glUV: a global UV-B radiation data set for macroecological studies


  • Michael Beckmann,

    Corresponding author
    1. Department of Computational Landscape Ecology, UFZ – Helmholtz Centre for Environmental Research, Leipzig, Germany
    2. Institute of Biology/Geobotany and Botanical Garden, Martin Luther University Halle Wittenberg, Halle, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Tomáš Václavík,

    1. Department of Computational Landscape Ecology, UFZ – Helmholtz Centre for Environmental Research, Leipzig, Germany
    2. Department of Ecology and Environmental Sciences, Faculty of Science, Palacký University Olomouc, Olomouc, Czech Republic
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Ameur M. Manceur,

    1. Department of Computational Landscape Ecology, UFZ – Helmholtz Centre for Environmental Research, Leipzig, Germany
    2. Department of Community Ecology, UFZ – Helmholtz Centre for Environmental Research, Halle, Germany
    Search for more papers by this author
  • Lenka Šprtová,

    1. Department of Ecology and Environmental Sciences, Faculty of Science, Palacký University Olomouc, Olomouc, Czech Republic
    2. Schwestern von Betlehem, Kloster Maria im Paradies, St. Veit im Pongau, Austria
    Search for more papers by this author
  • Henrik von Wehrden,

    1. Institute of Ecology/Faculty of Sustainability, Leuphana University Lüneburg, Lüneburg, Germany
    2. Center for Methods, Leuphana University Lüneburg, Lüneburg, Germany
    3. Research Institute of Wildlife Ecology, Vienna, Austria
    Search for more papers by this author
  • Erik Welk,

    1. Institute of Biology/Geobotany and Botanical Garden, Martin Luther University Halle Wittenberg, Halle, Germany
    2. German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig, Germany
    Search for more papers by this author
  • Anna F. Cord

    1. Department of Computational Landscape Ecology, UFZ – Helmholtz Centre for Environmental Research, Leipzig, Germany
    Search for more papers by this author


  1. Macroecology has prospered in recent years due in part to the wide array of climatic data, such as those provided by the WorldClim and CliMond data sets, which has become available for research. However, important environmental variables have still been missing, including spatial data sets on UV-B radiation, an increasingly recognized driver of ecological processes.
  2. We developed a set of global UV-B surfaces (glUV) suitable to match common spatial scales in macroecology. Our data set is based on remotely sensed records from NASA's Ozone Monitoring Instrument (Aura-OMI). Following a similar approach as for the WorldClim and CliMond data sets, we processed daily UV-B measurements acquired over a period of eight years into monthly mean UV-B data and six ecologically meaningful UV-B variables with a 15-arc minute resolution. These bioclimatic variables represent Annual Mean UV-B, UV-B Seasonality, Mean UV-B of Highest Month, Mean UV-B of Lowest Month, Sum of Monthly Mean UV-B during Highest Quarter and Sum of Monthly Mean UV-B during Lowest Quarter. We correlated our data sets with selected variables of existing bioclimatic surfaces for land and with Terra–MODIS Sea Surface Temperature for ocean regions to test for relations to known gradients and patterns.
  3. UV-B surfaces showed a distinct seasonal variance at a global scale, while the intensity of UV-B radiation decreased towards higher latitudes and was modified by topographic and climatic heterogeneity. UV-B surfaces were correlated with global mean temperature and annual mean radiation data, but exhibited variable spatial associations across the globe. UV-B surfaces were otherwise widely independent of existing bioclimatic surfaces.
  4. Our data set provides new climatological information relevant for macroecological analyses. As UV-B is a known driver of numerous biological patterns and processes, our data set offers the potential to generate a better understanding of these dynamics in macroecology, biogeography, global change research and beyond. The glUV data set containing monthly mean UV-B data and six derived UV-B surfaces is freely available for download at:


The complex and diverse effects of ultraviolet-B radiation (UV-B, wavelengths of 280–315 nm) on the physiology, distribution and population dynamics of numerous organisms are being increasingly recognized by studies that examine both natural and agricultural systems (see Fig. 1 and Table 1 for a summarized overview of UV-B effects on biological systems). Recent investigations of the environmental role of UV-B radiation (Jansen & Bornman 2012; Wargent & Jordan 2013) raised awareness of the possible negative, positive and regulatory impacts UV-B has on many organisms. Yet, the ecological role of UV-B has rarely been addressed in macroecological research compared to other climate conditions.

