Ken Norris, School of Animal and Microbial Sciences, University of Reading, Whiteknights, PO Box 228, Reading RG6 6AJ, UK (e-mail email@example.com).
1Demographic models are assuming an important role in management decisions for endangered species. Elasticity analysis and scope for management analysis are two such applications. Elasticity analysis determines the vital rates that have the greatest impact on population growth. Scope for management analysis examines the effects that feasible management might have on vital rates and population growth. Both methods target management in an attempt to maximize population growth.
2The Seychelles magpie robin Copsychus sechellarum is a critically endangered island endemic, the population of which underwent significant growth in the early 1990s following the implementation of a recovery programme. We examined how the formal use of elasticity and scope for management analyses might have shaped management in the recovery programme, and assessed their effectiveness by comparison with the actual population growth achieved.
3The magpie robin population doubled from about 25 birds in 1990 to more than 50 by 1995. A simple two-stage demographic model showed that this growth was driven primarily by a significant increase in the annual survival probability of first-year birds and an increase in the birth rate. Neither the annual survival probability of adults nor the probability of a female breeding at age 1 changed significantly over time.
4Elasticity analysis showed that the annual survival probability of adults had the greatest impact on population growth. There was some scope to use management to increase survival, but because survival rates were already high (> 0·9) this had a negligible effect on population growth. Scope for management analysis showed that significant population growth could have been achieved by targeting management measures at the birth rate and survival probability of first-year birds, although predicted growth rates were lower than those achieved by the recovery programme when all management measures were in place (i.e. 1992–95).
5Synthesis and applications. We argue that scope for management analysis can provide a useful basis for management but will inevitably be limited to some extent by a lack of data, as our study shows. This means that identifying perceived ecological problems and designing management to alleviate them must be an important component of endangered species management. The corollary of this is that it will not be possible or wise to consider only management options for which there is a demonstrable ecological benefit. Given these constraints, we see little role for elasticity analysis because, when data are available, a scope for management analysis will always be of greater practical value and, when data are lacking, precautionary management demands that as many perceived ecological problems as possible are tackled.
Demographic models are increasingly used in the management of endangered species (for a recent review see Beissinger & Westphal 1998). Models are applied to populations of such species in a variety of ways: the management of particular vital rates (Crowse, Crowder & Caswell 1987), the direct management of abundance by, for example, reintroduction or translocation (Green, Pienkowski & Love 1996) or management of the environment (Beissinger 1995; Root 1998). However, in each of these cases demographic models are used to predict how management might affect population growth or viability (i.e. extinction or quasi-extinction risk) and, as such, provide an ecological basis for decision-making.
Over recent years there has been considerable debate concerning the reliability of elasticity analysis and the role of potential alternatives (Mills, Doak & Wisdom 1999, 2001; Caswell 2000; Wisdom, Mills & Doak 2000; Ehrlén, van Groenendael & de Kroon 2001). On a more practical level, the results of an elasticity analysis only have value if management can manipulate the key vital rates identified (Caswell 2000; Ehrlén, van Groenendael & de Kroon 2001). Furthermore, certain authors have argued that it would be unwise to use elasticity analysis in isolation to target management because particular management options might be able to alter vital rates that have relatively small elasticities in a way that causes a greater increase in population growth than that achievable by targeting management solely towards vital rates with relatively large elasticities (Ehrlén, van Groenendael & de Kroon 2001). For example, Hiraldo et al. (1996) constructed a demographic model to examine the long-term viability of a lesser kestrel Falco naumanni (L.) population in southern Spain. Elasticity analysis showed that population growth was most sensitive to changes in the survival rate of adults, followed by juvenile survival. However, they also showed that population viability was substantially improved by plausible changes in the birth rate caused by management rather than by plausible changes in survival. This apparent inconsistency arose because there was only limited scope for increasing survival rates using management, whereas management measures had the potential to cause a much more dramatic increase in the birth rate (Hiraldo et al. 1996).
