In this work the influences of ozone, aerosols, and albedo on the clear sky UVA and UVB irradiance at a high-mountain station are investigated by using both routine spectral UV measurements from the high-mountain Sonnblick observatory in Austria (3106-m altitude) and theoretical simulations. The 501 single measurements and the model output show that calculations are on average higher than the measurements by 10%–12% at 305 nm, 8%–10% at 315 nm, and 5%–8% at 370 nm. The study of the fluctuations in UV irradiance constitutes the second part of this work. Columnar-ozone fluctuations lead to enhancements of UV irradiances of 560% at 305 nm and 69% at 315 nm as columnar ozone changes from 380 to 230 Dobson units. The radiative transfer model shows the same behavior with changes in columnar ozone and solar zenith angle as the measurements do. By using the measurement:model ratio, it therefore is possible to analyze the fluctuations in UV irradiance occurring at Sonnblick observatory that are not dependent on solar zenith angle or columnar ozone in order to trace the influence of albedo on UV irradiance. First, the maximum possible change in UV irradiance due to aerosols is simulated and shows that changes in aerosol optical depth could induce changes in UV irradiance of up to 5%. Second, the influence of various parameters on UV irradiance is examined. Cloud cover under the Sonnblick summit may enhance the UV irradiance on average by 6% in wintertime. The UV values in winter are on average 4.5% higher than the summer values for a fixed solar zenith angle. This result most probably is caused by changes in ground albedo resulting from larger areas being covered with snow during the winter. Simulations with the radiative transfer model suggest that the average albedo may be larger by 0.15–0.22 during wintertime as compared with summertime. Clouds under the summit may enhance the average albedo by 0.30 ± 0.15.
The stratospheric ozone layer is well known as a protective layer against harmful ultraviolet (UV) irradiance. Absorption of UV irradiance is an important characteristic of ozone. One therefore may expect an increase in the harmful UV irradiance at the ground because of the observed decrease in stratospheric ozone concentrations (WMO 1995). There have been many investigations of the UV–ozone relationship. Studies of long-term Robertson–Berger routine measurements (Scotto et al. 1988) at eight sites in the United States showed an unexpected decrease in biologically effective UV. Cloudiness and turbidity were found to have increased over the same period by appreciable amounts. Investigations by Weatherhead et al. (1997), however, indicate that serious doubts in the validity of the dataset used by Scotto et al. are justified. Frederick and Weatherhead (1992) found no long-term trend in ozone from the Dobson spectrophotometers at Bismarck, North Dakota, and Tallahassee, Florida, but found a decreasing trend in Robertson–Berger irradiance during the winter months when the Robertson–Berger signals were low. No explanation for this phenomena was found. An average 1% per year increase in Robertson–Berger measurements at the Jungfraujoch observatory (3576-m altitude in the Swiss Alps) was found over the years 1981–89 by Blumthaler and Ambach (1990). This increase, which is more in agreement with the ozone trend, may be explained by the lack of pollution at this altitude. Several studies have been undertaken using the routine spectral UV measurements sponsored by the National Science Foundation at Antarctica; southern Argentina; San Diego, California; and Point Barrow, Alaska. Frederick and Alberts (1991) found an increase in surface UVB by a factor of 2 in the austral spring of 1990 for periods of several days. Stamnes et al. (1992) found that UV irradiance at 305 nm was enhanced by a factor of 3–6 in October 1990 compared with the previous October, and found that summerlike UV values occurred 2 months early that year. Several other studies report increases in UVB because of very low stratospheric ozone values (e.g., Lubin et al. 1989; Kerr and McElroy 1993).
Despite many investigations of the UVB–ozone relationship, there are few UV measuring sites around the globe, so UV irradiance models are used to give wide-cover estimates of ground-level UV irradiances. Some radiative transfer models are stated to be absolutely accurate mathematically, with the accuracy of the model output being dependent only on the computer power, the number of streams used in the calculations, and the accuracy of the input data (Forster et al. 1993).
The main source of inaccuracy is therefore not the mathematical methods used to solve the radiative transfer equations but rather is the inaccuracy in the input data, which requires information about all variables that affect UV: ozone, single-scattering albedo (SSA), aerosol, and ground albedo. The resulting uncertainties in model output are detailed by Schwander et al. (1997) and Weihs and Webb (1997a). Large uncertainties result from the aerosol model input because of unknown characteristics of the aerosols, such as the single-scattering albedo and the phase function, and because of uncertainties in the determination of the aerosol optical depth. The uncertainty in model calculations due to aerosols is, however, larger for lower areas, with their very large aerosol concentrations, than it is for higher areas, where aerosol levels are very low.
