SEARCH

SEARCH BY CITATION

Keywords:

  • apparent survival probability;
  • capture–mark–recapture;
  • conservation ecology;
  • declining population;
  • demographic variation;
  • mortality

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1. Determining the precise timing and location of major demographic bottlenecks, such as periods of low survival, is key to identifying ecological causes of variation in population growth rate. Such understanding is key to designing efficient and effective mitigation.

2. In a protected population of red-billed chough Pyrrhocorax pyrrhocorax on Islay, Scotland, variation in population growth rate largely reflects among-year variation in first-year survival. First-year survival was unprecedentedly low during 2007–2010, threatening population viability.

3. We used colour-ring resightings to estimate monthly survival probabilities (Φm) throughout the first year from fledging for eight chough cohorts (totalling 519 individuals) representing the full observed range of variation in first-year survival. We thereby identify the time and location of recent low survival.

4. On average across all cohorts, Φm varied among months, being low during the first month after ringing (May–June, accounting for c. 24% of all first-year mortality) and high during the last 4 months of the first year (January–May, accounting for c. 6% of all first-year mortality). Most mortality (c. 70%) occurred after fledglings dispersed from natal territories.

5. The 2007–2009 cohorts experienced low Φm during July–December. This represents an additional low survival period compared to previous cohorts rather than decreased Φm across all months or further decreases through periods when Φm was low across all cohorts.

6.Synthesis and applications. These data have general relevance in showing that dramatically low annual survival, which needs to spark rapid management action, can reflect different and unanticipated periods of low survival rather than exaggeration of typical variation. With specific regard to conserving Islay’s chough population, our data show that sub-adult survival has recently been low during July–December, probably reflecting conditions on key grassland foraging areas. Managers aiming to increase population viability should increase invertebrate diversity, abundance and availability at these times and locations, thereby increasing foraging options available to choughs.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Diagnosing the exact demographic and ecological causes of variation in population growth rate (λ) is key to understanding population dynamics and hence to effective population management (Caughley 1994; Benton & Grant 1999; Franklin et al. 2000; Dobson & Oli 2001). Demographic studies of birds and mammals have shown that among-year variation in λ is often largely caused by among-year variation in juvenile or sub-adult survival, defined as survival from fledging or weaning through the year(s) prior to breeding (Peach, Siriwardena & Gregory 1999; Gaillard et al. 2000; Reid et al. 2004; Robinson et al. 2004; Morrison & Hik 2007). This is because sub-adult survival can vary substantially among years, and because λ can be sensitive to this variation (Franklin et al. 2000; Gaillard et al. 2000; Dobson & Oli 2001; Oli & Armitage 2004; Reid et al. 2004; Morrison & Hik 2007; Frederiksen et al. 2008). In such circumstances, understanding the causes of low or variable sub-adult survival is crucial to explaining observed population dynamics and designing appropriate mitigation (van der Jeugd & Larsson 1998; Peach, Siriwardena & Gregory 1999; Robinson et al. 2004; Wiens, Noon & Reynolds 2006; Yackel Adams, Shagen & Savidge 2006; Martín et al. 2007; Reid et al. 2008).

One valuable step towards diagnosing and managing the causes of low or variable sub-adult survival is to quantify the pattern of within-year variation in survival probability. For example, evidence that sub-adult survival was lowest during or soon after fledging or weaning (Fig. 1a) would focus attention on the causes of mortality at or near natal locations during the breeding season. Evidence that survival was lowest during or after dispersal to new foraging areas or several months subsequently (Fig. 1b,c) would focus attention on causes of mortality at these different locations and seasons. Alternatively, survival may be relatively constant throughout the sub-adult period (Fig. 1d), whether reflecting constant or varying causes, potentially implying that year-round management may be required or that the precise timing of short-term intervention may be less critical. Such patterns should ideally be quantified and compared across multiple cohorts experiencing different overall sub-adult survival rates. It would then be possible to determine whether all cohorts experience low survival at consistent times of year (with among-cohort variation in overall survival reflecting variation in the magnitude of this bottleneck, Fig. 1a), whether different cohorts experience low survival at different times of year (Fig. 1b), whether cohorts with low overall survival experience additional periods of low survival compared to cohorts with high overall survival (Fig. 1c), or whether different cohorts simply have consistently different survival probabilities (Fig. 1d). These different scenarios, which are not all mutually exclusive, would require management strategies that differ in their nature, timing and degree of among-year flexibility. Rigorous data describing among-cohort variation in the pattern of within-year variation in sub-adult survival are therefore required to inform targeted ecological research and management responses.

