Gilles Gauthier, Département de Biologie, Université Laval, Sainte-Foy, PQ, G1K 7P4, Canada (fax 418 656 2043; e-mail firstname.lastname@example.org).
1In recent decades, the populations of several goose species have increased dramatically and are severely impacting on their habitat. We examined the relative contribution of reproduction and mortality to changes in the dynamics of the entire North American population of greater snow geese Anser caerulescens atlanticus from 1970 to 1998.
2The total population increased 10-fold over this period, with two phases of rapid population growth in 1970–74 and 1984–98 separated by an intervening period of stagnation. The reproductive rate was estimated from age ratios in the autumn, survival from ring recoveries, and harvest rates from hunter surveys.
3Variations in population growth could not be explained by changes in reproduction, which was similar across the three periods (overall mean 26 ± 3% young in the autumn flock) with no evidence of density-dependence.
4Adult survival did not differ between the two periods of rapid population growth (0·84 ± 0·04 during 1970–74 vs. 0·80 ± 0·04 during 1990–96), thus providing no evidence of density-dependence effects either. The survival rates of young could only be estimated during 1990–96, when they varied greatly (mean 0·36 ± 0·12, annual range 0·11–0·48).
5Adult harvest rates were much higher during the period of no population growth (0·11 ± 0·01%) than before (0·04 ± 0·01%) or after (0·06 ± 0·01%). The increased harvest starting in 1975 was due to the re-opening of the US hunting season. Thus, reduced survival due to increased hunting mortality apparently caused the stagnation of growth between 1975 and 1983.
6We conclude that hunting mortality has had the most impact on recent population dynamics in the greater snow goose and, in the absence of density-dependent effects, hunting could be used to limit the growth of this population.
The management of overabundant wildlife species is a growing challenge in conservation biology. For instance, exploding populations of deer and geese in several parts of North America threaten habitat integrity (Ankney 1996; McShea, Underwood & Rappole 1997). The demographic causes of these population expansions are not always clear, and may lead to controversy in determining the management actions required, as in the mid-continent population of lesser snow geese Anser caerulescens caerulescens (Cooke et al. 2000; Rockwell & Ankney 2000). Ideally in these circumstances, adequate management should be based on a detailed understanding of the population dynamics.
In long-lived species, population growth is most sensitive to adult survival (Lebreton & Clobert 1991), and spatial and temporal variations in survival can strongly affect population dynamics. For example, Kanyamibwa et al. (1990) related the decline of white storks Ciconia ciconia to a decrease in survival rate apparently due to drought on the wintering grounds in West Africa. For wandering albatrosses Diomeda exulans, Croxall et al. (1990) suggested that the decline in population size was due to increased mortality from fishing lines.
Knowledge of systematic changes in survival rates and of factors influencing them is important for the management of hunted populations. The extent to which hunting affects survival is of considerable applied interest because hunting mortality can be either compensatory or additive to natural mortality (Nichols et al. 1984). Hunting mortality is compensatory to natural mortality when the risk of dying from natural causes decreases in response to the increase in hunting mortality (Boyce, Sinclair & White 1999). However, in geese, hunting mortality is largely additive to natural mortality, i.e. most geese killed by hunters would have otherwise survived to reproduce again (Gauthier et al. 2001). Under such circumstances, hunting should have a direct influence on population growth.
As in the mid-continent population of lesser snow geese (Abraham & Jefferies 1997), the greater snow goose Anser caerulescens atlanticus L. has increased considerably in number during the 20th century. Fewer than 10 000 birds were present in the early 1900s (Lemieux 1959), leading to strict protection. A year-round prohibition of hunting in the USA and an open season limited to the autumn migration through southern Canada allowed the population to increase to 40 000 birds in 1967 (Reed, Giroux & Gauthier 1998). Following further population growth, hunting was resumed on the wintering grounds in the USA in 1975. Despite increased hunting pressure, the population continued to grow, reaching 741 200 birds in spring 1998 (Reed, Giroux & Gauthier 1998). In other goose populations, similar population increases have brought density-dependent effects on survival and fecundity (Cooch & Cooke 1991; Pettifor et al. 1998, 2000).
