The link between springtime total ozone and summer UV radiation in Northern Hemisphere extratropics



[1] The link between stratospheric ozone decline and ultraviolet (UV) radiation increase at the Earth's surface is well established. In the Northern Hemisphere extratropics, stratospheric ozone is accumulated from autumn to spring as a result of transport from its source region in the tropics. The amount of accumulated ozone varies from year to year due to natural dynamical variability and chemical destruction by natural and anthropogenic substances. Observational and modeling studies show that these total ozone anomalies persist in the extratropics from spring to summer. Here we analyze time series of ground-based UV measurements and satellite retrievals of total ozone and UV radiation and demonstrate that there is a strong link between springtime total ozone and summer UV anomalies in the Northern Hemisphere extratropics. In some regions, the interannual variability in springtime ozone abundance explains 20–40% of the summer UV variability, and this relation can be used for improving seasonal UV forecasts. Using chemistry transport models, we estimate the influence of polar chemical ozone loss on the summer UV north of 35°N. We estimate that the massive Arctic ozone depletion 2011 increased the March–August cumulative erythemal clear-sky UV dose in the Northern Hemisphere extratropics by 3–4% compared to the climatology, with about 75% of the increase accumulated after the breakup of the polar vortex. This result strongly suggests that the effect of seasonal ozone anomaly persistence should be included in the assessment of the impacts of polar ozone losses.

1 Introduction

[2] Stratospheric ozone is the main absorber of ultraviolet (UV) radiation; its reduction leads to increase in UV radiation at the Earth's surface [Brasseur and Solomon, 1984; Kerr and McElroy, 1993; McKenzie et al., 1999; Diaz et al., 2003; de Laat et al., 2010]. An exposure to solar UV radiation leads to production of vitamin D in human skin [e.g., Engelsen et al., 2005]; however, excess exposure is associated with a number of adverse effects on human health, such as increased risks of skin cancer and cataracts, and also on terrestrial and aquatic ecosystems [Rousseaux et al., 1999; United Nations Environment Programme (UNEP), 2011]. The accumulation of man-made chemicals in the stratosphere during the twentieth century led to chemical ozone destruction, mainly in the polar regions due to their unique meteorology [e.g., Newman et al., 2007; Douglas et al., 2011], and to the decline of total ozone column globally. Thanks to the international efforts to stop the production and use of ozone-depleting substances (ODS), their atmospheric loading is slowly decreasing [e.g., Montzka et al., 2011]; however, the recovery of the ozone layer is not expected until at least mid-21st century [e.g., Eyring et al., 2007; Bekki et al., 2011].

[3] The current study is motivated by observations of anomalously high UV radiation levels at some Arctic UV stations during summer 2011 [Bernhard et al., 2011]. Unusually strong Arctic stratospheric polar vortex and low temperatures in winter-spring 2011 [Hurwitz et al., 2011] led to activation of ODS in heterogeneous reactions, followed by massive chemical ozone depletion [Manney et al., 2011; Sinnhuber et al., 2011]. Both weak transport and chemical loss contributed to anomalous low total ozone in early spring [Strahan et al., 2013]. Estimated ozone chemical loss reached 80% at 18–20 km by late March [Manney et al., 2011], and the estimated deficit of total ozone column due to chemical loss reached 47% at 80°N by early April [Adams et al., 2012]. Expectedly, anomalously low ozone was accompanied by enhanced UV radiation at the surface but, due to low Sun elevation at high latitudes, the UV radiation level remained relatively low in early spring [Bernhard et al., 2011]. However, a record-breaking level of UV radiation, with values falling into the high-exposure category according to World Meteorological Organization classification, was observed at some Arctic stations during summer 2011 [Bernhard et al., 2011]. For example, at Jokioinen (60.81°N, 23.50°E, see Table 1), the daily maximum UV index (UVI) reached the value of 7, while at the more northern site Sodankylä (67.40°N, 26.60°E, Table 1), the UVI reached the value of 6 for the first time in the observation record (since 1990). Before 2011, the maximum UVI values measured at these stations were 6 and 5 UVI, respectively. At high latitudes, summer UV increases are especially important, since this is the season when the dominant fraction of annual UV dose is received [Weatherhead et al., 2005].

Table 1. Ground-Based UV Stations Used in This Study
StationCountryCoordinatesInstrumentStart Year
  1. a

    Years 1996–1999 are missing.

  2. b

    Years 2005–2006 are missing.

AlertCanada82.50°N; 62.32°WBrewer spectrophotometer1995a
AndoyaNorway69.28°N;16.01°EMultiband filter radiometer GUV2000
BlindernNorway59.94°N; 10.72°EMultiband filter radiometer GUV1995
BergenNorway60.38°N; 5.33°EMultiband filter radiometer GUV1996
EurekaCanada79.99°N; 85.94°WBrewer spectrophotometer2001
FinseNorway60.60°N; 7.52°EMultiband filter radiometer GUV2003
JokioinenFinland60.81°N; 23.50°EBrewer spectrophotometer1995
KiseNorway60.78°N; 10.82°EMultiband filter radiometer GUV1996
LandvikNorway58.33°N; 8.52°EMultiband filter radiometer GUV1996
Ny-ÅlesundNorway78.92°N; 11.92°EMultiband filter radiometer GUV1995
ØsteråsNorway59.95°N; 10.60°EMultiband filter radiometer GUV1995
ResoluteCanada74.72°N; 94.98°WBrewer spectrophotometer1991b
SodankyläFinland67.40°N; 26.60°EBrewer spectrophotometer1990
TrondheimNorway63.42°N; 10.40°EMultiband filter radiometer GUV1996