Table 1. Summary table of UV-B-associated effects on different taxonomic groups. References cited within Table 1 can be found in Appendix S1
Taxonomic groupUV-B-associated effectStudies
MicrobesReduces microbial biodiversityKotilainen et al. (2009); Ballaré et al. (2011)
Alters taxonomic composition of bacterioplanktonManrique et al. (2012)
Affects coral–algal symbiosisHäder et al. (2011)
Reduces rate of litter decompositionPancotto et al. (2003)
Alters C:N ratioPancotto et al. (2003)
Changes C-storage in the ecosystemBallaré et al. (2011); Pancotto et al. (2003)
PlantsReduces aboveground biomassCaldwell et al. (2007)
Decreases plant heightCaldwell et al. (2007)
Alters leaf expansion and morphologyCaldwell et al. (2007)
Damages DNA, proteins and cellular compoundsParsons & Fry (2010); Björn (1996); Brosché & Strid (2003)
Reduces survival rateCaldwell et al. (2007)
Alters root lengthZaller et al. (2002)
Reduces photosynthetic yieldXiong & Day (2001); Jansen, Martret & Koornneef (2010)
Alters yield of agricultural plantsKoti et al. (2005); Caldwell et al. (2007); Mazza et al. (2012)
Alters the synthesis of shielding compoundsHarborne & Williams (2000); Jenkins (2009); Kuhlmann & Müller (2010)
Increases foliar hairsBeckmann et al. (2012); Skelton et al. (2012)
InvertebratesCauses genetic damage (larval krill) and alters dietary preference for algae with UV-B-absorbing amino acids (krill)Jarman et al. (1999); Dahms, Dubretsov & Lee (2011); Newman et al. (2000)
Damages DNA in embryos (sea urchins) and reduces biomass (benthic stream invertebrates)Lesser, Kruse & Barry (2003); Kelly, Bothwell & Schindler (2003)
VertebratesAlters habitat preferences for sites with lower UV-B radiation (frogs)Han et al. (2007)
Changes food preference (plantivorous fish)Häder et al. (2011)
Leads to skin lesions and altered pigmentation (whales)Martinez-Levasseur et al. (2011)
Increases embryo mortality and reduces larval survival (amphibians)Häkkinen, Pasanen & Kukkonen (2001); Broomhall, Osborne & Cunningham (2000)
Changes in immune functions (fish)Jokinen et al. (2008); Markkula et al. (2009)
Changes in calcium metabolism (birds)Stanford (2006)
Figure 1.

Conceptual model illustrating the effects of UV-B radiation on organisms in terrestrial and marine environments. The potential effects include direct detrimental effects, morphological and physiological changes, altered behaviour and site selection, and changes in diversity and distribution.

Climate conditions have long been recognized as major drivers of ecological processes and biogeographical distributions of organisms (Andrewartha & Birch 1954; Woodward 1987). As the role of climate factors is relevant particularly at continental and global scales (Pearson & Dawson 2003), globally continuous climate data sets are of increasing importance both for current macroecological studies and biogeographical modelling. The development of gridded climate surfaces based on temperature and precipitation measurements using spatial interpolation techniques (e.g. Hutchinson 1995) opened a new field of research within ecology. In the last decade, a surge of studies used global climate data to address a variety of ecological topics, including the conservation of endangered species (Pearson 2010; Wilting et al. 2010; Falcucci et al. 2013), prediction of biological invasions (Ibanez et al. 2009; Václavík et al. 2010; Ireland, Hardy & Kriticos 2013) or assessment of climate change impacts (Hijmans & Graham 2006; Randin et al. 2009; Tabor & Williams 2010).

