CRITERION A: HIGH DECLINE RATE
For criterion A, an estimate of current population size is compared with an estimate from the past or a projection for the future, and the change over the specified time period t is compared with threshold values for critically endangered, endangered, and vulnerable (Fig. 3). Population size is adjusted with the measure of “mature individuals” (IUCN 2001), which is specifically defined to reflect the size of the actual or potential breeding population. Because mature individuals of different species have very different average life spans (from hours to millennia), the period over which declines are measured is expressed in generation lengths. Generation length acts as a surrogate for turnover rates within populations. Long-lived species are at greater risk from increased annual adult mortality rates (measured as percentage of loss per year) than short-lived species because breeding adults experience this mortality over more years. Conversely, a long-lived species declining at the same rate as a short-lived one (measured as percentage of change per generation) shows smaller reductions over time (measured in years). The time window over which declines are measured is set to a minimum of 10 years because measuring changes over shorter time periods is difficult and does not reflect timescales for human interventions. The maximum projection into the future is 100 years, regardless of generation time, because of the uncertainties in predicting population sizes a long way into the future as might be necessary for long-generation species.
Figure 3. Different kinds of population decline used in criterion A. Each graph shows population size declining over time (black solid line) and the decline rate measured as number lost over the previous 10 years as a percentage of the starting number (gray solid line). The points in time marked−t and+t are the past and future points (dotted vertical lines), respectively, where assessment is made compared with the present (solid vertical line): (a) a constant number of individuals are lost in each time period; (b) a constant proportion are lost in each time period; (c) a declining proportion are lost in each time period; (d) an increasing number of individuals are lost in each time period.
Download figure to PowerPoint
The decline, measured as percentage of loss, can be estimated for the past (criterion A1 and A2), future (criterion A3), or a combination of the past and future (criterion A4). Because of difficulties in estimating population sizes in most natural populations, the criteria allow the assessor to use various kinds of direct and indirect evidence to estimate the decline rate, and the evidence is made explicit in the subcriteria. Criteria A1 and A2 listings may be derived from direct observation (population counts of some kind), which is obviously not feasible for future projections. For any of the criteria A1 through A4, declines may be determined on the basis of indices of abundance such as sightings or catch per unit effort. Assessment of rates of change in threatening processes may also be used for criterion A listings, on the basis of loss of habitat, levels of direct or indirect exploitation, and the effects of introduced taxa, hybridization, pathogens, pollutants, competitors, or parasites. Nevertheless, assessors need to use such indirect evidence cautiously. It is important to distinguish the decline in population size, with which we are concerned, from the underlying process. Harvesting, for example, is counteracted to some extent by density-dependent processes, such that the rate of decline is not linearly proportional to the number removed (Clark 1990). Similarly, a measured decline in habitat area cannot be straightforwardly translated to a decline in population size, especially if it involves losses of edges or lower-quality habitat (Lomolino & Channell 1995; Rodriguez 2002; Akçakaya et al. 2006).
Careful analysis of expected decline trajectories is needed because the shape of the decline-rate curve depends on the threat process involved (Fig. 3). Figure 3a shows a population declining by a constant amount each year such that the decline rate increases as the population becomes smaller. This might occur in a situation where interspecific competition, predation, or overexploitation leads to population reduction, but the reduction in population number is constant, perhaps related to the size of the predator or competitor population or the number of harvesting households. Here, past decline rates allow the species to qualify as vulnerable (decline >30%), but declines projected to continue on the same basis yield an endangered (>50%) categorization. If declines persist, this population will soon qualify for critically endangered (>80% decline) before going extinct.
Figure 3d shows the same process, but the depletion in population numbers is increasing over time. This situation is not unlikely, especially under habitat fragmentation and for species that provide consumer goods of high economic or social value where the value increases as the product becomes rarer or consumer tastes increase demand. Increased decline rates of smaller populations may be expected where inverse density dependence (known as Allee effects or depensation) is operating (Myers et al. 1995; Courchamp et al. 1999; Courchamp et al. 2006). In this situation, the decline rate increases even more quickly with time; thus, in a very short time, the species moves from nonthreatened through the categories vulnerable, endangered, and critically endangered until it goes extinct. In such cases, as portrayed in Figs. 3a & 3d, it may be appropriate to derive a higher threat category by forward projection with criterion A3, rather than on the basis of current rates.
