Animal Conservation (2004) 7, 1–8 © 2004 The Zoological Society of London

DOI:10.1017/S1367943004003476 Printed in the United Kingdom

The relationship between population size and temporal variability

in population size

David H. Reed1 and Gayla R. Hobbs2,3

1
The University of Mississippi, Department of Biology, PO Box 1848, University, MS 38677–1848, USA
2
Southwest Missouri State University, 901 S. National Avenue, Springfield, MO 65804, USA
3
Current address: Department of Statistics, Purdue University, 150 N. University Street, West Lafayette, IN 47907–2067, USA
(Received 9 December 2002; accepted 4 June 2003)

Abstract
The relationship between population size and temporal variability in population size was examined using
2387 populations of 203 species from the Global Population Dynamics Database. Population variability,
relative to population size, was assayed by regressing the standard deviation of population size against
mean population size. Linear regression produced slopes that were less than one in 165 of 203 species.
Thus, temporal fluctuations in population size became significantly weaker as population size increased.
Similarly, the slope was significantly less than one for a single regression including all 2387 populations,
regardless of taxonomic classification. The slopes of the regression lines did not differ for major
taxonomic groupings, but the Y-intercept was significantly lower for birds than for the other taxonomic
groupings. Factor analysis was used to examine the highly correlated parameters: data reliability,
population size and taxonomic lineage. Population size was obviously the most important parameter
affecting temporal variation in population size, data reliability was also very important, but taxonomy
was of little or no importance. The relationship between temporal variation in population size and mean
population size, predicted from the data, was used to predict the probability of extinction assuming a
normal distribution of population sizes. This simple model predicts that populations of vertebrates will
have to number in the thousands for long-term conservation to be effective.

length of the organism and the length of time that the

INTRODUCTION
Smaller populations suffer higher rates of extinction than
do larger populations (e.g., Brown, 1971; Jones &
Diamond, 1976; Diamond, Bishop & Van Balen, 1987;
Newmark, 1987; Primm, Jones & Diamond, 1998;
Richman, Case & Schwaner, 1988; Soulé et al., 1988;
Berger, 1990; Kindvall & Ahlén, 1992; Schoener &
Spiller, 1992; Pimm et al., 1993; Rosenzweig, 1995;
Foufopoulos & Ives, 1999), as do more variable
populations of a given size (Thomas, 1990; Vucetich et
al., 2000; Reed et al. 2003a). Population size is weighted
heavily in decisions concerning the estimated extinction
risk for populations of plants and animals (e.g., IUCN,
2000). Temporal variability in population size has
received a great deal of attention (Taylor, 1961; Taylor,
Woiwood & Perry, 1978, 1979, 1980; Taylor &
Woiwood, 1980), but is seldom given much weight in
decisions concerning levels of endangerment. This is
probably due to the fact that per year variability in
population size will be greatly affected by the generation
All correspondence to: D. H. Reed. Tel: 662–915–6647; Fax:
662–915–5144; E-mail: dreed@olemiss.edu.
organism is studied (Pimm & Redfearn, 1988; Inchausti
& Halley, 2001; Thomas, Wilson & Lewis, 2002; Reed et
al., 2003a), and because an empirical and experimental
link between variability in population size and extinction
risk has been elusive for statistical reasons (Vucetich
et al., 2000).
However, it is not population size per se that drives the
extinction process. Rather, it is the expectation concerning
the relationships among population size, population
variability and population fitness that leads us to believe that
small populations are in greater peril of extinction. Small
populations are viewed as more vulnerable to extinction,
owing in part to the expectation that they are also more
variable in relation to their size. Three forms of stochasticity
are widely held to influence population persistence:
demographic, environmental (including catastrophes) and
genetic (Shaffer, 1981). Each of these is expected to increase
in the strength of its effects with decreasing population size.
We elaborate on each of these below.
Demographic stochasticity is caused by sampling
variance among individual probabilities of death and
reproduction in a finite population. Demographic
stochasticity can increase the probability of extinction
2 D. H. REED AND G R. HOBBS
greatly, but is relevant to only very small populations
relative to the other two forms of stochasticity.
Demographic stochasticity probably increases the
probability of extinction significantly only in populations
of ≤25 individuals (Lande, 1993).
Environmental stochasticity results from perturbations
in the abiotic and/or biotic environment of the population.
It is unclear how best to model these perturbations.
Changes in climactic variables are autocorrelated through
time. Thus, environmental stochasticity may be seen as
the result of long-term cyclic (or directional) changes in
the global environment, with local stochasticity being due
to a number of lesser factors acting at smaller scales.
Catastrophes are defined as rare instances of very large
environmental perturbations, causing major reductions in
population size. Many (e.g., Ewens et al., 1987; Lande,
1993; Mangel & Tier, 1993) have modelled catastrophes
as separate from ‘normal’ environmental variation and
catastrophes presumably stem from a different
environmental source. However, Reed et al. (2003b)
present evidence contrary to this hypothesis, suggesting
that catastrophes are simply the extreme end of a
continuous distribution of environmental disturbances
with their frequency scaling to the generation length of
the organism.
Regardless of the nature of environmental stochasticity,
its effects should be dampened in larger populations. As
a population becomes larger, it will cover a larger
geographic area (all else being equal). Because the
correlation among population fluctuations between
adjacent sites declines with increasing distance among
sites, discrete populations or subpopulations among the
metapopulation become asynchronous in their fluctuations
with more distant neighbours and the entire population
remains relatively stable (Hanski & Woiwood, 1993;
Ranta, Kaitala & Lindström, 1995; Lindström, Ranta &
Lindén, 1996; Sutcliffe, Thomas & Moss, 1996;
Bjørnstad, Stenseth & Saitoh, 1999; Paradis et al., 2000;
Koenig, 2001).
Genetic stochasticity includes random genetic drift and
inbreeding depression. Inbreeding leads to inbreeding
depression in virtually all species studied thus far (e.g.,
Charlesworth & Charlesworth, 1999; Crnokrak & Roff,
1999; Hedrick & Kalinowski, 2000; Keller & Waller,
2002; Reed & Frankham, 2003). Random genetic drift
results when the product of effective population size (Ne)
and the coefficient of selection (s) against an allele is ≤1.0
(Crow & Kimura, 1970). Thus, selection is weaker in
small populations than in large populations, leading to the
possible fixation of deleterious alleles. Further, smaller
populations are less likely to give rise to rare beneficial
alleles that can have significant effects on the progress of
evolution (Elena, Cooper & Lenski, 1996; Burch & Chao,
1999) and those that do occur are more likely to be lost
through drift (Kimura 1983).
Inbreeding depression and random genetic drift pose a
threat to the persistence of populations of sizes much
greater than is generally believed (Franklin, 1980; Schultz
& Lynch, 1997; Reed, 1998; Reed & Bryant, 2000;
Whitlock, 2000; Brook et al., 2002; Reed & Frankham,
2003). An organism’s environment undergoes almost
constant perturbations. Thus, any factor that reduces
fitness, and therefore recovery time after a perturbation,
will make the population more susceptible to extinction
when further perturbations occur.
The threat of genetic stochasticity differs from the other
two forms of stochasticity by operating via the reduction
in mean population fitness, rather than directly affecting
population variability. In fact, the expected relationship
between genetic stochasticity and population variability
is not clear. Population models, built from extensive
demographic data representing 102 wild populations of
vertebrates, suggest that lower population growth rates
lead to larger temporal fluctuations in population size
(Reed et al., 2003a).
Here we present data on 2387 populations, from 203
species, concerning the relationship between population
size and population variability. In particular we examine
the following questions regarding the relationship
between population size and temporal variability in
population size: (1) What is the strength of the
relationship? (2) What is the form of the relationship (e.g.,
linear) and what does this tell us about the forms of
stochasticity affecting population variability? (3) Does the
strength and shape of the relationship vary among taxa?
(4) How does the relationship affect the probability of
extinction at various population sizes?
The empirical relationship between population size
and population variability has been well established
by the pioneering work of Taylor and his colleagues
(Taylor, 1961; Taylor et al., 1978, 1979, 1980; Taylor
& Woiwood, 1980). Our current study extends the results
of Taylor in a number of ways, most importantly by
including large numbers of long-term vertebrate
studies. Further, we provide valuable insights into the
Global Population Dynamics Database (NERC, 1999),
which is a valuable tool for conservation (Inchausti &
Halley, 2001).
METHODS
To quantify the relationship between population size and
population variability, we collected population census
data from the Global Population Dynamics Database
(GPDD: NERC, 1999). We searched the database for
census data on any species that had at least three
populations represented. Populations used in the analysis
were limited to those that were judged to be isolated and
little influenced by migration and that were measured in
a similar way so that individual series are comparable.
Data sets with more than one zero in the time series
were not included in the analysis. We calculated a
mean population size (N) and standard deviation in
population size. The standard deviation was calculated
from direct measurements of year to year variation in
population size.
Curve-fitting software was used to produce best-fit
lines, using population size as the independent variable
and the standard deviation in population size as the
dependent variable for all 2387 populations regardless of
 Population size and variability
species, for the 321 populations of birds separately, for
the 25 populations of crustaceans separately, for the 216
populations of fish separately, for the 1032 populations of
insects separately, for the 764 populations of mammals
separately, and for a number of combinations grouped
according to data reliability. The estimates of the
relationship between population size and the standard
deviation in population size, from the 108 most reliable
vertebrate data sets, were used to estimate extinction times
and compare these with estimates of extinction from other
population viability models.
Factor analysis and multiple regression were used to
assess the relative importance of data reliability, population
size and taxonomic status on population variability.
RESULTS
Relationship between population size and variability
in population size
Relative population variability was assayed by regressing
the standard deviation of population size against the mean
population size for each of the 203 species sets. All such
regressions produce relationships that are very linear.
Regressing the log10 transformed standard deviations
versus the log10 transformed means produced slopes that
were less than one in 165 of 203 species. This difference
is highly significant using a sign test (P < 0.001).
Similarly, regressing the standard deviation against
the mean produced a slope that was significantly less than
one for the entire data set (P < 0.001) regardless of
Fig. 1. The best-fit line for all 2387 populations regardless of taxon. Y = a + bx. Where Y is the log of the standard deviation of
population size, a = 0.0024, b = 0.9329 and x is the log of mean population size.
3
taxonomic class, and the slope was also less than one for
each of the five taxonomic classes individually (P < 0.001
for each).
Taxonomic specific relationships
The slope of the regression lines did not differ for the five
major taxonomic groups sampled (all classes with n ≥ 25).
The slopes are: 0.93 ± 0.004 (birds), 0.94 ± 0.013
(crustaceans), 0.92 ± 0.009 (fish), 0.93 ± 0.007 (insects)
and 0.93 ± 0.007 (mammals). However, the Y-intercept
was significantly lower for birds than for the other
taxonomic groupings. Thus the rate at which the standard
deviation in population size increases was less than one
and very similar for all taxonomic groups, but birds had
significantly lower standard deviations, for any given
population size, than did the other four taxonomic groups.
Figure 1 shows the best-fit line, for the entire 2387
populations, using linear regression (r2 = 0.9678, P <
0.0001, Y-intercept = 0.0024, b = 0.9329). Both the mean
population size and the standard deviation in population
size have been log10 transformed. The linear regression
statistics for the other taxonomic groups are as follows:
321 populations of birds (r2 = 0.9398, P < 0.0001, Y-
intercept = [–0.1979], b = 0.9329), 25 populations of
crustaceans (r2 = 0.9959, P < 0.0001, Y-intercept = 0.0939,
b = 0.9434), 216 populations of fish (r2 = 0.9839, P <
0.0001, Y-intercept = 0.0581, b = 0.9237), 1032
populations of insects (r2 = 0.9432, P < 0.0001, Y-intercept
= 0.0276, b = 0.9339) and 764 populations of mammals
(r2 = 0.9578, P < 0.001, Y-intercept = 0.0611, b = 0.9269).
4 D. H. REED AND G R. HOBBS
Table 1. Predicted CVN at different population sizes for various
Factors contributing to the relative magnitude of groupings. The numbers in parentheses, following the vertebrate
temporal variation category, refer to the probability of extinction at that population size.
These are based on the assumption of a normal distribution of
Table 1 shows the predicted coefficient of variation in population sizes through time, using the predicted standard deviation
population size (CVN ; which is the standard deviation in for the given population size, over a 250-year period. See the Results
population size divided by the mean population size) for section for more details of the model.
the various taxa, using the best-fit line for each, at different Population size
population sizes. 100 250 500 1000 2500
Besides population size and taxonomic status, the All 0.738 0.694 0.663 0.632 0.595
reliability of the census data can influence CVN. Further, Birds 0.465 0.437 0.417 0.398 0.374
Crustaceans 0.957 0.908 0.873 0.840 0.797
these variables are all highly correlated with each other. Fish 0.804 0.750 0.711 0.675 0.629
For instance, data sets with high scores for reliability also Insects 0.786 0.740 0.706 0.675 0.635
tend to be small, and birds and insects tend to have data Mammals 0.822 0.769 0.731 0.695 0.650
sets with high reliability scores and smaller population Vertebrates (best) 0.430 0.364 0.321 0.283 0.239
(92.3%) (52.8%) (20.2%) (5.7%) (0.3%)
sizes. Factor analysis was used to disentangle these
relationships (Table 2). Factor analysis shows that
population size was clearly the most important factor
Table 2. Results of factor analysis examining the relationship of
affecting CVN , independent of taxon and data reliability. population size, data reliability and taxonomic classification on
However, data reliability was also very important. temporal variation in population size. Clearly, population size is the
Taxonomic class matters very little. Multiple regression parameter most closely correlated wtih CVN.
was used to supplement the results from factor analysis Parameters Factor Loadings
(data not shown). While all three variables were
CVN 0.869
significant, population size was clearly the most important Population size –0.619
parameter in the multiple regression model as well, Reliability –0.164
explaining 31.7% of the variation in CVN by itself. Taxon 0.007
Both reliability and taxon were significant, but added
only 3.1% and 3.0%, respectively, to the total variance in
CVN explained by the model containing population Figure 2 shows what we consider to be the best model,
size alone. using the 108 most reliable data sets for vertebrates. The
The reliability of the data has a profound effect on the data set was limited to vertebrates for two reasons: (1) to
estimated magnitude of variation in population size. We allow comparison with populations modelled in previous
removed data sets in a hierarchical fashion starting with studies, and (2) because insect species are subject to
the lowest reliabilities and moving up. CVN decreased outbreaks that greatly increase CVN while probably having
dramatically with each removal of less reliable data sets. little effect on the probability of extinction. Figure 3

Fig. 2. Best-fit line for the 108 most reliable vertebrate populations. Y = a + bx. Where, Y is the log of the standard deviation of
population size, a = (–0.0021), b = 0.8179 and x is the log of mean population size.
 Population size and variability 5

provides the species, mean population size, standard
deviation in population size and reference for all 108 data
sets in the vertebrate model.
Relevance to extinction
The relationship between the standard deviation of
population size and population size, from our 108 most
reliable vertebrate data sets is Y = a + bx where Y is the log
of the standard deviation in population size, a is the
constant (–0.0021), b is the constant 0.8179 and x is the
log of mean population size. Thus, a population of 100
individuals is expected to have a standard deviation in
population size of 43. Any population of size N with a
standard deviation, s, can have its probability of reaching
zero (going extinct) estimated under the assumptions of a
normal distribution. The assumption that a series of
population sizes is normally distributed is probably a

Table 3. Species, mean population size, standard deviation in population size and references for the 108 populations of vertebrates judged to be
of the highest quality. The ID represents the Global Population Dynamics Database ID number for that population, except in the few cases where
another source is cited.

Species Mean Standard ID Species Mean Standard ID

deviation deviation
Accipiter nisus 208.4 74.48 3202 Parus caeruleus 73.6 23.84 3207
Accipiter nisus 165.2 37.24 3251 Parus caeruleus 77.5 21.16 3208
Accipiter nisus 95.2 48.68 3252 Parus caeruleus 67.2 19.67 3209
Accipiter nisus 64.6 15.52 3253 Parus caeruleus 75.2 28.82 3210
Anguilla anguilla 5417.1 1310.00 3249 Parus caeruleus 76.8 32.59 3211
Canis lupus 13.3 11.14 3605 Parus caeruleus 32.6 12.31 3212
Canis lupus 12.3 8.60 3607 Parus caeruleus 88.5 33.52 3213
Cervus elephus 40.2 20.26 31363 Parus caeruleus 33.3 7.30 3418
Connochaetes taurinus 19,194.0 3033.82 3640 Parus caeruleus 31.1 14.00 10,572
Corvus corone 15.2 3.53 333 Parus caeruleus 80.3 31.81 31,403
Corvus corone 15.9 3.70 1549 Parus major 24.8 7.82 1105
Corvus frugilegus 2144.0 159.51 2735 Parus major 6.2 2.16 1266
Corvus frugilegus 441.5 128.57 2741 Parus major 60.9 28.64 1348
Corvus frugilegus 257.2 37.87 2742 Parus major 133.4 58.95 1349
Cyanocitta cristata 17.4 10.75 1036 Parus major 27.6 17.60 1508
Cyanocitta cristata 33.7 20.20 1078 Parus major 68.9 28.49 3454
Dama dama 955.1 80.90 31,362 Parus major 107.4 42.61 6148
Dendroica kirtlandii 487.4 168.02 Dennis, Munholland Parus major 706.1 193.64 6190
& Scott, 1991 Parus major 69.8 24.84 10,571
Enhydra lutris 1629.2 285.81 3579 Parus major 84.8 20.49 31,404
Falco rusticolus 75.9 44.45 31,448 Parus montanus 60.8 36.34 1262
Ficedula albicollis 189.9 91.01 10,573 Parus palustris 5.5 2.28 1108
Ficedula albicollis 144.9 43.80 31,399 Parus palustris 40.1 11.55 31,402
Ficedula hypoleuca 170.6 91.90 1263 Phalacrocorax aristotellis 315.6 243.26 3201
Ficedula hypoleuca 67.8 23.90 1378 Phoca groenlandica 87.6 33.42 3344
Ficedula hypoleuca 139.5 29.36 10,520 Phoca vitulina 4563.2 662.66 3465
Ficedula hypoleuca 26.6 9.17 31,400 Phoca vitulina 1513.2 182.45 3466
Ficedula parva 19.5 4.82 31,398 Phoca vitulina 1500 418.33 3467
Fringilla coelebs 2.0 5.77 1125 Phoca vitulina 913.2 378.54 3468
Fringilla coelebs 42.8 30.04 1260 Phoca vitulina 31.8 11.21 3469
Fringilla coelebs 562.3 73.50 31,407 Picoides pubescens 8.7 4.32 1033
Fringilla monfifringilla 67.3 32.04 1264 Picoides pubescens 7 3.46 1061
Fulmarus glacialis 417.3 136.12 3203 Picoides pubescens 32.5 18.5 1077
Grus americanus 50.7 31.87 6176 Picoides pubescens 28.1 13.40 10,518
Gymnogyps californianus 31.0 14.55 Dennis et al., 1991 Picoides villosus 21.9 7.50 10,519
Hirundo rustica 66.4 35.20 3250 Rissa tridactyla 126.8 59.07 3264
Loxioides balleui 2875.0 1707.05 Dennis et al., 1991 Spizella arborea 19.2 17.12 1071
Lycaon pictus 28.7 16.83 3199 Spizella pusilla 4.1 3.18 1052
Macaca fuscata 17.2 6.19 2733 Spizella pusilla 270 111.81 1091
Macaca fuscata 222.8 88.50 2737 Sturnus vulgaris 43.4 36.08 329
Macaca fuscata 214 86.60 2740 Sturnus vulgaris 74.7 46.26 1045
Macaca sylvanus 15.2 4.39 2723 Sturnus vulgaris 8.8 13.89 1067
Macaca sylvanus 14.3 4.73 2724 Sturnus vulgaris 12.7 6.12 1123
Martes americana 157.1 37.43 6110 Sturnus vulgaris 20.3 19.78 1553
Melospiza melodia 91.5 34.51 3265 Sturnus vulgaris 66.3 29.79 31,409
Mirounga leonina 522.6 338.00 3256 Telespyza cantans 10329.9 3833.58 Dennis et al., 1991
Ovis canadensis 171.5 59.23 31,329 Tragelaphus strepsiceros 197.5 41.08 3267
Panthera leo 34.1 10.32 3198 Tragelaphus strepsiceros 143.2 26.19 3268
Papio cynocephalus 413.4 197.18 1376 Ursus arctos horribilis 80.1 14.92 1337
Parus atricapillus 48.7 23.67 1079 Vanellus vanellus 20.8 14.80 330
Parus atricapillus 97.3 37.66 1334 Vanellus vanellus 24.0 15.77 1546
Parus bipolar 63.5 40.51 1080 Vanellus vanellus 129.7 30.35 3646
Parus caeruleus 288.7 79.74 1106 Zenaida macroura 14.7 8.42 1024
Parus caeruleus 12.9 10.91 1506 Zenaida macroura 9.8 11.71 1055
Parus caeruleus 71.5 35.08 3206 Zenaida macroura 104.0 47.12 1073
6 D. H. REED AND G R. HOBBS
reasonable one (Reed et al., 2003b). Z = N1 – N2 / s where
Z is the Z-score corresponding to the area under the normal
curve, N1 is the mean population size, N2 is the threshold
size we want to estimate the probability of reaching (0 in
our case), and s is the standard deviation in population size.
For a population size of 100, Z equals 2.33. This
corresponds to a 1.0% chance, per year, of having a
population size of zero. The per year probability of going
extinct can be turned into an overall probability of
extinction by using the binomial expansion: P = [N! / x!
(N – x)!] px qN–x where P is the overall probability of an
event occurring x number of times given N number of
chances, N is the number of attempts, x is the number of
‘successes’, p the probability of the event occurring in any
given attempt, and q is 1 – p. In this case, the cumulative
probability of any population with a mean size of 100
going extinct during 250 years is approximately 92.3%.
(It is easier to solve the equation by setting x equal to zero
and solving for the probability of no extinctions in the
given timeframe and setting this number equal to 1 – P.)
Extinction probabilities calculated in this way are
presented in Table 1, directly under the estimates of CVN,
for the most reliable subset of vertebrate data.
DISCUSSION
The results of this study suggest the following about the
relationship between population size and temporal
variability in population size: (1) The relationship between
population size and temporal variation in population size
is negative, moderately strong, and very consistent among
different taxonomic groups. (2) The relationship between
population size and temporal variation in population size
is linear over the entire range of population sizes observed.
The standard deviation in population size increased in a
linear fashion with increasing population size and the
increase is significantly less than that for population size
itself. (3) Birds differed significantly from other
taxonomic classes in the magnitude of population
variability. However, this difference is primarily due to
differences in the quality of data for birds. Data sets in the
GPDD, with reliability scores less than three, generally
have large amounts of error variance. (4) The increase in
relative temporal variability with decreasing population
size is expected to have a profound effect on the
probability of extinction in populations of different sizes.
A very simple model relating the probability of extinction
with CVN gives predictions that are in broad agreement
with more sophisticated models of extinction. We
elaborate on each of these points below.
This study confirms a relationship, first demonstrated
by Taylor and colleagues and expected from theory, that
smaller populations have relatively more temporal
variation in population size than larger populations.
Whether the data is divided into groups of species or
higher taxonomic levels, the pattern of decreasing
temporal variation in population size with increasing mean
population size is consistent and highly significant.
The rate at which the standard deviation in population
size increases with increasing population size appears to be
linear across the entire range of population sizes observed.
Thus, all the models predict continuous decreases in the
standard deviation relative to the increases in population
size (i.e., slope less than one), even beyond the population
sizes where demographic stochasticity, or probably even
genetic stochasticity, will influence population variability.
Thus, all three forms of stochasticity seem to be important
in influencing the amount of temporal variation in
population size of natural populations.
The slopes of the regression lines predicted for each
taxonomic group are very homogeneous. However, birds
have significantly less temporal variation at a given
population size than do crustaceans, fish, insects or
mammals. For example, the expected CVN for birds at a
mean population size of 1000 is only 57.3% of that for
mammals at the same population size. Are bird
populations really less variable? Besides population size
and taxon, at least two other factors affect population
variability. One factor very likely to affect CVN is the
reliability of the data. The GPDD rates the reliability of
its data on a scale from 1 to 5, with 5 being the most
reliable. Statistical testing shows that reliability is strongly
correlated with the relative magnitude of temporal
variability. All the taxa differ significantly from each other
with respect to the quality of their data. Populations of
birds and insects had significantly higher reliability ratings
and smaller population sizes than crustaceans, fish and
mammals. In fact, all three variables (reliability, size and
taxon) are highly correlated with each other. Thus, we
turned to factor analysis to disentangle which variables
were truly responsible for influencing CVN . Population
size was the major factor correlated with CVN . Reliability
made a smaller, but highly significant contribution. The
effect of taxon was negligible. Thus, the answer seems to
be that bird populations are probably no less variable than
fish or mammal populations, and certainly not to the
degree suggested by an analysis that does not take into
account other correlated variables.
So why are insect populations more variable than bird
populations despite having better-quality data? One
reason could have to do with the length of the census
period. Population variability increases with increasing
census period and this increase is expected to scale to
generations. Using population models, Reed et al. (2003a)
have shown an approximately 25% increase in CVN with
each doubling of study length. Mean study lengths from
this data set are roughly 5.8 generations for birds, 5.4
generations for fish, 12.2 generations for insects and 10.2
generations for mammals. Thus, part of the explanation
for the greater variability in insect populations is the fact
that they are studied, on average, for more than twice as
many generations as birds. Further, insect populations are
prone to outbreaks that can greatly increase CVN.
Implications for conservation
More variable populations suffer from a greater risk of
extinction (Vucetich et al., 2000; Reed et al., 2003a). We
have confirmed and extended the work of Taylor (Taylor,
1961; Taylor et al., 1980; Taylor & Woiwood, 1980),
 Population size and variability
showing that smaller populations are more variable than
larger populations. This can be translated directly into
increased extinction risk. Using a simple model, based on
a normal distribution of population sizes with a standard
deviation predicted from the subset of the most reliable
data, we have predicted probabilities of extinction over a
250-year time-frame. These estimates will understate the
risk of extinction for several reasons. For example,
fluctuations in population size through time are positively
autocorrelated (Morales, 1999; Inchausti & Halley,
2001) and the completely random fluctuations in
our model probably underestimate the true risk of
extinction (Ripa & Lundberg, 1996; Johst & Wissel, 1997;
Morales, 1999)
These extinction times, predicted from increases in the
coefficient of variation in temporal fluctuations with
decreasing population size, can be compared to prior
research. Reed et al. (2003a) used population viability
models derived from extensive demographic data, on
natural populations for 102 vertebrates, to estimate
minimum viable population sizes (MVP). Using these
models, the mean estimated probability of extinction over
250 years for population sizes of 500, 1000 and 2500
would be 26.5%, 12.7% and 4.8%, respectively. The
predicted probability of extinction at the same three
population sizes from these data are 20.2%, 5.7% and
0.3%, respectively. Thomas (1990) used a similar logic
concerning the temporal variability of population sizes,
but not nearly as extensive a data set, and arrived at an
MVP of 5500. Thus, these models are in broad agreement
and underscore the growing consensus (Soulé, 1987;
Thomas, 1990; Reed & Bryant 2000; Shaffer, Watchman
& Snape, 2002; Reed et al. 2003a) that populations must
be conserved at population sizes that are greater than 2000
to be viable in the long term.
Acknowledgements
R. Palawan, B. Sanders and E. Sthapit helped with data
collection. S. Brewer, J. Gittleman, E. Linder and three
anonymous reviewers provided helpful comments on a
prior draft of the manuscript. Part of this research was
carried out while DHR was a visiting assistant professor
at Southwest Missouri State University.
REFERENCES
Berger, J. (1990). Persistence of different-sized populations: an
empirical assessment of rapid extinctions in bighorn sheep.
Conserv. Biol. 4: 91–98.
Bjørnstad, O. N., Stenseth, N. C. & Saitoh, T. (1999). Synchrony
and scaling in dynamics of voles and mice in northern Japan.
Ecology 80: 622–637.
Brook, B. W., Tonkyn, D. W., O’Grady, J. & Frankham, R. (2002).
Contribution of inbreeding to extinction risk in threatened species.
Conserv. Ecol. 6: 16.
Brown, J. H. (1971). Mammals on mountaintops: nonequilibrium
insular biogeography. Am. Nat. 105: 467–478.
Burch, C. L. & Chao, L. (1999). Evolution in small steps and
rugged landscapes in the RNA virus Φ6. Genetics 151: 921–927.
Charlesworth B. & Charlesworth, D. (1999). The genetic basis of
inbreeding depression. Genet. Res. 74: 329–340.
7
Crnokrak, P. & Roff, D. A. (1999). Inbreeding depression in the
wild. Heredity 83: 260–270.
Crow, J. F. & Kimura, M. (1970). An introduction to population
genetics. New York: Harper & Row.
Dennis, B., Munholland, P. L. & Scott, J. M. (1991). Estimation of
growth and extinction parameters for endangered species. Ecol.
Monogr. 61: 115–143.
Diamond, J. M., Bishop, K. D. & Van Balen, S. (1987). Bird
survival in an isolated Javan woodland: island or mirror? Conserv.
Biol. 2: 132–142.
Elena, S. F., Cooper, V. S. & Lenski, R. F. (1996). Punctuated
evolution caused by selection of rare beneficial mutations. Science
272: 1802–1804.
Ewens, W. J., Brockwell, P. J., Gani, J. M. & Resnick, S. I. (1987).
Minimum viable population sizes in the face of catastrophes. In
Viable populations for conservation: 59–68. Soulé, M. E. (Ed.).
New York: Cambridge University Press.
Foufopoulos, J. & Ives, A. R. (1999). Reptile extinctions on land-
bridge islands: life-history attributes and vulnerability to
extinction. Am. Nat. 153: 1–25.
Franklin, I. (1980). Evolutionary change in small populations. In
Conservation biology: an evolutionary–ecological perspective:
135–150. Wilcox, B. A. (Eds). Sunderland, MA: Sinauer
Associates.
Hanski, I. & Woiwood, I. P. (1993). Patterns of density dependence
in moths and aphids. J. Anim. Ecol. 62: 656–668.
Hedrick, P. W. & Kalinowskim S. T. (2000). Inbreeding depression
in conservation biology. Ann. Rev. Ecol. Syst. 31: 139–162.
Inchausti, P. & Halley, J. (2001). Investigating long-term ecological
variability using the global population dynamics database. Science
293: 655–657.
IUCN (2000). The 2000 IUCN red list of threatened species. Gland:
Species Survival Commission, IUCN.
Johst, K. & Wissel, C. (1997). Extinction risk in a temporally
correlated fluctuating environment. Theor. Pop. Biol. 52: 91–102.
Jones, H. L. & Diamond, J. M. (1976). Short-time base studies of
turnover in breeding birds of the California Channel Islands.
Condor 76: 526–549.
Keller, L. F. & Waller, D. M. (2002). Inbreeding effects in wild
populations. Trends Ecol. Evol. 17: 230–241.
Kimura, M. (1983). The neutral theory of molecular evolution.
Cambridge: Cambridge University Press.
Kindvall, O. & Ahlén, I. (1992). Geometrical factors and
metapopulation dynamics of the bush cricket, Metrioptera bicolor
Philippi (Orthoptera: Tettigoniidae). Conserv. Biol. 6: 520–529.
Koenig, W. D. (2001). Spatial autocorrelation and local
disappearances in wintering North American birds. Ecology 82:
2636–2644.
Lande, R. (1988). Genetics and demography in biological
conservation. Science 241: 1455–1460.
Lande, R. (1993). Risks of population extinction from demographic
and environmental stochasticity and random catastrophes. Am.
Nat. 142: 911–927.
Lindström, J., Ranta, E. & Lindén, H. (1996). Large-scale
synchrony in the dynamics of capercaillie, black grouse and hazel
grouse populations in Finland. Oikos 76: 221–227.
Mangel, M. & Tier, C. (1993). Dynamics of metapopulations with
demographic stochasticity and environmental catastrophes. Theor.
Pop. Biol. 44: 1–31.
Morales, J. M. (1999). Variability in a pink environment: why
‘white noise’ models can be dangerous. Ecol. Lett. 2: 228–232.
NERC (1999). NERC & Centre for Population Biology, Imperial
College. The Global Population Dynamics Database.
http://www.Sw.ic.ac.uk/cpb/cpb/gpdd.html.
Nei, M., Maruyama, T. & Chakraborty, R. (1975). The bottleneck
effect on genetic variability in populations. Evolution 29: 1–10.
Newmark, W. D. (1987). A land-bridge island perspective on
mammalian extinctions in western North American parks. Nature
325: 430–432.
8 D. H. REED AND G R. HOBBS
Paradis, E., Baillie, S. R. Sutherland, W. J. & Gregory, R. D. (2000).
Spatial synchrony in populations of birds: effects of habitat,
population trend, and spatial scale. Ecology 81: 2112–2125.
Pimm, S. L., Diamond, J., Reed, T. M., Russell, G. J. & Verner, J.
(1993) Times to extinction for small populations of large birds.
Proc. Nat. Acad. Sci. USA 90: 10,871–10,875.
Pimm, S. L., Jones, J. L. & Diamond, J. (1988). On the risk of
extinction. Am. Nat. 132: 757–785.