In lower areas generally a good agreement between measurements and calculations is reached; Forster et al. (1995) compared measurements from Reading, United Kingdom (300-m altitude), with discrete ordinate method model calculations and found an agreement within 10%. Wang and Lenoble (1994) described a comparison between their model, which was based on the discrete ordinate method, and the measurement data from an international spectrometer comparison in Panorama, Greece (385-m altitude), and found agreement within ±6% when the ratio of the measurements to the calculations is averaged over intervals of 10 nm. Zeng et al. (1994) compared calculations performed with a discrete ordinate method model with measurements made at Lauder, New Zealand. They found reasonable agreement between measured and calculated diffuse, direct, and global irradiances over the range 300–450 nm. The diffuse to direct ratio agreement between measurements and model calculations was found to be better than 8%. However, many parameters were assumed, and, as already shown by Forster et al. (1995), Zeng et al. (1994), Weihs and Webb (1997a,b), and Schwander et al. (1997), changing the input parameters in a reasonable way may produce large deviations in the model calculations, so there is a large probability of reaching agreement between calculations and measurements by using adequate input parameters. Mayer et al. (1997) and Weihs et al. (1997) compared measurements and calculations for a longer time period for clear-sky days and for many data points instead of comparing only several single scans. Mayer et al. performed their model–measurement comparisons at Garmisch-Partenkirchen, Germany (700-m altitude), for ground conditions without snow and using aerosol optical depth measurements performed at the site, and they found an agreement between measurements and calculations of −11% to 2%. Weihs et al. (1997) performed their comparisons at Reading, United Kingdom (300-m altitude), determined the aerosol optical depth by visibility measurements, and found an average absolute relative deviation of 12% at 305 nm and 10% at 380 nm between measurements and calculations.
In higher areas there is, as already mentioned, less influence of aerosols on UV irradiance, but there may be large changes in UV irradiance caused by changes in ground albedo. Blumthaler et al. (1993) reported a comparison between spectral UV radiation measurements performed at Jungfraujoch observatory (3576-m altitude) and calculations performed with the Green model (Green et al. 1974; Schippnick and Green 1982) and found a good correspondence, though no statements as to accuracy were made. Weihs and Webb (1997b) compared spectral UV measurements with discrete ordinate model calculations and found the model calculations to be about 10% higher than the measurements, also made at Jungfraujoch observatory. There have been few comparisons in high-mountain areas. There still is little understanding about the influence of ground albedo and topography on ground UV irradiance in these regions. These regions also offer a better opportunity to trace the influence of ground albedo on UV irradiance and to perform an accuracy study of UV models because it is less possible to reach agreement between the model and measurements by simply changing the aerosol model input parameters for an atmosphere with very low aerosol levels.
In this work, we try to quantify the influence of ozone, aerosols, and albedo on the UV irradiance at a high-mountain station. We used theoretical simulations from the discrete ordinate radiative transfer (DISORT) model of Stamnes et al. (1988) and routine spectral UV measurements from the Sonnblick observatory in Austria (3106-m altitude) to perform comparisons under clear sky conditions and to try to trace the influence of the ground albedo and stratospheric ozone on ground UV irradiance.
The measurements site
The Institut für Meteorologie und Physik (meteorological department) of the Universität für Bodenkultur (IMP_BOKU), Vienna, Austria, has been recording columnar ozone thickness at Sonnblick observatory since August 1993 and spectral UV irradiance since July 1994. The Sonnblick observatory (47.05°N, 12.95°E) is in southwest Austria at the border between Salzburg and Carinthia on a mountain summit at 3106-m altitude. It is surrounded by rock faces on the northern side and by a glacier on the southeastern side. The Sonnblick region and the Alps in general show a very pronounced topography. The valley adjacent to Sonnblick is 1300 m lower than the Sonnblick summit. Nearby summits in this region are at approximately the same altitude as Sonnblick and are surrounded by permanent glacier snow. The UV and ozone measurements were performed on a platform on the southern side of the observatory. The obstruction of the horizon was typically less than 2°. Only a metallic mast (15 cm thick, 7 m away) north of the instrument has a 20° elevation. The Sonnblick observatory is also a station of the Austrian Weather Office, which monitors the synoptic and climatological meteorological measurements of air pressure, wind speed and direction, temperature, humidity, and other weather events. The Austrian Weather Service also records the cloud and ground surface conditions. The cloud fraction above and below the observatory as well as the cloud type are estimated visually six times a day by the weather observer. The snowfall and the depth of the snow cover are measured once a day. The altitude of the clouds is estimated and assigned to the three usual categories of low cloud, midcloud, and high cloud. These data were made available for this study.
