CHANGES IN ULTRAVIOLET RADIATION
REACHING THE EARTH'S SURFACE
S. Madronich (USA), R. L. McKenzie (New Zealand),
M. M. Caldwell (USA), and L. O. Björn (Sweden)
Enhanced UV levels are clearly associated with the Antarctic springtime ozone reductions. Measurements show that maximum UV levels at the South Pole are reached well before the summer solstice, and DNA-damaging radiation at Palmer Station, Antarctica (64°S) during the springtime ozone depletion can exceed maximum summer values at San Diego, USA (32°N). UV increases at mid-latitudes are smaller. However, increases associated with the record low ozone column of 1992/93 in the northern hemisphere are evident when examined on a wavelength-specific basis.
Measurements in Argentina, Chile, New Zealand, and Australia show relatively high UV levels compared to corresponding northern hemispheric latitudes, with differences in both stratospheric ozone and tropospheric pollutants likely to be playing a role. Tropospheric ozone and aerosols can reduce global UV-B irradiances appreciably. At some locations, tropospheric pollution has increased since pre-industrial times, leading to decreases in surface UV radiation. However, recent trends in tropospheric pollution probably had only minor effects on UV trends relative to the effect of stratospheric ozone reductions.
Global ozone measurements from satellites over 1979/93 imply significant UV-B increases at high and midlatitudes of both hemispheres, but only small changes in the tropics. Such estimates however assume that cloud cover and tropospheric pollution have remained constant over these years. Under the current CFC phase-out schedules, global UV levels are predicted to peak around the turn of the century in association with peak loading of chlorine in the stratosphere and the concomitant ozone reductions. The recovery to pre-ozone depletion levels is expected to take place gradually over the next 50 years.
Fig.1.1. Increases in UV radiation in response to a 1 percent decrease in the total ozone column near 300 DU (1 DU = 2.69x1016 molec cm-2). Solid lines (right scale) give spectral irradiance changes, dotted lines (left scale) give percent changes. Values are for overhead sun (thick lines) and for a solar zenith angle of 70° (thin lines). From Madronich [1993c].
Accurate characterization of environmental UV radiation is difficult because of large geographical and temporal variations. However, significant progress has been made in recent years. Geographical and seasonal variations are now better understood in terms of earth-sun geometric factors, O3 amounts, cloudiness, various local and regional pollutants, and surface elevation and reflectivity. UV increases associated with recent O3 reductions have also been detected.
This update to the UNEP [1989, 1991] reports addresses the state of knowledge on environmental UV radiation as of August 1994, emphasizing the sensitivity to O3 changes for different biological and chemical photo-processes, the measurements of UV radiation at ground level and possible causes for its variations, and the implications of measured global O3 trends for UV radiation.
Various biological and chemical photo-processes respond differently to different parts of the UV spectrum. The relative effectiveness of different wavelengths must be known in order to assess the responses to O3 changes. The effective UV irradiance, E, or dose rate (exposure), is given by
where W() is the weighting function, or action spectrum, for a specific biological or chemical effect, and F() is the spectral irradiance, either computed or measured, for a given time and location. Hourly, daily, and yearly weighted doses may then be computed by time-integration of the dose rates.
The weighting procedure is required because solar UV-B increases steeply towards longer wavelengths, whereas the biological effectiveness often increases toward shorter wavelengths. As a result, weighted UV irradiances computed with different action spectra have different responses to atmospheric O3 changes. A commonly used measure of this dependence is the radiation amplification factor (RAF), defined by
where O3/O3 is the percent change in the ozone column, and E/E is the corresponding percent increase in weighted irradiance (instantaneous or an a time-integrated basis e.g., daily, yearly). The RAFs give the increase of available effective radiation in response to O3 reductions. They do not however measure the ultimate biological response, since this often depends non-linearly on the radiation exposure, and other factors, such as repair, time in the life cycle, whether the cell is dividing, etc., may also be important.
A compilation of RAFs for different biological and chemical processes is given in Table 1.1. The RAFs are useful for comparing the sensitivity of different processes to O3 changes. Action spectra that decrease strongly with increasing UV-B wavelengths have larger RAFs, while spectra with a significant UV-A tail tend to be less sensitive.
Several uncertainties in RAF values exist. In many cases, the original data used to derive action spectra are highly variable. For many spectra, insufficient data exist at the longer wavelengths, and if an exponential tail is used for extrapolation, the RAFs are significantly reduced (as indicated by the bracketed values in Table 1.1). The RAFs depend somewhat on the total O3 column, the altitude of the O3 perturbation, and the solar zenith angle (and therefore on latitude and season). They are, however, essentially independent of cloud cover, surface albedo, or local pollution.
The simple "percent rule" given above is valid only for small O3 changes. For large O3 changes, a more accurate relation is
where E1 and E2 are the weighted irradiances corresponding respectively to ozone columns (O3)1 and (O3)2. Figure 1.2 demonstrates the non-linear dependence of erythemal radiation on ozone reductions.
