1. Climate change and other human-driven environmental perturbations are causing reductions in biodiversity and impacting the functioning of ecosystems on a global scale. Metacommunity theory suggests that ecosystem connectivity may reduce the magnitude of these impacts if the regional species pool contains functionally redundant species that differ in their environmental tolerances. Dispersal may increase the resistance of local ecosystems to environmental stress by providing regional species with traits adapted to novel conditions.
2. We tested this theory by subjecting freshwater zooplankton communities in mesocosms that were either connected to or isolated from the larger regional species pool to a factorial manipulation of experimental warming and increased salinity.
3. Compensation by regional taxa depended on the source of stress. Warming tolerant regional taxa partially compensated for reductions in heat sensitive local taxa but similar compensation did not occur under increased salinity.
4. Dispersal-mediated species invasions dampened the effects of warming on summer net ecosystem productivity. However, this buffering effect did not occur in the fall or for periphyton growth, the only other ecosystem function affected by the stress treatments.
5. The results indicate that regional biodiversity can provide insurance in a dynamic environment but that the buffering capacity is limited to some ecosystem processes and sources of stress. Maintaining regional biodiversity and habitat connectivity may therefore provide some limited insurance for local ecosystems in changing environments, but is unable to impart resistance against all sources of environmental stress.
Anthropogenic ecological stressors, such as climate change, eutrophication, chemical pollution, habitat destruction and the introduction of invasive species, change environmental conditions globally and locally (Vitousek et al. 1997; Schindler 2001). Communities faced with increasingly stressful conditions may either respond in ways that maintain their structure and function (Walker et al. 2004), or experience losses of biodiversity and declines in their performance as ecosystems (Chapin et al. 1997). Losses of biodiversity and changes in the environment are presently occurring at unprecedented rates (Ricciardi & Rasmussen 1999; Sala et al. 2000; Pounds et al. 2006), altering the functioning of virtually all ecosystems (Chapin et al. 1997; Loreau et al. 2001; Halpern et al. 2008). Multiple ecological stressors can also interactively affect ecological communities in complex and unpredictable ways (Dukes et al. 2005; Christensen et al. 2006; Darling & Côté 2008). Understanding the properties that allow ecosystems to maintain function in the face of change, rather than collapse under multiple ecological stressors, is vital to minimizing the impacts of global environmental change on natural ecosystems and the services they provide to humanity.
The impact of environmental change on an ecological community is determined by the response of its resident species, as well as changes in composition because of dispersal and colonization from the regional species pool (Cottenie et al. 2003; Leibold et al. 2004). Beta diversity is maintained through filters imposed by abiotic conditions, interactions with local taxa and dispersal limitation (Shurin 2000). Ecological stressors that alter abiotic conditions or disrupt community stability may permit colonization of stress tolerant species from the regional species pool (Shurin 2001). Indeed, climate change is already driving range shifts by many species as the conditions that include their environmental niches move upward in elevation and latitude (Parmesan 1996). Species of virtually every taxonomic group are colonizing new habitats and becoming members of novel local ecological communities and ecosystems (Parmesan 2006).
The Spatial Insurance Hypothesis (Loreau, Mouquet & Gonzalez 2003) suggests that dispersal and colonization should provide spatial insurance against locally changing environments within a heterogeneous landscape. Species that are maladapted to new conditions could be replaced by new species from the region with traits that are favoured. This compensation could allow community function to be maintained if the new species perform similar roles to those they replace (Gonzalez & Loreau 2009). This theory predicts that dispersal by species in the region or beyond should promote the replacement of stress-sensitive species by more tolerant ones that can maintain ecosystem function.