Instrumentation at Sonnblick
Columnar ozone was measured with a sun-tracking Brewer MKIV (No. 093) spectrometer. The spectral irradiance between 280 and 500 nm was measured using a Bentham DM 150 double monochromator. The instrument was kept inside a temperature-stabilized, weather-protected box to minimize the temperature-dependent gain fluctuations. The box was partitioned into two parts. The right part contained all the electronics (stepper motor drive; high-voltage supply for the photomultiplier; signal amplifier; and the Proportional–Integral–Derivative (PID) controller, which switches heating and ventilators on and off according to the temperature inside the box). The left part of the box contained the monochromator and the highly temperature-sensitive photomultiplier. Because of heat-generating parts in the right part of the box, the temperature was stabilized to 30°C whereas the left part of the box was stabilized to 25°C. The box was made of wood and placed inside a metallic box to protect it against lightning. The Bentham spectrometer has a spectral resolution of 0.6 nm and a step width of 0.5 nm between 280 and 500 nm. Scans were taken every half hour from sunrise to sunset. The duration of a scan was 10 minutes. The entrance optic consisted of a quartz dome and a Teflon diffuser that was connected to the monochromator entrance by an optical fiber. At the beginning of each spectral UV measurement, the dark current and the signal offset were recorded and automatically deducted from the measurement signal. Data were stored in a personal computer that performed the automatic measurement schedule and were transferred by modem to IMP_BOKU in Vienna.
Two independent calibrations were performed with a portable 50-W lamp (from SCI TEC Company) and with a self-built portable 1000-W lamp calibrated to a primary standard 1000-W lamp, with a National Institute of Standards and Technology calibration certificate. These portable lamps are fixed during each calibration at a constant reproducible distance from the diffuser of the Bentham spectrometer. These lamps were mounted in a thick metallic housing to prevent ambient light leakage and to conduct the excess heat away. The intensities of the lamps mainly depend on the current, which was recorded to 6.5 digits of accuracy and kept constant to 6 digits of accuracy for the 1000-W lamp and to 4 digits of accuracy for the 50-W lamp. The voltage of the lamp, giving an indication of an eventual degradation of the lamp, is monitored for both lamps to 4.5 decimal places. In the 1000-W portable lamp, the temperature of the housing is recorded accurately and a thermostat switches the ventilation on and off to keep the internal lamp housed at a constant temperature so as to avoid any influence of air temperature on the bulb emission. In order to check the accuracy of the calibration standard of the institute, IMP_BOKU participated in the 1997 European Standardisation of Ultraviolet Spectroradiometry in Preparation of a European Network (SUSPEN) spectrometer intercomparisons. The results were satisfying and showed agreement between the IMP_BOKU spectrometer and the reference to within 2%–3% (Gardiner and Kirsch 1998).
The calibrations of the Sonnblick instrument were performed every fortnight. The 50-W and the 1000-W portable lamps showed an agreement from 280 to 380 nm to within 2%.
Cosine correction of the data
The angular response of the diffuser was measured in the laboratory. A cosine correction was calculated for the direct and diffuse parts of the incoming radiation. For the calculation of the cosine error of the diffuse irradiance, the algorithm used by Seckmeyer and Bernhard (1993) was applied, except for the hemispherical distribution of the sky radiance, which was calculated with the DISORT model and was not assumed to be isotropic. By calculating the diffuse–direct partitioning, the overall cosine error was calculated (Seckmeyer and Bernhard 1993). According to the simulations, the nonisotropic assumption produces up to 1.5% better accuracy than the isotropic assumption. The results of the cosine error of the diffuse, direct, and global irradiances obtained for the Bentham instrument are shown in Table 1 for three different solar zenith angles. The simulations show a lower cosine error for the diffuse irradiance at 305 nm (0%–5%) than at 370 nm (0%–10%). According to these results, correction multiplication factors for the global irradiance measurements of 1.05, 1.03, and 1.00 were taken at 305 nm for 70°, 50°, and 25° solar zenith angles, respectively. At 380 nm, correction factors of 1.11, 1.03, and 1.00 for 70°, 50°, and 25°, respectively, were taken. The spectral correction factors for the other wavelengths and for the other solar zenith angles were interpolated linearly.
Model and methods
The simulations used the DISORT model, which is based on the discrete ordinate method, with eight streams that, according to Forster (1994), are accurate to within 0.11%, and with 30 levels (an atmosphere with 30 layers), accurate to 1%. The DISORT model was validated within the scope of the 1997 model intercomparison of the European Scientific UV Data Management (SUVDAMA) project (van Weele et al. 1999, manuscript submitted to J. Geophys. Res.). Cases with the same well-defined input parameters were calculated by the different groups, and agreement of the IMP_BOKU DISORT model with the calculations of the other groups to within 1% in the UVA and to within 2% to 3% in the UVB was achieved.