Most climate data sets that are used to address macroecological questions (e.g. WorldClim, Hijmans et al. 2005; PRISM, Daly et al. 2000) are based on records from climate stations, which are recalculated into bioclimatic variables. For example, the recently published CliMond data base (Kriticos et al. 2012) contains 35 bioclimatic variables developed specifically for use in species distribution models. These climate surfaces are an indispensable source of information both for correlative and process-based approaches. Nevertheless, there is still room to augment these products. Most bioclimatic variables are created by combining temperature and precipitation variables and are therefore highly correlated, although recent efforts included data on radiation and soil moisture (Hutchinson et al. 2009; Kriticos et al. 2012). Furthermore, they are typically restricted to terrestrial environments and are based on interpolation, thus exhibiting higher levels of uncertainty in areas with low density of monitoring stations (Soria-Auza et al. 2010). To further advance our understanding of macroecological patterns and processes and to complement the existing climatologies, more ecologically relevant predictors are needed (Elith & Leathwick 2009). Ideally, these should cover both terrestrial and marine environments and contain new information not yet captured in existing climate data sets.

UV-B radiation, identified as being the biologically most relevant ultraviolet radiation (Madronich et al. 1998; Fig. 1), holds a great potential as a complementary ecological predictor. For example, increased UV-B radiation is known to alter photosynthetic capacity and reduce biomass production in plants (e.g. Xiong & Day 2001; Jansen, Martret & Koornneef 2010). UV-B radiation can also cause DNA and protein damage (Björn 1996; Brosché & Strid 2003), although some plant species are adapted to counteract the resulting stress by increasing the production of antioxidant enzymes (Jenkins 2009; Kuhlmann & Müller 2010) or by adapting leaf morphology (Yang, Yao & He 2008; Beckmann et al. 2012). In agricultural systems, UV-B radiation affects the yield of many crops (Koti et al. 2005; Caldwell et al. 2007) or fodder plants (Hofmann & Campbell 2012) and influences decomposition processes by altering microbial diversity in the soil (Ballaré et al. 2011). UV-B radiation also impacts the behaviour of a wide range of animals both in the terrestrial and marine environment. For instance, many amphibians avoid sites with high UV-B radiation because such exposure can kill them directly or cause sublethal effects such as slowed growth rate or immune dysfunctions (Anzalone, Kats & Gordon 1998; Blaustein et al. 1998, 2003; Belden & Blaustein 2002; Han et al. 2007). In marine environments, UV-B radiation drives the dynamics of larval krill, affects the coral–algal symbiosis and alters bacterial assemblages impacting higher trophic levels (Häder et al. 2011). These examples illustrate the crucial role UV-B radiation plays in ecological systems but, to date, globally conformal UV-B data have not been readily available for macroecological analyses (but see Lee-Taylor et al. 2010).

The levels of solar UV-B radiation have been monitored extensively in the last decades, using both ground-based and space borne instruments such as the Brewer spectrophotometer or the Total Ozone Mapping Spectrometer (TOMS), respectively. Ground-based monitoring stations alone are not able to cover the long-term trends in UV-B radiation at the global scale but satellite-based measurements, although affected by scattered reflections from the atmosphere and earth surface, can provide a global coverage of spatial data on UV-B radiation (Tanskanen et al. 2007). Since 2004, when the AURA satellite was launched, the Ozone Monitoring Instrument (OMI) has become a major source of UV-B information. Its validation with ground reference data showed that OMI provides reliable measurements at relatively high spatial and temporal resolutions (e.g. Cabrera et al. 2012; Krzyścin et al. 2012).

In an effort to fill the gap in availability of global UV-B data, we develop and test a global data set of surface UV-B radiation (glUV) that can be readily used for macroecological analyses. Our data set has a spatial resolution of 15 arc minutes and is based on daily measurements of remotely sensed UV-B levels collected by the Aura-OMI satellite mission in the period 2004–2013. We process and recalculate the measurements to derive six ecologically relevant UV-B variables, comparable to the existing sets of bioclimatic surfaces. To demonstrate the unique information that is captured by glUV but not contained in other existing climatologies, we assess our data for global and local correlations with selected variables from WorldClim and CliMond data bases as well as with sea surface temperatures for the marine environments. Finally, we highlight the relevance and applicability of our data for macroecological analyses. The data set glUV is freely available for download at:


