The jellification of north temperate lakes

(a) Have Holopedium abundances increased since pre-industrial times?

We examined long-term changes in the relative abundance of Holopedium within the pelagic cladoceran assemblages in lake sediments from 36 south-central Ontario lakes in 2007 [4], and 48 Nova Scotia lakes in 2002–2004 [5] (see electronic supplementary material, table S1). Sediment cores were collected from the deepest basin of each lake using a gravity corer, and sectioned at 0.25 cm intervals. We then employed the well-established ‘top/bottom’ palaeolimnological sampling design [30], wherein the chitinous cladoceran remains were isolated, identified and compared between a ‘top’ interval (surface sediments) representing present-day assemblages and a ‘bottom’ interval (sediment depth > 15.0 cm) representing assemblages present prior to the onset of lake acidification. Although this protocol cannot document the timing of changes, it clearly separates modern from historical assemblages, which is what the test of our first hypothesis required.

We counted and identified remains from a minimum of 90 individual cladocerans per sample [31,32], of which a mean of 78 and 61% were pelagic taxa in the Ontario and Nova Scotia lakes, respectively. This count is sufficient to characterize the taxonomic assemblage accurately [33]. We tested whether the contribution of Holopedium to planktonic cladoceran assemblages had changed since pre-industrial times using paired t-tests comparing Holopedium relative abundance in the pre-industrial versus recent sediment datasets. We excluded lakes where Holopedium was absent from both intervals.

(b) Are Holopedium increases consistent with recent lakewater Ca decline?

We tested our second hypothesis using two datasets: a longitudinal study of eight Ontario lakes sampled over a 30-year period [34], and a regional survey of 31 other Ontario lakes sampled in the 1980s and 2004 or 2005 [35] (see electronic supplementary material, table S2). Zooplankton were sampled during the ice-free season near the point of maximum depth by using a metered, 76 µm mesh, tow net. The net was deployed in four to seven vertical hauls and its contents were combined to produce a composite that corrected for the diminution of lake stratum volume with depth. A minimum of 250 crustacean zooplankton were enumerated in each composite, subsampling to ensure that no taxon comprised more than 10% of the total count. All Cladocera were identified to the species level [34].

To test whether the relative and absolute abundance of Holopedium had increased in the 31 lake regional dataset, we compared the 1980s versus 2004–2005 ice-free means using, respectively, a paired t-test and a Wilcoxon signed-rank test. The five lakes that were invaded by Bythotrephes between the two sampling periods were analysed separately. For the eight lake longitudinal dataset, we used Mann–Kendall trend tests to identify whether the absolute and relative abundances of Holopedium, daphniids and lakewater Ca concentrations had changed in a monotonic fashion over time when averaged across lakes in each year of the dataset. Prior to averaging absolute abundances, we standardized observations to a mean of 0 and s.d. of 1 across years in each lake.

(c) Discriminating among alternative mechanisms underlying increases in Holopedium abundance

We attempted to distinguish the causal processes underlying the long-term changes in daphniid and Holopedium abundances using ice-free season means from the eight longitudinal study lakes [34,36] (see electronic supplementary material, table S3). These lakes vary in acid-sensitivity [34] and have patterns of environmental change reflecting those of the broader region [17]. We needed composition and abundance data for crustacean zooplankton and phytoplankton, water chemistry, oxygen profiles and abundances of Chaoborus spp. to choose among our four hypothesized causal mechanisms. Briefly, zooplankton were enumerated from the volume-weighted composites as described above. Phytoplankton were sampled through the euphotic zone, and water chemistry samples were volume-weighted composites of all depths during isothermal periods, and of the mixed layers during periods of vernal and autumnal stratification. Dissolved oxygen profiles were generated at 1–2 m intervals from all depths. Finally, abundances of the zooplanktivore Chaoborus spp. were quantified in samples from night-time vertical hauls at 10 stations visited during the autumn in all lakes in 1986 and in one lake from 1987 to 1994 [37]. In the laboratory, a minimum of 300 cells, colonies or filaments of phytoplankton were counted within the Bacillariophyceae, Chlorophyceae, Cryptophyceae, Dinophyceae and Euglenophyceae. Dimensions of all counted algae were measured, and standing stocks were expressed as biovolumes [38]. We derived an index of ‘edible phytoplankton’ by summing phytoplankton biovolumes across these five families, which are preferentially consumed by zooplankton (see the electronic supplementary material).