One way to examine the potential value of prospective analyses, such as elasticity analysis and an assessment of the potential scope for management to influence population growth, is to consider how they may have influenced decisions taken in the past for particular endangered species. Comparisons of their predicted outcome with population growth actually achieved by management would be instructive. We used data from the critically endangered Seychelles magpie robin Copsychus sechellarum (L.), which responded favourably to a recovery programme that started in the early 1990s, to examine the role prospective analyses might have played had they been formally used to design management that was to be included during the subsequent recovery programme. To do this, we addressed several specific questions. (i) How did demography, population growth and abundance respond to the recovery programme? (ii) Prior to the recovery programme, how might an elasticity analysis have influenced management decisions? (iii) Prior to the recovery programme, what data were available to assess the scope for management to alter particular vital rates and increase population growth, and how might these have affected management decisions?
Seychelles magpie robin
The Seychelles magpie robin is a geographically isolated and morphologically distinct member of a predominantly Asian genus of chats that includes the shamas and Indian magpie robin Copsychus saularis (L.). It is perhaps the most critically endangered of the 12 bird species endemic to the Seychelles archipelago in the Indian Ocean and one of the rarest birds in the world. At the onset of continuous human settlement of the Seychelles archipelago in 1770 the magpie robin occurred on at least seven of the central, granitic, islands (Collar & Stuart 1985). It is likely that the species inhabited all the granitic islands with tall forest but early records are scarce.
Having evolved in an environment free from terrestrial mammals, the magpie robin was extremely vulnerable to the cats and rats that arrived with the early human colonists. The introduction of mammalian predators and the clearance of forest for plantation agriculture led rapidly to the species’ extinction on several islands. By 1880 the magpie robin had disappeared from the largest island, Mahé. The species survived into the early years of the twentieth century on Marianne, Aride and Frégate, as did an introduced population established around 1890 on the coralline island of Alphonse (McCulloch 1994). During the following 50 years the Marianne, Aride and Alphonse populations all succumbed to predators and habitat destruction. This left only the birds on Frégate Island (210 ha). Magpie robin numbers on Frégate were dramatically reduced following the introduction of cats to the island during the 1950s, and the total population may have comprised as few as eight individuals in 1965 (Penny 1968). Cats were eradicated from Frégate by 1981 but habitat degradation restricted the species’ subsequent recovery (Watson et al. 1992).
The original habitat of the magpie robin appears to have been mature coastal forest where the closed canopy maintained an open, unvegetated leaf litter. Magpie robins are strongly territorial as adults, occupying an area of 2–12 ha within which feeding and nesting activities occur (McCulloch 1996). The birds forage on the ground, in open areas of habitat, where they consume a range of invertebrates and small reptiles. The species has a low natural reproductive rate, producing only a single egg at each nesting attempt. Nesting frequency and the success of nesting attempts is influenced by territory quality, birds nesting more frequently and successfully on territories with high quality food supplies (Komdeur 1996).
BirdLife International has supported research on the magpie robin since 1988, when only 23 individuals survived, and from September 1990 has operated a recovery programme for the species (Rands & Komdeur 1989). A range of management measures were implemented on Frégate during the recovery programme (Table 1), designed to address real or perceived ecological problems (McCulloch 1996). Studies by Komdeur (1996) prior to the recovery programme had shown that food supplies available on territories could limit the birth rate, so supplementary feeding and habitat management were instituted to try to improve feeding conditions. Magpie robins nest in tree holes, but these were lacking in the predominantly exotic woodland on Frégate, so nest boxes were provided. Finally, concerns about the use of pesticides to control household insect pests were raised because fledged birds seemed more likely to disappear from territories inhabited by humans. As a result, a ban on the use of organophosphate and carbamide pesticides was implemented. Collection of data on magpie robin ecology, demography and abundance has been carried out continuously since 1988. Following establishment of the recovery programme, the abundance of magpie robins increased (Fig. 1). These data provide an opportunity to examine how demography changed over this time period, and give an interesting insight into how prospective analyses might have influenced management decisions taken when the recovery programme was established.