Detailed analyses of the population dynamics of white geese in North America became a pressing issue as new spring hunting seasons were recommended in 1998 to limit or reduce these populations (Batt 1997; Giroux et al. 1998a). Our objective in this study was to determine the relative contribution of reproduction and mortality to changes in the population dynamics of the greater snow goose over a 30-year period, and to look for evidence of density-dependent effects. We estimated the annual survival rate of geese in relation to age, sex and breeding locations from ring recoveries, and we used autumn age ratios as an index of reproductive rate. We also assessed long-term changes in harvest rates and examined the relationship between hunting mortality, survival and population growth.
The greater snow goose population has been surveyed every spring since 1969 by the Canadian Wildlife Service on the staging areas along the St Lawrence River, Quebec, Canada, using aerial photography. This survey yields an almost complete photographic coverage at a time when the entire population of this subspecies is present in a relatively small area (for details see Gauvin & Reed 1987). This survey thus estimates population size with a much greater accuracy than the mid-winter counts of most waterfowl populations (Reed, Giroux & Gauthier 1998). To characterize population growth, we divided the period into three time intervals following Reed (1990): the early period, from 1970 to 1974; the mid-period, starting with the re-opening of the US hunt in 1975 until 1983; and the late period, from 1984 to 1998, corresponding to a decline in harvest. We first modelled population growth from 1970 to 1998 as ln(Nt) vs. t using simple regression. We then adjusted separate regression lines for the three periods using dummy variables and their interactions in a multiple regression model. We used the Akaike’s information criterion adjusted for small sample size (AICc; Burnham & Anderson 1998) to select the model providing the best fit to the data. The slopes of the regression lines were used as estimates of growth rate.
From 1970 to 1998, the proportion of juveniles (i.e. young-of-the-year) in the autumn flock and brood sizes (since 1973) were determined by visual observations. These observations were made at several sites along the St Lawrence estuary during autumn staging (for details see Reed, Giroux & Gauthier 1998). The mean proportion of young and the mean brood size in the autumn were compared among the three periods with an analysis of variance.
Between 1970 and 1974, mass ringing of greater snow geese was conducted at several colonies throughout their Arctic breeding range by the Ministère de l’Environnement et de la Faune (Quebec Province, Canada). Sites included North Baffin, Bylot, Somerset, Devon, Bathurst, Axel-Heiberg and Ellesmere Islands (Nunavut) and ranged from 71° to 81°N. Geese were captured in summer when the adults were moulting, during two distinct periods, 3–6 July and 1–3 August, and probably included a mixture of breeding and non-breeding birds. Sex was determined by cloacal eversion. All birds received a metal US Fish and Wildlife Service ring. Too few goslings were ringed during this time period to estimate their survival rate. These ringing data were obtained from the Bird Banding Laboratory (USFWS, Washington, DC).
Between 1990 and 1996, we also ringed geese at one of the major breeding colonies, Bylot Island (Nunavut, 73°N 80°W; Hughes, Reed & Gauthier 1994). Each year, flightless geese (moulting adults with young) were captured in late summer (8–18 August). Birds were classified as adults (≤ 1 year old) or young based upon plumage. The same ringing procedure as above was applied, except that most adult females also received a yellow plastic neck collar with an alphanumeric code (Menu et al. 2000).
Ring recoveries were obtained from the Bird Banding Laboratory and came from hunting during the autumn migration in southern Quebec and on the wintering ground in eastern USA. Since 1990, some hunting guides from Quebec also reported rings directly to us.
survival rate analysis
We estimated survival with ring recovery analyses with the program mark (White & Burnham 1999). For the early period (1970–74) we used recoveries from 1970 to 1989, and for the late period recoveries were from 1990 to 1996. We used the general methodology described by Brownie et al. (1985) and followed their terminology. Recovery rate (fi) is the probability that a ringed bird alive at the time of ringing in year i will be shot and retrieved during the hunting season in year i, and its ring reported to a wildlife agency. Survival rate (Si) is the probability that a bird alive at the time of ringing in year i will survive to the time of ringing in year i + 1.