[4] Observation of high UV radiation in summer following large springtime chemical ozone depletion triggered the question of whether the two events are related. A persistence of ozone anomalies from spring to late summer has previously been demonstrated in the middle and high latitudes of both hemispheres [Fioletov and Shepherd, 2003, 2005; Tegtmeier and Shepherd, 2007; Weber et al., 2011]. These studies have demonstrated that after the cessation of the extratropical wintertime ozone buildup associated with the polar vortex breakup in spring [Fusco and Salby, 1999], the evolution of the extratropical ozone is driven by a relaxation toward the photochemical equilibrium until the beginning of the next buildup season in autumn. During this period, ozone supply from the tropical source region is absent while chemical processes within the extratropical stratosphere, mainly the catalytic reactions with nitrogen oxides [Brühl and Crutzen, 2000], lead to a net chemical ozone loss. Brühl and Crutzen [2000] estimated that the area-averaged chemical ozone loss in the 100–10 hPa layer north of 30°N during May–August is 31 Dobson unit (DU), i.e., about 10% of the averaged total ozone column. Such a loss rate is not enough to completely eliminate the springtime ozone anomaly. As a result, the abundance of the extratropical total ozone in summer is strongly linked to the amount of ozone accumulated in the extratropics by spring. Because the ozone loss rates vary strongly with altitude, latitude, and month [Brühl and Crutzen, 2000; Fahey et al., 2000], one could expect that transport and mixing within the extratropics would affect the persistence of the anomalies. For example, Tegtmeier and Shepherd [2007] found that an increase of the vertical diffusion coefficient in their model led to a faster decay of ozone anomalies compared to the model with a smaller diffusion coefficient.

[5] Although seasonal persistence of ozone anomalies is well documented, it has not been studied whether the persistence of ozone anomalies can considerably affect summer UV radiation. In addition to total ozone column, UV radiation near the ground is affected by cloud cover, aerosols, and surface albedo [e.g., Fioletov et al., 2001; Lakkala et al., 2003; Weatherhead et al., 2005]. Interannual variability in these factors can strongly affect UV radiation and may mask the effect of ozone anomaly persistence. Orsolini et al. [2003] described several episodes of high UV radiation in the Arctic in summer 2000 which they linked with the large polar ozone depletion in spring 2000; however, no attempts were made to statistically analyze the influence of spring ozone on summer UV. In this study, we investigate the link between springtime ozone anomalies and summer UV in the Northern Hemisphere extratropics. Additionally, we address the more specific question of what is the impact of springtime chemical ozone depletion on the summer UV radiation and use the depletion of 2011 as a case study. By analyzing observations and model simulations, we show that (1) there is a strong link between the springtime ozone and summertime UV anomalies in the Arctic and in the Northern Hemisphere midlatitudes and (2) the anomalous UV radiation in summer 2011 would not have occurred without chemical ozone destruction by ODS during the preceding winter-spring.

2 Observational Data

[6] We analyze ground-based UV measurements and satellite total ozone and UV data. The analysis of observations is aimed to statistically demonstrate the link between springtime total ozone and summertime UV. Ground-based UV measurements are from Finnish, Canadian, and Norwegian national UV networks (Figure 1 and Table 1). The analysis of ground-based UV is focused on high latitudes, approximately north of 60°N, where the dominant part of the polar ozone depletion takes place in spring. These measurements are described in Lakkala et al. [2008], Fioletov et al. [2004], and Johnsen et al. [2008, 2011]. The Finnish and Canadian stations are equipped with the Brewer spectrophotometers that measure the spectral UV irradiance from 290 to 325 nm (286.5 to 365 nm at Jokioinen) with a sampling interval of 0.5 nm. The Brewer instruments are regularly calibrated using standard lamps traceable to the National Institute of Standards and Technology for the Canadian instruments and to the Aalto University, Finland, for the Finnish instruments. The calibrations are performed once every 1–2 years for the Canadian instruments and monthly on average for the Finnish instruments. The overall uncertainty of these Brewer instrument measurements is estimated to be 5–6%. The Norwegian stations are equipped with GUV multiband filter radiometers from Biospherical Instruments Inc. The irradiance scale is maintained with a traveling reference instrument, traceable to the QASUME (Quality Assurance of Solar Spectral Ultraviolet Irradiance Measurements carried out in Europe project) European reference spectroradiometer. Blind test intercomparisons of UV index measurements of the network GUV instrument in Ny-Ålesund and the network traveling standard GUV in Oslo have shown close agreement with the QASUME unit [Gröbner et al., 2010]; see also materials available from the World Radiation Center ( The expanded uncertainty (coverage factor k = 2) of the GUV's UVI measurements is 6% [Aalerud and Johnsen, 2006].

Figure 1.

Locations of the ground-based UV stations.