Data source

We used UV-B data from the Ozone Monitoring Instrument (OMI) onboard the NASA EOS Aura spacecraft. OMI is a contribution of the Netherlands' Agency for Aerospace Programs (NIVR) in collaboration with the Finnish Meteorological Institute (FMI) to the EOS Aura mission of NASA. Since October 2004, OMI continues the Total Ozone Mapping Spectrometer (TOMS) record for total ozone and other atmospheric parameters with improved sensitivity and higher spatial resolution (Schoeberl et al. 2006). The Aura spacecraft orbits at 705 km in a sun-synchronous orbit with a 1:45 PM ±15-minute equator crossing time. OMI is a nadir-viewing instrument with a 2,600-km-wide-viewing swath and provides daily global coverage of the sunlit portion of the atmosphere. The spatial resolution of the instrument is 13 × 24 km in nadir and larger towards the edges of the swath (Tanskanen et al. 2006). OMI contains two spectrometers and measures reflected solar radiation in a selected range of the visible (350–500 nm) and UV spectrum (two bands at 270–314 and 306–380 nm, Levelt et al. 2006). The OMI measurements are used to calculate clear-sky surface UV irradiance, which is subsequently corrected for clouds and aerosols to obtain the OMI surface UV irradiance products (OMUVB, Tanskanen et al. 2006). The accuracy of the OMUVB products, assessed with ground reference data, ranges from 70% to 93%, depending on atmospheric and location-specific conditions (Tanskanen et al. 2006).

Data acquisition and processing

We acquired the Spectral Surface UV-B Irradiance and erythemal dose level-2G data product (OMUVBG, V003) at 15 arc-minute spatial resolution from the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC) by employing the Mirador interface ( The erythemal dose is an erythemally weighted estimate of daily UV-B radiation over the wavelengths measured by the OMI instrument given in J m² (for reasons of simplicity we use J m−2 d−1 for all derived data). We used all available data from October 1st 2004 to January 26th 2013 and converted them from the HDF5 format into ASCII files using the hd5dump tool ( provided by the HDF Group at the University of Illinois. All subsequent analyses were performed in R version 2.15.1 (R Core Team 2013). Since January 2009, some cross-track positions of the OMI sensor have been affected by a row anomaly that comprises a shift in the radiance signal depending on the position in the orbit. As the valid range for this product is between 0 and 11 000 J m−2, we flagged all other values as no data. From these data, we assessed the percentage of valid observations per pixel as a spatially explicit estimate of the sample size of our data set. For most areas, on average, more than 50% of the days within our study period provided valid observations. Due to the cyclical repeating overflights and the consequently shifting swath patterns, valid measurements were available intermittently but typically at least every other day (Fig. S1). With the exception of lower sample sizes towards the poles, where solar insolation is absent half of the year, we did not find any indication of spatial bias in available measurements (Fig. S1). We summarized the daily measurements into monthly mean UV-B erythemal daily dose values and averaged them across all years in the study period to account for inter-annual variability.

UV-B variables

We derived six biologically meaningful UV-B variables (Table 2) from the monthly mean layers, following methods similar to those used for the WorldClim (Hijmans et al. 2005) and CliMond (Kriticos et al. 2012) data sets. Our variables represent Annual Mean UV-B (UVB1), UV-B Seasonality (UVB2), Mean UV-B of Highest Month (UVB3), Mean UV-B of Lowest Month (UVB4), Sum of Monthly Mean UV-B during Highest Quarter (UVB5) and Sum of Monthly Mean UV-B during Lowest Quarter (UVB6). UV-B Seasonality is calculated as standard deviation of mean monthly values. The calculations of Sum of Mean UV-B during Highest/Lowest Quarter (UVB3 and UVB4) are based on the same algorithm used for computing the bioclimatic variables BIO16 and BIO17 (Precipitation of Wettest/Driest Quarter) in WorldClim (‘biovars’ method in the R package dismo, Hijmans et al. 2012). This algorithm computes all possible 12 combinations of quarterly sums including the ones around the turn-of-the-year and identifies the highest/lowest among these. For UVB3 to UVB6, the respective point in time (month or quarter, respectively) to which these variables refer is therefore spatially heterogeneous and may be different among adjacent cells. For reasons of consistency, the same units (J m² d−1) were used for all UV-B surfaces including UVB5 and UVB6.

