Portal de Periódicos da CAPES

Diel vertical migration of copepods in mountain lakes: The changing role of ultraviolet radiation across a transparency gradient

2014; Wiley; Volume: 60; Issue: 1 Linguagem: Inglês

10.1002/lno.10019

1939-5604

Janet M. Fischer, Mark H. Olson, Nora Theodore, Craig E. Williamson, Kevin C. Rose, Jin Hwan Hwang,

Isotope Analysis in Ecology

Limnology and OceanographyVolume 60, Issue 1 p. 252-262 ArticleFree Access Diel vertical migration of copepods in mountain lakes: The changing role of ultraviolet radiation across a transparency gradient Janet M. Fischer, Corresponding Author Janet M. Fischer Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaCorrespondence: janet.fischer@fandm.eduSearch for more papers by this authorMark H. Olson, Mark H. Olson Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaSearch for more papers by this authorNora Theodore, Nora Theodore Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaSearch for more papers by this authorCraig E. Williamson, Craig E. Williamson Department of Biology, Miami University, Oxford, OhioSearch for more papers by this authorKevin C. Rose, Kevin C. Rose Department of Biology, Miami University, Oxford, OhioSearch for more papers by this authorJin Hwang, Jin Hwang Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaSearch for more papers by this author Janet M. Fischer, Corresponding Author Janet M. Fischer Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaCorrespondence: janet.fischer@fandm.eduSearch for more papers by this authorMark H. Olson, Mark H. Olson Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaSearch for more papers by this authorNora Theodore, Nora Theodore Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaSearch for more papers by this authorCraig E. Williamson, Craig E. Williamson Department of Biology, Miami University, Oxford, OhioSearch for more papers by this authorKevin C. Rose, Kevin C. Rose Department of Biology, Miami University, Oxford, OhioSearch for more papers by this authorJin Hwang, Jin Hwang Department of Biology, Franklin and Marshall College, Lancaster, PennsylvaniaSearch for more papers by this author First published: 31 December 2014 https://doi.org/10.1002/lno.10019Citations: 40AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract The transparency regulator hypothesis (TRH) proposes that water transparency determines the relative importance of visual predation vs. ultraviolet radiation (UVR) in driving zooplankton out of the surface waters during the day. To test this hypothesis, we used a combination of observational and experimental approaches to examine the effects of UVR, photosynthetically active radiation (PAR), fish, temperature, and food resources on the daytime vertical distribution and diel vertical migration (DVM) of the copepod Hesperodiaptomus arcticus in a set of lakes in the Canadian Rocky Mountains. Across lakes, H. arcticus daytime vertical distribution increased with both UVR transparency and depth of food resources, and was not related to PAR, thermal structure, or the presence of fish. We also observed a strong positive relationship between UVR and H. arcticus daytime depth distribution in a single lake that varied in transparency over time. The presence of substantial DVM in H. arcticus in two transparent lakes coupled with the lack of DVM in a less transparent lake indicated a weakening effect of UVR in lower transparency conditions. Finally, we deployed open-bottomed vertical columns constructed of UVR-blocking and UVR-transmitting material overnight in two lakes of contrasting transparency. The following day, we observed significantly more copepods in UVR-blocking columns in the more transparent lake but no UVR effect in the less transparent lake. Collectively, our results provide multiple lines of evidence supporting the TRH by highlighting the changing role of UVR as a driver of DVM across lakes of varying transparency. Zooplankton in lakes and oceans around the world migrate up and down in the water column on a daily basis across complex vertical habitat gradients where light, temperature, food, and predation pressure vary with depth. Daytime and nighttime habitat choices are