Because geese were ringed over a large latitudinal range during the early period but at only one site during the late period, we were concerned with possible heterogeneity in survival rate related to breeding latitude, which would confound comparisons across the two time periods. To test this possibility, we analysed survival rates of adults in relation to breeding latitudes by grouping geese into those ringed south or north of 74°N (i.e. Lancaster Sound) for the early period. However, geese were ringed north of 74°N from 1971 to 1973, and south of 74°N in 1970–71 and 1973–74 (Table 1). Thus, geese were ringed both south and north of 74°N in only two of the five years (1971 and 1973). Data were too sparse to test the effects of both sex and breeding latitude simultaneously. Therefore, we conducted two separate analyses, one examining the effects of sex and time on survival and recovery rates, and one examining the effect of breeding latitude and time. This approach assumed that there was no interaction between sex and latitude. The sample size differed between the two analyses because some birds were of unknown sex.
Table 1. Number of greater snow geese ringed in the Canadian Arctic (Nunavut) and subsequently recovered during two time periods. During the period 1970–74, geese were ringed on several islands in the Arctic north and south of 74°N of latitude, but during the period 1990–96 they were all ringed on Bylot Island, 73°N
The recovery period for birds ringed in 1970–74 was 1970–89.
For the late period (1990–96), a previous analysis of females only (Menu et al. 2000) showed that neck collars did not affect survival but affected recovery rate (especially direct recovery, i.e. recovery in the year that the bird was ringed). Therefore, in our analysis we examined the effects of age (two age classes: young and adults), sex and time on survival and direct (f*) and indirect (f) recovery rates, and of neck collars on recovery rates.
For all analyses, we started with the most general model, which included all variables and their interactions. We then constructed reduced models by constraining some parameters to determine the most parsimonious list of parameters needed to model the data. We used the AICc to select the best model, i.e. the one with the lowest AICc, and other models were ranked relative to deviations from the best model (ΔAICc). When differences in AICc were small, we computed a weighted average of the parameters across models using AICc weights, which represent the weight of evidence in support of each model in the set of candidate models (the sum of all AICc weights = 1; Burnham & Anderson 1998). Differences in survival estimates coming from separate time periods were tested against χ2 (program contrast; Hines & Sauer 1989).
We tested the fit of the models to the data using the goodness-of-fit tests of Brownie et al. (1985) for ring recovery analyses (K. P. Burnham, personal communication). These tests were performed with software estimate and brownie. When the tests were significant, we corrected for this extra-binomial variation by a variance inflation factor, ĉ, which is the ratio of the χ2 goodness-of-fit test divided by the degrees of freedom (Lebreton et al. 1992). This variance inflation factor was used to adjust the deviance in the calculation of the AICc (which then becomes the quasi-likelihood AICc, QAICc), and √ĉwas used to adjust the standard errors of the estimates (these adjustments are made automatically by mark; White & Burnham 1999).
We estimated harvest rates independently from the ringing data. Harvest rate was calculated as the ratio of total number of geese harvested over the estimated total autumn population size. Total numbers of snow geese killed by sport hunting are estimated annually in Canada and USA through the national harvest surveys designed for all migratory waterfowl (Boyd & Finney 1978). Because greater snow geese are restricted to the Atlantic Flyway and overlap little with other snow goose populations during breeding, migration and winter, we defined the harvest for this population as the number of geese killed in Quebec south of latitude 47° 30′N, and in all the US states of the Atlantic Flyway (15 states reported a snow goose harvest, the major ones being New Jersey, Delaware, Maryland and North Carolina; Reed, Giroux & Gauthier 1998). Age ratios in the harvest were determined from tail fans of shot geese collected through the parts collection survey (Reed, Giroux & Gauthier 1998). Although the sample sizes were sometimes small (mean annual number of tails = 344, range 25–1081), age ratios in the harvest were highly correlated with the proportion of juveniles in the autumn flock (R2 = 0·86, n = 27, and R2 = 0·90, n = 22, for Quebec and USA harvest, respectively) and were thus considered accurate.