Table 1. Conservation management measures and their timing within the magpie robin recovery programme
June 1988–August 1990
Short-term experimental provision of food to specific territories (for details see Komdeur 1996)
September 1990–September 1992
Provision of food to territories in which breeding activity recorded and to poor quality (in terms of food availability) territories when monthly rainfall fell below 140 mm. Food supplement consisted of soil invertebrates when available, but largely consisted of boiled rice, coconut and lentils
October 1992 onwards
Daily feeding of all territories using cockroaches and small amounts of grated cheese. Food supplement corresponded to c. 20% of daily food demands of an individual bird
Provision of nest boxes Habitat management
April 1991 November 1991 onwards
Nest boxes provided throughout the island Scrub and undergrowth cleared from woodland areas within territories to improve access to leaf litter and soil. Limited programme of replanting with native tree species
Ban on the use of household insecticides
August 1991 onwards
Organophosphate and carbamide insecticides banned
All magpie robins have been ringed with individually unique colour combinations since 1988. New birds were usually ringed while in the nest, where this was accessible, or soon after fledging. The species is strongly territorial and territories are defended by groups of birds consisting usually, but not exclusively, of pairs and their offspring. Only a very small proportion of the population is ever found outside territories. Such individuals are usually wandering subadults or adults that have been ejected from their former territories. All territories were visited daily and all birds seen were recorded and identified from their coloured rings. Extra-territorial birds occasionally spent extensive periods in poorly accessible areas, where they might remain undetected for some time. Magpie robins are poorly adapted for long-distance flight and there is no evidence that they ever attempt to leave the island. Few remains of dead birds were found. Missing individuals were therefore assumed to be dead if they remained undetected for a maximum of 3 months. Death was assumed to have occurred in the week following the last sighting.
Most nests were discovered at the building stage during daily territory visits. All females without dependant offspring and not known to be nesting were followed for 45 min every 10 days to check for any breeding activity. Incubating birds would invariably return to the nest within this time. Once a nest was found, its continued activity would be checked daily from a safe distance. The distance at which an observer would be tolerated varied greatly among individual females. Females’ incubation regimes would be monitored during 1-h observations every third day. Similar observations were carried out after hatching to record the rate of food provision to the chick. A small number of nests where the adults were particularly confiding were occasionally subject to constant observation throughout the daylight hours. All accessible failed nests were investigated as soon as possible after cessation of activity to determine the cause of failure wherever possible. During the period of dependence after fledging, daily checks were made on the fate of the juvenile, presence being confirmed by sight, calls and parental behaviour.
We analysed demographic data collected immediately prior to (i.e. 1988 and 1989) and during the first 5 years (up to the end of 1995) of the recovery programme. We only used data collected prior to 1996 in our analyses because large-scale redevelopment work started on the island in late 1995 that impacted on the demography and population growth of the magpie robins (BirdLife Seychelles, unpublished data).
We first developed a simple two-stage matrix model in ramas stage (Ferson 1990) of the form:
where Y = the number of yearling females, A = the number of adult females (≥ 2 years), b = the mean number of female offspring fledged per paired female per year, c0 = the probability of becoming a paired female and breeding as a yearling (as not all females breed at age 1), s0 = the annual survival probability of first-year bird (i.e. in the first year post-fledging), s = the adult annual survival probability and t = time (in years). This model then formed the basis for all of our subsequent analyses, which were divided into three parts.
First, we analysed how the different parameters in the two-stage matrix model changed over the recovery programme, i.e. between 1988 and 1995, and evaluated their contribution to the observed population growth (Fig. 1). To estimate annual survival probabilities and to test for significant temporal increases in survival, we used log-linear hazard models (Crawley 1993). We used this approach because each bird is individually colour marked and resighted regularly (i.e. often on a daily basis), so we had relatively accurate estimates of the time of death for different individuals. For each individual and for each year it was alive, we constructed a censoring variable that described whether or not the bird died within a particular year (i.e. 1 if the bird died within a particular year, 0 if it survived). Next, we calculated the number of months each individual was known to have survived in each year. We then fit log-linear hazard models to these data in glim4 (Crawley 1993) by declaring the censoring variable as the response variable, using the log of time to death (months) as an offset and assuming a Poisson error distribution. We used this framework to test the hypothesis that survival had increased significantly over time by including year in the models as a continuous predictor variable. Separate analyses were conducted for the survival parameters s0 and s in the matrix model. Hazard (h) is related to survival (S) by the function S = exp(– h), so we have visually displayed the results of the log-linear hazard models as estimated survival rates. To examine time-series trends in the probability of a female becoming paired and breeding as a yearling (c0), we used COX regression analysis to examine time-series trends in the length of the pre-breeding period (weeks) for all females in the study whose birth year was accurately known (Armitage & Berry 1987). Females who died prior to breeding were included in this analysis as censored observations. Finally, the number of female offspring fledged per paired female per year was calculated by dividing the total number of offspring fledged each year by the total number of paired females present in the corresponding year. We multiplied the resultant value by 0·5 to estimate the number of female offspring (assuming a 50 : 50 sex ratio at fledging). Linear regression was used to examine time-series trends in the birth rate, by defining the number of offspring fledged female−1 year−1 as the response variable.
Then, we used the two-stage matrix model described above to investigate the relative contribution that the time-series trends in each particular vital rate made to the population growth observed during the recovery programme. Initially, we developed a baseline model that included the observed time-series trends in the vital rates defined above and used it to predict how abundance would be expected to change over time (i.e. from 1988 to 1995), starting with the 1988 abundance of female magpie robins and stage structure (i.e. numbers of fledged juveniles and adults). The model was run on time steps of 1 year. Next, for each of these vital rates in turn, we reran the model by holding a value constant at its 1988 value, and predicted how abundance would change over the same time period. In this way, we determined the importance of each vital rate in promoting the observed population growth.
Secondly, we used the matrix model to undertake an elasticity analysis. We calculated the population multiplication rate (λ), the stable age distribution (w), the reproductive value of each stage class (v) and sensitivities and elasticities of each matrix element using standard analytical methods as described in Benton & Grant (1999) and available in ramas stage. We estimated the parameters in the transition matrix using only data available prior to the recovery programme (i.e. 1988–89 inclusive) and the analytical methods described above.
Thirdly, we undertook a scope for management analysis to determine the possible impact management might have had on the birth rate and stage-specific survival probabilities, irrespective of the size of their elasticities. Data on the likely impact of management on demography prior to the recovery programme were sparse, but experimental studies of food supplementation conducted by Komdeur (1996) showed (i) an approximate doubling in the production of fledglings and (ii) an increase in the probability of nestlings surviving to 1 year of age from 0·25 to 0·5 (although this increase was not statistically significant) on territories provided with supplementary food. Therefore, we used the matrix model to estimate the population growth expected if these changes in vital rates could be promoted by food supplementation, by recalculating the elements in the transition matrix for the period prior to the recovery programme (i.e. 1988 and 1989).
There were no experimental studies available to examine the impact of management options on the annual survival probability of adults. However, there was widespread concern that the use of pesticides to control household insect pests was detrimental to adult survival prior to the recovery programme (McCulloch 1996), so we used a log-linear hazard model to compare the survival probability of adults living on territories that had or lacked human habitation, as a potential measure of their exposure to pesticides (using data from 1988 and 1989). We then generated estimates of adult survival that plausibly might be achieved by a pesticide ban, and used the matrix model to examine the implications for population growth in a similar way to our analysis of the possible impact of supplementary food on population growth.
time-series trends in vital rates and their contribution to population growth
Several aspects of demography changed from 1988 to 1995. The birth rate increased significantly over time (F1,71 = 4·51, P = 0·037, n = 72 female years; (Fig. 2a). The survival rate of first-year magpie robins also increased significantly over time ( = 7·21, P < 0·01; Fig. 2b), from values < 0·4 prior to 1992 to more than 0·75 from 1992 onwards. The adult survival probability showed a non-significant increasing trend ( = 1·75, P < 0·25) but the increase was much less rapid than in first-year birds (Fig. 2b). Our analysis showed no significant change in the length of the pre-breeding period over time (COX regression with YEAR as a covariate; YEAR: Wald = 0·08, P = 0·78, n = 16 females). We used the COX regression analysis to construct a survival function, which estimates the probability of a female being unpaired and not breeding at a given age (weeks). This showed that the probability of remaining a non-breeder at 1 year of age was approximately 0·9 (Fig. 2c), so the probability of becoming a paired female and breeding as a yearling was 0·1 and remained relatively constant during the recovery programme.
The two-stage matrix model was parameterized with the vital rates shown in Fig. 2, and assuming a constant probability over time of becoming a paired female and breeding as a yearling. Note that survival estimates for 1991 are lacking because no detailed data exist on individual resightings of fledged birds (Fig. 2b), so we simply used the average rates for adults and juveniles across the entire time series to estimate population growth in this year. The baseline model showed the rapid population growth that occurred after 1992 (Fig. 1), during which annual population multiplication rates (λ) varied from 1·26 to 1·52 (Fig. 3). When the adult survival rate was held constant at its 1988 value, the model predicted that the observed increase in the birth rate and annual survival probability of first-year birds also produced significant population growth after 1992, although specific annual growth rates were slightly lower than in the baseline model (λ ranged from 1·2 in 1995 to 1·39 in 1993). In contrast, if either the birth rate or annual survival probability of first-year birds was held constant at its 1988 value, the magpie robin population failed to grow, remaining at abundance levels comparable with those observed prior to the recovery programme or declining. This showed that the increase in the birth rate and annual survival probability of first-years over time acted together to promote the observed population growth in the magpie robin population, with changes in the annual survival probability of adults playing a relatively minor role.
Only a single female magpie robin fledged and was recruited into the breeding population in the period immediately prior to the recovery programme. This female fledged in 1988 and was 70 weeks (1·3 years) of age when she started nesting. This made it difficult to estimate the probability of becoming a paired female and breeding as a yearling (c0) directly from the demographic data for 1988 and 1989. To circumvent this problem we used all the data available over the study period to estimate this parameter, as our analysis revealed that it did not vary over time (see above).
The parameter estimates from Table 2 generated the following transition matrix:
Table 2. Parameter estimates of the two-stage matrix model for the period prior to the recovery programme (1988–89)
Adult annual survival probability
First-year annual survival probability
Probability of becoming a paired female and breeding as a yearling
Prior to the recovery programme λ = 0·974, suggesting a < 3% per annum decline in abundance. Adults dominated the stable age distribution (w) (adults: 0·94; yearlings: 0·06) and the reproductive values (v) were also higher for adults (adults: 1·12; yearlings: 1·0). Population growth was most sensitive to variation in the annual survival probability of adults, followed by adult fecundity and yearling survival to adulthood (Table 3). Note that sensitivity measures the impact on population growth of a small change in a transition matrix element relative to equal absolute changes in other elements, whereas elasticity refers to a proportional change in population growth resulting from a proportion change in a transition matrix element. Given that the vital rates used for the elasticity analysis were estimated over a short time period (i.e. < 2 years), we examined the impact on the elasticity analysis of variation in the vital rates, both in isolation and in combination (Table 4). This showed that the annual survival probability of adults had the largest elasticity across a broad range of vital rate values.
Table 3. Results of the elasticity analysis. Definitions of the particular vital rate symbols are given in Table 2
Recruitment of yearlings into adult population
Adult survival probability
Table 4. The impact of variation in vital rates on the elasticities of the matrix elements. Symbols and matrix model elements are defined in Table 2. The table shows how each vital rate was varied, and its subsequent impact on the elasticity analysis. If a cell is empty then the vital rate was maintained at the observed value given in Table 1
Elasticities of matrix elements
Recruitment of yearlings into adult populations
Adult survival probability
scope for management analysis
We first examined differences in the annual survival probabilities of adult magpie robins occupying territories with and without human habitation, because the elasticity analysis identified this as the ‘key’ vital rate in terms of population growth. Prior to the recovery programme no adults died on territories lacking human habitation (106 bird months, n = 6 birds) whereas the survival probability was lower on inhabited territories (0·812; 173 bird months, n = 13 birds), although the difference was not statistically significant (log-linear hazard model including habitation as a two-level factor: = 2·87, P < 0·1). As the overall survival probability of adults was 0·9227 during the same period, there appeared to be some limited scope to increase the adult survival probability by banning the use of household pesticides (assuming a causal link between survival and pesticide use) to probabilities of perhaps 0·95 or 0·99.
We estimated population growth expected from these changes in adult survival, and compared them with those expected from changes in the birth rate and juvenile survival probability promoted by the provision of supplementary food (see the Methods). Improving the annual survival probability of adults was sufficient to maintain stable levels of population growth or to promote a small increase in growth of c. 4% per annum (Fig. 4). However, even for relatively small increases in the survival probability of first-year birds and a doubling of the birth rate, population growth rates were higher than those achievable by targeting management solely to the adult survival rate. Even so, population growth was still lower than the annual rates observed during the recovery programme (Fig. 4). This remained the case even if the combined effects of improved adult (0·95) and first-year survival (0·5) and a doubling of the birth rate (0·6) on population growth were estimated (λ = 1, 1947).
Our analysis raises two important questions. First, how would a more formal use of simple demographic models have affected management decisions within the recovery programme for magpie robins developed in the early 1990s, and how would these decisions compare with the population growth actually achieved by the recovery programme? Secondly, what are the wider implications of the magpie robin study for the use of demographic models in designing management for endangered species using elasticity and scope for management analyses?
The management measures included in the magpie robin recovery programme were designed to address real or perceived ecological problems, rather than being explicitly targeted to improve particular vital rates (McCulloch 1996). Studies by Komdeur (1996) prior to the recovery programme had shown that food supplies available on territories could limit the birth rate, so supplementary feeding and habitat management were instituted to try to improve feeding conditions. Magpie robins nest in tree holes but these were lacking in the predominantly exotic woodland on Frégate, so nest boxes were provided. Finally, concerns about the use of pesticides to control household insect pests were raised because fledged birds seemed more likely to disappear from territories inhabited by humans. As a result, a ban on the use of organophosphate and carbamide pesticides was implemented. This suite of management measures resulted in significant population growth (Fig. 1) and our analysis has shown that this growth was driven primarily by a dramatic increase in the annual survival probability of first-year birds and an increase in the birth rate (Fig. 3). It is difficult to assess the relative impact of different management measures on vital rates (Table 1), as most were implemented on a non-experimental basis and applied to the entire population.
What insights does the demographic model provide on the ability of management to promote population growth? Our elasticity analysis showed that population growth was most sensitive to small perturbations in the annual survival probability of adult birds. It is recognized that the results of an elasticity analysis only have practical value if management can manipulate the key vital rates identified by the analysis (Caswell 2000; Ehrlén, van Groenendael & de Kroon 2001). We showed that, prior to the recovery programme, there was evidence consistent with the hypothesis that the use of household pesticides might be reducing adult survival, and by implication that a ban might result in an increase in survival probabilities. However, as annual survival probabilities were already high (0·9227), there was only limited scope for improving survival further, even though such changes would have stabilized population growth from < 3% decline per annum to a slight increasing trend (Fig. 4). More importantly, however, by undertaking a more complete analysis of the scope for management to affect other vital rates (i.e. birth rate and annual survival probability of first-year birds), we showed that higher rates of population growth were achievable by targeting management (in this case food supplementation) to these life-history stages, even though the vital rates associated with these stages had relatively low elasticities (Table 3). This provides an important practical illustration of the principle outlined by Ehrlén, van Groenendael & de Kroon (2001), who argued that it would be unwise to use elasticity analysis in isolation to target management because particular management options might be able to alter vital rates that have relatively small elasticities in a way that causes a greater increase in population growth than that achievable by targeting management solely towards vital rates with relatively large elasticities. Comparable examples are few, but other studies that have included both an elasticity analysis and an assessment of the scope for management to improve vital rates irrespective of their elasticity provide a similar conclusion (Doak, Karieva & Klepetka 1994; Hiraldo et al. 1996). While this may seem intuitively obvious, the important point illustrated by the magpie robin data is that the practical outcome of management in terms of realized population growth can be markedly different between the analytical approaches.