linked to fitness through effects on growth rates, survival, and reproduction (Lampert 1989; Hays 2003; Van Gool and Ringelberg 2003). In recent decades, the theory of zooplankton diel vertical migration (DVM) has largely focused on two biotic drivers: food availability and predation. According to this paradigm, zooplankton migrate up into the surface waters at night to feed, and back down to deeper depths to avoid visual predators during the day (Hays 2003; Van Gool and Ringelberg 2003). At the same time, there is a growing body of evidence suggesting that abiotic drivers such as damaging ultraviolet radiation (UVR) and temperature can also be important (reviewed by Williamson et al. 2011). The recently proposed transparency regulator hypothesis (TRH) integrates biotic and abiotic drivers of zooplankton DVM into a common conceptual model which postulates that their relative importance is regulated by water transparency (Williamson et al. 2011). Variation in water transparency to both UVR and photosynthetically active radiation (PAR, 400–700 nm, used interchangeably here with visible light) influences the depth distribution of food resources, the depth to which UVR damage can occur, the thermal profile, and the depth of a potential refuge from visually feeding planktivorous fishes. For these vertical gradients, the TRH explicitly differentiates between structural drivers of DVM, such as temperature and food, and dynamic drivers such as UVR and visual predation. Structural drivers create a vertical habitat gradient that does not show any strong or consistent variation over a 24 h period but can change markedly across seasons and locations. Thus, structural drivers vary too little on a diel basis to serve as proximate cues driving DVM, although they are important in determining the depths that promote growth, survival, and reproduction. In contrast, dynamic drivers include factors that show strong and systematic variation over a 24 h period and can thus act as proximate cues to directly drive DVM. The TRH posits that visual predation is the primary dynamic driver of normal (down by day) DVM of crustacean zooplankton in low transparency systems whereas UVR is the primary dynamic driver in high transparency systems. UVR is likely to be a more powerful driver of DVM in high transparency systems because these systems will have higher levels of potentially damaging UVR penetrating deeper in the water column as well as fewer visual predators due to lower productivity (Downing et al. 1990). Prior studies support several important components of the TRH, particularly that DVM is influenced by transparency and that UVR has the potential to be a more important DVM driver than visual predation in some lakes. Zooplankton in more transparent lakes exhibit either a greater DVM amplitude (Dodson 1990), a greater avoidance of surface waters (Alonso et al. 2004; Leech et al. 2005b), or a deeper daytime distribution (Wissel and Ramcharan 2003; Kessler et al. 2008). Experimental studies have demonstrated strong UVR avoidance behavior in some zooplankton taxa (Rhode et al. 2001; Fischer et al. 2006). In land-based mesocosm experiments that manipulated both UVR and visual predators, zooplankton had a deeper depth distribution in response to UVR and no clear downward migration in avoidance of fish (Hansson and Hylander 2009). Rose et al. (2012) found stronger effects of UVR than fish predation on Daphnia vertical distribution in 15 m deep mesocosms incubated in situ. Collectively, these studies provide only modest support for the TRH because they focus on a single or relatively small number of study systems or do not directly test for a UVR effect. Thus, a direct test of the TRH using observations of both structural and dynamic drivers in a larger set of systems that span a broad transparency gradient, along with coupled mechanistic experiments, is needed. In this study, we used a combination of observational and experimental approaches to examine how the role of UVR as a DVM driver varied as function of water transparency for the copepod Hesperodiaptomus arcticus in Canadian Rocky Mountain lakes. First, we quantified changes in the relative importance of UVR vs. other