1
1
Detecting Extinction in Sighting Data
Andrew Solow and Thomas Helser
Introduction
Extinction is probably the most dramatic, if not necessarily the most important,
form of change in a biological community. There is currently great concern that
human activities are directly or indirectly contributing to increased extinction
risks for a wide variety of plant and animal species (Diamond 1989). Extinctions
are rarely observed directly, and there is considerable uncertainty not only about
the overall rate of extinction (Smith et al. 1993) but also about the extinction of
individual species whose existence is known only through chance sightings.
This chapter summarizes some recent work on statistical methods for detecting
extinction from sighting data (Solow 1993a,b). The basic idea underlying these
methods is that it should be possible to base inference about the extinction of a
species on the time since the most recent sighting. Loosely speaking, the question
addressed in this work is, how long must a species go unsighted before it is
reasonable to conclude that it is extinct? The answer to this question depends on a
number of factors. One important factor is the way in which the observation effort
varies through time. For example, even a short period since the most recent
sighting may be significant evidence of extinction if the recent observation effort
has been high. Conversely, even a long period since the most recent sighting may
not provide significant evidence of extinction if the observation effort has been nil
or inefficient. Although it is possible to incorporate varying observation effort into
the methods discussed in this chapter, we assume that it is approximately constant
over the observation period. This would be the case, for example, for a fixed
observation program replicated over time. It would also be the case if sightings were purely accidental, such as opportunistic collections in museums and
herbariums.
A second important factor influencing the significance of the time since the
most recent sighting is variation in the population size of the species. We consider
two cases. In the first, the pre-extinction population is assumed to remain approximately constant for the period during which it was observed. This model would be
appropriate for a chronically small population subject to relatively rapid extinction due, for example, to the loss of critical habitat. Incidentally, the same model
1
Detecting Extinction in Sighting Data
Andrew Solow and Thomas Helser
Introduction
Extinction is probably the most dramatic, if not necessarily the most important,
form of change in a biological community. There is currently great concern that
human activities are directly or indirectly contributing to increased extinction
risks for a wide variety of plant and animal species (Diamond 1989). Extinctions
are rarely observed directly, and there is considerable uncertainty not only about
the overall rate of extinction (Smith et al. 1993) but also about the extinction of
individual species whose existence is known only through chance sightings.
This chapter summarizes some recent work on statistical methods for detecting
extinction from sighting data (Solow 1993a,b). The basic idea underlying these
methods is that it should be possible to base inference about the extinction of a
species on the time since the most recent sighting. Loosely speaking, the question
addressed in this work is, how long must a species go unsighted before it is
reasonable to conclude that it is extinct? The answer to this question depends on a
number of factors. One important factor is the way in which the observation effort
varies through time. For example, even a short period since the most recent
sighting may be significant evidence of extinction if the recent observation effort
has been high. Conversely, even a long period since the most recent sighting may
not provide significant evidence of extinction if the observation effort has been nil
or inefficient. Although it is possible to incorporate varying observation effort into
the methods discussed in this chapter, we assume that it is approximately constant
over the observation period. This would be the case, for example, for a fixed
observation program replicated over time. It would also be the case if sightings were purely accidental, such as opportunistic collections in museums and
herbariums.
A second important factor influencing the significance of the time since the
most recent sighting is variation in the population size of the species. We consider
two cases. In the first, the pre-extinction population is assumed to remain approximately constant for the period during which it was observed. This model would be
appropriate for a chronically small population subject to relatively rapid extinction due, for example, to the loss of critical habitat. Incidentally, the same model