Life history and demographic determinants of effective/census size ratios as exemplified by brown trout (Salmo trutta).

0.22 MB·English

by  Serbezov, Dimitar

Evolutionary Applications
EvolutionaryApplicationsISSN1752-4571
ORIGINAL ARTICLE
Life history and demographic determinants of
effective/census size ratios as exemplified by brown
trout (Salmo trutta)
Dimitar Serbezov,1 Per Erik Jorde,1,2 Louis Bernatchez,3 Esben Moland Olsen1,2 and Leif Asbjørn
Vøllestad1
1CentreforEcologicalandEvolutionarySynthesis(CEES),DepartmentofBiology,UniversityofOslo,Blindern,Oslo,Norway
2InstituteofMarineResearch,Flødevigen,Norway
3InstitutdeBiologieInte´grativeetdesSyste`mes(IBIS),PavillonCharles-Euge`neMarchand,Universite´ Laval,Que´bec,QC,Canada
Keywords Abstract
animalmating/breedingsystems,conservation
genetics,fisheriesmanagement,lifehistory A number of demographic factors, many of which related to human-driven
evolution. encroachments, are predicted to decrease the effective population size (Ne) rel-
ative to the census population size (N), but these have been little investigated.
Correspondence Yet, it is necessary to know which factors most strongly impact N, and how to
e
DimitarSerbezov,CentreforEcologicaland mitigate these effects through sound management actions. In this study, we use
EvolutionarySynthesis(CEES),Departmentof
parentage analysis of a stream-living brown trout (Salmo trutta) population to
Biology,UniversityofOslo,POBox1066,
quantify the effect of between-individual variance in reproductive success on
Blindern,0316Oslo,Norway.
Tel.:[phone]; the effective number of breeders (Nb) relative to the census number of breeders
fax:[phone]; (Ni). Comprehensive estimates of the Nb/N ratio were reduced to 0.16–0.28,
e-mail:[email protected] almost entirely due to larger than binomial variance in family size. We used
computer simulations, based on empirical estimates of age-specific survival and
Received:17November2011 fecundity rates, to assess the effect of repeat spawning (iteroparity) on N and
e
Accepted:19December2011
found that the variance in lifetime reproductive success was substantially higher
Firstpublishedonline:23January2012
for repeat spawners. Random family-specific survival, on the other hand, acts
doi:10.1111/j.1752-4571.2012.00239.x to buffer these effects. We discuss the implications of these findings for the
management of small populations, where maintaining high and stable levels of
N is crucial to extenuate inbreeding and protect genetic variability.
e
anthropogenic, and its maintenance is therefore a funda-
Introduction
mental concern in conservation biology. To be able to
The effective population size (N) is arguably one of the properly manage wild resources, genetic and evolutionary
e
most important parameters in evolutionary and conserva- considerations have to be integrated into management
tion biology as it determines the rate of genetic drift, the practices (Young 2004; Waples and Naish 2009).
rate of loss of diversity of neutral alleles and the rate of Because the magnitude of drift is inversely proportional
increase of inbreeding experienced by an isolated popula- to the effective size (e.g., Crow and Kimura 1970), small
tion (Charlesworth and Willis 2009). This is because it natural populations or populations experiencing fragmen-
establishes the relative evolutionary importance of sto- tation and loss of connectivity, often the focus of conser-
chastic (random genetic drift) and directional (migration vation programmes, are vulnerable to the random loss of
and selection) factors and is thus closely being linked to genetic variation. Knowledge of the effective size is there-
genetic diversity (Crow and Kimura 1970). Genetic varia- fore particularly important in such cases (Lande and Bar-
tion is the raw material for evolutionary change, allowing rowclough 1987; Nunney and Elam 1994). Anthropogenic
populations to evolve as a response to various environ- activities have recently resulted in many wild populations
mental challenges (Frankham 1996), be they natural or being seriously reduced in size, leading to significant
ª2012BlackwellPublishingLtd5(2012)607–618 607
Lifehistoryanddemographicdeterminantsofeffective/censussizeratios Serbezovetal.
reductions in N (Frankham et al. 2010). Supportive that female and male fecundity is age-independent (Nun-
e
breeding programmes may also be disadvantageous from ney 1993). Because N is based on lifetime reproductive
e
a conservation genetics perspective if the individuals success, such estimates are much more difficult to obtain
released into the wild originate from captive populations for wild populations of iteroparous organisms. Moreover,
with small N values (Ryman and Laikre 1991). Also, the to estimate generational N from demographic data, par-
e e
efficiency of natural selection is directly proportional to ents and their offspring should be counted at the same
N because genetic drift is weaker when N is large, and life stage. Family-correlated survival is expected to reduce
e e
this will determine the fate of natural populations N further compared to random survival (Crow and Mor-
e
(Charlesworth 2009). ton 1955; Hedgecock 1994). Little is currently known
Despite its fundamental importance in conservation about the relationships between life history plasticity,
genetics, it has been notoriously difficult to obtain reliable demographic change and N (Lande and Barrowclough
e
N estimates in the wild. Accurate ecological N estima- 1987; Nunney 1993), partly due to scarcity of the ecologi-
e e
tion requires quantification of a number of demographic cal data required for these analyses.