[7] In this study, we analyze the noontime UVI derived from the measurements. The UVI is defined as the erythemal UV dose rate expressed in mW/m2 and divided by 25.

[8] In order to get a global view of the total ozone and UV radiation, we use satellite measurements from Total Ozone Mapping Spectrometer (TOMS), Ozone Monitoring Instrument (OMI), and Solar Backscatter Ultraviolet SBUV/2 (SBUV). Both total ozone and UVI estimates are available from TOMS [McPeters et al., 1998] for the period 1979–2001 with gap in 1993–1996 and OMI [Levelt et al., 2006] for the period 2005–2011. As advised by the data providers, the TOMS data after 2001 are not used because of the calibration problems with the instrument ( The total ozone and UV measurements from these two instruments are analyzed in conjunction with each other. TOMS and OMI use similar algorithms for total ozone retrieval [Bhartia and Wellemeyer, 2002]. Here we use TOMS version 8 data and OMI collection 3 data, which is based on the TOMS version 8.5 algorithm. The algorithm versions 8 and 8.5 differ in the way they calculate the amount of ozone under the clouds and this leads to TOMS total ozone being about 2% higher than OMI under cloudy conditions (OMI TOMS ozone Readme file, available from TOMS and OMI also use similar algorithms for the UVI retrieval, which include the estimation of clear-sky UV irradiance and its correction for scattering by clouds and aerosols as well as for surface albedo. The differences between the two algorithms include the aerosol absorption, which is not taken into account in the OMI algorithm, and the estimation of surface albedo [Tanskanen et al., 2006]. The lack of aerosol absorption and different estimated surface albedo in snow-covered areas lead to larger UVI estimates by OMI [e.g. Ialongo et al., 2011]. Due to the differences in the algorithm, we analyze here the anomalies with respect to seasonal climatologies, which are calculated separately for TOMS and OMI, as discussed in section 4. In addition to the noontime UVI, we analyze the cumulative erythemal dose (expressed in J/m2), which is available from OMI data.

[9] We also use satellite total ozone data to complement the analysis of ground-based UV measurements. Because of the gaps in the TOMS measurements, we use the merged total ozone data set compiled from TOMS, OMI, and SBUV [Bhartia et al., 1996, 2004] measurements as described in Stolarski and Frith [2006]. The time series from 1995, the first year with sufficient ground UV measurements, to 2011 are used in the analysis.

3 Models

[10] Two chemistry transport models (CTMs) are used in this study to estimate the impact of polar chemical ozone depletion on summer UV. By getting consistent estimates from two independent models, we increase our confidence in the results.

[11] The first model is FinROSE [Damski et al., 2007], which includes 38 species with about 120 homogeneous reactions, 30 photodissociation processes, heterogeneous chemistry, as well as formation and sedimentation of polar stratospheric clouds. In this study, FinROSE is run with 35 vertical levels, a horizontal resolution of 6° × 3° in latitude and longitude and is driven by meteorological fields from ERA-Interim [Dee et al., 2011]. The model is run with full chemistry for the period from 1 January 1989 to 31 December 2011. The first 2 years of model run are considered spin-up and discarded. In order to isolate the effect of springtime chemical ozone destruction, another simulation is run, which is branched off the full chemistry simulation on 1 September 2010 and run until 31 December 2011. In this experiment, the formation of polar stratospheric clouds (PSCs) is switched off by setting the air temperature passed to the heterogeneous chemistry module of the model to 200 K. This setting has little influence on the reactions on the background aerosols, but prohibits formation of ternary liquids, nitric acid trihydrate (NAT), and ice PSCs, and thus prevents the activation of ozone-depleting halogen species and thus the chemical ozone destruction associated with it. In general, FinROSE has a negative total ozone column bias with respect to observations (about 10% at 60°N–90°N). The bias is stable through the time, and therefore, the model anomalies are comparable to those in observations.

[12] The second model is ATLAS, which is a global CTM based on a Lagrangian (trajectory-based) approach. A detailed description of the model can be found in Wohltmann and Rex [2009] and Wohltmann et al. [2010]. The model includes a gas-phase stratospheric chemistry module, heterogeneous chemistry on PSCs, and a particle-based Lagrangian denitrification module. The chemistry module comprises 47 active species and more than 180 reactions. Transport and chemistry in the model are driven by ERA-Interim reanalysis data. The model setup for the options of the chemistry scheme and microphysics is taken from the reference run of Wohltmann et al. [2012], except for the chemical initialization, which is based mainly on Microwave Limb Sounder data of December 2010. A similar set of two simulations—with and without heterogeneous chemical reactions in the polar vortex—is performed with ATLAS for the period from 1 December 2010 to 31 August 2011. In this experiment, the polar heterogeneous chemistry is switched off by setting the reaction rates for the key reactions ClONO2 + HCl/H2O and HOCl + HCl to zero. This procedure is expected to have similar results to that implemented in FinROSE except that in this method, the reactions rates are changed globally, while in the FinROSE, only the cold vortex area (with T < 200 K) is influenced. However, because of the large decrease of the reaction rates with increasing temperature, this difference has a negligible effect. The similarity of the two approaches was confirmed by a sensitivity run with FinROSE in which polar heterogeneous chemistry was switched off in the same way as in the ATLAS run. Only the simulation from December 2010 to August 2011 is available for ATLAS.