Fig. 1.2. Dependence of erythemally weighted UV radiation on O3 column changes. Measurements from South Pole, 1 February 1991 to 12 December 1992. Adapted from Booth and Madronich .
Table 1.1. Radiation Amplification Factors (RAFs) at 30°N.
|*Erythema Reference||1.1||1.2||McKinlay and Diffey, 1987|
|*Skin cancer in SKH-1 hairless mice (Utrecht)||1.5||1.4||deGruijl et al., 1993|
|*SKH-1 corrected for human skin transmission||1.2||1.1||deGruijl and van der Leun, 1994|
|Elastosis||1.1||1.2||Kligman and Sayre, 1991|
|Photocarcinogenesis, skin edema||1.6||1.5||Cole et al., 1986|
|Photocarcinogenesis (based on STSL)||1.5||1.4||Kelfkens et al., 1990|
|Photocarcinogenesis (based on PTR)||1.6||1.5||Kelfkens et al., 1990|
|Melanogenesis||1.7||1.6||Parrish et al., 1982|
|Erythema||1.7||1.7||Parrish et al., 1982|
|*Melanoma in fish||0.1||0.1||Setlow et al., 1993|
|Generalized DNA Damage||2.0||1.9||Setlow, 1974|
|Mutagenicity and Fibroblast Killing||[1.7]2.2||[2.7]2.0||Zolzer and Kiefer,1984;
Peak et al., 1984
|Fibroblast Killing||0.3||0.6||Keyse e al., 1993|
|Cyclobutane Pyrimidine Dimer formation||[2.0]2.4||[2.1]2.3||Chan et al., 1986|
|(6-4) photoproduct formation||[2.3]2.7||[2.3]2.5||Chan et al., 1986|
|HIV-1 activation||[0.1]4.4||[0.1]3.3||Stein et al., 1989|
|Damage to Cornea||1.2||1.1||Pitts et al., 1977|
|Damage to lens (cataract)||0.8||0.7||Pitts et al., 1977|
|Other effects on animal cells|
|*Occupational exposure limit||1.4||1.5||ACGIH, 1991|
|Immune Suppression||[0.4]1.0||[0.4]0.8||DeFabo and Noonan, 1983|
|*Cell mortality in Chinese hamster||Banrud et al., 1993|
|*Substrate binding in Chinese hamster||Banrud et al., 1993|
|Glycine leakage from E. Coli||0.2||0.2||Sharma and Jagger, 1979|
|Alanine leakage from E. Coli||0.4||0.4||Sharma and Jagger, 1979|
|Membrane bound K+-stimulated ATPase inactiv.||[0.3]2.1||[0.3]1.6||Imbrie and Murphy, 1982|
|Generalized plant spectrum||2.0||1.6||Caldwell et al., 1986|
|Inhibition of growth of cress seedlings||[3.6]2.8||3.0||Steinmetz and Wellmann, 1986|
|Isoflavonoid formation in bean||[0.1]2.7||[0.1]2.3||Wellmann, 1985|
|Inhibition of phytochrome induced anthocyanin
synthesis in mustard
|Anthocyanin formation in maize||0.2||0.2||Beggs and Wellmann, 1985|
|Anthocyanin formation in sorghum||1.0||0.9||Yatsuhashi et al., 1982|
|Photosynthetic electron transport||0.2||0.1||Jones and Kok, 1966|
|Overall photosynthesis in leaf of Rumex patientia||0.2||0.3||Rundel, 1983|
|*DNA damage in alfalfa||0.5||0.6||Quaite et al., 1992|
|Inhibition of motility (Euglena gracilis)||1.9||1.5||Häder and Worrest, 1991|
|*Inhibition of photosynthesis Phaeodactylum sp.||0.2||0.2||Cullen et al., 1992|
|*Inhibition of photosynthesis Prorocentrum micans||0.3||0.4||Cullen et al., 1992|
|*Inhibition of photosynthesis, in Antarctic community||0.8||0.8||Boucher and Prezelin, 1994|
|*Inhibition of photosynthesis
(Nodularia spumigena cyanobacteria
|0.2||0.2||Häder et al., 1994|
|*O3 + hnu --> O(1D) + O2||2.1||1.8||Madronich and Granier, 1994|
|*O3 + hnu --> O(3P) + O2||0.1||0.1||Madronich and Granier, 1994|
|*H2O2 + hnu --> OH + OH||0.4||0.4||Madronich and Granier, 1994|
|*HNO3 + hnu --> OH + NO2||1.1||1.0||Madronich and Granier, 1994|
|*NO2 + hnu --> O(3P) + NO||0.0||0.0||Madronich and Granier, 1994|
|*HCHO + hnu --> H + CHO||0.5||0.5||Madronich and Granier, 1994|
|*HCHO + hnu --> H2 + CO||0.2||0.2||Madronich and Granier, 1994|
|*CO production (Suwannee River)||0.3||0.3||Valentine and Zepp, 1993|
|*COS production (Gulf of Mexico)||0.2||0.2||Zepp and Andreae, 1994|
|*COS production (North Sea)||0.6||0.6||Zepp and Andreae, 1994|
|*Photodegradation of nitrate ions||1.1||1.0||Zepp et al., 1987|
|*Photodegradation of HCHO (Biscayne Bay)||1.3||1.1||Kieber et al., 19990|
|*Photoproduction of H2O2 in freshwater||0.1||0.1||Cooper et al., 1988|
|Yellowness induction in polyvinyl chloride||0.2||0.2||Andrady et al., 1989|
|Yellowness induction in polycarbonate||0.4||0.4||Andrady et al., 1989|
|Other weighting functions|
|Temple U. Robertson-Berger meter||0.8||0.7||Urbach et al., 1974|
|*Solar Light Robertson-Berger meter (Model 501)||1.2||1.1||M. Morys, priv. comm. 1994|
|*ozone cross section (273 K)||0.8||0.8|
|*UV-A (315-400 nm)||0.03||0.02|
|*UV-B (280-315 nm)||1.25||0.99|
|*UV-B (280-320 nm)||0.87||0.71|
|*simple exponential decay, one decade per 14 nm.||1.00||1.00|
Updated from UNEP (1991). (*) denotes change or new entry. Values in brackets show effect of extrapolating original data to 400 nm with an exponential tail, for cases where the effect is larger than 0.2 RAF units. RAFs computed on basis of daily integral.