Table 2. List of UV-B variables available in the glUV data set. All UV-B surfaces are based on daily satellite measurements conducted between 2004 and 2013 and are given in [J m−2 d−1]
UVB-Jan-DecMonthly Mean UV-B (12 layers)
UVB1Annual Mean UV-B
UVB2Annual UV-B Seasonality (standard deviation)
UVB3Mean UV-B of Highest Month
UVB4Mean UV-B of Lowest Month
UVB5Sum of Monthly Mean UV-B during Highest Quarter
UVB6Sum of Monthly Mean UV-B during Lowest Quarter

Comparison with existing climate data sets

To illustrate the potential relevance of the glUV data set and assess its additional information content, we tested the collinearity of all six UV-B variables with selected temperature, precipitation and solar radiation factors from the WorldClim and CliMond data sets for land and with Terra–MODIS Sea Surface Temperature (SST) for sea areas. We selected pairs of variables (Table 3) that allowed for meaningful (i.e. to not pair a sum with a standard deviation) and ecologically relevant comparisons. For example, we used no precipitation measure for the oceanic areas because water availability is not a limiting factor in this environment and hence not relevant to macroecological analyses. Also, to maintain the statistical independence of correlation tests, we selected only variables which were not collinear among themselves (i.e. certain values are derived from other variables in each data base). This careful selection of variable pairs for which correlation is tested also ensures that the Type I error rate is kept to its nominal rate (0·05). For each pair of a UV-B and climate variable, we calculated the Pearson's correlation coefficients (r) and corrected the respective significance tests for spatial autocorrelation (Dutilleul et al. 1993), as implemented in the R package SpatialPack (Osorio, Vallejos & Cuevas 2012). The number of cells to calculate Moran's I index (used to model the spatial autocorrelation) was selected based on Sturge's formula. For ocean areas, we removed missing temperature data found in Antarctica and along the coasts of the continents.

Table 3. Pairwise correlations for selected glUV and other bioclimatic variables. Significance tests were corrected for spatial autocorrelation
 Corr n k adj d.f.F-valueP-value
  1. For UV-B variables, see Table 2; SST = Sea Surface Temperature, BIO1 = Annual Mean Temperature, BIO4 = Temperature Seasonality, BIO15 = Precipitation Seasonality, BIO16 = Precipitation of Wettest Quarter, BIO17 = Precipitation of Driest Quarter (Hijmans et al. 2005), BIO20 = Annual Mean Radiation, BIO23 = Radiation Seasonality (Kriticos et al. 2012).

Terrestrial area
BIO1, UVB10·85247 0173613·0832·89<0·001
BIO4, UVB20·19247 0173650·601·970·1664
BIO15, UVB2−0·04247 0173671·050·0890·766
BIO16, UVB50·28247 0173639·633·310·0765
BIO16,UVB60·52247 0173633·8312·580·0012
BIO17, UVB5−0·07247 0173656·880·270·6039
BIO17,UVB60·07247 0173649·340·220·6397
BIO20, UVB10·82245 1553615·5732·18<0·001
BIO23, UVB20·06245 1553652·170·160·6910
Marine area
SST, UVB10·94668 4103818·93143·22< 0·001

In addition to global correlations and especially for cases where |r| > 0·7 (Dormann et al. 2013), we assessed the spatially explicit association of our UV-B variables with the selected climate data using the bivariate Moran's I metric. This method, also known as local indicator of spatial association (LISA; Anselin 1995), represents a local version of the correlation coefficient and shows how the nature and strength of the association between two variables varies across the study area (see Appendix S2 for additional information). Using OpenGeoDa version 1.2.0 (Anselin, Syabri & Kho 2006), we identified and mapped spatial associations of (i) high-high values, that is hot spots in which locations with high values of UV-B are surrounded by high values of a respective climate variable, (ii) low-low values, that is cold spots in which locations with low values of UV-B are surrounded by low values of another climate variable and (iii) high-low and low-high values, where the spatial association between the variables is negative (inverse). The strength of the relationship was measured at the 0·05 level of statistical significance calculated by a Monte Carlo randomization procedure based on 499 permutations. Associating UV-B values with climate factors in the neighbouring cells is important because simple cell overlap (used to calculate global correlations) can be affected by differences in spatial resolution, misregistration issues or noise in the data.


