Quality of UVR exposure for different biological systems along a latitudinal gradient

Transcription

1 ARTICLE TYPE CREATED USING THE RSC ARTICLE TEMPLATE (VER. 3.1) - SEE FOR DETAILS XXXXXXXX Quality of UVR exposure for different biological systems along a latitudinal gradient Maria Vernet, 1 Susana Diaz, 2,3 Humberto Fuenzalida, 4 Carolina Camilion, 2 Charles R. Booth, 5 Sergio Cabrera, 6 Claudio Casiccia, 7 Guillermo Deferrari, 2 Charlotte Lovengreen, 8 Alejandro Paladini, 3 Jorge Pedroni, 9 Alejandro Rosales, 9 and Horacio Zagarese 5 Received (in XXX, XXX) Xth XXXXXXXXX 0X, Accepted Xth XXXXXXXXX 0X First published on the web Xth XXXXXXXXX 0X DOI:.39/b000000x Exposure of organisms to ultraviolet radiation (UVR) is characterized by the climatology (annual cycle) and the variance (anomalies) of biologically-weighted irradiances at eight geographical locations in austral South America, from Net effect of UVR on biological systems is a result of the balance of damage and repair which depends on intensity and duration of irradiance and is modulated by its variability. The emphasis in this study is on day-to-day variability, a time scale of importance to adaptive strategies that counteract UVR damage. The irradiances were weighted with DNA- and phytoplankton photosynthesisaction spectra. Low latitude sites show high average UVR. For all sites, the frequency of days with above average irradiances is higher than below average irradiances. Persistence in anomalies is generally low (< 0.36 autocorrelation coefficient), but higher for DNA- than phytoplankton photosynthesis-weighted irradiances due to their higher correspondence to stratospheric ozone. Cloudiness and other factors with small wavelength dependence (i.e., aerosols and albedo) are highly correlated with UVR anomalies at low latitudes (24-33 S); ozone correlates higher at high latitudes ( S). Our results show that organisms in this region deal with several days of excess radiation and fewer, shorter and more intense periods of lower than average radiation. Relief from UVR stress (or higher frequency of days below the climatology) is more prevalent at high latitudes (54.5 S). Thus, lower latitudes are more stressful to organisms not only because of higher average UVR irradiance but also for the higher frequency of days above the climatology. Keywords: Ultraviolet radiation, latitudinal gradient, South America, DNA-weighted irradiance, phytoplankton-weighted irradiance, UVR stress, UVR relief Introduction The rapid ozone depletion in springtime over the Antarctic continent has created the necessity of keeping track of ultraviolet radiation (UVR) reaching the Earth surface, particularly over austral latitudes. 1 The southern cone of South America is one of the few inhabited regions where increased ultraviolet radiation could have a significant impact on terrestrial life. 2-5 At Ushuaia, Argentina (54.5º S) a scanning spectroradiometer has been in operation since 1988 as part of the U.S. National Science Foundation UV Radiation Monitoring Network. 6-9 A Brewer spectrometer belonging to INPE (National Institute for Space Research, Brazil) is installed at Punta Arenas, Chile.,11 Meteorological Services in Argentina and Chile also maintain respective broadband UV networks. 4 In this paper we present results from a network of multi-channel filter radiometers operating in Chile and Argentina since The time series from this network is still too short to detect trends in UVR. 13 Previous studies have shown the overall UV irradiance levels in this region, 3,4 the effect of latitude on annual UVR, 2 the influence of the Antarctic ozone hole in springtime at high and mid latitudes 2,14 and the relationship between atmospheric conditions and UVR. 14. Environmental variability in UVR is dependent on season, location as well as local atmospheric conditions, aerosols and gases. 16 Geometric factors (solar zenith angle and the Earth-Sun distance) are responsible for pronounced and systematic variations in irradiance. 17 Annual cycle in UVR at different latitudes is mainly determined by these factors, as well as the seasonal variability in stratospheric ozone. 18 Short-term variability in UVR is attributed to atmospheric conditions. 19, Clouds attenuate UV irradiance except under broken cloud conditions where an increase of up to 27% above clear sky values can be observed. 21,22 Spectral changes in UVR are influenced by atmospheric conditions. 23 The thinning of the ozone layer translates into increases in UVB (280 to 3 nm) radiation within the wavelength range of 295 to 3 nm, all other atmospheric parameters remaining constant. 6,24 UVA (3 to 0 nm) is more affected by atmospheric processes other than ozone. In South America, both UVA and UVB present a gradient with latitude. 2,4,14 Furthermore, a higher cross correlation between 5 nm and 70 3 nm at lower latitudes and higher cross-correlation between 5 nm irradiance and ozone at higher latitudes indicated the importance of cloud cover and other factors with relative small wavelength dependence (i.e., aerosols and albedo) in explaining UVR variability at lower latitudes (24º 75 S) and the importance of ozone at higher latitudes (54.5º S). 14 Although UVB radiation represents less than 0.8% of the total energy reaching the Earth s surface, it is responsible for almost half the photochemical reactions in aquatic environments. 80 UVB is known to affect a variety of cellular This journal is The Royal Society of Chemistry [year] Journal Name, [year], [vol],

2 processes and molecules in the marine environment and in terrestrial systems that is manifested as reduced productivity, habitats shifts or bio-geochemical cycle alterations. -28 On the other hand, although UVA may be damaging to cellular process, such as photosynthesis, 29 it is also involved in repair mechanisms. Thus, the ratio of UVB:UVA, driven by changes in the atmosphere or by differential attenuation of UVB and UVA in the water column 31 has been identified as key to our understanding of UV stress in aquatic ecosystems. The net effect of UVR is a balance between photochemical damage and biologically-driven processes of recovery and repair. 32 When considering UVR effects on organisms and biological systems, it is critical to understand how the environmental conditions affect the rate of UVR damage (i.e., by changing exposure) as well as the rate of repair (i.e., by activating enzymatic processes). In addition to the intensity and variability of surface radiation, UVR effects on organisms are dependent on the sensitivity of the living system to different wavelengths. Normalizing incident radiation by the corresponding Biological Weighting Function (BWF) allows for estimation of the potential damage. 