Few experiments have tested the theoretical predictions about how regional biodiversity affects the stability of communities under environmental change (Leibold & Norberg 2004; Loreau, Mouquet & Gonzalez 2003; Yachi & Loreau 1999; but see Forrest & Arnott 2007; Staddon et al. 2010). Here, we employed a large mesocosm experiment to test how the dispersal of freshwater zooplankton from a regional metacommunity affects the response of local communities to environmental changes resulting from multiple ecological stressors. We applied factorial stressors of climate warming (increase of c. 2·5 °C) and increased salinity (0·3 ppt) to a local plankton community in 1000 L field mesocosms. The diversity and composition of plankton communities in the experiment were comparable to those found in natural assemblages in the region. Our warming treatment followed the diurnal temperature fluctuations experienced by the ambient mesocosms (Fig. S1, Supporting information). This increase in temperature corresponds to the conservative range of predicted scenarios for northern hemisphere freshwater lakes and ponds under a forecasted doubling of atmospheric carbon dioxide (Magnuson et al. 1997). The salinity treatment is approximately half of the concentration that has been shown to be chronically toxic to freshwater life (Kaushal et al. 2005). Such increases in salinity are often caused by run off of road de-icing salt (Kaushal et al. 2005) as well as higher evaporation rates that accompany climate warming (Evans & Prepas 1996). The salinity and temperature stress treatments were crossed with two levels of dispersal; communities were either isolated (no dispersal facilitated) or connected to the regional species pool (assisted dispersal of plankton from 40 waterbodies in the region of Southwestern British Columbia). A number of ecosystem functions were measured in the mesocosms, including ecosystem productivity, leaf litter decomposition, sedimentation and periphyton growth. We also measured the standing stocks of zooplankton and phytoplankton biomass (as chlorophyll-a). This design allowed us to determine the potential of regional species to impart resistance and maintain ecosystem functioning in the face of significant and relevant levels of environmental change.
We predicted that reductions in biodiversity and biomass in response to the ecological stressors would be dampened when dispersal of regional organisms is facilitated. That is, we expected to see interactive effects of the dispersal and stress treatments on rates of ecosystem processes. We measured a diverse set of ecosystem function rates, some that we expected would be strongly affected by changes in zooplankton community, such as ecosystem production and sedimentation, and some that we did not expect zooplankton to affect directly, such as periphyton growth and leaf litter decomposition. We measured a range of ecosystem functions to compare their responses to stress and to zooplankton dispersal, and to test for potential indirect effects of zooplankton colonization.
Materials and methods
The experiment was conducted at the Experimental Pond Facility on the University of British Columbia Campus, Vancouver, BC. Forty 1000 L plastic cylindrical livestock watering tanks (Rubbermaid®, Sandy Springs, GA, USA) were used as experimental mesocosms. Tanks were filled with Vancouver, BC city water on 23 April 2008, and the bottoms were covered with rinsed, coarse sand as a benthic substrate. One litre of peat moss was added to each mesocosm on May 2 to increase the organic content of the water. Mesocosms were inoculated with plankton and nutrients from a nearby pond on May 8 by adding sediment and live plankton collected using a shovel and a 64 μm conical tow net, respectively. All collected sediment was mixed thoroughly prior to being added to ensure that mesocosms received roughly similar densities of plankton and to increase the chance that rare species were added to all tanks. Similarly, water containing plankton collected by conical net tow was thoroughly mixed and added in the same volume to all mesocosms. Counts of pre-treatment zooplankton samples collected on May 21 confirmed that there were no pre-treatment differences in community composition (anosim, R = −0·0107, P =0·648). Three mesocosms used to measure natural dispersal were established in similar fashion but received no peat moss, zooplankton, phytoplankton or sediment. These mesocosms were fertilized with 10 μg L−1 phosphorous as KH2PO4 and 160 μg L−1 nitrogen as NaNO3 on May 28 to stimulate primary production since they received no other sources of nutrients. Experimental communities were allowed to establish and stabilize for 5 weeks before treatments were applied. All mesocosms were left open to the air. We chose not to use mesh covering to prevent macroinvertebrate colonization because of previous experiments found this method to be ineffective (J. T. Ngai & J. B. Shurin, unpublished data). Random colonization by invertebrates may have led to increased variation among our replicates that would have decreased our statistical power but would not bias our results.
Application of treatments
The factorial treatments of warming, increased salinity and artificial dispersal were each replicated five times. The warming and salinity treatments were applied on June 17. The dispersal treatment was first applied on June 30 as described below, to allow the original community time to adjust to the new environmental conditions before new species were added. The dispersal treatment was repeated on July 15 and August 12 to allow multiple opportunities for colonization.
Tanks were warmed by 300 W submersible aquarium heaters (Hagen®, Montreal, Canada) running constantly so that heated mesocosms were c. 2·5 °C warmer than ambient mesocosms at all times and followed the same diurnal temperature fluctuations (Fig. S1, Supporting information). Artificial heaters, made of clear plastic tubing filled with sand sealed off from the water with rubber stoppers, were added to all ambient temperature mesocosms to control for the physical presence of the heaters.