Population size data used in the harvest rate calculation came from the spring aerial photo survey (see above), which is considered an accurate total population estimate (Reed, Giroux & Gauthier 1998). Two methods were used to estimate the size of the autumn population from the spring survey. First, the autumn population size of adults in year i was estimated using the preceding spring survey multiplied by the seasonal survival rate of adults (from May to October: over-summer survival, Ssu). Over-summer survival was calculated by assuming constant survival across a year and using the annual survival rate (S) estimated by the ring recovery analysis (Ssu = S5/12). With the second method, the autumn population was estimated with the following spring survey plus the total estimated harvest, assuming no natural mortality during the hunting season. For the two methods, the autumn population was divided into young and adult segments, using the proportion of young in the autumn flock, to estimate harvest rate for the two age classes separately (see details in the Appendix).
population size and reproduction
The spring population has increased 10-fold from 1970 to 1998 (Fig. 1). Although both the simple and multiple regression models of population growth were significant (R2 > 0·94, P < 0·001), the model with three different growth periods provided a better fit to the data (ΔAICc = 27·5 compared with simple regression). This confirmed that the growth rate of the population varied across the three periods (Table 2). From 1970 to 1974, the population increased rapidly from 89 620 to 165 000, for a mean annual growth rate of 14·7%. In contrast, there was virtually no increase from 1975 to 1983 (1·3% year−1). After 1983, the population started to grow again, from 185 000 in 1983 to 741 200 in 1998 (9·0% year−1).
Table 2. Demographic parameters of the greater snow goose population for the different time periods (mean ± SE)
Estimated from a multiple regression between ln(Nt) vs. t (see the Methods). F5,22 = 307, P < 0·001.
One-way anova among periods: F2,26 = 0·31, P = 0·737.
One-way anova among periods: F2,25 = 0·07, P = 0·934.
Comparison between periods: χ2 = 0·503, d.f. = 1, P = 0·478.
No ringing during these periods.
Estimated from birds ringed between 1990 and 1996. Adult survival is female only.
Estimated from birds ringed during 1970–74.
Estimated from birds ringed during 1970–74 (period 1984–89) and 1990–96 (period 1990–96, excluding females with neck collars).
One-way anova among periods: F2,25 = 15·9, P < 0·001 (Tukey test: period 2 > [period 1 = period 3]): first method.
One-way anova among periods: F2,25 = 9·60, P < 0·001 (Tukey test: period 2 > [period 1 = period 3]): first method.
Over the three decades, annual reproductive rate fluctuated widely, with the proportion of young in the autumn flock ranging from 0·4% to 48% (see Fig. 5 below). However, there was no difference among the three contrasting periods of population growth either in the proportion of young in autumn (overall mean 25·6 ± 2·5, SE) or in brood size (2·53 ± 0·05; Table 2). Furthermore, the highest reproductive rates occurred when the population was largest (Table 2). This suggests that changes in reproductive rate were not responsible for variations in population growth over the whole period.
survival and recovery rates
During the period 1970–74, 4624 adult geese were ringed, of which 614 (13·3%) were recovered between 1970 and 1989 (Table 1). The sex-specific goodness-of-fit test indicated that the model Sstfst adequately fitted the data (overall χ2 = 104·8, d.f. = 83, P = 0·054; males χ2 = 44·7, d.f. = 43, P = 0·399; females χ2 = 60·0, d.f. = 40, P = 0·022). A similar test with groups split according to breeding latitude was not possible due to the sparseness of the data. In the best model of the analysis with sex, survival was constant and equal between sexes (Table 3). The sum of AICc weights of models without a sex effect on survival was 0·781 compared with 0·218 for those with a sex effect. In the latter model, survival rate of males and females differed by less than 1% (0·836 ± 0·027 vs. 0·843 ± 0·027, respectively). In the best model of the analysis with breeding latitude, survival was constant over time but different according to breeding latitude (Table 3). However, the sum of AICc weights of models without a latitude effect was almost as high as the best model (0·458 vs. 0·536), providing only weak evidence that survival varied with breeding latitude. When averaged across all models, survival rate of geese breeding north of 74°N was slightly lower than those breeding further south (0·827 ± 0·043 vs. 0·838 ± 0·043, respectively). We therefore retained the latter value for comparison with survival during the late period because these geese were all ringed south of 74°N.