parameters that are difficult to obtain from natural popu- Salmonid fishes exhibit a wide diversity of breeding
lations in most species. Examples of these include systems (Fleming 1998; Fleming and Reynolds 2004), and
accounting for spatial substructuring and overlapping even different populations of the same species living
generations. In addition, a variety of factors related to its under different environmental conditions might have
interpretation complicate its application in wild popula- evolved different life history tactics and can thereby exhi-
tions. The ratio between the effective population size to bit different breeding behaviours (Gross 1991). Both sexes
the number of sexually mature adults in a population in salmonids show high variance in reproductive success
(census size), the N/N ratio, is of great potential interest. (Fleming 1998; Garant et al. 2001; Serbezov et al. 2010).
e
Different groups of organisms, however, seem to be char- However, reduced variance in individual reproductive
acterized by different N/N ratios, leading to confusion success at low breeder abundance (genetic compensation)
e
regarding the magnitude and the meaning of this parame- can lead to an increase in N and might be a realistic
e
ter (Nunney 1993; Frankham 1995). Also, if there is a aspect of salmonid breeding systems (Ardren and Kapu-
relationship between N and N, variable population size scinski 2003; Frasier et al. 2007). Understanding the rela-
e
can have a profound influence on long-term N because tionship between N and the mating system is therefore
e e
effective size for a series of generations is approximately important in allowing comparisons between populations
the harmonic mean of the single generation N (Wright in conservation contexts (Nunney 1993). Proper manage-
e
1938). This effect would be particularly relevant in popu- ment of such populations should account for those fac-
lations of conservational concern, where an apparently tors that may potentially strongly impact (i.e., reduce)
healthy census population may mask small N (Nunney the N and thus genetic variance. Indeed, in addition to
e e
and Elam 1994; Frankham 1995). climate change, salmonid fishes are exposed to a large
The most important factors responsible for reducing N number of human encroachments such as harvesting and
e
below the number of sexually mature adults in the popu- fragmentation leading to reduced connectivity. Many of
lation is a larger than binomial variance in individual these encroachments influence the breeding system, levels
reproductive success, unequal sex ratio and temporal fluc- of gene flow, population structure and overall fitness,
tuations in population size (Wright 1938). Polygamous thus impacting N and the overall level of genetic varia-
e
mating systems would result in high variance in repro- tion. To manage such populations, it is necessary to
ductive success that in turn leads to low effective popula- know which factors most strongly impact N and how to
e
tion size (Nunney 1993; Ardren and Kapuscinski 2003). mitigate these impacts through management actions. In
In real populations, the variance of progeny number per this study, we use extensive demographic and genetic
family generally exceeds the mean progeny number in data on a brown trout (Salmo trutta) population to
both sexes (Crow and Kimura 1970), so that the effective quantify the effects of variation in reproductive biology
number of individuals of either sex is less than the actual on N and N. We use empirical estimates of age-specific
b e
number present in any generation. However, as genera- survival (Carlson et al. 2008) and age- and sex-specific
tions overlap, there is no easy way of quantifying this fertility (Serbezov et al. 2010) in an individual-based sim-
effect on the per-generation N (Waples 2010). Estimates ulation model to estimate the variance in lifetime repro-
e
of reproductive success are commonly obtained using sea- ductive success and obtain a generational N estimate.
e
sonal measures, but these primarily estimate the effective Having this information, it is possible to evaluate what
number of breeders per year (N ) rather than the effective kind of population structure best maintains high and sta-
b
population size per generation (N). The N values are ble levels of N and thus help make informed manage-
e b e
however only useful in estimating N if it can be assumed ment decisions.
e
608 ª2012BlackwellPublishingLtd5(2012)607–618
Serbezovetal. Lifehistoryanddemographicdeterminantsofeffective/censussizeratios
types, combined with compatible length measurements,
Materials and methods
helped us confirm that an individual had been captured
Background before. Specifically, we found that age was underestimated
The study site is a small forest stream, Bellbekken, in by 1 year in 35% and by 2 years in 18% of individuals
South-East Norway (N: 61(cid:2)15¢, E: 11(cid:2)51¢) that was sam- that were 3 years or older. This uncertainty was incorpo-
pled extensively during the period 2002–2007 (see Olsen rated when calculating generation times and modelling
and Vøllestad 2003; Carlson et al. 2008; Serbezov et al. maturation probabilities, as described below. Ageing of
2010 for detailed information). Compared to nearby riv- juveniles was based on length and is thus not affected by
ers, Bellbekken has limiting growth conditions for fish errors in scale readings.
and is seldom, if ever, visited by anglers. The sampled
section starts at a small waterfall, which should prevent
Statistical analyses
upstream migration under most conditions, and there is a
weak but significant genetic differentiation between trout We calculated the mean (k) and variances (Vk) of off-
in the stream and the larger river below the waterfall spring numbers per male and per female parent for each
(Taugbøl 2008). spawning season using data presented in Fig. 1. Because
The analyses described herein utilize empirical data offspring were observed and counted when juvenile (age
described by Serbezov et al. (2010). Briefly, these were 0+ to 2+), and not at the same age as their parents,
based on a total of 451 (195 males and 256 females) effects of mortality until reproductive age are not fully
mature brown trout that were sampled (nondestructively) accounted for in these estimates of family size and their
from Bellbekken during the 2002, 2003 and 2004 spawn- variances. To investigate the effect of family-specific mor-
ing seasons, and 1856 juvenile trout sampled the follow- tality, we used the approach of Crow and Morton (1955)
ing years, aged 0+ (i.e., during their first year of life) to and adjusted k at different stages to their expected value
2+ years and representing birth years or cohorts 2003 of k = 2 (for a stable population) and recalculated Vk.
(834 individuals), 2004 (707) and 2005 (315). All fish We calculated the mean (k) and variance (Vk) of the
were genotyped at 15 microsatellite loci. The genotypes number of progeny for each sex for 1-year-old offspring
were used to individually assign (with 90% confidence or and for 3-year-old offspring separately. The age-specific
higher) offspring individuals to potential parents, using ratios of variance to mean family size R1 (= Vk1/k1) and
the MasterBayes software (Hadfield et al. 2006). Alto- R3 (= Vk3/k3) values were then scaled to their expected
gether 935 of the 1856 sampled offspring were assigned a values at k = 2 by assuming random survival from the
father and 723 were assigned a mother (Serbezov et al. last measured stage (Crow and Morton 1955; Waples
2010; table 3). Of the 451 potential parents, 78 males (of 2002), and these are denoted as R*1 and R*3. This proce-
195) and 116 females (of 256) were assigned one or more dure yields an indication of how much family-size vari-
offspring to them, whereas 117 males and 140 females ance through a particular life stage has contributed to
were not assigned with any of the sampled offspring. We the reduction of the effective population size (using
combined parentage assignment and sib-ship reconstruc- Ne/N = k/(1 + R*age) (Waples 2002).
tion to improve our pedigree inferences as follows. The Calculation of Ne per generation requires an estimate
software COLONY v2.0 (Wang 2004; Jones and Wang of the variance in lifetime reproductive success (Hill
2010) was used to partition the offspring cohort into full 1972, 1979). As we only have estimates of reproductive
and half-sib families and to infer their parental genotypes. success for single seasons, we used an individual-based
For the 2002 breeding season, a few large offspring half- simulation model for a population of 98 males and 108
sib families had no matching parents among the sampled females (the estimated average number of spawners in a
adults (Serbezov et al. 2010). These were assigned to gen- spawning season for the 2002–2007 seasons, see below).
erated parental genotypes and included in the calculations We estimated maturation probability, m(a), at a given
of the family-size variance as paternal half-sib families as age, a, from the observed proportions, o(a), at that age
they represent a large proportion of sampled offspring (Barot et al. 2004):
(Serbezov et al. 2010).
oðaÞ(cid:2)oða(cid:2)1Þ
Bellbekken has poor growth conditions for brown trout mðaÞ¼ ð1Þ
1(cid:2)oða(cid:2)1Þ
and age determination based on scale reading is therefore
uncertain. Genotype matching was used to quantify the Uncertainties in observed age readings were included in
error in age determination for adult fishes. Fish some- the model by randomly sampling without replacement
times loose their Passive Integrated Transponder (PIT)- fish from the distribution of observed individuals and
tags (Prentice & Flagg 1990), some of these individuals adding 1 or 2 years to the recorded age according to the
were tagged and genotyped again, and identical geno- observed error probabilities. This bootstrap technique was
ª2012BlackwellPublishingLtd5(2012)607–618 609
Lifehistoryanddemographicdeterminantsofeffective/censussizeratios Serbezovetal.
Males Females
Season 2002
5
0.
3
0.
1
0.
Season 2003
5
cy 0.
n
e
equ 0.3
Fr
1
0.
Season 2004
5
0.
3
0.
1
0.
0 5 15 25 35 45 55 65 0 5 15 25 35 45 55 65
Number of offspring
Figure1 Observed distribution of reproductive success (number of progeny assigned) for male and female brown trout in Bellbekken during
threeconsecutivespawningseasons.
performed 1000 times. Because we found only marginal respectively. The effective numbers of breeders were cal-
differences between estimates for males and females, we culated for each season and sex from the estimated mean
pooled data for both sexes (Fig. S1). The individual-based number of offspring (k) and the corresponding variance
simulations were done by following individuals through (V ) (Table 1). For females, this was done using:
k
their reproductive life, from 3 to 8 years of age, as all
N ¼k ðN (cid:2)1Þ
bfðiÞ f fðiÞ
assigned parents were within these ages. Yearly reproduc- ð2Þ
ð1þV =k Þ
tive success for mature fish was modelled by randomly kf f
sampling with replacement from the observed sex- and where N is the estimated number of sexually mature
fi
age-specific distributions of variation in reproductive suc- females in a spawning season i, and k and V are the
f kf
cess (Fig. 1). The number of offspring produced each year mean and the variance of the number of progeny pro-
was then summed to give an individual estimate of life- duced (Crow and Denniston 1988). The equation for
time reproductive success. Survival was modelled using male spawners is analogous. The number of sexually
the age-specific survival rate previously estimated based mature males and females in any given year was estimated
on capture–recapture estimates from our system (Carlson using the results from the three-pass electrofishing results
et al. 2008). This bootstrap technique was also performed and the standard Zippin depletion method (Zippin 1958;
1000 times. Bohlin et al. 1989). To obtain confidence intervals for the
estimated N ’s, we resampled the observed number of
b(i)
spawners, with replacement, from their reproductive suc-
Effective population sizes
cess. This bootstrap procedure was repeated 1000 times,
We estimated effective population sizes per season (N : and the 2.5% and 97% percentiles were taken as the
b
effective number of breeders) and per generation (N) lower and upper 95% confidence limits for N , respec-
e b(i)
from the seasons and lifetime distributions of family sizes, tively.
610 ª2012BlackwellPublishingLtd5(2012)607–618
Serbezovetal. Lifehistoryanddemographicdeterminantsofeffective/censussizeratios
Table1. Estimates ofN for2002–2004inBellbekkenbasedonsexratioandvarianceinthereproductivesuccess(basedonparentageassign-
b
mentanalysis).Shownaretheestimatednumbersofmalesandfemales,N isthenumberofbreederscorrectedforunequalsexratio(aftereqn
b
(1)),k ,V andk,V arethemeanandthevarianceinreproductivesuccessformalesandfemales,respectively,andN andN arethe
m m f f bm(i) bf(i)
respectiveeffectivenumbersofmalesandfemaleseachseason(aftereqns2and3).
Males Females Males+ N N (sexratio)/
b b
Season (N ) (N) females(N) (sexratio) N(est) k V N k V N N (95%CI) N /N
m f m m bm(i) f f bf(i) bi b
2002 94 132 226 220 0.97 4.2 118.2 13.5 2.4 22.4 30.0 37.3(26.2–49.3) 0.16
2003 85 103 188 186 0.99 4.8 94.0 19.6 2.9 18.0 40.7 52.9(41.7–69.2) 0.28
2004 96 118 213 211 0.99 1.4 11.3 14.7 0.8 1.8 25.1 37.0(28.2–45.6) 0.17
Average 91.7 117.7 209 205.7 0.98 3.5 74.5 15.9 2.0 14.1 31.9 39.8 0.20
An estimate of effective number of breeders in a year
Cost of reproduction
was obtained by combining female and male effective
numbers for that year: We investigated the effect of previous reproductive effort
on later reproductive success (i.e., cost of reproduction)
4ðN (cid:3)N Þ
N ¼ bfðiÞ bmðiÞ ð3Þ from estimates of individual reproductive success over the
bðiÞ ðN þN Þ
bfðiÞ bmðiÞ three consecutive spawning seasons. The individual repro-
ductive success in a given season may depend on sex,
We also used the actual number of spawners (N and
m body length, age, identity of the spawner and year
N) in place of effective numbers in this eqn (3) to quan-
f (because of e.g. environmental differences among years).
tify the relative effects on N of unequal sex ratio alone.
e Body length and age are highly correlated, and we used
The per-generation effective size, N, was estimated
e body length as covariate in the analysis because we have
from lifetime variances in family sizes obtained using the
higher confidence in the precision of body length mea-
simulations described above and based on Hill’s (1972,
surements than age (cf. ageing errors above). We use a
1979) approach, which is appropriate for species with
generalized linear mixed effects model implemented in
overlapping generations. We use Hill (1979), eqn 10:
the R package MCMCglmm (Hadfield 2010) with the fol-
lowing structure to model effects on reproductive success:
N ¼8NL=ðr2 þr2þ4Þ ð4Þ
e m f
y¼b Seasonþb Sexþb Lengthþz Spawner
where N is the average estimated number of spawners 1 2 3 1
þz Yearþe ð5Þ
(males plus females) in a season, L is the average genera- 2
tion length for males and females, and r2 and r2 are the
m f where Season (the number of the consecutive season an
lifetime variances in reproductive successes for males and
individual reproduces; from 1 to 3), Sex and fork Length
females, respectively. As presented by Hill (1972, 1979),
are fixed effects, and Spawner identity and Year (2002,
calculations of these variances include separate variance
2003, 2004) are random effects, and e is random residual
and covariance terms for parent–offspring sex. As second-
error. The response variable (y), specified in the model
ary sex characteristics are not developed until the fish
with Poisson distribution, is the estimated number of
approach maturity, offspring were not sexed in our study
progeny produced per individual and year. Significant
and we substituted r2 and r2 with V and V , respec-
m f km kf negative b values would be indicative of incurred costs
1
tively.
of previous reproduction on current reproductive output.
Sensitivity analysis Results
To assess how the N estimates depend on the values of Spawning brown trout found in Bellbekken were between
e
the model parameters, we re-analysed the model with 2 and 8 years of age. There was a tendency for females to
varying parameter values. First, we ran the simulations mature slightly younger than males (mean age 5.8 vs
using the age-maturation curve (Fig. S1) unadjusted for 6.1 years; t = 3.9, p = 0.0006). The generation length
521
ageing errors. Then, we ran the simulations with a range of the two sexes, defined as their average ages at the time
of values for the male and female variance in reproductive the offspring are born (the spring after a spawning event),
success (0.5–10 times the observed value) and age-specific was calculated from estimates obtained from the parent-
survival probabilities (from 40% less to twice the age data. The estimated generation time for males was
observed value). 6.5 years (SD = 1.1 years) and for females 5.9 years
ª2012BlackwellPublishingLtd5(2012)607–618 611
Lifehistoryanddemographicdeterminantsofeffective/censussizeratios Serbezovetal.
(SD = 1.0), for an average of L = 6.2 years (SD = 1.1), in values, but still considerably larger than unity. The com-
good agreements with estimates from other Scandinavian bined effects of all demographic processes, including
stream-resident brown trout populations (L = 6.2 – among-family variation in juvenile survival, led to a fur-
6.6 years, Palm et al. 2003). ther reduction (in most cases) of N relative to those pre-
b
sented in Table 1 and yielding final N /N ratios between
b
0.14 and 0.41 (Table 2). The reduction in N /N between
Effective spawning population size b
the 1+ and 3+ stages was smaller relative to the reduction
The estimated census number of breeders (N) in Bellbek- up to the 1+ stage, ranging from 32.41% further reduc-
ken from 2002 to 2007 was relatively stable, ranging from tion (for females in 2002) to a 9.36% increase (females
66 to 98 males and from 84 to 132 females. Based on 2003). The mean family size at enumeration at the 3+
these estimates, the effect on unequal sex ratios on N for stage (between 0.4 and 1.4; Table 2) is less than what
b
2002, 2003 and 2004 seasons was very small (eqn 3), and would be expected in a population of constant size,
the N /N ratios were negligibly different from k = 2. The Crow and Morton (1955) approach for scaling
b(sex ratio)
unity in all three spawning seasons (Table 1). k and V , however, also applies if k < 2, the result being
k
In contrast, including the variances in family sizes (V ) a scaled variance effective size that is equivalent to the
k
from the parentage assignment analysis yielded (eqs 2 and inbreeding effective size that is calculated with the
3)markedly reduced N estimates,with N /Nratiosinthe unscaled k and V . All in all, the mortality between 1+
b b k
rangefrom0.16to0.28(Table 1,right).Theimpactofvar- and 3+, although variable between sexes and seasons,
iance in reproductive success can be considered separately does seem to be rather random among families, acting to
for each sex and spawning season by comparing effective reduce the negative effect of the strong skew in individual
numbers(N andN )withcensusnumbers(N andN ). reproductive success on effective size and N /N ratios.
bf bm f m b
From such comparisons, it is clear that deviations from
binomial family-size distribution has a dramatic effect on
Repeat spawning
reducingeffectivenumbersforeithersex(cf.Table 1),and
the low N /N ratios were not caused by a single sex only. A third of the spawners in Bellbekken were observed to
b
This was observed in all three seasons and thus appears to be sexually mature during more than one season; a few
beageneralfeatureinthisbrowntroutpopulation. individuals of each sex were even observed to be mature
Effects of among-family variation in survival on esti- during five consecutive seasons (Fig. 2A). There was no
mated lifetime means and variances in family sizes were difference between the sexes in the proportion of iterop-
assessed by the ratio of variance to mean progeny num- arous (repeat spawning) individuals (v2 = 0.68, df = 1,
bers (R: Table 2), calculated separately for age 1 and 3 p = 0.41) or in the levels of realized iteroparity (being
i
progeny. R values were always larger than 1, implying assigned offspring in more than one season) (v2 = 0.94,
1
that family-size variances were larger than the binomial. df = 1, p = 0.33) (Fig. 2B). On average, iteroparous males
This was more pronounced for males which, in addition increased their reproductive success by 52% and females
to family-specific survival in the early juvenile stages, also by 38% relative to semelparous fish. The only parameter
reflect the reproductive skew in this population (see Serb- from the mixed linear effect model (eqn 5) that had sig-
ezov et al. 2010). The adjusted R* values (indicating nificant effect on reproductive success was body length
1
what the index of variability would be if survival was ran- (mean ± SD: 0.09 ± 0.02). There seem to be quite large
dom from that point on to an adult population with among-year variation in the number of offspring
k = 2) were in general smaller than the unadjusted R (mean ± SD: 21.20 ± 58.49) that masks any possible
1
Table2. Demographic estimates used in analysing family-correlated survival. Shown are the estimated census spawner numbers (N), mean (k)
and variance (V) in family sizes for each sex and the corresponding age-specific index of variability (R =V ) and the scaled to their expected
k i k/k
valueatk=2age-specificindexofvariability(R*).Thesevaluesaregivenfor2002and2003andforthe1+and3+ageclasses.Therightmost
columnshowsthepercentagechangebecauseoffamily-sizevariationthathasoccurredintheN /Nratiobetween1+and3+stages.
b
1+oldfish 3+oldfish
%Changebetween
Season N k V R R* N /N=2/1+R* k V R R* N /N=2/1+R* thetwostages
1 1 1 1 b 1 3 3 3 3 b 3
Males 2002 94 3 57.7 19.4 13.4 0.14 1.4 12.3 9.1 13.0 0.14 )2.8
2003 85 3.6 60.5 16.8 9.8 0.19 0.6 2.5 4 10.8 0.17 8.49
Females 2002 132 1.4 8.4 6.2 8.6 0.21 0.6 2.8 4.7 13.2 0.14 32.41
2003 103 2.4 12.1 5 4.4 0.37 0.4 0.6 1.6 3.9 0.41 )9.36
612 ª2012BlackwellPublishingLtd5(2012)607–618
Serbezovetal. Lifehistoryanddemographicdeterminantsofeffective/censussizeratios
(A) tion length) and N = 196 (average estimated number of
0.8 22000023 spawning individuals for the 2002–2007 period) in Hill’s
2004 eqn (4), NeV was estimated to be 104.3 ± 44.9 (SD), and
4 2005 a N /N ratio of 0.53 ± 0.24 (SD).
0. eV
To evaluate the consequences of repeat spawning on
0 effective size in brown trout, we introduced obligate mor-
0. 1 season 2 seasons 3 seasons 4 seasons 5 seasons tality (semelparity) after a maturation event in the model.
This had the effect of reducing the number of offspring
(B)
8 Males per breeder substantially for both sexes (3.8 ± 8.8 vs
0. Females 1.8 ± 5.2 for males, t-test: p < 0.0001, 2.7 ± 4.7 vs
4 1.8 ± 3.7 for females, t-test: p < 0.0001), but also reduced
0. its variance (F-test to compare two mean-standardized
variances: F = 2.8, p < 0.0001 for males;
0 1,13460
0. 1 season 2 seasons 3 seasons F = 1.6, p < 0.0001 for females). This reduction in
1,13480
variance in family size resulted in a N estimate for the
eV
Figure2 (A) Proportions of brown trout individuals that were hypothetical semelparous situation of 272.0 (SE = 112.7),
observed to have been sexually mature during different consecutive
more than twice the estimate with repeat spawning
breeding seasons (1–5 seasons). The four categories (2002–2005)
(Fig. 3). The increased family variance however led to a
denote the first year a spawner was observed to be sexually mature.
increased variance (uncertainty) in the N estimate (F-
(B)Proportionsofassignedspawnersofbothsexesin1–3consecutive eV
seasons. test to compare two mean-standardized variances:
F = 5.7, p < 0.0001).
1,999
effects of previous reproduction on the current offspring
production, and no cost of reproduction was detected. Sensitivity analysis
The N estimates were relatively sensitive to changes in
Per-generation effective size eV
the variance in reproductive success, especially at lower
Using the lifetime variances in estimated family sizes values compared to the obtained values (Fig. 4A), but
(r2 = 82.6 and r2 = 22.7), L = 6.2 (the average genera- very little with respect to the value of the age-specific
m f
Iteroparity Semelparity
mean = 104.3
2 2
0. 0. mean = 269.4
y
c
n
e
u
q
e
Fr
1 1
0. 0.
0 100 200 300 400 0 100 200 300 400 500 600 700
Effective population size (N )
e
Figure3 Distribution ofsimulation-basedestimatesofN fortheiteroparousmodel(A)andthesemelparousmodel(B).Thelifetimevariancein
eV
reproductivesuccessissimulatedbasedonsex-andage-specificfertilitiesfrombrowntroutinBellbekken.
ª2012BlackwellPublishingLtd5(2012)607–618 613
Lifehistoryanddemographicdeterminantsofeffective/censussizeratios Serbezovetal.
0 0
0 0
7 5
0
0
6
0
0
4
0
0
5
0
D 00 D 30
S 4 S
N ± e 300 N ± e00
2
0
0
2
0
0
0 1
0
1
0 0
–2 0 2 4 6 8 10 –0.4 –0.2 0.0 0.2 0.4 0.6 0.8
Delta variance Delta survival
Figure4 Sensitivity of the N estimates to changes (% change from the observed values, vertical dashed line) in reproductive success variance
e
(A)andage-specificsurvivalestimates(B).
survival estimates (Fig. 4B). The N estimates were also dom at older juvenile stages with respect to families.
eV
rather insensitive to the shape of the age-specific proba- Simulation results indicate that iteroparity, on the other
bility of maturation curve (Fig. S1), giving slightly higher hand, increases the variance in reproductive success, lead-
estimate when using the age-determination-error unad- ing to a decreased effective population size. Iteroparity
justed maturation probabilities (128.1 ± 51.2). however also decreased the variance of the effective popu-
lation size, which might lead to a decreased risk of demise
because of stochastic population events.
Discussion
The genetic parentage assignments and the exhaustive
In this study of an exhaustively sampled stream-resident sampling should allow for robust estimates of N . Never-
b
brown trout population, we were able to estimate effec- theless, nearly half of the offspring were not assigned a
tive number of spawners per season and generation and parent, and there was some uncertainty in the ageing of
to quantify several factors that impacted the N /N and mature fish, which might bias the generational N esti-
b e
N/N ratios. This is the first study to our knowledge that mate. We do not expect that the high proportion of
e
uses parentage genetic assignments data to estimate effec- unsampled parents would have significant impact on the
tive population sizes in a natural salmonid population, r2 and r2, as there is no reason to suspect that the un-
m f
yielding separate estimates for per season and per genera- sampled spawners produce a different family-size distri-
tion. Earlier attempts on this species have relied on bution. Also, the N estimate seemed to be insensitive to
e
temporal change in allele frequencies, yielding only per- ageing errors as analyses done without making age adjust-
generation estimates (Jorde and Ryman 1996; Laikre et al. ments (see Materials and methods section) yielded esti-
1998, 2002; Hansen et al. 2002; Palm et al. 2003; Hegg- mates that were only slightly higher. The N estimate also
e
enes et al. 2009). While these approaches to estimate per- seems to be quite insensitive to higher family-size variance
generation N yielded congruent results, our approach has and survival values than the observed ones, but increase
e
the great advantage of also providing estimates of per sea- significantly at lower variance values (Fig. 4). This result
son N and of the factors determining effective sizes. As is expected for the family-size variance, whereas decrease
b
expected, the effect of the variance in reproductive success in survival seems to increase family-size variance and thus
led to the largest reduction in N relative to the census to lower generational N value.
e. e
size (N), whereas unequal sex ratios had only a marginal We arrive at N /N estimates of 0.16–0.28, whereas the
b i
effect. The impact of family-size variation might be buf- N/N value was around 0.5 from our simulation model.
e
fered with age as juvenile survival seems to be rather ran- Becausecensussizeisoftentheonlyavailabledemographic
614 ª2012BlackwellPublishingLtd5(2012)607–618
Serbezovetal. Lifehistoryanddemographicdeterminantsofeffective/censussizeratios
parameter from wild populations, obtaining taxon-specific revealed that the effect of body size on siring offspring is
N /N and N/N would be useful for monitoring changes significant (see also Serbezov et al. 2010). Larger individu-
b i e
in genetic diversity. The N /N values are within the range als leave more offspring compared to smaller individuals,
b i
of those reported for other salmonids, both semelparous and this difference widens as individuals reproduce
(0.1–0.4, (Hendry and Stearns 2004)) and iteroparous repeatedly. Such a scenario conforms with the fact that
(0.11–0.31, (McLean et al. 2007)). Generational N/N is family-size variance increases with decrease in survival in
e
theoreticallyexpectedtobearound0.5foriteroparousspe- our sensitivity analysis. Other studies have also conjec-
cies with overlapping generations (Nunney 1993) and tured that iteroparity and multiple paternity can increase
rangebetween0.25and0.75(NunneyandElam1994;Wa- the variance in reproductive success and therefore reduce
iteandParker1996).Ina recentreview,PalstraandRuzz- N (Karl 2008). In the closely related Atlantic salmon
e
ante (2008) reported a median N/N value of 0.14 for a (Salmo salar), repeat spawners can also produce dispro-
e
rangeoforganisms,butthatthisratioisactuallyhigherfor portionate share of recruitment (Mills 1989).
small populations. The formula for effective size of the Small populations may fluctuate in numbers for a vari-
population assumes no environmental stochasticity, stable ety of reasons, including demographic stochasticity, in
age structure and constant population size. However, in addition to natural or human-driven environmental per-
organismswithoverlappinggenerations,N/Nratiosbelow turbations. If population fluctuations are sufficiently large
e
0.1 apparently require extreme conditions, one of which and not rapidly attenuated, they could lead to loss of
mightbeselection(Nunney1993,1996;NunneyandElam genetic diversity and possibly population extinction
1994). Comparisons of estimated N (39.8, on average) (McCauley 1991). Complexity, roughly defined as the
b
and N (104) verifies that for iteroparous species N per degree of iteroparity, is positively correlated with stability
e e
generation is not simply a product of per season N and in terms of ability to recover from perturbations (Demet-
b
generation length (approximately 6 years), as is the case rius et al. 2004). Here, the N estimates obtained by
e
forsemelparousspecies(Waples1990). assuming strict semelparity led to increased variance in
the N estimates, implying that such a population might
e
be more exposed to stochastic demographic processes.
Family-specific survival
There were signs of demographic/genetic stochasticity/
Survival at later juvenile stages (older than 1+) was rather perturbations in our system. The 2005 cohort appears to
random with respect to families and occasionally led to be relatively much reduced in samplings from fall 2006,
increases in N /N (Table 2). Similar pattern has also been despite similar sampling effort and catchability (data not
b i
found in other salmonid studies (Hedrick et al. 2000). shown). The estimated total number of individuals in the
More pronounced family-correlated survival might be stream varied up to 45% between the seasons investi-
expected for younger juvenile stages than what we can gated. This considered, and in the light of the size of the
make inference for, when the juveniles belonging to a sib- population, demographic stochasticity might not be
ship group are especially vulnerable and tend to be uncommon in this system. Nevertheless, the system seems
spatially clumped (Serbezov et al. 2010). We observed, to be resilient to perturbations as it shows relatively high
however, marked year-to-year variation in the amount by levels of genetic variation (data not shown). Iteroparity,
which N /N is reduced between 1+ and 3+ life stages. which can be viewed as a ‘bet-hedging’ strategy to deal
b
Stochastic environmental or demographic events would with environmental uncertainty, thus might be an effec-
explain such year-to-year variation. Year-to-year variabil- tive strategy in such a system which might regularly expe-
ity in juvenile survival has been shown to dominate over rience years with poor recruitment and low number of
size-dependent survival in this system (Carlson et al. breeders (Gaggiotti and Vetter 1999). Some support for
2008). this comes from empirical studies on long-lived fish spe-
Our simulation results show that the iteroparous sys- cies (e.g., Diaz et al. 2000; Lippe et al. 2006), which
tem has higher variance in reproductive success that con- reported the maintenance of high genetic diversity despite
sequently lowered the N value. This was despite the fact relatively small population sizes. Patterns of iteroparity
e
that iteroparous individuals are more productive because between steelhead trout (Oncorhynchus mykiss) popula-
of higher cumulative fecundity and lifetime fitness (see tions reveal that the levels of iteroparity are highest at lat-
also Fleming and Reynolds 2004). In polygamous systems, itudinal extremes and suggest that it might be a more
iteroparity can be thought to lead to subordinate individ- beneficial strategy at harsher/less predictable conditions
uals having more opportunities to be successful, reducing (Stearns 1976; Roff 2002; Einum and Fleming 2007).
the between-individual variance in reproductive success From a conservation or management standpoints, it is
and thus increasing N (Nunney 1993). We could not important to understand how populations might evolve
e
find any significant cost of reproduction, but our results in response to environmental changes and how these
ª2012BlackwellPublishingLtd5(2012)607–618 615
Lifehistoryanddemographicdeterminantsofeffective/censussizeratios Serbezovetal.
responses might then feed back on population persistence 19:3193–3205 and can be found at Dryad: doi:10.5061/
and productivity. Among other things, this requires an dryad.1597.
understanding of the factors affecting effective population
size. Here, our results emphasize the importance of
Literature cited
obtaining knowledge of important life history traits and
delineating their impact on evolutionary dynamic pro- Ardren,W.R.,andA.R.Kapuscinski.2003.Demographicandgenetic
estimatesofeffectivepopulationsize(Ne)revealsgeneticcompensa-
cesses such as those reflected by N. The life history and
e
tioninsteelheadtrout.MolecularEcology12:35–49.
demographicparametersthataremostessentialtoapopu-
Barot,S.,M.Heino,L.O’Brien,andU.Dieckmann.2004.Estimating
lation’s conservation can thus be better identified, espe-
reactionnormsforageandsizeatmaturationwhenageatfirst
cially in species with complex life histories, allowing reproductionisunknown.EvolutionaryEcologyResearch6:659–678.