Laboratory grade NaCl was added to increase the salinity to 0·3 ppt from an ambient concentration of <0·1 ppt. This concentration was chosen because it is approximately half of the level that is chronically toxic to freshwater organisms (Kaushal et al. 2005) and thus can be considered a non-lethal stress. Evaporation caused salinity levels to reach a maximum of 0·6 ppt in the warming + salinity treatment and 0·5 ppt in the salinity treatment at ambient temperature in late July, levels that may have been chronically toxic to freshwater organisms. Elevated salinity is a natural side effect of increased evaporation at elevated temperatures in pond ecosystems. Four hundred litres of water from a nearby experimental pond that had been filled from the same source as our mesocosms was pumped into each tank on July 22 to offset evaporation. All added water was filtered through a 64 μm mesh to prevent introduction of new zooplankton. The same volume of water was added to tanks in all treatments to mimic a rain event rather than restoring the ponds to their original volumes. The warming treatments, therefore, retained higher salinities than the ambient tanks after the addition of water.
Concentrated 60 mL aliquots of live plankton were added to each dispersal treatment mesocosm during each of the three dispersal events. The plankton were collected from 40 lakes and ponds in Southwestern British Columbia, spanning a salinity gradient of 0·0–2·8 ppt, and a temperature gradient of 17·0–23·6 °C at the time of collection. Plankton were collected from each waterbody using a 64-μm conical plankton net pulled through the water column either vertically or horizontally depending on waterbody morphometry. One litre of sediment was also collected from the middle of each waterbody using an Ekman sampler to obtain species present in the sediment bank. Both littoral and pelagic species were collected during the sampling although not in a consistent manner in all sites. Plankton were transported back to the experimental ponds in coolers. Time to transport varied from <20 min to 3 days depending on the location of the waterbody. This was an unavoidable consequence of sampling 40 widely dispersed waterbodies to obtain maximum regional diversity.
Plankton from all 40 regional waterbodies were mixed together in a bucket prior to dispersal. Zooplankton were dispersed within 15 min to minimize the stress of being held in high densities. An average of 755 individuals (range among dates: 226–1410) with an average SE of 63·6 (range among dates: 12·8–156·8) were added to the tanks in each dispersal event. This represented an addition of about 1·5% of the total estimated density of zooplankton individuals already present in the mesocosms. Dispersing individuals included copepods, cladocerans and rotifers, some of which were already present in our local community. In total, 16 identified taxa were added in the dispersal treatments and all but two successfully colonized the mesocosms. The regional species pool contains more than 16 taxa, but we sampled only 40 waterbodies and some taxa may have not survived transport back to the experimental site. Additionally, many taxa were not identified to species level and so we likely underestimated regional species richness. Eight of the 16 introduced regional taxa were not found in the local community in pre-dispersal treatment samples (Table S1, Supporting information). Identical inocula of plankton were heat-killed and added to each non-dispersal treatment mesocosm to control for added nutrients and water.
Communities of regional plankton were maintained in five tanks during the experiment so that regional plankton only needed to be collected once. The composition of the regional species pool likely changed over time in the tanks, although Shannon diversity and richness in our dispersal treatment additions did not change significantly over time (Shannon diversity, P =0·066; richness, P =0·478, Fig. S2, Supporting information). Maintaining the species pool in this way was necessary because of the logistical challenge of sampling 40 widely situated waterbodies multiple times over the course of the experiment. The high salinity lakes and ponds were sampled last and were not stored in the holding tanks prior to the first dispersal event so that they would not be exposed to low salinity prior to being added to the experiment.
The average of 755 individuals added in each of our dispersal treatments represents a high but not unnatural rate of dispersal. Vanschoenwinkel et al. (2008) collected over 850 viable zooplankton propagules from 17 different taxa in 28 days in nine windsocks placed near temporary mountain ponds. While this study was conducted in a windy environment and only a small number of these propagules would have landed in waterbodies, it demonstrates the potential intensity of zooplankton dispersal. Additionally, the 16 taxa introduced in the dispersal treatment are comparable to the numbers found in studies of natural colonization of mesocosms and newly formed ponds (Cáceres & Soluk 2002; Cohen & Shurin 2003; Louette, De Meester & Declerck 2008). At the end of our experiment, three taxa (Daphnia, Ceriodaphnia and calanoid copepods), that were common in both the regional and local species pools, had colonized the three mesocosms that had not been initially seeded with plankton. Therefore, natural levels of dispersal into our mesocosms were not as high as those found by the above studies, and the 16 taxa that we introduced were around 5×, the level of background dispersal naturally occurring into our zooplankton-free mesocosms.