Table 3. Modelling of survival (S) and recovery rates (f) for greater snow geese ringed in the Arctic during the early period (ringing: 1970–74; recoveries: 1970–89), adults of both sexes. (a) Effect of sex and time modelled (breeding latitudes pooled). (b) Effect of breeding latitude (north vs. south of 74°N) and time modelled (sex pooled). For each model, the relative corrected Akaike’s information criterion (ΔAICc), its weight, the number of estimable parameters (No. parameters) and deviance are given. Only the most parsimonious models (AICc weight > 0·001) and the general model are shown for each analysis
g: breeding latitude (south vs. north of 74°N).
t3: time-dependence reduced to three homogeneous periods: 1970–74, 1975–83, 1984–89.
(a) Effect of sex and time
(b) Effect of breeding latitude and time
During the period 1990–96, 19 185 young and adult geese were ringed, of which 1308 (6·8%) were recovered (Table 1). The sex-specific goodness-of-fit test indicated that the model Ssatfsatf*st did not adequately fit the data (overall χ2 = 44·9, d.f. = 26, P = 0·012; males χ2 = 20·7, d.f. = 15, P = 0·148; females χ2 = 24·2, d.f. = 11, P = 0·012). This model, however, was slightly less general than our starting model as we could not include the neck collar effect on f in this test (Table 4). For subsequent analysis, we used a ĉof 1·726 (44·87/26). The best model had survival rates different between adults and young, equal between sexes, and time-dependent in young only (Table 4). There was strong support for most of these effects as the sum of QAICc weights across models was c. 1·0 for the age effect, 0·99 for no sex difference in young, 0·81 for time-dependence in young and 0·90 for constant survival in adults. There was weaker support for no difference between sexes in adults (QAICc weights = 0·65). Survival rates of adults were estimated at 0·797 ± 0·044 in females and 0·789 ± 0·045 in males, a very slight difference (average of top three models in Table 3). Adult survival rate did not differ significantly between the periods 1990–96 and 1970–74 (Table 2). Average survival rate of young was 0·356 ± 0·124 and point estimates ranged from 0·112 to 0·484 (Fig. 2a).
Table 4. Modelling of survival (S) and recovery rates (f) for greater snow geese ringed on Bylot Island during the late period (1990–96), young and adults of both sexes. For each model the relative corrected Akaike’s information criterion (ΔQAICc), its weight, the number of estimable parameters (No. parameters) and deviance are given. Only the 11 most parsimonious models (AICc weight > 0·002) and the general model are shown (ĉ = 1·726)
f*: direct recovery rate (i.e. in the year that the bird was ringed).
a: age-dependent (young and adults).
n: neck collar dependent (adult females only).
Y: time or sex effect present in young only.
A: time or sex effect present in adults only.
f*N: direct recovery rate different in neck-collared females only.
Variations in recovery rates can provide an index of variations in harvest rates assuming no change in reporting rates by hunters over time. During the period 1970–89, recovery rates were time-dependent and did not differ between sexes or breeding latitudes (sum of AICc weights for models with these effects c. 1·0; Table 3). Extrapolating the survival rate obtained during 1970–74 beyond this period allowed us to examine patterns of recovery rates between 1975 and 1989. Recoveries increased sharply in 1975 upon re-opening of the US hunt (Fig. 3). Precision of recovery rate estimates declined over time because no new birds were ringed after 1974.