Using an SEM approach, we then estimated a series of equations (2.1)–(2.6) describing a causal network that simultaneously tested our four hypothesized mechanisms.

(i) Declining populations mechanism

We estimated the probability of observing that a random cladoceran individual in each lake i in year j was a Holopedium , independent of the other species in the community. We did so assuming that the number of Holopedium individuals annually counted in each lake could be described by a binomial distribution with N_ij total number of planktonic cladoceran individuals

Thus, the relative abundance of Holopedium within the planktonic cladoceran assemblage could increase simply because the absolute number of non-Holopedium Cladocera declined (i.e. N_ij becomes smaller over time). Relative increases over time could also arise because both and N_ij increased. Alternatively, there could be an increase in the absolute abundance of Holopedium, calculated independently of any other taxa in the assemblage. The probability would correspondingly have increased over time.

(ii) Less competition mechanism

To test whether the number of Holopedium increased with food availability (ξ_ij), and whether ξ_ij increased as the dominant Cladocera in our lakes (Daphnia spp.) declined with falling lake Ca, we first modelled as an inverse-logit of ξ_ij

2.1

where α⁽¹⁾ is the estimated mean probability of observing Holopedium across all lakes and years, γ₁ is the effect of food availability, γ₂ is an estimated effect accounting for predation by Chaoborus spp. (Chaob_ij, square-root transformed, individuals m⁻³) and ν_i⁽¹⁾ and ν_j⁽¹⁾ account for random variation among the i lakes and j years and are each drawn from a normal distribution with a mean of 0 and separately estimated standard deviation. Although we did not observe values directly for ξ_ij, a latent variable, we parametrized it using observed data.

We assumed that ξ_ij was approximately N(η_ij,σ_ξ), and modelled mean food availability η_ij in each lake i in year j given abundances of potential Daphnia competitors. Exploitative competition for limiting resources, namely food, is a major factor structuring zooplankton communities [14]. Although larger zooplankton species can exclude smaller species because their greater body reserves survive low food periods, this competitive hierarchy depends on how the abundances of species vary with levels of predation and abiotic conditions [39]. For Daphnia, the abundances of species, and thus their competitive effects, depend on species-specific Ca requirements [14]. We therefore classified Daphnia species into either relatively Ca-rich (D. dubia, D. longiremis, D. mendotae, D. pulicaria and D. retrocurva) or Ca-poor (D. ambigua and D. catawba) groups using interspecific differences in body Ca content [11,40] and prevalence thresholds in field surveys [20,40] (electronic supplementary material, figure S1). However, one single metric cannot summarize the responses of all Ca-rich daphniids because body-size and Ca sensitivity still vary within this group (electronic supplementary material, figure S2). We additionally allowed for different responses within this group by summarizing community composition with an index D_ij that reflects both the relative abundances of species and their differences in Ca sensitivity [41]. The resulting D_ij is an ‘effective’ diversity measure; e.g. a community with three species and D_ij = 2.5 is slightly less diverse than a community of three equally abundant taxa with totally dissimilar Ca requirements (see the electronic supplementary material). This combination of approaches essentially allows both linear and nonlinear interactions between Holopedium and Daphnia in our model, without assuming any specific effect. Finally, we could not ignore herbivorous copepod species (Cop_ij), as their biomass (μg dry weight m⁻³) can account for the majority of zooplankton biomass in our study lakes, and temporal changes in their densities will influence food available for Cladocera [6,42]. η_ij was then a function of the probability of observing Ca-poor and Ca-rich daphniids , the composition of Ca-rich daphniid communities (D_ij) and Cop_ij

2.2

where γ₃–γ₆ are estimated effects and σ_ξ is the estimated s.d.