UV-B variables

The twelve monthly mean UV-B erythemal daily dose values (see Fig. 2 for quarterly examples from January, April, July and October) illustrate the pronounced global intra-annual dynamics of UV-B intensities. While UV-B intensity generally decreases towards higher latitudes, its seasonality increases at the same time, following climate seasons. During winter in the Northern and Southern Hemisphere (January and July, respectively), UV-B irradiation is generally lowest in the mid-latitudes and declines to zero beyond the polar circles where solar insolation is absent during the polar night. However, these seasonal patterns do not simply reflect latitudinal gradients but also show topographic and climatic variations, driven by cloud cover, at the regional scale. In addition, notable differences in the range of UV-B values exist between the Northern and Southern Hemisphere, with generally higher values in the Southern Hemisphere (compare insets in Fig. 2).

Figure 2.

Examples of four monthly mean UV-B layers: January, April, July and October. The two insets on the right show finer-scale variation in the data for the same latitudinal ranges on the Northern and Southern Hemispheres. White areas are caused by the absence of solar radiation due to astronomical polar night and twilight.

At the global scale, mean UV-B irradiation is highest at lower latitudes, as it is illustrated by Annual Mean UV-B (UVB1, Fig. 3a). However, larger high-altitudinal areas such as the Andes and Tibetan Plateau are clearly detectable. Intra-annual UV-B Seasonality (UVB2, Fig. 3b) is generally lowest near the equator and highest in mid-latitudes although many mountain ranges are characterized by a high Annual UV-B Seasonality in comparison to surrounding areas at lower altitudes. Although the Mean UV-B of Lowest Month (UVB4, Fig. 3d) and the Sum of Monthly Mean UV-B during Lowest Quarter (UVB6, Fig. 3f) indicate latitudinal gradients, more local effects are evident within the tropical areas, for example in the Amazon Basin and in Central Africa. The Mean UV-B of Highest Month (UVB3, Fig. 3c) and the Sum of Monthly Mean UV-B during Highest Quarter (UVB5, Fig. 3e) further show that UV-B values can reach similar monthly and quarterly maxima across large regions (approximately between 30°N and 30°S).

Figure 3.

Main UV-B data layers available in glUV: (a) Annual Mean UV-B (UVB1), (b) Annual UV-B Seasonality (UVB2), (c) Mean UV-B of Highest Month (UVB3), (d) Mean UV-B of Lowest Month (UVB4), (e) Sum of Monthly Mean UV-B during Highest Quarter (UVB5), (f) Sum of Monthly Mean UV-B during Lowest Quarter (UVB6).

Comparison with existing climate data sets

The comparison of our UV-B layers with selected existing climate data sets revealed significant global correlation (P < 0·05) between three pairs of layers (Table 3). Significant positive correlation (P < 0·001) was found between Annual Mean UV-B (UVB1) and Annual Mean Temperature (BIO1) with r = 0·846, between Annual Mean UV-B and Sea Surface Temperature with r = 0·940 and between Annual Mean UV-B and Annual Mean Radiation (BIO20) with r = 0·821. For all other comparisons of UV-B and climate data considered, we did not observe significant or strong correlations (i.e. |r|<0·7; Dormann et al. 2013). The scatterplots reported in Fig. S3 confirm the correlation analysis.

The LISA analyses revealed regionally variable spatial associations between UV-B data and selected climate factors (Fig. 4, Fig. S2). For the combination of variables that had low global correlation, the patterns of hot spots, cold spots and negative relationships varied substantially across the globe. However, even for the three pairs of variables with |r| > 0·7, the pattern of local association was highly diverse. The spatial association of Annual Mean UV-B with Annual Mean Temperature (Fig. 4a), Sea Surface Temperature (Fig. 4b) and Annual Mean Radiation (Fig. 4c) exhibited clustering of high-high values in tropical and subtropical regions and low-low values in boreal and (ant)arctic zones. However, large areas of non-significant or negative association dominated especially in mid-latitudes. These statistically non-significant correlations and combinations of high-low and low-high values indicate locations in which the UV-B data provide new information that is not captured by other climate variables.

Figure 4.