researcherstodesignimprovedmanagementandconserva- Bohlin,T.,S.Hamrin,T.Heggberget,G.Rasmussen,andS.Saltveit.
tion programmes. Namely, we documented large effect of 1989.Electrofishing–theoryandpracticewithspecialemphasison
thevarianceinreproductivesuccessonreducingtheN /N salmonids.Hydrobiologia173:9–43.
b i
Carlson,S.M.,E.M.Olsen,andL.A.Vøllestad.2008.Seasonalmortal-
ratio in a wild salmonid population. However, this
ityandtheeffectofbodysize:areviewandanempiricaltestusing
reductionmaybebufferedwithageofthejuvenilesasfam-
individualdataonbrowntrout.FunctionalEcology22:663–673.
ily-specific survival seems to be rather random at older
Charlesworth,B.2009.Effectivepopulationsizeandpatternsofmolec-
juvenilestages.Simulationresultsalsoindicatedthatitero- ularevolutionandvariation.NatureReviewsGenetics10:195–205.
parity increases the variance in reproductive success, lead- Charlesworth,D.,andJ.H.Willis.2009.Thegeneticsofinbreeding
ing to a decreased effective population size. At the same depression.NatureReviews.Genetics10:783–796.
time, iteroparity decreases the variance of the effective Crow,J.F.,andC.Denniston.1988.Inbreedingandvarianceeffective
populationnumbers.Evolution42:482–495.
population size, which might lead to a decreased risk of
Crow,J.,andM.Kimura.1970.AnIntroductiontoPopulationGenet-
demisebecauseofstochasticpopulationevents.Theevolu-
icsTheory.Harper&Row,NewYork,NY.
tionary implications of effective population size therefore
Crow,J.F.,andN.E.Morton.1955.Measurementofgenefrequency
depend on detailed knowledge of its interactions with life driftinsmallpopulations.Evolution9:202–214.
history and population dynamics. These insights may be Crozier,L.G.,A.P.Hendry,P.W.Lawson,T.P.Quinn,N.J.Man-
attained only through carefully considering age structure tua,J.Battin,R.G.Shawetal.2008.PERSPECTIVE:potential
in empirical investigations of N (Waples 2010). Finally, responsestoclimatechangeinorganismswithcomplexlifehistories:
e
evolutionandplasticityinPacificsalmon.EvolutionaryApplications
environmental change might produce conflicting selection
1:252–270.
pressures in different life stages in organism with complex
Demetrius,L.,V.MatthiasGundlach,andG.Ochs.2004.Complexity
life histories, which will interact with plastic (i.e., nonge-
anddemographicstabilityinpopulationmodels.TheoreticalPopu-
netic)changesinvariousways(Crozieret al.2008).Inthe lationBiology65:211–225.
case of the resident brown trout population studied here, Diaz,M.,D.Wethey,J.Bulak,andB.Ely.2000.Effectofharvestand
weinferhighlevelsofresilience,whichresultsfromacom- effectivepopulationsizeongeneticdiversityinastripedbasspopu-
bination of plastic life history traits such as promiscuous lation.TransactionsoftheAmericanFisheriesSociety129:1367–
1372.
matingsystemanditeroparity.Specifically,theN seemsto
e
Einum,S.,andI.A.Fleming.2007.Ofchickensandeggs:diverging
be pretty resilient with respect to age-specific maturation
propagulesizeofiteroparousandsemelparousorganisms.Evolution
probabilities and survival rates. Nevertheless, the factors
61:232–238.
identifiedinthispapertoreducetheNeshouldbecarefully Fleming,I.A.1998.Patternandvariabilityinthebreedingsystemof
consideredinconservationandmanagementapplications. Atlanticsalmon(Salmosalar),withcomparisonstoothersalmonids.
CanadianJournalofFisheriesandAquaticSciences55:59–76.
Fleming,I.A.,andJ.D.Reynolds.2004.Salmonidbreedingsystems.
Acknowledgements InA.P.Hendry,andS.C.Stearns,eds.EvolutionIlluminated,pp.
264–294.OxfordUniversityPress,Oxford.
We thank the Norwegian Research Council for financial
Frankham,R.1995.Effectivepopulation-sizeadultpopulationsize
supportoverseveralyears.Wealsothankthelargenumber
ratiosinwildlife–areview.GeneticalResearch66:95–107.
of graduate students who have participated in fieldwork Frankham,R.1996.Relationshipofgeneticvariationtopopulationsize
overtheyears.Thelargeamountoflaboratoryworkwould inwildlife.ConservationBiology10:1500–1508.
have been impossible without the kind assistance of Lucie Frankham,R.,J.D.Ballou,andD.A.Briscoe.2010.Introductionto
Papillon,VickyAlbertandAnnetteTaugbøl. ConservationGenetics.CambridgeUniversityPress,Cambridge.
Frasier,D.J.,M.M.Hansen,S.Ostergaard,N.Tessier,M.Legault,
andL.Bernatchez.2007.Comparativeestimationofeffectivepopu-
Data archiving statement lationsizesandtemporalgeneflowintwocontrastingpopulation
systems.MolecularEcology16:3866–3889.
The raw data used for the parentage assignment analyses Gaggiotti,O.E.,andR.D.Vetter.1999.Effectoflifehistorystrategy,
were archived for Serbezov et al., Molecular Ecology environmentalvariability,andoverexploitationonthegenetic
616 ª2012BlackwellPublishingLtd5(2012)607–618