Mean food availability η_ij is directly proportional to and . As these probabilities will vary with lakewater Ca, we modelled them using binomial distributions based on the number of individuals of Ca-poor and Ca-rich daphniids, respectively, counted in lake i in each year j

2.3

2.4

where α^(k) is the estimated mean probability of observing individuals of life group k across all lakes and years, γ₇, γ₉, γ₁₀ are estimated effects of lake Ca, γ₈ and γ₁₁ are estimated effects accounting for the fact that predation by Chaoborus influences daphniid abundances, and and account for random variation among the i lakes and j years and are drawn from zero-mean normal distributions with separately estimated s.d. for each source of variation associated with each life group k. We allowed the response of Ca-rich daphniids to lakewater Ca to be nonlinear, as observed in mid-summer surveys of 304 regional lakes [40], by including the quadratic effect . There was no reason to do so for Ca-poor taxa as these are not positively associated with lakewater Ca [40].

(iii) More food mechanism

We then defined food availability ξ_ij as increasing with edible phytoplankton biovolume (P_ij, mm³ · m⁻³), and drew P_ij from a Poisson distribution modelled as

2.5

where α⁽⁴⁾ is mean biovolume across all years and lakes, γ₁₂–γ₁₃ are estimated effects of the number of sampling events and total phosphorus (TP), and ν_i and ν_j account for variation among the i lakes and j years, respectively, and are drawn from zero-mean, normal distributions with separately estimated s.d. We expected P_ij to increase primarily with TP, so we did not include other chemistry variables in equation (2.5) because they covaried with TP (e.g. Pearson's correlation with DOC = 0.68).

(iv) Increasing predation mechanism

Finally, we considered whether the relative abundances of Holopedium and their competition for food could be changed by Chaoborus, a principal predator of Daphnia. We modelled the densities of Chaoborus spp. (Chaob_ij) as a function of the thickness of the hypolimnetic water layer with less than 3 mg l⁻¹ O₂ (O_ij) and concentration of DOC in lakewater (DOC_ij), which both reduce predation from fish [24]

2.6

where α⁽⁵⁾ is the estimated mean density of Chaoborus, γ₁₄–γ₁₅ are estimated effects of O₂ and DOC and accounts for random variation among the i lakes and is drawn from a zero-mean normal distribution with estimated s.d. Chaoborus spp. were measured only in one lake outside of 1986, and so we could not adequately estimate both variation among lakes and years (total n = 16). Thus, we only accounted for variation among lakes. For all the other year × lake combinations in equations (2.1), (2.3) and (2.4), we let Chaob_ij take on mean observed values.

We did not directly consider how variation in predation on Chaoborus spp. themselves influenced daphniid abundances. We expect Chaoborus populations are likely to experience similar top-down control among lakes and whole-lake manipulative work on nearby lakes has shown that changes in piscivorous fish communities do not influence the overall importance of Chaoborus as zooplanktivores [43]. We did not model copepod biomass as a function of Chaoborus densities as we were not interested in estimating the associated effects. Any changes in copepod biomass associated with Chaoborus was reflected in the raw data and allowed us to determine the influence of copepods on food availability.

(a) Have Holopedium abundances increased since pre-industrial times?