Sampling and analysis
Chlorophyll and salinity were sampled semi-monthly starting June 16, the day before the initial application of treatments. Chlorophyll-a concentrations were measured using the in-vivo fluorescence method on a Trilogy fluorimeter (Turner Designs, Sunnyvale, CA, USA). Salinity, pH and conductivity were measured using a handheld probe (YSI®, Yellow Springs, OH, USA). Zooplankton were sampled on June 16, the day before the initial application of treatments, and twelve weeks later on September 9. Zooplankton samples were taken by collecting 10 L of water, sub-sampled from at least 10 haphazardly chosen locations around the tanks using an integrated sampler constructed from PVC pipe and filtering it through a 64-μm sieve. Samples were preserved in 70% ethanol. A separate integrated sampler was used for each mesocosm, and the 64 μm sieve was rinsed thoroughly between samples to prevent contamination of organisms between tanks.
Zooplankton samples were counted using a dissecting microscope at 10× magnification and at 60× magnification for identification and measuring body length. Zooplankton samples were subsampled so that at least 500 individuals of each taxa were counted. All individuals of taxa with fewer than 500 individuals were counted in each sample. Biomass of each taxa was estimated by measuring length of 10 haphazardly selected individuals per sample and using length–mass regressions for crustaceans and length–volume regressions for rotifers (Dumont & Balvay 1979; McCauley 1984; Johnston & Cunjak 1999). Zooplankton taxon richness was rarefied to the number of individuals in the sample with the fewest animals to allow comparison between samples with different densities of individuals (Gotelli & Colwell 2001). Shannon diversity was calculated for the entire zooplankton community, as well as for the crustacean zooplankton (rotifers excluded) (Krebs 1998).
Ecosystem function measurements
Daily cycles in dissolved oxygen concentrations were measured using a handheld probe (YSI®) at dawn, and dusk, on August 6 and again on September 10 to estimate ecosystem production and respiration. Net ecosystem productivity was calculated as the increase of oxygen from dawn to dusk (Downing & Leibold 2002). Sampling dates were analysed separately to determine whether the ecosystem rate responses to the treatments varied over time. Dissolved oxygen concentrations are influenced by water temperature, but we consider this a natural consequence of our warming treatment.
Pre-weighed, oven-dried Alnus rubra leaves in 0·5 mm mesh bags were deployed on August 18 and left anchored to the bottom in the centre of the mesocosms for 4 weeks. The litterbags were then removed, dried at 40 °C for 2 days and reweighed. Decomposition was calculated as the per cent difference in mass over time (Bärlocher 2005).
Pre-weighed 60-mL centrifuge tubes were fixed, uncapped, to the bottom of the mesocosms in an upright position between June 30 and September 8 to measure sedimentation rates. Tubes were capped during zooplankton sampling and when water levels were topped up to avoid accidental deposition of sediment. Following the experiment, tubes were placed uncapped in a drying oven at 40 °C until all water had evaporated, prior to weighing. The sedimentation rate was calculated as dry weight of sediment deposited per day that the tubes were in the mesocosms.
Unglazed 25 cm2 clay tiles were deployed between July 18 and September 10 to measure periphyton growth rates. Tiles were removed from the mesocosms and scrubbed with a toothbrush to brush off periphyton into distilled water, which was filtered through a Whatman GF/C filter paper. Filter papers were analysed for chlorophyll after cold extraction in acetone using the non-acidification method on a Turner Trilogy fluorimeter (Turner Designs; Welschmeyer 1994). Periphyton growth was calculated as the total concentration of chlorophyll divided by the number of days the tiles were in the mesocosms.