Determination of the model input parameters
The required input parameters are the vertical profiles of air pressure, temperature and ozone; the aerosol optical depth; the aerosol SSA; the aerosol phase function;the total columnar ozone; the ground air pressure; and the solar zenith angle. The air pressure is measured by the Austrian Weather Office at the Sonnblick observatory. Air pressure and temperature input then are calculated for each atmospheric layer. These profiles are defined according to hybrid coordinates. The pressure from each atmospheric layer easily can be changed, and the altitude above sea level therefore easily can be implemented. For the ozone input parameters, the columnar ozone and a yearly averaged ozone profile measured by the Brewer MKIV at Sonnblick were used. The profile was scaled to match the input parameter total columnar ozone. Values of absorption cross section from Bass and Paur (1985) and the Rayleigh cross section of Nicolet (1984) were used. The temperature dependence of the ozone cross section is taken into account by Bass and Paur (1985).
The zenith angle of the sun was determined in the usual way by using well-known astronomical formulas. The extraterrestrial irradiance spectrum is the one from the Solar Ultraviolet Spectral Irradiance Monitor (SUSIM) flown on the third Atmospheric Laboratory for Applications and Science (ATLAS-3; M. VanHoosier 1996, personal communication) and was corrected for the change in Sun–Earth distance. An SSA of 1 and a Henyey–Greenstein aerosol phase function with an asymmetry parameter of 0.73 (IAMAP 1984) were used for the whole wavelength range and for the troposphere and the stratosphere.
Aerosol optical depth input parameters
A NOLL MSD90 sun photometer was used for the determination of the aerosol optical depth at Sonnblick in summer and winter 1992. The NOLL sun photometer has five channels at 368, 500, 675, 778, and 862 nm and was calibrated to the sun by applying the Langley plot method. In December 1992, measurements were carried out on four consecutive days; optical depth values were gathered on two consecutive days in April 1992 and on two days in July 1992. At that altitude the aerosol optical depth was low, and daily-average aerosol optical depth at 368 nm between 0.045 and 0.06 was measured, with maximum and minimum single optical depth values of 0.07 and 0.03. The largest fluctuations occurred during single days, but no change larger than 0.015 in the daily-averaged optical depths occurred from day to day. The Weller and Leiterer (1988) aerosol optical depth measurements on the Mongolian plateau (3000-m altitude) and the Blumthaler et al. (1997) aerosol optical depths measurements at Jungfraujoch, Switzerland (3576 m), and at Izaña, Canary Islands (2367 m), showed average values in the UV wavelength range that are of the same order of magnitude (about 0.01–0.08), not higher than those measured at the Sonnblick observatory. Routine measurements of the aerosol optical depths that have been performed for two years at Zugspitze (2964 m) in the German Alps (Albold 1998) showed that on average on clear days (without cloud obstruction) the daily averages of aerosol optical depths are slightly higher in summer by about 0.01–0.015 compared with the winter values. The values for the other wavelengths were extrapolated by using the Ångström equation and a wavelength exponent of 1.
Determination of the ground albedo
The “effective” albedo that affects the surface UV irradiance is an averaged albedo of the surroundings. The reflected irradiance from the surroundings is reflected back to the earth by the atmosphere and may therefore cause an enhancement of surface UV irradiance. Spatially averaged measurements have been performed mainly by satellites; however, the resolution has an order of magnitude between 50 and 500 km [e.g., 50 km × 50 km for a view from the nadir for the Total Ozone Mapping Spectrometer (TOMS) (Herman and Celarier 1997; Eck et al. 1987)], which may be too large for modeling purposes.
Most measurements of surface reflectivity are performed in the visible range; there are, however, a few groups that reported measurement of the earth reflectivity in the UV wavelength range. Eck et al. (1987) selected for their investigations TOMS minimum reflectivity values at 370 nm over a 3-month period so as to exclude reflectivity enhancement from aerosols and clouds. The albedo values found by Eck et al. for the alpine regions are about 2%. Eck et al., however, mentioned having problems with the altitude correction algorithm over mountainous regions, which might have introduced a ±3% inaccuracy. However, their values refer to an average, whose account of seasonal changes is insufficient. Herman and Celarier (1997) used a longer time period of 14.5 yr of TOMS measurements to deduce seasonal changes in UV albedo. Their measured surface reflectivity values range from 4%–8% in the summer months to values larger than 16% in winter months. However, this study focused especially on snow-free conditions, and no statement as to the exact reflection values of the alpine region was made.
An average albedo of the surroundings of Sonnblick was calculated by weighting reflections from different surface types by the areas covered by the respective types. On a map of the Sonnblick region, a grid (500 m × 500 m) in the 5 km surrounding Sonnblick was drawn, and the type of surface that predominates in each one of these squares was determined. Then the percentages of the total area covered by the different types of surfaces such as primitive rock, grassland, fields, forests, and rock faces was calculated (Table 2). Primitive rock makes up 62% of the horizontal area, glacier 16%, and rock faces approximately 12% of the surface. Then different scenarios such as summertime with old glacier snow, normal wintertime, or wintertime with very fresh snow were assumed by changing the respective albedo values; and an average albedo Aa was calculated by using the following equation:
where Ax is the albedo of the respective type of surface x and Sx is the percentage of the surroundings covered with x.
