Download International Association for Ecology

International Association for Ecology
Variation in Predation Pressure as a Mechanism Underlying Differences in Numerical Abundance
between Populations of the Poeciliid Fish Heterandria formosa
Author(s): Jean M. L. Richardson, Margaret S. Gunzburger and Joseph Travis
Source: Oecologia, Vol. 147, No. 4 (Apr., 2006), pp. 596-605
Published by: Springer in cooperation with International Association for Ecology
Stable URL: http://www.jstor.org/stable/20445860
Accessed: 30-10-2015 16:42 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
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 [email protected].
Springer and International Association for Ecology are collaborating with JSTOR to digitize, preserve and extend access to
Oecologia.
http://www.jstor.org
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
Oecologia (2006) 147: 596-605
DOI 10. 1007/s00442-005-0306-y
Jean M.
L. Richardson
-Margaret
S. Gunzburger
Joseph Travis
Variation in predation pressure as a mechanism underlying differences
in numerical abundance between populations of the poeciliid fish
Heterandria formosa
Received: 31May 2005/ Accepted: 4 November
? Springer-Verlag 2005
2005 /Published online: 10 December 2005
Abstract We explored whether a variation in predation
and habitat complexity between conspecific populations
can drive qualitatively different numerical dynamics in
those populations. We considered two disjunct popula
tions of the least killifish, Heterandria formosa, that ex
hibit long-term differences in density, top fish predator
species, and dominant aquatic vegetation. Monthly
censuses over a 3-year period found that in the higher
density population, changes in H. formosa density
exhibited a strong negative autocorrelation structure:
increases (decreases) at one census tended to be followed
by decreases (increases) at the next one. However, no
such correlation was present in the lower density pop
ulation. Monthly census data also revealed that preda
tors, especially Lepomis sp., were considerably more
abundant at the site with lower H. formosa densities.
Experimental studies showed that the predation by
Lepomis gulosus occurred at a much higher rate than
predation by two other fish and two dragonfly species,
although L. gulosus and L. punctatus had similar pre
dation rates when the amount of vegetative cover was
high. The most effective predator, L. gulosus, did not
discriminate among life stages (males, females, and
juveniles) of H. formosa. Increased predation rates by L.
gulosus could keep H. formosa low in one population,
Communicated by David Post
S. Gunzburger
J.M. L. Richardson M.
Department of Biological Science,
Florida State University,
Tallahassee, FL 32306-4340, USA
J. Travis
J. M. L. Richardson (3)
Department of Biological Sciences, Brock University,
500 Glenridge Ave., L2S 3A1, St. Catharines, ON, Canada
E-mail: [email protected]
Tel.: + 1-905-6885550
Fax: + 1-905-6881855
Present address: M. S. Gunzburger
United States Geological Survey,
Florida Integrated Science Center,
7920 NW 71st St., Mexico, FL 32653-3701, USA
thereby eliminating strong negative density-dependent
regulation. In support of this, changes in H. formosa
density were positively correlated with changes in vege
tative cover for the population with a history of lower
density, but not for the population with a history of
higher density. Our results are consistent with the
hypothesis that the observed differences among natural
populations in numerical abundance and dynamics are
caused in part by the differences in habitat complexity
and the predator community.
Keywords Density Habitat complexity
Population dynamics -Population limitation
Species-specific predation
Introducffon
The debate over the relative importance of density
dependent and density-independent processes in deter
mining numerical dynamics is a long-standing one in
ecology (Sinclair 1989; Ellner and Turchin 1995; Sale
and Tolimieri 2000; Cooper 2001).While ecologists have
searched among species for patterns in the prevalence,
mode, and strength of density dependence (Hassell et al.
1976; Ellner and Turchin 1995;Kendall et al. 1998), the
literature on variation among conspecific populations in
numerical dynamics and variation in the strength and
mode of genuine regulation ismore limited (see refer
ences in Leips and Travis 1999). Yet, because different
levels of regulation are themost straightforward avenue
throughwhich density-dependent selection can generate
intraspecific variation in life history (Charlesworth
1994), these data can provide insight into the role of
density-dependent selection in nature. Despite the
prominence that notions of density-dependent life-his
tory evolution have attained (reviewed byMueller 1997),
there are few studies of natural populations that attempt
to delineate rigorously the relative importance of den
sity-dependent and density-independent processes and
attempt to discern if different density regimes are likely
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
597
to be the selective agents responsible for divergence in
life histories or other critical characters (Law et al. 1977;
Primack and Antonovics 1982;Bradshaw and Holzapfel
1989;Ohgushi 1995; Rossiter 1995).
Populations of the least killifish, Heterandria for
mosa, appear to be prime candidates for density
dependent life-history evolution. The least killifish is a
poecilliid native to lentic habitats, including lakes,
ponds, and river edges, in the coastal plain of the
southeastern United States. Populations in north
Florida exhibit consistent, remarkable differences in
abundance and patterns of numerical dynamics (Leips
and Travis 1999). Levels of genetic variation in
molecular markers are correlated with historical density
estimates (Soucy and Travis 2003), suggesting that
density regimes have been different among these pop
ulations for a long period. Ratios of predator density to
fish density vary widely among these populations, with
the highest ratios observed in the populations exhibit
ing the lowest densities (Leips and Travis 1999). Spatial
patterns of genetic variation among H. formosa popu
lations are consistent with those expected in popula
tions unconnected by gene flow (Baer 1998; Soucy and
Travis 2003). Genetically based life-history differences
among populations resemble the expectations from
density-dependent selection theory: females from the
populations with lower densities are larger in body size
and produce more, larger broods of smaller offspring
(Leips and Travis 1999; Leips et al. 2000). When den
sity is higher at a single location, females are smaller,
have lower lipid levels, and produce fewer, smaller
broods, but this effect was seen only in fluctuations in
theWacissa River (WR) population, the one with the
highest observed densities (Leips and Travis 1999).
These patterns are consistent with different levels of
regulation, with theWR population being the only one
studied with the potential for regulation in the strictest
sense (Sale and Tolimieri 2000) and with conspecific
density
acting
as a selective
agent
in the WR
popula
as a whole,
Field observations
We performed monthly censuses in two populations, TP
andWR, previously studied by Leips and Travis (1999)
thatwe chose based on known differences inH. formosa
densities and potential predator densities. See Leips and
Travis (1999) for these background data and for a de
tailed description of each habitat. TP has a persistently
low density of H. formosa and a predator community
dominated by warmouth sunfish, Lepomis gulosus. The
-5 ha pond has moderate density of submerged vege
tation, dominated by Myriophyllum laxum, a plant with
soft stems and fine, feather-like leaves (Leips and Travis
1999).WR has a high density of H. formosa and lower
densities of its predator community, which is dominated
by the spotted sunfish L. punctatus. The spring-fed river
contains still water areas with a high density of sub
merged vegetation, dominated by Hydrilla verticillata,
which has a solid structurewith strap-like pointed leaves
arranged in a whorl pattern along the length of its stem
(Leips and Travis 1999). Populations of H. formosa in
these two locations are genetically distinct (Baer 1998;
Leips and Travis 1999; Leips et al. 2000; Soucy and
Travis 2003).
We estimated average densities of H. formosa and
potential predator species as well as amount of vegeta
tive cover in the littoral habitat used by H. formosa with
methods described in detail in Leips and Travis (1999).
In brief, we visited each sitemonthly for 3 years (Sep
tember 1999-September 2002) and used nine replicate
placements of a 0.5-iM2 throw-trap to estimate H. for
mosa density and cover at each visit (three settings of the
trap in each of three sites at each location, TP and WR).
Cover was estimated visually as the percent surface area
of the bottom obscured by submerged vegetation within
the trap (the consensus assessment of two researchers
was
tion but not in the others.
Taken
Methods
these results
suggest
that the fac
tors influencing population dynamics may differ among
H. formosa populations and that the system might well
reflect the conditions necessary for density-dependent
selection to drive population divergence. We hypothe
sized that low-density sites, such as Trout Pond (TP),
have numerical dynamics that reflect limits placed on the
population through predation pressure, while a high
density population site, such as WR, has numerical
dynamics that reflect population regulation around the
carrying capacity of the environment. In this study, we
explore these possibilities by (1) collecting additional
monthly census data from two localities, TP andWR,
including estimates of relative densities of predator
species, (2) comparing predation rates on H. formosa of
themost abundant potential predators, and (3) consid
ering the interactive effect of vegetation type and
quantity (which also differed substantially between
locations) on predation rates of top fish predators.
used). We
estimated
density
of H. formosa
and all
other vertebrates and macroinvertebrates by removing
in the trap, hand
all animals
and
caught
sorting,
the total number
of each taxa. The 3-year
counting
period over which data were collected represents a span
of between
seven
and nine
generations
of H.
formosa,
based on developmental patterns in laboratory, meso
cosm studies (Leips et al. 2000; J. Travis and others,
unpublished data), and calibrated studies of otolith
increments (J.Travis and R. Allman, unpublished data).
Mean density estimates of H. formosa and potential
predators caught in the trapwere calculated separately
for each location. Our
for winter and summer months
and in particular
census methods
the size of the throw
to provide accurate estimates
trap we used were designed
of H. formosa densities. Potential predators differ sub
in body size and habits from H. formosa and
stantially
accurate
not provide
thus our sampling
regime may
measures
at each cen
of absolute density for predators
sus. However, for our purposes it is the density of pre
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
598
in one
dators
site
relative
site
to the another
is
that
important and since identical trapping techniques were
used at both the sites, these data, once averaged over the
> 30 sample
dates,
provide
an accurate
estimate
of rel
ative predator density.
To investigate the potential presence of density
dependent effectswithin eachH. formosa population, we
used a time series analysis, first log-transforming and
differencing the data to produce stationary series, and
then following themethods for stationary series, as de
scribed in Pankratz (1991) and Chatfield (1996). We
looked for autocorrelations within each H. formosa
population for the rate of change in fish density at
through
monthly
lags ranging from t = (i-12)
t=
1 year before
(1+ 12) (from
to 1 year after)
to look
for density dependence within the population.
Comparison of predator effects
To determine the effectiveness of different predators, we
quantified predation rates of five potential predators
most
that were
common
in field
data:
L. punctatus
(spotted sunfish), L. gulosus (warmouth), Aphredoderus
sayanus (pirate perch), and two species of Aeshnid
dragonfly larvae,Anaxjunius and Coryphaeschna ingens.
Because prey behavior experiments revealed no differ
ences in anti-predator behavior or predation escape rates
between the twoH. formosa populations (J.Richardson,
unpublished
we used WR
data),
H. formosa
in all pre
dation experiments.
We
established
a stock population
of H. formosa
in a
large ( 1000 1)holding tank in the greenhouse at the
FSU Biological Research Facility in Tallahassee. This
tank was
filled with
aged well water
and had constant
filtration and aeration. Approximately
40 1 of pond
(Hydrilla) were
water and 19 1of well-rinsed
vegetation
also added to the tank. H. formosa in this stock tank had
ad libitum food through the continual presence of a disc
of
and daily addition
of "Tetra Time Release Meals"
approximately 5ml of ground TetraMin? flake food,
supplemented with freeze-dried daphnia.
L. gulosus were caught in throw traps duringmonthly
censuses
at TP and L. punctatus
were
seined
from WR.
Aphredoderus sayanus were collected from WR with
throw traps. All fish were housed individually in 40 1
aquaria,
each with
a sponge
filter, a piece of pvc pipe
for
cover and vegetation appropriate to their origin (Hyd
rilla or Myriophyllum). Anax and Coryphaeschna were
collected with a dipnet from several different ponds in
Leon County, FL and were housed individually in 1.5 1
plastic tubs in the greenhouse with vegetation for cover
and perching. All predators were maintained on a diet of
H. formosa.
Predation rates of each predator were quantified
using unpatterned blue wading pools, 99 cm in diameter
and 15 cm high ('.0. 115 in3)with a thin layer of sand on
the bottom. Wading pools were emptied of water and re
filled between trials and each prey and predator indi
vidual was used only once. Myriophyllum (the amount
that could be loosely packed into a 1.5-1 bucket) was
added
to each pool
and
spread
out
in one-half
of
the
pool; another half of the pool was open water. We did
six replicates for each predator except C. ingens for
which we did four replicates.
Experiments with Aphredoderus sayanus, Anax ju
nius, and C. ingens started with nine female, six male,
and six juvenile H. formosa as prey. Experiments with
L. punctatus and L. gulosus as predators were carried
out in the context of a larger experiment (see Fish
predator and vegetation effects), which used eight fe
male, fivemale, and five juvenile H. formosa per pool.
H. formosa were haphazardly sorted into containers of
the appropriate number of females, males, and juve
niles; these sorted fish were assessed visually by two of
the authors for similarity of size distribution, and fish
substitutions made as necessary to equalize the size
distribution. Once sufficient numbers of H. formosa had
been sorted, they were placed into pools. Predators
were
added
1 h later and shade cloth was placed
on top
of pools to reduce light intensity and disturbance to the
experiments. After 48 h, predators were removed from
pools, pools were drained, and H. formosa remaining
were counted.
We analyzed both predation rates and selectivity by
predators on each H. formosa stage (male, female, and
juvenile) separately. Predation rate was defined as the
number of prey eaten divided by the number of prey
available (i.e., initially placed into the pool). Residual
analysis revealed these data to be heteroscedastic and
this could not be corrected by any of several transfor
mations tried. Thus, differences among predators in
predation rates were analyzed using a nonparametric
Kruskal-Wallis test. Predation rate data were also
combined with field density estimates to calculate a
predation pressure index for each predator at each
location; predation pressure index = average density of
predator x average per capita predation rate (Gunz
burger and Travis 2004).
We defined selectivity as the number of prey eaten
from a given stage (female or juvenile) divided by the
total number of prey eaten (calculated separately for
each predator individual). Males were not included in
the statistical analysis because their proportion is
determined by the first two.We also calculated the ex
pected proportions if females and juveniles are con
sumed in proportion to their abundance. For juveniles,
the expected proportion consumed was 0.28 for L.
punctatus
and
L.
gulosus
and
0.29
for
aeshnids
and
Aphredoderus sayanus. For females, the expected pro
portion consumed was 0.44 for L. punctatus and L.
gulosus and 0.43 for aeshnids and Aphredoderus sayanus.
To analyze whether predators were selectively preying
on different stages, we subtracted the observed from the
expected selectivity. As with predation rate, heterosce
dascity in these data could not be corrected by trans
formation of the data and we analyzed data using a
nonparametric Kruskal-Wallis test.
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
599
Fish predator and vegetation effects
In this experiment, we assessed the potential for vege
tation differences between habitats to modify the influ
ence of predator limitation on prey population
dynamics.We used the dominant sunfish and vegetation
from each site: L. gulosus and Myriophyllum for TP; L.
punctatus and Hydrilla forWR. We also considered two
levels of vegetation density. Thus, we had a three-factor
crossed design of predator species by vegetation species
by vegetation density.
Trials were run in nine wading pools inside the
greenhouse. Each pool was dedicated to its predator
treatment to ensure that no cross-treatment contamina
tion of predator odors could occur; vegetation used in
pools was likewise dedicated to a particular predator
treatment. Between trials, pools were randomly re-posi
tioned within the greenhouse to avoid confounding po
sition and predator treatment.We did seven runs to
complete six replicates of each of the eight treatment
combinations (we did not use all nine pools in each run).
Treatment combinations were randomly assigned to each
run,with the constraint that each predator was starved
for 48 h before being used in a trial. Predators were not
fed outside of experimental trialsunless theyhad failed to
eat any prey during the trial.We had six individuals of
each predator fish species and randomly assigned treat
ment order experienced by each fish; each fishwas used
once in each of the four vegetation treatments.The mean
size of individual fishpredators did not differ between the
two species (L. gulosusmean body length= 57.7 mm; L.
punctatusmean body length= 61.3 mm, two-tailed t-test
on ln(length) gives P > 0.30).
Vegetation amounts were standardized as one loosely
packed 1.5-1 bucket of vegetation for "low" and two
closely packed 1.5-1 buckets of vegetation for "high".
For both treatments, vegetation was confined to one
half of the pool. Hydrilla vegetation was collected from
WR. Myriophyllum was collected initially from TP and
supplemented with collections fromMoore Lake (Leon
County, FL, USA). All vegetation was carefully washed
to remove any invertebrates.
For each run of the experiment, H. formosa were
dipnetted from the greenhouse stock tank and sorted by
stage/sex. Once sufficient numbers were collected, fish
were haphazardly separated into groups of eight fe
males, fivemales, and five juveniles. Size distributions of
fish for each pool were equalized, as described above. To
determine if size-selective predation was occurring, we
photographed the H. formosa in each pool with a ruler
10.98 + 0.164. Prey assignment to pools was haphazard.
Prey were added to pools in themorning and predators
were added 1h later. Pools were then covered with two
layers of shade cloth and left undisturbed. After 48 h,
predators were removed and prey fish remaining were
counted. Pools were then drained, re-positioned, re-fil
led, and given 18-24 h to come into temperature equi
librium before the next trial.
Treatment differences in the number of fish surviving
were analyzed using a stratified analysis of contingency
tables (proc "freq" with "cmh" option in SAS). Size
selective predation was tested separately for L. gulosus
and L. punctatus predators with paired t-tests comparing
pre-predation mean prey sizewith post-predation mean
prey size. All statistics were performed in SAS for
Windows 8.01 (SAS Institute, Inc., Cary, NC 1999).
We found that density of cover does affect predation
rates (see Results) and therefore further analyzed our
field data using a time series analysis to look for cross
correlations between change in vegetative cover and
change in H. formosa density. Because predation rate
varies with the amount of vegetative cover, we propose
that change in vegetative cover correlateswith change in
predation rate and thus change in vegetative cover can
be used as a surrogate for changing predation rates.We
again used the following data transformation:
Zj =
logxj - logx1_l
where
xj is the value
of
the time series at time t=j
for
both H. formosa density and vegetative cover. We
estimated cross-correlations between the monthly rate
in cover
of change
at time
t=j
and
the monthly
rate
of change in fish density at monthly lags ranging from
t=j-12
through t=j+12
(from 1 year before to
1 year after). When significant autocorrelations were
present, we filtered the series by fitting ARIMA
models (Box et al. 1994) and used residuals to test for
cross-correlations (Pankratz 1991; Box et al. 1994;
Chatfield 1996).
Results
Field observations
Average densities of H. formosa fluctuated substantially
in both populations,
in each case by more
than an order
of magnitude, although neither population exhibited any
substantial overall net changes (Fig. 1). The density of
before adding them to pools and again at the end of the
trial to look for any change
in mean
size of each fish
H. formosa
tude lower
stage such selection would impose (photographs were
used tominimize handling of H. formosa).We measured
pre- and post-trial sizes using digital imaging software
for ten L. punctatus and 12 L. gulosus trials.Mean body
lengths of photographed prey added to pools were
rating data previously collected on these two sites (Leips
and Travis 1999; Soucy and Travis 2003). Time series
(mean
male,
+
SE
in mm):
n= 109,
females,
15.29+0.167;
n=
176,
15.59+0.268;
juveniles,
n= 107,
analysis
at TP was consistently
than that at WR
(Fig.
revealed
that at WR,
an order of magni
1, Table 1), corrobo
changes
in log(H. formosa
density) were significantly autocorrelated with changes
in log(H. formosa density) during the previous time
period. At TP, no significant autocorrelation structure
was present for changes inH. formosa density.
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
600
300
a
Wacissa
punctatus density divided by H. formosa density is 0.275
100
River
for TP and 0.00137
90
250
80
Percent cover
200
70
60
150
50
and
40
100
30
E
50-20
10 ~
IC
Density
w t 0
C~~~~~~~~~~~~~
for WR
in winter
and 0.0253
for TP
and 0.00085 forWR in summer; Table 1). This differ
ence in the number of potential predators per prey is
consistent with previous data from these sites (Leips and
Travis 1999). Vegetative cover at each monthly census
was greater inWR than TP (Table 1).Vegetative cover
also decreased between the latter part of the first year
-
>
0
t
the latter part of the third year, but
this effect was
similar at both the sites (Fig. 1). The decrease in vege
tation was caused in part by a drought that reduced
water levels throughout the region, drawing themargin
of the littoral zone downward faster than the typical
littoral zone vegetation could follow.
As expected, our sampling regime could not provide
accurate measures of absolute density for predators at
each census. This was clearly the case because the esti
mated density per trap was < 1 individual, individual
throw trap estimates thus being highly susceptible to
sampling error.We therefore could not do time series
analysis on potential predator species to look for cor
relations between predator and prey dynamics. The
relative difference in predator densities based on all
census data provides robust evidence that the relative
abundance of predators differs between the two loca
tions (Table 1).
dit
axes
Pond
co~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~i
0
90
1
p8
IL
6
1nu
14
80
h
-70
12
-60
10
50
8
40
6
30
00
4
20
2
20
bon TroePon
10
100eene inteseiso2oeta
0
Comparison of predator effects
0
Mar-00
Sep-99
1 Average
cover
percent
Fig.
Mar-01
of H. formosa
density
censuses
from monthly
2002
September
Sep-00
at two northern
Sep-01
Mar-02
Sep-02
(number
per in2) and
done from September
Florida
aWacissa
locations:
The time series analysis of field data done above suggests
thatWR H. formosa show density-dependent population
dynamics, but TP H. formosa do not. Further, popula
average
1999 to
River
tion
(WR) and b Trout Pond (TP). Density was estimated each month
with
three
0.5-in2
for three different
sites for
throw-trap
samples
on nine
location
is based
Error
(i.e., each average
samples).
bars are absent
for the purpose
that H. formosa
of clarity. Note
axes in (a) and (b) differ
in scale
density
ulate
each
We
clear differences
in the species of potential
present at each site (Table 1). The number of
per prey is two to three orders of
predators
higher in TP than inWR
(e.g., L. gulosus + L.
found
predators
potential
magnitude
densities
are much
lower
at TP
and
predator
numbers per prey are much higher, leading us to spec
that the TP H. formosa
population
may
be limited
by predation. We explored this possibility by comparing
the predation rates of different predators under con
trolled conditions. Predators differed in consumption
rates across all stages of H. formosa prey (females:
X24=19.80, P<0.001; males: X24=14.17, P<0.007;
juveniles: X24= 13.72, P < 0.009). While these three tests
are not independent, they are statistically significant
Table 1 Summary of field data from Trout Pond (TP) (Leon County, FL) andWacissa River (WR) (Jefferson County, FL) from monthly
0.5-m2 box-trap samples from September 1999 to September 2002
Locality
Percentage cover
Density
(per mi2)
H. formosa
Winter (September-March)
TP
35 (25)
WR
59 (17)
Summer (April-September)
TP
33 (20)
WR
56 (20)
Aeshnid
Aphredoderus sayanus
L. gulosus
L. punctatus
2.9 (2.2)
102 (47)
0.75 (0.81)
0.16 (0.28)
0
0.52 (0.40)
0.80 (0.77)
0.02 (0.1)
0
0.12 (0.36)
8.3 (6.2)
106 (71)
0.40 (0.48)
0.16 (0.33)
0
0.96 (0.84)
0.21 (0.3)
0.01 (0.05)
0
0.08 (0.12)
Each month, three box traps samples were taken from each of three sites atWR and TP. Average percent cover of submergent aquatic
vegetation (estimated visually) and average density (all animals collected per sample were counted live in the field) of H. formosa and four
predators (odonate larvae: Aeshnidae spp., pirate perch: Aphredoderus sayanus, and sunfish: L. gulosus, and L. punctatus) are reported.
Data are presented as mean (SD)
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
601
Fig. 2 Proportion of H.
formosa females, males and
juveniles eaten (mean ? 1 SE)
by five different predators (n= 6
for Anax junius, Aphredoderus
sayanus, L. punctatus, L.
gulosus; n = 4 for C. ingens) in
wading pools with low density
of Myriophyllum as cover.
Missing error bars indicate zero
0.7
* Females
A Males
0.6
a)
CZ 05
*Juveniles
-
cu
00.4
variance
0.3
-
o
c 0.2
0
0
O 0.1
0
-0.1 A. junius
C. ingens
A. sayanus
L. punctatus
L. gulosus
Predator
even with themost conservative Bonferroni correction.
All prey classes experienced the greatest predation rates
from L. gulosus; mortality was the lowest from Aph
redoderussayanus for males and juveniles, and from L.
punctatus for females (Fig. 2). The predation pressure
index, which incorporates both predation rates of each
predator and density of the predator at a location, also
indicates that L. gulosus generate a comparatively larger
predation pressure on H. formosa than any other pred
ator, especially during thewinter months (Table 2).
Female H. formosa mortality was less frequent than
expected by chance from C. ingens, Aphredoderus
sayanus and L. punctatus and in proportions similar to
expected from Anax junius and L. gulosus (Fig. 3). This
led to a significant difference among predators in selec
tivity for female H. formosa (X24= 11.55, P<0.02).
H. formosa juveniles appeared to be selectively preyed
upon by Aphredoderus sayanus and L. punctatus (Fig. 3),
but variance was high and selectivity for juveniles did
not differ among predators (x24= 4.27, P < 0.37).
Table 2 Predation pressure index [average predator density (per
mi2) x predation rate (number of prey consumed per 48-h trial)] for
each location in each season
Locality
Aeshnid Aphredoderus
sayanus
Winter (September-March)
TP
2.28
0
WR
0.50
0.34
Summer (April-September)
TP
1.22
0
WR
0.50
0.64
L. gulosus L. punctatus Total
7.42
0.19
0
0.14
9.70
1.17
1.96
0.09
0
0.09
3.18
1.32
Values for Aeshnid are calculated using the mean value of the
predation rate estimated for Anax and Coryphaeschna. The "total"
value given in the last column is the sum of the four predator effects
and gives an indication of overall predation pressure for that
location
Fish predator and vegetation effects
Habitat complexity modifies realized predation rates in
nature. Thus, to improve our ability to assess the
mechanisms underlying the differences in population
dynamics between populations, we experimentally ex
plored the potential interaction among vegetation type,
vegetation amount, and top (most abundant and great
est potential for impact) fish predator on predation
rates. H. formosa mortality was greater with L. gulosus
than L. punctatus at low vegetation, whether it was
Myriophyllum (X21= 53.80, P < 0.0001), or Hydrilla
(X2= 43.79, P < 0.0001). Prey capture rates were also
higher for L. gulosus than L. punctatus with high Hyd
rilla (X21= 10.79, P<0.001). With high Myriophyllum,
however, differences in capture rates were reduced
(X2I=3.51, P<0.061) (Fig. 4).
Analysis of vegetation effects within each predator
type reveals that predation by L. gulosus was signifi
cantly reduced in high vegetation treatments regardless
of vegetation type (Hydrilla: X21 = 30.99, P < 0.001;
Myriophyllum: X I=43.81, P<0.001). In L. punctatus,
increased vegetation led to reduced predation with
Hydrilla (X21= 4.69, P < 0.030) but not with Myrio
phyllum (X21=0.86, P<0.35).
While these results suggest an interaction between
vegetation and predator, the effect was not statistically
significant (Breslow-Day test for Homogeneity of odds
ratio, X23 = 5.6, P < 0.133). However, visual inspection
of the data shows that high amounts of Myriophyllum
reduced L. gulosus predation to levels similar to those of
L. punctatus (Fig. 4).
Predator species and vegetation effects were similar
for male, female, and juvenile classes of prey fish
(Fig. 4). Analysis of H. formosa size revealed that
L. gulosus preyed selectively on larger males (mean
size ? SE, pre-predation = 15.52 ? 0.38 mm, post-pre
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
602
Fig. 3 Selectivity (difference
between observed proportion of
prey eaten and offered) for
females, males and juveniles of
H. formosa (mean ? 1 SE), by
each predator (Anax junius, C.
ingens, Aphredoderus sayanus,
L. punctatus, and L. gulosus).
All trials were done inwading
pools with low density of
Myriophyllum as cover. Missing
error bars indicate zero variance
0.8
* Females
C)
0
0.6
.?
0
.
a
-
0
-
0.2
CD
0 Males
O Juveniles
0
0
0
0-0.2
A.s
Tuis
C nes
A aau
A. junius
C. ingens
A. sayanus
uou
.pnttsL
C-,
a)
-0.6
L. punctatus
L. gulosus
With
dation = 14.72?0.35;
t11=3.48, P<0.006).
L. gulosus as predators, neither juvenile nor female
H. formosa mean size changed with predation (females:
t11==0.49, P>0.63,
juveniles: t11==0.73, P>0.48).
L. punctatus predation had no effect on H.formosa mean
females: tg=0.77,
size (males: tg=0.10, P>0.92,
changes in density and changes in percent cover is
consistent with the importance of vegetation and pre
dation pressure to the dynamics at TP. In the WR
population, with abundant vegetation and fewer, less
efficient predators, time series dynamics could be ex
plained solely through an autocorrelation of changes in
H. formosa density, consistent with this population
P > 0.46; juveniles:
tg= 0.87, P > 0.41).
Based on these data showing that predation ratewas being more affected by within species regulation than
correlated with density of cover, we used our field data by predators or vegetative cover.
A weaker correlation between changes inH. formosa
on vegetative cover to look for patterns in change of
vegetative cover with change inH. formosa density, as a density and changes in vegetative cover atWR is con
method of assessing the effect of changing predation rate sistentwith higher levels of vegetative cover, slower rates
on H. formosa density. Time series analysis of the cor of change in cover, lower densities of predators, and the
relation between change in vegetative cover and change less effective predation of L. punctatus when compared
in H. formosa density revealed differences between to L. gulosus. These factors may combine to keep pre
locations
(Fig.
5). At WR,
a time
series
analysis
that
dation
rates at WR
at negligible
levels and render other
controlled for the autocorrelation of changes inH. for
mosa density, found no significant cross-correlation be
tween change in vegetative cover and change in fish
forces, especially intraspecific density, more important in
affecting H. formosa dynamics. Microsatellite data
indicate that heterozygosity at microsatellite genes is
in H.
much higher inWR than inTP, providing evidence that
density
(r = 0.23, P> 0.25). In contrast,
changes
formosa density were cross-correlated with current the differences in density we observed betweenWR and
TP have persisted for many years (Soucy and Travis
cover (r = 0.39, P < 0.05). (This re
change in vegetative
sult remained significant with correction for a margin
2003). The importance of density dependence to popu
ally significant autocorrelation in vegetative cover lation dynamics inWR is also supported by population
differences in density-dependent changes in reproductive
changes incorporated into themodel.)
traits. Leips and Travis (1999) found that both the
number of broods and the size of broods (each corrected
Discussion
for maternal size) were negatively correlated with den
sity inWR,
but not
in TP. These
data
combined
imply
Our results are consistent with the hypothesis that that density is a prominent selective force forWR H.
predation
pressure, mediated
through habitat formosa, while it plays a lesser, or even negligible, role in
complexity, is a driving force of H. formosa population TP H. formosa.
dynamics. The TP population had lower H. formosa
While the potential importance of habitat complexity
densities, less vegetative cover, and higher densities of a is well recognized, empirical data on the influence of
more efficient predator
than that of WR. The lack of a habitat complexity on predation rates are still lacking
significant autocorrelation for changes in H. formosa (Almany 2004). Our finding that vegetative cover affects
densities and a significant cross-correlation between prey survivorship during piscivore predation are quali
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
603
c
0o5cm
L. gulosus
a-L. punctatus All combined
a
0.5
a
Wacissa
-1.5
*
Q_ 6
+
00.
-1
0.3
-
?
+
0.1
_
0~~~~~
0.
Ie
-
C
0
-1.5
r= 0.23,
CD*
Females
{
b
Eg
-2.5
CZ
0.1
_
C
1
0)
a) 0.3
u)
-
P >0.25
.
.
-1
?
River
0
-0.5
1
0.5
Trout Pond
b
1.5
0)
6
i
c~~~~~~~~~~
oC
a)0 -
Males
C
cu~~~~~ 0
0~~~~
*t 0.5 -
-1.5
0o
X
-.5
1.5
i
0.3
r= 0.39,
.C
-2.5
-3
{
0.1
{
i
High
Hydrilla
Low
-1
0
1
2
3
in cover (i toi+1)
Fig. 5 Correlations between the monthly change in the logarithm
of average percent cover and themonthly change in the logarithm
of average density of H. formosa (number per mi2) at aWR and b
TP. For WR, a significant autocorrelation at lag + 1 was first
removed from the data. Note that x-axis scales differ
l
l
Low
-2
Change
Juveniles
d
P< 0.05
I
High
Myriophyllum
Fig. 4 Proportion of H. formosa (a all combined, b females, c
males, and d juveniles) eaten by L. gulosus and L. punctatus
predators during 48 h wading pool trialswith low or high amounts
of Hydrilla or Myriophyllum. The proportion of fish eaten is out of
18 for all stages combined and out of 8, 5, and 5 for females, males
and juveniles, respectively. Means and SE are given; missing SE
bars indicates zero variance for that treatment combination
While the two Lepomis predators that we used are
congeners and did not differ in size, differences in their
predation ratesmay be due to any number of differences
between the predators. In particular, we suspect differ
ences in predation rates reflect differences in gape size
between the two species. Diet content analysis confirms
that wild L. gulosus include H. formosa in their diet
(M. Aresco and J. Travis, unpublished data). For our
purposes of considering predation pressure as a mech
anism underlying differences in population dynamics,
on H. formosa popu
the potential
impact of predators
lations is a key, not any reciprocal effect on the predator
It was not the goal of this study to under
population.
tatively similar to those of other studies (Dionne and
Folt 1991; Jordan 2002, but see Flynn and Ritz 1999). stand inter-trophic, predator-prey population dynamics.
Studies of coral reef fishes have found an interaction
Theory and empirical work in population dynamics
between habitat complexity and predators on recruit has focused on consumer-resource systems (Persson
ment rates (Beukers and Jones 1997;Almany 2004). In et al. 1998; Post et al. 1999; De Roos and Persson 2002).
our study, each sunfish species had predation rates de Only recently has the role of predators on the popula
creased to a greater extent with increases in the density tion dynamics of a generalist consumer been considered
of its co-occurring vegetation relative to the same (De Roos and Persson 2002). H. formosa is a generalist
increase in the density of the dominant plant species in omnivore and thus on this basis these models seem rel
that
the second population. This effect is especially clear for evant. Size-structured population models
L. gulosus from TP, where a high density of Myrio
incorporate population feedback on individual life-his
phyllum from TP reduced their predation rate to a level tory traits reveal that density-dependent effects on
in the same treatment
similar to that of L. punctatus
maturation time can substantially affect predicted pop
the
existence in nature of ulation dynamics (reviewed in De Roos et al. 2003a).
result
suggests
(Fig. 4). This
subtle effects of vegetation type on predation rates that However, two key assumptions of thesemodels do not
extend beyond simple density of cover.
accord with our knowledge of H. formosa phenology.
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
604
Firstly, the models are based on populations that are
likely to be highly regulated by juvenile growth retar
dation,
such as the one that occurs
in fish or amphibians
predator. Under this scenario, L. punctatus, Aphredode
rus sayanus, and Coryphaeschna leftwith H. formosa for
a longer timewould have shown no selectivity.
Although the most efficient predator, L. gulosus,
preyed on H. formosa sexes in proportion to their
abundance, consideration of size selectivity within males
shows that L. gulosus selected for larger males. L.
gulosus was also the only predator tested that showed a
(nonsignificant) trend toward selectivity for females,
which are larger thanmales. This may reflect an attempt
tomaximize energy return per unit effort through opti
that produce very large numbers of eggs (De Roos et al.
2003b). H. formosa is viviparous and exhibits a high
degree of superfetation, resulting in extended clutch
overlap and the continuous release of small numbers of
offspring (Travis et al. 1987). In collections every
3 weeks fromMarch to November atWR, all females
collected were gravid (Travis et al. 1987). This continu
ous reproduction over several months (females have
of greater
broods of three to five offspring every 1-2 weeks over a mal foraging or it may be a consequence
period of 8-12 weeks) leads to a population structure activity levels in larger individual prey thatmake them
that lacks any cohort effects, as indicated by synchro more likely to be detected and eaten. A recent model
nicity in juvenile and adult abundances (data not suggests that size-selective predation on the largest prey
can also lead to the type of positive
feedback on prey
shown).
The second way in which H. formosa dynamics are density seen with size-selective predation on small prey,
but thismodel again assumes no regulation of the adult
to the models
contrary
of De Roos et al. (2003a) is that
density-dependent effects appear to occur most strongly stage (DeRoos et al. 2003b). The potential impact of
size of offspring
additional predators that consume less prey, but select
at the adult stage in H. formosa. While
is unaffected by food level, interbrood interval, the for smaller prey (such as our data suggest may occur in
number of broods and thus the total number of offspring theH. formosa system) is unknown.
Results of this study demonstrate the potential for
increasewith increasing food availability to H. formosa
females (Travis et al. 1987). Mesocosm studies of H. differences in predation rate to explain differences in
formosa found that all components of female repro population density in nature, but more general inter
duction decreased with increasing density (Leips et al. pretations require caution for a couple of reasons.
2000). Juvenile recruitment also decreases with increas Firstly, we have no replication at the population level. It
ing density (Leips et al. 2000), which may indicate some is not feasible to do this scope of detailed work on many
density-dependent regulation at the juvenile stage. populations. However, ongoing work is being done to
However,
it is also
true that size at maturity
is pheno
typically plastic in H. formosa, mean and minimum
adult sizes decrease with increasing density (Leips et al.
2000), and that time to maturity varies little between
populations (J. Travis and R. Allman, unpublished
data). Thus, H. formosa populations are not strictly size
structured
as modeled
by De Roos
et al. (2003a).
In fact,
collect
the same field census
data on a larger number
of
field sites twice yearly (J. Travis, unpublished data).
Field census data on several low and high-density H.
formosa sites will indicate whether the patterns of cor
relation between H. formosa density, predator presence,
and
seen
vegetation
in our
two populations
can
be
generalized to other populations. Secondly, the two
the nearly immediate response in inter-brood interval
populations
that females make with changes in resource (Travis et al.
to mature
at
1987), as well as the ability of females
etative cover and predators present. Leips and Travis
(1999) had replicate river and lake populations in their
detailed study of H. formosa population dynamics and
smaller sizes are expected to stabilize population
dynamics in these systems (Leips and Travis 1999).
Theoretical work also indicates that differences in
predator selectivity of different prey stages or sizes crit
ically affect prey population dynamics.We found that the
most effective predators (L. gulosus and Anax) preyed
upon H. formosa individuals in proportion to their
abundance. However, the bias observed in other preda
tors toward predation on smaller juveniles (Aphredoderus
sayanus, L. punctatus) or males (Coryphaeschna) could
lead to very different population dynamics. Theoretical
models to date (Persson and Eklov 1995; Persson et al.
1998; Post et al. 1999;De Roos et al. 2003a, b) consider
only a single generalist predator; the role of additional
predators with size-selective but lesser predation rates is
unclear. It is also possible that selectivity observed in this
study simply reflects the biases of predators to catch
easier (smaller) prey prior to preying on larger prey. If
true, a more efficient predator would equalize capture
rateswith available prey types sooner than a less efficient
showed
as well
we used differ abiotically
that at
least
some
aspects
of
the
as in veg
life-history
differences observed between populations could not be
attributed
to river versus
lake habitat.
Field
data
col
lected on several spring-fed river and lake siteswill allow
us to better assess whether some of the population dy
namic effects we observe are attributable to abiotic
factors (J. Travis, unpublished data). Even if abiotic
factors prove important, wild L. gulosus frequently prey
on H. formosa (M. Aresco, unpublished data), and thus
clearly influenceH. formosa populations in nature.
Heterandria
formosa
populations
provide
a rare and
valuable opportunity to further our understanding of
the mechanistic basis for differences in population
dynamics among populations of the same species. There
is evidence for both genetically based life-history differ
ences in this system (Leips and Travis 1999) and for
persistent differences in population densities (Soucy and
Travis 2003). Here, we demonstrate that not only are H.
formosa population densities consistently different in
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
605
two populations, but that the population dynamics also
differ.More importantly, we show that the surrounding
community (e.g., vegetation and predators) has the po
tential to drive the population dynamics of a species. If
this pattern proves to be applicable more generally, it
elucidates a mechanism for among-population differ
ences in density-dependent effects that is likely to be
widespread. Given the ability of variation in levels of
population regulation to generate intraspecific differ
ences in life history (Charlesworth 1994), differences in
population regulation that are generated through dif
ferences in the surrounding community have the po
tential to be a major route by which population
divergence occurs.
Dionne M, Folt CL (1991) An experimental analysis of macrophyte
growth forms as fish foraging habitat. Can J Fish Aquat Sei
48:123-131
Ellner S, Turchin P (1995) Chaos in a noisy world-new methods
evidence
and
atory
on
style
time-series
Am
analysis.
Nat
145:343-375
(1999) Effect of habitat complexity and pred
success
the capture
of fish
on aggregated
feeding
prey. JMar Biol Assoc UK 79:487^*94
Gunzburger MS, Travis J (2004) Evaluating pr?dation pressure on
green
across
larvae
treefrog
a
habitat
gradient.
Oecologia
140:422-429
Hassell MP, Lawton JH, May RM (1976) Patterns of dynamical
in single-species
behaviour
J Anim
populations.
Ecol
45:471-486
Jordon F (2002) Field and laboratory evaluation of habitat use by
rainwater
killifish
in the St.
(Lucania parva)
Estuaries
25:291-298
ary, Florida.
Johns
River
Estu
Kendall BE, Prendergast J, Bjornstad ON (1998) The macroecol
of
ogy
taxonomic
population
dynamics:
in population
cycles.
patterns
Acknowledgements This work was supported by National Science
Foundation award DEB 99-03925 to JT.M. Richards and several
anonymous reviewers provided helpful feedback on previous ver
sions of themanuscript. We thankM. Aresco, R. Fuller, L. Horth,
and B. Shoplock for help with the field censuses and M. Aresco for
help with the predation experiments. We are grateful toK. Graffius
Ashcraft for facilities support for the predation experiments. This
work was performed under Florida State University's ACUC Pro
tocol 9321 and complies with the current laws of the USA.
from
Flynn AJ, Ritz DA
Ecol
Lett
and
biogeographic
1:160-164
Law R, Bradshaw A, Putwain P (1977) Life-history variation in
annua.
Poa
31:233-246
Evolution
Leips J, Travis J (1999) The comparative expression of life-history
traits
and
to
its relationship
the numerical
of
dynamics
four
populations of least killifish. J Anim Ecol 68:595-616
Leips J, Travis J, Rodd FH (2000) Genetic influences on experi
mental population dynamics of the least killifish. Ecol Monogr
70:289-309
Mueller LD (1997) Theoretical and empirical examination of
selection.
density-dependent
Annu
Rev
Ecol
28:269-288
Syst
Ohgushi T (1995) Adaptive behavior produces population stability
in herbivorous lady beetle populations. In:Cappuccino N, Price
References
Almany GR
(2004) Does
and
pr?dation
increased habitat complexity
in coral
competition
reef fish
assemblages?
reduce
Oikos
106:275-284
Baer
CF
II. Gene
formosa.
flow
US
and
fresh
biogeogra
phy within the St. Johns River drainage. Heredity 81:404?411
Beukers JS, Jones GP (1997) Habitat complexity modifies the im
pact
of
on
piscivores
a coral
reef
fish
population.
Oecologia
114:50-59
Box GEP, Jenkins GM, Reinsei GC (1994) Time series analysis,
and
forecasting
3rd
control,
edn.
Prentice-Hall,
Englewood
Cliffs
Bradshaw W, Holzapfel C (1989) Life-historical consequences of
Wyeomia smithii. Am Nat
B
Charlesworth
(1994)
University
Cambridge
in
selection
density-dependent
the
pitcher
plant
mosquito,
133:869-887
Evolution
Press,
populations.
Cambridge
Chatfield C (1996) The analysis of time series, an introduction.
Chapman
and Hall/CRC
Boca
Press,
Raton
Cooper G (2001) Must there be a balance of nature? Biol Philos
16:481-506
De Roos AM, Persson L (2002) Size-dependent life-history traits
promote
catastrophic
of
collapses
Acad Sei 99:12907-12912
De Roos AM, Persson L, McCauley
size-dependent
of populations
the structure
Ecol
De Roos AM, Persson L, Thieme HR
effects
in top predators
feeding
on
Proc R Soc Lond B 270:611-618
Proc
predators.
Nati
E (2003a) The influence of
on
traits
life-history
and communities.
top
Lett
and dynamics
6:473-487
(2003b) Emergent Allee
structured
prey
populations.
models.
piscivorous
and
perch
juvenile
perch
and
roach.
Ecology
76:70-81
Persson
Leonardsson
L,
K,
DeRoos
tensen B (1998) Ontogenetic
of a size-structured
dynamics
AM,
in structured
processes
whole-lake
experiments.
Primack RB, Antonovics
Gyllenberg
M,
Chris
scaling of foraging rates and the
consumer-resource
Popul Biol 54:270-293
Post JR, Parkinson EA, Johnston NT
fish populations:
Ecol Monogr
model.
Theor
(1999) Density-dependent
interaction
in
strengths
69:155-175
J (1982) Experimental ecological genetics
in plantago.
7. Reproductive
effort
L. Evolution
lanceolata
36:742-752
in populations
of Plantago
(1995) Impact of life-history evolution on population
Rossiter M
in age-structured
synthesis.
Wiley, New York
Persson L, Eklov P (1995) Prey refuges affecting interactions be
tween
in a south-eastern
structure
(1998) Population
fish, Heterandria
water
new approaches
P (eds) Population
and
dynamics:
San Diego,
Academic,
pp 303-320
A
with
Pankratz
(1991) Forecasting
dynamic
regression
dynamics:
Cappuccino
proaches
the presence
of maternal
effects.
predicting
new
Price P (eds) Population
N,
dynamics:
and synthesis.
San Diego,
Academic,
pp 251-275
In:
ap
Sale PF, Tolimieri N (2000) Density dependence at some time and
place? Oecologia 124:166-171
Sinclair ARE (1989) Population regulation in animals. In: Cherret
JM
(ed) Ecological
of
understanding
concepts,
the natural
the contribution
world.
In: 29th
of ecology
symposium
to an
of
the
British Ecological Society. Blackwell, Oxford, pp 197-241
Soucy S, Travis J (2003)Multiple paternity and population genetic
structure
in natural
populations
of
the poeciliid
fish, Heteran
driaformosa. J Evol Biol 16:1328-1336
Travis J, Farr JA, Henrich S, Cheong RT (1987) Testing theories of
clutch
formosa.
overlap
Ecology
the reproductive
68:611-623
with
This content downloaded from 206.224.223.242 on Fri, 30 Oct 2015 16:42:53 UTC
All use subject to JSTOR Terms and Conditions
ecology
of Heterandria