Estimation and limitation of numbers of floaters in a Eurasian Sparrowhawk population

Transcription

1 lbk (21) 143, Estimation and limitation of numbers of floaters in a Eurasian Sparrowhawk population IAN NEWTON* & PETER ROTHERY Centre for Ecology and Hydrology; Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK Over a 2-year period, the numbers of Eurasian Sparrowhawk Accipiter nisus nests found in a 2 k m 2 area in south Scotland remained relatively stable (mean 33.3 pairs, CV = 1.6%). Nest numbers fluctuated from year to year in a manner expected of a population subject to density-dependent regulation. The numbers of non-breeders (floaters) could not be counted directly, but the number of female floaters was estimated, using known mortality rates, from the numbers of females recruited to the breeding population each year at different ages. Female floater numbers were estimated by two methods: Method A assumed that birds bred for the first time in their first, second or third year, in the same ratio as they were found breeding for the first time in the study area; and Method B assumed that all third-year birds found breeding for the first time in the study area had bred previously, unknown to us, outside the area. Under Method A, floaters consisted of some 1-year and some 2-year birds; while under Method B, floaters consisted only of some 1 -year birds. Under both methods, the estimated number of female floaters fluctuated greatly from year to year, but under Method A they averaged.9 per female breeder, and under Method B they averaged.28 per female breeder. The Method B estimate was most consistent with other data and with the finding that some birds found breeding in the study area were likely to have bred previously outside the area (because of territory changes). Moreover, the mean and variance in female floater numbers estimated by Method B were similar in magnitude to the values obtained in a simulation model. In this model, breeding density was given a fixed ceiling, while breeding success was allowed to vary from year to year within the limits observed in the study area. It was concluded that (a> recruitment of floaters to the breeding sector was density dependent with respect to breeder numbers, i.e. broadly spealung, floaters filled gaps in the territorial system left by the deaths and movements of established breeders; and (bj floater numbers themselves were probably not regulated in a density dependent manner, but depended on whatever was the balance between annual additions (from reproduction and immigrationj and subtractions (from mortality, emigration and entry to the breeding sector). Many bird populations in the breeding season consist of territorial breeding birds and non-territorial, usually non-breeding birds, called floaters (Brown 1969, Smith 1978, Smith et al. 1991, Newton 1998). That the defence behaviour of the territorial birds prevents the floaters from acquiring a nesting territory has been inferred from observations. It has also been confirmed repeatedly in experiments in which removed territorial birds were rapidly replaced by known floaters, which took over the territory and bred there in the same year. Corresponding author ine@ceh.ac.uk Such removal experiments have now been conducted on more than 5 species of birds (review Newton 1992), including Eurasian Sparrowhawks Accipiter nisus (Newton & Marquis 1991). Collectively, they show the widespread role of territorial behaviour in limiting bird breeding densities, and in producing a surplus of floaters, able to brwd in the conventional manner only when they can get a nesting territory. In this paper, we estimate by indirect methods the numbers of female floaters associated with a breeding population of Sparrowhawks. Secondly, we assess the influence of breeding numbers and output on female floater numbers. Thirdly, we test which (if any) of 21 British Ornithologists Union

2 Floaters in a Sparrowhawk population 443 these demographic parameters vary in a density dependent manner, and thus help to regulate breeding density and total density. Analysis is restricted to females because only female breeders were caught in sufficient numbers to estimate the various parameters reliably. We know of no other study in which interactions between different components in a population have been examined in this way, although attempts to calculate the potential size of the floating component of bird populations from knowledge of breeding success and breeder survival were made by Brown (1 969) and Hunt (1998), and from radiotagging by Kenward et al. (1 999). Evidence for territorial behaviour in Eurasian Sparrowhawks Accipiter nisus, and the association between territory sizes (as assessed from nest spacing) and food supply, was given in earlier publications (Newton et al. 1977, 