The resulting albedo values for Sonnblick shown in Table 2 range between 0.12 and 0.88 and therefore are higher than the values obtained by Herman and Celarier (1997). One explanation may be that the resolution of one TOMS pixel is 10 times the size of the area chosen in this study. Second, the effects of the topography are not taken into account in these calculations. Large areas of this mountainous region are in the shade and therefore only reflect the diffuse part of the global irradiance. According to Janza et al. (1975), shadow areas have a brightness that is about 10 times lower than that of areas illuminated by the sun. In regard to general brightness, Janza et al. pointed out that the lower brightness of the shadow areas was compensated for almost fully by the higher brightness of the sloping surfaces exposed to sunlight. They are, however, referring to the visible or the infrared radiation that has proportionally much more direct irradiance than UV irradiance has, and mention only the compensation of the reflection of the direct sun. Janza et al.’s statement is, however, contradictory to the findings of Thomas (1997). Thomas (1997) performed simulations of the average albedo by using a three-dimensional model and found that the albedo in mountainous regions is lower than in flatter regions in the visible and infrared wavelength ranges. In regard to reflection of diffuse irradiance, locations in the bottom of a valley suffer from an obstruction of the horizon and therefore do not get the same amount of diffuse irradiance as do places up on the mountains. One therefore can expect additionally that the amount of reflected UV irradiance to the upper atmosphere is less for a mountainous region than for a flat region with the same type of surface and albedo. In this respect, future efforts should be focused on the influence of topography on the effective albedo.
Methods of investigation
The investigation of the characteristics of the UV radiation at Sonnblick was carried out by comparing the UV measurements with UV model calculations made with the DISORT model.
To exclude the influence of some slight wavelength shifts and the influence of the spectral characteristics (slit function) of the instruments, the integrals over the wavelength ranges 304–306 nm, 314–316 nm, and 360–380 nm were used. The first two intervals were kept so short because of the fast-changing absorption characteristics of ozone within this wavelength range. In the UVA wavelength range, the dependence on the aerosol characteristics shows a small wavelength dependence that should, however, not disturb the investigation. More than 500 clear sky scans from 1996 to 1998 with solar zenith angles smaller than 74° were used for this study. The selection of the clear-sky days was done by using the cloud fraction and sunshine duration measurements of the Austrian Weather Office, which are performed six times per day. Only data with a cloud coverage fraction of 0/8–1/8 were used. The UV data taken at times between the cloudiness readings of the Austrian Weather Office were selected only if the fractional cloudiness was less than 2/8 at the readings before and after the UV measurements. The time taken for the instrument to scan was taken into account for the model calculations where different solar zenith angles were used for the respective wavelengths.
Comparison of absolute UV levels of model calculations with measurements
The absolute UV levels of the measurements first were compared with the model calculations before the influence of the various parameters—ozone, albedo, and aerosol—was quantified by using the fluctuation in the measurement-to-model ratios related to these parameters.
The results of the comparisons between the model calculations and the measurements are shown in Figs. 1a–c for the respective wavelength intervals 304–306, 314–316, and 360–380 nm. These wavelength intervals will be referred to from now on only as 305, 315, and 370 nm. These calculations were performed with a constant Ångström turbidity coefficient of 0.0268 and with a wavelength exponent of 1. This average optical depth is large for a high-mountain location and is equivalent to an optical depth of 0.067 at 370 nm. A constant ground albedo of 0.18 was used for all the calculations. Figures 1a–c show the ratios of the measured UV intensities to the calculated UV intensities as a function of the cosine of the solar zenith angle. The ratios are shown for summer (April–October) and winter (November–March) to trace clearly an eventual solar zenith angle dependence of the measurements. The winter values in general are higher than the summer values [this result will be discussed in section 3b(3)iii]. There is a systematic deviation between measurements and model calculations. The model values are approximately between 15% higher and 2% lower than the measurements at 370 nm, between 16% higher and 5% lower at 315 nm, and between 25% higher and 10% lower at 305 nm. The ratios of the 370-nm values show no significant dependence on the solar zenith angle, whereas the ratios at 305 and 315 nm show a slight solar zenith angle dependence. The average ratios and their standard deviations are shown for 305, 315, and 370 nm in Table 3. The 305-nm standard deviation is 7.7%, almost twice the standard deviation at 370 nm. The UV spectrometers have a much lower accuracy in the UVB than in the UVA wavelength ranges because of the very low signal in the UVB. On the other hand, the use of an averaged vertical ozone and temperature profile for each month may introduce an additional error in the measurement: model (m:m) ratios. To find out whether an agreement between measurements and model calculations can be reached, calculations with a maximum aerosol optical depth of 0.08 at 370 nm were performed (according to our measurements and to values found in the literature). The average m:m ratios show that this change in optical depth does not produce any change larger than 1% in the m:m ratio (Table 3). Model calculations with an albedo value of 0.04, which is [according to the values of Herman and Celarier (1997)] the summer albedo values in Alpine regions, then were done. The enhancement of the m:m ratio is between 1.54% at 305 nm and 2.3% at 315 and 370 nm, with yearly average m:m ratio values of 0.90, 0.92, and 0.95 at 305, 315, and 370 nm, respectively.