Pimm, S. L., & Redfearn, A. (1988). The variability of population
densities. Nature 334: 613–615.
Ranta, E., Kaitala, V. & Lindström, J. (1995). Synchrony in
population dynamics. Proc. R. Soc. Lond. Ser. B 262: 113–118.
Reed, D. H. (1998). Population size, selection, and mutation
accumulation. Ph.D. thesis. University of Houston, Texas.
Reed, D. H. & Bryant, E. H. (2000). Experimental tests of minimum
viable population size. Anim. Conserv. 3: 7–14.
Reed, D. H. & Frankham, R. (2003). Correlation between fitness
and genetic diversity. Conserv. Biol. 17: 230–237.
Reed, D. H., O’Grady, J. J. Brook, B. W., Ballou, J. D. & Frankham,
R. (2003a). Estimates of minimum viable population sizes for
vertebrates and factors influencing those estimates. Biol. Conserv.
113: 23–34.
Reed, D. H., O’Grady, J. J., Ballou, J. D. & Frankham, R. (2003b).
The frequency and magnitude of catastrophic die-offs in
populations of vertebrates. Anim. Conserv. 6: 109–114.
Richman, A. D., Case, T. J. & Schwaner, T. D. (1988). Natural and
unnatural extinction rates of reptiles on islands. Am. Nat. 131:
611–630.
Ripa, J., & Lundberg, P. (1996). Noise colour and the risk of
population extinctions. Proc. R. Soc. Lond. Ser. B 263: 75–77.
Rosenzweig, M. L. (1995). Species diversity in space and time.
Cambridge: Cambridge University Press.
Schoener, T. W. & Spiller, D. A. (1992). Is extinction rate related
to temporal variability in population size? An empirical answer
for orb spiders. Am. Nat. 139: 1176–1207.
Schultz, S. T. & Lynch, M. (1997). Mutation and extinction: the
role of variable mutational effects, synergistic epistasis, beneficial
mutations, and degree of outcrossing. Evolution 51: 1363–1371.
Shaffer, M. L. (1981). Minimum population sizes for species
conservation. BioScience 31: 131–134.
Shaffer, M. L., Watchman, L. H. Snape III, W. J. & Latchis, I. K.
(2002). Population viability analysis and conservation policy. In
Population viability analysis: 123–142. Beissinger, S. R. &
McCullough, D. R. (Eds). Chicago: University of Chicago Press.
Sinclair, A. R. E. (1996). Mammal populations: fluctuations,
regulations, life history theory and their implications for
conservation. In Frontiers in population ecology: 127–154. Floyd,
R. B., Sheperd, A. W. & De Barro, P. J. (Eds). Melbourne: CSIRO
Publishing.
Soulé, M. E. (1987). Introduction. In Viable populations for
conservation: 1–10. Cambridge: Cambridge University Press.
Soulé, M. E., Bolger, D. T., Alberts, A. C., Wright, J., Sorice, M.
& Hill, S. (1988). Reconstructed dynamics of rapid extinctions of
chaparral-requiring birds in urban habitat islands. Conserv. Biol.
2: 75–92.
Sutcliffe, O. L., Thomas, C. D. & Moss, D. (1996). Spatial
synchrony and asynchrony in butterfly population dynamics. J.
Anim. Ecol. 65: 85–95.
Taylor, L. R. (1961). Aggregation, variance, and the mean. Nature
189: 732–735.
Taylor, L. R. & Woiwood, I. P. (1980). Temporal stability as a
density-dependent species characteristic. J. Anim. Ecol. 49:
209–224.
Taylor, L. R., Woiwood, I. P. & Perry, J. N. (1978). The density-
dependence of spatial behaviour and the rarity of randomness. J.
Anim. Ecol. 47: 383–406.
Taylor, L. R., Woiwood, I. P. & Perry, J. N. (1979). The negative
binomial as an ecological model for aggregation and the density-
dependence of k. J. Anim. Ecol. 48: 289–304.
Taylor, L. R., Woiwood, I. P. & Perry, J. N. (1980). Variance and
the large scale spatial stability of aphids, moths, and birds. J. Anim.
Ecol. 49: 831–854.
Thomas, C. D. (1990). What do real populations dynamics tell us
about minimum viable population sizes? Conserv. Biol. 4:
324–327.
Thomas, C. D., Wilson, R. J. & Lewis, O. T. (2002). Short-term
studies underestimate 30–generation changes in a butterfly
metapopulation. Proc. R. Soc. Lond. Ser. B 269: 563–569.
Vucetich, J. A., Waite, T. A., Qvarnemark, L. & Ibargüen, S.
(2000). Population variability and extinction risk. Conserv. Biol.
14: 1704–1714.
Whitlock, M. C. (2000). Fixation of new alleles and the extinction
of small populations: drift load, beneficial alleles, and sexual
selection. Evolution 54: 1855–1861.