structural and dynamic drivers of DVM in 14 lakes that spanned a broad gradient in transparency. While we compared lakes with and without fish, we focused on fishless lakes to examine the effects of changing UVR transparency in the absence of confounding changes in predation pressure from visually feeding planktivores (Kessler et al. 2008). Second, we examined changes in the vertical distribution of H. arcticus in a single lake that varied strongly in transparency within a single season. Third, we quantified the magnitude of DVM through day and night sampling in three lakes that differed in transparency and in the presence vs. absence of fish. Fourth, we conducted a set of in situ experiments to provide a mechanistic test of the role of UVR as a DVM driver in two fishless lakes that differed in transparency. Collectively the results of these multiple approaches provide support for the TRH and demonstrate an increasing importance of UVR in driving the daytime vertical distribution and DVM of copepods with increasing water transparency. Methods Study sites We examined factors influencing the vertical distribution of H. arcticus in a set of lakes in the Central Canadian Rockies Ecosystem (Table 1). All lakes are within the boundaries of Banff and Yoho National Parks except for Sunlite Lake, which was less than one kilometer west of the Yoho park boundary. Elevations ranged from 1644 m to 2423 m above sea level, surface areas ranged from 0.028 km2 to 2.75 km2, and maximum depths ranged from 6 m to 51 m (Table 1). Bow, Moraine, O'Hara, and Redoubt all had resident populations of naturalized or native salmonid species, and the other study lakes were fishless. Three of these currently fishless lakes (Herbert, Annette, and Pipit) had been previously stocked with fish and have recovered to their naturally fishless state. The copepod H. arcticus is generally the top invertebrate predator in our study lakes, except for a small number of lakes where amphipods have been observed. Table 1. Location and characteristics of study lakes. Latitude and longitude are in decimal degrees north and east, respectively. Sample date refers to date of sample used in cross-lake analysis. Z1%320 is the estimated depth where UV320 irradiance is 1% of the subsurface irradiance. Lake Latitude Longitude Elevation (m) Area (km2) Zmax (m) Fish Kd320 (m−1) Z1%320 (m) KdPAR (m−1) Sample date Amiskwi 51.58 116.63 2088 0.060 12.9 No 0.575 8.0 0.270 23 Jul 2008 Annette 51.35 116.20 1965 0.053 11.7 No 0.306 15.1 0.175 21 Aug 2013 Bow 51.67 116.45 1940 2.750 51.0 Yes 0.834 5.5 0.344 03 Sep 2013 Eiffel 51.32 116.23 2271 0.135 10.6 No 0.148 31.1 0.120 31 Jul 2009 Hamilton 51.45 116.57 2140 0.050 13.3 No 1.238 3.7 0.600 21 Jul 2008 Herbert 51.45 116.13 1644 0.057 11.4 No 2.470 1.9 0.317 08 Aug 2012 Moraine 51.30 116.18 2072 0.413 20.7 Yes 0.458 10.1 0.203 14 Sep 2013 O'Hara 51.35 116.32 2014 0.344 35.9 Yes 0.671 6.9 0.209 13 Sep 2013 Oesa 51.35 116.28 2285 0.162 35.4 No 0.414 11.1 0.211 28 Jul 2008 Opabin 51.33 116.32 2270 0.032 10.9 No 1.261 3.7 0.471 25 Jul 2008 Pipit 51.62 115.68 2210 0.080 20.4 No 0.665 6.9 0.159 30 Jul 2008 Redoubt 51.47 116.07 2393 0.191 8.3 Yes 0.851 5.4 0.237 13 Aug 2013 Sentinel 51.32 116.22 2423 0.028 6.0 No 1.557 3.0 0.312 01 Aug 2009 Sunlite 51.60 116.68 2285 0.051 24.7 No 0.539 8.5 0.230 24 Jul 2008 Cross-lake comparison For our cross-lake comparison, we sampled 14 lakes at mid-day in summer (June to September) between 2008 and 2013 (Table 1). At each lake, we collected zooplankton by towing vertically at discrete depth intervals using a closing bongo net consisting of two Wisconsin style (opening size reduced for increased collection efficiency) plankton nets with 0.1 m diameter openings and 48 μm mesh. Most lakes were sampled at one or two meter depth intervals through the water column at or near the deepest part of the lake. Depth intervals for sampling were larger in deeper lakes, such as Sunlite Lake and Lake Oesa. Zooplankton from both nets were combined and immediately preserved in 90% ethanol for later enumeration under a dissecting microscope. For each lake, we used weighted mean depth (in m) to represent the vertical distribution of H. arcticus (Worthington 1931). This value is calculated by first multiplying the proportions of H. arcticus individuals (copepodids plus adults) collected in each depth interval (relative to the total number of individuals collected) by the mid-depth of the interval. The resultant values for each depth are then summed across all depths to determine the average depth of all H. arcticus collected in a lake. Vertical profiles of underwater light were collected using a BIC Cosine submersible ultraviolet (UV) radiometer (Biospherical Instruments). The BIC is a medium bandwidth radiometer with both a submersible and a deck reference cell that measure UV irradiance (in units of μWatts cm−2) at 0.25 s intervals in wavelength bands centered at 320 nm and 380 nm with a bandwidth of 8–10 nm at 50% of peak response. In addition, the BIC measured irradiance of PAR (400–700 nm) in μmol m−2 s−1. Diffuse attenuation coefficients (Kd) for 320 nm, 380 nm, and PAR were estimated as the slope of the natural log of irradiance vs. depth. The domains of these regressions were constrained to the log-linear portion of the relationship, beginning just below depths where wave focusing or boat shadows affected measurements and ending before irradiance levels reached detection limits. R2 values were greater than 0.95 for all regressions and averaged 0.998. We also measured chlorophyll a (Chl a) fluorescence (as a proxy for algal biovolume) and temperature as a function of depth using profiling sondes. For lakes sampled from 2008 to 2012, we used a Turner Designs C6 multisensor platform, and in 2013 we used a Yellow Springs Instruments Exo 2 Sonde. Both instruments measured fluorescence and temperature at one second intervals as they were lowered at a rate of ∼ 30 s per m through the entire water column. Chl a fluorescence measurements were averaged for 0.5 m depth intervals, and the sum of proportions of total Chl a fluorescence at each interval times the mid depth of the interval was used to calculate the weighted mean Chl a fluorescence depth. Because all of our study lakes exhibited deep chlorophyll maxima, nonphotochemical quenching in the surface waters should have minimal effects on estimates of weighted mean Chl a fluorescence depth (Hamilton et al. 2010). We used forward selection multiple regression to examine the cross-lake relationships between weighted mean H. arcticus depth and the environmental variables Kd320, KdPAR, weighted mean Chl a fluorescence depth, and five measures of water temperature calculated from thermal profiles. Three of these measures (maximum temperature, minimum temperature, and the difference between maximum and minimum temperature) were used to represent the strength of the thermal gradient. Because many of our lakes were only weakly stratified, we were not able to calculate traditional thermal metrics such as thermocline depth and instead estimated the depth at which temperature was 5% below maximum and the depth at which temperature was 5% greater than the minimum. Associations between significant predictor variables and weighted mean H. arcticus depth were also evaluated by Person product-moment correlation. To determine whether the presence of fish affected H. arcticus weighted mean depth, we used a two sample t-test to compare mean residuals from the final model between lakes with and without fish. Seasonal study of Lake Oesa We sampled Lake Oesa weekly at mid-day during the 11 week period from 16 July 2013 to 24 September 2013 to examine short-term changes in vertical distribution of H. arcticus within a single system that varied strongly in transparency due to fluctuations in turbidity associated with precipitation events that washed in glacial flour from the surrounding watershed. Sampling protocols were the same as described above for the cross-lake comparison. Similarly, we used forward selection multiple regression and correlation to evaluate factors influencing weighted mean H. arcticus depth among sample dates in Lake Oesa. Because Kd320 was the only environmental variable significantly correlated with weighted mean H. arcticus depth (see Results), we conducted additional analyzes to examine the effects of variation in incident