The effect of the three treatments and their interactions on zooplankton community composition was analysed using redundancy analysis (RDA). Zooplankton community data were transformed using a Hellinger transformation to reduce the influence of the large number of zeros that typically occur in community data (Legendre & Gallagher 2001). Significance of each treatment and interaction was determined using Monte Carlo permutation tests on the model results of the RDA. The responses of the local and regional taxa to the individual stressors were compared with a t-test of the standardized taxa responses to the individual stressors obtained from the RDA. The effect of the treatments on the ecosystem function rates and standing stocks was tested using three-way factorial anova (α = 0·01 – Bonforroni correction). Ecosystem function rates and standing stock data were log transformed prior to analysis to normalize variance. Chlorophyll concentrations were analysed using repeated measures anova using spss (SPSS Inc., 2010, Chicago, IL, USA). All other analyses were performed using R (R Development Core Team 2008), using the ‘vegan’ package.
Zooplankton diversity and richness
Warming increased both rarefied zooplankton richness (W, P =0·002, Fig. 1a) and zooplankton Shannon diversity (W, P = 0·009, Fig. 1b) but the increase in Cladoceran zooplankton Shannon diversity was marginal (excluding rotifers, W, P = 0·066, Fig 1c). Dispersal caused an increase in rarefied zooplankton richness (D, P =0·018, Fig. 1a) and Cladoceran zooplankton Shannon diversity (D, P =0·006, Fig. 1c) but had no effect on overall zooplankton Shannon diversity (D, P =0·120, Fig. 1b). Salinity did not have any detectable direct effect on either richness or diversity. However, salinity marginally reduced the positive effect of dispersal on Shannon diversity (D, P =0·120, Fig. 1b). There were no other significant interactions between the treatments.
Zooplankton community composition
The warming, salinity and zooplankton dispersal treatments had independent and interactive effects on zooplankton community composition, shown in the RDA ordination (Fig. 2, Table 1). Taxa plotted in the direction of the treatment arrows in Fig. 2 increased in abundance with that treatment or combination of treatments (e.g. cyclopoid copepods increased with salinity), while taxa plotted in the opposite direction declined in abundance (e.g. Daphnia declined with warming). The results of the corresponding univariate tests can be found in the Supporting Information (Table S2). Here, we describe the trends shown in the RDA because we are interested in the effects of treatments on overall community composition rather than the abundance of individual taxa. Daphnia and Chaoborus abundance were negatively associated with warming and high salinity. The dominant local calanoid copepod declined in the salinity treatment while cyclopoid copepod abundance increased. The genera Scapholeberis, Ceriodaphnia, Bosmina, Diaphanosoma, Polyphemus and Chydorus were all positively associated with the warming, dispersal and warming with dispersal treatments. Rotifers showed no association with the dispersal treatment, and their contribution to zooplankton community biomass was negligible. The warming × dispersal interaction explained 5% of the total variation in zooplankton composition and 20% of the explained variation (Table 1).
Table 1. Redundancy analysis of the Hellinger transformed biomass (μg L−1) of the zooplankton community composition (17 taxa)
Treatment or interaction effect
λ: The proportion of zooplankton community variance explained by each treatment or interaction effect.
P-values based on permutation tests. Statistical significance (P < 0·05) indicated in bold.
Warm × Salt
Warm × Disp.
Salt × Disp.
Warm × Salt × Disp.
Zooplankton species response to stress
The 17 taxa found in the experiment exhibited a wide range of responses to climate warming and salinization (Fig. 3). Regional taxa (those that were introduced from the regional pool through the dispersal inocula and were not present in the tanks in the no-dispersal treatment) were more tolerant on average to warming than local taxa present in the initial community. That is, the average response score to the warming treatment in the RDA was more positive for regional than local taxa (t-test, P =0·042). There was no difference in average response to salinity between local and regional taxa (t-test, P =0·674).
Total zooplankton biomass
Increased salinity reduced zooplankton community biomass (S, P =0·036), and this effect was synergistically greater when combined with warming (W × S, P =0·009, Fig. 4a). Dispersal did not directly or interactively affect total zooplankton biomass.
Ecosystem function rates
Warming increased net ecosystem production in August (W, P =0·005, Fig. 5a), decreased net ecosystem production in September (W, P =0·005, Fig. 5b), increased periphyton growth (W, P =0·001, Fig. 5e) and had no effect on leaf litter decomposition (W, P = 0·448, Fig 3c) or sedimentation (W, P = 0·731, Fig. 5d). Salinity had no direct or interactive effect on any of the ecosystem function rates. Dispersal had no direct effects on ecosystem function rates but negated the positive effect of warming on net ecosystem production in August (W × D, P =0·003, Fig. 5a). Dispersal did not alter the effect of warming on September net ecosystem production (W × D, P =0·926, Fig. 5b) or periphyton growth (W × D, P =0·793, Fig. 5e). All corresponding univariate tests are reported in the Supporting Information (Table S2).
Warming and salinity both caused chlorophyll-a concentrations to increase (W, P =0·025; S, P =0·019, Fig 4b). Dispersal did not directly affect chlorophyll-a concentrations (D, P =0·703) but interactively negated the positive effect of salinity (S × D, P =0·034). Chlorophyll concentrations differed between sampling dates (P =0·001) but the treatment effects were similar between sampling dates (i.e. none of the treatment-by-time interactions were significant in the rmanova). Chlorophyll concentrations on each sampling date are shown in Fig. S3. Chlorophyll-a concentration and total zooplankton community biomass averaged across all dates for each treatment combination were strongly negatively correlated among treatments (n = 8, Pearson’s R = −0·887, P =0·003).
Our findings indicate that regional biodiversity and habitat connectivity affect the response of local ecosystems to environmental change. Most experimental studies of biodiversity effects on ecosystem function deal with small spatial scales and single trophic levels (Hooper et al. 2005; Srivastava & Vellend 2005; Duffy et al. 2007). The discrepancy between the scope of these experiments and the broad geographic scales of extinction limits our understanding of how biodiversity loss may impair ecosystems. Our experiment addresses the potential for regional diversity to impart community resistance to environmental change.
Environmental change, caused by warming and salinization, resulted in compositional changes in the local zooplankton community. Notably, warming caused declines in Daphina, the dominant local cladoceran, while salinization resulted in declines in the dominant calanoid copepod and increases in cyclopoid copepods. This presence of salt tolerant taxa in the local community contrasted with the lack of heat tolerant taxa, suggesting some local adaptive capacity to resist increases in salinity but not environmental warming (Fig. 3).
The dispersal treatment added additional species not present in the local community, increasing both overall zooplankton richness and crustacean diversity. The addition of these regional species mediated the warming-induced compositional changes by introducing heat tolerant cladoceran taxa (Scapholeberis, Ceriodaphnia, Bosmina and Diaphanosoma) that compensated for losses in the local Daphnid. The interactive effects of warming and dispersal on zooplankton species composition shown in the RDA analysis indicate that dispersal promotes species turnover in changing environments. However, the addition of regional taxa had little effect on the response of the community to salinization. This was surprising because our sampling of the regional pool included naturally saline ponds (max salinity: 2·8 ppt) in the interior of British Columbia. It is possible that regional species capable of providing compensation may have been lost because of stress during collection. Tolerance to handling does not necessarily correlate with tolerance to environmental stress. This may be one reason why we did not find compensation in the case of salinity. It may also be the case that compensation within the local community was sufficient to exclude colonization by regional species.
Warming and salinity had synergistic effects on zooplankton biomass. This suggests that neither the local nor regional species pools contained enough taxa that could withstand both stressors to maintain performance, limiting the resistance of the ecosystem to the combined stress. The buffering capacity provided by regional diversity and dispersal breaks down in the face of multiple combined sources of stress. The lack of species with co-tolerance to both stressors allows for little compensation by regional biodiversity and resulted in synergistic losses in zooplankton community biomass. As many environmental changes often occur simultaneously, including warming and salinization (Schindler 2001; Vinebrooke et al. 2004), this is likely to factor in the response of freshwater systems to ongoing global changes.
Environmental change caused changes in ecosystem function as predicted. However, this change only occurred with warming and not with salinity, and only for two of the five measured rates (ecosystem productivity and periphyton growth). The lack of ecosystem function change with salinity may have been related to the compensation witnessed within the local zooplankton community, although this cannot be confirmed. The effect of warming was also not consistent across time; in August, there was a positive effect on ecosystem production, while in September, the effect was negative. We suspect that the increase in productivity witnessed in August was the result of reductions in the dominant grazer Daphnia, allowing the phytoplankton biomass to increase. The cause of the negative response of production to warming in September is unknown. Supporting this theory, phytoplankton biomass appeared to be top-down controlled by zooplankton grazing as chlorophyll-a (a proxy for phytoplankton biomass) was inversely related to zooplankton community biomass (Fig. 4a,b).