image

Figure 1.  Hypothetical scenarios of within- and among-year variation in sub-adult survival probability. Solid and dashed lines illustrate survival probability per unit time for two different cohorts. Among-cohort variation in overall sub-adult survival could reflect variation in (a) the magnitude of a common low survival period, (b) the magnitude and timing of low survival, (c) the number of low survival periods or (d) constant survival probability.

Download figure to PowerPoint

Such data, however, are remarkably scarce, precluding either specific or general conclusions regarding the timing, location and mechanism of sub-adult mortality. This reflects the logistical difficulty of regularly monitoring the survival of sufficient individuals throughout the sub-adult period. Individuals are often only marked some time after fledging or weaning, meaning that mortality during or immediately after fledging or weaning remains unquantified (e.g. Peach, Siriwardena & Gregory 1999; Martín et al. 2007). Conversely, studies that focus on measuring survival during the immediate post-fledging or post-weaning period often only follow individuals for a few days or weeks rather than full years and hence cannot quantify variation in survival probability occurring after the initial period (e.g. Anders et al. 1997; Naef-Daenzer, Widmer & Nuber 2001; Yackel Adams, Shagen & Savidge 2006). Studies that do continue often only monitor individuals infrequently, meaning that the precise timing of mortality still cannot be pinpointed (e.g. van der Jeugd & Larsson 1998; Ward et al. 2004). Many studies cannot distinguish mortality from dispersal, but these processes have critically different implications for large scale population dynamics and management. Even without dispersal, variation in the probability of encountering a surviving individual must be explicitly quantified to ensure that estimated variation in survival is not biased by variation in encounter probability. Furthermore, due to accumulating mortality, sample sizes for individual cohorts will decrease with time from fledging or weaning, meaning that power to detect variation in survival will also decrease (Wiens, Noon & Reynolds 2006). High encounter probabilities and sufficient initial sample sizes are therefore imperative. Identifying the basis of among-cohort variation in sub-adult survival, and hence variation in λ, therefore requires that sufficient individuals are marked at or prior to fledging or weaning and encountered with high probability at regular intervals throughout the period of interest, and that such comprehensive studies are repeated across multiple cohorts experiencing a wide range of variation in overall sub-adult survival. However, the most comprehensive studies in terms of duration, monitoring frequency and encounter probability typically follow relatively few individuals from few cohorts, meaning that the magnitude and basis of among-cohort variation in survival, which may be critical to understanding and managing population dynamics, cannot be accurately quantified (e.g. Ward et al. 2004; McIntyre, Collopy & Lindberg 2006; Wiens, Noon & Reynolds 2006; Daunt et al. 2007; Kerbiriou & Julliard 2007; Martín et al. 2007).

Accordingly, we quantified among-cohort variation in monthly sub-adult survival probabilities in a protected population of red-billed choughs Pyrrhocorax pyrrhocorax on Islay, Scotland, where among-year variation in sub-adult survival accounts for a substantial proportion of variation in λ and recent low sub-adult survival is causing heightened conservation concern.