33 This approach has shown that the sensitivity of a living system can change by a factor of more than a 0 between 0 and 0 nm. 34 Variability in sensitivity to a certain radiation energy among biological systems is also large. The BWF for erythema changes by three orders of magnitude between 0 and 3 nm. The BWF expressing the sensitivity of fish larvae to UV radiation changes only by 1 order over the same wavelength range whereas unprotected DNA is more than 4 orders of magnitude more sensitive at 0 nm. This high variability in sensitivity to UVR is combined with variability in intensity and spectral composition to estimate potential UVR damage to biological systems. Plankton organisms are particularly sensitive to UVR variability due to their floating habitat. Mixing of aquatic organisms from surface to depth (and vice versa) can augment or reduce net damage by altering the balance in damage and repair at different depths. 36 Antarctic phytoplankton incubated under sunlight showed a wide range of inhibition under varying UVB. 37 A reduced photosynthetic performance was evident under full sunlight. They neutralized net UVR damage when incident radiation was not very high (i.e., a cloudy day) thus allowing repair processes to be equivalent to damage; in addition, they recovered photosynthetic efficiency when a cloudy day followed a sunny day. Thus day-to-day UVR variability is important in estimating net damage. In this study we characterize the inter day changes in UVR for an eight year series ( ) at eight locations in South America. The sites encompass a large diversity in latitude and climate (24º 54.5º S). In order to characterize exposure and to ascertain the main environmental factors affecting it, we present anomalies with respect to the annual mean. The source of variability is analyzed with correlations to ozone and 3 nm irradiance (representing clouds and other atmospheric factors with relative small wavelength dependence). 38 The day-to-day variability is analyzed with Markov chains and the persistence of UVR is determined by autocorrelations. Our results show that organisms in this region deal with several days of excess radiation and fewer, Site Latitude Longitude Altitude Location Ushuaia S W SL CADIC Punta Univ. de S 70. W SL Arenas Magallanes Trelew 43. S. W SL Univ. de la Patagonia Bariloche S W ~0 m CRUB Valdivia S W SL Univ. Austral de Chile Buenos Aires 34. S W SL INGEBI Santiago S 70. W 543 m Univ. de Chile Jujuy 24. S.01 W 10 m Univ. Nac. de Jujuy Table 1. The UVR network in South America, with eight multi-channel GUV-511 radiometers (Biospherical Instruments Inc., USA). CADIC: Centro Austral de Investigaciones Científicas; CRUB, Centro Regional Universitario de Bariloche; INGEBI, Instituto Nacional de Genética y Biotecnología. shorter and more intense periods of lower than average radiation. A latitudinal gradient in relief from UVR stress (or higher frequency of days below the climatology) is observed. Materials and Methods a. UV irradiances Irradiances were obtained from measurements provided by the IAI Radiation Network (IAI RadNet). At present, this network is composed of eight multi-channel radiometers (GUV-511, Biospherical Instruments Inc., USA) located in Chile and Argentina (Table 1). The instruments were installed during the 1990 s by different national and international efforts and are still in operation. The stations geographical distribution provides information on UVR at the regional scale (sub-tropical to sub-antarctic). Data from 1995 to 02 are presented in this study. The GUV-511 is a temperature stabilized multi-channel radiometer, which measures downwelling irradiances with moderately narrow bandwidth (near nm) at approximately 5, 3, 3 and 380 nm, plus Photosynthetic Available Radiation (PAR; nm). 39 Under a teflon diffuser, the GUV-511 hosts the five filters and corresponding photodiodes which are arranged with the 5 nm sensor at the center and the other four around it. In normal operation all channels can take a reading every 0.5 seconds and store an average of 1 readings per minute. One advantage of this type of instrument is that the four wavelengths in the UVR are observed simultaneously so fast sky changes can be tracked. All instruments collected one-minute average, 24 hours a day. Raw data were processed applying software provided by Biospherical Instruments Inc., and modified by the Laboratory of UV and Ozone, CADIC, Ushuaia. During data processing, quality and consistency of the data were checked, night values (instrument internal noise) were subtracted, and calibration constants were applied. In cases where the temperature controller failed or the operational temperature of the instrument differed from the temperature 2 Journal Name, [year], [vol], This journal is The Royal Society of Chemistry [year]

3 Erythema (CIE) DNA (Setlow) Plants (Caldwell) Fish (Hunter) Phyto (Neal) at the moment of the calibration, temperature corrections were applied to the calibration constants. Any observed changes in radiometric data were assumed to be linear. Therefore, calibration constants applied between two calibrations dates were calculated by linear interpolation between the two closest calibrations, unless an abrupt known change had occurred (i.e. filter replacement). The GUV-511 were periodically sun calibrated, usually once a year, with a traveling reference GUV-511 (RGUV). 39 The reference radiometer was calibrated, under solar light, before and after traveling, against the spectroradiometer SUV-0 of the U.S. National Science Foundation UV Monitoring Network, installed in San Diego, California, USA. 41 Tests performed during the calibration in year 00 indicated that the GUVs of the network had an error between 8 and 13%, for the 5 channel and Solar Zenith Angle (SZA) smaller than degrees with respect to the SUV-0 from San Diego. For the 3 nm channel, the error was smaller than 5% for all SZA. Details of calibration can be found in Diaz et al. 12,14 The biologically weighted irradiances presented in this paper were based on these measurements and calculated using a multi-regression model (see below). b. Ozone Daily total ozone column was obtained from satellites Nimbus-7, Meteor-3 and Earth Probe, version 8, provided by NASA Goddard Space Flight Centre. 