Commonly used weighting functions are based on biological action spectra. Frequently a weighting function at the effects level is the composite of more than one spectrum at the molecular level, and is modified by absorbing molecules filtering the radiation before it reaches its target. Therefore the weighting functions, and data derived from them such as the RAFs of Table 1.1, depend on various independent factors and should be used with caution, taking into account the conditions under which the experiments were performed. When labels such as "DNA-effective radiation" are used, this simply describes the integrated irradiance potentially effective in causing effects (e.g., naked DNA damage) but does not necessarily mean that the effect (DNA damage) will ensue. The actual effect will depend on the sensitivity of the particular organism and several other factors.
Traditionally, action spectra have been developed for very different purposes than evaluating biological effects of O3 reductions. Action spectra allow the photobiologist to draw some conclusions regarding the biological pigment or molecule that absorbs the radiation and mediates the effect within an organism. The criteria often used to develop action spectra are directed to this traditional use in photobiology and these, along with many technical constraints, limit the usefulness of action spectra as weighting functions. A photobiologist may wish to know how an absorbing molecule is acting, with as little interference by other substances in the organism as possible, e.g., how DNA is absorbing radiation and acting to mediate an effect. However, for evaluating the biological consequences of O3 reduction it is more important to know how a molecule such as DNA in its normal state within the organism is affected by the radiation and how its effect may be altered by the activity of other absorbing molecules.
Action spectra are usually developed by exposing the biological material to radiation of only one wavelength (or a narrow range of wavelengths) at a time and then measuring the effect. Again, for traditional photobiological purposes this is quite suitable. However, organisms in nature are exposed simultaneously to radiation at all wavelengths in the entire solar spectrum reaching the earth's surface, and the radiation in the UV-A and visible wavebands is orders of magnitude more intense than in the UV-B. Under such conditions several chromophores may be acting to cause interacting effects in the organism. For example, the UV-A and visible radiation can ameliorate the effects of UV-B in many organisms [e.g., Caldwell and Flint, 1994]. For practical purposes, action spectra are usually developed with only hours of irradiation at each wavelength: whereas in nature organisms are usually exposed over periods of days, months or longer to the full sunlight spectrum. This too can cause interacting effects that might not be predicted from an action spectrum.
Another limitation of UV-B action spectra is that they may be arbitrarily restricted to a certain waveband where data were collected. While these still serve the traditional photobiological purposes, they can limit the usefulness for the O3 reduction question. For example, if data for action spectra are restricted to the UV-B, but in reality there is some, even very low, effectiveness in the UV-A, this can change the resulting radiation amplification factors when these action spectra are used as weighting functions.
These limitations and qualifications of weighting functions based on action spectra must be borne in mind when considering the significance of biologically effective radiation. Use of such weighting functions is still more appropriate than the use of unweighted UV-B radiation, but qualification is required. In this chapter, several weighting functions are employed in considering the effect of O3 reduction on the global distribution of "effective" UV radiation. While these are illustrative, they need to be interpreted with a knowledge of the uncertainties inherent in the weighting functions.
Measurements of environmental UV radiation still present some difficulty, especially for the detection of long-term trends since high accuracy and stability are required. Significant advances have been made recently in assessing data quality through instrument characterizations, intercomparisons, and data re-analysis.