UVR due to cloud cover and seasonal changes in solar angle. UVR exposure at a particular depth depends on water transparency and incident UVR. For each sample date, we incorporated both of these environmental variables into a single metric using the equation Ez = E0e−Kd320Z, where Ez is UV320 irradiance at depth Z (in μWatts cm−2), E0 is UV320 irradiance at the surface, and Kd320 is the attenuation coefficient at 320 nm. We used mean surface irradiance during the UVR profile to represent E0 and calculated Ez for one meter depth increments from 1 m to 10 m. By calculating irradiance at depth rather than using direct measurements, all estimates were based on the same surface irradiance averaged over duration of the profile. We then compared these measures of irradiance at depth with Kd320 alone to determine if they were better predictors of weighted mean depth of H. arcticus. Day and night vertical distributions of H. arcticus On the same date as our daytime collection, we also quantified the nighttime vertical distribution of H. arcticus by sampling at least one hour after sunset in three study lakes: Annette, Moraine, and Herbert. Each lake represented a different set of environmental conditions because Annette and Moraine were highly UVR transparent whereas Herbert had low UVR transparency, and Moraine had fish whereas Annette and Herbert were fishless (Table 1). In Annette and Moraine, we also collected zooplankton and measured surface irradiance at dusk (defined as 20–30 min prior to sunset). Day and dusk surface irradiance values are presented in Table 2. To evaluate whether H. arcticus depth distributions changed between day and night (and dusk in Annette and Moraine), we compared cumulative distributions as a function of depth between two samples using Kolmogorov–Smirnov (K–S) tests. Table 2. UV320 and PAR irradiance in three lakes sampled for day and night vertical distributions of H. arcticus. In Lake Annette and Moraine Lake, irradiance was measured during the day and at dusk. In Herbert Lake, irradiance was only measured during the day. na is "not applicable." UV320 (μWatts cm−2) PAR (μmol m−2 s−1) Lake Day Dusk Day Dusk Annette 17.28 1.08 1250.5 45.6 Moraine 7.96 0.84 293.6 54.5 Herbert 15.74 na 970.8 na UVR avoidance experiments To examine the role of UVR as a driver of vertical migration in H. arcticus, we conducted a pair of short-term field experiments in Lake Annette and Herbert Lake, two fishless lakes with contrasting transparency (Table 1). We used open-bottomed cylindrical columns (0.3 m diameter, 800 mm long) constructed of plastic material that either transmitted or blocked UVR to create ambient UVR and UVR-shielded treatments. Ambient UVR columns were made with Aclar® (Honeywell), which transmits 100% of PAR and 99% of ultraviolet A (UVA, 320–399 nm), and 98% of ultraviolet B (UVB, 295–319 nm) through water. UVR-shielded columns were made with Courtgard® (CPFilms), which transmits 95% of PAR but only 9% of UVA and no UVB and has a sharp wavelength cutoff at 400 nm. At the bottom of each column, we attached a 0.32 m long skirt of clear polyethylene material which was encircled by a drawstring that could be pursed shut when the experiment was terminated. Polyvinyl chloride (PVC) rings filled with sand attached at the top and bottom held the columns open during deployment. In each lake, we deployed three replicates of ambient UVR and UVR-shielded treatments in random order along an east-west linear array. Each replicate was suspended vertically with the top positioned ∼ 0.4 m below the water surface from a float line. Columns were deployed in the afternoon of 08 August 2012 in Herbert Lake and 11 August 2012 in Lake Annette and allowed to sit overnight. Because the columns were open at the bottom, zooplankton were able to enter during their nighttime migration upward. The following morning between 10:30 h and 11:00 h we pursed each column shut and tied the bottoms off to enclose any zooplankton that remained within the cylinder. Skies had 0% cloud cover on the day of collection for both experiments. We pumped all water from within each column through a 48 μm net to collect the zooplankton. All zooplankton collected from a column were immediately preserved in 90% ethanol for later enumeration under a dissecting microscope. Differences in mean number of H. arcticus per column in ambient UVR and UVR-shielded treatments were analyzed with a two sample t-test. One replicate was lost from the ambient UVR treatment in the Lake Annette experiment due to experimenter error. Results Cross-lake comparison The daytime vertical distribution of H. arcticus varied among our 14 study lakes from weighted mean depths of 1.62 m in Sentinel Lake to 14.95 m in Sunlite Lake. Forward selection multiple regression using the set of eight environmental variables to predict weighted mean H. arcticus daytime depth produced a significant final model (F2,11 = 11.09, p = 0.002), and included Kd320 (p = 0.016) and weighted mean Chl a fluorescence depth (p = 0.015) as significant predictors. Because Kd320 and weighted mean Chl a fluorescence depth were not correlated across lakes (r = −0.26, n = 14, p = 0.39), these two predictor variables had independent effects on H. arcticus vertical distribution. Importantly, KdPAR was not included in the final model despite low colinearity with the significant predictor variables (tolerance = 0.69). Across lakes, weighted mean depth was negatively correlated with Kd320 (r = −0.65, n = 14, p = 0.012), indicating that H. arcticus were deeper as UVR transparency increased (Fig. 1a). H. arcticus depth was also positively correlated with weighted mean Chl a fluorescence depth (Fig. 1b, r = 0.65, n = 14, p = 0.011). Both of these correlations remained significant after Bonferroni correction for multiple comparisons. Mean residuals of the multiple regression model did not differ between fish and fishless lakes (t-test, t = 0.74, degrees of freedom [df] = 12, p = 0.48), indicating that fish did not influence vertical distribution (Fig. 1). Figure 1Open in figure viewerPowerPoint Weighted mean H. arcticus depth vs. (a) the attenuation coefficient Kd320, and (b) weighted mean Chl a depth. Each point represents one lake sampled at mid-day. Seasonal study of Lake Oesa Transparency in Lake Oesa in 2013 was highly dynamic, with Z1%320 (the estimated depth where UV320 irradiance is 1% of the subsurface irradiance) varying from 8.2 m to 19.8 m over the course of the ice-free season. A forward selection multiple regression with the eight environmental variables included as potential independent variables was significant (F1,9 = 5.10, p = 0.05), and included Kd320 as the only significant predictor of weighted mean H. arcticus daytime depth. As observed in the cross-lake comparison, weighted mean H. arcticus depth was negatively correlated with Kd320 over the 11 sample dates (Fig. 2a, r = −0.60, n = 11, p = 0.05). However, the correlation with weighted mean Chl a fluorescence depth was not significant (r = 0.52, n = 11, p = 0.10). Figure 2Open in figure viewerPowerPoint Weighted mean H. arcticus depth vs. (a) Kd320, and (b) UV320 irradiance at a depth of five meters. Each point represents one mid-day sample for Lake Oesa in 2013. Across the 11 sample dates, weighted mean H. arcticus daytime depth was highly correlated with UV320 irradiance at five meters (Fig. 2b, r = 0.91, n = 11, p = 0.0001), a result that is robust to Bonferroni correction for multiple comparisons across ten depths. We observed similar strong relationships with UV320 irradiance at 4 m and 6 m; however, the correlation weakened at both shallower and deeper depths (Fig. 3a). In the top three meters of the water column, irradiance at depth was significantly correlated with surface irradiance whereas irradiance at depths greater than six meters was significantly correlated with Kd320 (Fig. 3b). Between 4 m and 6 m, irradiance was not correlated with either surface irradiance or Kd320 and therefore reflects the combined effects of both variables. Figure 3Open in figure viewerPowerPoint Correlation coefficients (r) between UV320 irradiance at a given depth and (a) surface irradiance on the left axis and Kd320 on the right axis and (b) weighted mean H. arcticus daytime depth. Each correlation was based on 11 sample dates from Lake Oesa in 2013. Absolute values of r greater than 0.6 (shown with the dotted line) were significant at α = 0.05. Correlations between UV320 irradiance at a given depth and Kd320 in panel a were plotted on a reversed scale. Day and night vertical distributions of H. arcticus In Lake Annette, a highly UVR transparent fishless lake, we observed a strong upward shift in vertical distribution from day to night on 21 August 2013 (Fig. 4a). Weighted mean H. arcticus depth was 8.7 m during the daytime, and no individuals were collected in the top six meters. In contrast, weighted mean depth was 4.1 m at night and 84% of individuals were collected in the top six meters. Cumulative distributions as a function of depth differed significantly between the two time periods (K–S test, Dmax = 0.87, p < 0.001). At dusk, weighted mean H. arcticus depth was 5.3 m, and 60% of individuals were collected in the top six meters (Fig. 4a). The cumulative distribution at dusk differed from the daytime distribution (K–S test, Dmax = 0.79, p < 0.001) but not the nighttime distribution (K–S test, Dmax = 0.30, p > 0.05). Figure 4Open in figure viewerPowerPoint Cumulative distributions of H. arcticus as a function of depth (in m) in (a) Lake Annette, a high transparency fishless lake; (b) Moraine Lake, a high transparency lake with fish; and (c) Herbert Lake, a lower transparency fishless lake. In Lake Annette and Moraine Lake, samples were collected at mid-day, dusk, and at night after sunset on 21 August 2013 and 14 September 2013, respectively. In Herbert Lake, samples were collected at mid-day and at night on 08 August 2013. Cumulative distributions represent summed proportions of the total number of H. arcticus collected in all samples with each successive depth interval from the surface to the bottom. Changes in vertical distribution from day to night in Moraine Lake (a highly UVR transparent lake with fish) were similar to Lake Annette (Fig. 4b). Weighted mean depth was 13.8 m in the day and 3.9 m at night, and cumulative distributions of the two samples differed significantly (K–S test, Dmax = 0.64, p < 0.001). As we observed in Lake Annette, H. arcticus completed most of their upward migration by dusk (Fig. 4b). Weighted mean depth at dusk was 5.4 m, and the cumulative distribution was significantly different from the daytime distribution (K–S test, Dmax = 0.61, p < 0.001) but not the nighttime distribution (K–S test, Dmax = 0.37, p > 0.05). In contrast to the diel changes observed in Annette and Moraine, H. arcticus did not change vertical distribution in Herbert Lake (a fishless lake with lower UVR transparency) on 08 August 2012 (Fig. 4c). Cumulative distributions did not differ between the daytime and nighttime samples (K–S test, Dmax = 0.59, p > 0.05). UVR avoidance experiments Patterns of UVR avoidance by H. arcticus in our in situ column experiments differed between our two study lakes. In Lake Annette, where UVR transparency was high, no individuals were collected in any of the ambient UVR columns whereas the UVR-shielded columns had 30.67 ± 8.83 ( X ¯ ± 1 standard error [SE]) individuals (Fig. 5). Ln-transformed mean densities in the two treatments differed significantly (t-test, t = 7.32, df = 3, p = 0.005). In Herbert Lake, where UVR transparency was lower, we found no evidence of UVR avoidance. H. arcticus were found in both ambient UVR and UVR-shielded columns (Fig. 5), and means did not differ between treatments (t-test, t = 0.61, df = 4, p = 0.58). Figure 5Open in figure viewerPowerPoint Mean number of H. arcticus collected per mesocosm in Ambient-UVR and UVR-shielded treatments in (a) Lake Annette, a high transparency fishless lake; and (b) Herbert Lake, a lower transparency fishless lake. Error bars represent ±1 SE. Discussion While the literature is replete with detailed case studies of DVM in single systems (Rautio et al. 2003; Jung et al. 2004; Winder et al. 2004), the lack of UVR data in many studies combined with differences in methodology and species composition make it difficult to use a meta-analytic approach to examine changes in the role of UVR or other DVM drivers over a transparency gradient. To our knowledge, there have only been a few previous studies of the relative importance of UVR and predation as DVM drivers