42,43 The time series from July 1995 to December 02 was used for this study. c. Calculation of irradiance weighted by biological action spectra Biologically weighted irradiances are easy to calculate from spectral measurements, but calculation from multi-channel radiometers is not direct. Dahlback 44 proposes a method to estimate the biologically weighted irradiances combining the measurements of four UV channels with a radiative transfer model. BSI, 46 and later de La Casinière 47 propose a multiregressive equation to derive the biologically weighted doses from the GUV-511 irradiances, applying a look up table. Here we use a similar approach but with some modifications. In our method, we do not use a look up table, but include all four UV channels in the multi-regressive equation and determine a function that affects each channel. Also, we add a term to take into account the influence of SZA. Indeed, we Ushuaia (6/28/00 to 12/31/00) Ch 5 (x 1 ) Ch 3 (x 2 ) Ch 3 (x 3 ) Ch 380 (x 4 ) 90-SZA (x 5 ) SZA 70 x 1 a 2e x b 2e x 2 x 3 x 4 a 5e x b 5e x 5 SZA >70 ln x 1 ln x 2 ln x 3 ln x 4 k 5e x l 5e x m 5e x 5 + n 5e SZA a 1d x b 1d x 1 a 2d x b 2d x 2 a 3d x b 3d x 3 a 4d x b 4d x 4 a 5d x b 5d x 5 SZA > ln x 1 ln x 2 ln x 3 ln x 4 k 5d x l 5d x m 5d x 5 + n 5d SZA 70 a 1p x b 1p x 1 a 2p x b 2p x 2 a 3p x b 3p x 3 a 4p x b 4p x 4 a 5p x b 5p x 5 SZA > 70 ln x 1 ln x 2 ln x 3 ln x 4 k 5p x l 5p x m 5p x 5 + n 5p SZA 70 x 1 a 2f x b 2f x 2 x 3 x 4 a 5f x b 5f x c 5f x 5 SZA > 70 ln x 1 ln x 2 ln x 3 ln x 4 k 5f x l 5f x m 5f x 5 + n 5f SZA x 1 x 2 x 3 x 4 a 5ph x b 5ph x 5 SZA > ln x 1 ln x 2 ln x 3 ln x 4 k 5ph x l 5ph x m 5ph x 5 + n 5ph Table 2. Function used to fit data from each channel in the GUV-511 and solar zenith angle with the weighted irradiance obtained from the spectroradiometer. Eryth: erythema; CIE: Commission International de l Eclairage; DNA: Desoxyribo-nucleic acid; Fish: fish larvae; Phyto: phytoplankton photosynthesis; SZA:solar zenith angle. use the complement of the SZA in the equation to ensure the condition that the weighted irradiance tends to zero when each of the variables tend to zero. It has been observed that, in calibrating multichannel instruments, ozone and SZA are critical factors, mainly for the 5 channel 44. By introducing a multi-regressive method in calibrating the radiometers, improvements were observed, mainly in the UVB channel. 12 A similar approach is used here by adding the above mentioned changes, thus including the effect of ozone and SZA. Then, the equation we propose for weighted irradiances is: I w = A f(x 1 ) + B f(x 2 ) + C f(x 3 ) + D f(x 4 ) + E f(x 5 ) [1] where I w is the weighted irradiance, A, B, C, D and, E are the regression coefficients determined with least square methods and f(x i ) is a function of the irradiance measured by channel i, except f(x 5 ), which is a function of (90-SZA). The function that affects each channel was determined as the best fit of the scatter plot of the biologically weighted irradiance obtained from the SUV-0 against the irradiance measured by the GUV-511 channel under consideration. For 70 SZA, the scatter plot is done with the difference between the weighted irradiance and irradiance channel 5, against the complement of the SZA (90-SZA). For larger SZA we used the logarithm of the irradiance to improve the quality of the fitting, increasing the influence of 75 smaller irradiance values. ln(i w ) = K f(x 1 )+L f(x 2 )+M f(x 3 )+N f(x 4 )+O f(x 5 )+P [2] where K, L, M, N, O and P are the coefficients of the multiregressive equation determined with least square methods. 80 In order to develop and test the proposed method we used simultaneous spectral and multichannel data from Ushuaia (54 o 49 S, 68 o 19 W) and San Diego (32 o N, 117 o 11 W). The selected sites show important differences in the patterns of ozone and cloud variability. Ushuaia experiences large 85 daily ozone variations during the spring because of the This journal is The Royal Society of Chemistry [year] Journal Name, [year], [vol],

4 presence of the ozone hole, and is under rather cloudy skies. San Diego presents a smaller variation in ozone and smaller SZA in the summer than Ushuaia and preponderance of clear skies, although with more influence of aerosols as consequence of its larger population (about 3 millions versus ~,000 for Ushuaia). At both sites, the spectral data was obtained with spectroradiometers SUV-0, 48 and the multichannel data with a GUV-511 radiometer. Measurements at both sites were performed under all Coef weather conditions, which included different cloud cover, large solar zenith angles, extreme ozone conditions and various albedo situations. The model was solved for Ushuaia combining instantaneous spectral data weighted by a given action spectrum and multichannel data, for the period July 1 st to December 31 st, 00, for all solar zenith angles smaller than 85 degrees. Once the equations and coefficients were determined, the model was tested for San Diego, using data for the period July 9 th to December th, Data for the radiometers GUV are obtained each minute, while the spectroradiometer SUV-0 performs one scan each minutes. Thus, 4 concurrent values per hour could be obtained. We solved the multi-regressive equation for four biologically weighted irradiances: 1) parameterization of DNA damage; 49 2) general plant response; 3) erythema; 51 and 4) phytoplankton photosynthesis. 34 Table 2 shows the fitting functions for each channel and for the complement of SZA, where: a 2e = , b 2e = a 5e = , b 5e = k 5e = , l 5e = , m 5e = , n 5e = a 1d = , b 1d = a 2d = , b 2d = a 3d = , b 3d = a 4d = , b 4d = a 5d = , b 5d = k 5d = , l 5d = , m 5d = , n 5d = a 1p = , b 1p = a 2p = , b 2p = a 3p = , b 3p = a 4p = , b 4p = a 5p = , b 5p = k 5p = , l 5p = , m 5p = , n 5p = a 2f = , b 2f = a 5f = , b 5f = , b 5f = k 5f = , l 5f = , m 5f = , n 5f = a 5ph = , b 5ph = k 5ph = , l 5ph = , m 5ph = , n 5ph = and where the first subscript corresponds to the term of the equation and the second to the effect. The regression coefficients corresponding to equation 1 and 2, which Erythema (CIE) DNA (Setlow) Plants (Caldwell) Fish (Hunter) resulted of solving the multi-regressive equation for the four action spectra are shown in Table 3. The coefficients of determination (r 2 ) for the regression and RMS errors between the values of the biological weighted irradiances calculated with the SUV and the GUV are shown in Table 4, for hourly averaged values and SZA smaller than 85 degrees. The multi-regressive model to calculate UV weighted irradiances based on the multi-channel filter radiometers presented here has universal application. It was developed and tested at two sites that present very different atmospheric and ground characteristics, for a period of near six month (winter to summer), during which very extreme situations were present (i.e., ozone hole overpass at Ushuaia, which produces % ozone depletion, large SZA, snow cover in Ushuaia during winter, etc.), and using two sets