Across both study regions, the relative contribution of Holopedium to planktonic cladoceran assemblages clearly increased since pre-industrial times (i.e. pre-1850 or prior to the onset of acid deposition). These increases were evident in both of our palaeolimnological surveys as the relative abundance of Holopedium increased in 25 of the 35 Ontario lakes and 17 of the 23 Nova Scotia lakes in which it was present (paired t-test, t₃₄ = 2.7, p = 0.012 and t₂₂ = 2.7, p = 0.013, respectively; figure 2a,b). Furthermore, Holopedium appeared (i.e. was present in the ‘top’ samples, but was absent from the pre-industrial samples) in 15 Ontario lakes and 11 Nova Scotia lakes, while disappearing from only two and one lakes, respectively, suggesting that increases in absolute abundances have also occurred.

• Download figure
• Open in new tab
• Download PowerPoint

(b) Have Holopedium abundances increased during the recent period of lakewater Ca decline?

Holopedium has become more abundant in the 31 south-central Ontario lakes between the 1980s and 2004–2005 as Ca declined by a median of 25% [17]. Holopedium relative abundances increased in 17 of the 26 lakes that have not been invaded by Bythotrephes; these increases ranged from 3 to 37% (paired t-test, t₂₅ = 2.8, p = 0.009, figure 2c). In each of the five invaded lakes, Holopedium was more abundant (by 5–30%) in 2004–2005 than in the 1980s (paired t-test, t₄ = 3.5, p = 0.024, figure 2c). The absolute abundance of Holopedium also increased; median abundance in the lakes doubled, from 304 animals m⁻³ in the 1980s to 607 animals m⁻³ in 2004–2005 (W₃₀ = 123, p = 0.013, figure 2d).

Increasing absolute abundances of Holopedium in the eight longitudinal study lakes mirror the trends in the regional dataset. The relative abundance of Holopedium increased in these lakes as lakewater Ca declined (figure 3a; Mann–Kendall test: τ = 0.41, p = 0.002 and τ = −0.71, p < 0.001, respectively). This arose partly because there were lower absolute abundances of both Ca-rich daphniids and all other Cladocera as Ca declined (τ = −0.48 and −0.55, respectively; p < 0.001 for both). Greater absolute abundances of the two Ca-poor daphniid species over this period (τ = 0.50, p < 0.001; figure 3b) could not offset these declines, i.e. the trend in absolute abundance of all non-Holopedium Cladocera was negative (τ = −0.48, p < 0.001). The increased relative abundance of Holopedium in the longitudinal study lakes was also attributable to an increase in its absolute abundance in Harp, Heney and Red Chalk Lakes from 1981 to 2009 (τ > 0.37, p < 0.010), and in Dickie Lake prior to recent dust suppressant (CaCl₂) additions [48], which raised lakewater Ca levels (one-tailed test for increasing trend: τ = 0.33, p = 0.029; see electronic supplementary material, table S3).

• Download figure
• Open in new tab
• Download PowerPoint

(c) Discriminating among alternative mechanisms underlying increases in Holopedium abundance

Our SEM revealed that the relative rise of Holopedium in the eight lakes of the longitudinal study could be explained by both declines in non-Holopedium Cladocera (‘declining populations’ mechanism) and increases in absolute numbers of Holopedium. The latter changes were specifically associated with reduced interspecific competition for food and declines in Cladocera predicted by the ‘less competition’ and ‘increasing predation’ mechanisms (figure 4).

• Download figure
• Open in new tab
• Download PowerPoint

Falling lakewater Ca was the primary explanation for declines in dominant Daphnia spp. that have favoured Holopedium. The probability of observing Daphnia increased with lakewater Ca in the SEM, with the effect much stronger for Ca-rich as opposed to Ca-poor species (95% CIs: 1.09–1.14 and 0.13–0.25, respectively). Thus, as Ca has fallen over time, so too have Ca-rich daphniids (trends reported above), thereby favouring a greater proportional representation of Holopedium within communities (i.e. lower N_ij predicted by ‘declining populations’ mechanism).

