Measurements of solar UVR started in the first decades of the 20th century with chemical detectors [8], where the changing of the color of a solution was an indicator of UV-B irradiance. Physical measurements of UVR use photoelectric methods to give quantitative information about the intensity and spectral distribution of UVR. The first extensive data sets originate from the 1960's, from Davos (Switzerland) [9], but it was not until the 1990's that more of such high quality spectral measurements were made at further locations world-wide. Since the 1970's a greater number of UV measurements have been made with a different type of detector, sensitive to a broad wavelength range in the UV-B (and to a lesser extent in the UV-A), the so-called broadband erythemal detectors. They have been used in many parts of the world for monitoring over periods of several years, but systematic, long-term observations are very rare and again were mainly established in the 1990's in response to stratospheric ozone depletion. More recently, estimations of UVR based on measurements with instruments on satellite platforms allow more global coverage.

2.3.1 Ground-based measurements Instrumentation

Different types of UV instruments are used for measuring solar UVR depending on the objectives of the measurements and on the required accuracy. Instrument types differ in the extent of spectral information that they provide between high (e.g., spectroradiometers), moderate (e.g., multifilter radiometers) or low (e.g., broadband radiometers, dosimeters) spectral resolution. In addition, time integration is usually done by frequent sampling, whereas with dosimeters a time integration is inherent in the measurement.

The entrance optics of all types of instrument are usually designed to measure global irradiance, that is radiation falling on a horizontal surface with a spatial weighting according to the cosine of the zenith angle of the radiance. This response is achieved by the so-called cosine-diffuser. Flat pieces of Teflon provide a first approximation to the cosine-law, but for higher quality measurements more care is necessary (i.e., by using specially shaped Teflon). As an alternative, integrating spheres are used in some cases as input optics. For measurements of actinic fluxes (the radiation on the surface of a small sphere) instead of irradiances, the input optic has to have a uniform sensitivity in all directions, which is realized by a 47i-head, or a 27i-head for one hemisphere only.

Spectral instruments: The highest amount of information about UVR can be gathered with spectral instruments. These instruments use a monochromator to disperse (separate) the incoming radiation into small spectral intervals, in most cases by using a diffraction grating, and then measure the signal at individual wavelengths with a photoelectric detector. The spectral resolution is determined by the geometry of the entrance- and exit-slits of the monochromator, which define the slit function. It has usually the shape of a triangle and is characterized by the full-width at half of the maximum signal (FWHM). Typically values for FWHM are between 0.2 and 2 nm.

As a consequence of the steep decline of the solar spectrum at UV-B wavelengths, high stray-light rejection is necessary to avoid erroneous signals from higher wavelengths (in the UV-A and visible), where the intensity of the solar radiation is several orders of magnitude higher than the UV-B. Therefore, sensitivity outside the ideal slit function should be as small as possible. This can be achieved best by using a double-monochromator, where a second monochromator follows at the exit slit of the first one, and also gives the additional advantage of higher spectral resolution.

The wavelength setting of a monochromator may be carried out by rotating the grating(s), which directs radiation of the individual wavelengths to the exit slit, where it is measured with a photomultiplier tube. This type is called a scanning spectroradiometer, and it is the usual practice for double mono-chromators. For single monochromators it is also possible to mount a diode array or a CCD-element at the exit slit, providing information on the intensity at individual wavelengths without any mechanically moving parts. Furthermore, the information about all wavelengths is taken simultaneously, whereas scanning radiometers usually need a time in the range of 1 to 10 minutes to measure a wavelength range of about 100 nm. However, for measurements of solar UV-B radiation the best results are currently achieved with scanning spec-troradiometers with an attached photomultiplier, providing superior stray light rejection and a high dynamic range.

Spectroradiometers for field measurements of solar radiation are complex and sensitive instruments, usually they require temperature stabilization and they need frequent control of their calibration with respect to wavelength and to irradiance. Therefore maintenance of these instruments is a challenging task and needs significant manpower and experienced operators.

The calibration of the output signal of a spectroradiometer to the incoming radiation is usually done by comparison with a 1000 W quartz tungsten halogen lamp. Standard lamps of this type are calibrated by national calibration laboratories, giving the irradiance (W m-2) at defined wavelengths and at a defined distance. However, even from the National Reference Laboratories world-wide, the uncertainty of the calibration of these lamps in the UV range is in the order of 2-4% (95% range), which ultimately limits the possible accuracy of solar UV measurements. Adding transfer uncertainties, leads to a best-achievable final uncertainty of solar UV measurements of about 5% (95% range) [10]. A level of agreement of about 5% has also been found between several UV-measurement groups in the most recent spectroradiometer intercomparisons, where up to 20 different spectroradiometers were simultaneously measuring solar UVR [11].

Broadband instruments: Broadband instruments measure solar irradiance in a specified wavelength range, typically 20 nm to 100 nm wide. This range is defined by the construction of the detector and it results from a combination of different optical elements such as filters and photoelectric sensors. The output signal of broadband instruments corresponds to the integral of the incident irradiance multiplied by the spectral response of the detector. Therefore, any information about the detailed spectral structure of the incident solar radiation is lost. On the other hand, the measurement is instantaneous and thus allows rapid changes in irradiance to be followed, due to fast moving clouds for example.

One of the most commonly used type of broadband detectors for measurement of solar UVR, the so-called Robertson-Berger type detector [12,13], has a spectral sensitivity which is adapted to the standardized erythema action spectrum of the human skin [14]. It has the maximum of its sensitivity around 297 nm, then sensitivity decreases steeply to about 320 nm and in the U V-A range the sensitivity is about 1000 times smaller than at the maximum in the UV-B. Thus these detectors give a direct measure for the biologically relevant irradiance. However, as no one available detector has a spectral sensitivity perfectly matched with the erythema action spectrum, corrections are necessary to get a standardized output from these instruments. These corrections depend on the variation of the solar spectrum, mainly correlated with solar zenith angle and with total atmospheric ozone content, and are specific to an individual instrument. This conversion from detector based units into absolutely defined erythemally weighted units may not be necessary if relative variations of UVR are observed over longer time scales at one station only and no absolute comparison is made with results from other detectors.

Currently a great number of detectors of this type is in use world wide, but the calibration to a common reference is often not possible. To help address this problem the World Meteorological Organization has organized two (1995,1999) international intercomparisons of broadband detectors to encourage homogen-ization of the data between different sites [15,16]. Meanwhile, in the USA a central calibration facility has been established [17] to serve the different organizations that operate broadband instruments, thus helping to maintain a constant quality of data.

As the absolute calibration of broadband detectors is usually based on field intercomparisons with spectroradiometers and an additional conversion of raw data into standardized erythemally weighted units is necessary, the overall uncertainty of broadband detectors is higher than that for spectroradiometers. Under careful operation and frequent recalibration, an uncertainty of about 7-8% (95% range) might be achievable. This is the absolute uncertainty of an individual detector - if similar detectors are used in a network, then the relative uncertainty between these individual detectors might be, at best, in the order of 2-3% [18]. It is important to mention that although the price of broadband detectors is relatively low; these detectors also need a significant amount of quality control and quality assurance to give reliable results. Operating a net work of broadband meters at a high level of quality is a challenging task. Long-term stability cannot be expected by itself, but it has to be verified by the operator. In addition it now appears that internal humidity of the detectors might significantly affect the response and therefore careful maintenance of the attached desiccant is important.

Moderate bandwidth instruments: Instruments that measure solar radiation with a bandwidth between about 2 and 20 nm are called moderate bandwidth instruments. In the most cases, interference filters are used in combination with photodiodes. For absolute calibration the spectral sensitivity of the detector has to be known with high accuracy over the whole wavelength range of solar radiation, in order to avoid distortions from secondary transmission regions far away from the central wavelength. Filters in the UV-B range around 300 nm are particularly sensitive to this problem, and the long-term stability of the transmission of the interference filters also has to be tested carefully. The small size of filters and diodes allows several channels to be combined in one instrument, thus offering simultaneous measurements with moderate bandwidths over a broad wavelength range, usually with 4-8 channels from the UV to the visible range. Therefore spectral effects of solar radiation (in the range of a few nm) can be investigated under all weather conditions.

Data from multi-filter, moderate bandwidth instruments are often post-processed in combination with radiative transfer modeling, which allows the reconstruction of the full solar spectrum. In a second step, the calculated solar spectrum can be weighted with any biological weighting function, i.e. for determination of erythemal doses [19].

Biological UV dosimeters: Dosimeters for UVR measure the integrated dose over the time of exposure. A well utilized dosimeter based on a polymer and with a response spectrum similar to that of erythema is polysulfone film, developed in the 1970's [20]. More recently biological UV dosimeters have been developed in the 1990's. These are based directly on a specific biological reaction (e.g., DNA damage), the inactivation of bacterial spores or bacteriophages, a photochemical reaction in the in situ photosynthesis of vitamin D, or on uracil molecules [21,22]. The biological material is exposed to solar UVR and its response is then measured (often by a slow analysis process some time after irradiation). In order to quantify the reading of these dosimeters, the spectral response function of the specific reaction has to be known, as well as exposure geometry, linearity and stability to environmental parameters. Usually the dynamic range of these detectors is rather small. On the other hand, for some biological applications it is an advantage to collect directly the dose (total over time) of the biologically relevant component of incident radiation fluxes. However, the conversion into absolute radiative quantities may still have uncertainties greater than 10% [23]. Results

The effects of the different parameters that determine solar UVR at the Earth's surface are illustrated with the results of UV measurements, demonstrating the great natural variability of the solar UVR.

Solar elevation: The dependence of solar UVR on solar elevation is shown in

Figure 7 for a cloudless day at Izana Observatory, Canary Islands, Spain (latitude 28.3°N, longitude 16.5°W, altitude 2367 m above sea level). Solar elevation at noon on 18.07.1995 was 80.3°, total ozone 282 DU and aerosol optical depth 0.06 at 350 nm. Therefore the measured irradiance was close to the expected maximum for these solar elevations, due to the relatively low ozone content, the very low amount of aerosols and the high altitude. Higher values could only be expected if the surroundings had been covered with snow: in fact under the measurement conditions the albedo of the surrounding was very low. From Figure 7 it can be seen that for erythemally weighted irradiance the noon value was 350 mW m~2, which corresponds to an UV-Index of 14 (1 UV-Index = 25 mW m~2). In addition to global erythemal irradiance, the diurnal variation of spectral irradiance at 302 nm and at 320 nm is shown in relative units, normalized to the maximum of erythemal irradiance. This shows that for shorter wavelengths the change of irradiance with solar elevation is steeper than for longer wavelengths, and the variation of erythemal irradiance as a broadband integral with a central wavelength around 310 nm (dependent on solar elevation) lies in between. Therefore the range of variation of irradiance for solar elevation from 20° to 80° is about a factor 6 at 320 nm and a factor 70 at 302 nm, showing that at low solar elevations the relative contribution of shorter wavelengths is much lower than at higher solar elevations. The reason for this effect is the absorption in the ozone layer, which is more pronounced at low

1 1



j S"',

'yz y .

' / ''' / / j


320 nm -> / ! / i ' L x

/<r- 3l02 nm / j

;/ / /

' \

- yf J / L .


/ ; j




Figure 7. Dependence of global erythemal irradiance (solid curve) on solar elevation for a clear sky day (18 July, 1995) at Izana Observatory (Canary Islands, Spain, 28.3°N, 16.5°W, 2367 m above sea level) with total ozone 282 DU and aerosol optical depth 0.06 at 350 nm. Spectral irradiances at 302 nm and 320 nm (dashed curves) on the same day are given in relative units, normalized to the maximum of erythemal irradiance.

solar elevations and more effective for short wavelengths in the UV-B.

From Figure 7 the seasonal variation of UVR at noon can also be estimated, remembering that noon solar elevation increases from winter solstice to summer solstice by about 47°. Annual maximum noon solar elevation (se(max)) can be calculated for any latitude higher than 23.5° as (se(max)) = 113.5° —latitude, while for latitudes lower than 23.5° the maximum noon solar elevation is of course 90°. As already mentioned, the absolute values of erythemal irradiance shown in Figure 7 correspond to a situation with extremely high irradiance; however, relative variations of irradiance with solar elevation can be estimated from these measurements.

The change in solar elevation can also be interpreted in terms of changing latitude. At 20° latitude northwards of the station Izana, the noon intensity would be (under all other identical conditions) 100 mW m-2 less, which is about 30% smaller. Moving another 20° northwards would reduce erythemal irradiance again by about 125 mW m~2, which means that irradiance at 68°N is about 50% less than at 48°N.

Ozone: The relation between changes of total ozone column and the corresponding changes in UVR is well established through measurements. That means that if ozone decreases then UVR increases, when all other parameters that influence UVR are constant. The shorter the wavelength in the UV-B range, the stronger is the increase due to ozone decrease, as a consequence of the spectral shape of the ozone absorption cross-section. For broadband spectral ranges, the simplification by using a power law to describe the relation between irradiance (I) and ozone (0),

is justified for a broad range of applications. Therefore the Radiation Amplification Factor (RAF) can be used to estimate the effect of changing ozone on UVR for various weighting functions. The RAF depends slightly on solar elevation and on absolute ozone content, and one should consider that the linear relation between variation of irradiance (A I in %) and ozone (A O in %)

becomes significantly erroneous if ozone variations greater than about 20% occur. Some RAF's (calculated for daily totals, July, 30°N, 305 DU) [25] are given in Table 2.

Sensitivity in the UV-A range of an action spectrum contributes significantly to the RAF (by reducing it) because the intensity of solar irradiance in the UV-A range is about 1000-fold higher than in the UV-B range. In addition, small uncertainties in the action spectrum in the UV-A, where the determination of the action spectrum is often more difficult, result in significant uncertainties of the overall RAF.

Aerosols: The amount, type and optical characteristics of aerosols are usually not very well known when UV measurements are made. This is the cause of the greatest uncertainties in the comparison of UV measurements with results of radiative transfer models under cloudless skies. The determination of the aerosol

Table 2. Radiation amplification factors (RAF) for various action spectra, calculated for daily totals of solar radiation in July, 30°N, 305 DU ozone amount

Action spectrum


Skin erythema [14]


Skin cancer[26]


DNA damage [27]


Photokeratitis [28]


Cataract [28]


Phytoplankton motility [29]


J O('D) Photolysis [30]


optical depth, which is the basic parameter used to quantify the actual aerosol amount, is usually carried out by sunphotometric measurements. With these instruments the intensity of direct solar irradiance, either spectral or in specific wavelength ranges, is measured and from the known extraterrestrial spectrum and the known attenuation of direct solar irradiance by molecules (Rayleigh-scattering) the aerosol optical depth is derived. If absorption by gases takes place too (in the UV by ozone), this quantity must also be known in order to derive aerosol optical depth. The vertical distribution of aerosols can be determined with lidar systems, which are becoming more portable and now feasible additions to measurement campaigns. The separation between tropospheric aerosols and stratospheric aerosols is important, because they affect solar irradiance differently.

The separation between scattering and absorbing component of the aerosols is characterized by the single scattering albedo. For in situ measurements (at the ground or from an airplane) instruments are available to provide this information. Attempts have recently been made to derive the single scattering albedo from radiance measurements of the scattered radiation of the sky in combination with radiative transfer calculations [31,32], which gives reasonable results for higher aerosol optical depth (greater about 0.4). In a similar way, the spatial characteristics of the forward scattering of the aerosols can be derived. Therefore measurements of clear sky radiance distributions at wavelengths in the UV range have the potential to provide valuable information about the atmosphere.

The most complete information about aerosols would be vertical profiles of their size distribution and their chemical composition, which would allow the complex refractive index to be determined. However, such detail is usually not available.

The regional and temporal distribution of aerosols is strongly variable, therefore measurements of their effect on solar irradiance are primarily local information. In general, in addition to a tropospheric and stratospheric background aerosol load local enhancements typically occur due to urban pollution or natural events (e.g., forest fires). An example of the effect of aerosols on UVR is shown in Figure 8, where the attenuation of UV-B irradiance by varying aerosols is shown [33]. The measurements near the city of Athens (Greece) in summer 1996 were first simulated with a radiative transfer model, where the aerosol optical depth was independently measured and the aerosol single scattering




154 156 158 160 162 164 166 168 170 172 174 176

Day of year

Figure 8. Attenuation of erythemal UVR (measured with a Bentham spectroradiometer) by aerosols on several days in summer 1996 near Athens, Greece. Ratio of model calculations including the actual aerosols relative to calculations without aerosols.

albedo was the only free parameter to fit the model to the measurements. In this case, a variation of single scattering albedo between 0.85 and 0.98 was found, consistent for two independent spectroradiometric measurements. Comparing the model result for calculated spectra with and without aerosols shows great temporal variability and reductions of up to 35% due to the aerosols. Only situations without any observable clouds are illustrated here.

Worldwide enhancements of aerosols in the stratosphere are observed after big volcanic eruptions (i.e., by the volcano Pinatubo in 1991), which led to a decrease in direct solar irradiance and an increase in diffuse irradiance. This effect could be measured especially well at a high mountain station, where the disturbance by urban aerosol pollution is very small. In this case, diffuse solar radiation was increased nearly twofold, while global (direct and diffuse) radiation was reduced by about 4% [34].

Albedo: Model calculations show for increasing albedo an increase of solar irradiance with decreasing wavelength in the UV-A and a maximal effect around 320 nm (Figure 9). In the UV-B range, the effect of increasing albedo is less pronounced due to absorption of the reflected radiation by tropospheric ozone. The amplification of global irradiance at 320nm for an increase of albedo by 0.1 is between 3% and 4%. Therefore a change of the terrain from snow free (average UV albedo 0.03) to a fresh snow covered terrain (UV albedo 0.9) will result in an increase at 320 nm of about 30%, and of about 18% at 400 nm. These numbers are valid for cloudless sky. If there are clouds, then the effect of surfaces with high albedo is enlarged due to multiple reflections between the cloud and the surface.

The calculations above are valid for the assumption of a homogeneous, horizontally unlimited surface with a specific albedo. Of course this is not true in reality. Recently, 3D radiative transfer models became available, using the

—1—1—1—1—1—1—1—1—1—1—1—1—1—1—1—1—1—1— k • A

1k 1 !i ' 1 * L


1 M :l t • • ; 1» J t l\ •• *»

. }

\ >

* : f

• • • •


' 1


Bentham i i i i i i i i i i i i i i i

« i

154 156 158 160 162 164 166 168 170 172 174 176

Day of year


Figure 9. Increase of spectral global irradiance as a consequence of an increase of albedo by 0.1 for three different levels of albedo.


Figure 9. Increase of spectral global irradiance as a consequence of an increase of albedo by 0.1 for three different levels of albedo.

Monte Carlo method to calculate irradiance for any inhomogeneous surface with variable albedo. This can be applied to specific terrains to show the local variability [35]. For general application in the standard ID radiative transfer models, an effective albedo is assumed. This is the value for a homogeneous surface that has the same effect on modeled irradiance as the real, in-homogeneous distribution. In Alpine areas with snow coverage, values of effective albedo in the range 0.3 to 0.8 have been derived, depending on local topography and ground vegetation.

3D models were also used to derive the radius of significance around a measurement site, i.e. the area over which the albedo still has an influence on local UVR. It was found that for clear sky conditions this radius can extend to 20-30 km [36], whereas it is significantly smaller in the presence of cloud.

So far the effect of changing albedo was discussed only on global irradiance, that means on horizontal detector surfaces. If the surface is tilted or if actinic detectors (2n or 4k) are used, then significantly higher effects due to high albedo have to be expected.

Altitude: The intensity of solar radiation increases with increasing altitude under cloudless conditions due to the smaller amount of scattering molecules and particles at higher altitudes. That means that direct irradiance is increasing strongly, whereas diffuse radiation is nearly constant or slightly decreasing with altitude. As the separation of global irradiance into direct and diffuse components is dependent on wavelength (with a higher diffuse component at shorter wavelengths), the increase of global irradiance with altitude is wavelength dependent. In addition, because of the wavelength dependence of the scattering coefficients, the increase with altitude is more pronounced at the shorter wavelengths, and the absorption by tropospheric ozone will also result in an enhanced increase of irradiance at shorter wavelengths in the UV-B range with increasing altitude. As so many parameters act together, it is not possible to state only one number for the increase of irradiance with altitude, but it is necessary to give a range of values, depending on the specific conditions.

Usually the term "altitude effect" is used to describe the percentage increase of irradiance for an increase in altitude by 1000 m. Model calculations with an atmosphere without any aerosols and without any ozone give an altitude effect of about 6% in the UV-A and 9% at 310 nm. Measurements of UVR at different altitudes in the Chilenean Andes have indicated altitude effects of 8% in the UV-A and 9% in the UV-B [37]. These values are related to relatively clean air at all altitudes with probably very low tropospheric ozone pollution. Significantly higher values for the altitude effect are found in the Alps, where a marked gradient in pollution from the lower altitude to the higher altitude is typical. Several studies [38,39] have shown values in the UV-A of about 10% and at 310 nm of about 15% to 20% (Figure 10). If the higher altitudes are covered by snow, while the lower altitudes are snow free (which is a typical situation in the Alps), then the altitude effect for UVR is increased by about 10%.

Clouds: The discussion of solar UV variability in preceeding paragraphs has assumed cloud free conditions. However, in most parts of the world this is the exception, usually there are either broken cloud fields or a more or less homogeneous cloud cover. In general, clouds reduce solar radiation, but the amount of reduction is extremely variable due to the variable nature of clouds. For the estimation of the effect of clouds on solar irradiance, the most important par-

Figure 10. Increase of spectral global irradiance for an increase in altitude by 1000 m, based on measurements of three spectroradiometers (ATI, DEZ, FRG) at different altitudes (1200 m, 1750 m, 2964 m) relative to an instrument at a valley station (Garmisch-Partenkirchen, Germany, 730 m). The solid line is the average of the individual measurements.

Figure 10. Increase of spectral global irradiance for an increase in altitude by 1000 m, based on measurements of three spectroradiometers (ATI, DEZ, FRG) at different altitudes (1200 m, 1750 m, 2964 m) relative to an instrument at a valley station (Garmisch-Partenkirchen, Germany, 730 m). The solid line is the average of the individual measurements.

ameter is not the amount of clouds, which cover the sky, but it is the coverage of the sun by clouds. Investigations have shown that even in situations when 70% or 80% of the sky are covered by clouds, the global solar irradiance is only marginally reduced, as long as the sun is not covered by clouds. The reduction by complete cloudiness, when the direct sun is not visible, can range up to more than 90%: average values are about 75% at sea level and about 50% at high altitudes, as there the optical thickness of the clouds is usually smaller.

Within clouds, solar radiation is mainly scattered on small water droplets. This scattering process is only slightly dependent on wavelength, therefore the color of clouds appears white or grey or dark, depending on the thickness, as long as the illuminating sun is "white" (when the solar elevation is low and the sun is reddish, then clouds also may appear reddish). However, the effect of clouds on global irradiance at the ground is dependent on wavelength: irradiance at shorter wavelengths is less attenuated than at longer wavelengths. The reason is the different distribution between direct and diffuse irradiance at different wavelengths. If a cloud is blocking the direct sun, then the reduction of visible global irradiance is much higher than that of UVR, where already a great part of global irradiance is diffuse. From radiation measurements together with detailed cloud observations at the High Alpine Research Station Jungfraujoch (3576 m above sea level, Switzerland) it is found that UV-A and erythemally weighted UVR are attenuated about in the same amount, whereas total solar radiation (300 nm to 3000 nm) is attenuated about 40% more [40] (Figure 11). From measurements of spectral irradiance above and below a cloud layer in Garmisch-Partenkirchen (Germany) again a spectral dependence of the effect of clouds was derived, where the transmission of the cloud was in the UV-B 57% and in the UV-A 45% [41].

The multiple scattering of photons in the cloud and between the cloud and the ground enlarges the average photon path length. Therefore any absorption by tropospheric ozone or aerosols is amplified and thus the spectral shape of the spectrum is significantly modified by clouds.

Global distribution: When comparing measurements of UVR at the Northern and Southern hemisphere, significantly higher values are found on the Southern hemisphere at the same solar elevations in the respective summer time. There are several reasons for this: firstly, the Earth-sun distance has its minimum in the Southern hemisphere summertime. Secondly, the average aerosol load in the planetary boundary layer and in the free troposphere is significantly lower in the Southern hemisphere. Thirdly, tropospheric ozone levels are lower for the Southern hemisphere. All three facts act together in increasing the level of UV-B irradiance on the Southern hemisphere in relation to the Northern. In a measurement campaign in 1990/1991 [42] it was shown that erythema doses in summertime at mid-latitudes (about 48°) in New Zealand were about double those in Germany in the corresponding summer. At that time, systematically lower total ozone content was also observed in the Southern summer relative to Northern summer.

Worldwide maxima of measured irradiances for erythemally weighted UV are found in the high altitude region of Northern Argentina, where altitudes around



LU 1

01 23456789 10


Figure 11. Ratio of global radiation fluxes (normalized to cloudless conditions) for erythemal radiation, UV-A radiation and total solar radiation in dependence on cloudiness, separated for cases when the sun is free or when the sun is totally covered by clouds. Bars indicate + 3 standard deviation of the mean.

3500 m are regularly populated. In summertime, when the sun is close to the zenith, a maximal UV Index of 20 was measured at 22° South and 3500 m above sea level [43], at a time when the total ozone amount had relatively low values (236 DU, the annual average at this site is about 260 DU).

2.3.2 Space-born measuremen ts

Since 1978 spectral measurements of sunlight backscattered from the Earth to the space have been made from satellites, starting with NASA's Nimbus-7 satellite with the TOMS (Total ozone mapping spectrometer) instrument. This spectrometer was designed to measure backscattered UVR at six wavelengths in the UV-B and UV-A and to derive total column ozone amount from these radiance data. This application is well established and the uncertainty is well determined [44]. The combination of the space-born measurements of backscat-tered sunlight at several wavelengths in the UV range with a radiative transfer model also allows an estimate of the UV spectrum at the Earth's surface [45]. UV estimates based on measurements of TOMS instruments on board of various satellites are available now for November 1978 to May 1993, August 1991 to December 1994 and August 1996 to present. Surface UVR has also been estimated from the Earth Radiation Satellite (ERS-2) with the GOME (Global Ozone Monitoring Experiment) instrument [46].

The great advantage of estimating surface UVR from space-born measurements is that results for the whole globe can be derived. Thus a global daily image of the geographic distribution of UV irradiance in the UV-B and UV-A range is calculated. However, from the principle of this method it is clear that not all relevant parameters for the radiative transfer calculations can be measured by the instrument on the satellite. When ozone is known from the well established ozone retrieval, then the remaining main parameters are ground albedo, aerosols and clouds. From the long-term satellite measurements, a global climatology for albedo has been derived, based on the minimum values of reflectivity observed, which is assumed to be the real ground reflectivity without any clouds. This is done for a wavelength in the UV-A range to avoid any interaction with the absorption by ozone. The only difficulty is the separation between cloud-free snow-covered terrain and snow-free terrain with clouds, which is the source of an additional uncertainty. The determination of tropospheric aerosols is possible for absorbing aerosols above about 1.5 km [47], whereas the estimation of the effect of aerosols in the boundary layer remains uncertain. Also, the interaction of changing ground albedo and changing amounts of aerosols close to the ground remains a problem.

The great natural variability of cloudiness in space and in time requires further consideration. As the satellite has an overpass of a given place at regular intervals, ranging from hours to days, the information about cloudiness is only a snapshot. Combining the results of several satellites may improve the temporal resolution, but temporal averaging will still be necessary. Therefore, the most valuable results of space-born retrievals are monthly averages of solar irradiance. Only in specific regions, where cloudiness is very low over time scales of days, can instantaneous values be derived with reasonable confidence. The regional distribution of cloudiness results in a similar problem as the temporal variation. The minimum spatial resolution of data derived from satellite measurements is given by the pixel size of the instrument on the satellite. This ranges from about 100 x 100 km2 for the TOMS instrument down to about 1 x 1 km2 for the cloud information from the AVHRR (Advanced Very High Resolution Radiometer) instrument of NOAA (National Oceanic and Atmospheric Administration). Clouds smaller than the pixel size of the instrument cannot be detected.

Comparisons between UV estimates derived from satellite data and ground-based measurements at a few stations in Europe have shown that for daily doses there is only a bias between data sets of about 5%, but the standard deviation is in the order of about 30% [48]. For monthly doses, the scatter becomes signifi cantly smaller, but differences of up to 30% are observed, depending on the site of the comparison [49,50]. The differences seem to correlate with tropospheric extinction and with seasonal changes in regional snow cover.

0 0

Post a comment