It is clear from our work that an elasticity analysis in isolation would have been unable to target management effectively to promote increased population growth because of constraints in improving the ‘key’ vital rate using management. Assessing the scope for management to improve vital rates irrespective of their elasticities would have enabled targeted management to achieve increased population growth. However, only a limited range of potential management options could be formally examined (specifically supplementary feeding and a pesticides ban) due to limited data on the plausible impact of management on particular vital rates. As a consequence, although this analysis identified management that could promote increased population growth, the predicted growth rates were lower than those actually achieved by the recovery programme (Fig. 4). This implies that the recovery plan had a greater positive impact on vital rates than the more limited range of management options for which data were available for our analysis prior to the recovery programme. This view is supported by the vital rate data shown in Fig. 2 because both the annual survival probability of first-year birds and, to a lesser extent, the birth rate exceeded the values expected from the supplementary feeding experiment conducted by Komdeur (1996) (see the Methods). This highlights an important limitation, imposed by available data, on the ability of a scope for management analysis to estimate population growth actually achievable by all practical management options.
What are the general implications of the magpie robin study? Although a number of studies have used elasticity analysis to identify key vital rates in terms of population growth (Crowse, Crowder & Caswell 1987; Doak, Karieva & Klepetka 1994; Escos, Alados & Emlen 1994; Heppell, Walters & Crowder 1994; Maguire, Wilhere & Dong 1995; Marschall & Crowder 1996; Plissner & Haig 2000) and some studies have then estimated the potential population growth that might be achieved by focusing management on the key vital rate (Crowder et al. 1994), our analysis and other studies (Hiraldo et al. 1996) argue strongly that this is insufficient in isolation to generate effective management. This is because there is often limited scope for management to improve population growth by improving the ‘key’ vital rate (Caswell 2000), and because management has the potential to influence population growth by improving vital rates with low elasticities. In our view, this means that demographic models should be used to undertake a comprehensive assessment of the scope for management to affect all vital rates (and subsequent population growth) for which there are available data on the actual or potential impact of management on specific vital rates. If this is not done, there is the possibility that a potentially beneficial management option might not be considered simply because it involved a vital rate that had a relatively low elasticity.
The only important issue then becomes whether there are sufficient data to permit a comprehensive scope for management analysis to be undertaken. Often sufficient information will not be available, given the paucity of demographic data on endangered species, even for relatively well-studied taxa such as birds (Green & Hirons 1991). Furthermore, it is worth noting that the data available for magpie robins, although limited in many respects, are remarkably detailed in comparison with other endangered species. Even in this case, the discrepancy between population growth rates estimated by the scope for management analysis and that realized by the actual recovery plan highlights the limitations that a demographic modelling analysis experiences due to uncertainties in how the full range of management options might impact on vital rates. On a practical level, this inevitably means that certain management decisions will need to be taken on a precautionary basis. That is, management measures that alleviate perceived ecological problems have a potentially important role to play. It will not be possible or wise only to institute management measures that have a demonstrable ecological benefit. The successful recovery programme for the Seychelles magpie robin and other avian examples, particularly from Mauritius (Jones et al. 1995; Jones & Swinnerton 1997; Safford & Jones 1998), show that identifying perceived ecological problems and designing management to alleviate these problems can be a profitable approach to endangered species management. Under these circumstances, demographic models can play a useful role in examining the possible short-term impacts that particular management options might have on population growth, but they should not be used in isolation.
We thank the management and staff of Frégate Island for their help and support. Jan Komdeur, Adam Gretton and Neil McCulloch undertook magpie robin monitoring between 1988 and 1995. BirdLife International, the Royal Society for the Protection of Birds and latterly BirdLife Seychelles supported the magpie robin work. Rhys Green and Mark Beaumont provided constructive criticism of an earlier version of the paper.