of radiometer and spectro-radiometers, with very good results. When comparing the results of the model proposed here with a calculation of weighted irradiances based only on a coefficient for each channel (instead of a function) and not including the SZA term, we observed our method to give better estimates, mainly for SZA larger than o, which is particularly important at higher latitudes (e.g., for DNA the RMS error diminished by half, for SZA between o and 85 o, data not shown). In this paper, we applied the abovementioned methodology to obtain biologically weighted irradiances from the IAIRadNet. d. Annual cycle Phytoplankton (Neal) A B C D E K L M N O P Table 3. Coefficients obtained when solving the multi-regressive equation for calculating weighted irradiances with the GUV-511. For acronyms see Table 2 legend. To determine climatology we estimated the annual cycle using a procedure described in Wilks 52, which is a more general version of the Fast Fourier Transform (FFT) spectral analysis. Wilks approach has the advantage that equidistant data are not needed because coefficients and phases corresponding to the harmonic components are obtained via multivariate analysis. 14 Times series of noon values were selected from the databases of biologically weighted irradiances. Then, the time series average was calculated for each Julian day and the annual cycle was inferred thereof. The annual component of a time series may be represented by a cosine function, as follows: y t = y + C 1 cos 2πt n φ 1 [3] where y t is sampled value on day t; y is mean value of the 4 Journal Name, [year], [vol], This journal is The Royal Society of Chemistry [year]

5 annual cycle; t is time; n is number of days in a cycle, in this case 3; C 1 is amplitude of the annual cycle; and φ 1 is phase. Equation [3] may be re-formulated taking into account that: cos( α φ ) = cos( φ )cos( α) sin( φ )sin( α) [4] l l + Given that A 1 = C 1 cos(φ 1 ) B 1 = C 1 sin( φ 1 ) and that x 1 = cos(α) x 2 = sin(α) equation [4] is transformed into a regression with two predictors y t = y + A x + B [5] 1 1 1x2 Usually, the annual cycle is not a pure cosine. It is rather the summation of the first (n = 3), the second (n = 180) and the third (n = 1) harmonics. Taking this into account, and following the above-explained procedure, a multivariate equation with six predictors is obtained which, when solved, provides the annual cycle. The independent term of the equation corresponds to the mean of the annual cycle or continuous component in the Fast Fourier Transform. Tests confirmed that this calculation of the annual cycle represented well the data (not shown). The same procedure was used to calculate the annual cycle for ozone. Since the seasonal variability in UV radiation is very pronounced, the annual cycle calculated for irradiance would present large errors in winter, due to the small weight that winter irradiance presents when solving the multivariate equation using least squares. In order to avoid this problem, logarithmic transformation was applied. 14 e. Anomalies Anomalies for total column ozone, irradiance at 3 nm and biologically-weighted irradiance (DNA and phytoplankton) were calculated as deviations from the annual cycle as normalized_anomaly = V d V J [6] V J where V d is the time series value at date d and V J is the annual cycle value at Julian Day J, corresponding to date d. Gaps in data collection were filled with the annual cycle. Usually, the anomaly is normalized by dividing it by the standard deviation. 53 In this case we preferred to use the value of the annual cycle for normalization because, as consequence of ozone depletion, the standard deviations for DNA- and phytoplankton photosynthesis weighted-irradiances and total column ozone are perturbed at higher latitudes. f. Correlations and autocorrelations Atmospheric variables often exhibit statistical dependence l Erythema (CIE) DNA (Setlow) Plants (Caldwell) Fish (Hunter) Phyto (Neal) with their own past or future values. In the terminology of atmospheric sciences, this dependence through time is known as persistence. Then, persistence can be defined as the existence of a positive statistical dependence among successive values of a variable, or among successive occurrences of a given event and can be characterized in terms of serial correlations or temporal autocorrelations. 22 We performed persistence studies based on time series of the normalized anomalies for time lags of one to seven days. Additionally, cross correlations of the normalized anomalies of biologically weighted irradiances with the normalized anomalies of ozone and 3 nm irradiance were carried out with the aim of determining the influence of different parameters on the weighted irradiances. g. Markov chains - transition probabilities To evaluate the damage-repair probabilities at each site, we also estimated persistence in UVR by converting the time series of anomalies into a Markov chain and calculating transition probabilities. 52,54 In order to perform the analysis, the values of the anomaly time series were converted into discrete values, representing a first-order Markov chain. A Markov chain can be considered as being based on collection of states of a model system. For each time period, the Markov chain can either remain in the same state or change to one of the other states. The behavior of a Markov chain is governed by a set of probabilities for these transitions called the transition probabilities. These probabilities specify likelihoods for the system being in each of the possible states during the next time period. For first-order Markov chain the transition probabilities controlling the next state of the system depend only on the current state of the system: they are conditional probabilities of the type P r ppt today ppt yesterday 90 Ushuaia (7/1/00 to 12/31/00) Hourly Average Hourly Average R 2 Error RMS (%) R 2 Error RMS (%) SZA <SZA< <SZA<80 (85) 7.92 (9.66) 8.82 (9.82) SZA <SZA< <SZA<80 (85) (23.38) (.64) SZA <SZA< <SZA<80 (85) 6.48 (8.27) (18.24) SZA <SZA< <SZA<80 (85) 7. (9.07).39 (11.05) SZA <SZA< <SZA<80 (85) 6.82 (7.49) 8.27 (9.95) Table 4. RMS error and determination coefficients between the hourly averaged irradiances weighted by different action spectra calculated from spectral measurements and using multi-channel instrument. Model developed for Ushuaia and tested for San Diego. [7] San Diego (7/9/1999 to 12//1999) This journal is The Royal Society of Chemistry [year] Journal Name, [year], [vol],

6 Figure 1. Biological Weighting Functions (BWF) of key molecules (Desoxyribo-nucleic acid or DNA), general plant response, human skin (erythema), organisms (fish larvae) or ecosystem processes (phytoplankton photosynthesis) showing the relative sensitivity to ultraviolet radiation (UVR). The values are normalized to 1 at 0 nm. DNA is the most sensitive to UVB (280 3 nm) with small sensitivity to UVA (3-0 nm). Phytoplankton photosynthesis has the lowest sensitivity to UVB and highest to UVA such that effective irradiance at 390 nm is only 0 times lower than at 0 nm. In contrast, DNA effective irradiance at 3 nm is 0 times less effective than at 0 nm. Other molecules or biological processes have intermediate sensitivity. Analyses in this paper correspond to UVR weighted by the BWF (calculated as Radiation * BWF) from DNA and phytoplankton photosynthesis representing the two extremes in UVA and UVB sensitivity. where the probability Pr that a value be present today (ppt today ) given a determined value yesterday (ppt yesterday ). The simplest kind of discrete random variable pertains to the situation of dichotomous (yes/no) events. In our case, we will consider a two state first-order Markov chain. The system will be in state 1 (event 1), if the anomaly is bigger than zero (irradiance larger than climatology), and in state 0 (event 0), if the anomaly is smaller or equal to zero (irradiance smaller or equal to climatology). Three states could have been considered: larger than 0, equal to 0 and smaller than 0, but, in this analysis no anomaly is equal to 0, then, the value 0 was included in state 0. Taken this into account, the probability that a day with irradiances smaller than, or equal to, the annual cycle (anomaly < 0) be followed by a day with irradiance smaller than, or equal to, the annual cycle is P 00 ; and if followed by a day with irradiance higher than annual cycle is P 01. In the same way, the probability that a day with irradiance larger than the annual cycle (anomaly > 0) be followed by a day with irradiance larger than the annual cycle is P 11 ; and if followed by a day with irradiance smaller than or equal to the annual cycle is P. Results a. South American climate Climate is a key issue to understand the short-term and seasonal variability in UVR. We summarize here the main characteristics at each location. The study sites in South America, between the Capricorn Tropic (24º S) and sub- Antarctic conditions (54.5º S) experience very different climate conditions (Table 5). Jujuy, on the eastern slope of the Andes, receives most of its rainfall during the summer months by convective activity. Santiago and Valdivia belong to the Mediterranean type of climate with a dry summer and most rain falling in winter associated with frontal passages that become more frequent the higher the latitude. The main difference between these two sites is the annual precipitation, almost an order of magnitude larger in Valdivia, so that even in the driest month 66 mm are received. Punta Arenas is part of a cold steppe that develops on the eastern side of the Andes with a fairly scanty and homogeneous precipitation regime throughout the year. Ushuaia, although located inside the Andes shares this condition. Trelew aridity is due to the rain shadow on the lee (eastern) side of the Andes receiving its annual 6 mm in a fairly well distributed manner through the year. Bariloche is located at the foothills of the Andes on the lee side. Buenos Aires has a fairly moist and mild climate with no major rainfall season. Site Annual rain (mm) Minimum (mm) month Maximum (mm) month b. UVR variability under clear sky conditions Cloud cover (tenths) Jujuy 6 0 (July) 183 (January) n.a. Buenos Aires 59 (July) 118 (March) 4 even Trelew 6 8 (August) 21 (May) 5 even Ushuaia 549 (October) (March) 6 even Santiago (January) 80 (June) 2 to 6 Valdivia (February) 396 (June) 4 to 7 Punta Arenas 3 24 (October) 43 (May) 6 to 7 Table 5. Precipitation and cloud cover in the localities where GUV radiometers are deployed. Organisms at any given location and living for at least one year are exposed to a range of variability in effective UVR of several orders of magnitude. Using clear sky weighted 70 irradiance at Ushuaia, S, several characteristics in the spectral UVR are of note. As expected, the difference in the overall sensitivity in UVB vs. UVA of the different weighted irradiances is similar to the shape of the Biological Weighting Function (BWF, Fig. 1) given that incident 75 radiation is the same for all. In winter, the maximum irradiance or the wavelength with highest effective irradiance is 312 nm for DNA-weighted irradiances, 3 nm for fish larvae, general plant-weighted irradiances, and 3 nm for phytoplankton photosynthesis ( June; Fig. 2a). 80 Phytoplankton photosynthesis-weighted irradiances are an order of magnitude higher than DNA-weighted irradiances in the UVB and four orders of magnitude in the UVA (3 nm). Although the relative sensitivity between UVB and UVA in the different biological systems is maintained throughout the 85 seasons, there is a shift to lower maximum effective wavelength in the summer. At this time, maximum wavelength of DNA-weighted irradiance is 3 nm, 8 nm for erythema- and general plant-weighted irradiances, and 3 nm for phytoplankton photosynthesis-weighted 90 irradiances (Fig. 2b). Also as expected, higher irradiances are experienced in the summer, 6-9 with an increase of two orders of magnitude for DNA-weighted irradiances at its maximum wavelength of 3 nm and an increase of one order of magnitude for phytoplankton photosynthesisweighted irradiances at 3 nm (Fig. 2b). The 95 other 6 Journal Name, [year], [vol], This journal is The Royal Society of Chemistry [year]

7 Figure 2. Spectral Irradiance at different seasons and for varying stratospheric ozone concentrations. Clear sky spectral UVR between 295 and 390 nm weighted by BWF of different systems (DNA, open squares; General Plant response, triangles; Fish Larvae survival, grey squares; Erythema or skin sensitivity, black squares; Phytoplankton Photosynthesis, diamond at a) austral winter; b) austral summer and c) austral summer under % ozone reduction. All plots for Ushuaia, Argentina at 54 o 49 S, 68 o 19 W. SZA depicts Solar Zenith Angle. maximum sensitivity, 0 nm for DNA-weighted irradiances, 5 nm for erythema- and general plant-weighted irradiances, and 3 nm for phytoplankton photosynthesis. There is also an increase of almost one order of magnitude for DNA effective irradiance with smaller increases for biological systems experience intermediate increase in effective irradiances. Finally, the effect of ozone is experienced at the lowest UVB wavelengths, near 0 nm, with no changes in the effective UVA radiation (Fig. 2c). Under conditions similar to the ozone hole (% ozone reduction), there is a further