During the period 1990–96, recovery rates differed between age classes but not between sexes, and were time dependent; direct recovery rates was also different in neck-collared birds (sum of QAICc weights for models with all those effects was 0·94). Average recovery rates of adults were about half of those for young (0·029 ± 0·004 vs. 0·064 ± 0·007, respectively), except for direct recovery rates of collared females (0·059 ± 0·010) which were twice as high as those of leg-ringed only females (Fig. 2b,c). Comparisons among the three contrasting periods of population growth showed that average adult recovery rate (excluding neck-collared birds) was higher during the low growth phase period (1975–83) than during the periods of rapid growth (1970–74 and 1984–96; Table 2).
The estimated total legal harvest increased significantly from 1970 to 1996 (R2 = 0·43, P < 0·001) but annual variation was considerable (Fig. 4a). This annual variation can be largely explained by the proportion of young in the autumn flock as harvest is much larger in years of good reproduction (R2 = 0·50, P < 0·001; Fig. 4b). Harvest rates estimated by the two methods (see the Appendix) were very similar (Fig. 5). Over the period 1970–96, the mean harvest rate was 0·074 ± 0·007 for adults but more than six times higher for young (0·470 ± 0·047), indicating that young birds were much more vulnerable to hunting than adults. The estimated harvest rate of young was aberrant in 1972 (> 1), most likely due a very small sample size for estimating age ratios in the harvest because of a record low production of young that year. Harvest rate increased considerably following the re-opening of the US hunt in 1975, especially in adults (Fig. 5). During the period of low population growth (1975–83), harvest rate of adults was three times higher than during the period 1970–74, and 1·5 times higher in young birds (Table 2). After 1984, harvest rate declined abruptly to a level similar to the period 1970–74 (Fig. 4 and Table 2). Temporal trends in harvest rates determined with the hunter surveys were consistent with those estimated with ring recovery data for the adults (Fig. 3).
population size and reproduction
Despite a stagnation of growth between 1975 and 1983, overall the greater snow goose population increased 10-fold between 1970 and 1998 and this growth has not levelled off in recent years. Our two measures of reproductive rate were remarkably similar during the three contrasting periods of population growth, suggesting that a reduction in reproductive rate was not the cause of reduced population growth in the late 1970s. Moreover, our data provide no evidence of density-dependent effects acting either on the proportion of breeding adults (based on proportion of young in the autumn flock) or on the reproductive success of breeding adults (based on brood size in the autumn) during the whole 30-year period. Annual variations in the proportion of juveniles and the mean brood size in the autumn can be explained by unpredictable climatic factors affecting goose reproduction (Gauthier, Rochefort & Reed 1996; Skinner et al. 1998) and variable mortality occurring during the migration between fledging and the autumn staging areas (Francis et al. 1992a; Menu 1998). The latter factor may also account for the large variation observed in annual survival of young.
Several studies have reported density-dependent effects on some components of reproduction, including clutch size, gosling growth and early survival of young in other increasing goose populations (Cooch et al. 1989, 1991; Cooch & Cooke 1991; Francis et al. 1992b; Loonen, Oosterbeek & Drent 1997; Pettifor et al. 1998, 2000). In greater snow geese, Reed & Plante (1997) showed a decline in body size, mass and condition of juvenile birds over the period 1975–94 which they attributed to increasing densities on the brood-rearing areas and reduced per capita food availability. However, it appears that this density-dependent effect has not yet affected reproductive rate. Pettifor et al. (1998) argued that density-dependent effects on vital rates were difficult to detect at the population level in barnacle geese Branta leucopsis because, as some colonies filled to capacity and experienced strong density-dependent effects, birds founded new colonies where density was low and growth rate high. However, in greater snow geese, growth rate at the large colony of Bylot Island was similar to the whole population over the last 15 years, and there is no evidence of expansion of birds into new colonies (Reed, Giroux & Gauthier 1998).
There were no differences in adult survival rates between 1970–74 and 1990–96, two periods when the population was growing rapidly. Although survival rate tended to be lower during the late period (0·797 vs. 0·838 for the early period), Gauthier et al. (2001) estimated survival rate of adult females at 0·828 during 1990–98 using neck collar resighting data. Moreover, as some non-breeders were ringed during the early period but not during the late one, we cannot exclude the possibility that the survival of geese was biased high during the early period because non-breeders do not incur any cost of reproduction. Thus, all the evidence suggests the absence of density-dependent effects on adult survival even though the population has increased 10-fold over the past 30 years.