The results obtained in this study are comparable to the results obtained by Mayer et al. (1997) in Garmisch-Partenkirchen (700 m) and by Weihs and Webb (1997b) at the Jungfraujoch (3576 m). Mayer et al. found an m:m ratio between 0.9 and 1.02, whereas Weihs and Webb found the measurements to be 10% lower than the calculations. This discrepancy is explained by the uncertainties of both the measurements and the model. Uncertainties in model calculations are due to errors in the model input parameters. At 305 and 315 nm, the largest source of error lies in uncertainties in the vertical temperature and ozone profiles. For example, model simulations show that a change of 10 K in stratospheric temperature may induce changes in ground UV irradiance of 6% at 305 nm and 2% at 315 nm. Other sources of error are input parameters such as the aerosol single-scattering albedo or the aerosol phase function (Weihs and Webb 1997b). However, in this investigation these sources are very small [smaller than 1% if realistic variations in aerosol characteristics according to IAMAP (1984) are assumed] because of the very small aerosol concentrations at higher-altitude sites.
The systematic deviation between the model and the measurements may be explained, in general, by the uncertainty of UV instruments in measuring absolute UV levels. The systematic error of the calibrations of the spectrometer and of the secondary standard lamp calibration in the laboratory—taking into account systematic error of the primary 1000-W lamp, calibration errors, and temperature effects—is around ±5% (Forster 1994; Huber 1994). The calibration certificate of the primary standard 1000-W lamp states an error of ±3% (Huber 1994). According to Woods et al. (1996), UV measurements in space are encountering the same problems. Woods et al. (1996) state the uncertainty of the ATLAS-2 extraterrestrial spectrum (measured during a space shuttle mission) as 4%–8%, where the systematic uncertainty of the irradiance standard alone is 4%.
The reproducibility of UV instruments relating to time is, with approximately 2% uncertainty, much better (Woods et al. 1996; Huber 1994). Tests of the random errors of the lamp and spectrometer performed by IMP_BOKU in the field were, with maximum deviations of around 2%, in the same order of magnitude. The deviations of the single data points therefore may be attributed, with a certainty of 2%, to real changes of the radiative characteristics of the atmosphere.
The values that are far from the average may be due to external factors such as fast-changing weather situations (clouds) or other unusual events that obscure the optical aperture from the incoming irradiance.
Except for the investigation of the influence of ozone on ground UV irradiance, further analysis of measurements will be performed only at 370 nm because of the large uncertainty in model calculations in the short UVB wavelength range due to uncertainties in temperature and ozone profiles.
Comparison of computed UV fluctuations with measured UV variations
Comparison of UV fluctuations as a function of columnar ozone
The measurements and model calculations (58 single measurements and calculations) both were compared as a function of columnar ozone (Fig. 2) at 305 nm for solar zenith angles between 69° and 71°. Figure 2 shows that the model and measurements have the same dependence on ozone. Columnar ozone at Sonnblick follows the typical seasonal fluctuations with maximum values of about 380 DU in spring and minimum values near 230 DU in autumn. The minimum UV values are 0.3 μW m−2 at 305 nm and 7.6 μW m−2 at 315 nm, whereas maximum UV values of 2 μW m−2 at 305 nm are larger by a factor of almost 6 and of 12.9 μW m−2 at 315 nm are larger by a factor of 1.7.
The model simulates the dependence of the UV irradiance on the solar zenith angle and on columnar ozone with sufficient accuracy. We assume therefore that the deviations of the m:m ratios are attributable mostly to changes in aerosol optical depth and surface albedo.