decrease in the wavelength of erythema-, general plant-, or phytoplankton photosynthesisweighted irradiances. In summary, the sensitivity of the different biological systems (Fig. 1) affects wavelength and intensity of exposure as well as the balance between UVB and UVA (Fig. 2). Seasonal changes in UVR affect overall intensity as well as the wavelength of maximum effect, with lower maximum effective wavelengths in the summer. Finally, ozone influences mostly DNA effective irradiance. Lower ozone decreases wavelength of maximum sensitivity and increases irradiance, similar to the effect of smaller SZA (summer). Based on these results we choose DNA- and phytoplankton photosynthesis-weighted irradiances to characterize the UVR exposure to organisms living in the southern cone of South America. These two BWFs represent extremes in UVB:UVA sensitivity, DNA mostly sensitive to UVB changes and phytoplankton photosynthesis very sensitive to UVA. c. UVR annual cycle The UVR climatology at each site is defined as the annual cycle of the biologically weighted irradiances (smooth line in Fig. 3). As expected, higher average DNA-weighted irradiances are observed at lower latitudes year round. 2 From those, arid areas at mid latitudes (Santiago de Chile, Fig. 3c) have higher average summer irradiances than wet forested locations (Jujuy, Fig. 3a) in spite of the latter being further north and experiencing higher UVR irradiance under clear skies (smaller SZA) (Table 6). A large difference between summer and winter is evident at all locations. This seasonal contrast is more pronounced at higher latitudes (Table 6, compare Jujuy with Ushuaia); furthermore, low latitude sites present measurable UVR in the winter while no UVR is detected south of 42º S (i.e., Trelew, Fig. 3f). Daily changes in irradiance can be observed by comparing DNA-weighted irradiances (thin line in Fig. 3) with the annual cycle. Several patterns in irradiance variability are apparent. Most sites show lower than average irradiances (below the annual cycle) that are larger in magnitude and shorter in duration than above average irradiances; in contrast, above average irradiances usually last longer (several days) and have moderate magnitude (Fig. 3b, 3e, 3f). The observed variability pattern, characteristic of the 8- year time series, is attributed to the fact that the 3-day annual cycle is based on average values. The bias towards higher than average irradiances means that the integral of above average irradiances is similar to the integral of below average irradiances, which is sensitive not only to the duration but also to the magnitude of irradiance events (see also Anomalies section). The presence of anomalous periods is also evident; for example in Bariloche (Fig. 3e), consistent below average irradiances lasted for several weeks in the springtime of Similarly, long sequences of days with above average irradiances can occur at several locations, as seen in Bariloche in January and February. It is also of note a large variability in irradiance in Jujuy, Buenos Aires, Valdivia and Trelew (Figs. 3a, 3b, 3d and 3f) where days of very high (maximum) irradiance are followed by days of very low irradiance; the resulting pattern is one of large oscillations. For example, the ratio of maximum/minimum irradiance in summer is 1238 for Buenos Aires, in comparison to a minimum of 71 in Santiago de Chile (Table 6). The maximum ratio is seen in Ushuaia (1661). Such variability could be attributed to the high irradiances found in summer during clear days combined with characteristic thick cloudiness in overcast days. Finally, high short peaks in irradiance during springtime, previously attributed to the ozone hole, can be observed as far north as Valdivia and Bariloche (39.5 S; Figs. 3d, 3e). 2,14 This journal is The Royal Society of Chemistry [year] Journal Name, [year], [vol],

8 a) DNA Annual Cycle Summer Irradiance Winter Irradiance Site Max Min Max/Min Max Min Max Min Jujuy Buenos Aires Santiago Valdivia Bariloche Trelew Punta Arenas Ushuaia Table 6. Irradiances in South America, weighted by DNA and Phytoplankton photosynthesis action spectra, in units of μw cm -2. Maxima (Max) in the Annual Cycle are observed between 21 December and 13 January; minima (Min) between 22 and June. Annual maxima irradiances are measured between 2 December and 17 January, with the exception of Jujuy where there is an early maximum in 6 November 1997; the minima in irradiance are between 7 June and 7 July with the exception of Jujuy where there is a late minimum in 29 September Phytoplankton photosynthesis-weighted irradiances at the eight localities show very similar annual cycles and d. UVR anomalies variability than for DNA-weighted irradiances (Fig. 4). The Anomalies can further describe the degree of day-to-day resulting regional pattern is one where summer irradiances variability of weighted irradiances (Figs. 5 and 6). Only one are highest at mid latitude locations (Figs. 4c, 4d, 4e), year is shown in the figures to allow enough detail in the intermediate at low or mid latitudes (Figs. 4a, 4b and 4f) and graph to understand the differences among sites while lowest at higher latitudes (Figs. 4g and 4h). The subsequent statistical analyses are based on the anomalies of characteristic feature of several days with higher than the entire time series ( ). In contrast to deviations average irradiances interspersed with few days of very low irradiance is also present here and is even more pronounced for the phytoplankton- than for a) DNA Mean +/- Standard Deviation Percentage of Total DNA-weighted irradiances, in particular at mid Positive Negative Positive Negative Site Anomalies Anomalies Anomalies Anomalies latitudes (Figs. 4b, 4d, 4e and 4f). Several features are characteristic of this irradiance which is more heavily weighted by UVA (compare with Figs. 1 and 2). In general, the magnitude of the phytoplankton b) Phytoplankton Annual Cycle Summer Irradiance Winter Irradiance Site Max Min Max/Min Max Min Max Min Jujuy Buenos Aires Santiago Valdivia Bariloche Trelew Punta Arenas Ushuaia photosynthesis weighted irradiances is three to four times higher than DNA-weighted irradiances. This change is most obvious in the winter when higher-latitude sites show measurable phytoplankton-weighted UVR (Figs. 4e, 4f, 4g and 4h) but not DNA-weighted irradiances (Figs. 3e, 3f, 3g and 3h). The overall effect is one of reduced seasonality (see Annual Cycle Max/Min in Table 6). Another major difference is the lower latitudinal gradient in the annual cycle observed in phytoplankton-weighted irradiances. While for irradiances weighted with the DNA action spectrum the average summer irradiance between lowest and highest latitudes has a factor of ~2.5 (0. μw cm -2 in Jujuy and 0.21 μw cm -2 in Ushuaia, Figs. 3a and 3h; Table 6), it is ~1 for irradiances weighted with the phytoplankton photosynthesis action spectrum (2.2 μw cm -2 in Jujuy and 1.7 μw cm -2 in Ushuaia, Figs. 4a and 4h). In addition, irradiance variability is lower (oscillations of day- to-day irradiance are less pronounced) for phytoplankton photosytnthesis-weighted irradiance than for DNA-weighted irradiances (Table 6). A similar latitudinal gradient is observed when comparing 5 nm and 3 nm irradiances. 