In mid-continent lesser snow geese, a population at least five times larger than that of greater snow geese, Cooke et al. (2000) reported an increase in adult survival during a 24-year period characterized by a continuous population growth. This suggests that adult survival is relatively insensitive to density-dependent natural mortality in geese, in contrast to juvenile survival for which there is evidence for density-dependence (Cooch & Cooke 1991; Loonen, Oosterbeek & Drent 1997). This is probably a fairly general pattern in large birds and mammals (Gaillard, Festa-Bianchet & Yoccoz 1998).
We found weak evidence that adult greater snow geese breeding north of 74°N had a slightly lower survival than those breeding further south. In lesser snow geese, birds from northern colonies also tended to have lower survival rates than those from more southern colonies (Francis et al. 1992a). Variations in survival rate with latitude could be due to differential costs of reproduction and/or differences in migration length, although Gauthier et al. (2001) found little evidence that migration increased mortality risk in greater snow geese.
hunting, survival and population growth
The re-opening of the US hunt for greater snow geese in 1975 clearly led to an increase in harvest rate, especially for adults. The proportion of adults in the US harvest is higher than in Quebec because fewer young are still alive to be shot when the flock arrives in the US. Moreover, although young geese are more vulnerable to the gun than adults, this effect attenuates by the time they reach the US. The re-opening of the US hunt coincided with the beginning of the period of stagnation in population growth, suggesting that hunting mortality was the cause. This is consistent with the observation that hunting mortality is largely additive to natural mortality in geese (Francis et al. 1992b; Rexstad 1992; Gauthier et al. 2001). Because adult survival is the vital rate to which population growth is most sensitive in long-lived species such as geese (Lebreton & Clobert 1991; Gauthier & Brault 1998), it is likely that a reduction in adult survival was the primary factor accounting for the stabilization of the population in the late 1970s. Unfortunately, we cannot test this hypothesis directly because we have no survival estimate for during the period 1975–85 as no ringing was conducted at that time.
Ring recovery rates are often used as an index of hunting mortality although changes in recovery rates could be due to changes in reporting rate, harvest rate or both (Brownie et al. 1985). Variations in recovery rates of adult birds between the three periods were similar to those of the harvest rate, which provides further evidence that hunting mortality increased after 1974. Because no ringing took place between 1975 and 1989, recovery and survival rates were not separately identifiable unless we assumed that adult survival was constant throughout the period 1970–89. However, if increased hunting mortality led to a decrease in survival after 1974, then our estimates of recovery probability were biased downward. Therefore, our conclusion of an increase in recovery rate after 1974 is conservative.
The reoccurrence of population growth from around 1984 coincided with a reduction in harvest rate, especially in adults, thus suggesting that a decrease in hunting mortality was the primary cause. However, contrary to 1975, the change in harvest rate did not result from an obvious change in hunting regulations (Reed, Giroux & Gauthier 1998). The reasons for this decline are not clear as other factors besides regulations also affect hunting success. Although a high proportion of young in autumn is generally associated with good hunting success, one or two poor hunting seasons (e.g. due to weather conditions) despite a high production of young may have been sufficient to allow the population to escape control by hunters, thus triggering growth. During the same period, geese expanded their range during the autumn migration and winter (Reed, Giroux & Gauthier 1998). They are now more dispersed, especially in agricultural fields, where hunting may be less efficient than in the traditionally used marshes of the St Lawrence estuary (G. Gauthier, personal observation; Giroux & Bédard 1986).
The accuracy of our harvest rate estimates depends on the accuracy of the estimates of spring population size, total harvest and adult survival rate. Total harvest is likely to be the parameter with the largest error. Because the greater snow goose has a relatively small and geographically restricted population, hunter surveys may be less accurate than for some other waterfowl (Boyd & Finney 1978). This could explain the unrealistically high harvest rates of young estimated in 1972 and 1978. The precision of harvest estimates probably improved over time, with increased harvest yielding larger samples. However, temporal trends should not be strongly affected by this problem.