We found support for our prediction that dispersal of regional zooplankton could reduce the magnitude of changes in ecosystem function by introducing stress tolerant taxa that compensate for losses in function in stress-sensitive local species. Dispersal of regional taxa negated the positive effect of warming on productivity in August, likely through the introduction of heat tolerant regional cladocerans that compensated for the loss of grazing by local taxa such as Daphnia.Blake & Duffy (2010) report a similar grazer diversity-induced stabilizing effect, although not related to dispersal, on algal biomass when multiple stressors were applied to an experimental seagrass ecosystem. Our dispersal treatment caused an increase in community richness and diversity of crustacean zooplankton independent of the stress treatments, but dispersal only affected ecosystem function in conjunction with warming. That is, ecosystem productivity was only affected by dispersal under warmed conditions, while dispersal caused an increase in zooplankton community richness in all conditions. This suggests that the greater diversity of zooplankton that resulted from dispersal did not affect ecosystem function except when it allowed for compensation by the regional taxa for local species that declined under stress from warming. This result demonstrates the potential for dispersal of regional taxa to buffer local ecosystem function under changing environments, as is predicted by the Spatial Insurance Hypothesis (Loreau, Mouquet & Gonzalez 2003).
Our results also demonstrate that the potential for regional biodiversity to provide spatial insurance is limited. We found no such buffering effect for periphyton growth, and the other ecosystem function that was affected by the stressors. This is not surprising since periphyton growth is a benthic process, while zooplankton are more associated with limnetic habitats. Dispersal of organisms such as benthic macroinvertebrates that play larger role in grazing periphyton may have had a similar stabilizing effect. The stabilizing effect of dispersal on ecosystem productivity was also seasonal; the interactive effect with warming was apparent in August but not in September. We believe that this temporal discrepancy is because of the fact that warming decreased productivity in September rather than increasing it, as was the case in August. Zooplankton grazing, therefore, could not potentially stabilize ecosystem productivity. The contrast between the range of responses to the salinity and temperature treatments by different zooplankton taxa may explain why certain ecosystem functions, particularly primary production, were more influenced by warming than salinity. In addition, fewer members of the local community responded positively to warming than to salinity (Fig. 2), which may be why regional connectivity affected the ecosystem response to warming but not to salinity. These results indicate that maintaining high regional diversity and habitat connectivity can offer some limited buffering capacity against environmental change. However, regional diversity is not likely to buffer all processes against all forms of environmental stress.
Our different measurements of ecosystem function responded idiosyncratically to environmental change. Hector & Bagchi (2007) show that as more ecosystem rates are measured, the importance of biodiversity to their maintenance becomes more apparent as those species responsible for maintaining one function may not be the same as those that maintain another. While our dispersal treatment only affected one aspect of ecosystem function (ecosystem productivity in August), it may still have complex indirect effects on others through trophic interactions and interspecific competition that arise as a consequence of stabilization of productivity. Measuring ecosystem rates over longer time periods could illuminate the potential for such effects as they could potentially arise over multiple generations.
This study provides evidence for the potential for regional biodiversity to maintain ecosystem resistance to environmental change. However, while some ecosystem functions can be buffered by the introduction of regional species, this was not consistent across all seasons and did not extend to all ecosystem functions. Therefore, the functional compensation and ecological resistance provided by dispersal may be limited. Although ecosystem function often declines with biodiversity, several studies have shown that very few species are able to maintain ecosystem function under stable conditions (Cardinale et al. 2006; Downing 2005; but see Hector & Bagchi 2007; Hooper et al. 2005). However, in a dynamic and changing environment, higher biodiversity may be necessary to provide insurance and maintain function under novel conditions. Regional species represent the pool of potential functional biodiversity at local scales. Their dispersal increases the potential for compensation when changing conditions impair the performance of the local community. Fragmentation that isolates habitats by reducing dispersal and the loss of regional biodiversity may therefore decrease ecosystem resistance in the face of environmental changes (Staddon et al. 2010). Our study provides evidence for this role of regional diversity, but also highlights its limited capacity to stabilize ecosystems. Maintaining regional diversity and habitat connectivity may therefore form one component of a larger strategy aimed at conserving ecosystem resistance under environmental change.
We thank Michaela Martin, Anita Norman, Blake Matthews, Steven Declerck and Roger Gendron for assistance in the field, and Mark Vellend and three anonymous reviewers for helpful comments on the manuscript. Funding was provided by the Natural Sciences and Engineering Research Council of Canada.