The trajectory in Fig. 3c shows a case where the population numbers decline over time, but the decline rate is decreasing. Given enough time, the population can stabilize and may even recover. The decline rate progressively decreases so that the species that originally qualified as critically endangered moves through the categories endangered and vulnerable until eventually it is not threatened. This pattern is expected under a variety of situations. For example, it could be the outcome of managed harvesting programs. Here managers may seek to reduce the population size until it reaches the density at which productivity is increased or maximized, and the harvest is then stabilized at a sustainable level, at which point there should be no further decline in population. Because the IUCN rules require listing under the category of highest threat, such species are listed according to past, not future, projected rates. Nevertheless, when there is confidence that the trend is declining, criterion A1 is used, rather than A2, and its more exclusive thresholds for listing allow the reduced extinction risk to be reflected (see later).
In Fig. 3b the decline rate is constant and the change in population size is reducing over time, perhaps as a function of the interaction between population change and density-dependent processes, for example, when the effects of exploitation, predation, or competition are reduced with the abundance of the species. The population in Fig. 3b always qualifies as vulnerable and will never qualify for any higher threat category under criterion A until it goes extinct. In practice, a species showing this pattern qualifies for higher threat categories under criteria B, C, and D once the population size or the geographic range reaches sufficiently low levels to meet the thresholds in these categories.
Criteria A1 and A2 differ according to whether or not the causes of decline are reversible, understood, and have ceased. Criterion A1 can only be used if all these conditions are met (IUCN 2006). If they are, the threshold decline rate is higher than in A2. Criterion A1 is likely to be used for species following trajectory 3c, where they are under good management and should have lower extinction risks than species with the same current decline rate but where future decline rates cannot be expected to be reduced.
The criteria do not specify how information on temporal changes in population size should be used to calculate a past decline rate or project a future one. It may be appropriate to use some statistical method to calculate the decline rate, such as fitting a least-squares regression line and estimating the decline rate from the slope of the line. Nevertheless, it may often be inappropriate and impractical to do this. If populations show nonlinear trends within the 3-generation assessment period, such as when an increase in population is followed by a decrease, fitting such a regression line could be misleading. Also, for many species there are no systematic data on population size, and the assessor may need to make a determination of trends on the basis of extremely limited information. Often, the best that can be done is to use the estimated number at the beginning and end of the 3-generation census period. The potential problems in this approach are self-evident, but it is less obvious that even with apparently good information, it may be difficult to make a robust estimate of population trends. Accurate measurements of changes in population size depend critically on the quantity and quality of available data (Taylor 1995). Over limited time spans or with small numbers of surveys, it is possible either to fail to detect a real decline (Type II error) or to detect a decline when actually there is none (Type I error). Although statistical techniques such as power analysis can be used to support assessments (Taylor & Gerrodette 1993), they do not solve the problem if the situation is both extremely uncertain and potentially serious (Colyvan et al. 1999). In situations of uncertainty, the assessor should use the best available information and combine formal data analysis with expert judgment (Colyvan et al. 1999), for which methods are now available (Akçakaya et al. 2000).
CRITERION B: SMALL RANGE AREA AND DECLINE
Criterion B allows a species to qualify as threatened when its geographical range is very restricted and when other factors suggest that it is at risk. In some situations, population size may not be measurable or relevant to an elevated extinction risk, for example, when species are restricted to small areas or to habitat remnants that are themselves disappearing. Although this criterion was originally developed for plants, the drafting group considered this criterion applicable to other species, especially those at high densities within restricted areas or habitats.