To estimate harvest rate, we assumed the same survival rate throughout, whereas we suggested that the increased hunting pressure during 1975–84 reduced adult survival. However, as no hunting occurs from spring to autumn, the summer survival rate should be less affected. With the second method of estimating harvest rates, assuming no natural mortality during autumn and winter could have led to an overestimation of harvest rates. Overall, because the two methods gave similar results, these biases were probably not severe.
management of overabundant goose populations
Over the past 50 years, several goose populations, including the greater snow goose, have increased dramatically, and some of them are now severely impacting their habitat to the point of threatening its integrity (Kerbes, Kotanen & Jefferies 1990; Iacobelli & Jefferies 1991; Kotanen & Jefferies 1997; Giroux et al. 1998b). As found in this study, density-dependent effects on either survival or reproduction were not strong enough to stop population growth, even though some of these populations apparently exceeded the carrying capacity of their natural habitats, especially on the Arctic breeding grounds. It has been hypothesized that this situation occurred because wintering and spring migrating geese benefit from an agricultural food subsidy in the south, thereby enabling them to maintain a positive population growth despite a shortage of summer food in the Arctic (Abraham & Jefferies 1997).
In accordance with this hypothesis, there is evidence that fat storage in spring improved during the 1980s in greater snow geese, a period when geese increasingly used farmland, especially cornfields (Gauthier, Giroux & Bédard 1992). Improved fat storage in spring has also been reported in barnacle geese after they switched from natural habitats to agriculture land for feeding in Norway (Prop & Black 1998). Even though this may enable geese to arrive in better condition in the Arctic for breeding, this has not translated into measurable effects on the reproductive rate of greater snow geese in recent decades. Alternatively, improved feeding condition in agricultural lands in winter may have reduced natural mortality, especially in young.
Change in hunting mortality appears from our data to be the primary factor affecting the population growth of greater snow geese over the past 30 years. A re-opening of hunting in the US in 1975 caused an increase in harvest, which led to a stagnation of growth for about 10 years. However, a recent decline in hunting mortality was probably responsible for the renewal of rapid population growth. Why hunting mortality declined in the late 1980s and 1990s is unclear, although it may be related to increased use of farmland where geese are more difficult to hunt. In view of our failure to find density-dependent effects acting on survival or reproduction of greater snow geese, management actions aimed at increasing hunting mortality may be warranted to limit the growth of this population and preserve the integrity of its natural habitats in the Arctic and elsewhere (Gauthier & Brault 1998; Giroux et al. 1998a).
The research was funded by the Natural Science and Engineering Research Council of Canada, the Fonds pour la Formation des Chercheurs et l’Aide à la Recherche of the Quebec Government, the Arctic Goose Joint Venture of the Canadian Wildlife Service and the Department of Indian Affairs and Northern Development. We thank the Pond Inlet Hunters and Trappers Association for their assistance and the Polar Continental Shelf Project (PCSP) for generously providing all the logistic support. We also thank G. Picard for supervising the ringing from 1990 to 1996 and the numerous people who helped with this operation. We are grateful to J. D. Heyland for having ringed geese from 1970 to 1974, and J. Nichols for his constructive comments. This is PCSP contribution no. 025-01.
estimation of harvest rates in greater snow geese
The variables used in the calculation are as follows.
NGi: spring population in year i
PFiy: proportion of young in autumn in year i
Hi: total harvest in year i (HQiin Quebec; HUiin the USA)
ARi: harvest age ratio in year i (proportion of young in the harvest; ARQiin Quebec; ARUiin the USA)
Ssu: over-summer survival
If we define NFi as the autumn population size in year i, we can calculate the size of the autumn population of adult (NFai) and young (NFyi) using two different methods:
First method: NFai=NGi·Ssu
NFyi=NFai· (PFyi/(1 −PFyi))
Second method:NFai=NFi· (1 −PFyi)
We can then calculate harvest rate in year i (HRi) for both adult (HRai) and young (HRyi), using estimates of autumn population size calculated with either method, as follows: