaller rate of increase in radiance when moving toward the direction of the sun. The horizon brightening was more pronounced at Lauder, as a consequence of less attenuation by aerosols.
After each measurement of sky radiance, the polarization of diffuse sky radiance was also measured. For this purpose, a polarization filter was put in front of the sky radiance optics and at each point on the sky this polarization filter was rotated 4 times by 45°. From these measurements, the degree and the direction of polarization in the two planes on the sky for the three wavelengths 450, 350, and 310 nm could be derived. These measurements are compared with model calculations with the libRadtran model (Mayer and Kylling 2005), assuming an aerosol-free atmosphere. In Fig. 8, qualitatively, the general features are well reproduced by the model. The agreement between modeled and measured polarizations is satisfactory for elevation angles close to the solar elevation angle (SZA about 45°). However, there appears to be a small step in the immediate vicinity on either side of the sun (around elevation angle 50°), probably caused by the small difference in time between the measurements. The agreement becomes poorer when the elevation angle increases toward the zenith, and especially when the angular distance from the sun approaches 90°. For these angles, the measured polarization is only ∼80% of that calculated. The difference cannot be explained by the small amount of aerosols present. At the moment, no explanation for this difference between measurement and modeling can be given. A new version of the model for calculating polarization effects including aerosols is in preparation, and this might possibly reduce the differences.
During a campaign at Lauder, New Zealand, spectral solar radiation fluxes were compared over 7 days between two completely independently calibrated spectroradiometer systems. Individual measurements of global irradiance showed differences of less than 5%, whereas the spectrally averaged ratio was almost exactly matching. Also, the average ratio (62 data points) of erythemally weighted irradiance was almost perfectly at 1.0 with a standard deviation of 2.7%. This very good result is also in agreement with the comparison of the two calibration lamps, which agreed on the 1% level. For actinic flux, the agreement between the two systems was not as good, with deviations of about 10% for the photolysis rate for ozone and about 7% for nitrogen dioxide. This might indicate that the absolute calibration and the angular response of this type of detector are not yet fully characterized. The derivation of actinic flux from global irradiance showed for different methods agreement better than 10%, which is expected from the uncertainties of the individual methods. The quality of the calibration of spectral direct sun measurements was proven indirectly by comparing the derived aerosol optical depth with measurements from standard instruments used within the Baseline Surface Radiation Network. When the aerosol optical depth at 350 nm was about 0.054, which is typical for the very clean air at the measurement site at Lauder, the difference between the results of the two methods was less than 0.01. This is a very satisfying result, justifying the usage of the derived values of aerosol optical depth in radiative transfer calculations to compare the measurements with results of the model calculations.
Measurements of sky radiance under situations with high and low amounts of aerosols show at the polluted site less pronounced brightening of the horizon and relative higher levels near the sun (up to about 40° from the position of the sun). The additional measurements of polarization of sky radiance provide a unique opportunity to verify model calculations under pristine conditions at a relatively low altitude, where polarization is dominated by Rayleigh scattering and only marginally affected by aerosols.
Corresponding author address: Mario Blumthaler, Division for Biomedical Physics, Innsbruck Medical University, Muellerstr. 44 A-6020, Innsbruck, Austria. Email: firstname.lastname@example.org