Local indicator of spatial association (LISA) between Annual Mean UV-B and Annual Mean Temperature (a), Annual Mean Sea Surface Temperature (b) and Annual Mean Radiation (c). The pattern shows how the nature and strength of the association between two variables varies across the globe. High-high clusters show locations of high values of UV-B being surrounded by high values of a respective climate variable. Low-low clusters show locations of low values of UV-B being surrounded by low values of a respective climate variable. High-low and low-high clusters show inverse spatial association.


The glUV data set provides spatial data on UV-B radiation at 15 arc-minute resolution, covering the entire globe and including both the terrestrial and marine environments. glUV makes bioclimatic UV-B variables as well as monthly mean UV-B data available for a wide range of potential applications to the scientific community. In combination with existing and widely used global climatologies such as the WorldClim (Hijmans et al. 2005) or the CliMond (Kriticos et al. 2012) data sets, glUV provides a more comprehensive set of predictors suitable for macroecological studies and beyond.

Temporal and spatial variability of UV-B

While glUV is congruent with known patterns of global UV-B distribution (e.g. the difference between Northern and Southern Hemisphere), it offers further details on temporal variability and spatial relationships, which were not available as climate surfaces before, yet might prove useful in ecological modelling studies. For example, while high intra-annual seasonality can be found in both hemispheres (Fig. 2), seasonality is generally lower in tropical regions and the Northern hemisphere (Fig. 3b). Although correlations of UV-B with other climatic variables can be positive at the global scale, the local correlations reveal a more complex pattern, showing large areas without significant spatial association (Fig. 4). Altitudinal extremes, for example the Tibetan Plateau, where high UV-B irradiation may be correlated with low temperature values, also demonstrate negative spatial association. Similarly, large ocean currents are altering the expected spatial correlation of Annual Mean UV-B and Mean Sea Surface temperature (Fig. 4b), for example the Humboldt and Benguela currents (high-high correlation) or the Gulf Stream (low-low correlation). glUV also captures the current hemispheric difference in UV-B exposure, with generally higher values in the Southern Hemisphere (Fig. 2). This pattern can be related mostly to the earths' elliptical orbit whereby the closest earth–sun distance occurs in the beginning of January when day lengths are longer in the Southern Hemisphere. The elliptical orbit changes over time, and a complete cycle takes an evolutionary relevant period of ~30 000 years (Hay, DeConto & Wold 1997).

The comparisons of glUV with other climate data bases have further shown significant correlations between several variables, but also some surprising lack of correlation, notably between Annual UV-B Seasonality (UVB2) and Radiation Seasonality (BIO23, Kriticos et al. 2012). While this suggests that the UV-B fraction in the total radiation does not vary as much as the total radiation, it may also hint at discrepancies of the data sources (i.e. satellite and ground-based measurements or models). More research is required to explain this result.

The spatial variability of short-term influences on UV-B exposure, including human-related changes in UV-B intensities, is not directly identifiable within glUV. For example, the effects of ozone depletion are known to have substantially increased UV-B irradiation in specific areas; however, these began much earlier than 2004, when the collection of data used to produce glUV started (Solomon et al. 1986). Similarly, effects of human activities that lead to decreasing UV-B irradiation (Herman et al. 1997) may not be detected in glUV. For example, Watanabe et al. (2012) show by hindcasting UV-B irradiation back into pre-industrial times that UV-B has been decreasing across the Northern Hemisphere, while at the same time it increased in the Southern Hemisphere. Therefore, when using glUV, it should be taken into account that the data set presents relatively recent UV-B conditions only.

Potential sources of uncertainty and error within glUV may arise from the temporal and spatial resolution and the varying sample size used to produce the data set. However, between the polar circles (i.e. in areas that are of most interest for biogeographical studies), the samples cover a sufficient range between 50% and 80% of available days. Other limitations of glUV might be attributed to specific effects and problems of the OMI instrument. For example, areas covered by snow and ice are often not distinguishable from clouds in UV-B readings. Therefore, in regions with temporary snow or ice or highly heterogeneous surface albedo, the UV-B irradiance estimates have a higher potential uncertainty (Tanskanen et al. 2006, 2007). In highly populated areas, the OMI instrument, and therefore glUV, may also overestimate UV-B irradiation as the fine-scale distribution of aerosols caused by human activities is not captured by the sensor (e.g. Cabrera et al. 2012). Furthermore, several row anomalies, affecting the quality of the data, have occurred since the start of the AURA mission and resulted in error flagging and recalculation of the OMI data (detailed information on anomalies is available at Despite these limitations, our data set provides the opportunity to address current macroecological questions related to UV-B – organism interactions, which is also underpinned by the outcomes of comparing glUV with other ecological covariate data sets.