Choughs are listed on Annex 1 of the EU Wild Birds Directive, reflecting their restricted distribution, small population sizes and multiple perceived threats, and are consequently the focus of conservation management across their European range (e.g. Bignal, Bignal & McCracken 1997; Blanco, Tella & Torre 1998; Johnstone, Whitehead & Lamacraft 2002; Kerbiriou et al. 2006; Kerbiriou & Julliard 2007). Islay’s resident chough population has been monitored since 1981, allowing the demographic basis of variation in λ to be quantified (Bignal, Bignal & McCracken 1997; Reid et al. 2004). During 1983–2006, the probability of survival from fledging to age one (first-year survival, Φ1) varied among years, from c. 0·20 to 0·65 (Fig. 2), and accounted for a substantial proportion of among-year variation in λ (Reid et al. 2004, 2008). However, Φ1 has subsequently decreased substantially; mean Φ1 for the 2007–2009 cohorts was only 0·10, compared to 0·42 for the 1983–2006 cohorts (Fig. 2). A matrix projection model for this population predicts λ ≈ 1·00 and hence stable population size given the mean demographic rates observed during 1983–2006, but only λ ≈ 0·87 and hence rapid population decline given these same rates but Φ1 = 0·10 (as observed during 2007–2009, Appendix S1, Supporting information). Even optimistic scenarios of increased breeding success and second-year and adult survival still predict λ < 1·0 given Φ= 0·10, and the degree to which mean breeding success and adult survival would have to increase to achieve population stability is probably unrealistic (Appendix S1, Table S1 and Fig. S1). Continued low Φ1 is therefore likely to cause rapid decline of Islay’s chough population. Indeed population size decreased c. 20% between censuses in 2007 and 2010, from c. 55 to c. 45 pairs, largely reflecting low recruitment from the 2007 and 2008 cohorts. Understanding the causes of among-year variation in Φ1, and particularly of the recent low Φ1, is therefore an immediate conservation priority. However, a statistical model that explained c. 80% of among-year variation in Φ1 during 1983–2005 as a function of weather and tipulid larvae abundance (Reid et al. 2008) did not predict the low Φ1 observed during 2007–2009 (Fig. 2), suggesting that different and as yet unidentified constraints are now operating. Quantifying the pattern of within-year variation in sub-adult survival, and hence determining the exact timing and location of high recent mortality, is a key first step towards to generating and testing appropriate hypotheses explaining this discrepancy and designing appropriate mitigation.

image

Figure 2.  Estimated first-year survival probabilities (Φ1) for chough cohorts fledged in 1983–2009 inclusive (symbols and solid line, extended from Reid et al. 2004, 2008), and predicted Φ1 for each cohort based on weather and tipulid abundance (dashed line, extended from Reid et al. 2008). Filled symbols indicate the eight cohorts for which monthly survival probabilities (Φm) were estimated (Fig. 3). Estimates are shown ± 1SE.

Download figure to PowerPoint

We used colour-ring resightings to estimate monthly survival probabilities for eight chough cohorts throughout the first year from fledging. These cohorts, fledged in 1984, 1985, 1986, 2005, 2006, 2007, 2008 and 2009, encompass the full range of variation in Φ1 observed during the long-term demographic study, including the recent low rates (Fig. 2). We first estimated mean survival through each month across all eight cohorts combined, thereby determining whether survival probability varied among months, was particularly low in specific months, or was relatively consistent across all months (Fig. 1). Secondly, we estimated monthly survival probabilities for all eight cohorts individually, thereby quantifying whether among-cohort variation in overall first-year survival reflected variation in the magnitude and/or timing of low survival periods and identifying the key periods concerned. We demonstrate that the recent low Φ1, which threatens population viability, primarily reflected an additional and unexpected period of low survival that was not experienced by previous cohorts, and discuss the implications for targeted mechanistic research and efficient population management.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study system

Choughs feed primarily on soil invertebrates and are closely associated with agricultural systems where mixed, low-intensity grazing creates diverse grassland habitats (Bignal & McCracken 1996; Bignal, Bignal & McCracken 1997; Blanco, Tella & Torre 1998; Johnstone, Whitehead & Lamacraft 2002; Kerbiriou et al. 2006). Each year, breeding has been monitored at a sample of Islay nest sites; mean breeding success is 2·0 ± 0·2(SE) fledglings per attempt (Reid et al. 2004). In mid-late May each year during 1983–2009, c. 2 weeks before fledging, chicks were marked with unique combinations of coloured rings. Resightings of ringed individuals collected during spring and summer each year during 1984–2010 allowed annual apparent survival (Φ) and resighting (p) probabilities to be estimated using capture–mark–recapture (CMR) approaches based on a 15 April–15 July encounter period (Fig. 2, Reid et al. 2003, 2004, 2008). Islay-ringed choughs are extremely rarely observed elsewhere, even in an adjacent population on Colonsay (c. 8 km away). Dispersal from Islay is therefore probably rare, meaning that variation in Φ primarily reflects variation in survival rather than emigration (Reid et al. 2003).

Monthly resightings

In most years restricted observer availability meant that autumn and winter resightings were insufficient to quantify among-month variation in sub-adult survival within each year. However, for eight cohorts (519 individuals) fledged in 1984–1986 and 2005–2009 choughs were resighted each month throughout the first year post-ringing, allowing monthly apparent survival probabilities (Φm) to be estimated while controlling for monthly resighting probabilities (pm). Choughs remain on their natal territories for some days or weeks post-fledging before moving to traditional sub-adult foraging and roosting areas where they remain until acquiring breeding territories, typically aged 2–3 years (Bignal, Bignal & McCracken 1997; Reid et al. 2003). On ≥ 1 day during approximately 15th–20th of each month, all or most sub-adult areas were searched for colour-ringed individuals. Breeding territories were also searched during the months immediately post-fledging (typically June–July). This regime resulted in a generally high pm (although resightings in some months were less comprehensive due to adverse weather and/or reduced observation effort), providing appropriate data for discrete-time survival analysis.

Analyses

Individual encounter histories were compiled from May in each individual’s natal year (the month of ringing) to May the following year, comprising 13 occasions and 12 monthly intervals. However, since Φm and pm for the final interval and occasion cannot be independently estimated when fully parameterised models are fitted, an additional capture occasion was included for each individual, coded as one and zero if an individual was ever or never seen after age one. This specification maximizes the final resighting probability and hence power to estimate Φm throughout the first year. The additional Φp product is a nuisance parameter with no useful biological interpretation, and estimates are not reported.

CMR models were fitted (using program MARK) for all eight cohorts individually and across all eight cohorts combined. Initial models included full month- and cohort-dependence in both Φm and pm. Since observation effort and conditions varied among months and models with constant pm were poorly supported, full month- and cohort-dependence in pm were retained in all final models. To test whether Φm varied among months, both within each cohort individually and across all cohorts combined, the support for models where Φm was constrained to be constant was compared with that for fully month-dependent models, and for models with random (rather than fixed) month effects. However, since Φm was sometimes estimated as 1·0, interpretation of random effects models was uncertain. These models were therefore considered as subsidiary evidence, and process variance in Φm was not formally estimated. Sample sizes for some months were small, particularly in later months for cohorts that experienced low initial Φm. Statistical power to detect among-month variation in Φm was then low. However, since the monitored choughs comprised substantial proportions of each cohort, observed variation in Φm is biologically relevant in the context of Islay’s population, and our initial aim was to estimate Φm rather than test specific hypothesised patterns of variation. After quantifying variation in Φm within each individual cohort and across all cohorts, we then explicitly quantified how Φm differed between the recent (2007–2009) cohorts that experienced very low overall Φ1 and the previous (1984–1986 and 2005–2006) cohorts that experienced higher overall Φ1 (Fig. 2).

Bootstrap goodness-of-fit tests showed that final models fitted the data (> 0·11) with relatively little overdispersion (median variance inflation factor ≤ 1·22). Models were judged better supported than rivals if the difference in AIC, adjusted for small sample sizes (ΔAICc), exceeded 2·0 (Burnham & Anderson 1998). Full details are available as online Supporting information Appendix S2, Table S2 and S3. Since full month-dependence in pm was retained in all models, statistics describing the distribution of estimated pm are presented (means ± 1SE). Statistics for Φm were estimated directly from constrained CMR models. Effects of sex on Φm were not modelled because fledgling sexes were generally unknown. To visualize the proportion of each cohort estimated to be alive in each month after ringing (equalling the probability that an individual will survive to a particular month), cumulative survival curves were calculated as the product of successive estimates of Φm for each individual cohort and all eight cohorts combined. Monthly survival rates have previously been estimated for 1983–85 (Bignal et al. 1987). However, this study considered fledged individuals in years of unusually high Φ1 (Fig. 2), and did not formally control for variation in encounter probability.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Totals of 54, 92, 70, 53, 57, 62, 61 and 70 fledglings were colour-ringed in 1984, 1985, 1986, 2005, 2006, 2007, 2008 and 2009, respectively, comprising c. 50–75% of each fledgling cohort.

Resighting probability

Mean pm was 0·73 ± 0·03 across all 96 observation months (median 0·82, IQR 0·54–0·95), and pm varied from 0·17 in March 2005 to 1·00 in 21 different months (Fig. 3). Mean pm was 0·53 ± 0·06, 0·55 ± 0·06, 0·59 ± 0·09, 0·65 ± 0·08, 0·96 ± 0·01, 0·88 ± 0·04, 0·75 ± 0·06 and 0·88 ± 0·04 for the 1984–2009 cohorts, respectively. Across all eight cohorts, mean pm was reasonably consistently high across all 12 calendar months, varying from 0·63 to 0·81 (Fig. 4). Therefore, although pm varied substantially among individual months it was > 0·15 in all 96 months, > 0·5 in 76 (79%) months and > 0·8 in 49 (51%) months, providing sufficient power to estimate Φm for all months considered.

image

Figure 3.  Estimated monthly resighting probabilities (pm, open symbols & dashed lines) and apparent survival probabilities (Φm, filled symbols & solid lines) for chough cohorts fledged in 1984, 1985, 1986, 2005, 2006, 2007, 2008 and 2009. Estimates are shown ± 1SE. Y-axis scales are standardized to facilitate comparison among cohorts. X-axis labels indicate the observation month and end of each survival period (hence ‘May’ indicates pm in May and Φm for April–May).

Download figure to PowerPoint

image

Figure 4.  Mean estimated monthly resighting probabilities (pm, open symbols & dashed lines) and apparent survival probabilities (Φm, filled symbols & solid lines) estimated across all eight chough cohorts combined. Estimates are shown ± 1SE. Mean pm is the mean of the eight cohort-specific estimates for each month. Mean Φm is the global modelled estimate for each month. X-axis labels indicate the observation month and end of each survival period (hence ‘May’ indicates pm in May and Φm for April–May).

Download figure to PowerPoint

Survival probability

Across all eight cohorts combined, a model with month-specific Φm was much better supported than a model with constant Φm across all months (ΔAIC= 40·0). Overall Φm therefore varied among months, from c. 0·83 to 0·98, being highest during June–July and January–May and lowest during May–June and November–January (Figs 4 and 5).

image

Figure 5.  Cumulative survival curves for all cohorts combined (black line) and each individual cohort (1984: dark blue; 1985: light blue; 1986: green; 2005: grey; 2006: purple; 2007: red; 2008: orange; 2009: yellow). The box indicates the approximate timing of dispersal from natal territories to sub-adult flocks. Estimates of overall first-year survival differ quantitatively from those in Fig. 2 due to different encounter periods, but show qualitatively similar variation.

Download figure to PowerPoint

However, these estimates of Φm across all eight cohorts masked substantial among-cohort variation in month-specific survival (Figs 3 and 5). Mean Φm was 0·90 ± 0·01 across all 96 months, and varied from 0·56 in Nov–Dec 2007 to 1·00 in 26 different months (Fig. 3). Models with full among-month variation in Φm were better supported than models with constant Φm for the 1985, 1986 and 2009 cohorts, demonstrating that Φm varied among months for these cohorts (ΔAICc≥ 2·2). Models with random month effects were as well or better supported than constant Φm models for the 1984, 1985, 1986, 2007, 2008 and 2009 cohorts, further suggesting that Φm varied among months for these cohorts. In contrast, models with full among-month variation in Φm were poorly supported for the 2005 and 2006 cohorts (ΔAICc≥ 14·9), as were random effects models. However, even for these two cohorts, Φm varied from 0·80 to 1·0 and 0·84 to 0·98, respectively. The lack of statistical support for among-month variation may therefore reflect low power (particularly for 2005 given the relatively small cohort size and low pm). Overall, therefore, there was evidence that Φm varied among months for most individual cohorts.

A model that combined the eight cohorts into high (1984, 1985, 1986, 2005, 2006) and low (2007, 2008 & 2009) survival groups with full month-dependence in Φm within each group was much better supported than models that included all eight cohorts individually (ΔAIC= 67·0) or that combined all eight cohorts into a single group (ΔAIC= 83·1). Estimates suggested that Φm tended to be lower for the low-survival cohorts than the high-survival cohorts during May–January, but that Φm was similar across both groups during January–May (Fig. 6). However, models where Φm differed between the high- and low-survival cohorts during May–June and June–July were no better supported than constrained models where Φm did not differ between the two groups (ΔAICc≤ 0·9). There was therefore no strong statistical evidence that Φm differed between high- and low-survival cohorts during the first 2 months after ringing. Models where Φm was constrained to be the same across the high- and low-survival cohorts during October–November and each month during December–May were to some degree better supported than models where Φm differed between the two groups (ΔAICc≥ 1·0). There was therefore no evidence that Φm differed between high- and low-survival cohorts during these months. Since estimates of Φm for the two groups during December–May were very similar (Fig. 6), this reflected small effect size rather than solely low statistical power due to the relatively small number of individuals that survived after December. In contrast, for each month during July–October and November–December, models where Φm was constrained to be the same across the high- and low-survival cohorts were less well supported than models where Φm differed between the two groups, demonstrating that Φm was higher for the high-survival cohorts than the low-survival cohorts during these months (Fig. 6, ΔAICc≥ 3·4).

image

Figure 6.  Estimated monthly apparent survival probabilities (Φm) for high survival cohorts (1984–2006, filled symbols) and low survival cohorts (2007–2009, open symbols). X-axis labels indicate the end of each survival period (hence ‘May’ indicates Φm for April–May).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

A major step towards successfully managing the dynamics of any population is to determine the identity and ecological causes of demographic constraints on λ (Caughley 1994; Franklin et al. 2000). When sub-adult survival is identified as one such constraint (as in choughs), quantifying temporal variation in sub-adult survival probability, both within and among cohorts, is key to pinpointing periods and locations of high and low survival, to developing and testing biological hypotheses explaining observed variation, and hence to designing efficient management that maintains or increases survival through bottleneck periods or locations (van der Jeugd & Larsson 1998; Ward et al. 2004; McIntyre, Collopy & Lindberg 2006; Wiens, Noon & Reynolds 2006; Yackel Adams, Shagen & Savidge 2006; Martín et al. 2007). However, few studies have quantified within-year variation in sub-adult (or indeed adult) survival with high temporal definition across multiple cohorts experiencing different overall survival rates, precluding identification of specific or general bottlenecks and hence judicious targeting of research and management effort. Our analyses of resightings of 519 chough fledglings from eight cohorts demonstrated substantial variation in monthly apparent survival probabilities (Φm) during the year after fledging and substantial among-cohort variation in the pattern and magnitude of this variation. Recent low first-year survival (Φ1), which threatens population viability, primarily reflected an additional period of markedly low survival compared to previous cohorts.

Among-month variation in survival

Across all 519 fledglings, Φm varied from 0·83 during May–June to 0·97 during February–March. This difference in survival probability of c. 0·14 between the lowest and highest survival months represents a substantial biological effect, and demonstrates the existence of distinct periods of low average survival (Fig. 1a–c) rather than a constant rate of attrition throughout the first year (Fig. 1d).

Mean Φm was low during the first month from ringing (May–June), representing the immediate pre- and post-fledging period; this interval accounted for c. 24% of the total mean first-year mortality of c. 71% experienced across the eight studied cohorts. All eight cohorts experienced some degree of low survival during this month, comprising c. 10% (1986) to 40% (1985) of the total first-year mortality experienced by that cohort. Furthermore, since exact ringing dates varied through mid-late May, the May–June interval averaged slightly shorter than subsequent monthly intervals. Values of Φm for May–June will therefore slightly overestimate the true survival probability compared to subsequent months. Low survival during the immediate pre- and post-fledging period might generally be expected, since inexperienced offspring may be poor foragers and/or vulnerable to predation and hence particularly sensitive to low prey availability or high predator density (Naef-Daenzer, Widmer & Nuber 2001; Wiens, Noon & Reynolds 2006; Berkeley, McCarty & Wolfenbarger 2007; Daunt et al. 2007). However, while the 17% probability of mortality estimated for May–June (and 24% of first-year mortality accounted for) is substantial, it is lower than estimated by studies that focussed on measuring post-fledging survival. These often report ≥ 50% mortality within the first few days or weeks after fledging (e.g. Anders et al. 1997; Naef-Daenzer, Widmer & Nuber 2001; Robinson et al. 2004; Yackel Adams, Shagen & Savidge 2006; Berkeley, McCarty & Wolfenbarger 2007; Martín et al. 2007), although post-fledging survival is obviously much higher in species with higher overall first-year survival (e.g. van der Jeugd & Larsson 1998; McIntyre, Collopy & Lindberg 2006).

The converse interpretation of our data is that in choughs, c. 76% of the total first-year mortality occurred after the first month from ringing and therefore after the typical time of fledgling dispersal from natal territory to sub-adult foraging and roosting areas (Fig. 5). Despite our generally high monthly pm and corresponding confidence in estimates of Φm, uncertainty around estimates of Φm might be expected to increase for later months due to the inevitable decrease in cohort size with time since ringing. However, estimates of Φm for January–May were relatively consistently high across all cohorts, giving Φm≥ 0·94 for these 4 months with relatively small standard errors. This period therefore accounted for only c. 6% of the total mean first-year mortality. Such increased survival probability with months since fledging has been observed in other species (e.g. McIntyre, Collopy & Lindberg 2006; Martín et al. 2007) and may reflect ameliorated environmental conditions in spring, increased expertise at foraging and/or predator avoidance, or initial selection against individuals with intrinsically low survival probabilities (van der Jeugd & Larsson 1998).

Among-cohort variation in monthly survival

Estimates of Φm for months other than May–June and January–May were less consistent across cohorts. Across the five cohorts (1984–2006) where overall Φ1 was high, Φm was relatively high during June–December but lower during December–January. The latter month accounted for c. 28% of the total first-year mortality experienced by the five high-survival cohorts and coincides with a widely expected mid-winter survival bottleneck; low mid-winter survival is predicted in temperate regions since thermoregulatory costs are high and foraging is constrained by daylight, increasing sensitivity to prey availability and predation risk (Robinson et al. 2004; Daunt et al. 2007). In contrast, across the three cohorts (2007–2009) where overall Φ1 was very low, Φm was conspicuously low during July–October and November–December. Since Φm was as low as c. 0·65, the overall probability of survival through the 5-month July–December period was only c. 0·20, accounting for almost two-thirds (c. 63%) of the substantial mortality experienced by these three cohorts. By contrast, individuals in the five high-survival cohorts had an overall probability of c. 0·68 of surviving through this same 5-month period. The very low overall Φ1 experienced by the 2007–2009 cohorts, which if continued is likely to render Islay’s chough population unviable, therefore primarily reflected additional late summer and early winter periods of low Φm that were not experienced by previous cohorts (akin to Fig. 1c). It did not primarily reflect decreased baseline survival across all months (Fig. 1d) or substantial decreases in survival during the immediate post-fledging (May–July) or mid-winter (December–January) periods when Φm was, and might be expected to be, relatively low across all cohorts.

Conclusions & management implications

Comparisons of variation in monthly survival probability (Φm) allow key mortality periods to be identified, targeted hypotheses explaining observed mortality to be generated and tested and hence efficient mitigation designed. Our analyses suggest that one appropriate time for management aimed at increasing overall first-year survival in Islay’s choughs may be the fledging period (May–June), when some degree of low survival affected most cohorts and mean Φm was only 0·83. On Ouessant, France, Φm for the first month from ringing was > 0·95 across 73 choughs from six cohorts, representing only c. 5% of total first-year mortality (Kerbiriou & Julliard 2007). This comparison suggests that it may be biologically realistic to increase Φm on Islay. Post-ringing nest checks and resightings on Islay suggest that most May–June mortality occurred post- rather than pre-fledging (E. Bignal et al., unpublished data). In other bird species, post-fledging survival can increase with fledgling condition (e.g. Naef-Daenzer, Widmer & Nuber 2001; although see Anders et al. 1997). Managing for increased fledgling condition, whether through maintaining appropriate grassland habitat within breeding territories and hence increased parental condition and provisioning, or through direct provision of supplementary food, may therefore be beneficial on Islay.

In contrast, our analyses suggest little potential direct benefit of management aimed at increasing sub-adult survival during the last months of the biological year (January–May). Observed Φm is already consistently high during these months, and any management-induced increase in Φm through this time could only benefit the relatively few individuals that survive to January. This does not preclude the possibility that appropriate management during this period could benefit breeding adults.

Dramatic decreases in key demographic rates that are likely to cause substantial decreases in λ and rapid population decline need to spark rapid management responses. Our analyses show that the very low first-year survival experienced by the 2007–2009 chough cohorts primarily reflected low survival through a different and unexpected period, July–December, from that experienced by previous cohorts. July–December should clearly be a primary focus for urgent ecological investigation and development of management strategies targeted at increasing sub-adult survival in Islay’s choughs.

Low Φm during late summer and autumn is unlikely to reflect increased post-fledging dispersal away from Islay since no dispersers have been observed elsewhere, and unusual numbers of choughs were found dead on Islay during this period. Low Φm during late summer is perhaps surprising since high prey abundance, mild weather and ample foraging time might all be expected. However, in Ouessant’s choughs, survival was low during July–August (Φm ≈ 0·55) and October–December (Φm ≈ 0·7–0·8), attributed to decreased invertebrate biomass and switches in main prey type (Kerbiriou & Julliard 2007). By analogy, the recent low July–December survival on Islay may therefore reflect altered foraging conditions compared to previous years, potentially reflecting changes in invertebrate populations, weather and/or grassland management, or interactions among these variables that influence prey abundance and/or availability. The low July–December survival occurred after fledglings had dispersed from natal territories to traditional sub-adult foraging areas, which comprise specific grazed coastal dunes and fields that are cut for silage during the post-fledging period (Bignal, Bignal & McCracken 1997). An urgent goal should therefore be to determine the primary causes of mortality in these areas and their underlying ecological mechanisms (which may be lagged in time and space, Reid et al. 2008). These areas have recently experienced substantial changes in grassland management, partly driven by agricultural intensification and agri-environmental incentives. Although the short- and long-term consequences of these actions for the abundance and availability of the chough’s invertebrate prey remain largely unquantified, management that creates long grass swards can render prey unavailable to foraging choughs (e.g. McCracken & Tallowin 2004). In addition, late summer (July–September) rainfall on Islay was substantially higher on average during 2007–2009 than during previous years. This may have reduced invertebrate activity and abundance and increased vegetation height during the period when high sub-adult mortality occurred. For example, wet weather may reduce colonization of freshly deposited cowpats by Aphodius spp. (e.g. Webb et al. 2010), thereby reducing the abundance of larvae on which sub-adult choughs often feed. A recent tendency towards warmer springs may have altered the abundance or phenology of other invertebrate prey, as observed in tipulids in English uplands (Pearce-Higgins et al. 2010), although there is as yet no evidence of similar changes in abundance in southern Scottish grasslands. Alternatively, warmer springs may have increased rates of vegetation growth and therefore reduced the availability of soil invertebrates to foraging choughs in late summer and early autumn. One primary hypothesis is therefore that the recent low survival reflects food shortage due to changes in grassland management and/or weather. This hypothesis requires thorough test through targeted experiments (for example by manipulating grassland management and/or providing supplementary food) and further correlative analysis of links between Φm and management, weather and invertebrate abundance. Effective conservation management for choughs may ultimately require greater recognition and better understanding of links between grassland management and invertebrate abundance and availability, especially in the context of agri-environmental schemes. Managing grasslands to generate a diversity of habitats supporting high invertebrate abundance and availability, thereby increasing the foraging options available to choughs at any time, may provide the best practical means of ameliorating impacts of changes in weather that cannot be managed directly (Reid et al. 2008). However, the possibility that sub-adult mortality may have other as yet unidentified causes, such as increased predation or disease, cannot be discounted.

These analyses have general relevance in demonstrating that the precise timing and location of major demographic changes which threaten population viability, such as unusually low survival, cannot be assumed to match previously known or postulated bottlenecks. Rather, detailed monitoring, covering a range of different years, may be required to pinpoint the timing, location and causes of demographic change and inform the most efficient and effective management responses. The analyses also have specific importance to conserving choughs on Islay in identifying the key periods and locations for targeted ecological investigation and the development of management actions aimed at increasing sub-adult survival and hence ensuring the viability of this protected population.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank the Islay land-owners and farmers who allowed access to nest sites, everyone who contributed to data collection, and RSPB, SNH, NERC, University of Aberdeen, the Royal Society and the Scottish Government Rural and Environment Research and Analysis Directorate (RERAD) for funding.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Appendix S1. Matrix projection models used to estimate asymptotic population growth rate.

Appendix S2. Results of capture-mark-recapture analyses.

Table S1. Definitions of matrix projection model terms and specified parameter values.

Table S2. Capture-mark-recapture models describing apparent monthly survival probability for eight chough cohorts.

Table S3. Capture-mark-recapture models where month-specific survival was constrained to be identical across cohorts with relatively high and low first-year survival probabilities.

Fig. S1. Four stage-class matrix projection model.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_1973_sm_supporting-files.doc99KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.