1986, Marquis & Newton 1982). STUDY AREA AND METHODS The study area of 2 k m2 was centred on the town of Langholm, in Eskdale, south Scotland (55'1 O", 3"O'W). The area is predominantly pastoral, and Sparrowhawks nested in the patches of (mainly coniferous) woodland scattered through it. Each year during , all the woodland was searched in an attempt to find all the Sparrowhawk nests. Details of breeding success were taken from each nest, and as many breeding adults as possible were trapped for ringing and identification. Breeders that had been ringed as chicks could be aged accurately at whatever age they were first caught in later life. Unringed breeders were recognized as 1-year-olds if they wore the brown dorsal juvenile plumage acquired in the nest, and as 2-yearolds if they wore blue-grey dorsal adult plumage but with occasional retained brown feathers. (Birds ringed as chicks, and hence of precisely known age, invariably had retained brown feathers at this age.) All unringed birds that were first caught in blue-grey dorsal plumage, but with no retained brown feathers, were classed as 3-year-olds, even though some of them may have been older than 3 years. Most of the nesting females were caught each year. Sparrowhawks were resident in the area year-round, and each pair raised no more than one brood per year, containing 1-6 young. On average, about half of all breeding attempts failed each year, mostly at the egg stage. While the numbers of breeding Sparrowhawks in each year were known from the numbers of nests, floaters could not be identified as such or counted, without the use of expensive radiomarlung (Kenward 1978, Kenward et al. 1999). Their numbers therefore had to be estimated indirectly from birds recruited each year to the nesting population. The annual breeding population itself consisted of established breeders that had survived from the previous year and new breeders recruited in the current year. From the age composition of the new breeders, and from knowledge of the annual mortality of different age groups (Newton & Rothery 1997), we could back-calculate to estimate the numbers of floaters in each previous year. Earlier work had suggested that the floating population was made up of 1 -year-old and 2-year-old birds, all Sparrowhawks in Eskdale breeding at the latest by their third year (Newton 1986). Birds which were recruited into the breeding population as 2-year-olds in year t were assumed to be survivors of the 1-year-olds in the floating population of year t-1, and birds recruited as 3-year-olds were assumed to be survivors of the 2-year-old floaters in year t-1. The estimated number of floaters was obtained by applying the average annual survival rates of l-year-old and 2-year-old birds to the number of 1 -year-old and 2-year-old recruits. The latter figures were obtained from observed proportions among trapped birds applied to the total annual nest count (Table 1). Thus, if seven out of 28 trapped breeders were 1-year-old recruits, and the total nest number was 36, then the total number of 1-year-old recruits was taken as 7/28 x 36 = 9. The same procedure was applied to 2-year-old and 3-year-old recruits, and the sum of the three estimates gave the total number of new recruits for that year. The numbers of 1 -year-old floaters in year t (fir) were thus estimated as (ficc+,, + n2(t+l,)/sl and the numbers of 2-year-old floaters (f2j in year t as n3ct+ll/s2, where n,, and n3c were the numbers of 2- and 3-year-old birds recruited in year t, and S, and S, were average survival rates of 1-year-olds and 2-year-olds. The above procedure, here called Method A, also gave maximum estimates of the age of first breeding, and hence of the size of the floating (prebreeding) population. The estimates were maxima because the method took insufficient account of the movements of adults. While up to 7% of known surviving adults nested on the same territories in successive years, others changed from one nesting territory to another. Mostly they moved to a nearby territory, but occasionally further afield. This meant that a small proportion of breeders might have left the study area completely each year to breed elsewhere and that others might have moved in. The latter would have been recorded as first-time breeders in the area, when in fact they had bred previously elsewhere, but unknown to us. 21 British Ornithologists' Union, Ibis, 143,

3 ~~ ~ ~~~~ ~~~~ ~ ~~ ~ ~ ~ ~ ~ ~~~ ~~~~~~~ ~~~ ~ ~ ~~~ ~~ ~~~~ ~~ ~ Newton & f? Rothery movements would have led to overestimates of the numbers of birds which did not breed until 2 or 3 years old (and hence to overestimation of the total numbers of floaters in the population), and also to underestimates of breeder survival (which is measured as year-to-year persistence in the study area). In an attempt to address this problem, we made a second series of estimates of non-breeder numbers assuming that all third-year birds had bred before; i.e., that all birds bred for the first time in their first or second year (here called Method B). The actual totals of floaters were likely to lie between these two sets of estimates. To assess their likely validity, we examined both sets in relation to other information, and in relation to the findings from a population simulation model. RESULTS Trends and density dependence in nest numbers During , the number of nests found each year remained relatively stable, around a mean of 33.3 (CV = 1.6%, Table 1). Linear regression of log nest numbers against time showed no long-term upward or downward trend (b = -.54, se = k.44, r2 = 7.8%, P =.23). Inspection of the annual nest counts revealed that numbers fluctuated from year to year in a manner expected of a regulated population: in general, years with high nest numbers were followed by large proportionate declines, and years with low nest numbers by large proportionate increases. Bulmer's (1975) test applied to the annual nest counts showed strong evidence of density dependence (R =.51 7, P <. 1). A potential problem in testing for density dependence arises from possible errors in nest counts (i.e. if any nests were missed), which would tend to inflate the statistical Type I error and lead to spurious statistical significance. To assess the effect of any missed nests, the null distribution of R was simulated using the random walk model (Bulmer 1975), but with counts subject to estimation error. The approach Lve took assumed that each nest was detected lvith a constant probability of.9, so that on average 9%) of nests were detected, although this proportion might vary from year to year. For independently located nests, the standard deviation in the nest count was then equal to t'(.1 x.9 x =.3tiN (for N = 3, cv = 5.9%). In applying Bulmer's test for the simulation exercise, the variance in the change in log nest numbers was estimated by subtracting from the observed variance the amount expected from the sampling error. The P- value was then estimated as the proportion of values of Table 1. Annual nest numbers, and numbers of established and new breeders among trapped females, Eskdale, south Scotland Numbers trapped of Numbers of Established ~~ New breeders of following ages (years) nests breeders British Ornithologists' Union. Ib6, 143,

4 ~~ Floaters in a Sparrowhawk population 445 R less than, or equal to, that observed in 5 series of simulated data with counts subject to estimation error. This gave P =.7, i.e. strong evidence of density dependence. Hence, we conclude that, even if up to 1% of nests were missed each year (which seems to us unlikely), this would not have affected our conclusion that the pattern of year-to-year fluctuations in nest numbers reflected density dependence. In other words, we could safely conclude that breeding density was regulated. Estimated numbers of floaters On both methods of estimation, the numbers of female floaters varied greatly from year to year, but on average the estimates were much higher on Method A than on Method B (which assumes that all birds bred in their first or second year). On Method A the mean annual value was about 29 birds or.9 female floaters per female breeder, while on Method B the mean was about 9 birds or.28 female floaters per female breeder (Table 2). Clearly, the way that third-year birds were treated in the analyses made a big difference to the estimated size of the floating population. Any analysis of density dependence in the estimated floating population is problematic because errors of estimation could be relatively large and hard to quantify. In consequence, P-values are likely to be too low, leading to spurious statistical significance. Moreover, on Method B, the estimates of floater numbers are low, with one zero value (Table 2). On Method A, Bulmer s test for density dependence in floater numbers gives R =.812, P <.5, and on Method B it gives R =.447, P <.1. Bearing in mind that these P-values are likely to be too small, the estimates from Method A show little evidence for density dependence but those on Method B are more suggestive. Density dependence and inter-relationships between demographic variables In an attempt to find how estimated floater numbers varied in relation to reproduction and other aspects of population demography, we examined some relevant relationships in Table 3. Using estimates for female floater numbers from both methods, statistically significant negative correlations emerged between the per capita breeding output for the total female population (i.e. YJ(B, + FJ) and the total female population (Br + FJ, where for year t, Y, denotes the total number of young produced, and B, and F, denote the numbers of female breeders and floaters respectively. This rela- Table 2. Estimates of female floater numbers calculated in two ways: (A) taking records at face value and assuming new 3-year breeders in the study area had not bred before, and (B) assuming all 3-year breeders had bred before (i.e. all birds bred for the first time in their first or second year). Year Mean sd se NO.^ k13.9 k3.2 Method A No. per breeder r.46 ro.1 aestimated to nearest whole number. NO.^ Method B No. per breeder k4.21 k.15 k.97 k.3 tionship implied that reproduction was density dependent with respect to the total size of the population, a finding that might be expected in a system in which Table 3. Correlations between different demographic variables for both methods of estimating the floating female population. Relationship Method A Method B -.12 ns.2 ns.27 ns -.4 ns -.66**.15ns -.39 ns -.62 ns.31 ns.36 ns -.12 s.72**.17 ns S ns.35 ns.3 ns.31 ns.11 s B,, breeding ( = nest) numbers year t; F,, estimated floater numbers year t; Y,, total young produced year f; S,, adult survival year t; PI,, proportion of 1-year olds in breeding population year f; MI,, number of 1-year olds in breeding population year t; F,, = estimated 1-year olds in floating population year t. P <.5; **P <.1; **P < British Ornithologists Union, Ib~s, 143,

5 Newton & P Rothery breeder numbers remained fairly stable, while floater numbers fluctuated greatly from year to year. However, the interpretation of this correlation is complicated by the presence of F, in both the x-variable and y-variable, so that any measurement error in F, would induce a negative correlation. This could explain why the correlation is higher for Method A (which gives higher F, values) than for Method B. Using the floater estimates from Method B, a significant positive correlation emerged between the breeding female population and the total female population. However, this could again simply be a reflection of the fact that the number of female breeders occurs in both the x-variable and y-variable, and that by Method B, estimated floater numbers form a relatively small proportion of the total of breeders plus floaters. No significant relationships emerged between reproduction or adult female survival in one year and estimated floater numbers in the following year, nor between the proportion of 1-year-old females in the breeding female population and the estimated numbers of female floaters present in the same year. The absence of statistically significant correlations between the estimated number of female floaters and the total production of young in the previous year was surprising. However, the correlation depends on the variation in juvenile survival and on the sampling errors in the estimated number of floaters, both measured relative to the \,ariation in production of young. Such variation could have reduced the strength of any correlation that might otherwise have existed between floater numbers and breeding success. Also, in a relatively short series, sampling errors in the estimated correlations were relatively large which further limited the scope for detecting an effect. The same points could be made about the lack of significant correlation between production of young and the estimated total number of first-year birds (floaters plus recruits) in the following year. Consistency of estimates of floater numbers with other demographic variables One way to check the likely accuracy of the estimates of floater numbers was to find how consistent they were with other, more precisely measured, demographic parameters, assuming a closed population. Again the calculation refers to females only. In the simplest case, the size of the floating population in time t + 1 is obtained as the sum of the surviving floaters from year t plus the number of surviving offspring from year t minus the number of birds recruited into the breeding population to replace breeders which died during year t: Where for year t, F, denotes the number of female floaters, B, denotes the number of female breeders, Yr denotes the total number of young produced, s,, denotes the average survival of female floaters (which remained in the floating population), s,, denotes the average survival of female juveniles (fledging to April of first year), and siit denotes the average annual survival of female breeders. For the above model, a linear regression through the origin of the variable F,,, + (1 ~ sbr) B, against the variables F, and Y, gave estimates of the floater and juvenile survival rates. These could then be compared with values based on other data. By Method A, mean floater survival (+se) was calculated at.33 (k.2) and mean juvenile survival (.se) from fledging to the first April of life at.74 (1_.14). By Method B, the equivalent estimates were.52 (k.26) and.38 (ko.6). By Method B the survival of floaters calculated in the model at.52 is close to the value of.49 for survival of 1 -year-old breeders using capture-recapture data (although the standard error is large). Also, the estimated juvenile survival calculated in the model at.38 is close to the value of.37 derived from the proportion of birds recovered dead in their first year of life (see below), allowing ten years (the maximum recorded lifespan) for all recoveries to come in. Overall, then, this fairly crude analysis supports the estimates of floater numbers based on Method B, because they compare more favourably than those from Method A with independent estimates of survival obtained previously in other ways. Simulation models for estimating the f1oater:breeder ratio A further test of the likely accuracy in estimates of floater numbers is to see how these numbers compare with those expected from the measured survival and breeding rates of breeders. In a closed population, the potential size and structure of the floating component would depend on the pattern of age-dependent survival and breeding success, and on the relative recruitment rates of I-year, 2-year and 3-year birds from among the floaters. Consider a simple case in which older birds have priority in recruitment, i.e British Ornithologists' Union, Ibis, 143,

6 Floaters in a Sparrowhawk population 447 year-olds become breeders only if spaces remain after recruitment of 3-year-olds, and 1 -year-olds only if spaces remain after recruitment of 2- and 3-year birds. Assume also that the size of the breeding population is fixed. On this basis, we calculated the equilibrium breeding and floating populations using the known Eskdale pattern of age-dependent survival and breeding success (from Newton & Rothery 1997). Breeding success in the study area could be measured accurately (annual mean 2.72 young per female, sd = k.4 se = +.9), but the survival estimates (like all capture-recapture estimates) measured continued residence in the study area. Because breeders that left the study area could not be retrapped, the estimates obtained fell short of actual survival. We therefore examined what would be the effect on estimated floater numbers by raising in the model the survival estimates of all age groups by 1% (i.e. multiplied by a factor of 1.1). In our judgement, 1% is the maximum discrepancy likely between measured and actual values. The estimate of juvenile survival (fledging to first April) is based on recoveries of nestlings ringed in the study area and later found dead. From a total of 15 dead female recoveries, 66 were in this stage of life. Assuming that the reporting rate is the same for all ages, the juvenile female mortality is estimated as 66/15 =.63, with corresponding survival of.37. This figure may be an underestimate because the reporting rate in raptors is sometimes higher for young birds than for older ones (Newton 1979, Kenward et al. 1999). So further simulations were carried out after increasing the value by 1% to.41. Simulations using the estimated pattern of survival in female breeders, together with juvenile female survival rates of.37 or.41, showed that the population could not sustain itself. Increasing the survival rates of breeding females by 1% produced a viable breeding population with a surplus of floaters. Table 4 shows the resulting age-structure in the breeding and floating segments of the population, under both estimates of juvenile survival, while Table 5 gives a summary of the population characteristics at equilibrium, again under both estimates of juvenile survival. The floater:breeder ratio is evidently very sensitive to changes in juvenile survival. However, the values for the floater:breeder ratio at equilibrium are much closer to those obtained using Method B than to those using Method A (compare estimates in Table 5 with those in Table 2). Using the data in Table 1, the average proportion of first-year breeders was calculated as.18 (range -.43), which is close to the figure of.17, estimated in Table 5 under a juvenile survival (s,) of.41. Similarly, from Table 1 the average proportion of 2- year-olds found breeding for the first time was calculated as.14 (range -.26). The corresponding figure in the model population is obtained as the proportion breeding as 2-year-olds minus those 1 -year-old breeders which survived to become 2-year breeders. This gives:.24 - (.3 x.535) =.8 for s, =.37;.28 - (.17 x.535) =.189 for sj =.41. The figures for mean breeding success estimated in the model as 2.41 and 2.57 for each value of juvenile survival are both close to the measured value of 2.72.Similarly, the average survival values in the model Table 4. Estimated female breeder and floater population densities using model scenarios based on age-dependent values for survival and breeding success in Eskdale. Age-dependent survival rates obtained as 1.1 times the Eskdale estimates to allow for emigration of breeders from the study area. Juvenile survival values taken as.37 based on ring recoveries of dead birds and as.41 to allow for possible underestimates from likely higher reporting rates of first-year birds. Total breeder density held constant at unit density, and floater densities at different ages estimated accordingly. Juvenile survival.37 Observed young ~. raised per pair Calculated annual Floater Breeder Age per year survival x 1.1 density density Juvenile survival.41 Floater Breeder density density British Ornithologists' Union, Ibis, 143,

7 Newton & I? Rothery Table 5. Summary of female population characteristics at equilibrium on two different rates of juvenile survival (fledging to first April). Juvenile survival Floater: breeder ratio Proportion 1 -year old breeders.3.17 Proportion 2-year old breeders Proportion of 1 -year olds which bred Mean breeding success (young per pair) Mean annual survival of breeders of.62 and.64 are both close to the value of.59 estimated in an earlier analysis for female Sparrowhawks in Eskdale using capture-recapture data over a shorter period of years (Newton et al. 1993). In general, therefore, this simulation based on data from the study population again gives estimates of female floater numbers which are more consistent with the female floaterbreeder ratio obtained under Method B than with that obtained under Method A. Stochastic models for estimating the f1oater:breeder ratio The above models are deterministic so that they settle down to fixed numbers of breeders and floaters. To get some idea of possible stochastic effects, we ran the model with mean breeding success varying between years by randomly selecting from the observed values. Such simulations showed that variable breeding success could induce substantial annual variation into the observed mean floater:breeder ratio. The standard deviations on the annual estimates were fairly stable and similar to the observed value of.14 using Method B. Overall, therefore, it looks as though the other data gathered for this population are more consistent with the female floaterbreeder ratio of.28 produced under Method B than with the higher value of.92 produced under Method A. DISCUSSION Our analysis used data from females because, since they are easier to catch than males, their demographic parameters could be estimated more precisely. However, Sparrowhawks breed monogamously, and overall survival rates of males do not differ significantly from those for females (Newton et al. 1993), so we have no reason to suppose that conclusions based on females alone cannot be extrapolated to males. An assumption in the estimation of floater numbers was that survival rates measured on breeders over years 1-2 and years 2-3 of life were also applicable to nonbreeders. We could not check this assumption, but by analogy with some other bird species (e.g. Smith 1976, Winker et al. 199), survival of non-breeders might have been lower. The effect of such an error would have been to produce slightly inflated estimates of the numbers of floaters. In any case the size of the floater population, estimated from the age at recruitment to the breeding population, is somewhat notional, for we have no independent estimate of numbers. Some floaters may have moved into the area just before their first breeding. On Method A, the mean ratio of female floaters: breeders was.92 and on Method 8 (assuming all birds had bred before their third year), the mean ratio was about.28. On both methods the ratio varied greatly from year to year, which was consistent with known variations in breeding success. In an age-structured population model, the floater:breeder ratio estimated by Method B was most consistent with other measured parameters, notably juvenile and 1-2 year survival rates. On this model, all birds would have bred for the first time in their first or second year, and all floaters would have been first-year birds. Another check came from Sparrowhawk carcasses, the oviducts in which revealed the proportions that had laid by different ages wyllie & Newton 1999). Among 47 females examined in their second winter of life, 43% had a convoluted oviduct, implying that they had laid in their first year. This figure lies between the estimates of 32% and 67% obtained in our population model assuming juvenile survival rates of.41 and.37, respectively. Similarly, among the carcasses of 1 18 females examined in their third and later winters of life, 97% had laid (implying that practically all females had laid in their first or second year). This was again more consistent with the scenario envisaged in Method B than in Method A. The carcasses came from various parts of Britain, with very few from the study area itself. The findings from Method B were thus consistent with other information. On the findings from Method A, the population of the study area would not have been self-sustaining. While the study area might have been a 'sink', with breeder numbers and high floater numbers maintained by net immigration, this seemed unlikely from the relatively high quality of the local habitat. 21 British Ornithologists' Union, Ibis, 143,

8 Floaters in a Sparrowhawk population 449 In the regulation of the Eskdale Sparrowhawk population, two main density dependent processes emerged, confirming earlier findings (Newton 1986). First, nest numbers fluctuated from year to year within fairly narrow limits, and in a manner expected of a regulated population. Almost certainly, this regulation was achieved by the territorial behaviour of the breeding pairs which resulted in their uniform and consistent spacing through the woodland of the study area. This behaviour also helped to limit the number of new recruits that could be added to the breeding population each year. To some extent, new breeders occupied gaps left by the deaths or emigration of existing breeders, so that recruitment was inversely density dependent with respect to the numbers of surviving breeders (Newton 199 1). In the second density dependent relationship, the numbers of young produced per nest each year was inversely correlated with the total population of breeders plus floaters. The greater the total population size, the lower the per capita production of young. This relationship arose because the breeders were fairly stable in numbers (and had limited breeding output), while the floater numbers fluctuated greatly from year to year (and had no recorded breeding output). The number of young produced per nest was not density dependent with respect to breeding numbers alone. Only slight and unconvincing evidence was obtained that floater numbers fluctuated in a density dependent manner, and no evidence was found that the annual survival of breeders was density dependent with respect to either breeding numbers or to the total numbers of breeders plus floaters. Hence, the regulation of the population could have been achieved by the territorial behaviour of breeders (in relation to habitat and food-supply), and floater numbers could have been determined primarily as the balance between inputs (from reproduction) and outputs (from deaths and entry to the breeding sector), modified by movements in and out of the area. Bulmer, M The statistical analysis of density dependence. Biometrics 31: Hunt, W.G Raptor floaters at Moffats equilibrium. Oikos 82: Kenward, R.E Radio transmitters tail-mounted on hawks. Ornis Scand. 9: Kenward, R.E., Marcstrom, V. & Karlbom, M Demographic estimates from radiotagging: models of age-specific survival and breeding in the Goshawk. J. Anim. Ecol. 68: Marquiss, M. & Newton, Habitat preference in male and female Sparrowhawks. lbis 124: Newton, Population Ecology of Raptors. Berkhamsted: Poyser. Newton, The Sparrowhawk. Calton: T. & A.D. Poyser. Newton, The role of recruitment in population regulation. Proc. lnt. Ornithol. Congr. 2: Newton, Experiments on the limitation of bird numbers by territorial behaviour. Biol. Rev. 67: Newton, I Population Limitation in Birds. London: Academic Press. Newton, Marquiss, M Removal experiments and the limitation of breeding density in Sparrowhawks. J. Anim. Ecol. 6: Newton, I., Marquiss, M., Weir, D.N. & Moss, D Spacing of Sparrowhawk nesting territories. J. Anim. Ecol. 46: Newton, 1. & Rothery, P Senescence and reproductive value in Sparrowhawks. Ecology 78: Newton, I., Wyllie, Mearns, R.M Spacing of Sparrowhawks in relation to food supply. J. Anim. Ecol. 55: Newton, I., Wyllie, 1. & Rothery, P Annual survival of Sparrowhawks Accipiter nisus breeding in three areas of Britain. lbis 135: Smith, S.M Ecological aspects of dominance hierarchies in Black-capped Chickadees. Auk 93: Smith, S.M The underworld in a territorial sparrow: adaptive strategy for floaters. Am. Nat. 112: Smith, J.N.M., Arcese, P. 81 Hochachka, W.M Social behaviour and population regulation in insular bird populations: implications for conservation. In Perrins, C.M., Lebreton, J.-D. & Hirons, G.J.M. (eds) Bird Population Studies: Relevance to Conservation and Management; Oxford: Oxford University Press. Winker, K., Rappole, J.H. & Ramos, M.A Population dynamics of the Wood Thrush in southern Veracruz, Mexico. Condor 92: Wyllie, 1. & Newton, Use of carcasses to estimate the proportions of female Sparrowhawks and Kestrels which bred in their first year of life. lbis 141: REFERENCES Brown, J.L Territorial behaviour and population regulation in birds. Wilson Bull. 81 : Received 25 October 1999; revision accepted 22 August 21 British Ornithologists Union, lbis, 143,