Applicability and relevance for future research in ecology and evolution

Some of the key research questions that may be addressed with glUV are how UV-B irradiation influences spatial patterns of population performance and species distribution. It has been recently pointed out that there is a need for more ecologically and functionally relevant covariates beyond temperature and rainfall (Elith & Leathwick 2009) to refine species distribution modelling. Due to its manifold effects on organisms (see Fig. 1 and Table 1 for an overview), UV-B may serve as an environmental factor that links functional response traits to species occurrence patterns (Beck et al. 2012). As demonstrated recently by Skelton et al. (2012) for the Proteaceae in Southern Africa, plant traits related to radiation protection (e.g. leaf pubescence) may be a common phenomenon among such large taxonomically related groups. Whether the presence, absence or shape of such traits helps explain species occurrence in areas of different UV-B irradiation is one of the questions that may be addressed with glUV. Furthermore, linking UV-B radiation to the current or past distribution of species may help uncover similar patterns in the distribution of functional response traits. To understand the role of UV-B in driving species distribution, glUV may be used in explanatory and predictive species distribution modelling. As the glUV layers describing temporal variability are only partly correlated with existing data sets, we think the data set might contribute to improved models of distributions for those species that are affected significantly by direct or indirect UV-B exposure.

Regarding biological invasions in regions of high UV-B intensity, UV-B may act as an environmental factor that enhances the spread of exotic taxa with functional UV-B response traits or that prevents the spread of UV-B-sensitive species. Thereby, future predictions of spreading exotic species may be enhanced and refined by implementing glUV. A recent study suggested that functional UV-B response traits partly aided the invasion of Hieracium pilosella in New Zealand (Beckmann et al. 2012). Correlative species distribution models could usefully explore the utility of glUV in addition to climatic covariates in explaining the distribution patterns of invasive organisms.

Combining glUV and historical data on species or trait distribution may also help modelling micro-evolutionary processes. For example, herbaria records may be used to identify the distribution and development of UV-B response traits in relation to global UV-B irradiation patterns in recent history. However, while addressing the effects of long-term patterns of UV-B intensities (e. g. hemispherical differences) on (micro-)evolutionary processes provides valid and interesting research questions, the glUV data set has to be used with caution, as it covers only the most recent past. Hindcasting methods as presented by Watanabe et al. (2012) may provide more suitable data to address these questions.


The glUV data set represents a potentially useful addition to existing climate data sets that may help refine macroecological analyses and bioclimatic modelling. It has been developed with bioclimatic modelling applications in mind, and its familiar data structure should allow for an easy use by modellers. By adding new information on ecologically relevant conditions, glUV will help address research questions related to the range-limiting capacities that UV-B radiation has on terrestrial and marine life. However, the applicability of glUV is not limited to bioclimatic modelling. We see its potential use, for example, for planning experiments along UV-B gradients or for reanalysing existing species distributions. Studies investigating the intra-annual variability of UV-B radiation may also benefit from implementing the provided monthly mean UV-B data sets used to derive the bioclimatic UV-B variables. glUV may even find applications outside the realms of ecology and biology, for example, in estimating material degradation rates or addressing UV-B-related health issues.


We acknowledge the use of data drawn from the Goddard Earth Sciences Data and Information Center. We thank P. Brandt for help with the initial analysis of the satellite records, D. Kriticos for insightful comments and M. Václavíková for assistance with figure preparation. The project was supported by grant 01LL0901A: Global Assessment of Land Use Dynamics, Greenhouse Gas Emissions and Ecosystem Services–GLUES (German Federal Ministry of Education and Research). MB, AC and TV would like to express their sincere regrets over LS leaving a promising scientific career and joining a monastery instead.

Data accessibility

The glUV data set is accessible online at Raw satellite data used to create the data set (OMUVBG, Level 2, V003) are available online from the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC).