28
29
Table 4. Results of four cases of different stochastic projections for long-billed curlew. Stochastic fluctuations have
the greatest effect when acting on first-year transitions (Case 3).
Case 1
Case 2
Case 3
Case 4
Input factors:
Affected cells
All the F
ij
All the P
ij
P
21
and F
11
(first year)
P
21
and F
11
(first year)
S.D. of random normal distribution
1/4
1/4
1/4
1/8
Output values:
Deterministic λ
1.003
1.003
1.003
1.003
# Extinctions/100 trials
1
3
3
0
Mean extinction time
1,667
1,445
1,445
N.a.
# Declines/#
survived pop
34/99
56/97
62/97
3/100
Mean ending population size
5.6 X 10
6
461,697
185,499
3.9 X 10
6
Standard deviation
4.6 X 10
7
2.5 X 10
6
533,737
1.0 X 10
7
Median ending population size
26,204
3,405
2,544
815,138
Log λ
s
0.0004
-0.0005
-0.0011
0.002
λ
s
1.0004
0.9995
0.9989
1.002
% reduction in λ
0.26
0.35
0.4
0.09
Table 3. Reproductive values (left eigenvector). Reproductive values can be thought of as describing the “value”
of an age class as a seed for population growth relative to that of the first (newborn or, in this case, fledgling) stage.
The reproductive value of the first age-class or stage is, by definition, 1.0. The peak reproductive value (second-year
females) is highlighted.
Age Class
Description
Reproductive value
1
Fledglings/first-year females
1.0
2
Second-year females
2.2
3
“Older adult” females
1.4
Table 2. Stable age distribution (right eigenvector). At the census, 63 percent of the individuals in the population
should be fledglings. An additional 17 percent will be yearlings (females beginning their second year). The rest will
be “older adult” females in their third year or older.
Stage
Description
Proportion
Mean age (± SD) Variant 1
1
Fledglings (to yearling)
0.63
0 ± 0
2
Second-year females
0.17
1 ± 0
3
“Older adult” females
0.20
2.7 ± 1.1
size of 10,000 distributed according to the SSD of
the deterministic model. Beginning at the SSD helps
to avoid the effects of transient, non-equilibrium
dynamics. The overall simulation consisted of 100 runs
(each with 2,000 cycles). We calculated the stochastic
growth rate, logλ
S
, according to Eqn. 14.61 of Caswell
(2001), after discarding the first 1,000 cycles in order to
further avoid transient dynamics.
The stochastic model (Table 4) produced two
major results. First, only high levels of stochastic
fluctuations had appreciable detrimental effects. Low-
level stochastic fluctuations (Case 4, SD of one eighth)
resulted in no extinctions and only three declines.
Second, varying the first-year transitions had the
greatest detrimental effects (Case 3, three extinctions
and 65 declines). The difference in the effects of which
arc was most important is predictable largely from the
elasticities. λ was most elastic to changes in the first-
year transitions. This detrimental effect of stochasticity
occurs despite the fact that the average vital rates remain
the same as under the deterministic model - the random
selections are from a symmetrical distribution. This
apparent paradox is due to the lognormal distribution of
stochastic ending population sizes (Caswell 2001). The
lognormal distribution has the property that the mean
28
29
exceeds the median, which exceeds the mode. Any
particular realization will therefore be most likely to end
at a population size considerably lower than the initial
population size. These results indicate that populations
of long-billed curlew are somewhat vulnerable to
stochastic fluctuations in first-year survival or fertility
(due, for example, to annual climatic variation or to
human disturbance) when the magnitude of fluctuations
is high. Nevertheless, the relatively even elasticity
values (Figure 11) in the life cycle of long-billed
curlews may, to some extent, help buffer them against
environmental stochasticity. Pfister (1998) showed
that for a wide range of empirical life histories, high
sensitivity or elasticity was negatively correlated with
high rates of temporal variation. That is, most species
appear to have responded to strong selection by having
low variability for sensitive transitions in their life
cycles. Long-billed curlews, however, may have little
flexibility in reducing variability in first-year transition
rates. Variable early survival, and perhaps fertility, is
likely to be the rule rather than the exception.
Potential refinements of the models
Clearly, improved data on survival rates and
age-specific fertilities are needed in order to increase
confidence in any demographic analysis. The most
important “missing data elements” in the life history for
long-billed curlew are for first-year transitions, which
emerge as vital rates to which λ is most sensitive as well
as most elastic. Data from natural populations on the
range of variability in the vital rates would allow more
realistic functions to model stochastic fluctuations. For
example, time series based on actual temporal or spatial
variability would allow construction of a series of
“stochastic” matrices that mirrored actual variation. One
advantage of such a series would be the incorporation of
observed correlations between variations in vital rates.
Using observed correlations would improve on our
“uncorrelated” assumption by incorporating forces that
we did not consider. Those forces may drive greater
positive or negative correlation among life history
traits. Other potential refinements include incorporating
density-dependent effects. At present, the data appear
insufficient to assess reasonable functions governing
density dependence.
Summary of major conclusions from matrix
projection models:
v
Survival accounts for 65 percent of the total
“possible” sensitivity, with first-year survival
as the most important (47 percent of total)
followed by first-year fertility (22 percent
of total). Any absolute changes in first-year
rates will have major impacts on population
dynamics.
v
First-year survival (e
21
= 30 percent) and
first-year fertility (e
11
= 19 percent) account
for almost 40 percent of the total elasticity.
Proportional changes in first-year transition
rates will have a major impact on population
dynamics.
v
The reproductive value of “older” females
is relatively low. Thus yearling females
appear to be the key reservoir of population
dynamics under the model formulated here.
v
Stochastic simulations echoed the elasticity
analyses in emphasizing the importance of
first-year survival and fertility to population
dynamics. In comparison to life histories of
other vertebrates, long-billed curlews appear
slightly less vulnerable to environmental
stochasticity (because of the buffering effect
of a relatively even importance of different
vital rates, as assessed by the sensitivities
and elasticities).
Community ecology
Predators and habitat use
Predator response to grazing or to fragmentation
of prairie habitats and how this might influence
reproductive success of long-billed curlews have not
been studied. Trees are not a historical element of the
mixed-grass and shortgrass prairie landscapes, and their
presence (e.g., plantings, treerows, windbreaks) may
result in increased predation by providing perches for
avian predators such as magpies, ravens, and raptors.
Parasites and disease
Aspergillosis killed 15 percent (two of 13) of
prefledglings during one season in Idaho (Redmond
and Jenni 1986). Three species of lice from curlews
were reported in Texas and New Mexico studies
(Cummingsiella longistricola, Lunaceps numenii
numenii, and
Austromenopon crocatum; Wilson
1937, Butler and Pfaffenberger 1981). Other records
of ectoparasites include a chigger (Toritrombicula
dupliseta; Loomis 1966) and a species of louse
(Cummingsiella ovalis; Malcomson 1960).