Falling abundances of Ca-rich Daphnia have also reduced competition for food, supporting the ‘less competition’ mechanism. The SEM revealed that Ca-rich but not Ca-poor Daphnia spp. reduced food availability (95% CIs: −0.06 to ≤ −0.01 and −0.32 to 0.17, respectively), and the probability of observing Holopedium increased with more food (95% CI: 1.01–1.17). Thus, as Ca-rich species declined, there was more food available for Holopedium. For example, a decline in lakewater Ca from 3 to 1 mg l⁻¹ corresponded with an approximately 34% increase in the absolute probability of observing Holopedium at mean levels of all other covariates (e.g. TP and Chaoborus spp. densities) when the pathways influenced by Ca were followed from start to finish in our SEM (figure 4). Copepods have similarly declined over time (τ = −0.40, p = 0.002), increasing food available for Holopedium (95% CI for their effect on food: −0.15 to −0.12). However, food availability did not increase in all lakes despite declines in non-Holopedium zooplankton. This was because the total food supply, determined by the abundance of edible phytoplankton, depended on TP concentrations (95% CI: 0.02–0.08), which have declined over time across lakes (τ = −0.33, p < 0.014). These declining trends lend no support to the ‘more food’ mechanism.

Although Ca-rich Daphnia have declined, the net effect on food availability could be offset by changes associated with their species composition (95% CIs for composition were negative in five lakes and positive in three; electronic supplementary material, table S4). Thus, overall food availability increased only in two lakes (Heney and Red Chalk), explaining why the absolute number of Holopedium has increased in these two lakes but not the others. There was no increase in food availability in Harp Lake despite increases in Holopedium reported earlier, because we only included years prior to Bythotrephes invasion in the SEM (1980–1992; electronic supplementary material, table S3). Absolute abundances of Holopedium only increased when we also considered the years after the invasion.

Finally, the SEM supported our ‘increasing predation’ mechanism, which predicted that Holopedium increased in relative abundance because they suffered less from predation by Chaoborus than did Daphnia (95% CIs for effect on Ca-rich Daphnia, Ca-poor Daphnia, Holopedium: −0.21 to −0.11, −0.20 to −0.09, −0.26 to 1.01, respectively). Declines in Daphnia from predation will have also increased food availability, promoting greater absolute abundances of Holopedium (figure 4). These changes may continue into the future as Chaoborus densities increased with the thickness of hypoxic water (95% CI: 0.23–13.6), which has increased over time (τ = 0.30, p = 0.023), and most strongly in recent years.

(a) Drivers of increased relative and absolute abundances

Holopedium abundance increased principally because Ca-rich daphniid populations fell with declining lakewater Ca. The survival, growth, development and reproduction of daphniids is known to be reduced at Ca levels below 1.5–2 mg l⁻¹ [49,50]. Delayed maturation at low Ca consequently limits the population growth of daphniids consistent with our ‘declining populations’ mechanism. Daphniids are also more vulnerable to at least one key predator (Chaoborus) at low Ca because their ability to produce anti-Chaoborus defences (larger bodies, more rigid carapaces and projecting neck teeth) is compromised [50]. Thus, greater predation on daphniids at low Ca, predicted by our ‘increasing predation’ mechanism, likely further increased the relative abundance of Holopedium.

Our SEM also suggests that ‘less competition’ for resources with other herbivorous zooplankton, principally daphniids, is linked to increasing absolute abundances of Holopedium as Ca levels decline. Competition between Holopedium and Daphnia is well documented [6–8,14], and a comparison of recent daphniid and Holopedium fecundity supports the competition mechanism. In the regional-scale monitoring survey, the average clutch size of Holopedium was twice that of its most common daphniid competitors (electronic supplementary material, figure S2), suggesting it was better able to secure the limiting food resources that reproduction requires. Among daphniids, only the relatively rare D. dentifera and D. pulicaria, which are much larger and thus more likely controlled by fish predation, equalled Holopedium's fecundity.