[13] Both CTMs are focused on stratospheric chemistry; therefore, only stratospheric ozone column is available as a model output. Total ozone column at each grid point and each day needed for UV calculations is calculated by adding the tropospheric ozone column taken from ERA-Interim. ATLAS only calculates ozone column between the 350 K and 1900 K isentropic levels; therefore, the ERA-Interim ozone column above 1900 K is also added to the ATLAS outputs. Only clear-sky UV estimates based on solar zenith angle and ozone column information [Allaart et al., 2004] are available from the models. Since the influence of ozone depletion on UV is estimated here as a difference between two CTM runs with identical meteorology including cloud distribution, the use of cloud-corrected values would not change the CTM-based estimates qualitatively. The estimations based on the cloud-corrected UVI values would, however, be somewhat smaller because the cloud-correction procedure is multiplicative. Both noontime UVI and cumulative erythemal dose are calculated from the model outputs.

4 Methods

[14] In this study, we use anomalies calculated with respect to the climatological mean seasonal cycles. For ground-based UV measurements, both daily and monthly mean UVI anomalies are analyzed. The climatologies are calculated for the periods between the start of the measurements, which differ between the stations (see Table 1), and until 2010. Because the distributions of noontime UVI have a negative skew corresponding to cloudy days, i.e., the distributions are not normal, the monthly or seasonal averages of the noontime UVI anomalies do not necessarily provide sufficient information about the UVI during the analyzed period. Therefore, we perform a more detailed assessment of the distributions. In addition to the mean values, other parameters of climatological distributions of noontime UVI are also estimated for each station and for each Julian day separately and over the same periods as the climatological mean values. For example, the 90th percentile for 1 June at a particular station is defined as the noontime UVI value which was not exceeded at 1 June in 90% of years for which measurements are available until 2010. Similarly, the maximum UVI value for 1 June is defined as the maximum value measured at the station on 1 June until 2010. We next define the threshold exceeding fraction, EF, as

display math

and N is the number of measurement days during the analyzed season (i.e., during June–August in a particular year). EF shows the relative number of days when a certain UVI threshold was exceeded. The 50th, 75th, 90th, and 95th percentiles and the maximum values from the daily noontime UVI distributions are used as the thresholds in the EF calculations.

[15] For satellite data, the monthly total ozone climatology is calculated for the period 1979–2010 using the available data from both TOMS and OMI. For satellite UVI, the monthly climatology is calculated separately for TOMS (1979–2001) and OMI (2005–2010) periods. By calculating the UVI climatologies separately, we aim to avoid the discontinuity in the time series between the two instruments. (Without this procedure, the time series would be biased high during the OMI period, as explained in section 2.) We consider that this homogenization is sufficient for our purposes and use the anomaly time series combined for the whole period. This is equivalent to the assumption that any remaining differences between the two periods are small compared to the interannual variability.

[16] We focus the analysis of satellite data on area-averaged diagnostics. Following Fioletov and Shepherd [2003, 2005], we study the area north of 35°N and analyze area weighted total ozone and UVI anomalies for both midlatitudes (35°N–60°N) and polar latitudes (north of 60°N). Since the satellite measurements stop at the terminator, the northern boundary of the polar averaging area varies between 59°N in December (thus, the December polar means are not calculated) and 88°N in the summer months. When averaging over latitudinal belts, attention is paid to the fact that measurement coverage varies from year to year. Therefore, a screening procedure is applied to ensure that for each Julian day, the same areas are used for the averaging every year. The procedure is especially important for UVI which varies strongly with latitudes, and the differences in area coverage can lead to spurious interannual variability in the area-averaged diagnostics. When model outputs are compared with observations, they are first interpolated onto the observational grid and then masked with the observation coverage before calculating area averages.

5 Results

5.1 The Link Between Spring Ozone and Summer UV

[17] In this section, we examine the link between springtime ozone and summer UV radiation. We first analyze ground-based UV measurements from the Arctic stations and focus on summertime (June–August) UV radiation in 2011, when the Arctic ozone loss was exceptionally high. The analyzed stations are listed in Table 1. Their locations are shown in Figure 1. Daily noontime UVI anomalies for the stations are shown in Figure 2. The daily anomalies in summer 2011 were mainly positive at the majority of the stations; however, when averaged monthly (Figure 2, cyan lines) or over the summer (Table 2), the anomalies are both positive and negative. The summer averaged anomalies are positive at 8 out of 14 stations. When averaged over all stations, the relative summer UVI anomaly in 2011 is 2%.

Figure 2.

Noontime UVI anomaly for 2011 (red dots) calculated with respect to the climatological mean seasonal cycle. For each station, the climatology is calculated for the period from the first year of observations to 2010. The dark shading marks the 25–75% interval, and the light shading mark the min-max interval of noontime UVI calculated for each Julian day and for the same period as the climatology. The cyan lines show the monthly mean anomalies for 2011. Note the different UVI scales for the panels.

Table 2. Summer (June–August) 2011 Averaged Noontime UVI Anomaly and Exceeding of Noontime UVI Thresholds at the Ground-Based Stationsa
StationN Measurement Days in Summer 2011Summer 2011 Averaged Anomaly (%)Percentage of Summer Days in 2011 When Nth Percentile (or Maximum) Value Was Exceeded
  1. a

    For each station, the percentiles and the maximum values are calculated over the observation period until 2010.

Average 2.358.036.419.112.810.5

[18] Table 2 also shows the EF statistics for summer 2011. It shows that the number of days with relatively high or very high UVI in summer 2011 was large and higher than what could be expected to occur by chance at all but two stations. For example, the 90th percentile of the noontime UVI distributions at Kise was exceeded in ~40% days during June–August 2011. By definition, the 90th percentile is expected to be exceeded by chance during a season in about 10% of days only.

[19] The summer UVI at the two exceptional stations—Alert and Finse—is modulated by variations in snow cover, which strongly enhances UV radiation. Alert is the most northern analyzed station (82°N) at which snow often lasts through the end of June (not shown). Snow cover can increase the ratio between measured UV and estimated clear-sky UV above unity [e.g., Fioletov et al., 2001]. Observations of such high ratio values at 324 nm, where ozone absorption is low, usually indicate the presence of snow. In summer 2011, high values (>1.2) of the ratio between measured and estimated clear-sky UV were observed at Alert until mid-June only and decreased afterward (not shown), suggesting an early snowmelt in 2011. Consistent with this explanation, the daily UVI anomalies at Alert were negative mainly during June 2011 but positive afterward (Figure 21). Finse, the other station with low UVI anomaly in summer 2011, is the mountain station (1200 m above sea level) located near a glacier. At this station, low snow albedo was observed in August 2011 (not shown), also suggesting the lack of snow-reflected UV radiation as the explanation for low UVI anomaly in summer 2011.

[20] Overall, the case study of 2011 not only showed that the number of summer days with anomalously high noontime UVI was large at the majority of Arctic stations but also demonstrated the importance of other factors such as snow cover.

[21] We next look at the interannual variability of the summer UV radiation using the ground-based measurements and relate it to springtime ozone anomalies. Figure 3 shows the time series of June–August mean noontime UVI and EF statistics averaged across all stations. Also shown in Figure 3 are the March total ozone anomalies from the merged data set area averaged north of 35°N. The averaging north of 35°N is used following Fioletov and Shepherd [2005] who showed that the total ozone amount accumulated by spring over the extratropics (over 35°N–80°N in their study) is a better predictor of summertime ozone both in midlatitudes and polar latitudes than springtime ozone in either midlatitudes or polar latitudes. After vortex breakup, ozone is well mixed through the extratropics, and therefore, springtime ozone in both midlatitudes and polar latitudes influences summertime ozone everywhere in the extratropics. However, before the breakup, the correlation between ozone in midlatitudes and polar latitudes is weak, and therefore, neither of them separately is the optimal predictor of summertime ozone. March total ozone anomalies are used as a predictor of summertime UVI because this is the month of the annual maximum in the extratropical total ozone. Given the large correlation between March and April total ozone anomalies [e.g., Fioletov and Shepherd, 2003; Tegtmeier and Shepherd, 2007], the use of the April anomalies would not qualitatively modify our results. The time series start in 1995, prior which the UV data are only available from two stations. High covariability between the summer UV and springtime ozone is evident. In particular, the EFs for the 75th and 90th percentiles in summer 2011 were the highest during the analyzed period. The highest summer mean noontime UVI anomaly was in 1997, another year with very low springtime ozone [e.g., Newman et al., 1997; Fioletov et al., 1997]. The correlation coefficients between the March total ozone anomalies and summer UVI statistics vary between −0.63 and −0.66 and are statistically significant according to a two-sided t test with p = 0.01.

Figure 3.

The time series of June-August UVI threshold exceeding fractions (%, upper left axis) and mean noontime UVI anomalies (%, lower left axis) averaged across the Arctic stations. Black diamonds and squares show the exceeding fractions for the 90th and 75th percentiles, respectively. Black circles show the mean summer relative UVI anomalies. Red circles show the 35–90°N March total ozone anomaly time series from the merged ozone data set (DU, right axis). The numbers at the upper axis show the number of UV stations used to make the respective year average. Note the reversed axis for the total ozone.

[22] To demonstrate the link between springtime ozone anomalies and summer UV for larger areas and over longer periods, we analyze the satellite measurements. Figure 4 shows the seasonal evolution of the total ozone and UVI anomalies for the polar latitudes (>60°N) and midlatitudes (35°N–60°N) for the period 1979–2011. Since one of this study's motivations is to quantify the impact of chemical ozone losses on summer UV, we highlight years with large chemical ozone loss, although in principle, the ozone anomaly persistence does not depend on how the springtime anomalies are created [Tegtmeier and Shepherd, 2007]. In year 2011, low ozone values persisted from spring to summer both in the polar latitudes and midlatitudes. The UVI anomalies in 2011 were record high in midlatitudes in July and September. According to cloud optical thickness measurements available from the OMI, the cloud cover in summer 2011 was not anomalously small (not shown) and therefore could not alone explain the large UVI anomalies. There was a temporal weakening of the anomalies in April–May in the polar region and in May–June in midlatitudes. Inspection of Figure 4 also reveals a month-to-month variations in the relative strength of the anomalies during some other years. Such variability may be expected due to changing meteorological conditions, which contribute a noise to the anomaly persistence. For example, in April 2011, the weakening of the polar anomaly coincided with the strengthening of the midlatitude anomaly, which suggests that this was associated with the shift of the polar vortex to midlatitudes and that the ozone-poor air mass was still confined within the vortex and not well mixed with the surroundings.

Figure 4.

Seasonal evolution of TOMS/OMI satellite-derived monthly mean total ozone column anomalies averaged over (a) 60°N–90°N and (b) 35°N–60°N for each year between 1979 and 2011 when the measurements were available. (c, d) The same as in Figures 4a and 4b but for monthly mean cloud-corrected noontime UV index anomaly. Red, blue, green, and orange colors highlight the 2011, 2005, 2000, and 1997 years, respectively.

[23] The years with the largest Arctic springtime chemical ozone loss, apart from 2011, are 1996, 2000, and 2005 [Manney et al., 2011; Rex et al., 2006]. Due to the lack of data, the year 1996, as well as the years 1993 and 1995 when ozone depletion was also large, is not shown. In both 2000 and 2005, the summer UVI anomalies were usually positive, although in year 2005, the anomaly was not particularly large, partly because the spring total ozone was only moderately low due to strong mixing across the vortex edge and also an early final warming [Manney et al., 2006]. In 2000, the UVI was record high in the polar latitudes in June and in midlatitudes in July. The Arctic springtime ozone was exceptionally low also in 1997 due to reduced transport despite only moderate chemical ozone loss in that year [Manney et al., 1997]. The UVI in summer 1997 was record high in the polar latitudes in June and in August–September. These results indicate that transport and mixing played a role comparable to that of chemical loss in determining the variability in the Arctic springtime total ozone amount during recent years [Tegtmeier et al., 2008].

[24] Figure 5a shows the correlation function between the March total ozone anomalies north of 35°N and total ozone and UVI anomalies in the following months in the polar latitudes and midlatitudes. A statistically significant correlation persists between March total ozone and total ozone in the following months until autumn. The same effect has previously been noted by Fioletov and Shepherd [2003, 2005] and Tegtmeier and Shepherd [2007] in the merged TOMS/SBUV time series, by Tegtmeier and Shepherd [2007] in chemistry-climate model (CCM) simulations, and by Weber et al. [2011] in the merged GOME-1/SCIAMACHY/GOME-2 time series based on the Global Ozone Monitoring Experiment (GOME) 1 and 2 and Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) data. Here, we extend their results by showing that spring total ozone also influences the UV radiation during all months from spring through autumn. Figures 5b–5c show the scatterplots of noontime UVI averaged over June–August and over polar latitudes (north of 60°N) and midlatitudes (35°N–60°N) against March monthly mean extratropical (north of 35°N) total ozone. High correlation coefficients in both latitudinal belts indicate that a significant fraction of the observed summer UVI variability in the Northern Hemisphere extratropics is related to the variability of springtime total ozone.

Figure 5.

(a) Correlation functions between the TOMS/OMI March extratropical (north of 35°N) total ozone and monthly mean total ozone and noontime UVI in polar latitudes (north of 60°N) and midlatitudes (35–60°N) from March to December. Note that for UVI, the absolute value of the correlation coefficient is plotted (red lines). Scatterplots of the June–August noontime UVI for (b) polar latitudes and (c) midlatitudes versus March extratropical total ozone. Also shown in Figures 5b and 5c are the correlation coefficients between the spring ozone and the summer UV.

[25] The relationship with spring total ozone suggests a possibility of improving seasonal UV radiation forecasts. To elaborate this possibility further, we show in Figure 6 the map of correlation coefficients between March extratropical total ozone and June–August mean noontime UVI at each grid point. The correlation coefficients are, in general, smaller than those in Figures 5b–5c because at the smaller scales, the influence of local factors such as clouds or snow cover (in high latitudes) becomes more important and masks the effect of seasonal ozone anomaly persistence. Nevertheless, areas with correlations coefficients exceeding in absolute values 0.5 are evident, for example, in Eastern Europe, Mediterranean, and southern and southwestern USA. In these areas, the influence of springtime total ozone on summer UVI is considerable, likely because these regions have fewer clouds and therefore are more exposed to clear-sky conditions. The coefficient of determination [e.g., von Storch and Zwiers, 1999, p. 151] indicates that in these regions, 20–40% of the interannual variability in the summer UV radiation is related to the variability in spring ozone. For such regions, summer UV forecasts including springtime extratropical total ozone content may be more skillful.

Figure 6.

Correlation coefficients between the TOMS/OMI March extratropical total ozone and June–August noontime UVI.

5.2 The Impact of Arctic Chemical Ozone Loss on Summer UV

[26] We next estimate the influence of springtime ozone depletion on the seasonal evolution of ozone layer and UV radiation by using chemistry transport model simulations. Figure 7 demonstrates the ability of CTM FinROSE to reproduce the seasonal persistence of ozone anomalies and their influence on clear-sky UV radiation. The figure shows the same diagnostics as those in Figure 5, except that clear-sky UVI is used instead of cloud-corrected UVI because the latter is not available from the model. Comparison of Figures 5 and 7 shows that in some years, springtime ozone anomalies simulated by FinROSE deviate from the observations. However, the overall agreement with observations is fairly good, and the correlation coefficient between the observed and simulated springtime ozone anomalies during the overlapped years is 0.83. FinROSE is capable of simulating seasonal persistence of ozone anomalies; however, the simulated anomalies decay faster than the observed ones. A similar effect was noted by Tegtmeier and Shepherd [2007] in a version of their model with a strong vertical diffusion, suggesting that an excessive mixing maybe the reason for the faster anomaly decay in FinROSE. Nevertheless, Figure 7 suggests that FinROSE can be used for estimating the influence of springtime ozone on summer clear-sky UV. Faster anomaly decay in the model implies that the negative ozone anomaly caused by chemical ozone loss would disappear faster, and its influence on the seasonal UV radiation would be underestimated compared to the real atmosphere.

Figure 7.

(a) Correlation functions between the FinROSE March extratropical total ozone and monthly mean polar latitudes (north of 60°N) and midlatitudes (35°N–60°N) total ozone and noontime clear-sky UVI from March to December. Note that for UVI, the absolute value of the correlation coefficient is plotted (red lines). Scatterplots of the June–August noontime clear-sky UVI for (b) polar latitudes and (c) midlatitudes versus March extratropical total ozone.

[27] Figure 8 shows the seasonal evolution of simulated and observed ozone anomalies in 2011 calculated with respect to the period 1991–2010. The FinROSE run with full chemistry reproduces the ozone minimum in March in the polar latitudes (Figure 8a) and a negative ozone anomaly through summer in polar latitudes and midlatitudes (Figure 8a-b). Note that in midlatitudes, the largest ozone anomaly was observed in April, when the polar vortex containing the ozone-depleted air shifted to midlatitudes and broke down [Manney et al., 2011; Hurwitz et al., 2011]. The FinROSE run with full chemistry captures this anomaly well. When heterogeneous polar ozone loss reactions are excluded in the other experiment, the model does not capture the magnitude of the observed polar ozone negative anomaly in spring, and the evolution of total ozone anomalies in the following months is remarkably different from the observed one. In this experiment, the simulated ozone anomaly is also negative in late winter–early spring, which is consistent with a reduced poleward ozone transport by the meridional circulation [Fusco and Salby, 1999; Nikulin and Karpechko, 2005; Strahan et al., 2013], reduced mixing through the vortex edge [Manney et al., 2011; Strahan et al., 2013], and also an elevated tropopause which contributed to the negative Arctic ozone anomaly in February [Manney et al., 2011]. However, the magnitude of the March polar ozone anomaly is only 58% of that in the full chemistry run. After the shift of the polar vortex to midlatitudes in April followed by the vortex breakup [Hurwitz et al., 2011; Manney et al., 2011], the polar ozone anomaly becomes positive and remains so until the end of summer, indicating that the anomalously low ozone in summer 2011 can only be reproduced when springtime ozone depletion is taken into account. The midlatitude anomalies in this run remain small through the simulation (Figure 8b).

Figure 8.

Seasonal evolution of monthly mean total ozone anomaly in 2011 averaged over (a) 60°N–90°N and (b) 35°N–60°N. The anomalies are derived from TOMS/OMI retrievals (red line) and FinROSE model simulations with full chemistry (black line/circles) and without heterogeneous chemistry (black line/stars). (c, d) The same as in Figures 8a and 8b but for clear-sky UVI.

[28] The observed and simulated evolution of clear-sky UVI anomalies in 2011 (Figures 8c–8d) is consistent with that of total ozone, but some difference are to be expected since the UV anomalies are calculated with respect to the OMI period (2005–2010) when the observed clear-sky UVI is available. The FinROSE run with full chemistry reproduces the positive UV anomaly in summer both in polar latitudes and midlatitudes. In the experiment without heterogeneous polar ozone loss, the UVI anomalies in summer are mostly negative in both latitudinal bands, consistent with the positive ozone anomalies.

[29] The difference between the two model simulations can be used to estimate the ozone deficit and UV radiation excess due to polar ozone loss. Figure 9 shows these diagnostics estimated from the FinROSE and ATLAS runs. First, we note a remarkable agreement between the two models, which increases the confidence in provided estimations. The deficit of ozone column peaks in the polar latitudes in March and in midlatitudes in April and then decreases toward the end of summer due to photochemistry [Tegtmeier and Shepherd, 2007]. The UV radiation excess peaks in the polar latitudes in June and in midlatitudes in May. This delay reflects the fact that the increase of Sun elevation from spring to summer has a larger effect on the UV radiation than the reduction of ozone deficit does.

Figure 9.

(a) Seasonal evolution of monthly mean total ozone deficit in 2011 north of 60°N (dashed) and over 35°N–60°N (solid) estimated from FinROSE (circles) and ATLAS (triangles). (b) The same as in Figure 9a but for clear-sky UVI excess.

[30] Finally, we estimate the influence of polar ozone loss on the cumulative seasonal erythemal clear-sky UV radiation dose. Figure 10 shows the latitudinal distribution of cumulative erythemal UV radiation dose excess due to ozone loss as estimated by FinROSE and ATLAS. The agreement between the two models is good although the ATLAS-estimated dose excess is smaller than that from FinROSE, with maximum difference of ~20% south of 45°N. The cumulative seasonal (March–August) UV dose excess estimated from FinROSE reaches 3–4% of the climatological UV dose. Figure 10 also demonstrates that the estimated UV dose excess is comparable in magnitude with the satellite-retrieved March–August erythemal UV dose anomaly for 2011. Since the observed extratropical UV anomaly in 2011 was one of the largest during the satellite period (see Figure 5), this result suggests that the springtime polar ozone loss can strongly modulate the interannual variations in cumulative seasonal UV radiation dose. It is important to stress that less than 25% of the seasonal (March–August) cumulative erythemal clear-sky UV radiation dose excess is accumulated during the period of polar ozone loss in March–April (Figure 10). The largest contribution comes later in the season as a result of the seasonal ozone anomaly persistence.

Figure 10.

March–August (black) and March–April (gray) cumulative erythemal UV dose (ED) excess in 2011 as a function of latitude estimated from FinROSE (circles) and ATLAS (triangles) based on clear-sky irradiances. Red line shows March–August 2011 cumulative erythemal UV dose anomaly based on satellite-derived clear-sky irradiances.

6 Discussion and Conclusions

[31] We have demonstrated that the total atmospheric ozone amount accumulated in the Northern Hemisphere extratropics by spring strongly influences the solar UV radiation penetrating to the Earth's surface during the following summer season. This result is a direct consequence of the seasonal ozone anomaly persistence documented previously in several papers starting with Fioletov and Shepherd [2003]; however, the implication of this effect for the UV radiation is a novel result of the current study. The signal of spring ozone in the summer UV interannual variability appears strongest when the UV radiation is averaged over large areas because in this case, the noise arising from the interannual meteorological variability is minimized. The local influence on the UV radiation includes, in particular, variability in cloud cover and surface albedo, and their influence may mask the influence of ozone variability, as we have demonstrated in this paper in the case study for 2011. Also, the thickness of ozone column at a particular location and time is influenced by local meteorology such as tropopause height and transport in the lower stratosphere [Orsolini et al., 1998; Steinbrecht et al., 1998; Petzoldt, 1999; Hood et al., 2001; Koch et al., 2005], which may further mask the effect of seasonal ozone anomaly persistence. Nevertheless, we have demonstrated that in some regions such as Eastern Europe, Mediterranean, or southern USA, the March total ozone averaged over extratropics explains as much as 20–40% of the interannual variability in June–August UV radiation, which suggests the possibility of improving seasonal UV forecasts.

[32] The information about UV radiation is relevant, for example, for agriculture because UVB radiation may damage crops. Morrison [1997] showed that an increase of UVB radiation equivalent to 5% decrease in total ozone column may lead to Canadian crop yield losses worth of several million dollars. Since different crops have different sensitivity to the radiation [Krupa and Kickert, 1989], UV forecasts, if issued early enough, may be used to take measures for protecting crops. Such a possibility warrants feasibility studies. Some studies suggest that predictors of springtime ozone anomaly may be found already in the preceding fall season [Kawa et al., 2005]. The advance information about summer UV radiation may also influence human behavior.

[33] We have further demonstrated that the polar ozone chemical depletion has a strong influence on the summer UV. In principle, the relationship between summer UV radiation and springtime total ozone does not depend on what has caused the spring ozone anomaly. However, assessing the effects of chemical ozone loss provides relevant information for the Parties to the Montreal Protocol. Although long-term changes in summer UV have previously been detected and linked to global ozone declines [e.g., Herman et al., 1996; Herman, 2010], our study is probably the first one to estimate the impact of polar ozone loss on summer UV increases. We have estimated that the cumulative March–August clear-sky UV dose excess due to the massive Arctic ozone loss in 2011 was about 3–4% of the climatological cumulative UV dose in the Northern Hemisphere extratropics. Such an increased amount of UV radiation under cloud-free conditions may have implications for ecosystems; however, estimation of possible impacts is beyond the scope of the present study. Our results also suggest that a positive springtime total ozone anomaly would be followed by a negative summer UV radiation anomaly, a result whose implications may also be considered.

[34] Our results show that the high UV radiation levels observed in summer 2011 in the Arctic would be unlikely without the polar ozone loss in spring. Furthermore, we have demonstrated that only 25% of cumulative March–August UV dose excess related to the polar ozone loss in heterogeneous reactions with ODS is accumulated during spring. The rest is accumulated later in the year, after the heterogeneous ozone-destroying reactions have ceased, as a result of the seasonal ozone anomaly persistence. This result strongly suggests that the effect of seasonal ozone anomaly persistence should be included in the assessment of the impacts of polar ozone losses.


[35] This work was supported by the Academy of Finland under grants 134325, 128261, 214695, 132851, and 259537. We thank the OMI and TOMS teams for providing the data and Richard Stolarski for providing access to the merged TOMS/SBUV ozone data. ECMWF is acknowledged for providing ERA-Interim data. We thank Susann Tegtmeier for useful discussions and three anonymous reviewers for their comments. Tuomo Smolander is acknowledged for help with the snow data.