180
Lloyd Goldwasser, Scott Ferson, and Lev Ginzburg
However, we also made a few modeling choices that may be considered optimistic. We assumed that there will be no further habitat loss due to human
impacts from logging or other intrusions. This assumption may be unrealistic in
the face of the continuing ecological pressure exerted by the human population,
but investigating the consequences of general anthropogenic habitat destruction
and other land-use effects is beyond the scope of the present study. We also
assumed that there will be no loss of breeding birds to emigration out of the
Olympic Peninsula population. Finally, we assumed that Allee effects do not
become important until the population falls to 15 or fewer breeding pairs. This
assumption could be relaxed by using an appropriately larger quasi-extinction
level or perhaps by modeling Allee effects directly.
Some phenomena that have been included in the modeling exercise turned out
not to have large effects on the demography of the owls. In particular, we observed
that rare catastrophic windstorms probably affect owl extinction risk very little,
unless they are of a magnitude or frequency much larger than has been seen in the
historical record. Likewise, the increase in available breeding territories as a result
of the maturation of young forests did not have a very large impact on extinction
risk either. It did, however, introduce the potential that some trajectories actually
increase over the century-long simulation rather than every single trajectory inexorably declining.
One of the most important assumptions made about the model was that measurement error is negligible. Although this assumption is commonly made, it
represents the untenable belief that the numeric values used in the model are
known precisely. In fact, because they are estimated from empirical studies, their
measurement errors may be considerable. The next section considers this issue
explicitly.
Measurement Error
Most population viability analyses estimate the risk of extinction by computing
the probability of extinction under random environmental variability. This approach is sufficient only if the demographic rates and the magnitude of their
year-to-year fluctuations are known exactly. However, such knowledge is rarely
available in empirical situations, and even in fairly unchanging environments
there can be considerable uncertainty associated with the estimate of each
demographic rate. Indeed, Burnham and associates (1994; Burnham, personal
communciation) found the variance in adult fecundity due to measurement error
to be an order of magnitude greater than that due to year-to-year fluctuations. The
effect of such uncertainty is to blur any projected population trajectory, even when
the environmental variability is known completely. This blurring increases with
time, rendering long-term projections particularly susceptible to uncertainties in
the original estimates. More precise estimates of the demographic rates would
result in narrower bundles of trajectories and a longer horizon for making meaningful population projections.
Lloyd Goldwasser, Scott Ferson, and Lev Ginzburg
However, we also made a few modeling choices that may be considered optimistic. We assumed that there will be no further habitat loss due to human
impacts from logging or other intrusions. This assumption may be unrealistic in
the face of the continuing ecological pressure exerted by the human population,
but investigating the consequences of general anthropogenic habitat destruction
and other land-use effects is beyond the scope of the present study. We also
assumed that there will be no loss of breeding birds to emigration out of the
Olympic Peninsula population. Finally, we assumed that Allee effects do not
become important until the population falls to 15 or fewer breeding pairs. This
assumption could be relaxed by using an appropriately larger quasi-extinction
level or perhaps by modeling Allee effects directly.
Some phenomena that have been included in the modeling exercise turned out
not to have large effects on the demography of the owls. In particular, we observed
that rare catastrophic windstorms probably affect owl extinction risk very little,
unless they are of a magnitude or frequency much larger than has been seen in the
historical record. Likewise, the increase in available breeding territories as a result
of the maturation of young forests did not have a very large impact on extinction
risk either. It did, however, introduce the potential that some trajectories actually
increase over the century-long simulation rather than every single trajectory inexorably declining.
One of the most important assumptions made about the model was that measurement error is negligible. Although this assumption is commonly made, it
represents the untenable belief that the numeric values used in the model are
known precisely. In fact, because they are estimated from empirical studies, their
measurement errors may be considerable. The next section considers this issue
explicitly.
Measurement Error
Most population viability analyses estimate the risk of extinction by computing
the probability of extinction under random environmental variability. This approach is sufficient only if the demographic rates and the magnitude of their
year-to-year fluctuations are known exactly. However, such knowledge is rarely
available in empirical situations, and even in fairly unchanging environments
there can be considerable uncertainty associated with the estimate of each
demographic rate. Indeed, Burnham and associates (1994; Burnham, personal
communciation) found the variance in adult fecundity due to measurement error
to be an order of magnitude greater than that due to year-to-year fluctuations. The
effect of such uncertainty is to blur any projected population trajectory, even when
the environmental variability is known completely. This blurring increases with
time, rendering long-term projections particularly susceptible to uncertainties in
the original estimates. More precise estimates of the demographic rates would
result in narrower bundles of trajectories and a longer horizon for making meaningful population projections.