7
2
Inferring Threat from Scientific
Collections: Power Tests and an
Application to Western Australian
Acacia Species
Mark Burgman, Bruce R. Maslin, David Andrewartha,
Marie R. Keatley, Chris Boek, and Michael McCarthy
Introduction
The classification of threat by the World Conservation Union (IUCN 1994) is used
widely in conservation biology, particularly to assist in developing priorities for
management. Originally, the classification was based on subjective estimates and
qualitative classes of threat. Mace and Lande (1991) specified quantitative thresholds for risks and time horizons for categories of threat. These criteria were
extended in the IUCN (1994) revised criteria. They may be static (e.g., population
size fewer than 250 mature individuals), retrospective (e.g., an observed decline
of 50% in the past 10 years), or prospective (e.g., probability of extinction of 20%
within the next 20 years), and may include the absolute population size and
distribution, the size of the adult population, or changes in population size or
range. Quantitative criteria such as trends in range or abundance may be costly to
estimate because they involve collection of field information at repeated intervals
over time, parameter estimation, and predictive modeling.
When dealing with rare species, it may be necessary to make inferences about
the decline or extinction of a species on the basis of a handful of collections or
opportunistic field observations (Solow and Helser, this volume). If the abundance of a species declines or its range contracts through time, then this may be
represented by less frequent collections and relatively long periods during which
the species is not observed or collected. Often, these data will be the only information available to establish conservation priorities.
The first developments of formulas for inferring extinction from observation
data were proposed by Solow (1993; see Solow and Helser, this volume). Burgman et al. (1995) explored the use of some more general tests in these circumstances. McCarthy (1998) extended Solow’s test to include variation in the intensity of the observation process. Deviations from the assumptions of the tests will
occur if species ranges or abundances decline through time. The tests are designed
to detect such changes. However, it is difficult to know how sensitive the equations will be to changes in the underlying abundance of a species, even if all the
assumptions of the methods are met. The use of these statistical formulas in
2
Inferring Threat from Scientific
Collections: Power Tests and an
Application to Western Australian
Acacia Species
Mark Burgman, Bruce R. Maslin, David Andrewartha,
Marie R. Keatley, Chris Boek, and Michael McCarthy
Introduction
The classification of threat by the World Conservation Union (IUCN 1994) is used
widely in conservation biology, particularly to assist in developing priorities for
management. Originally, the classification was based on subjective estimates and
qualitative classes of threat. Mace and Lande (1991) specified quantitative thresholds for risks and time horizons for categories of threat. These criteria were
extended in the IUCN (1994) revised criteria. They may be static (e.g., population
size fewer than 250 mature individuals), retrospective (e.g., an observed decline
of 50% in the past 10 years), or prospective (e.g., probability of extinction of 20%
within the next 20 years), and may include the absolute population size and
distribution, the size of the adult population, or changes in population size or
range. Quantitative criteria such as trends in range or abundance may be costly to
estimate because they involve collection of field information at repeated intervals
over time, parameter estimation, and predictive modeling.
When dealing with rare species, it may be necessary to make inferences about
the decline or extinction of a species on the basis of a handful of collections or
opportunistic field observations (Solow and Helser, this volume). If the abundance of a species declines or its range contracts through time, then this may be
represented by less frequent collections and relatively long periods during which
the species is not observed or collected. Often, these data will be the only information available to establish conservation priorities.
The first developments of formulas for inferring extinction from observation
data were proposed by Solow (1993; see Solow and Helser, this volume). Burgman et al. (1995) explored the use of some more general tests in these circumstances. McCarthy (1998) extended Solow’s test to include variation in the intensity of the observation process. Deviations from the assumptions of the tests will
occur if species ranges or abundances decline through time. The tests are designed
to detect such changes. However, it is difficult to know how sensitive the equations will be to changes in the underlying abundance of a species, even if all the
assumptions of the methods are met. The use of these statistical formulas in