Influence of the aerosol optical depth on the UV irradiance
Because routine aerosol optical depth measurements are lacking for the period of this investigation, the analysis in further sections will be based first on the assumption of a random distribution of the aerosol optical depth with time (no seasonal dependence) and then on the assumption of a summer aerosol optical depth that is 0.015 larger than the winter aerosol optical depth (Albold 1998) [see section 2c(2)]. The impact of maximum fluctuations in aerosol optical depth on the ground UV irradiance however, has, to be assessed. The influence of aerosol optical depth changes on the UV irradiance therefore was simulated with the DISORT model. The choice of the aerosol optical depth values used was based on the optical depth measurement campaigns performed at Sonnblick, which showed aerosol optical depth measurements that were, as already mentioned in section 2c(2), in good agreement with optical depth values recorded by other scientific groups in high-mountain areas. The range of the aerosol optical depth that was assumed for the current simulations was therefore between 0.03 and 0.08. The resulting maximum relative variations in UV irradiance that were obtained are shown in Table 4. The changes in UV irradiance range from 3% at 305 nm to 4.7% at 315 and 370 nm. These results underline the fact that aerosols have a low influence on UV irradiance in high mountainous regions. An increase in aerosol optical depth of 0.015 during summertime [according to section 2c(2)] would induce a decrease in UV irradiance of between 0.75% at 305 nm and 1.5% at 370 nm.
Investigations of UV fluctuations due to changes in albedo
Influence of clouds below the summit
According to Eck et al. (1987), the reflectivity of clouds in the UV at 370 nm is similar qualitatively to the visible reflectivity of clouds. The cloud reflectivity that was measured with TOMS (Eck et al. 1987) shows a considerable increase with cloud-top height. The minimum cloud reflectivities are 51.5 ± 16.2% for low clouds and 76 ± 13.5% for high clouds. This range is in agreement with the values of Frederick and Abrams (1981), who found typical average cloud albedo to lie in the range 50%–60% in the Atmosphere Explorer-E (AE-E) satellite measurements taken at 330–340 nm.
From these results, one may expect, for conditions with clouds below the summit, some increase in average albedo primarily during summertime when the surrounding albedo reaches minimum values [see section 2c(3)].
The influence of a cloud layer below the summit on clear-sky UV irradiance for winter and summertime was investigated. The measurement:model ratio was compared to the fraction of clouds below the summit for 370 nm and is shown in Fig. 3 for winter and summertime, as a function of cloud fraction. Cloud cover below the summit influences the ground UV irradiance only at fractions over 2/8. The average m:m ratio at 370 nm is about 3% higher over the whole year and about 6% higher in wintertime. No significant increase is shown during summertime, although there were some events in summer with 6/8 cloud cover below the Sonnblick summit. These “valley cloud events” are relatively rare, especially in summer. About two-thirds of the clear-sky scans were performed during conditions without any cloud cover below the Sonnblick summit. For this investigation, 54 scans with 1/8 cloud cover, 70 with 2/8, and 61 with cloud cover fractions of greater than 2/8 were used. In summer, only 18 clear-sky scans with a cloud cover of over 2/8 below the summit were taken. That small number considerably reduces the significance of the analysis of the summer measurements. Extreme enhancement of UV radiation caused by changes in albedo may, however, be studied only by looking at single cases. Three high-UV irradiance events happened, for instance, in winter 1997 and 1998, in which clear-sky days with cloud cover of 3/8–6/8 during the whole day showed m:m ratios up to 1.02, which means values 9% higher when compared with the mean winter UV levels.
Seasonal UV fluctuations due to changing ground albedo
The surrounding areas of Sonnblick that are covered with glaciers make up 16% of the surface whereas in wintertime one may expect as much as 88% of the surface to be covered with snow, if one assumes that no snow covers the rock faces (Table 2). Figures 1a–c indicate higher UV values in winter than in summer. To trace a seasonal effect, all the cases with valley cloud cover larger than 2/8 were removed. The average values and the respective standard deviations of the remaining 447 scans are shown in Table 5 as a function of the season for the three wavelength intervals. Because of the considerable uncertainties in calculations at 305 and 315 nm mentioned in section 3a, which are in accordance with the large standard deviations of the m:m ratios, the values at 370 nm are the best indicator as to the investigated effects. At 370 nm, winter values are, on average, 4.6% higher than the summer values. This difference may be attributed to changes in ground albedo.
The m:m ratios therefore were plotted as a function of the snow depth as measured by the weather observer at a given place on the Sonnblick glacier and are shown in Fig. 4. An enhancement of UV irradiance is shown between snow depths of 100 and 200 cm, but UV irradiance remains at a constant level for snow depths larger than 250 cm. Thus, snow depth is only a partial indicator of the extent of snow coverage of surrounding areas. Fresh snowfalls may, for instance, remain on the glacier but melt at lower valley locations. The conditions (e.g., dirtiness, age) of the snow are also factors that are independent of the total snow amount but that highly influence the reflective characteristics of snow. Investigations about the influence of the depth of new snowfalls (the same day or the day before) on UV irradiance were analyzed but no meaningful connection could be traced. Decrease in UV radiation, however, seems dependent on the aging of the snow. A decrease of about 3% in the m:m ratio is shown for days on which the last snowfall occurred more than 3 days previously (Table 6).
Overview and analysis of the changes in UV
The order of magnitude of the presumed changes in albedo was estimated by changing the albedo values in the DISORT model simulations until the aimed-for change in UV irradiance was reached. Before these simulations, it was confirmed that the relative change in UV radiation due to changes in albedo was dependent on neither solar zenith angle nor on the level of the albedo, but only on the relative change in albedo. Results are shown for a random distribution of the aerosol optical depth relating to the time, and for aerosol optical depths that are 0.015 larger in summer than in winter. A change in albedo of 0.1, for instance, will induce a change in UV irradiance of approximately 1.1% at 305 nm, 2% at 315 nm, and 2% at 370 nm. Table 7 shows the presumed changes in albedo for 370 nm. A maximum fluctuation in average ground albedo between 0.12 and 0.88 [see section 2c(3)] would induce an 18% change in UV irradiance. The observed increase of 4.5% in average UV irradiance during wintertime represents an increase in albedo of approximately 0.22 if constant aerosol optical depth is assumed and 0.15 if a yearly course of the aerosol optical depth is assumed. Clouds below the summit may enhance the albedo by approximately 0.30, with a maximum increase of 0.45. The values stated in Table 7 are approximate values; according to the standard deviation of the UV values, one may expect an uncertainty in albedo of ±0.15. The magnitude of the changes in UV irradiance that occur at Sonnblick observatory are shown in Table 8. Ozone has the largest impact on UV irradiance at 305 nm and 315 nm with changes in UV radiation of 560% and 78%, respectively. Aerosols account for maximum changes in UV irradiance on the order of 5%, whereas ground albedo–induced changes in UV are on the order of 4.5%. A cloud cover below the Sonnblick summit can induce changes in UV irradiance of up to 9%.
The goal of this study was to characterize the UV irradiance at the high-mountain Sonnblick site and study the effect of ground albedo on ground UV irradiance. For that purpose, comparisons with the radiative transfer model DISORT were used to study the nonsolar zenith angle–dependent changes in clear-sky UV irradiance. The comparisons between measurements and models showed that the measured UV levels are lower by approximately 12% at 305 nm, 10% at 315 nm, and 8% at 370 nm than the values calculated with the DISORT model. These discrepancies are attributable mainly to systematic discrepancies in the calibration of the IMP_BOKU spectrometer to the instruments that performed the ATLAS-3 (extraterrestrial spectrum) measurements. Considerations of the uncertainty in measurements from UV instruments show that m:m comparisons are not suited for the retrieval of the average effective albedo. The stability of UV instruments relative to time, however, is much better; thus, relative changes in UV may be used for further analysis. The m:m comparisons show the same dependence of the DISORT model and measurements on solar zenith angle at 370 nm and on ozone at 305 nm. The ratios of the measurements to the model calculations for 501 scans therefore were taken to study the influence of the ground albedo on UV. Because of larger uncertainties in model calculations at 305 and 315 nm caused by uncertainties in input parameters, the investigations were performed mainly at 370 nm. On the assumption that the aerosol optical depth is constant over the whole year, an increase in albedo of 0.22 in wintertime as compared with summertime was obtained. This seasonal change in albedo would be only about 0.14 if an increase of the aerosol optical depth of 0.015 in summer (relative to winter) was assumed.
By assuming an average summer albedo between 0.04 (Herman and Celarier 1997) and 0.12 (maximum value for summertime calculated in this study), we estimate that the average effective winter albedo amounts to approximately 0.18–0.34. That albedo is much lower than the effective albedo that one would estimate (usually between 0.4 and 0.9 because of the high reflectivity of snow) in snow-covered regions. This discrepancy probably will be explained by looking at the effect of topography on the average effective albedo. Horizon obstruction in the bottom of valleys and shading from the direct sunlight, as well as directional reflectivity and slope inclinations, are factors that should be taken into account for the accurate determination of the albedo in mountainous regions.
It also is important to point out that the low albedo-induced increases in clear-sky surface UV irradiance shown in our study do not imply that people on snowfields are not exposed to higher UV levels, because most of the additional UV stress is reflected directly from the ground to the people.
This work has been funded by the Austrian Ministry for Environment and Transport and the European Commission, DG XII, within the scope of the projects SUVDAMA and CUVRA. We thank the Austrian Weather Office for making the meteorological data available to us. We also would like to acknowledge the help of the Sonnblickverein and that of the weather observers at Sonnblick observatory.
Corresponding author address: Dr. P. Weihs, Institut für Meteorologie und Physik, Universität für Bodenkultur, Türkenschanzstrasse 18, A-1180 Vienna, Austria.