This criterion does not simply use range area as a surrogate for population size. Although there is a very broad positive correlation within and across species between total geographic range size and population numbers, there is much variation and the details can alter according to the spatial scale at which the species is assessed (Gaston 1994a; Gaston et al. 2000; Blackburn et al. 2006). In some cases, species may qualify under population size and range size criteria, but more often the 2 measures will operate somewhat independently. Many species that qualify as threatened under criterion B cannot qualify on the basis of population size. Conversely, some species (e.g., many marine mammals) cannot qualify under criterion B however close they are to extinction because the ranging patterns of individuals exceed the critical thresholds.
The measurement of range area is complicated (Gaston 1991; Gaston 1994a, 1994b, 1994c, 2003; Maurer 1994). The criteria consider 2 quantities, extent of occurrence (EOO) and area of occupancy (AOO) (sensu Gaston 1991). Extent of occurrence is defined as the area contained within the shortest continuous boundary that can be drawn to encompass all the known, inferred, or projected sites of occurrence of a species. This measure could be strongly influenced by cases of vagrancy and by marked discontinuities or disjunctions within the overall distribution of a species, both of which should be excluded. What constitutes a discontinuity or disjunction has deliberately been left vague, but of particular concern here are extents of occurrence composed of broad environments that are totally unsuitable for the species to occupy or often even to disperse into. For example, it would be inappropriate to include intervening areas of ocean when estimating EOO for a forest-dwelling species occurring at sites on 2 continents. The IUCN guidelines provide additional details on estimating EOO (IUCN 2006).
Area of occupancy quantifies the area within the EOO where the species is found. Species are hardly ever continuously distributed throughout their EOO. As applied in the criteria, AOO is the smallest area essential at any stage to the survival of existing populations of a species (e.g., colonial nesting sites, feeding sites for migratory species). The size of the AOO for a species inevitably depends on the spatial scale at which it is measured: the finer the resolution, the smaller the resultant area (Gaston 1991). There has been much debate over how this issue can best be resolved (Keith et al. 2000; Hartley & Kunin 2003).
Although no scale of measurement is specified in the criteria, the rules state that the scale should be appropriate to relevant biological aspects of the species and should be measured on a grid (or equivalents). The guidelines give more specific advice on avoiding problems of scale when using AOO (IUCN 2006). In general, spatial scales used for measuring ranges should reflect the movement and dispersal patterns of the species in question, and exceedingly fine or very coarse resolutions will lead to inappropriate listings under criterion B.
The measurement of EOO and AOO has been thought difficult for species with linear ranges (e.g., intertidal, stream, and riverine species). These range areas tend to be very small because one dimension (e.g., the width of the intertidal zone or the river) is so limited. In fact, species that depend on linear habitats are particularly vulnerable because a threat can rapidly affect an entire area (e.g., a single upstream pollution event may easily affect a whole river downstream). On balance, therefore, areas of linear ranges are thought to provide a fair reflection of risk.
Unlike population decline rates and population sizes, there is no strong theoretical framework to associate given range areas (which may contain hugely different numbers of individuals) with different levels of risk of extinction. Therefore, although a range-area-based criterion was regarded as essential to the listing of many groups of organisms (for which population data are either not available or not of foremost importance in determining extinction risk), the choice of critical thresholds for criterion B has been plagued with difficulties from methodological and biological standpoints. The final decisions were made largely on an iterative basis of trial and error, and empirical testing by SSC experts using data on a variety of relevant species. This resulted in the maintenance of a constant ratio of cut-off values for EOO and AOO (a difference of a factor of 10) in each of the categories critically endangered, endangered, and vulnerable and cut-offs, respectively, of 100 km2, 5,000 km2, and 20,000 km2. All these areas, for EOO and AOO, are comparatively small, reflecting that for this criterion, risk of extinction is associated with range area itself.
Unless extremely small (see criterion D), limited range size is not sufficient on its own for a species to qualify as threatened. Many species have persisted successfully for long periods within small global ranges and have a low risk of extinction (Gaston 1994a; Gaston 2003). To qualify under criterion B, therefore, a species must also exhibit at least 2 of 3 other symptoms of risk. To avoid overlisting, the conditions were made difficult to meet. There must be some evidence the population is or is projected to be in continuing decline, severely fragmented, limited to a few locations, or subject to extreme fluctuations. Empirical and theoretical studies suggest that all these conditions will increase the likelihood of extinction.
Commentary on criterion B suggests that it may be overly inclusive, with the threshold values set so high that a large number of species are inappropriately listed as threatened (Keith 1998). In fact, for certain small natural areas, such as oceanic islands, where the total area under analysis is small, there is little habitat heterogeneity, and threats are pervasive, all endemic species may justifiably qualify as threatened. Nevertheless, species cannot be listed as threatened solely on the basis of small range areas, so the number of such cases is limited. More often the area under assessment is small because it is a politically defined subunit within a wider area, in which case the assessment should include the status of the species outside the area (IUCN 2003). Similarly, it has been suggested that the different criteria should give similar threat assessments across species and that the numbers listed in the categories of threat should be evenly spread (Keith 1998). Nevertheless, we see no reason a priori why either of these should follow because the criteria are intended to operate independently of one another and threats are expected to vary between species and habitats.
CRITERION C: SMALL POPULATION SIZE AND DECLINE
Criterion C focuses on populations that are numerically small and in continuing decline and is the most straightforward of all to place in a theoretical framework. The choice of threshold sizes for the number of mature individuals is derived from theoretical values for minimum viable populations (see above) adjusted to reflect timescales appropriate for the species. The initial condition is that the population must number fewer than 10,000 mature individuals (for vulnerable), 2,500 mature individuals (for endangered), and 250 individuals (for critically endangered). The steep ramping down of critical population sizes reflects what is known from theoretical studies about the general relationships between population size and time to extinction under various kinds of environmental and demographic stochasticity (Lande 1993, 1998).
A population in continuing decline may immediately qualify if the population size meets the threshold values above and the population is declining sufficiently fast (criterion C1). If a decline is known or expected, but is not measurable or sufficiently severe to meet the C1 threshold, the species may qualify instead under C2. For this to happen, its population must exist entirely or almost as a single unit, have relatively small subpopulations, or experience extreme fluctuations in size. Species cannot qualify for criterion C simply by meeting the population size threshold and being in decline. The additional conditions are more difficult to meet in criterion B than in criterion C because there is direct evidence in C that the population size is already small, which is not necessarily the case in B. Therefore, although criteria B and C are comparable, the difference between range areas and population sizes as entry points to the criteria mean that the subcriteria and conditions should not be the same in each (Keith 1998).
CRITERION D: VERY SMALL POPULATION SIZE
Criterion D allows species to be listed as threatened without evidence that there has been, is, or will be a decline of some sort. It was developed because theoretical models show that numerically small populations can have relatively high extinction risks solely from internal processes. The term demographic stochasticity has been used to describe the process whereby random variation among individuals in demographic vital rates or random variation in sex ratio alone can lead to population extinction (Goodman 1987; Lande 1993), the importance of which is supported empirically by a number of studies on very restricted populations (Kokko & Ebenhard 1996; Legendre et al. 1999). Although demographic stochasticity is generally unimportant for populations with effective population sizes over about 100 individuals, its deleterious effects are amplified by life history and behavioral differences among species (Sorci et al. 1998; Legendre et al. 1999). Hence, the threshold numbers used in the criteria are larger. For vulnerable, this means any species with fewer than 1000 mature individuals can qualify. The equivalent figures for endangered and critically endangered are 250 and 50. The scaling of these values reflects the relationship between population size and extinction time (Fig. 1).
Criterion D has a subcriterion D2 that is present only in the vulnerable category. Subcriterion D2 allows species to qualify solely on the basis of a very restricted distribution (i.e., it is the range-area equivalent of D1). Subcriterion D2 is conceptually distinct, however, because it is implicit in its definition that it is not restricted range alone that should be used to list species under this category. Rather, it is evidence that the species is actually threatened because of its very restricted distribution. The D2 subcriterion has sometimes been misused, mainly through applying the numerical thresholds mentioned in the first part of the definition without reference to the second part. Summary tables of the criteria, increasingly used by assessors instead of the full text, tend only to include the numerical guidelines, and this may have increased the extent of misinterpretation.
Subcriterion D2 does not extend into the higher-risk categories because the justifications for listing are even more problematic at higher levels of risk. Although D2 is justified under the precautionary principle at the relatively low levels of risk embraced by vulnerable, this is not so for endangered and critically endangered. Some users believe D2 should be extended to allow listings higher than vulnerable for extremely restricted species (Seddon 1998), whereas others find D2 overly inclusive and are critical that it apparently fails to recognize that for many species rarity is a natural state and only certain kinds of rare species are actually liable to go extinct (de Lange & Norton 1998). During the criteria review, the conditions for D2 were tightened to avoid overlisting, but it remains among the most inconsistently applied elements of the IUCN criteria.
CRITERION E: UNFAVORABLE QUANTITATIVE ANALYSIS
Criterion E allows the assessor to use any kind of quantitative analysis for assessing the risk of extinction, which is then compared with the extinction-risk thresholds given for each of the categories. These quantitative thresholds are expressed as the probability of extinction within a given time frame. The time frame is measured in years or generations as in the formulation of criterion A, with whichever of the 2 is longer. Justifications for the thresholds are essentially the same as in Mace and Lande (1991), except that the time frame for critically endangered has changed from 5 to 10 years, for consistency with the other criteria, and the future time frame is capped to 100 years as in criterion A.
The term quantitative analysis was chosen carefully to avoid the impression that this criterion necessarily involves a population viability analysis (PVA). Criterion E can be used in any case where a robust estimate of extinction risk can be derived. Often this might be done without detailed information on population dynamics and is derived from information on the status of the habitat. For example, consider the situation in which a species is endemic to an area and is forest dependent and forestry rights have been sold to allow the entire area within which this species lives to be cleared within 20 years. Such a species would certainly qualify as endangered and even possibly as critically endangered because there is at least a 50% chance that the critical habitat areas will be cleared in the first 10 years. Many similar cases in which criterion E can be used involve land-use changes and expected levels of exploitation. It can also be used if there is a high risk of invasion by a species whose presence would be disastrous for the resident species.
More commonly, however, a PVA would be involved in the assessment. The rules dictate that the structure of the model and the data used in the analysis be made explicit, and standards have been developed (IUCN 2001, 2006). There are several potential difficulties with widespread use of PVA modeling in red list assessments. First, despite the requirements that the assumptions be made explicit, it is in practice difficult to list and justify the background to a PVA analysis without lengthy documentation. Listings under criterion E would thus require longer justification than listings under other criteria.
Second, PVA outcomes can be very sensitive to the levels of some input variables. For example, expected changes in habitat availability, the incidence and severity of catastrophes, levels of mortality, and the interaction between population size and inbreeding depression might each determine the extinction risk category on their own when set to plausible, although improbable, values in a PVA model. It will be hard for IUCN to monitor and guarantee standards when accuracy depends on validating many such sensitive variables (Mangel & Tier 1994; Ludwig 1996, 1999).
Conversely, PVA models may not be precautionary in the absence of good information if they assume favorable values for key parameters (e.g., Armbruster et al. 1999). Although carefully constructed PVAs built on reliable data can apparently predict risks of decline accurately (Brook et al. 2000), many practitioners suggest that PVA is best used as a way of assessing the relative risks of different processes or the relative benefits of different management strategies, but not the absolute risk of extinction (Akçakaya & Burgman 1995; Beissinger & Westphal 1998). We concur with this view and recommend use of criterion E for simple and explicit modeling on the basis of reliable and sufficient data, rather than on the basis of outcome of detailed models with uncertain parameters and a large number of assumptions.