14 Jujuy 0.5 +/ / % 34% Buenos Aires / / % 32% Santiago / / % % Valdivia / / % 37% Bariloche / / % 39% Trelew / / % 36% Punta Arenas / / % % Ushuaia / / % % b) Phytoplankton Mean +/- Dev Percentage of Total Site Positive Negative Positive Anomalies Anomalies Anomalies Jujuy / / % 34% Buenos Aires / / % 31% Santiago / / % % Valdivia / / % % Bariloche / / % % Trelew / / % 29% Punta Arenas / / % % Ushuaia 0.0 +/ / % % Negative Anomalies Table 7. Statistics of the positive and negative UVR anomalies ( ), a) for DNA- and b) for phytoplankton photosynthesis-weighted irradiances. 8 Journal Name, [year], [vol], This journal is The Royal Society of Chemistry [year]

9 Figure 3. Seasonal variability in UVR-weighted irradiances in South America Daily noon DNA-weighted UVR ( nm) at different localities from 28 to S, in units of μw cm -2, from 1 July 1998 to June The thin straight line depicts the climatology or annual cycle. Jujuy 24. o S,.01 o W; Buenos Aires 34. o S, o W; Santiago o S, 70. o W; Valdivia o S, 73.1 o 4 W; Bariloche o S, o W; Trelew 43. o S,.05 o W; Punta Arenas o S, 70. o W; Ushuaia o S, o W. of weighted irradiances to the annual cycle, the overall magnitude of the anomalies of DNA- and phytoplankton photosynthesis weighted irradiances is more similar, with the exception of high-latitude sites (Figs. 5g and 5h). The sequence of positive and negative anomalies in time follows the pattern already described for irradiances (Figs. 3 and 4) with several days of positive anomalies interspersed with short, pronounced negative anomaly periods (see Figs. 5e and 5f). On average for the 8-year time series, ~% of the anomalies are positive and % are negative (Table 7). This is not an artifact of the annual cycle calculation as the average and standard deviation of the positive and negative anomalies is similar at each locality, with the exception of Punta Arenas and Ushuaia where the positive anomalies for DNA-weighted irradiances present higher values and variability, as might be expected by the increased UVB during periods of influence of the Antarctic ozone hole. 6,7 Statistics for the period indicate this is a representative year (data not shown). The presence of extended periods with positive or negative anomalies, mentioned already for the irradiances (Figs. 3 and 4), are seen for most localities, with the exception of Buenos Aires. These periods break up the temporal development of positive and negative anomalies described above, can last from 6 to 0 days and had an average extension of + 9 days for positive anomalies and + days for negative Site P 00 P 01 P 11 P Jujuy Buenos Aires Santiago Valdivia Bariloche Trelew Punta Arenas Ushuaia a) DNA Site P 00 P 01 P 11 P Jujuy Buenos Aires Santiago Valdivia Bariloche Trelew Punta Arenas Ushuaia b) Phytoplankton Table 8. UVR Anomaly Transition Probabilities a) for DNAweighted irradiance anomalies and b) phytoplankton photosynthesisweighted irradiance anomalies. The system will be in state 1 (event 1), if the anomaly is bigger than zero (irradiance larger than the climatology), and in state 0 (event 0), if the anomaly is smaller or equal to zero (irradiance smaller or equal to the climatology). All transition probabilities are significant at α = This journal is The Royal Society of Chemistry [year] Journal Name, [year], [vol],

10 Figure 4. Seasonal variability in UVR-weighted irradiances in South America Daily noon phytoplankton photosynthesis-weighted UVR ( nm) at different localities from 28 to S from 1 July 1998 to June The thin straight line depicts the climatology or annual cycle. Latitudes and Longitudes of each location can be found in Table 1 and Figure 3 legend. anomalies during For example, at Bariloche, periods of positive anomalies were observed in the summer between 12 January and 13 February 1999 and also in the fall from 19 April to May. Negative anomalies were almost continuous in the spring, from 22 October to 3 December 1998 (Figs. 3e and 5e). The magnitude and characteristic of the anomalies at each site present interesting geographical differences. Most noticeable, some locations have pronounced anomalies (i.e., Figs. 5b, 5g and 5h) while others are less variable (Fig. 5c). Several mid-latitude locations present a seasonal cycle in the phytoplankton-weighted anomalies: Buenos Aires, Santiago, Valdivia and Bariloche show lower positive anomalies in summer (Figs. 6b, 6c, 6d and 6e). There is also a difference among locations with respect to negative anomalies: Santiago de Chile and Bariloche often show a predominance of several days of positive anomalies (Figs. 5c, 6c and 5e, 6e) whereas Jujuy, Valdivia and Ushuaia (Figs. 5a and 6a, 5d and 6d, 5h and 6h) present more frequent negative anomalies. e. UVR anomaly autocorrelations In order to characterize further the day-to-day variability in weighted irradiance we used autocorrelation to estimate the similarity in anomaly values in a seven day period (lag = 1 to 7). Autocorrelations of DNA-weighted anomalies show low and consistent values (between 0.26 and 0.36, Fig. 7) at all sites in consecutive days (lag = 1). Autocorrelations suggest predictability in consecutive days is about -%. For lags larger than one, the values of the autocorrelations diminish, more rapidly at mid-latitude (i.e., Buenos Aires, Fig. 7b) than at higher-latitude sites (i.e., Ushuaia, Fig. 7h). Autocorrelations of phytoplankton photosynthesis weighted anomalies show a marked difference among sites, varying between 0.04 and 0.27 (lag = 1); in general, there is a lower autocorrelation than for DNA-weighted anomalies, in agreement with the dependence of 5 nm irradiance on ozone. 14 The slope of the autocorrelation function can indicate the degree of persistence in the system. In general the persistence is low and lowest for phytoplankton photosynthesis weighted anomalies (the latter presents a steeper slope). At lower latitudes, the values for phytoplankton- and DNA-weighted irradiance differ less than for higher-latitude sites (the slope of the lines overlap, i.e., Buenos Aires). As only DNA-weighted irradiances are highly sensitive to stratospheric ozone concentration (as shown in Fig. 2c), the higher similarity in persistence between both weighted irradiances at lower-latitude sites can be attributed to a higher influence of atmospheric conditions on DNA-weighted irradiances. In contrast, only ozone concentrations seem to have a higher influence on DNA- Journal Name, [year], [vol], This journal is The Royal Society of Chemistry [year]