Several intercomparisons among different spectro-radiometers showed substantial differences among instruments [Gardiner et al., 1993; McKenzie et al., 1993; Kirk et al., 1994], especially at the shortest wavelengths where the solar spectrum is steepest, and therefore problems of dynamic range, stray light rejection, and wavelength calibration are most severe. At the present time, agreement to no better than about +/-5 percent can be expected for wavelengths longer than ca. 310 nm, and the agreement is worse at shorter wavelengths.
Intercomparisons with broad-band and filter instruments are more difficult, due in part to calibration ambiguities that arise when the spectral shape of the solar spectrum changes under different solar zenith angle, O3 column, and other atmospheric conditions [DeLuisi and Harris, 1983]. Extensive re-examination has been carried out for the most commonly used broad-band instrument, the Robertson-Berger (RB) meter. Its temperature coefficient (ca. 1 percent K-1) has been determined [Johnsen and Moan, 1991; Dichter et al., 1993; Blumthaler, 1993], and a new generation of temperature-stabilized instruments is now available. The spectral response of the RB instruments was found to be stable over more than a decade, although with some differences between different instruments [DeLuisi et al., 1992]. A review of calibration records by Kennedy and Sharp  did not identify any significant problems. However, DeLuisi  found calibration shifts in the long-term data record of the RB meter located at Mauna Loa. The magnitude, timing, and direction of these shifts are such as to produce an apparent negative trend in UV comparable to the decreasing UV trends reported by Scotto et al.  for RB meters located in the continental USA over 1975-85. Smith and Ryan  have also identified substantial variations between different RB instruments. Until a full re-analysis of the calibrations of the RB meter network is carried out, trends derived from RB meters must be viewed with caution.
Very little UV irradiance data in the Antarctic are available before the discovery of the springtime O3 hole. Baker-Blocker et al.  reported measurements over 1979-81, but because their broad-band instrument was heavily weighted in the UV-A, no conclusions can be drawn about the pre-ozone hole UV-B record. The number of Antarctic UV measurements has increased greatly in recent years [Lubin and Frederick, 1989, 1991; Lubin et al., 1989, 1992; Stamnes et al., 1990, 1992; Frederick and Alberts, 1991; Smith et al., 1992a,b; Beaglehole and Carter, 1992a,b; Booth et al., 1993, 1994; Roy et al., 1994; Frederick and Lubin, 1994; Helbling et al., 1994]. The effect of the O3 hole on UV levels is now clearly established. Figure 1.3 shows the UV radiation in two different wavelength bands, 298-303 nm (very sensitive to O3) and 338-342 nm (relatively insensitive to O3), measured at the south pole between early 1991 and early 1994. While the 338-342 nm band maximizes near summer solstice as expected, the 298-303 nm values reach their highest values in November, and show clearly the effects of the O3 hole. Figure 1.4a shows that 1991 springtime DNA-damage-weighted UV radiation measured at the south pole was much higher than measurements obtained in Barrow, Alaska, for the same solar zenith angle. Visible radiation measurements at the same locations (Figure 1.4b) effectively demonstrate that the higher UV at the south pole was not due to other atmospheric factors such as less cloud cover. Although the natural UV levels at high latitudes are usually smaller than at low latitudes, measurements now show (see Figure 1.5) that during the springtime ozone depletion the DNA-damaging radiation at Palmer Station, Antarctica (64°S) can exceed maximum summer values at San Diego, USA (32°N).
Fig. 1.3. Hourly irradiance integrated over 298-303 nm (top) and over 338-342 nm (bottom), at the South Pole between 1991 and 1994. Vertical dotted lines mark the summer solstices. From Booth et al. .
Fig. 1.4. Comparison of 1991 springtime radiation at South Pole and Barrow (Alaska). (a) DNA-weighted UV irradiance; (b) 400-600 nm integrated irradiance. From Booth et al. .
Northern hemisphere polar regions experienced anomalously low O3 levels in 1992 and 1993 (see below). Spectral measurements obtained at Barrow during the spring of 1991-94 show the expected increases in UV radiation associated with low O3, and a return to more normal levels in 1994 [R. Booth, private communication, 1994].
Measurements at many locations are providing a more clear characterization of the geographical variations of ground-level UV. The RB meter network [Scotto et al., 1975, 1988; Cotton, 1990] has confirmed the general higher UV-B levels at lower latitudes in the US. Other RB meter measurements also confirm the broad latitudinal differences, e.g., in Russia [Garadzha and Nezval, 1987], Switzerland [Blumthaler and Ambach, 1990], Malaysia [Ilyas, 1987], and New Zealand [Zheng and Basher, 1993], though detailed comparisons of local effects (e.g., pollution, cloudiness) has not yet been carried out. Multifilter measurements at 39°N, 77°W have been made since 1975 by Correll et al.  and are continuing, but direct comparison to other instruments is difficult because of the different spectral response.
Fig. 1.5. DNA-weighted noon irradiances measured in 1993 at Palmer Station, Antarctica (64°S) and in San Diego, USA (32°N). (Courtesy R. Booth, 1994).
