Variability and predictability in a zooplankton community: The roles of disturbance and

dispersal
Author(s): Jessica FORREST and Shelley E. ARNOTT
Source: Écoscience, Vol. 14, No. 2 (2007), pp. 137-145
Published by: Taylor & Francis, Ltd.
Stable URL: https://www.jstor.org/stable/42901846
Accessed: 13-02-2022 16:43 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Taylor & Francis, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to
Écoscience

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
 ÊTOSCI ENJCE 14(2): 137-145(2007)

Variability and predictability in a zooplankto

community: The roles of disturbance
and dispersal1
Jessica FORREST2 & Shelley E. ARNO TT3, Department of Biology, Queen's University, Kingston,
Ontario K7L 3N6, Canada, e-mail: arnotts@biology.queensu.ca

Abstract : Disturbances are expected to increase variability and/or decrease predictability in ecosystems, while dispersal
(or immigration) may have positive or negative effects on ecosystem stability. We conducted a mesocosm experiment with
pelagic zooplankton to examine the independent and interactive effects of a nutrient pulse and immigration from neighbouring
lakes on (a) temporal variation in the mesocosm communities and (b) variation among spatial replicates. Surprisingly, nutrient
enrichment had no significant effect on among-replicate variability in algal or zooplankton biomass, and it decreased both
among-replicate variability in zooplankton community composition and temporal variability of algal biomass (measured
as the coefficient of variation). The nutrient pulse increased temporal variation in zooplankton biomass but decreased
temporal change in relative species abundances. However, both these effects depended on dispersal, suggesting a context-
dependent role of dispersal in community stability. We discuss the importance of considering multiple measures and
aspects of community variability.
Keywords : dispersal, immigration, stability, variability, zooplankton.

Résumé : Dans les écosystèmes, on suppose généralement que les perturbations augmentent la variabilité et/ou diminuent
la prédictibilité alors que la dispersion (ou l'immigration) peut avoir des effets positifs ou négatifs sur la stabilité des
écosystèmes. Nous avons réalisé une expérience en mésocosme avec du plancton pélagique afin d'étudier les effets
individuels et les interactions d'un apport de nutriments et de l'immigration à partir de lacs voisins sur (a) la variation
temporelle dans les communautés du mésocosme et (b) la variation entre les réplicats dans l'espace. Étonnamment,
l'enrichissement en nutriments n'avait pas d'effet significatif sur la variabilité entre les réplicats dans la biomasse d'algues
ou de zooplancton et diminuait à la fois la variabilité entre les réplicats dans la composition de la communauté de zooplancton
et la variabilité temporelle de la biomasse d'algues (mesurée par le coefficient de variation). L'apport de nutriments
augmentait la variation temporelle dans la biomasse de zooplancton mais diminuait le changement dans le temps dans
l'abondance relative des espèces. Cependant, ces deux effets dépendaient de la dispersion, ce qui suggère que le rôle de la
dispersion dans la stabilité des communautés dépend du contexte. Nous discutons de l'importance de considérer plusieurs
mesures et aspects de la variabilité des communautés.
Mots-cles : dispersion, immigration, stabilite, variabilité, zooplancton.

Nomenclature'. Stemberger, 1979; McLaughlin et al, 2005.

Introduction

The causes of temporal variation in populationsrepresent

and natural systems, it should be possible to make pre-
communities have been of theoretical and practical inter-dictions about variability in natural systems based on model
est for a long time but remain controversial (Loreau et results.
al., However, the complexity of biological communities
can of
2002). Part of the difficulty arises from the proliferation defy analytical modelling, and, often, population-level
different conceptions of stability and variation, with results
some cannot be extended to the level of the whole commu-
nity (Loreau et al., 2002).
of the key theoretical concepts (e.g., eigenvalues, equilib-
ria) being difficult to apply in natural systems (Loreau et Mathematical models have indicated that dispersal may
al., 2002). While natural populations may not have easily
have a stabilizing effect on population dynamics, in that
defined equilibria, it is possible to measure temporal vari-the addition of a fixed number of individuals in a simple
ability (e.g., as the coefficient of variation) in natural popu-
population growth model can simplify complex or chaotic
lations and argue that reduced variability is analogous to
dynamics and produce a stable equilibrium (Holt, 2002).
greater stability in the mathematical sense (Steiner, 2005).
However, experimental demonstrations of analogous phe-
Then, to the extent that mathematical models accurately nomena are rare (although see Holyoak & Lawler, 1996),
and some models show a potentially destabilizing role of
immigration; there appear to be no general rules governing
»Ree. 2006-05-11; acc. 2006-10-02. when immigration should stabilize or destabilize population
Associate Editor: Marco Rodriguez. dynamics (Stone & Hart, 1999).
2Present address: Department of Ecology and Evolutionary Biology, University of
Toronto, Toronto, Ontario M5S 3G5, Canada. At a community level, dispersal (or immigration from
3Author for correspondence. source areas) should also be an important regulator of tem-

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
Forrest & Arnott: Zooplankton community variability

community-
poral variability, but its effects are difficult to predict and wide covariance contributes to reduced variabil-
might depend on species identity. More diverse communi-ity in total community biomass. On the other hand, a more
negative total covariance also indicates increased temporal
ties (both real and simulated) tend to show reduced temporal
variation at the level of the whole community (Doak variation
et al, in community composition {cf. Figure 1 in Micheli
1998; Tilman, 1999; Loreau et al. , 2002), so additionetofal.,
new1999).
species through immigration could be expected to reduce
Theoretically, it seems possible that a disturbance
community variability. During periods of environmental
{e.g., an experimental nutrient pulse) could increase or
change, in particular, introduction of new species mightcovariances among species, depending on whether
decrease
promote compensatory dynamics among species and help
species tend to respond similarly or differently to the dis-
to stabilize aggregate community properties such as total
turbance. During a nutrient pulse, immigration from other
biomass (Loreau, Mouquet & Gonzalez, 2003; Vinebrooke
habitats might reduce covariances, either by introducing
et al.f 2003). In the absence of other disturbances, however,
differently adapted competitors that might covary negatively
species introductions {e.g., of predators) can have destabi-
with resident species or simply by disrupting synchro-
lizing effects on communities (Halpern et al., 2005).nous dynamics. An alternative view is that immigration,
by introducing
Perturbations are generally expected to (and often do) both new species and new genotypes of
resident
increase the variability of biological systems (Odum, Finn species, might increase covariances by causing
& Franz, 1979; Warwick & Clarke, 1993; Rusak et al., 2001), species to respond in a similar, positive way to
multiple
and nutrient pulses, resulting from anthropogenic nutrients.
inputs However, we know of no experimental tests of
and/or natural mixing and runoff, are common perturba-
these ideas, and it is unclear how immigration and nutrient
tions in aquatic habitats (Meyer & Likens, 1979;enrichment
Cooke should affect population variability of individ-
& Prepas, 1998). Many studies of the effects ofual nutri-
zooplankton species. In addition, the effect of nutrient
enrichment on variability among experimental units has
ents consider naturally occurring gradients or longer-term
experimental manipulations rather than short-term not been addressed, although spatial variability seems as
pulses.
For instance, simple predator-prey models show important
unstable as temporal variability as a potential indicator of
dynamics with increased resource availability (Rosenzweig,
ecosystem stress (Warwick & Clarke, 1993).
1971), and although these models may not adequately rep- We tested whether dispersal and disturbance could
resent real communities, they have been used to predict a
influence community- and population-level variability in
more general tendency for enrichment to increase variability
an aquatic system. We conducted a mesocosm experiment
(Cottingham, Rusak & Leavitt, 2000; Steiner et al., 2005).
with a pelagic zooplankton community, in which nutrient
There is some evidence that increased primary productiv-
levels and immigration rate were manipulated in a facto-
ity can increase temporal variation in aquatic communities:
rial design. Treatment effects on community composition
in a survey of small ponds, Steiner (2005) foundand a weak
diversity are considered elsewhere (Forrest & Arnott,
positive relationship between the temporal variability of
2006). Here, we demonstrate that a nutrient pulse affects
zooplankton biomass and an aggregate measure of pond
short-term variability in phytoplankton and zooplankton
productivity. Cottingham, Rusak, and Leavitt (2000) docu- differently and that the effect on zooplankton
biomass
mented an increase in variability and a decrease in biomass
predict- is influenced by zooplankton immigration. To
ability of the phytoplankton assemblage of a lake during
facilitate comparison with previous studies, we focus
long-term experimental eutrophication. A nutrienton pulse is
temporal variability in biomass, but we also consider
therefore expected to increase community variabilityamong-replicate
by act- variability, an important but often over-
ing as a disturbance that also increases resource availabil-
looked aspect of variability, as well as variation in species
ity - although the increase in resource availability composition,
may not and discuss how different measures of vari-
be maintained over a sufficiently long time-period ability
to causecan lead to different conclusions.
a long-term change in variability.
Unfortunately, "variability" has a variety of conceptual Methods
and mathematical definitions, and perception of variability
Experimental design and sampling
can be influenced by sampling frequency and spatial extent
(Gaston & McArdle, 1994), making it difficult to compare Between 22 June and 27 July 2004, we conducted a
2x2
results among studies. For example, variability can be factorial-design mesocosm experiment in George
inter-
Lake,
preted as measuring magnitude of change through Killarney Provincial Park, Ontario, Canada
time,
variation around an average temporal trajectory, or (46° 01. 36' N, 81° 24.71' w). Mesocosms were 16 chemi-
variation
callyused
among experimental units. Several recent studies have inert plastic bags 1 m in diameter and 10 m deep,
in which
the temporal coefficient of variation (CV) as a measure of we manipulated zooplankton immigration rate
variability (Bai et al., 2004; Lepš, 2004; Caldeiraandet nutrient
al, levels. On 17 June, prior to the start of the
2005). The CV of an aggregate community property experiment,
(such we filled mesocosms with epilimnetic water
as biomass) is a function both of variances of allfrom the surrounding lake, filtered through 50-ļim mesh.
compo-
nent species and the covariances among all pairs ofZooplankton
species, were collected on the following day by taking
18-m vertical
scaled to the total biomass of the community (Lehman & tows with a 50-ļim mesh net from an area near
the enclosures;
Tilman, 2000). Negative covariances can reflect negative animals were added to enclosures at ambient
densities, after removal of aquatic mites. Each week, start-
interactions between species or preferences for different
environmental conditions; in any case, a more negative
ing on 23 June, we introduced colonist animals from 6 other

138

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
 ÉCOSCIENCE, vol. 14 (2), 2007

lakes to half of the mesocosms at a rate of approximately excluded from all analyses of compositional variability but
10% of the mesocosm population size per week. Heat- were included in analyses of biomass variability.
killed animals were added as a control to the remaining To quantify aggregate temporal variability, we calculat-
mesocosms. The source lakes were located within 20 km of
ed the temporal coefficients of variation (CVs) of total chlo-
George Lake and were chosen to encompass a broad range rophyll a and total zooplankton biomass for each replicate
of productivity and to maximize the diversity of the colonistover all post-treatment sampling dates. Because both total
species pool; although there was substantial overlap in spe-chlorophyll a and zooplankton biomass tended to increase
cies composition with George Lake, including several of thein response to the nutrient pulse, we also used a detrended
dominant species, an additional 15 new species (not previous-measure of temporal variability: we calculated the tempo-
ly recorded in George Lake) were detected in the colonist ral CVs of the residuals of the best-fit quadratic regression
pool. This dispersal treatment was crossed with a nutrientof loge-transformed zooplankton biomass or chlorophyll a
pulse treatment, consisting of the one-time addition of N versus time for each treatment. These detrended CVs rep-
and P fertilizer (NH4N03 and Na2HP04; Fisher Scientific, resent the variation (through time) in the deviation of each
Pittsburgh, Pennsylvania, USA) at a level of 10 jigP-L-1 replicate from that treatment's average trajectory; they were
and a N:P ratio of 40:1 by mass. This level of enrichment calculated as the temporal standard deviation of the residu-
exceeds the strictest provincial water-quality guidelines forals divided by the mean absolute value of the residuals. We
oligotrophic lakes like George Lake (Ontario Ministry of used 2-factor ANOVA to test for treatment effects on CVs
Environment and Energy, 1994) and was chosen to repre- and detrended CVs.
sent a severe but realistic perturbation for our study system Temporal variation in species composition was evalu-
(P levels in George Lake in July 2004 were 3-5 jxg-L-1).
ated for each replicate by (a) calculating the Bray-Curtis
Nutrients were added to half the mesocosms (denoted in thedistance (Quinn & Keough, 2002) between initial com-
text as the + nutrients treatment; untreated mesocosms aremunity composition (on the first post-treatment sampling
denoted as the 0 nutrients treatment) during the first week date) and final community composition and (b) calculating
of the experiment, after the first addition of colonists. Each
the mean Bray-Curtis distance between all adjacent pairs of
week, we took 2 (pseudo-) replicate samples of water (usingpost-treatment sampling dates. The former measures com-
an 8-m integrated tube sampler) and zooplankton (using a positional change over the course of the entire experiment,
15 -cm-diameter, 50-ļim mesh net) from each enclosure andwhile the latter measures average week-to-week change.
pooled these 2 samples for each enclosure to give one set of To understand the role of individual species' responses
initial (pre-treatment) samples and 5 sets of post-treatment
in overall community variability, we used the temporal
samples. Water samples were analyzed for chlorophyll CV a of the abundances of 6 of each of the most abundant
using a TD-700 fluorometer (Turner Designs, Sunnyvale, crustacean and rotifer species (without detrending); note
California, USA), following 24-h methanol extraction. To that all 12 were established species in George Lake. We
calculate zooplankton biomass, we enumerated crustaceans conducted ANOVAs for each species, applying a sequen-
and rotifers in subsamples of a volume that depended on tial Bonferroni correction for the number of species tested
species relative abundances in each sample; this protocol(Hochberg, 1988). For each replicate, we also decomposed
increases chances of detecting rare species without distort- the temporal CV of total biomass into its mathematical
ing calculations of individual species' densities or relative components by calculating summed variances, summed
abundances. We then estimated biomass by measuringcovariances among all pairs of species, and total biomass
a subset of animals in each sample and using published to understand the role of each of these in determining
length-weight regressions (crustaceans) or using publishedthe overall effect of each treatment on the CV (the CV
weights (rotifers). A more detailed dçspription of fieldis a positive function of variances and covariances but is
and laboratory methods is provided elsewhere (Forrest & inversely related to total biomass). We conducted 2-factor
Arnott, 2006). ANOVAs on each of these variables.

Measures of variability and statistical analysis We used RM-ANOVA to test for differences among
treatments in chlorophyll a across weeks. We used loge-
To quantify among-replicate variability in total chloro-
transformation to improve variance homogeneity when nec-
phyll a and zooplankton biomass, we performed a median-
essary. Statistica v.6.0 (StatSoft Inc., 2001) was used for
log Levene's test by calculating the absolute deviation of
all statistical analyses except NMDS, which was performed
each replicate from the median loge-transformed value for
using PC-Ord v.4.17 (McCune & Mefford, 1999).
that treatment in each week and conducting a repeated-mea-
sures analysis of variance (RM- ANOVA) on these standard- Results
ized deviates. This method has been shown to be robust to
departures from normality (Schultz, 1985). Among-replicate Chlorophyll a
variability in species composition was assessed by first Nutrient enrichment caused a strong increase in phyto-
performing non-metric multidimensional scaling (NMDS;plankton density and/or biomass, indicated by a marked dif-
described in detail in Forrest & Arnott, 2006) on speciesference in chlorophyll a concentrations between + nutrients
densities in the final week of the experiment and then run-and 0 nutrients enclosures, first apparent on the second post-
ning a median-Levene's test on the untransformed NMDS treatment sampling date and persisting until the end of the
axis scores (a single axis was found to be significant in thisexperiment (RM-ANOVA, nutrients Fx 12 = 93.9, P< 0.0001;
case). Unidentifiable immature copepods (nauplii) weretime X nutrients F4 48 = 15.2, Huýnh-Feldt-adjusted

139

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
Forrest & Arnott: Zooplankton community variability

dency in
P< 0.0001). Pre-treatment chlorophyll concentrations for species to respond similarly to changing condi-
tions.
enclosures were 0.94 ± 0.28 (ig L-1 (mean ± SD); in Summed covariances of introduced species (those tha
nutri-
ent-enriched enclosures these were 2.67 ± 0.91 ļj.g-L1never
on theoccurred in 0-dispersal enclosures) were significantly
last sampling date. lower than the average for all other species (Mann-Whitney
Across all weeks, there was no significant effect ř/8 8 for all + dispersal enclosures = 64, P = 0.0008), but not
of either nutrients or dispersal on among-replicate significantly
vari- less than zero (mean ± SE = -0.0001 ± 0.034).
Although
ability in total chlorophyll a, although a non-significant there was no significant effect of dispersal on any
component of the CV (P >0.1), it tended to reduce both vari-
nutrients x time interaction (Table I) reflected a tendency
for among-replicate variability to increase throughancestime and
to covariances in the presence of nutrients (Figure 2)
A
a greater extent in + nutrients enclosures (comparison ofnon-significant tendency (Tukey's HSD, P = 0.18) for dis-
linear trends, Fxxl = 4.98, P = 0.046). persal to increase variability in the absence of nutrients was
driven mainly by an apparent weak negative effect of disper
Nutrient enrichment significantly decreased the
sal on total biomass (ANOVA on loge-biomass, Fx X2 = 3.0,
temporal CV of total chlorophyll a (ANOVA, nutrients
P = 0. 1 1) without a corresponding negative effect on variances
Fx 12 = 6.15, P = 0.03) because mean chlorophyll a con-
or covariances. This "effect", however, was driven by chance
centration increased to a greater extent than did its standard
initial differences among treatments in total biomass rather
deviation, which increased with enrichment (ANOVA on
than any real effect of dispersal (i.e., there was no significan
loge-SD, nutrients Fx 12 = 35.7, P < 0.0001). Temporal vari-
time x dispersal treatment interaction). Nevertheless, these ini-
ability of detrended chlorophyll a, measured as the CV of
tial differences should not distort the overall pattern in the CV.
the residuals of quadratic regressions of loge-chlorophyll a
There was no effect of either treatment on week-
versus time, was not significantly affected by either treat-
ment (ANOVA, P > 0.4). to-week variation in zooplankton community composi-

Zooplankton
Table I. Repeated-measures median-log Levene's tests for
There was no effect of either treatment or their among-replicate
interac- variability in total chlorophyll a and total zoo-
tion on among-replicate variability in total zooplankton
plankton biomass.
biomass (Table I), but nutrients significantly decreased
chlorophyll a zooplankton biomass
among-replicate variation in species composition, as reflected
Source df F P F P
in the coordinates on the single significant NMDS axis
Between-subjects
(NMDS r 2 = 0.90; median-Levene's test on NMDS axis
nutrients 1 1.36 0.27 0.12 0.74
scores, Fx 12 = 5.45, P = 0.04; Figure 1). However, by
dispersal 1 0.81 0.39 1.11 0.31
chance, 3 of our experimental replicates (one in nutrients
each of x dispersal 1 0.04 0.85 0.00 0.95
3 treatments) had unusual species compositions from
error the 12
beginning of the experiment: one enclosure had high
W num-
ITHIN- SUBJECTS (HUYNH-
time 4 1.72 0.16 1.08 0.38
bers of the predator Leptodora kindtii and correspondingly
time x nutrients 4 2.38 0.07 0.77 0.55
low numbers of its bosminid prey, and 2 enclosures had
time x dispersal 4 0.84 0.51 0.77 0.55
extremely low densities of rotifers. To verify that these
timeaber-
x nutrients x dispersal 4 1.78 0.15 1.26 0.30
rant replicates were not responsible for the significant error
result, 48

we re-ran the analysis without them. The significant effect was

maintained (and, in fact, became stronger). Either way, there
was no effect of dispersal or a dispersal x nutrients interaction
2.5 -i

on among-replicate compositional variability (P > 0.25).

2.0 - -T-
The temporal CV of total zooplankton biomass (without
detrending) increased significantly with nutrient enrichment
1.5-
(ANOVA, nutrients FXX1 = 17.2, P = 0.001), and there was
a significant nutrients x dispersal interaction (Fx n = 4.85,
5®

8 1.0-
P = 0.048), indicating a destabilizing effect of nutrients 5«

only in the absence of dispersal (Tukey's HSD, /* = 0.003;

.22 ■
Figure 2). However, these effects disappeared after remov-
TU
.22 g 0.5- Ļ^J ■ A -1-
ing the increasing trend: the detrended CV, using residuals
S °-
On
from loge-biomass quadratic regressions, was not signifi-
°-
£
cantly related to either treatment (ANOVA, P > 0.4).
-0.5 - - I - i
The increase in temporal C V of total zooplankton biomass
with enrichment (before detrending) was a result of strong řžn
positive effects of nutrients on both the summed variances -io- řžn LJ
and summed covariances among species (ANOVA on loge-
-1.5-1
transformed variables, Fx 12 = 1 16.6, P < 0.0001 for variances, + nutrients 0 nutrients
Fx 12 = 38.3, P < 0.0001 for covariances), which overwhelmed
Figure 1. Variation among replicates in final zoop
thè positive effect of nutrients on total biomass (ANOVA on
composition, as reflected in NMDS axis scores. Box-
loge-biomass, Fx X2= 101.9, P< 0.0001; Figured).an,
Summed
25% quartiles, and range of values. Black square
triangles
covariances were positive in all replicates, indicating a ten- = 0 dispersal (n = 4).

140

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
 ÉCOSCIENCE, vol. 14 (2), 2007

60n tion (ANOVA on mean Bray-Curtis distan

Figure 3a). However, nutrients decreased
50- T positional change between initial and
dates (ANOVA, Fì ^ = 9.79, P = 0.009).
40 - nutrients x dispersai interaction (ANOVA
? ^ P = 0.02) indicated that the stabilizing effec
occurred only in the presence of dispersa
> 3°- P = 0.007, Figure 3b).
U
20- There were both positive and negative
effects of nutrients on the temporal variabil
10-
species, but no effects of dispersal (Table
showed signs of a nutrients x dispersal i
these effects were not significant after Bon
0-1
tion (Table II, Figure 4). Nutrient effects on
200 -1 of abundance appeared to be unrelated to eit
group or direction of nutrient effects on m
for example, although nutrients increased th
/^s
-1-

' 150 " T both Gas tropus s ty lifer (a rotifer) and Eu

fi
a T A cen (a cladoceran), the former decreased i
response to nutrient enrichment while the
0
¥ 100- II 1 (Figure 4; Forrest & Arnott, 2006). However
a

*C
03
a)
1 50- 1.0 -p

■ s 0.8-
OH

200 -,

0.6-

ír 150 - 0.4-
® A
S
• wo T
■ ï
r loo- _L „ 0.2-
o
e a
«

■S
«
i
-i-
Vi

'■3 OH
5°-
Ī b)
5 10 r

oj
« 0.8-
200 1

I
0.6 - s g
-T 15°" -r T i
= ■
M _L 0.4-
£
s 100-
ä
S 0.2-
#g
2 Ä
w 50- M
0-1
+ nutrients 0 nutrients

oj
Figure 3. Temporal variability in zooplankton community
+ nutrients 0 nutrients measured as Bray-Curtis distances between samples. Bray
can range from 0 (identical composition) to 1.0 (maximum
Figure 2. Components of temporal variability a) Meanofweek-to-week dissimilarity and b) dissimilarity
total zooplankton
biomass in experimental enclosures. Black squares = + dispersal;
(post-treatment) white
and final samples. Black squares = + disp
triangles = 0 dispersal (n = 4). Error bars are ± angles
SE. = 0 dispersal (n = 4). Error bars are ± SE.

141

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
Forrest & Arnott: Zooplankton community variability

Table II. Results of 2-factor ANOVAs on temporal CVs (calculated without detrending). F ratios have 1 and 12 df. The direction
ment effect is indicated by a + or - sign after the appropriate P- value. Values in bold are significant after sequential Bonferroni

nutrients dispersal nutrients x dispersal

Species F P F P F P
Copepoda
Diacy clops thomasi (Forbes) 162.70 <0.0001 (-) 2.65 0.13 (-) 2.41 0.15
Leptodiaptomus minutus (Lilljeborg) 0.77 0.40 (+) 0.48 0.50 (-) 7.62 0.02
Cladocera
B osmina s. str. sp. 7.24 0.02 (+) 1.92 0.19 (-) 1.56 0.24
Eubosmina longispina (Leydig)* 37.60 <0.0001 (+) 2.27 0.16 (-) 0.27 0.62
Eubosmina tubicen (Brehm) 71.40 <0.0001 (+) 1.59 0.23 (-) 0.30 0.59
Holopedium gibberum Zaddach 5.36 0.04 (-) 1.08 0.32 (+) 5.13 0.04
Rotifera
Conochilus sp. 8.54 0.01 (-) 0.13 0.72 (-) 0.03 0.86
Gastropus stylifer Imhof 13.60 0.003 (+) 1.13 0.31 (+) 0.16 0.69
Kellicottia longispina (Kellicott) 1.08 0.32 (-) 0.07 0.80 (-) 0.41 0.53
Keratella cochlearis (Gosse) 44.40 <0.0001 (-) 0.06 0.81 (-) 1.10 0.31
Keratella taurocephala Myers* 0.47 0.51 (+) 0.18 0.68 (-) 0.01 0.92
Polyarthra spp.

* The assump

Figure 4. Tem
and triangles
on y-axes.

species showing the greatest change in abundance withdance was driven by directional change. Exceptions are
nutrients (whether positive or negative) were also thosePolyarthra spp. and Leptodiaptomus minutus , abundant taxa
showing a positive effect of nutrients on the CV, indicatingthat increased markedly following nutrient enrichment but
that much of the effect on temporal variability of abun- showed no significant effect of nutrients on CV. Species

142

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
 ÉCOSCIENCE, vol. 14 (2), 2007

that were significantly less variable with the nutrient pulse however, immigration tended to decrease (rather than faci
(Diacyclops thomasi , Keratella cochlearis) were those whose tate) compositional change in the presence of the nutrien
abundances were maintained relatively constant with nutri- pulse, while tending to increase change in the absence
ents but tended to decline without nutrients. Dispersal had no nutrients, a pattern similar to that observed for zooplankto
significant effects on the CV of individual species (Table II). biomass variability.
The destabilizing effect of nutrients on zooplankton
Discussion biomass was only observed in the absence of dispersal.
Effects of disturbance and dispersal on community the + nutrients enclosures, dispersal tended to increase st
VARIABILITY
bility, owing to a tendency for dispersal to reduce variance
and covariances among species. This effect, too, may
Contrary to expectations, temporal variability of been
have chlo-produced by the higher proportion of predator
rophyll a decreased following the nutrient pulse, while
species in the colonist pool: predators likely (a) covari
zooplankton biomass variability increased - atnegatively
least basedwith their prey and (b) dampened populatio
on CVs calculated in the usual way, without prior defend-
increases in nutrient-enriched enclosures, thereby reducing
ing. It is possible that this difference resulted from increasedHolyoak and Lawler (1996) showed a similar
variances.
regulation of algal biomass by zooplankton grazing follow- effect of dispersal in a predator-prey micr
stabilizing
ing nutrient enrichment; herbivorous zooplankton, cosminmetacommunity.
con-
trast, were only weakly regulated by predators (see below).
In the final week of the experiment, enclosures that
However, the timeframe of the experiment likely played a
had received a nutrient pulse were more similar to ea
role in the perception of variability. The increase in phyto-
other in terms of zooplankton species composition, bu
plankton biomass in response to nutrient enrichment began
less similar in terms of algal biomass, than were undi
within a week of nutrient addition, while the zooplankton
turbed enclosures, while the nutrient pulse had no effect
biomass response only became pronounced a week later.
on among-replicate variability in zooplankton biomas
Thus, we might have observed a reduction or reversal in the
Interestingly, increased compositional similarity follow
effect of the nutrient pulse on zooplankton had it been pos-
ing the nutrient pulse occurred despite differences amon
sible to monitor community dynamics over a longer period.
replicates in algal biomass and did not depend on dispe
The nutrient pulse produced a stabilizing sal,
effect on
demonstrating that a consistent, predictable response
some zooplankton species, apparently because it to
prevented
disturbance was possible even in closed communitie
population declines. Overall, nutrients decreased
Thiscompo-
is an unexpected result, given that perturbations are
sitional change in the zooplankton community over the
expected to increase among-replicate variability (Warwick
course of the experiment - a result that runs counter to that1993) and that predictable responses of zoo
& Clarke,
of Rusak et al (2001), who found increased compositional
plankton to habitat alterations have elsewhere been shown
change following a perturbation - although the shorteron the availability of colonists (Cottenie & D
to depend
(subannual) temporal scale of the present study, as well
Meester, 2004; Leibold & Norberg, 2004). Instead, pertu
as the difference in the type of perturbation, may partly
bation seems to have driven increased homogenization
explain the different results. local zooplankton communities.
In the absence of the nutrient pulse, dispersal tended to
In summary, disturbance was not uniformly destabiliz-
destabilize zooplankton biomass, owing to a tendency for
ing to the aquatic ecosystem under study. The nutrient pul
dispersal to increase covariances (and, to a lesser extent,the temporal variability of algal biomass an
decreased
variances) relative to total biomass. Although abundances
the mecha- of some zooplankton species, while increas
nism for this effect is unclear, a greater relativeing
abundance
the temporal variability of total zooplankton biomass
of predators (or large omnivores) in the colonist pool
over the time-span of the experiment - although only in th
compared to the resident assemblage (Forrestabsence & Arnott,
of immigration. The role of immigration appears t
2006) may have played a role: Halpern et al. (2005) havedriven by an increased predator/omnivore pres
have been
suggested that introductions of predators canence destabilize
rather than by increased species richness. Contrary
herbivore communities. Given the increased species rich-
expectations, the disturbance failed to increase tempor
ness in + dispersal enclosures (Forrest & Arnott,variation
2006) andin community composition, and it decreased het
the expected negative relationship between species richness among replicate zooplankton communities,
erogeneity
and community variability (Doak et al., 1998;effect Loreau et did not depend on community openness. Studie
that
al., 2002), it would have been reasonable to predict lowerover longer time-scales and under a variety
conducted
variability in the + dispersal treatment. However,environmental
the effects conditions should help determine the gene
of immigration on variability appear to depend ality
on context
of these results.
and on the identity of the introduced species. Dispersal
Variability and scale in mesocosm studies
increased species richness mainly by adding rare species,
which are unlikely to contribute substantially to a Studies
reduc- of plankton community dynamics may be car-
ried out
tion in total community variability (Cottingham, at scales ranging from cattle tanks up to whole
Brown
& Lennon, 2001). Dispersal had no significantlakes.
effectsTheon
former approach allows greater replication and
temporal variation in community composition control over characteristics of the study system, but may
(i.e., dis-
tance between initial and final assemblages), be
indicating
more difficult to extrapolate to the complex dynamics
of large
that it was not a major driver of compositional natural habitats. In contrast, whole-lake experi-
change;

143

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
Forrest & Arnott: Zooplankton community variability

ments may sacrifice replication for greater realism.between

Large, fluctuations about a mean and directional change
within-lake mesocosms are a compromise approach that
within the time-span of interest.
allows for larger spatial scale, larger population sizes, and
This study highlights the need to consider multiple
closer approximation to the habitat complexity of real lakes
components of variability in community ecology. Different
than can be achieved in cattle tanks, while still allowing
metrics may lead to very different interpretations of the
for greater replication than whole-lake studies. However,effects of disturbance and the importance of dispersal in
even in mesocosms as large as those used in this study, the
regulating variability. Thus, for example, although zoo-
duration of experiments is limited by periphyton growth plankton biomass variability, in terms of temporal CV,
and deterioration of the enclosures. Because measures of
appeared to increase in response to the nutrient pulse (an
variability are highly dependent on the temporal scale of effect modified by dispersal), the actual change in biomass
sampling (McArdle, Gaston & Lawton, 1990), it is possible followed a relatively regular, rather than erratic, trajectory
that a much longer experiment, in which zooplankton and (no treatment effect on detrended CV), and final biomasses
algal biomass would have had time to return to pre-distur-were no less predictable with the nutrient pulse than without
bance levels, would have resulted in different conclusions. (no effect on among-replicate variability). Also, variation in
Nonetheless, we believe that a 5 -week study with week-one species or trophic level may be unrelated to variation
ly sampling should give adequate resolution at least forin other species or trophic levels (cf. Halpern et al., 2005).
rapidly-reproducing asexual rotifers and cladocerans; fur- Additional thorough investigations of community-level
thermore, variability in total zooplankton biomass at this variability should deepen our understanding of the ways in
time-scale may be relevant to the growth of larger con-which different components of variation are linked.
sumers (e.g., larval fish). Similarly, a more severe perturba-
tion than our nutrient pulse would undoubtedly have altered
some of our results. For example, Cottingham and Schindler
Acknowledgements
(2000) observed some qualitatively different patterns in We thank Killarney Provincial Park for permission to conduct
pond community responses to "large" ( > 150 jig P L-1) and research in the park, the Cooperative Freshwater Ecology Unit
"small" (10 jigP-L-1) nutrient pulses. However, a pulse ofin Sudbury (especially J. Heneberry, B. Keller, and G. Morgan)
10 jig P L-1 represented a> 100% increase over ambient for logistical support, and M. Cuke, A. Derry, A. Kalyniuk, M.
levels in our study lake. A larger pulse might have elic-Neytcheva, and A. Strecker for assistance in the field. Comments
from C. Eckert, T. Wheeler, and three anonymous reviewers great-
ited different responses by zooplankton and algae but would
ly improved the manuscript. Funding was provided by a Natural
have been less realistic for our oligotrophic study system.
Sciences and Engineering Research Council (NSERC) Discovery
Such trade-offs between realism and the need to produce Grant to S. E. Arnott and a Julie Payette-NSERC Postgraduate
detectable effects are unavoidable.
Scholarship to J. Forrest.

Components and measures of variability

Literature cited
The perception of temporal variability in this study
Bai,in
depended strongly on the presence of temporal trends Y., X. Han, J. Wu, Z. Chen & L. Li, 2004. Ecosystem stabil-
ity and compensatory effects in the Inner Mongolia grassland.
both chlorophyll a and total zooplankton biomass. Even
Nature, 431: 181-184.
temporal variation in composition was affected by consider-
ing average week-to-week change (no treatment effect) asCaldeira, M. C., A. Hector, M. Loreau & J. S. Pereira, 2005.
Species richness, temporal variability and resistance of biomass
opposed to total change from beginning to end of the experi-
production in a Mediterranean grassland. Oikos, 110: 115-123.
ment (effect of nutrients and a significant nutrients x disper-
sal interaction). Removal of the trends in chlorophyll a andS. E. & E. E. Prepas, 1998. Stream phosphorus and
Cooke,
nitrogen export from agricultural and forested watersheds on
zooplankton biomass by using residuals from regressions
the Boreal Plain. Canadian Journal of Fisheries and Aquatic
resulted in the loss of any significant treatment effect on
Sciences, 55: 2292-2299.
temporal variability. Detrending is not straightforward, how-
Cottenie, K. & L. De Meester, 2004. Metacommunity structure:
ever, and may not be justifiable. We chose to use quadratic
Synergy of biotic interactions as selective agents and dispersal
regression in this case, as this provided reasonably goodas fit
fuel. Ecology, 85: 114-119.
to most temporal trends (mean r2 = 0.66; range 0.44-0.85);
Cottingham, K. L., B. L. Brown & J. T. Lennon, 2001. Biodiversity
but there was no a priori reason to choose a quadratic curve.
may regulate the temporal variability of ecological systems.
A poor fit (low r2) for a given treatment will mean larger
Ecology Letters, 4: 72-85.
residuals and therefore likely a higher standard deviation of
Cottingham, K. L., J. A. Rusak & P R. Leavitt, 2000. Increased eco-
residuals. We avoided this problem by using the temporal system variability and reduced predictability following fertili-
CV of residuals, but this may be an overly conservative
sation: Evidence from palaeolimnology. Ecology Letters, 3: 340-348.
approach because it reduces the possibility of attributing
Cottingham, K. L. & D. E. Schindler, 2000. Effects of grazer com-
large deviations from the trend to temporal variability. munity
An structure on phytoplankton response to nutrient pulses.
even more conservative approach, which would remove Ecology,
any 81: 183-200.
trend without resorting to arbitrary curve-fitting, would
Doak,beD. F., D. Bigger, E. K. Harding, M. A. Marvier, R. E.
to examine the temporal standard deviation in a given rep-
O'Malley & D. Thomson, 1998. The statistical inevitabil-
licate 's deviation from the treatment median for each sam-
ity of stability-diversity relationships in community ecology.
pling date. However, trends are an important component of
American Naturalist, 1 5 1 : 264-276.
variability because short-term monotonie trends may form J. & S. E. Arnott, 2006. Immigration and zooplankton
Forrest,
community responses to nutrient enrichment: A mesocosm
part of longer-term fluctuations and, more simply, because
experiment. Oecologia, 150: 119-131.
they reflect change. Nonetheless, it is worth distinguishing

144

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms
 ÉCOSCIENCE, vol. 14 (2), 2007

Gaston, K. J. & B. H. McArdle, 1994. The temporal variabil- Meyer, J. L. & G. E. Likens, 1979. Transport and transforma-
ity of animal abundances: Measures, methods and patterns. tion of phosphorus in a forest stream ecosystem. Ecology, 60:
Philosophical Transactions of the Royal Society of London, B, 1255-1269.
Biological Sciences, 345: 335-358. Micheli, F., K. L. Cottingham, J. Bascompte, O. N. Bjornstad, G.
Halpern, B. S., E. T. Borer, E. W. Seabloom & J. B. Shurin, 2005. L. Eckert, J. M. Fischer, T. H. Keitt, B. E. Kendall, J. L. Klug
Predator effects on herbivore and plant stability. Ecology & J. A. Rusak, 1999. The dual nature of community variability.
Letters, 8: 189-194. Oikos, 85: 161-169.
Hochberg, Y., 1988. A sharper Bonferroni procedure for multiple Odum, E. P., J. T. Finn & E. H. Franz, 1979. Perturbation theory
tests of significance. Biometrika, 75: 800-802. and the subsidy-stress gradient. Bioscience, 29: 349-352.
Holt, R. D., 2002. Food webs in space: On the interplay of dynamic Ontario Ministry of Environment and Energy, 1994. Water
instability and spatial processes. Ecological Research, 17: Management: Policies, Guidelines, Provincial Water Quality
261-273. Objectives of the Ministry of Environment and Energy. Queen's
Printer for Ontario, Toronto, Ontario.
Holy oak, M. & S. P. Lawler, 1996. The role of dispersal in preda-
Quinn, G. P. & M. J. Keough, 2002. Experimental Design and
tor-prey metapopulation dynamics. Journal of Animal Ecology,
65: 640-652. Data Analysis for Biologists. Cambridge University Press,
Cambridge.
Lehman, C. L. & D. Tilman, 2000. Biodiversity, stability, and
Rosenzweig,
productivity in competitive communities. American Naturalist, M. L., 1971. Paradox of enrichment: Destabilization
156: 534-552. of exploitation ecosystems in ecological time. Science, 171:
385-387.
Leibold, M. A. & J. Norberg, 2004. Biodiversity in metacommu-
nities: Plankton as complex adaptive systems? Limnology Rusak,
& J. A., N. D. Yan, K. M. Somers, D. J. McQueen & C. W.
Oceanography, 49: 1278-1289. Ramcharan, 2001. Differences in the variability of crustacean
zooplankton communities between manipulated and reference
Lepš, J., 2004. Variability in population and community biomass in
lakes. Archiv für Hydrobiologie Special Issues Advances in
a grassland community affected by environmental productivity
Limnology, 56: 171-186.
and diversity. Oikos, 107: 64-71.
Schultz, B. B., 1985. Levene's test for relative variation.
Loreau, M., N. Mouquet & A. Gonzalez, 2003. Biodiversity as spa-
Systematic Zoology, 34: 449^56.
tial insurance in heterogeneous landscapes. Proceedings of the
National Academy of Sciences of the USA, 100: 12765-12770. StatSoft, 2001. Statistica (data analysis software system), Version
6. StatSoft Inc., Tulsa, Oklahoma.
Loreau, M., A. Downing, M. Emmerson, A. Gonzalez, J. Hughes,
Steiner,
P. Inchausti, J. Joshi, J. Norberg & O. Sala, 2002. A new look C. F., 2005. Temporal stability of pond zooplankton
assemblages. Freshwater Biology, 50: 105-112.
at the relationship between diversity and stability. Pages 79-91
Steiner,
in M. Loreau, S. Naeem, & P. Inchausti (eds.). Biodiversity and C. F., Z. T. Long, J. A. Krumins & P. J. Morin, 2005.
Ecosystem Functioning: Synthesis and Perspectives. Oxford Temporal stability of aquatic food webs: Partitioning the effects
University Press, Oxford. of species diversity, species composition and enrichment.
McArdle, B. H., K. J. Gaston & J. H. Lawton, 1990. Variation Ecology
in Letters, 8: 819-828.
Stemberger, R. S., 1979. A Guide to Rotifers of the Laurentian
the size of animal populations: Patterns, problems and artefacts.
Journal of Animal Ecology, 59: 439-454. Great Lakes. Environmental Monitoring and Support
McCune, B. & M. J. Mefford, 1999. Multivariate AnalysisLaboratory,
of US Environmental Protection Agency,
Cincinnati, Ohio.
Ecological Data, Version 4.17. MjM Software, Gleneden
Beach, Oregon. Stone, L. & D. Hart, 1999. Effects of immigration on the dynamics
McLaughlin, P. A., D. K. Camp, M. V. Angel, E. L. Bousfield, of simple population models. Theoretical Population Biology,
55: 227-234.
P. Brunei, R. C. Brusca, D. Cadien, A. C. Cohen, K. Conlan,
Tilman, D., 1999. The ecological consequences of changes in
L. G. Eldredge, D. L. Felder, J. W. Goy, T. Haney, B. Hann,
R. W. Heard, E. A. Hendrycks, H. H. Hobbs, J. R. Holsinger,biodiversity: A search for general principles. Ecology, 80:
1455-1474.
B. Kensley, D. R. Laubitz, S. E. LeCroy, R. Lemaitre, R.
F. Maddocks, J. W. Martin, P. Mikkelson, E. Nelson, W. A.
Vinebrooke, R. D., D. W. Schindler, D. L. Findlay, M. A. Turner,
M. Paterson & K. H. Mills, 2003. Trophic dependence of eco-
Newman, R. M. Overstreet, W. J. Poly, W. W. Price, J. W. Reid,
system resistance and species compensation in experimentally
A. Robertson, D. C. Rogers, A. Ross, M. Schotte, F. R. Schram,
acidified Lake 302S (Canada). Ecosystems, 6: 101-113.
C.-T. Shih, L. Watling, G. D. F. Wilson & D. D. Turgeon, 2005.
Common and Scientific Names of Aquatic Invertebrates Warwick,
from R. M. & K. R. Clarke, 1993. Increased variability
the United States and Canada: Crustaceans. American Fisheries as a symptom of stress in marine communities. Journal of
Society, Special Publication 31, Bethesda, Maryland. Experimental Marine Biology and Ecology, 172: 215-226.

145

This content downloaded from 44.235.20.64 on Sun, 13 Feb 2022 16:43:40 UTC
All use subject to https://about.jstor.org/terms