Upland land use predicts population decline in a globally near-threatened wader

Authors


Summary

  1. Changes in large-scale land use may fragment and degrade habitats, affecting animal species adapted to these habitats. In the UK uplands for example, changes in sheep and game management, and afforestation, have altered the configuration of internationally important moorland habitat and are predicted to have increased predation pressure for a globally unique suite of breeding birds of international conservation importance.
  2. Some of these upland bird species have declined, with particular concern over ground-nesting waders. Using resurveys of the rapidly declining Eurasian curlew Numenius arquata as a focal species of global conservation concern, we investigate whether upland land use predicts low nesting success and population decline.
  3. Curlew population changes over an 8- to 10-year period were positively related to gamekeeper density (a surrogate of predator control intensity) and inversely to the area of woodland surrounding sites, as a likely source of predators to adjacent open ground. Model predictions suggest that increasing woodland cover from 0% to 10% of the land area within 1 km of populated sites requires an increase in human predator control effort of about 48%, to a level associated with high-intensity grouse production, to achieve curlew population stability.
  4. Curlew nesting success, known to be a key driver of population trends, was also positively related to gamekeeper density and inversely to woodland area surrounding sites, providing a plausible mechanistic link between land use and population change.
  5. Synthesis and applications. Upland land use is associated with curlew declines, with predation a likely mechanism, and this may apply to other breeding waders. The removal of isolated woodland plantations from otherwise unafforested landscapes may help reduce predation pressure across a range of systems including moorland. However, direct predator control may also be important to conserve ground-nesting birds in these landscapes, for example, where moorland management and forestry coexist as major land uses. Predator control may also mitigate climate change effects by enhancing wader productivity, particularly where climate effects coincide with changing land use. Emerging land uses in open landscapes, including native woodland restoration and wind farms, require careful siting to minimize further impacts on open-area breeding birds.

Introduction

The influence of changing land use on biodiversity is a fundamental issue in ecology. When the land uses that have shaped a habitat shift in extent or intensity, or new land uses introduced, this may have important consequences for the animals adapted to these habitats. An example is the semi-natural, unenclosed, grazed moorland of the UK uplands, which supports a globally unique community of breeding birds of international conservation importance (Thompson et al. 1995). These open landscapes have been created in the long term by loss of native forest to timber and fuel wood harvesting, grazing and burning (Stevenson & Birks 1995) and shaped more recently by grazing by domestic livestock (principally sheep Ovis aries) and red deer Cervus elaphus, and management for shooting of red grouse Lagopus lagopus scoticus, requiring vegetation burning and predator control.

Some upland-breeding birds have declined over recent decades, with particular concern over ground-nesting waders including lapwing Vanellus vanellus, dunlin Calidris alpina and Eurasian curlew Numenius arquata (Sim et al. 2005). The causes of declines remain unclear, but wader breeding success is a major driver of population change, and predation is in turn often the main cause of breeding failure (MacDonald & Bolton 2008). Declines may therefore have been driven by changes in land use altering predation pressure, affecting the quality of nesting or foraging habitat influencing productivity or interactions between these (Evans 2004).

In these upland systems, the impacts of grazing and burning on vegetation structure and composition may affect nesting and foraging habitat quality (Pearce-Higgins et al. 2009a), and there is landscape-scale and experimental evidence that removal of predators of red grouse may benefit other ground-nesting birds, including waders (Tharme et al. 2001; Fletcher et al. 2010). Significant afforestation has also occurred in the UK uplands since the 1940s; initially conifers for commercial timber, and although still the dominant woodland type, current policies seek to increase and diversify woodland cover for a range of additional environmental, economic and social benefits including carbon sequestration, native woodland restoration, fuel and timber production and flood management (Scottish Government 2009). Afforestation causes direct habitat loss and fragmentation of remaining open ground (Ratcliffe 2007); moreover, such fragmented wooded landscapes host high abundances of mesopredators including red foxes Vulpes vulpes and corvids Corvus spp. relative to continuous open landscapes (Valkama, Currie & Korpimäki 1999). Afforestation is associated with edge effects on adjacent open ground across a range of systems (Ries et al. 2004) with, for example, planting of cereal steppe leading to higher avian nest predation in surrounding agricultural land, by altering habitat configuration and through the association of predators with woodland (Reino et al. 2010). Increased predation pressure on open ground adjacent to woodland, coupled with potential behavioural avoidance of edge habitats, may therefore have added to direct losses from afforestation for waders.

In this study, we bring together measures of grazed vegetation structure and composition, predator control and woodland cover to test their influences on changes in abundance, and nesting success, of an upland-breeding wader over 8-10 years in two study regions with contrasting overall population trends. Specifically, we predict that (i) declines in grazing intensity are associated with increases in wader abundance; (ii) gamekeeper density, a surrogate of predator control intensity, is positively related to nesting success and population change; and (iii) afforestation is associated with greater predator abundance on adjacent open ground and is inversely related to wader nesting success and population change.

Our focal species is the Eurasian curlew (hereafter curlew), a species of urgent conservation concern, now globally near-threatened because of widespread declines across the breeding range (BirdLife International 2013). Our study sites are in the UK, which was previously estimated to hold 39% of the European breeding population, but where breeding numbers have shown a long-term decline of 61% from 1970 to 2010, a recent rapid 44% decline from 1995 to 2010 (Eaton et al. 2012) and a catastrophic decline in Northern Ireland (Birdwatch Ireland 2011).

Although our study is focussed for these conservation reasons, we expect results to be applicable across the breeding range of curlew, where large-scale land uses such as farming and forestry impact on the species (Valkama, Currie & Korpimäki 1999); to other waders of conservation concern exposed to the same land-use effects, including lapwing and dunlin (Pearce-Higgins et al. 2009a; Eaton et al. 2012); and more widely where open-area birds of conservation concern are threatened by emerging global pressures including afforestation of open ground (Allan et al. 1997; Di Giacomo & Krapovickas 2001; Brennan & Kuvlevsky 2005; Reino et al. 2010), changes to extensive grazing systems (Dong et al. 2011) and widespread increases in generalist predator abundances (Macdonald & Reynolds 2008; van der Vliet, Schuller & Wassen 2008; EBCC 2012).

Lastly, there is increasing interest in predator control to enhance bird productivity, for adaptive management to increase resistance to climate change (Pearce-Higgins 2011), with a need to understand how changes in predator control intensity influences bird populations, alongside changes in other land uses acting within the same landscape.

Materials and methods

Study Design

Work was undertaken in two regions (Fig. 1) with differing curlew population trends: south Scotland (declining) and the South Pennines (increasing) (Sim et al. 2005). Previous studies had established sites (ranging from 1·8 to 2·3 km2 in area) in these regions, within which curlew abundance and a range of environmental correlates were measured (Pearce-Higgins & Grant 2006; Buchanan, Pearce-Higgins & Grant 2007). All sites were on unenclosed moorland between 280 and 610 m asl, dominated by acid grassland (Molinia caerulea/Nardus stricta) and heather Calluna vulgaris. Sites were originally selected to be at least 200 m from the nearest woodland edge, due to an assumption that woodland edge effects may confound bird abundance (see 'Discussion') (Pearce-Higgins & Grant 2006).

Figure 1.

Location of study sites within the two regions.

In south Scotland, 71 sites were first surveyed in 1999 or 2000, selected using a stratified random sample based upon the extent of heather cover (Pearce-Higgins & Grant 2006). Sites were considered for resurvey in 2009 if they had held two or more pairs of curlew during the initial surveys, to increase the likelihood that some sites retained curlew, to enable estimates of current nesting success to be obtained. A random sample of 41 of these sites was then selected, stratified by the seven hill ranges distinguished in Pearce-Higgins and Grant (2006), to ensure a representative geographical spread.

In the South Pennines, the initial sample of 37 sites in 2002 was also a random sample of moorland, stratified by heather cover (Buchanan et al. 2007). Thirty-six sites were resurveyed in 2010, omitting one for which access was refused. Although three of the 36 sites had no pairs recorded in 2002, these were resurveyed to provide data on nesting success in the current period, in the event that pairs were recorded in 2010 in this increasing population. These sites were excluded from analyses of population change to avoid confounding correlates of population change in occupied sites with those of colonization of unoccupied sites.

During resurveys, we repeated the initial data collection methods of Pearce-Higgins and Grant (2006) and Buchanan et al. (2007). These comprised curlew abundance and environmental correlates, which formed explanatory variables to test the hypotheses of drivers of curlew population change. We also estimated curlew nesting success and fox activity during the resurveys; these data were not collected in the previous surveys.

Bird Surveys to Determine Abundance

Breeding curlew were surveyed using a three-visit method (O'Brien & Smith 1992); survey periods were 13 April–14 May, 10 May–31 May and 27 May–18 June. Visits were a minimum of seven days apart, with one visit per site commencing at dawn or terminating at dusk. Surveys were not undertaken during rain or strong winds. On each visit, the entire site was covered to within 100 m along parallel transects 200 m apart, recording bird locations and behavioural codes. Either a single bird or a pair comprised a breeding pair, excluding those overflying. At least two visits per site were conducted by different observers and the mean count of pairs across the three visits comprised the site population estimate, to account for potential observer variation in curlew detection (Pearce-Higgins & Grant 2006).

Nesting Success

For 2009–2010 only, the proportion of pairs hatching young was estimated per site, during the last of the three standard survey visits and on two subsequent visits approximately two weeks apart (Grant et al. 2000). Adults with dependent young perform vigorous alarm-calling behaviour towards intruders, and the maximum single-visit count of alarm-calling pairs from these three later visits comprised the number of pairs hatching young, expressed as the proportion of the maximum number of breeding pairs from the first three survey visits.

Correlates of Population Change and Nesting Success

Environmental measures were collected from sites following methods established in previous surveys. Variables are summarized as follows (see Table 1 for descriptions and Appendix S1 in Supporting Information for full methods) and unless otherwise stated were remeasured in the current surveys to assess the magnitude of change relative to the previous survey.

Table 1. Univariate tests of explanatory variables for (a) curlew population change between two surveys 8–10 years apart, n = 72 plots in south Scotland and the South Pennines; (b) curlew nesting success in current survey period 2009–10, n = 61 sites. Variables are the value per site in the initial survey (1), current survey (2), absolute change between surveys, or fixed value across both surveys. Analysis of population change considered all measures; nesting success considered only current measures. Significant terms at < 0·05 (in bold) were considered for multivariate modelling
Variable typeVariable name(a) Population change(b) Nesting successDescription
SlopeSEF1, 70 P SlopeSEχ21 P
  1. a

    Many sites (30 of 72) had no woodland surrounding sites in initial surveys, particularly the South Pennines (25 of 32 sites; South Scotland 5 of 40), making a continuous variable of initial woodland area uninformative; this was converted to woodland present/absent within 1-km radius during initial surveys. Because many sites had woodland created between surveys, we considered current woodland area and change in woodland area as continuous variables.

  2. b

    Fox activity only recorded in current surveys.

Land useSheep 1−0·004080·004160·960·330    Percentage of sample points with sheep droppings
Sheep 2−0·007860·005601·970·1650·0156 0·00748 4·40 0·036 As above
Sheep change−0·0002740·005060·000·957    Difference between 1 and 2
Grazing 1−0·006200·003133·930·051    Percentage of sample points with bitten vegetation
Grazing 20·0100 0·00403 6·20 0·015 0·0106 0·00531 4·07 0·044 As above
Grazing change0·0005400·00405-0·020·894    Difference between 1 and 2
Gamekeeper density 14·963·551·950·167    Gamekeeper density (gamekeepers km−2) as index of predator control
Gamekeeper density 2 9·01 2·81 10·24 0·002 13·8 3·74 13·98 <0·001 As above
Gamekeeper change 9·27 3·74 6·16 0·016     Difference between 1 and 2
Burning 10·005590·01370·170·685    Percentage of vegetation sample points in burnt ground
Burning 20·01780·01082·730·1030·01430·009182·420·120As above
Burning change0·02330·01382·860·095    Difference between 1 and 2
Woodland pres/ab 1a0·405 0·178 5·18 0·026     Presence/absence of woodland within 1-km radius buffer around site
Woodland area 2a0·0236 0·00838 7·94 0·006 0·0539 0·0154 14·60 <0·001 Percentage area of 1-km radius buffer around site comprising woodland
Woodland area changea−0·02530·01393·340·072    Change in woodland area between 1 and 2
Predator abundanceCrow 10·03040·04120·540·463    Maximum number of groups of one or more carrion crows
Crow 2−0·02830·04320·430·513−0·02080·04250·240·623As above
Crow change−0·04430·03631·490·226    Difference between 1 and 2
Fox 2b−0·3220·1773·290·074−0·2310·2860·660·418Number of fox scats per km per 100 days
VegetationHeight 1−0·001290·01530·010·933    Mean vegetation height (cm)
Height 2−0·01680·01491·260·2650·009920·02000·250·620As above
Height change−0·01300·01370·9000·346    Difference between 1 and 2
Height heterogeneity 1−0·02460·02530·940·336    Mean difference in vegetation height between successive sample points
Height heterogeneity 20·001680·02270·010·9410·01970·02820·490·485As above
Height heterogeneity change0·02060·02230·850·359    Difference between 1 and 2
Density 1−0·2240·3690·370·545    Mean number of visible white marks on cane at 0, 10, 20, 30 and 40cm
Density 20·2000·3220·380·537−0·3530·4260·690·406As above
Density change0·2980·2881·080·303    Difference between 1 and 2
Heather/grass 1−0·1140·2840·160·689    Ratio of heather/grass cover within site (0=100% grass, 1=100% heather)
Heather/grass 20·04850·2850·030·8650·732 0·373 3·89 0·049 As above
Heather/grass change1·030·7042·140·148    Change in ratio of heather/grass
Tall rush cover 1−0·04220·03181·760·189    Percentage cover of tall Juncus spp as surrogate for site wetness
Tall rush cover 2−0·04020·02313·030·086−0·04930·03122·600·107As above
Tall rush cover change−0·03410·03261·100·299    Difference between 1 and 2
TopographicAltitude−0·0003950·001120·120·7260·0008950·001490·360·549Mean altitude of site
Slope>5°−0·005360·003532·310·133−0·002420·003680·430·512Percentage of plot with ground sloping >5o

Land use measures comprised grazing (index of sheep density; index of pressure from all grazers), woodland (percentage cover of a 1-km radius buffer around the boundary of a site comprising woodland) and grouse management (gamekeeper density (index of predator control intensity (Tharme et al. 2001)); area of vegetation burning).

Finer scale biological measures comprised vegetation structure (height, density, height heterogeneity), vegetation composition (ratio of heather to grass within a site; cover of tall rushes Juncus spp as a surrogate for soil moisture) and predator abundance (carrion crows Corvus corone; fox activity (recorded in current period only)). Site topographical data, assumed to have remained fixed between the two survey periods, comprised mean site altitude and gradient.

Analysis

Influence of woodland cover on predator abundances

We first examined whether afforestation is associated with greater predator abundance on adjacent open ground. Because the relationship between predators and woodland was likely to be influenced by gamekeeping intensity, we used a subset of sites (n = 31) with no gamekeeping in the current surveys (the only time period with fox data) and examined the correlation between area of woodland within a 1-km radius of sites, and abundance indices of foxes and crows.

Associations between curlew population change and grazing, predator control and woodland cover

The response variable was the log ratio of change per site, with 0·5 added to values to account for some zero counts in the current period, calculated as ln(previous count+0·5/current count+0·5) (Fletcher et al. 2010). This response variable had a normal distribution and model checking confirmed normality of residuals. Two sites with missing data for some explanatory variables were removed, yielding 72 sites (southern Scotland = 40; south Pennines = 32). Explanatory variables were site-specific measures describing land use, finer scale biological measures and topography (Table 1). Where a variable had been measured in both surveys, both the initial and current values were available, and we also calculated the absolute magnitude of change by subtracting the initial value from the current. We considered initial, current and change measures of explanatory variables where available, or a single value for those that remained fixed between surveys. This yielded 36 candidate explanatory variables (Table 1). We reduced the list of candidate variables by employing a two-stage approach to identify important terms (Pearce-Higgins et al. 2009b). Each variable was first fitted in a univariate regression model with normal error structure and identity link. Variables with < 0·05 were identified using F tests and pairwise correlations between these examined; where variables were strongly correlated (r > 0·5), the weaker of the two variables in explaining curlew population change was excluded. Remaining variables were then fitted in a multiple regression model, simplified using backwards stepwise deletion to a minimum adequate model (MAM) of variables significant at < 0·05, testing for interactions between variables in the MAM. We examined regional differences in the slope of remaining explanatory variables by testing the addition of variable*region interaction terms to the MAM.

Stepwise modelling has been criticized in favour of approaches using information criteria and model averaging (Whittingham et al. 2006). A subsequent study validated stepwise deletion for model selection and established that it performed as well as other methods of producing predictive models (Murtaugh 2009). Moreover, and importantly for this study, model averaging was not a feasible alternative, due to the need to consider a relatively large number of variables (and with no means of reducing this list based upon existing knowledge of curlew biology), and in particular the need for post hoc testing of interactions between explanatory variables in the MAMs and interactions of these variables with the two-level region factor. We minimize the problems associated with stepwise approaches by considering correlations between explanatory variables and accounting for these as described above.

We then used the variables that were important in the MAM of population change (current gamekeeper density and current woodland area) to further examine the influence of woodland cover and predator control on curlew populations. Using the parameter estimates from the MAM, we examined the minimum level of gamekeeper density, which predicted curlew population stability (log ratio of zero) under three scenarios of woodland cover (0%, 5% and 10% of land within a 1-km radius of sites). Using eqn (eqn 1), we calculated curlew population change (y) in relation to current gamekeeper density (x1, using 100 values of x1 increasing at equal increments within the range of current values 0–0·1 keepers km−2), where b1 equals the slope of current gamekeeper density from the model and the intercept a. Current woodland area x2 was held initially at 0, with the slope of woodland cover from the model as b2; this was repeated for values of x2 of 5 and 10.

display math(eqn 1)

Associations between curlew nesting success and grazing, predator control and woodland cover

Using binomial regression models with logit link, the response variable was the number of pairs hatching young per site as the numerator, with maximum number of pairs per site from the first three visits as denominator, minus number of pairs hatching young (in R terminology; ‘successes/failures’, Crawley 2007). Sites were included if they had at least one pair of curlew during the resurveys and no missing values in any relevant explanatory variables, yielding 61 sites (southern Scotland = 33, South Pennines = 28). Explanatory variables comprised measures from the current resurvey period, including those fixed between survey periods. This yielded 14 variables (Table 1), with the modelling stages repeated as described above for population change, using chi-square tests in place of F tests, testing for correlations between explanatory variables and regional interactions with terms in the MAM.

Analyses were conducted in R 2.15.0 (R Core Development Team 2012).

Results

Influence of Woodland Cover on Predator Abundances

For a subset of sites (n = 31) with no gamekeeping in the current surveys, the area of woodland surrounding study sites showed a highly significant positive correlation with the fox abundance index (r = 0·52, = 0·003), but no clear correlation with crow abundance (r = 0·11, = 0·556).

Associations between curlew population change and grazing, predator control and woodland cover

The minimum adequate model (MAM) of population change contained two significant variables: current gamekeeper density (positive) and the current area of woodland surrounding study sites (negative) (Table 2, Fig. 2); there was no significant interaction between these variables (F1,68 = 1·82, = 0·182). There was no significant interaction with region for either variable in the MAM (gamekeeper density: F1,68 = 2·05, = 0·157; woodland area: F1,68 = 0·07, = 0·794).

Table 2. Minimum adequate model (MAM) of curlew population change between surveys 8–10 years apart, n = 72 sites across two regions
VariableSlopeSEFd.f. P
MAM
Current gamekeeper density8·282·719·331, 690·003
Current woodland area−0·02120·007967·08 1, 690·010
Nonsignificant terms deleted
Current grazing index−0·003690·004200·261, 670·383
Initial presence/absence of woodland−0·1960·1751·251, 680·267
Figure 2.

Correlates of curlew population change over an 8- to 10-year period. Plots show fitted relationships where log ratio of zero equals no population change between surveys. Southern Scotland = filled circles, South Pennines = open circles.

Model predictions from this data set suggest that levels of gamekeeper density predicted to achieve curlew population stability under varying scenarios of woodland area were as follows: 0% woodland within a 1-km radius of sites=0·054 gamekeepers per km2 (equivalent to one gamekeeper for every 18–19 km2), 5% woodland = 0·067 per km2 (one per 14–15 km2) and 10% woodland = 0·080 per km2 (one per 12–13 km2) (Fig. 3).

Figure 3.

Gamekeeper densities predicted to achieve curlew stability (log ratio = 0), under three scenarios of woodland area within a 1-km radius of populated sites on adjacent open ground.

Associations between curlew nesting success and grazing, predator control and woodland cover

The mean percentage of pairs with young per site was 38% (range 0–100%, n = 61 sites), with a mean of 24% in south Scotland (range 0–67%, n = 33 sites) and 56% in the south Pennines (range 0–100%, n = 28 sites). The MAM of nesting success contained two significant variables: gamekeeper density (positive) and woodland area surrounding sites (negative) (Table 3, Fig. 4); there was no significant interaction between these variables (χ21 = 0·43, = 0·510). There was no significant interaction between gamekeeper density and region (χ21 = 2·23, = 0·135), but a significant interaction between woodland area and region (χ21 = 7·56, = 0·006), with a more negative relationship between woodland area and nesting success in the south Pennines (Fig. 4) (slope −0·120 ± 0·0394), and the slope for south Scotland not differing significantly from zero (slope −0·0125 ± 0·0429).

Table 3. Minimum adequate model (MAM) of curlew nesting success in current survey period 2009–2010, n = 61 sites across two regions
VariableSlopeSEχ2d.f. P
MAM
Current gamekeeper density11·13·848·5110·004
Current woodland area−0·04440·01579·1310·003
Nonsignificant terms deleted
Current sheep index−0·009570·007741·5410·215
Current ratio of heather to grass−0·7550·3923·7610·053
Figure 4.

Correlates of curlew nesting success in 2009–2010. Plots show fitted relationships with single solid line where there was no interaction between a variable and study region and separate lines where the regional interaction was significant. Southern Scotland = filled circles and dashed line, South Pennines = open circles and dotted line.

Discussion

Curlew nesting success and population change were inversely associated with the exposure of sites to woodland, yielding the first quantitative evidence of indirect effects of afforestation on this species of global conservation concern. Afforestation will also have caused direct habitat loss, and while studies quantifying the impacts on upland birds are lacking, an estimate from south Scotland suggests that 5000 curlew pairs were lost since the 1939–1945 war, from replacement of open ground with conifer woodland (Ratcliffe 2007). In the present study, as no woodland was planted on study sites, these impacts probably arise via edge effects on adjacent open ground. A likely mechanism is increased predation pressure, in particular from foxes, the abundance of which was correlated with woodland area surrounding sites. There was less evidence for a relationship between carrion crows and woodland area, even though this species will also breed in woodland and forage over surrounding open ground (Stroud et al. 1987; Ratcliffe 2007). Because study sites were originally selected to be at least 200 m from woodland edge (Pearce-Higgins & Grant 2006), these results may underestimate the magnitude of adverse woodland edge effects on curlew nesting success and population change.

These results are consistent with a previous spatial study showing that in a fragmented farmland landscape containing woodland, the abundance of foxes and crows was 2–3 times higher, and curlew nest predation rates four times higher, than that in a continuous farmland landscape without woodland (Valkama, Currie & Korpimäki 1999). Other breeding waders also show negative relationships between measures of productivity or population change and exposure to upland woodland edge (Finney, Pearce-Higgins & Yalden 2005; Amar et al. 2011a). These results are also consistent with a wider body of evidence, linking edge effects in fragmented landscapes to nest predation across a range of avian taxa (Batary & Baldi 2004).

In contrast, gamekeeper density (a surrogate of predator control intensity) was positively associated with curlew population change and nesting success. This effect was stronger than that of direct predator abundance measures (foxes and crows) and could reflect the imprecision of survey methods in estimating fox and crow abundance as indices of their predatory impact. However, gamekeepers also routinely control other predators (e.g. mustelids), and gamekeeper density may therefore better represent the overall intensity of predator control.

In contrast to afforestation and predator control, there was little evidence for associations of grazing and its effects on vegetation condition on curlew. Indices of grazing pressure generally declined across sites between the two survey periods (see Table S1, Supporting information), consistent with wider reductions in stocking densities (e.g. SAC 2008), with some evidence of associated changes in vegetation structure (e.g. nonsignificant increase in vegetation height and significant increase in height heterogeneity, Table S1). Reductions in grazing have shown benefits for other upland birds (Calladine, Baines & Warren 2002; Amar et al. 2011b); however, it appears that for curlew, other land uses are currently exerting stronger pressures through their likely influence on predation pressure. The lack of effects of grazing and associated habitat measures within the range of this study may also reflect the relatively weaker associations between variation in curlew breeding abundance and measures of habitat structure and composition than other upland waders (Pearce-Higgins & Grant 2006).

Breeding success is a key determinant of wader population changes, with predation, typically from mesopredators such as foxes and corvids, in turn a major cause of breeding failure (MacDonald & Bolton 2008). When considered in conjunction with previous studies of upland waders and the influence of predator control on their breeding success and breeding densities, this study suggests that interactions between landscape structure (the configuration of woodland and open ground) and predation pressure may be an important influence on population change for curlew in the UK uplands and may be for other waders such as lapwing and dunlin. This evidence can be summarized as follows: (i) the recent catastrophic declines of curlew (and other waders) in Northern Ireland (Birdwatch Ireland 2011), where previous detailed studies have shown that high nest predation is the probable cause of curlew declines, at least (Grant et al. 1999); (ii) higher breeding densities of waders on land managed for grouse than those on land not managed for grouse, due at least in part to predator control (Tharme et al. 2001); (iii) positive effects of experimentally deployed predator control on wader breeding success and population change (Fletcher et al. 2010); (iv) lower declines in waders such as lapwing where grouse management was most intensive and larger declines in areas with high densities of carrion crow (Amar et al. 2011a); and (v) the lack of other environmental effects on nesting success and population change in this study (grazing, habitat, topography).

Management Implications

Afforestation and, in some areas, declines in predator control for grouse shooting are likely to have increased avian and mammalian nest predators in open upland habitats. These effects may not act in isolation, for example, grouse management may be abandoned close to afforested areas (Robertson, Park & Barton 2001), presumably because the source of predators reduces the efficiency of predator control; this may compound woodland edge predation effects for waders on surrounding open ground.

Conservation measures for breeding waders and other ground-nesting birds should consider means to reduce nest predation rates. Although based on correlative data, our model predictions suggest that the workforce required to achieve curlew population stability through predator control increased from one gamekeeper per 18–19 km2 of land to one per 12–13 km2, if woodland increased from 0% to 10% of land area within a 1-km radius of populated sites. These are high workforce requirements, typical of high-intensity grouse production (Hudson & Newborn 1995), and will be expensive to fund over large scales (Sotherton, Tapper & Smith 2009). Payments are available through agri-environment schemes for predator control for bird conservation, although, using Scotland as an example, currently only on certain sites (e.g. EU-designated Natura sites) or for selected other species (black grouse Tetrao tetrix and capercaillie Tetrao urogallus, Scottish Government 2012). At present, therefore, benefits of predator control for ground-nesting upland birds such as waders are usually a by-product of grouse management and must therefore be set against the disbenefits arising from the illegal killing of raptors and growing evidence of wider environmental costs of high-intensity grouse production (Thompson et al. 2009; Grant et al. 2012).

There is therefore also a need to consider reducing predation pressure through landscape-scale manipulations of land use pattern, both in these upland landscapes and more widely where afforestation of open habitats is emerging as a threat to open-area birds through edge effects (Brennan & Kuvlevsky 2005; Reino et al. 2010), for example through the removal or redistribution of existing woodland away from important breeding bird areas. In one such landscape in northern Scotland, removal of commercial conifer plantations for blanket bog restoration could also benefit waders breeding on surrounding open ground through reduced edge effects (J.D. Wilson, R. Anderson, S. Bailey, J. Chetcuti, N.R. Cowie, M.H. Hancock, C.P. Quine, N. Russell, L. Stephen, & D.B.A. Thompson, in review). Similar interventions in other landscapes for breeding waders would allow testing of whether these populations can be maintained and recovered without intensive predator control. Furthermore, future afforestation of open systems in the UK and elsewhere, to deliver a range of objectives (e.g. Scottish Government 2009), requires careful siting away from existing breeding bird strongholds, so that populations are not further affected by fragmentation of open ground and woodland edge predation effects.

Further understanding the drivers of high predation pressure from mesopredators such as foxes and crows in disturbed landscapes would also be beneficial; in addition to afforestation, these may include loss of apex predators and superabundance of non-native prey such as rabbits Oryctolagus cuniculus and released game birds (Lees, Newton & Balmford 2013; Pasanen-Mortensen, Pyykönen & Elmhagen 2013).

Lastly, there is interest in the use of predator control to enhance wader productivity, to increase resistance to climate change (Pearce-Higgins 2011). Upland birds may be particularly vulnerable to climate change effects (Pearce-Higgins et al. 2009a), and predator control may play an important role in reducing the severity of predicted population decline associated with temperature increases, particularly where climate change effects occur alongside other changes in land use. Upland waders also face threats from emerging land uses including wind farms (Pearce-Higgins et al. 2009b, 2012), which should be sited carefully.

Acknowledgements

This work was funded by RSPB and Natural England through the Action for Birds in England partnership. We thank landowners and gamekeepers for land access and management information, RSPB field staff for data collection and Emma Teuten for map production. Comments from Andy Brown, Ian Johnstone, Jen Smart, Pat Thompson and two referees improved drafts.

Ancillary