Declining TP levels did not support our ‘more food’ mechanism, but they could be promoting greater absolute abundances of Holopedium by reducing the competitive effects of Daphnia. First, daphniids likely have to eat more than Holopedium to get the P they need, because their P content is twice as high [12,13]. This should make them more vulnerable than Holopedium to falling food levels, particularly if they are poorer competitors. Thus, the competitive effects of daphniids may further depend on whether food supplies exceed the levels required to meet basic metabolic needs. Holopedium's advantage could also be enhanced if algal P content declines with falling TP levels [51]. Additional data on nutritional thresholds and phytoplankton community composition would help to test these explanations.

Many environmental changes other than Ca decline influence zooplankton, though they are unlikely to be the dominant factor explaining the rise of Holopedium. For example, the SEM suggests that recent declines in hypolimnetic oxygen levels due to climate warming [25] can promote abundances of Chaoborus, which prey more heavily on daphniids than Holopedium (‘increasing predation’ mechanism). However, predation risk is not increasing universally, as factors such as the spread of the piscivorous bass across Ontario [52] may lower the abundance of some planktivores, in turn reducing pressure on both Daphnia and Holopedium. Levels of coloured DOC [17,18] have also been rising in many Ontario lakes, though we found no evidence in our SEM that this has been large enough to benefit Holopedium. Holopedium abundances have increased over a broad range of DOC concentrations in both Ontario and Nova Scotia (1.4–12.7 mg l⁻¹) [4,5], so high DOC is clearly not a requirement of Holopedium increases. While our survey data also suggest the Bythotrephes invasion is promoting Holopedium (figure 2), Holopedium have increased in the majority of our study lakes in the absence of this invasion.

One factor we excluded from our SEM was pH, because lake acidity was an unlikely factor in the observed increases in Holopedium. Although Holopedium is much more acid-tolerant than Daphnia [14,53], and its relative abundance increased in Nova Scotia lakes that declined in pH [22], there have also been widespread reductions in lake acidity in south-central Ontario [34]. We might have expected increases in Daphnia abundance at the expense of Holopedium in Ontario, if lake acidity was the key driver, but the reverse was observed.

(b) Consequences of an increasingly jellied future

Our results signal a shift in the ecological state of temperate softwater lakes of eastern North America. Holopedium is widespread in these lakes [54], and we have shown both its relative and absolute abundances are increasing with widespread Ca decline. This will likely reduce vertical energy and nutrient transport in lake food webs [55]. Both the declining availability of daphniid prey and the relative ease with which planktivorous fish consume Holopedium [56], suggest that this taxon will increasingly dominate the diet of zooplanktivorous fish. Although Holopedium and daphniids have similar fatty acid content and composition [57], Holopedium has much lower P and Ca content than its daphniid competitors [11–13]. Thus, with its increasing dominance, fewer essential nutrients will be transferred to planktivorous fish [58]. The co-occurring copepods have low P content matching Holopedium [12,59]. These changes may also be quite long-lasting if acid deposition rates continue to exceed mineralogical weathering rates [60,61], reducing exchangeable Ca levels in soils, and ultimately in downstream waters.

As Ca declines, the rise in the absolute abundance of Holopedium may also have socioeconomic consequences, because increased concentrations of pelagic jelly may impede the withdrawal of lakewater for residential, municipal and industrial uses. The costs of operating infrastructure for water users will therefore rise as Ca continues to decline and densities of filter-clogging jelly capsules increase [62]. In Ontario, 20% of government-monitored drinking water systems draw only from surface waters within landscapes containing lakes with Ca concentrations less than or equal to 3.5 mg l⁻¹ (electronic supplementary material, figure S4), levels that favour Holopedium (electronic supplementary material, figure S3). Our results show that the continued jellification of pelagic food webs, due to ongoing declines in lakewater Ca, and exacerbated by other environmental stressors, has the potential to negatively impact the functioning of, and services provided by, temperate lakes in eastern North America.