11 Figure 5. Anomalies of DNA-weighted irradiances Normalized anomalies for the eight studied sites in South America, from 1 July 1998 to June Latitudes and Longitudes of each location can be found in Table 1 and Figure 3 legend. weighted irradiances further south and here the persistence diverges. In summary, a low persistence in UVR is characteristic of this region, as implied by the see-saw pattern of irradiance about the mean observed in Figs. 3 and 4. At higher latitudes, the influence of ozone is associated with higher persistence 14 and as expected, DNA-weighted irradiances are more sensitive to this influence. f. UVR anomaly transition probabilities Another way to characterize the irradiance anomalies is by estimating the probability of finding a certain UVR anomaly given an anomaly on the previous day, or transition probabilities (Table 8). In general, both action spectra weighted irradiances show similar transition probabilities at all locations. The value of the probabilities indicates a very interesting irradiance pattern for the region. P 11 (the probability of having two consecutive days with irradiance larger than the annual cycle) is largest everywhere (~0.7). In addition, P (the probability of having a day with irradiance higher than the annual cycle followed by a day with irradiance lower than the annual cycle) is lowest everywhere (~0.3). These probabilities support the observation made before that several days of higher than average UVR are followed by shorter periods of lower than average irradiance (Figs. 3 to 6 and Table 7). A latitudinal gradient is also observed in these two transition probabilities, with P 11 lower and P higher at higher-latitude sites, implying a larger probability of a day with high irradiance being followed by a day of low irradiance towards the south (e.g., higher frequency of cloudy or low ozone days). On the other hand, P 00 (the probability of finding two consecutive days with irradiances equal or below the annual cycle) and P 01 (the probability of a day with irradiance equal to or lower than the annual cycle followed by a day with irradiance larger than the annual cycle) show similar values, for both biologically-weighted irradiances at all locations (~0.5). Thus, the chance of switching from low to high UVR in two consecutive days is about the same as staying low. Some latitudinal differences are observed between the action spectra. At lower-latitude sites, P 01 is higher for the DNAand lower for the phytoplankton-weighted irradiances; at lower latitudes there is a higher probability of having a day with above average UVR after a day with low UVR. The opposite is observed for P 00 where larger probabilities for DNA-weighted irradiances are observed at lower latitudes, i.e., two consecutive low-irradiance days are more common at higher latitudes. In summary for this region, 70% of the time a day with high UVR is followed by another day with high UVR. And this pattern is more prevalent at low latitudes. If a day has low UVR, there is a % chance of having low UVR during the next day. The chance of a second low UVR day is higher at higher latitudes. Thus, we suggest that DNA would have This journal is The Royal Society of Chemistry [year] Journal Name, [year], [vol],

12 Figure 6. Anomalies of Phytoplankton Photosynthesis-weighted irradiances - Normalized anomalies for the eight studied sites in South America, from 1 July 1998 to June Latitudes and Longitudes of each location can be found in Table 1 and Figure 3 legend. more probability of repair at higher- than at lower-latitude sites (P 00 and P larger). Higher latitudes provide also more relief from UVR stress for phytoplankton photosynthesis (P is highest at high latitudes) but the period of recovery is shorter than for DNA (P 00 is lower at higher latitudes). g. Importance of Atmospheric Conditions and Ozone on Irradiance Anomalies Cross-correlation of biological-weighted irradiance anomalies can be used to ascertain importance of environmental variables (Table 9). Influence of cloudiness and other atmospheric processes (albedo, aerosols, etc.) on UVR variability was estimated with 3 nm irradiances. The effect of stratospheric ozone on UVR was estimated from ozone column height in Dobson units. In general, weighted irradiances have a high correlation with 3 nm, higher for phytoplankton (r = ~1.0) than for DNA (r = ~0.8) and with decreasing influence with latitude for DNAweighted irradiance anomalies but high everywhere for phytoplankton-weighted irradiance anomalies. In contrast, influence of ozone is present everywhere for DNA (r = ~0.4), and increases with latitude (r = 0.), but not for phytoplankton (r = ~0.01), as expected from the high UVA sensitivity of phytoplankton seen in Fig. 2. Thus,` cloudiness and other relatively wavelength independent factors are a better predictor of DNA damage by UVR at lower latitudes (24 S) and both ozone and 3-nm play an important role at higher latitudes (54.5 S). 14 Cloudiness is the only predictor for phytoplankton photosynthesis UVR damage at all latitudes. There is a small but significant cross correlation between 3 nm irradiance and stratospheric ozone column (Table ). The low correlation implies little overlap between these two phenomena although it is known that 3 nm irradiance and ozone column respond to atmospheric pressure (cyclonic and anticyclonic systems). During a high-pressure system (anticyclonic) there is good weather and clear skies resulting in high irradiance, including high 3 nm irradiance. At the same time, a high pressure system causes a decrease in ozone DNA DNA Phytoplankton Phytoplankton Site 3 nm Ozone 3 nm Ozone Jujuy Buenos Aires ns Santiago Valdivia Bariloche Trelew ns Punta Arenas Ushuaia Table 9. Cross-correlation coefficients (r) of biological-weighted irradiances to ascertain importance of environmental variables, for lag = 0. Irradiance anomalies at 3 nm are an index of cloudiness (and other wavelength independent atmospheric factors). Ozone column anomalies were estimated from TOMS NASA satellite. All values are statistically significant for α = 0.01 unless otherwise indicated. (ns) not statistically significant. 12 Journal Name, [year], [vol], This journal is The Royal Society of Chemistry [year]