Spectral measurements show higher summertime values of UV-A and UV-B radiation in Lauder (New Zealand) and Melbourne (Australia) compared to Neuherberg (Germany) [Seckmeyer and McKenzie, 1992; McKenzie et al., 1993], due to the yearly cycle of the sun-earth distance, and to lower stratospheric O3 levels in the southern hemisphere and higher tropospheric pollutant levels (O3 and aerosols) in Germany. Enhanced southern hemisphere UV-B levels are also evident from measurements in Ushuaia, Argentina (55°S) [Diaz et al., 1991, 1994; Frederick et al., 1993b], where in December 1991 the average noontime clear sky radiation at 306.5 nm was 45 percent larger than calculated from the O3 climatology of the previous decade.
Additional mid-latitude spectral measurements [Bais et al., 1993, 1994; Gardiner et al., 1993; Blumthaler, 1993; Blumthaler et al., 1993, 1994; Seckmeyer et al., 1994; Booth et al., 1993; Ito, 1993; Kerr and McElroy, 1993; Kirk et al., 1994; Cabrera et al., 1994] have recently contributed to a growing UV data base, and systematic compilation of the data, though not yet achieved, should provide a reasonable picture of mid-latitude UV distributions.
The importance of clouds to surface UV is well established. Analysis of the long-term RB meter data from various locations in the US shows that monthly average UV levels are reduced by 10-50 percent, depending on season and location [Frederick and Snell, 1990; Frederick et al., 1993a]. Empirical parameterizations based on the fraction of occluded sky observed from the surface have been developed in Malaysia, the United States, Sweden, and Australia [Ilyas, 1987; Cutchis, 1980; Josefsson, 1986; Paltridge and Barton, 1978; Ito, 1993]. The results are scattered and non-linear, with as much as 70-80 percent reductions at full cover, and average reductions in the 10-50 percent range. The parameterizations should be viewed as highly approximate and applicable only to the cloud types that are characteristic of the different locations of the studies.
Tropospheric aerosols, e.g., sulfate, reduce UV levels significantly in polluted regions [Liu et al., 1991]. Seckmeyer and McKenzie  estimated that aerosols had only a minor effect in the contrast between UV irradiances in New Zealand and Germany. However, Seckmeyer et al.  measured substantial UV-A and UV-B reductions on a day with high turbidity, compared to a day with low turbidity. Cabrera et al.  found larger UV-A gradients with elevation in relatively polluted areas compared to pristine regions of Chile. Justus and Murphey  analyzed RB meter data over 1980-84 for Atlanta, Georgia, and attributed the observed 10 percent decrease to local or regional aerosols. Thus, although an effect appears to exist, its magnitude is still not well defined.
The effect of stratospheric sulfate aerosols on surface UV irradiance has been of increased interest since the eruption of Mt. Pinatubo in June 1991. Scattering of the incoming UV radiation by the aerosol may decrease surface irradiance at long wavelengths, but may also change the photon pathlengths through stratospheric O3, resulting in increased surface irradiance under some conditions (short wavelengths, large solar zenith angle [Michelangeli et al., 1992; Davies 1993]). The net effect on biologically weighted radiation is expected to be a relatively small decrease [Madronich et al., 1991; Vogelmann et al., 1992]. Spectral measurements show a marked increase in the diffuse/direct UV ratio but little effect on the total radiation [McKenzie, 1993; Blumthaler and Ambach, 1994]. Stratospheric aerosols also influence surface UV levels indirectly through their effects on stratospheric O3 chemistry.
Tropospheric Ozone and other Gaseous Pollutants
Brühl and Crutzen  suggested that tropospheric O3 may be a somewhat more effective absorber of UV radiation than stratospheric O3, due to enhancement of photon pathlengths by scattering in tropospheric air. Frederick et al. [1993a] found a negative correlation between surface O3 concentrations and RB meter readings in Chicago (USA). Cabrera et al.  found larger increases with surface elevation for UV-B than for UV-A, consistent with a larger tropospheric O3 column at the lower elevations. Other tropospheric gases (especially NO2 and SO2) may attenuate UV in some urban areas [Frederick et al., 1993a; Bais et al., 1993], but are probably not important in less polluted regions.
Surface Albedo and Elevation
Surface reflections affect UV radiation both through direct reflections toward a target, and by enhancing the diffuse down-welling radiation. The relatively sparse measurements of surface reflectivity (albedo) in the UV range have been reviewed recently by Madronich [1993a,b] and Blumthaler . Values usually fall below 10 percent for vegetation, but are highly variable for ice (7-75 percent) and snow (20-100 percent). High reflections may be of some importance to the geographical and seasonal UV distributions because they apply preferentially to colder climates.
UV levels are expected to be increase with increasing surface elevation above sea level because of the thinner overhead atmosphere. Measurements at remote locations in Chile show increases of 4-10 percent per km [Cabrera et al., 1994], in agreement with model calculations for unpolluted air [Madronich, 1993a]. Other locations show much larger vertical gradients, up to 40 percent per km near Santiago, Chile [Cabrera et al., 1994], and 9-23 percent per km in the Swiss Alps [Blumthaler, 1993], presumably because the lower elevations experience more tropospheric ozone, aerosol turbidity, and possibly lower surface albedo.
Several long-term records have been obtained using RB meters, and are subject to the cautions mentioned above. In Moscow, Garadzha and Nezval  found a 12 percent decrease in the RB meter measurements of UV radiation over 1968-83, with a concurrent 15 percent increase in turbidity and a 13 percent increase in cloudiness. RB meter measurements taken over 1974-85 at eight different sites in the USA showed UV decreases of 0.5 percent and 1.1 percent per year [Scotto et al., 1988]. RB meter data obtained at a station in the Swiss Alps (3.6 km above sea level, 47°N) showed increases of 0.7+/-0.2 percent per year over 1981-89 [Blumthaler and Ambach, 1990], persisting at 0.7+/-0.3 percent per year over 1981-91 [Blumthaler, 1993]. Zheng and Basher  have found increases of about 0.6 percent per year in RB meter data over 1981-90 in New Zealand, anticorrelated with O3 column data. Increases at high elevations and southern latitudes, together with possible decreases in industrialized northern hemisphere regions, are consistent with a role of local pollution, although RB meter calibration shifts may have also played a role (see above).
Multi-filter measurements by Correll et al.  over 1975-90 at a single site in Maryland (USA) indicate that the maximum monthly mean UV-B irradiance was 13 percent higher in 1983-89 than for the entire data record, with an overall increase of 35 percent from 1977/78 to 1985, much larger than expected from actual O3 reductions. However lower values were observed after 1987, and may be indicative of the role of the atmospheric factors such as cloud variability.
Kerr and McElroy  monitored spectral UV radiation in Toronto (Canada) from early 1989 through August 1993, a period during which ozone levels changed by -4.1 percent per year during winter (December-March) and by -1.8 percent per year in summer (May-August). The corresponding UV changes were strongly wavelength-dependent, as shown in Figure 1.6, with the greatest increments occurring at the shortest wavelengths as expected from ozone reductions. Using a similar instrument, Zerefos et al.  found statistically significant increases of +9.7 percent per year at 305 nm and +0.1% per year at 325 nm at Thessaloniki (Greece) between November 1990 and November 1993. These large increases span a relatively short time and are influenced by the anomalously low ozone of 1992/93, so that they are better interpreted as a perturbation rather than a trend [Michaels et al., 1994; Kerr and McElroy, 1994], but they demonstrate that ozone-induced changes in UV can be detected over a period of several years despite variability due to cloudiness and local pollution, particularly at the shortest wavelengths.
Seckmeyer et al.  found larger UV-B and simultaneously lower UV-A levels in 1993 relative to 1992 in Germany. The UV-A changes were attributed to the different average cloud cover, and the UV-B enhancements are consistent with independently measured lower O3 values, 322 Dobson Units (DU) in 1993 compared to 342 DU in 1992, May-July averages. Enhanced peak RB meter values, measured in Innsbruck (Austria) during 1993 winter/spring, were also found by Blumthaler et al. .
Limited evidence for long-term changes in the spectral distribution of surface UV radiation comes also from data collected by ground-based O3 monitoring networks (see above). Most of these determine O3 through a measurement of the ratio of UV at several wavelengths, either from the direct solar beam or the zenith sky. Long-term negative trends in O3 reported by such networks are therefore indicative of long-term shifts of surface UV towards shorter wavelengths.
Trend detection remains a problem due to the sparsity of reliable long-term data. It is likely that the opportunity to measure the historical natural UV baseline levels (i.e., pre-ozone depletion) has already been lost over most of the globe, and will not return until the ozone layer returns to its natural state.
The geographical coverage possible with radiative transfer models is limited only by the available atmospheric data which, if derived from satellites, can span the entire globe. Therefore, such models are an important complement to UV measurements. Models are also an essential aid in identifying the causes of observed UV changes, to carry out sensitivity studies, and ultimately to predict future UV environments under different atmospheric O3 reduction scenarios.
The theory of scattering and absorption of atmospheric radiation is well established, and there is no scientific doubt that, all other factors being held constant, O3 reductions are accompanied by predictable increases in surface UV radiation. However, models require as input the optical characteristics of the atmosphere, and if these are poorly known, as is often the case with pollutants and clouds, model-calculated irradiances can be in serious error.
For cloud-free and low-aerosol sky conditions and known O3 column, model irradiances generally fall within the experimental errors of both broad-band meters [e.g., Jokela, 1993] and spectroradiometers [e.g., McKenzie et al., 1993; Kirk et al., 1994; Zeng et al., 1994; Wang and Lenoble, 1994]. The theoretical relationship between O3 reductions and UV increases has also been confirmed in numerous studies [Stamnes et al., 1988, 1990, 1992; Roy et al., 1990, 1994; McKenzie et al., 1991; Smith et al., 1992b; Seckmeyer and McKenzie, 1992; Bais et al., 1993; Frederick et al., 1993b; Holm-Hansen et al., 1993].
Fig. 1.6. Impact of low ozone over Toronto, Canada in 1992/1993 compared with earlier years. The top panel shows the median daily UV spectral irradiance for the summers of 1989 and 1993, and the winters of 1989-90 and 1992-93. The middle panel shows spectral irradiance ratios for summer (1993 divided by 1989) and winter (1992-93 divided by 1989-90). The bottom panel compares the observed changes as a function of wavelength with the ozone absorption spectrum; the log of the winter ratio is used because the transmission of UV radiation depends on the exponent of the ozone absorption coefficient. Adapted from Kerr and McElroy .
Validation of the models is still a problem in the presence of clouds and tropospheric pollutants. Commonly used UV models are incompatible with the empirical formulations based on fractional cloud cover discussed in section above, either because they don't consider clouds at all, or idealize them as hori-zontal layers of homogeneous vertical optical depth that do not allow for partial coverage nor for realistic cloud morphology [Frederick and Lubin, 1988; Madronich 1990, 1993a,b]. Use of satellite images to estimate both areal coverage and optical depths is promising [Lubin et al., 1994; Gautier et al., 1994] but requires additional development and evaluation. Other highly variable factors such as aerosols and gaseous pollu-tants are seldom well characterized, which presents some difficulty for accurate modeling [Liu et al., 1991].
The recent changes observed in atmospheric O3 are described in the WMO  report and summarized only briefly here. Measurements of the total O3 column are made mainly by optical means from ground-based and satellite-based instruments. In addition, vertical profiles of O3 concentrations are made from balloon-borne instruments. The rather large data set has elucidated the geographical and seasonal distribution of O3. Ground-based Dobson instruments, developed in the 1920's, have increased in number and have been providing fairly wide geographic coverage since the 1960's. Other instruments, including the M-83 and M-124 filter ozonometer, the Brewer, and the SAOZ spectrometer, have shorter data records. True global coverage began in the late 1970's with the deployment of the Nimbus 7 satellite carrying the Total Ozone Mapping Spectrometer (TOMS) and the Solar Backscatter Ultraviolet spectrometer (SBUV). The SBUV instrument ceased functioning in June 1990 and the TOMS in May 1993. The Nimbus 7 TOMS data have been analyzed extensively, and the version 6 data appear to be reliable at least through May 1990, but there is concern about additional calibration drifts after that date. SBUV/2 (on the NOAA-11 satellite) was launched in January 1989 and another TOMS instrument was launched in August 1991 (on the Russian Meteor 3 satellite). Both of these instruments are still functioning. The combined SBUV and SBUV/2 data records appear to be suitable for trend determinations. The Meteor 3 TOMS data have not yet been properly assessed for consistency with the earlier systems. Other satellite O3 monitoring systems (TOVS, LIMS, SAGE I and II) are at the present time less suitable due to incomplete detection of the total O3 column, and some still unresolved calibration issues.
Statistically significant negative trends in total O3 are found at all latitudes except possibly in the tropics, where ground-based measurement show no trend. Recent Dobson trends are larger (more negative) than long-term trends, suggesting greater O3 reductions in the more recent years. The years 1992 and 1993, in particular, exhibited large negative anomalies in O3, possibly related to the eruption of Mt. Pinatubo, with record low values measured at northern mid-latitudes. Total O3 in early 1994 has largely returned to the trends observed before the Pinatubo eruption. The springtime Antarctic O3 loss, first observed in the late 1970's, has continued as an annual event, and record low values (for the first time less than 100 DU) were observed at the South Pole in late September and early October 1993. The 1992 and 1993 O3 holes appeared earlier and covered larger areas than those of earlier years.
Tropospheric O3 accounts for about 1/10th of the total O3, and tends to be highly variable, being associated with more polluted regions. Surface O3 concentrations in Europe may have increased 3-4 fold since last century, and doubled since the 1950's. However, surface values may not be representative of the free troposphere. Analysis of vertical profiles from O3 sondes indicates increases of about 10 percent per decade in the northern mid-latitudes, with strong regional patterns and largest increases over European and Japanese stations [WMO, 1994]. There is also some evidence for a slowing of the trends during the 1980's. In the southern hemisphere, surface O3 levels show little or no trends, except for the polar regions where a decrease of about 7 percent per decade has been reported.
The Nimbus-7 TOMS O3 data (version 6) over 1979/92 have been analyzed by Madronich and de Gruijl [1993, 1994] to infer the corresponding changes in UV radiation weighted by several biological action spectra. The TOMS data after May 1990 may be affected by a calibration drift which is not considered in the version 6 data. Here, the calculations of biologically weighted UV changes are repeated using the combined SBUV and SBUV/2 data, from January 1979 through December 1993.
The procedure to calculate trends in biologically effective doses is essentially unchanged. Briefly, the atmosphere is taken to be cloudless and aerosol-free, with 10 percent surface albedo, and standard vertical profile of O3 scaled to the total column from SBUV(/2). The down-welling spectral irradiance is calculated at 1 nm intervals (279.5-399.5 nm), then integrated over wavelength using several biological action spectra to compute dose rates. This is repeated every 15 minutes for the 15th day of each month of the O3 data record (January 1979-December 1993) in 10° latitude increments from 70°S to 70°N. The results are time-integrated to compute daily and yearly doses, and trends with their standard deviations are calculated by linear least-squares fitting.
Figure 1.7 shows the trends in daily doses for UV radiation weighted by the action spectrum for in vitro DNA damage [Setlow, 1974]. The results are generally similar to those obtained with TOMS data over 1979-89 [Madronich, 1992] and over 1979-92 [Madronich and de Gruijl, 1993, 1994]. Relative increases (percent per decade, Figure 1.7a) are significant in both hemispheres at middle and high latitudes, and are largest in winter and spring. The relative increases in the tropics are small and probably not significant. Daily dose increments (J m-2 day-1 per decade, Figure 1.7b) are shifted toward summer and lower latitudes, i.e., toward higher natural UV levels.
Table 1.2 summarizes the increases over the 15-year SBUV(/2) data record for the annual doses for DNA damage, erythema induction in humans [McKinlay and Diffey, 1987], and skin cancer induction in laboratory mice [de Gruijl et al., 1993] (see also Table 1.1). The largest increases are observed at high latitudes of the southern hemisphere. Significant increases are also found in the northern high latitudes, and the mid-latitudes of both hemispheres.
Some regions, particularly in the Northern Hemisphere, have experienced increased tropospheric pollution (mostly sulfate aerosol and ozone) during the last century. It has been estimated that the corresponding UV (DNA-weighted) could have been reduced by 6-18 percent from the sulfate aerosol increases [Liu et al., 1991] and by 3-10 percent from the tropospheric ozone increases [UNEP, 1991] in some industrialized regions. However, no direct information exists on pre-industrial stratospheric ozone, precluding accurate estimates of the net UV changes. More recent tropospheric ozone trends in industrialized regions are estimated to contribute at most 2 percent per decade to the DNA-weighted UV, compared to +5 to +11 percent per decade from mid-latitude ozone reductions [UNEP, 1991]. Sulfur emissions have recently decreased in some regions while increasing in others [NRC, 1986], and the corresponding UV changes are expected to reflect such local variations.
Fig. 1.7. Trends in daily DNA-damaging radiation based on SBUV(/2) total O3 measurements over 1979-93. Heavy shading indicates regions where trends differ from zero by less than one standard deviation (1s), light shading by more than 1s but less than 2s (a) Fractional trends, in percent per decade relative to 1979-93 average values; (b) energy trends, J m-2 day-1 per decade with action spectrum normalization at 300 nm.
Table 1.2. Annual UV exposures estimated from stratospheric ozone measurements between 1979 and 1993(a).
(a) Computed using monthly and zonally averaged ozone column data measured from SBUV and SBUV/2 satellite instruments between January 1979 and December 1993. Annual doses (only erythemal is shown) are 1979-93 averages. Percent changes are expressed relative to the 1979-93 annual averages. Uncertainties are one standard deviation.
Long term changes in cloud cover may also affect ground-level UV radiation. Surface observations at a few continental and marine locations suggest some long term cloudiness increases, but there is yet no confidence that such changes have occurred on global scales [IPCC, 1991]. The implications of such changes for surface UV levels have not been quantified. Global space-based observations of clouds [e.g., Rossow et al., 1991] are becoming available, but the record is still of insufficient length for trend detection.
With continued full compliance to the Montreal Protocol and its amendments, it is expected that stratospheric chlorine will peak around the year 1998, with a slow recovery over the subsequent 50 years [WMO, 1994]. Model simulations of stratospheric ozone indicate that the peak global ozone depletions will also occur in the next several years, with recovery over the next half-century. Relative to 1960, the maximum ozone depletion expected at northern mid-latitudes is 12-13 percent in winter/spring, and 6-7 percent in summer/fall. The depletion at southern mid-latitudes will be approximately 11 percent (all seasons). However, it should be noted that there are some differences among the predictions of different models, and between modeled and observed trends to date. Furthermore, model predictions may be altered significantly by events such as volcanic eruptions of magnitude comparable to that of Mt. Pinatubo.
The corresponding changes in biologically-weighted UV irradiances can be estimated using the RAFs in Table 1.1. The estimated UV increases for erythema induction and DNA damage are, respectively, 15-17 and 29-32 percent (northern hemisphere mid-latitudes, winter/spring), 8-9 and 12-15 percent (northern hemisphere mid-latitudes, summer/fall), and 15 and 25 percent (southern mid-latitudes, all seasons).