UV irradiance at the surface of the Earth depends on several variable factors identified in Section 5.1. These factors include scattering (Rayleigh, aerosol, and cloud) and absorption (ozone, aerosol, and pollution) processes that occur in the atmosphere, as well as variations in extraterrestrial solar flux and ground reflectivity. Variability in each of these factors combines to produce large fluctuations in UV irradiance-for example, between corresponding months of different years (Weatherhead et al., 1997). This large variability makes it difficult to quantify systematic decadal changes in UV irradiance and interpret them in terms of cause and effect with instruments other than well-maintained spectroradiometers.
It is important to understand and quantify the individual effects of the many variables affecting UV irradiance at the surface of the Earth. The contributions of the different variables can be studied with the use of radiative transfer models. Several models have been developed for a variety of applications (Dave, 1965; Frederick and Lubin, 1988; Stamnes et al., 1988; Madronich, 1992; Ruggaber et al., 1994; Forster, 1995; Herman et al., 1996). These models use extraterrestrial solar spectral irradiance (Mentall et al., 1981; Neckel and Labs, 1984; Kaye and Miller, 1996; Woods et al., 1996) as input and simulate the physical processes that occur as radiation is scattered and absorbed by the atmosphere and at the surface of the Earth. Model output includes global (direct plus diffuse) and diffuse spectral radiation at the Earth's surface in a form that can be compared with measurements.
The comparison of UV measurements with model simulations is an important exercise for checking both the accuracy of the model and the quality of measurements. Once it is demonstrated that the measurements and model results are in good agreement for a wide range of conditions, a reliable simulation of the transfer of UV radiation through the atmosphere is possible. The model can then be used with confidence to extend a measurement series in time and in space (between ground-based stations), provided measurements of all variables affecting surface UV irradiance are available. The model can also estimate future levels of surface UV irradiance by using predictions of variables such as aerosols or ozone that may change as a result of increased air traffic. Models are used in this context in Section 5.4.
Comparisons between models and measurements are best evaluated by two approaches. One is the comparison of irradiance as a function of wavelength normalized to a certain wavelength-usually in the UV-A, where ozone has a negligible effect. This comparison emphasizes the response of irradiance to variables that are strongly wavelength-dependent (such as ozone absorption or wavelength error) and minimizes effects that are weakly wavelength-dependent, such as clouds, aerosols, or the absolute calibration of instrument responsivity. The second approach is the comparison of absolute irradiances at the wavelength of the normalization to quantify effects that are weakly wavelength-dependent. In any case, comparison of measured irradiances with modeled irradiances suffers from the limited availability and quality of data necessary to describe the atmosphere (Schwander et al., 1997).
Several studies have compared measured irradiances with model simulations. Measurements made under clear sky conditions and no snow cover demonstrate quite convincingly that lower ozone values result in higher UV irradiance levels at the surface of the Earth; these data sets have been used to quantify the dependence statistically (McKenzie et al., 1991; Booth and Madronich, 1994; Kerr et al., 1994). When these measured dependencies are compared with those computed for clear-sky conditions, no aerosols, and low ground reflectivity (no snow cover), reasonable agreement has generally been found (McKenzie et al., 1991; Wang and Lenoble, 1994; Forster et al., 1995).
The availability of aerosol optical depth measurements in the UV has allowed studies of the effects of particulates on ground-level irradiance. Mayer et al. (1997) compare clear-sky UV spectral data obtained at Garmish-Partenkirchen, Germany, between 1994-96 with model simulations. The model simulations use ozone and aerosol optical depth measurements as inputs. Systematic differences between measured irradiance spectra and model results were between -11% and +2%. It was necessary to introduce ground-level aerosols into the model to achieve agreement to within 5%. From total ozone, aerosol optical depth, and spectral UV irradiance measurements made under clear-sky conditions at Toronto between 1989-91, Kerr (1997) demonstrates that most of the observed variability of UV irradiance between 300 and 325 nm can be explained by ozone and aerosols. The remaining unexplained variability is 4% at 300 nm and 2% at 325 nm. Comparison of the observed dependence of UV irradiance on aerosol optical depth with model results suggests that typical aerosols over Toronto are slightly absorbing (Krotkov et al., 1998). The model also shows that a single scattering albedo of about 0.95 for aerosols gives the best agreement with the Toronto data.
Surface UV irradiance is also reduced by atmospheric sulfur dioxide (SO2), which has strong absorption features at UV wavelengths and occurs both naturally from volcanic emissions and anthropogenically from industrial sources (Zerefos, 1997; Kerr et al., 1998). The presence of SO2 can interfere with the measurement of ozone and estimates of ozone and UV trends at sites affected by local air pollution (Bais et al., 1993; De Meur and De Backer, 1993). However, measurements made at several sites in less-polluted situations suggest that the effects of SO2 on UV over wider areas are small (Fioletov et al., 1997).
A method developed recently to calculate surface spectral UV irradiance uses Total Ozone Mapping Spectrometer (TOMS) satellite measurements of ozone and UV reflectivity with a radiative transfer model (Eck et al., 1995; Herman et al., 1996; Krotkov et al., 1998). Comparison of model results with ground-based measurements made at Toronto under clear skies indicates agreement of absolute irradiance to about 2% after correction for the angular response of the ground-based instrument.
The effects of surface albedo have been considered in the UV-A (324 nm), where there is negligible ozone absorption, by observing the difference between measurements made with and without snow cover at several sites (Wardle et al., 1997). The presence of snow was found to enhance irradiance differently from one site to another. The minimum enhancement was 8% at Halifax, Canada; the maximum was 39% at Churchill, Canada. The difference between these two sites is likely to be a result of differences in the surrounding terrain and snow texture. For example, the clean snow on the flat terrain around Churchill would result in a higher average surface albedo than at Halifax, where snow would be dirtier in the suburban areas and not present on nearby open water. Model results show an enhancement of about 50% for an albedo of about 1 (Deguenther et al., 1998; Krotkov et al., 1998). Although there are no direct measurements of albedo available when snow is present, the model gives quite reasonable effective albedo values of about 20% at Halifax and 90% at Churchill. Model results of Deguenther et al. (1998) show that irradiance values are affected by surface albedo (snow cover) at distances up to 40 km; most of the dependence is influenced by albedo within a radius of 10 km.
Variability in cloud cover is the largest contributor to short-term changes in surface UV irradiance. It is possible to include the effects of clouds in radiative transfer calculations to various levels of approximation. However, routinely available observational data do not allow a rigorous characterization of cloud optical properties. Measurements show that UV spectral transmittance depends on cloud type, cloud thickness, and whether there are absorbers within the cloud. Although detailed quantification of these dependencies requires further research, some general conclusions can be made. The effects of thin clouds are weakly (<1% per nm) wavelength-dependent, with only broad wavelength features (Seckmeyer, 1989; Seckmeyer et al., 1996; Kylling et al., 1997; Mayer et al., 1998a,b). Under heavy convective clouds-when the amount of radiation is reduced by more than 90%-there is enhanced wavelength dependence as a result of increased absorption due to a longer pathlength through ozone within the cloud (Brewer and Kerr, 1973; Fioletov and Kerr, 1996). The effects of changes in stratospheric ozone on surface UV irradiance through all types of sky conditions (clear and cloudy) have been quantified from statistical analysis of data sets several years in length (Kerr and McElroy, 1993; Wardle et al., 1997). Algorithms that use ozone and reflectivity information from TOMS are able to include the effects of clouds in simulations of surface UV irradiance (Eck et al., 1995; Frederick and Erlick, 1995; Herman et al., 1996), although the results should be interpreted as averages over the large areas covered by the sensor's field of view.
Increased air traffic is expected to lead to changes in the abundances of ozone, NOx, SO2, and aerosols, as well as the frequency of cirrus clouds. In general, comparisons of observations with calculations have indicated that radiative transfer models can simulate the effects of gaseous absorbers quite reliably. Greater uncertainties are associated with the treatment of aerosols and cirrus because of the need to specify optical properties and perhaps fractional sky coverage. In the latter case, cirrus can lead to local increases in UV irradiance even though the area-averaged effect is a decrease. Models can simulate both non-absorbing (water or sulfate) and absorbing (carbon) aerosols, although the absorption properties of realistic aerosol types, which consist of mixtures of various chemical components, are not well known.
Other reports in this collection